FYI Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "CFPB_ResearchConference"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, January 7, 2015 2:08:09 PM Subject: RE: Paper submission for Conference ­ Research on Consumer Finance Hi Jennifer, We’re very excited and thankful to get your submission. We’ll be back in touch about the conference agenda, and whether your submission was included, by mid-March. Thanks! David Kastelman From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, December 30, 2014 11:26 AM To: CFPB_ResearchConference Subject: Paper submission for Conference - Research on Consumer Finance I am writing in reference to the CFPB Conference - Research on Consumer Finance. Abstract: Using payday-lender administrative data matched to borrower credit attributes from a national credit bureau, I find that borrowers who engage in protracted refinancing (“rollover”) activity have better financial outcomes (measured by changes in credit scores) than consumers whose borrowing is limited to shorter periods. These results are robust to an alternative definition of a “rollover” that ignores out-of-debt periods of 14 days between successive loans. Also, exploiting interstate differences in rollover regulation, I find that, while regulation has a small effect on longer-term usage patterns, consumers whose borrowing is less restricted by regulation fare better than consumers in the most restrictive states, controlling for initial financial condition. These findings directly contradict key assumptions about this market, raise significant policy questions for federal regulators, and suggest the appropriateness of further study of actual consumer outcomes before the imposition of new regulatory rollover restrictions. Please find my completed research paper attached. The paper can also be found on the SSRN: http://papers.ssrn.com/sol3/papers.cfm? abstract_id=2534628 In advance, thank you for your consideration. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - I think I am going to submit the paper to this conference: http://files.consumerfinance.gov/f/201411_cfpb_call-forpapers.pdf Just wanted to check in with you first. Have you given to KSU? http://tinyurl.com/ksuwhyIgive Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? And so it starts... Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "G.E.K."  Sent: Tuesday, December 9, 2014 1:53:07 PM Subject: Re: your thesis ­ PAYDAY LOAN ROLLOVERS AND CONSUMER WELFARE Hi G.E.K. thanks for your note. The main finding of the paper was that for those at the lowest ends of the credit spectrum (<550), payday borrowing activity is correlated with increases in credit scores. The premise of the CFPB's policy is that the two are negatively related. I showed mathematically that not only is that not the case, but in some segments credit scores improve with payday borrowing activity. I would be happy to speak in more detail about the study if needed. Kindest Regards. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "G.E.K."  To: jpriestl@kennesaw.edu Cc: Pmcbain@aol.com Sent: Tuesday, December 9, 2014 1:49:59 PM Subject: your thesis ­ PAYDAY LOAN ROLLOVERS AND CONSUMER WELFARE You are kidding, right? ‘Borrowers who engage in protracted refinancing (“rollover”) activity have better financial outcomes (measured by changes in credit scores) than consumers whose borrowing is limited to shorter periods.’ That your statement is much akin to concluding that a drunk who drinks constantly is in better health than another who drinks twice as much on alternate days to make up for lost time and volume. What is the saying about statistics… ‘lies, damn lies, and ………’ (in this case - bad conclusions). Hi Hilary I loaded the paper yesterday - its status is "Under SSRN Review". I don't know how long that takes. Would you like for me to load the paper in other locations as well? My website? Our Dept website? Forward to Microbuilt? FactorTrust? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks Hilary. Will do. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, December 5, 2014 6:12:18 PM Subject: RE: Priestley Release Shell Here it is – both Word and ready-to-upload .pdf format. Please use the abstract verbatim as the SSRN abstract, if you don’t mind doing so. HM From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, December 05, 2014 4:18 PM To: Hilary B. Miller Subject: Re: Priestley Release Shell Hilary - can you forward to me the paper with the embargo stamp taken off? I will then load it into SSRN. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Amy Cantu"  To: "Jennifer Lewis Priestley (jpriestl@kennesaw.edu)"  Cc: tturne88@kennesaw.edu, "Hilary Miller"  Sent: Friday, December 5, 2014 2:24:59 PM Subject: FW: Priestley Release Shell Dr. Priestley, Tim and I have discussed announcing the study to media next Tuesday, Dec. 9. Can you let us know your plans for putting this up on SSRN? We would like to include the link to the full study in the announcement. Thanks, Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Amy Cantu Sent: Wednesday, December 03, 2014 1:11 PM To: 'Tim Turner' Cc: Tammy DeMel; Jennifer Lewis Priestley; 'Hilary Miller' Subject: RE: Priestley Release Shell Thanks, Tim. What are your thoughts on releasing this next Tuesday, Dec. 9? Amy From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Tuesday, December 02, 2014 3:27 PM To: Amy Cantu Cc: Tammy DeMel; Jennifer Lewis Priestley Subject: Re: Priestley Release Shell Amy: Thanks for sending this over. We'll take a look at it and get it back to you as soon as possible. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu From: "Amy Cantu"  To: "Tim Turner"  Sent: Tuesday, December 2, 2014 2:49:53 PM Subject: RE: Priestley Release Shell Hi Tim, The attached includes some other minor changes. Can you let me know if these are all good by you? Thank you! Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Thursday, November 20, 2014 4:13 PM To: Amy Cantu Subject: Priestley Release Shell Hi Amy: Here's the shell we reworked here for your review. Let me know your thoughts. Thanks. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu That timing is fine...I will let you know as soon as I load it. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Amy Cantu"  To: "Jennifer Lewis Priestley (jpriestl@kennesaw.edu)"  Cc: tturne88@kennesaw.edu, "Hilary Miller"  Sent: Friday, December 5, 2014 2:24:59 PM Subject: FW: Priestley Release Shell Dr. Priestley, Tim and I have discussed announcing the study to media next Tuesday, Dec. 9. Can you let us know your plans for putting this up on SSRN? We would like to include the link to the full study in the announcement. Thanks, Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Amy Cantu Sent: Wednesday, December 03, 2014 1:11 PM To: 'Tim Turner' Cc: Tammy DeMel; Jennifer Lewis Priestley; 'Hilary Miller' Subject: RE: Priestley Release Shell Thanks, Tim. What are your thoughts on releasing this next Tuesday, Dec. 9? Amy From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Tuesday, December 02, 2014 3:27 PM To: Amy Cantu Cc: Tammy DeMel; Jennifer Lewis Priestley Subject: Re: Priestley Release Shell Amy: Thanks for sending this over. We'll take a look at it and get it back to you as soon as possible. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu From: "Amy Cantu"  To: "Tim Turner"  Sent: Tuesday, December 2, 2014 2:49:53 PM Subject: RE: Priestley Release Shell Hi Tim, The attached includes some other minor changes. Can you let me know if these are all good by you? Thank you! Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Thursday, November 20, 2014 4:13 PM To: Amy Cantu Subject: Priestley Release Shell Hi Amy: Here's the shell we reworked here for your review. Let me know your thoughts. Thanks. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu I have a few small comments - see attached. Overall, I think it reads well. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Amy Cantu"  To: "Jennifer Lewis Priestley (jpriestl@kennesaw.edu)"  , tturne88@kennesaw.edu Cc: "Tammy DeMel" , "Hilary Miller"  Sent: Wednesday, November 19, 2014 5:21:51 PM Subject: RE: Press release shell Dr. Priestley - Attached for your review and consideration is a draft press release announcing your recent study. Tim - Could we possibly connect again via phone to discuss the mechanics of the release and how we, CFSA and the foundation, can best augment your media outreach efforts? Thank you, Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Monday, November 17, 2014 10:58 AM To: Amy Cantu Cc: Tammy DeMel; Jennifer Lewis Priestley Subject: Re: Press release shell No problem, Amy. Was just checking in on it since initially Wednesday was our deadline to get the release out the door. Thanks for the update. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu From: "Amy Cantu"  To: "Tim Turner"  Cc: "Tammy DeMel"  Sent: Monday, November 17, 2014 10:13:09 AM Subject: RE: Press release shell Hi Tim – Thanks for reaching out. We are refining the release to make sure we portray the findings in the best way. We’ve asked the research foundation to review it as well. I should have our draft to you and Ms. Priestley in the next day or two. My apologies for the delay. Best, Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Tim Turner [mailto:tturne88@kennesaw.edu] Sent: Friday, November 14, 2014 4:24 PM To: Amy Cantu Cc: Tammy DeMel Subject: Press release shell Hi Amy: Just checking in to see where you were with the press release shell. Let me know if you need help with anything. Thanks. Tim Tim Turner, Public Relations Professional Kennesaw State University University Relations 1000 Chastain Road MD 9103 Kennesaw GA 30144 (470) 578-3057 tturne88@kennesaw.edu All of that sounds fine. Tim is a great resource for you. He and I spoke this morning, he has read the paper and is very familiar with the context. As we progress, please do not hesitate to contact me if there is anything I can help with. Jen :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Amy Cantu"  To: "Jennifer Lewis Priestley" , "Hilary B. Miller"  Cc: "Tiffany Capuano" , tturne88@kennesaw.edu Sent: Monday, November 10, 2014 11:47:02 AM Subject: RE: Release of paper Thank you, Jennifer. I spoke with Tim Turner in university relations this morning regarding the mechanics of the release. My team is currently drafting the press release for your consideration. I will be back to you with that by end of week. Best, Amy __ Amy Cantu Communications Director CFSA 515 King St., Suite 300 Alexandria, VA 22314 703.842.2092 (direct) acantu@cfsaa.com CFSA www.cfsaa.com From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, November 06, 2014 1:49 PM To: Hilary B. Miller Cc: Amy Cantu; Tiffany Capuano Subject: Re: Release of paper Hi Hilary All of that sounds fine. Amy, Tiffany Capuano copied here - will be reaching out to you to begin the conversation regarding the webinar... I am looking forward to having the paper "see the light of day". :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Cc: "Amy Cantu (acantu@cfsaa.com)"  Sent: Thursday, November 6, 2014 12:40:30 PM Subject: RE: Release of paper Jen – Here is the final version of the paper in our files. I have copied my colleague Amy Cantu on this message (her phone number is (703) 842-2092). Amy and I will take a first stab at initial language for a release, which we will then send to you for approval and, assuming you agree, you can show to your media relations team. We would also like to coordinate a webinar, at our expense, to accompany the release of the paper. So Amy, you and the media relations team will need to coordinate on timing of a number of different steps. The paper should continue to be embargoed until the actual agreed release date. Thank you so much for your help with this. Regards, Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, November 06, 2014 9:11 AM To: Hilary B. Miller Subject: Re: Release of paper Hilary Just got off the phone with our Media Relations person. We can handle the PR and the release of the paper (universities are pretty good at this...these guys have walked this path before). Here is what I need from you: 1. The final copy of the paper that was distributed to the CFPB (I know I have it, but in the interest of version control I just want to be certain). 2. The contact person at the Consumer Credit Research Foundation that our PR people can coordinate with regarding the message. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, November 5, 2014 3:10:44 PM Subject: RE: Release of paper Jen – Sorry I missed your window. Here’s what I was calling about: We received no feedback from the CFPB about your paper. Although they told us they would be calling you with comments and suggestions, apparently they did not. I think it is reasonable to assume that they either have none, or that they want to hold their fire until after the paper is “out” so that they can get a publicity benefit from making their criticisms public. Either way, it is now approaching time to release the paper. We would like to work with you on the mechanics of release. My question for you is whether your institution will issue a press release regarding the publication of your paper – which we would be happy to draft for their review. Once released, you could put the paper up on SSRN and circulate it to various journals for publication. Happy to have a call to discuss, but the $64 question is whether the press release could come from your end rather than ours. We would greatly prefer this approach. The question is timely and important, so it seems that the school might want to crow over it. Thanks. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, November 05, 2014 1:30 PM To: Hilary B. Miller Subject: Re: Release of paper Hi Hilary. thanks for the note. I am available now...until about 2:45. Then, I will be available again later this evening - after 6. Let me know if that works for you. I will be on my cell - 404-229-3216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Wednesday, November 5, 2014 12:07:05 PM Subject: Release of paper Jen – Do you have time to talk today by phone? Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Hi Hilary All of that sounds fine. Amy, Tiffany Capuano - copied here will be reaching out to you to begin the conversation regarding the webinar... I am looking forward to having the paper "see the light of day". :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Cc: "Amy Cantu (acantu@cfsaa.com)"  Sent: Thursday, November 6, 2014 12:40:30 PM Subject: RE: Release of paper Jen – Here is the final version of the paper in our files. I have copied my colleague Amy Cantu on this message (her phone number is (703) 842-2092). Amy and I will take a first stab at initial language for a release, which we will then send to you for approval and, assuming you agree, you can show to your media relations team. We would also like to coordinate a webinar, at our expense, to accompany the release of the paper. So Amy, you and the media relations team will need to coordinate on timing of a number of different steps. The paper should continue to be embargoed until the actual agreed release date. Thank you so much for your help with this. Regards, Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, November 06, 2014 9:11 AM To: Hilary B. Miller Subject: Re: Release of paper Hilary Just got off the phone with our Media Relations person. We can handle the PR and the release of the paper (universities are pretty good at this...these guys have walked this path before). Here is what I need from you: 1. The final copy of the paper that was distributed to the CFPB (I know I have it, but in the interest of version control I just want to be certain). 2. The contact person at the Consumer Credit Research Foundation that our PR people can coordinate with regarding the message. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, November 5, 2014 3:10:44 PM Subject: RE: Release of paper Jen – Sorry I missed your window. Here’s what I was calling about: We received no feedback from the CFPB about your paper. Although they told us they would be calling you with comments and suggestions, apparently they did not. I think it is reasonable to assume that they either have none, or that they want to hold their fire until after the paper is “out” so that they can get a publicity benefit from making their criticisms public. Either way, it is now approaching time to release the paper. We would like to work with you on the mechanics of release. My question for you is whether your institution will issue a press release regarding the publication of your paper – which we would be happy to draft for their review. Once released, you could put the paper up on SSRN and circulate it to various journals for publication. Happy to have a call to discuss, but the $64 question is whether the press release could come from your end rather than ours. We would greatly prefer this approach. The question is timely and important, so it seems that the school might want to crow over it. Thanks. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, November 05, 2014 1:30 PM To: Hilary B. Miller Subject: Re: Release of paper Hi Hilary. thanks for the note. I am available now...until about 2:45. Then, I will be available again later this evening - after 6. Let me know if that works for you. I will be on my cell - 404-229-3216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Wednesday, November 5, 2014 12:07:05 PM Subject: Release of paper Jen – Do you have time to talk today by phone? Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Hilary Just got off the phone with our Media Relations person. We can handle the PR and the release of the paper (universities are pretty good at this...these guys have walked this path before). Here is what I need from you: 1. The final copy of the paper that was distributed to the CFPB (I know I have it, but in the interest of version control I just want to be certain). 2. The contact person at the Consumer Credit Research Foundation that our PR people can coordinate with regarding the message. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, November 5, 2014 3:10:44 PM Subject: RE: Release of paper Jen – Sorry I missed your window. Here’s what I was calling about: We received no feedback from the CFPB about your paper. Although they told us they would be calling you with comments and suggestions, apparently they did not. I think it is reasonable to assume that they either have none, or that they want to hold their fire until after the paper is “out” so that they can get a publicity benefit from making their criticisms public. Either way, it is now approaching time to release the paper. We would like to work with you on the mechanics of release. My question for you is whether your institution will issue a press release regarding the publication of your paper – which we would be happy to draft for their review. Once released, you could put the paper up on SSRN and circulate it to various journals for publication. Happy to have a call to discuss, but the $64 question is whether the press release could come from your end rather than ours. We would greatly prefer this approach. The question is timely and important, so it seems that the school might want to crow over it. Thanks. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, November 05, 2014 1:30 PM To: Hilary B. Miller Subject: Re: Release of paper Hi Hilary. thanks for the note. I am available now...until about 2:45. Then, I will be available again later this evening - after 6. Let me know if that works for you. I will be on my cell - 404-229-3216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Wednesday, November 5, 2014 12:07:05 PM Subject: Release of paper Jen – Do you have time to talk today by phone? Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. I will check with our media relations people tomorrow. "Hilary B. Miller" wrote: Jen – Sorry I missed your window. Here’s what I was calling about: We received no feedback from the CFPB about your paper. Although they told us they would be calling you with comments and suggestions, apparently they did not. I think it is reasonable to assume that they either have none, or that they want to hold their fire until after the paper is “out” so that they can get a publicity benefit from making their criticisms public. Either way, it is now approaching time to release the paper. We would like to work with you on the mechanics of release. My question for you is whether your institution will issue a press release regarding the publication of your paper – which we would be happy to draft for their review. Once released, you could put the paper up on SSRN and circulate it to various journals for publication. Happy to have a call to discuss, but the $64 question is whether the press release could come from your end rather than ours. We would greatly prefer this approach. The question is timely and important, so it seems that the school might want to crow over it. Thanks. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, November 05, 2014 1:30 PM To: Hilary B. Miller Subject: Re: Release of paper Hi Hilary. thanks for the note. I am available now...until about 2:45. Then, I will be available again later this evening - after 6. Let me know if that works for you. will be on my cell - 404-229-3216 I Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ________________________________ From: "Hilary B. Miller" > To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)" > Sent: Wednesday, November 5, 2014 12:07:05 PM Subject: Release of paper Jen – Do you have time to talk today by phone? Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue - Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 2063727 • hilary@miller.net • bio • v-card download ________________________________ This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding taxrelated penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Hi Hilary. thanks for the note. I am available now...until about 2:45. Then, I will be available again later this evening - after 6. Let me know if that works for you. I will be on my cell - 404-2293216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Wednesday, November 5, 2014 12:07:05 PM Subject: Release of paper Jen – Do you have time to talk today by phone? Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Nope. Radio silence. "Hilary B. Miller" wrote: Jen Have you received any feedback, comments or other communications regarding your paper from the CFPB or anyone else? Hilary Sure - I am around today...and tomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: https://analytics.kennesaw.edu/~jpriestl/ department page: https://analytics.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Thursday, July 24, 2014 9:34:14 AM Subject: FW: The package has been delivered Jen – The subject line is a not-very-secret coded message to reflect that your paper was handdelivered this morning to David Silberman, who is Associate Director for Research, Markets and Regulation at the CFPB. They have known it was coming, I think, but this is their first look. They will likely duplicate and circulate it internally, and your phone will soon start to ring. I am meeting with Jesse Leary, who is their lead economist on payday, at the end of next week, and this will also be a topic for discussion then. Let’s chat briefly when you have a moment, please. Hilary From: joisheffield@yahoo.com [mailto:joisheffield@yahoo.com] Sent: Thursday, July 24, 2014 9:14 AM To: Dennis Shaul; Charles Halloran; Hilary B. Miller Subject: The package has been delivered He was appreciative of the manner in which delivered and stated that. He glanced at the first few pages and said he was looking forward to reading it. I made the points you conveyed to me Hilary and told him I hope that this would encouraged the bureau to dig deeper into this area. Excellent. Drive safely. Talk to you when you return. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, June 30, 2014 9:42:18 AM Subject: Re: Default Trends Thanks for this. I’m going to be thinking about how to proceed. Miller family is going on  National Lampoon vacation through 7/8. Talk next week. Thank you again!! From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Sunday, June 29, 2014 at 10:48 PM To: Hilary Miller Subject: Default Trends Hi Hilary. Take a look at the attached. I separated all borrowers out by their starting Vantage Score (<560, 560-580, 581-600, 600+). This is a simple way to control for exogenous factors. Then I looked at how their scores changed over time - comparing defaulters to non defaulters. Let me know if you want to catch up on Monday. I will be out of pocket the rest of the week. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary. Take a look at the attached. I separated all borrowers out by their starting Vantage Score (<560, 560-580, 581-600, 600+). This is a simple way to control for exogenous factors. Then I looked at how their scores changed over time - comparing defaulters to non defaulters. Let me know if you want to catch up on Monday. I will be out of pocket the rest of the week. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I am doing some more detailed analysis - controlling for starting point - but in the interim, I thought I would send this over to you - this is the overall change in Vantage score over the four year period for those who defaulted versus not. As you can see, the slopes are almost exactly parallel - indicating that the rate of change is the same. In other words, while those who default start and end at a lower vantage score relative to those who do not default, defaulting on payday loans does not appear to impact financial welfare (defined as vantage score). If defaulting did have an impact, then we would expect to see significantly different slopes/rates of change. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Got it. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:38:29 PM Subject: RE: Defaults on Payday Loans We want to control for non-default factors, which in this context means to me comparing outcomes for defaulters with the outcomes similarly initially scored non-defaulters. I leave the methodology to you. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:36 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans Im on it. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:36:14 PM Subject: RE: Defaults on Payday Loans Yes. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:35 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans This is a mock up (not real data)...is this aligned with what you are thinking? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:29:37 PM Subject: RE: Defaults on Payday Loans I’m sure they are. But that’s not the question! The question is whether defaulters have worse declines in credit scores after default than similarly situated non-defaulters. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:28 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans Sure - The pattern that I think is emerging (but I will verify) is that bad credit risks are bad credit risks - across all products. I will be more responsive this time. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:25:43 PM Subject: RE: Defaults on Payday Loans This is useful, but for the most part it answers the wrong question. In the last few paragraphs, it begins to zero in on the issue we care about. As a reminder, we are not interested in predicting defaults, or in who defaults. Rather, we are investigating whether the fact of having defaulted makes a difference to a consumer’s welfare after the default. We are making this because the CFPB has asserted that defaults are harmful to consumers, which really seems unlikely given that the consequences of most defaults are that the borrower retains the loan proceeds without being subject to collection action and without any bureau derogatory report. So, it would be useful to look at changes in credit scores (or other outcome variables, such as delinquencies on other debts, which are likely to be similar) in the time following default. Perhaps we could compare these changes with the changes in scores of non-defaulters with similar initial credit scores. Would you mind taking another stab at this, please? Sorry if we miscommunicated about it. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:19 PM To: Hilary B. Miller Subject: Defaults on Payday Loans Hi Hilary I owe you a huge apology. I had a panic attack this morning when I realized that I forgot to send this to you before I went away last week. I have also been heads down working with a group of trial lawyers - using social media data to try to predict jury verdicts...I am so sorry that this is so late... I actually set a grad student loose on the default and payday loan variables to see if he could find something clever. As you will see, his findings are fairly straight forward...as the number of payday loans taken out increases, the probability of a default decreases. In addition, at the end, you will see that vantage scores are lower for people who default. Sorry again for forgetting to get this over to you. Let me know if you would like to discuss. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Im on it. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:36:14 PM Subject: RE: Defaults on Payday Loans Yes. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:35 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans This is a mock up (not real data)...is this aligned with what you are thinking? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:29:37 PM Subject: RE: Defaults on Payday Loans I’m sure they are. But that’s not the question! The question is whether defaulters have worse declines in credit scores after default than similarly situated non-defaulters. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:28 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans Sure - The pattern that I think is emerging (but I will verify) is that bad credit risks are bad credit risks - across all products. I will be more responsive this time. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:25:43 PM Subject: RE: Defaults on Payday Loans This is useful, but for the most part it answers the wrong question. In the last few paragraphs, it begins to zero in on the issue we care about. As a reminder, we are not interested in predicting defaults, or in who defaults. Rather, we are investigating whether the fact of having defaulted makes a difference to a consumer’s welfare after the default. We are making this because the CFPB has asserted that defaults are harmful to consumers, which really seems unlikely given that the consequences of most defaults are that the borrower retains the loan proceeds without being subject to collection action and without any bureau derogatory report. So, it would be useful to look at changes in credit scores (or other outcome variables, such as delinquencies on other debts, which are likely to be similar) in the time following default. Perhaps we could compare these changes with the changes in scores of non-defaulters with similar initial credit scores. Would you mind taking another stab at this, please? Sorry if we miscommunicated about it. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:19 PM To: Hilary B. Miller Subject: Defaults on Payday Loans Hi Hilary I owe you a huge apology. I had a panic attack this morning when I realized that I forgot to send this to you before I went away last week. I have also been heads down working with a group of trial lawyers - using social media data to try to predict jury verdicts...I am so sorry that this is so late... I actually set a grad student loose on the default and payday loan variables to see if he could find something clever. As you will see, his findings are fairly straight forward...as the number of payday loans taken out increases, the probability of a default decreases. In addition, at the end, you will see that vantage scores are lower for people who default. Sorry again for forgetting to get this over to you. Let me know if you would like to discuss. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? This is a mock up (not real data)...is this aligned with what you are thinking? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:29:37 PM Subject: RE: Defaults on Payday Loans I’m sure they are. But that’s not the question! The question is whether defaulters have worse declines in credit scores after default than similarly situated non-defaulters. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:28 PM To: Hilary B. Miller Subject: Re: Defaults on Payday Loans Sure - The pattern that I think is emerging (but I will verify) is that bad credit risks are bad credit risks - across all products. I will be more responsive this time. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, June 25, 2014 12:25:43 PM Subject: RE: Defaults on Payday Loans This is useful, but for the most part it answers the wrong question. In the last few paragraphs, it begins to zero in on the issue we care about. As a reminder, we are not interested in predicting defaults, or in who defaults. Rather, we are investigating whether the fact of having defaulted makes a difference to a consumer’s welfare after the default. We are making this because the CFPB has asserted that defaults are harmful to consumers, which really seems unlikely given that the consequences of most defaults are that the borrower retains the loan proceeds without being subject to collection action and without any bureau derogatory report. So, it would be useful to look at changes in credit scores (or other outcome variables, such as delinquencies on other debts, which are likely to be similar) in the time following default. Perhaps we could compare these changes with the changes in scores of non-defaulters with similar initial credit scores. Would you mind taking another stab at this, please? Sorry if we miscommunicated about it. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, June 25, 2014 12:19 PM To: Hilary B. Miller Subject: Defaults on Payday Loans Hi Hilary I owe you a huge apology. I had a panic attack this morning when I realized that I forgot to send this to you before I went away last week. I have also been heads down working with a group of trial lawyers - using social media data to try to predict jury verdicts...I am so sorry that this is so late... I actually set a grad student loose on the default and payday loan variables to see if he could find something clever. As you will see, his findings are fairly straight forward...as the number of payday loans taken out increases, the probability of a default decreases. In addition, at the end, you will see that vantage scores are lower for people who default. Sorry again for forgetting to get this over to you. Let me know if you would like to discuss. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I just read the edits - fine with me. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, June 10, 2014 10:19:44 AM Subject: Paper with Trans Union comments Jen – See the attachment. The few changes are on pp. 11-12 only. HM Hi Hilary - my 2:00 meeting was cancelled today - so I have time to speak today between 1:30 and 3:00 if that works better than 9 - 10. Just let me know. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - my 2:00 meeting was cancelled today - so I have time to speak today between 1:30 and 3:00 if that works better than 9 - 10. Just let me know. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Mr. Baines Hilary Miller (copied here) has asked me to send you a dataset that he and I have been working with over the last few months. The data is too large to email. So I wanted to check with you regarding your preferred method of receipt. I recently sent the data as SAS Files on two CDs via Federal Express to another client. If that works for you, please forward your mailing address and I will get the data out to you this week. If you have an alternative method of receipt, just let me know. Kindest Regards. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? All of that sounds fine. Just let me know when you want to connect. I will be in and out of meetings on Monday. Pretty open Tuesday morning. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Saturday, June 7, 2014 4:50:42 PM Subject: RE: Paper Will do. Want to talk to you about the "default" issue and see if we can coordinate with Mann. Also, need to have you send a copy of the dataset to another consultant who will be using it for a completely unrelated purpose. Will send you details on Monday. Thanks so much. Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Saturday, June 07, 2014 11:24 AM To: Hilary B. Miller Subject: Re: Paper I will defer to you. If your travels ever bring you to Atlanta, please do let me know. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Friday, June 6, 2014 6:03:38 PM Subject: RE: Paper Jen -We have not yet provided the paper to the CFPB and would like to continue its embargoed status until the CFPB has had a chance to review and consider it. We are awaiting the right strategic moment to slip it in there. It is, obviously, your paper, but we do not want it to be distributed. I will be happy to send you a "clean" copy in any event. Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, June 06, 2014 9:00 AM To: Hilary B. Miller Subject: Paper Hi Hilary, I trust all is well. I wanted to check in on the status of the paper and to see if I could get a clean (non embargoed) copy. Thanks. Jen Sent from my iPhone I will defer to you. If your travels ever bring you to Atlanta, please do let me know. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Friday, June 6, 2014 6:03:38 PM Subject: RE: Paper Jen -We have not yet provided the paper to the CFPB and would like to continue its embargoed status until the CFPB has had a chance to review and consider it. We are awaiting the right strategic moment to slip it in there. It is, obviously, your paper, but we do not want it to be distributed. I will be happy to send you a "clean" copy in any event. Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, June 06, 2014 9:00 AM To: Hilary B. Miller Subject: Paper Hi Hilary, I trust all is well. I wanted to check in on the status of the paper and to see if I could get a clean (non embargoed) copy. Thanks. Jen Sent from my iPhone Hi Hilary, I trust all is well. I wanted to check in on the status of the paper and to see if I could get a clean (non embargoed) copy. Thanks. Jen Sent from my iPhone Hi Hilary - sorry for the delayed response... First - no changes. I am good with this. I have hired a body guard and hired an accountant. Second - we received the check. thank you. Third - I checked with Ronald - he has the data. I told him I would be happy to help him navigate the data (it took me a few weeks to really become comfortable with it)... let me know if you need anything else from me... Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, May 19, 2014 12:15:28 PM Subject: RE: Edited paper Jen – Here is an edited (not redlined) version of the paper. Minor changes were made at several places. Please let me know what you think of this version and of any further changes you think appropriate. We are continuing to embargo the paper while we consider whether we want to slip it quietly into the CFPB before some more general distribution. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, May 19, 2014 12:09 PM To: Hilary B. Miller Subject: Re: Edited paper Hilary I sent the data disks to Ronald Mann - I will be following up with him today to ensure that everything was received in good order. Is the paper ready? Could I trouble you for a copy? Is there something you would like for me to do in terms of distribution? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, May 12, 2014 12:43:59 PM Subject: Edited paper Jen – My comments (redlined) are noted in the attachment. Please let me know what you think and provide any necessary adjustments. Hilary Data received. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Ronald Mann"  To: "Jennifer Lewis Priestley"  Sent: Monday, May 19, 2014 1:11:56 PM Subject: Re: Following Up Thanks!  CDs received and data transferred off the discs to my computer with no  apparent problems!  Will let you know if I have any trouble, and thanks so much for your generosity in sharing the results of your labor. On Mon, May 19, 2014 at 1:09 PM, Jennifer Lewis  Priestley  wrote: Hi Ronald. I understand that the CDs were delivered on Friday. I want to ensure that the data was accessible and everything was received in good order. Just let me know if I can help in any way. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Spam Not spam Forget previous vote ­­  Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570 Hilary I sent the data disks to Ronald Mann - I will be following up with him today to ensure that everything was received in good order. Is the paper ready? Could I trouble you for a copy? Is there something you would like for me to do in terms of distribution? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, May 12, 2014 12:43:59 PM Subject: Edited paper Jen – My comments (redlined) are noted in the attachment. Please let me know what you think and provide any necessary adjustments. Hilary Great. Thanks for saving me from embarrassing myself. :) Sent from my iPhone On May 12, 2014, at 11:22 AM, "Hilary B. Miller" wrote: Yes. He may need some handholding, which I would appreciate greatly – don’t kill yourself. He goes by “Ronald,” not “Ron.” :) From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, May 12, 2014 11:22 AM To: Hilary B. Miller Subject: Re: Three things Hilary Attached is the invoice. I will reach out to Ron Mann early tomorrow. I assume he just needs the raw/merged dataset? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, May 12, 2014 10:22:35 AM Subject: Three things Jen – 1. I am working on an edited version of your paper. I should have it done today. I will send it back to you for your further review, but I think this is very nearly the end. 2. I have spoken with Ronald Mann about some of the default-related issues we unearthed in this database. He is an extremely sharp guy and he would be a great collaborator with you on the “second study,” if you would be interested in working with him. In any event, I would like to share the combined dataset with him. Would you please arrange to send him a link or other means by which he can FTP or download it? His email address is rmann@law.columbia.edu. 3. Finally, since you have mercifully not already done so, would you please bill CCRF for the “first study”? We’ll get that paid right away. Thanks. Regards, Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Hilary Attached is the invoice. I will reach out to Ron Mann early tomorrow. I assume he just needs the raw/merged dataset? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, May 12, 2014 10:22:35 AM Subject: Three things Jen – 1. I am working on an edited version of your paper. I should have it done today. I will send it back to you for your further review, but I think this is very nearly the end. 2. I have spoken with Ronald Mann about some of the default-related issues we unearthed in this database. He is an extremely sharp guy and he would be a great collaborator with you on the “second study,” if you would be interested in working with him. In any event, I would like to share the combined dataset with him. Would you please arrange to send him a link or other means by which he can FTP or download it? His email address is rmann@law.columbia.edu. 3. Finally, since you have mercifully not already done so, would you please bill CCRF for the “first study”? We’ll get that paid right away. Thanks. Regards, Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. All of that sounds fine. I will get an invoice out today. Shall I email it? Or is there an address where I should mail it formally? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, May 12, 2014 10:22:35 AM Subject: Three things Jen – 1. I am working on an edited version of your paper. I should have it done today. I will send it back to you for your further review, but I think this is very nearly the end. 2. I have spoken with Ronald Mann about some of the default-related issues we unearthed in this database. He is an extremely sharp guy and he would be a great collaborator with you on the “second study,” if you would be interested in working with him. In any event, I would like to share the combined dataset with him. Would you please arrange to send him a link or other means by which he can FTP or download it? His email address is rmann@law.columbia.edu. 3. Finally, since you have mercifully not already done so, would you please bill CCRF for the “first study”? We’ll get that paid right away. Thanks. Regards, Hilary Hilary B. Miller • Law offices of Hilary B. Miller • 500 West Putnam Avenue Suite 400 • Greenwich, Connecticut 06830-6096 • voice: (203) 399-1320 • fax: (914) 206-3727 • hilary@miller.net • bio • v-card download This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. Hi Hilary I think I incorporated all of the comments. Take a look and let me know what you think. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Thursday, May 1, 2014 3:02:23 PM Subject: FW: New paper I’m glad we didn’t wait too long to get these comments – they are not that helpful. We’re not going to start from scratch. Take what you want from them. HM From: Gregory Elliehausen [mailto:gregory.elliehausen@frb.gov] Sent: Thursday, May 01, 2014 3:00 PM To: Hilary B. Miller Subject: RE: New paper My review is attached. One of the tables was not formatted correctly in the last version. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Sorry for the delay. I spent some time over the weekend and today making some edits to reflect the reviewers comments. I think the paper flows and reads better now (I hate it when reviewers are right). :) See attached. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, April 29, 2014 5:08:57 PM Subject: RE: Dataset question Yeah, this is what I have. I don’t have definitions for the “customer input” fields (i.e., the ones that didn’t come from Trans Union). I’m trying to replicate these fields and don’t know how they are defined. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, April 29, 2014 5:00 PM To: Hilary B. Miller Subject: Re: Dataset question See if this is what you are looking for. I will get a revised draft to you this evening. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, April 29, 2014 4:49:34 PM Subject: Dataset question Jen – As part of the materials you received from me or otherwise, do you have a dictionary for the data fields that were supplied by the lenders (they were labeled “Customer Input”)? If so, would you send me what you have or can find, please? Hilary Call me and let me see if I can help. 404-229-3216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, April 29, 2014 5:08:57 PM Subject: RE: Dataset question Yeah, this is what I have. I don’t have definitions for the “customer input” fields (i.e., the ones that didn’t come from Trans Union). I’m trying to replicate these fields and don’t know how they are defined. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, April 29, 2014 5:00 PM To: Hilary B. Miller Subject: Re: Dataset question See if this is what you are looking for. I will get a revised draft to you this evening. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, April 29, 2014 4:49:34 PM Subject: Dataset question Jen – As part of the materials you received from me or otherwise, do you have a dictionary for the data fields that were supplied by the lenders (they were labeled “Customer Input”)? If so, would you send me what you have or can find, please? Hilary See if this is what you are looking for. I will get a revised draft to you this evening. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, April 29, 2014 4:49:34 PM Subject: Dataset question Jen – As part of the materials you received from me or otherwise, do you have a dictionary for the data fields that were supplied by the lenders (they were labeled “Customer Input”)? If so, would you send me what you have or can find, please? Hilary Excellent. I will be in my office. 770-423-6107. Just call when you are ready. Sent from my iPhone On Apr 22, 2014, at 5:42 PM, "Hilary B. Miller" wrote: I have a 9-11. Let’s try to speak at 11. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, April 22, 2014 5:06 PM To: Hilary B. Miller Subject: Re: Priestley - Rollovers Ok. thanks. Do you have some time to talk tomorrow? I have availability 10:30 - 12. Then in the late afternoon - between 4 and 5:30 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, April 22, 2014 4:00:25 PM Subject: FW: Priestley ­ Rollovers Here are Victor’s comments. They are more comprehensive, more oriented toward the analytical presentation, and more useful than Ronald’s -- but also more daunting to implement. On reflection, I agree with him regarding the value of reversing the order of the two principal findings. As with the previous comments, view these as suggestions rather than commands. Feel free to email or call him directly if you want to discuss it with him. I am still awaiting another set of comments from Greg Elliehausen. Based on previous work with him, I don’t expect anything soon. If you are comfortable doing so, please dive in with what you’ve got and we can fill in from there. From: Victor Stango [mailto:vstango@ucdavis.edu] Sent: Tuesday, April 22, 2014 3:00 PM To: Hilary B. Miller Subject: Re: Priestley - Rollovers Here you go. Let me know if you want to discuss. From: "Hilary B. Miller" Date: Mon, 7 Apr 2014 14:53:22 -0400 To: Victor Stango Subject: Priestley - Rollovers Victor — This is a review draft and I would greatly appreciate your comments, as we discussed. Thank you! Regards, Hilary Ok. thanks. Do you have some time to talk tomorrow? I have availability 10:30 - 12. Then in the late afternoon - between 4 and 5:30 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Tuesday, April 22, 2014 4:00:25 PM Subject: FW: Priestley ­ Rollovers Here are Victor’s comments. They are more comprehensive, more oriented toward the analytical presentation, and more useful than Ronald’s -- but also more daunting to implement. On reflection, I agree with him regarding the value of reversing the order of the two principal findings. As with the previous comments, view these as suggestions rather than commands. Feel free to email or call him directly if you want to discuss it with him. I am still awaiting another set of comments from Greg Elliehausen. Based on previous work with him, I don’t expect anything soon. If you are comfortable doing so, please dive in with what you’ve got and we can fill in from there. From: Victor Stango [mailto:vstango@ucdavis.edu] Sent: Tuesday, April 22, 2014 3:00 PM To: Hilary B. Miller Subject: Re: Priestley - Rollovers Here you go. Let me know if you want to discuss. From: "Hilary B. Miller" Date: Mon, 7 Apr 2014 14:53:22 -0400 To: Victor Stango Subject: Priestley - Rollovers Victor — This is a review draft and I would greatly appreciate your comments, as we discussed. Thank you! Regards, Hilary I ran the models and the relevant analysis with the 14 day definition - no changes to relative position or to significance. This is logical and makes sense given that over half of all rollovers take place within the first 2 days. I am appending this analysis to an Appendix. I will edit the text as appropriate and get a revision to you tomorrow evening (the writing always takes me a lot longer than the analysis). Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, April 21, 2014 12:40:20 PM Subject: RE: Rollover Impact 1. With regard to the first two rows of this table, Sandler explains where the data came from (and some limits on obtaining a full sample from certain states and one operator). You can paraphrase this information from her paper. 2. For the paper to have maximum usefulness, you should consider re-running the principal analyses using a 14-day rollover definition and reporting the results as an alternative finding in an appendix (as Fusaro and Cirillo do). Mann is correct that, in light of the CFPB’s “Data Point,” policymakers are going to be almost single-mindedly focused on a 14-day “lookback” period. 3. The fees in the table are generally correct. FL should be $15/100 (possibly with a footnote); TX is unlimited. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, April 21, 2014 10:28 AM To: Hilary B. Miller Subject: Re: Rollover Impact Hi Hilary I went through all of Mann's comments...and provided my thoughts in the attached. I make reference to a v9 of the paper in the interest of "version control" I can let you be the keeper of the working draft...but I was making some edits in the paper to reflect some of Mann's comments. We can use v9 (I will send later today) or I can just put them in red and you can pick them up as needed. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you:   1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff.   2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore.   3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that.   4. Let's wait for the remaining comments before we start any drafting.   I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this.   HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss.     ­­ Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212-854-1570 Will do. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, April 21, 2014 1:13:25 PM Subject: Re: Rollover Impact Yes. Also any of the other analyses that you think should be alternatively stated on this  basis.  From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Monday, April 21, 2014 at 1:10 PM To: Hilary Miller Subject: Re: Rollover Impact thats fine. I will run the GEEs with a 14 day defined Rollover and add this to an Appendix. Right? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, April 21, 2014 12:40:20 PM Subject: RE: Rollover Impact 1. With regard to the first two rows of this table, Sandler explains where the data came from (and some limits on obtaining a full sample from certain states and one operator). You can paraphrase this information from her paper. 2. For the paper to have maximum usefulness, you should consider re-running the principal analyses using a 14-day rollover definition and reporting the results as an alternative finding in an appendix (as Fusaro and Cirillo do). Mann is correct that, in light of the CFPB’s “Data Point,” policymakers are going to be almost single-mindedly focused on a 14-day “lookback” period. 3. The fees in the table are generally correct. FL should be $15/100 (possibly with a footnote); TX is unlimited. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, April 21, 2014 10:28 AM To: Hilary B. Miller Subject: Re: Rollover Impact Hi Hilary - I went through all of Mann's comments...and provided my thoughts in the attached. I make reference to a v9 of the paper in the interest of "version control" I can let you be the keeper of the working draft...but I was making some edits in the paper to reflect some of Mann's comments. We can use v9 (I will send later today) or I can just put them in red and you can pick them up as needed. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you:   1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff.   2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore.   3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that.   4. Let's wait for the remaining comments before we start any drafting.   I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this.   HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss.     ­­ Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570   thats fine. I will run the GEEs with a 14 day defined Rollover and add this to an Appendix. Right? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, April 21, 2014 12:40:20 PM Subject: RE: Rollover Impact 1. With regard to the first two rows of this table, Sandler explains where the data came from (and some limits on obtaining a full sample from certain states and one operator). You can paraphrase this information from her paper. 2. For the paper to have maximum usefulness, you should consider re-running the principal analyses using a 14-day rollover definition and reporting the results as an alternative finding in an appendix (as Fusaro and Cirillo do). Mann is correct that, in light of the CFPB’s “Data Point,” policymakers are going to be almost single-mindedly focused on a 14-day “lookback” period. 3. The fees in the table are generally correct. FL should be $15/100 (possibly with a footnote); TX is unlimited. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, April 21, 2014 10:28 AM To: Hilary B. Miller Subject: Re: Rollover Impact Hi Hilary I went through all of Mann's comments...and provided my thoughts in the attached. I make reference to a v9 of the paper in the interest of "version control" I can let you be the keeper of the working draft...but I was making some edits in the paper to reflect some of Mann's comments. We can use v9 (I will send later today) or I can just put them in red and you can pick them up as needed. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you:   1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff.   2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore.   3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that.   4. Let's wait for the remaining comments before we start any drafting.   I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this.   HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss.     ­­ Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570   Hi Hilary I went through all of Mann's comments...and provided my thoughts in the attached. I make reference to a v9 of the paper - in the interest of "version control" I can let you be the keeper of the working draft...but I was making some edits in the paper to reflect some of Mann's comments. We can use v9 (I will send later today) or I can just put them in red and you can pick them up as needed. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you: 1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff. 2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore. 3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that. 4. Let's wait for the remaining comments before we start any drafting. I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this. HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss. ­­  Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570 Read through his comments - all easy enough. I will go through and respond to the notes in a word document this evening. Then we can just copy/paste the responses (with appropriate edits) into the paper. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you: 1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff. 2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore. 3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that. 4. Let's wait for the remaining comments before we start any drafting. I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this. HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss. ­­  Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570 Understood. I await the academic review that says "The paper is perfect. I agree with everything - please do not make any changes" :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley Ph. D."  Sent: Thursday, April 10, 2014 10:09:22 AM Subject: Fwd: Rollover Impact Here are Ronald's comments. Before you read them, I have some thoughts for you: 1. Many of his comments are short, simple fixes. For example, we need to describe how the lender dataset was selected, how the 29K survivors are representative of the  original 37K, and and expand the robustness test of 2 vs 14 days. Easy stuff. 2. Just ignore his comments about Caskey and the big picture question. He doesn't get  it. That's not what we sought to study and I don't think it matters as a policy matter  anymore. 3. His most important comment relates to Florida. We need to supplement the  discussion re 2006 and explain away these results. There is a large unobserved  process going on here. Let's talk about how to address that. 4. Let's wait for the remaining comments before we start any drafting. I appreciate getting these comments from him. As an academic scholar, I'm sure you're  impervious to this kind of feedback ­­ all in the interest of better science. Thanks again  for all your hard work on this. HM   Begin forwarded message: From: Ronald Mann  Date: April 10, 2014 at 4:40:22 AM PDT To: "Hilary B. Miller"  Subject: Rollover Impact Some comments are attached.  Standing by to discuss. ­­  Ronald Mann Albert E. Cinelli Enterprise Professor of Law Columbia Law School 435 W. 116th Street New York, NY 10027 rmann@law.columbia.edu 212­854­1570 I trust he was generally pleased? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley Ph. D." Sent: Thursday, April 10, 2014 9:35:01 AM Subject: Paper Jen -I have sent your paper to a few reviewers in confidence -- Ronald Mann, Victor Stango and Gregory Elliehausen. I just received Mann's comments and will pass them along to you in a separate email. H _______________________________ Hilary B. Miller 500 West Putnam Avenue - Suite 400 Greenwich, Connecticut 06830-6096 (203) 399-1320 (voice) (203) 517-6859 (cell) (914) 206-3727 (fax) (sent from iPad) See attached. I made a small edit to f.11 on page 13 and then the larger addition to f.12 on page 14. I am about to go offline for a few hours - but I will be back online this evening. You can always call me at 404-229-3216 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, April 7, 2014 1:08:09 PM Subject: Re: A couple more questions ... Yes, I would add this to the “robust” footnote. From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Monday, April 7, 2014 at 1:05 PM To: Hilary Miller Subject: Re: A couple more questions ... Hilary I am still reading through this - I have a few very minor edits. But I wanted to share something with you that I thought might be relevant. Specifically, on page 14 - f.12... I discovered that - Of all borrowers who are defined as having a "rollover" under the 14 Day definition, 55% took their rollover within two days, and 94% took their rollover within 7 days. So, under our definition, we are still capturing well over half of all rollovers under a commonly accepted, but much looser definition. It also highlights that if people are going to do rollover, they do it right away - which I believe supports Mann's findings. If you think this is noteworthy, I will append relevant verbiage to f.12 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, April 7, 2014 10:46:06 AM Subject: RE: A couple more questions ... Here is a revised draft (both redlined and clean versions). Please advise any additional comments. HM From: Hilary B. Miller Sent: Saturday, April 05, 2014 10:39 PM To: Jennifer Lewis Priestley Subject: Re: A couple more questions ... This is great. I’ll get you another draft tomorrow. Thank you! From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Saturday, April 5, 2014 at 8:46 PM To: Hilary Miller Subject: Re: A couple more questions ... Hi Hilary I provided my thoughts on your points below. If you want me to append them to the doc, just let me know: 1. We have retained the language in the body (now deleted from the abstract) that “…longer-term borrowers have better outcomes than consumers whose borrowing is restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days” came from originally, since there is no state with a 30day limitation. Let’s delete that here, okay? Is there some better or more accurately descriptive way we can explain this? I like the generality, so I’m not pressing for more detail. FROM JLP: I agree. I believe this was left over from an earlier iteration. Since we restricted a rollover to <= 2days, this came in "under the radar" for all states, but I don't think we need to get into that discussion here. 2. This raises a somewhat larger question, which is how we chose to classify states as “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft if you agree. FROM JLP: Conceptually I agree with you - but we never actually put the states into distinct categories. We only make reference to Texas as the least restrictive (I suppose that would be a category of 1). And then more generally make references to "states with strict rollover restrictions" (Florida, Kansas, Oklahoma). Or states with database requirements (Florida and Oklahoma). I actually like the looser comparisons, because it allows us more flexibility in our discussions - like with California - without boxing ourselves in a corner and then having to defend why we are countering our own position. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean. Do you concur? JLP: Its hard to say. Looking at Figure 3 in his paper, the distribution is heavily skewed with values out to 250+. A mean calculation is very sensitive to outliers - where a median is not. I would expect that you are correct - the value of 36 appears to be a median rather than a mean - which I would expect would be much higher. But, ultimately without the actual data or the descriptive stats, we cant know for certain. 4. I see that you added additional language about how VantageScores treat mortgages, but I’m not sure that this added language is relevant to this market (which you seem to point out in the parenthetical). JLP: I added the verbiage because it was identified as a regularly referenced difference between the two scores. Ultimately, I would expect that relatively few people in this segment would have a mortgage. So, while I don't think this point is directly relevant to the paper, it generally explains a frequently cited distinction. 5. I’m going to draft a footnote regarding the metric you created, which is percentage of loans rolled over. I think this needs explanation and it seems like a very simple and powerful tool. JLP: Yep. Given that we needed this value for each borrower in the dataset, this was the only way I knew to do it. So, again, I took the total number of loans (INPUT_SEQNUM) for each customer (KEYFLAG). This was the denominator. Then, I determined how many of those loans were "rollovers" defined as <=2 days between the date paid for the previous loan and the date opened for the next loan. The number of loans that met this definition became the total number of rollovers for each customer. The ratio of the two values was the percent of loans rolled over by customer. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to subsets of 29K. What should we say about that? JLP: All analytical software works on a "complete case basis" for multivariate analysis (BTW - I used BASE SAS v9.3 if that is needed). So, what this means is that we could have 37,000 borrowers, but only 29,000 had populated values for each variable required for analysis. If 10 variables are included in something like the GEE, and an obs only has 9 of the variables populated, the entire obs gets dropped because it is not a "complete case". In the end, the borrowers that dropped out were not statistically different (looking at credit score) from the ones that were retained for modeling. I could have imputed missing values - but given the overall number of obs, I did not think it was necessary. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Priestley"  Sent: Saturday, April 5, 2014 10:11:58 AM Subject: A couple more questions ... I tend to mull, re­read, and fidget a lot. Here are a couple more questions for you:   1. We have retained the language in the body (now deleted from the abstract) that “… longer­term borrowers have better outcomes than consumers whose borrowing is  restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days”  came from originally, since there is no state with a 30­day limitation. Let’s delete that  here, okay? Is there some better or more accurately descriptive way we can explain  this? I like the generality, so I’m not pressing for more detail.   2. This raises a somewhat larger question, which is how we chose to classify states as  “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how  they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft  if you agree.   3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean.  Do you concur?   4. I see that you added additional language about how VantageScores treat mortgages,  but I’m not sure that this added language is relevant to this market (which you seem to  point out in the parenthetical).   5. I’m going to draft a footnote regarding the metric you created, which is percentage of  loans rolled over. I think this needs explanation and it seems like a very simple and  powerful tool.   6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to  subsets of 29K. What should we say about that?   Will await your response and incorporate these matters in a redlined revised draft  later today or tomorrow.   Hilary Hilary I am still reading through this - I have a few very minor edits. But I wanted to share something with you that I thought might be relevant. Specifically, on page 14 - f.12... I discovered that Of all borrowers who are defined as having a "rollover" under the 14 Day definition, 55% took their rollover within two days, and 94% took their rollover within 7 days. So, under our definition, we are still capturing well over half of all rollovers under a commonly accepted, but much looser definition. It also highlights that if people are going to do rollover, they do it right away - which I believe supports Mann's findings. If you think this is noteworthy, I will append relevant verbiage to f.12 Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Monday, April 7, 2014 10:46:06 AM Subject: RE: A couple more questions ... Here is a revised draft (both redlined and clean versions). Please advise any additional comments. HM From: Hilary B. Miller Sent: Saturday, April 05, 2014 10:39 PM To: Jennifer Lewis Priestley Subject: Re: A couple more questions ... This is great. I’ll get you another draft tomorrow. Thank you! From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Saturday, April 5, 2014 at 8:46 PM To: Hilary Miller Subject: Re: A couple more questions ... Hi Hilary I provided my thoughts on your points below. If you want me to append them to the doc, just let me know: 1. We have retained the language in the body (now deleted from the abstract) that “…longer-term borrowers have better outcomes than consumers whose borrowing is restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days” came from originally, since there is no state with a 30day limitation. Let’s delete that here, okay? Is there some better or more accurately descriptive way we can explain this? I like the generality, so I’m not pressing for more detail. FROM JLP: I agree. I believe this was left over from an earlier iteration. Since we restricted a rollover to <= 2days, this came in "under the radar" for all states, but I don't think we need to get into that discussion here. 2. This raises a somewhat larger question, which is how we chose to classify states as “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft if you agree. FROM JLP: Conceptually I agree with you - but we never actually put the states into distinct categories. We only make reference to Texas as the least restrictive (I suppose that would be a category of 1). And then more generally make references to "states with strict rollover restrictions" (Florida, Kansas, Oklahoma). Or states with database requirements (Florida and Oklahoma). I actually like the looser comparisons, because it allows us more flexibility in our discussions - like with California - without boxing ourselves in a corner and then having to defend why we are countering our own position. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean. Do you concur? JLP: Its hard to say. Looking at Figure 3 in his paper, the distribution is heavily skewed with values out to 250+. A mean calculation is very sensitive to outliers - where a median is not. I would expect that you are correct - the value of 36 appears to be a median rather than a mean - which I would expect would be much higher. But, ultimately without the actual data or the descriptive stats, we cant know for certain. 4. I see that you added additional language about how VantageScores treat mortgages, but I’m not sure that this added language is relevant to this market (which you seem to point out in the parenthetical). JLP: I added the verbiage because it was identified as a regularly referenced difference between the two scores. Ultimately, I would expect that relatively few people in this segment would have a mortgage. So, while I don't think this point is directly relevant to the paper, it generally explains a frequently cited distinction. 5. I’m going to draft a footnote regarding the metric you created, which is percentage of loans rolled over. I think this needs explanation and it seems like a very simple and powerful tool. JLP: Yep. Given that we needed this value for each borrower in the dataset, this was the only way I knew to do it. So, again, I took the total number of loans (INPUT_SEQNUM) for each customer (KEYFLAG). This was the denominator. Then, I determined how many of those loans were "rollovers" defined as <=2 days between the date paid for the previous loan and the date opened for the next loan. The number of loans that met this definition became the total number of rollovers for each customer. The ratio of the two values was the percent of loans rolled over by customer. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to subsets of 29K. What should we say about that? JLP: All analytical software works on a "complete case basis" for multivariate analysis (BTW - I used BASE SAS v9.3 if that is needed). So, what this means is that we could have 37,000 borrowers, but only 29,000 had populated values for each variable required for analysis. If 10 variables are included in something like the GEE, and an obs only has 9 of the variables populated, the entire obs gets dropped because it is not a "complete case". In the end, the borrowers that dropped out were not statistically different (looking at credit score) from the ones that were retained for modeling. I could have imputed missing values - but given the overall number of obs, I did not think it was necessary. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Priestley"  Sent: Saturday, April 5, 2014 10:11:58 AM Subject: A couple more questions ... I tend to mull, re­read, and fidget a lot. Here are a couple more questions for you:   1. We have retained the language in the body (now deleted from the abstract) that “… longer­term borrowers have better outcomes than consumers whose borrowing is  restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days”  came from originally, since there is no state with a 30­day limitation. Let’s delete that  here, okay? Is there some better or more accurately descriptive way we can explain  this? I like the generality, so I’m not pressing for more detail.   2. This raises a somewhat larger question, which is how we chose to classify states as  “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how  they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft  if you agree.   3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean.  Do you concur?   4. I see that you added additional language about how VantageScores treat mortgages,  but I’m not sure that this added language is relevant to this market (which you seem to  point out in the parenthetical).   5. I’m going to draft a footnote regarding the metric you created, which is percentage of  loans rolled over. I think this needs explanation and it seems like a very simple and  powerful tool.   6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to  subsets of 29K. What should we say about that?   Will await your response and incorporate these matters in a redlined revised draft  later today or tomorrow.   Hilary Hi Hilary I provided my thoughts on your points below. If you want me to append them to the doc, just let me know: 1. We have retained the language in the body (now deleted from the abstract) that “…longer-term borrowers have better outcomes than consumers whose borrowing is restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days” came from originally, since there is no state with a 30day limitation. Let’s delete that here, okay? Is there some better or more accurately descriptive way we can explain this? I like the generality, so I’m not pressing for more detail. FROM JLP: I agree. I believe this was left over from an earlier iteration. Since we restricted a rollover to <= 2days, this came in "under the radar" for all states, but I don't think we need to get into that discussion here. 2. This raises a somewhat larger question, which is how we chose to classify states as “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft if you agree. FROM JLP: Conceptually I agree with you - but we never actually put the states into distinct categories. We only make reference to Texas as the least restrictive (I suppose that would be a category of 1). And then more generally make references to "states with strict rollover restrictions" (Florida, Kansas, Oklahoma). Or states with database requirements (Florida and Oklahoma). I actually like the looser comparisons, because it allows us more flexibility in our discussions - like with California - without boxing ourselves in a corner and then having to defend why we are countering our own position. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean. Do you concur? JLP: Its hard to say. Looking at Figure 3 in his paper, the distribution is heavily skewed with values out to 250+. A mean calculation is very sensitive to outliers - where a median is not. I would expect that you are correct - the value of 36 appears to be a median rather than a mean - which I would expect would be much higher. But, ultimately without the actual data or the descriptive stats, we cant know for certain. 4. I see that you added additional language about how VantageScores treat mortgages, but I’m not sure that this added language is relevant to this market (which you seem to point out in the parenthetical). JLP: I added the verbiage because it was identified as a regularly referenced difference between the two scores. Ultimately, I would expect that relatively few people in this segment would have a mortgage. So, while I don't think this point is directly relevant to the paper, it generally explains a frequently cited distinction. 5. I’m going to draft a footnote regarding the metric you created, which is percentage of loans rolled over. I think this needs explanation and it seems like a very simple and powerful tool. JLP: Yep. Given that we needed this value for each borrower in the dataset, this was the only way I knew to do it. So, again, I took the total number of loans (INPUT_SEQNUM) for each customer (KEYFLAG). This was the denominator. Then, I determined how many of those loans were "rollovers" defined as <=2 days between the date paid for the previous loan and the date opened for the next loan. The number of loans that met this definition became the total number of rollovers for each customer. The ratio of the two values was the percent of loans rolled over by customer. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to subsets of 29K. What should we say about that? JLP: All analytical software works on a "complete case basis" for multivariate analysis (BTW - I used BASE SAS v9.3 if that is needed). So, what this means is that we could have 37,000 borrowers, but only 29,000 had populated values for each variable required for analysis. If 10 variables are included in something like the GEE, and an obs only has 9 of the variables populated, the entire obs gets dropped because it is not a "complete case". In the end, the borrowers that dropped out were not statistically different (looking at credit score) from the ones that were retained for modeling. I could have imputed missing values - but given the overall number of obs, I did not think it was necessary. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Priestley"  Sent: Saturday, April 5, 2014 10:11:58 AM Subject: A couple more questions ... I tend to mull, re­read, and fidget a lot. Here are a couple more questions for you: 1. We have retained the language in the body (now deleted from the abstract) that “… longer­term borrowers have better outcomes than consumers whose borrowing is  restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days”  came from originally, since there is no state with a 30­day limitation. Let’s delete that  here, okay? Is there some better or more accurately descriptive way we can explain  this? I like the generality, so I’m not pressing for more detail. 2. This raises a somewhat larger question, which is how we chose to classify states as  “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how  they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft  if you agree. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean.  Do you concur? 4. I see that you added additional language about how VantageScores treat mortgages,  but I’m not sure that this added language is relevant to this market (which you seem to  point out in the parenthetical). 5. I’m going to draft a footnote regarding the metric you created, which is percentage of  loans rolled over. I think this needs explanation and it seems like a very simple and  powerful tool. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to  subsets of 29K. What should we say about that? Will await your response and incorporate these matters in a redlined revised draft  later today or tomorrow. Hilary :) Sent from my iPhone On Apr 5, 2014, at 10:48 AM, "Hilary B. Miller" wrote: No rush. You are always so prompt with turnaround. From: Jennifer Priestley Date: Saturday, April 5, 2014 at 10:44 AM To: Hilary Miller Subject: Re: A couple more questions ... All of those points are easy enough to address...I have a few kid things to do this morning...will get back to you this afternoon. Sent from my iPhone On Apr 5, 2014, at 10:12 AM, "Hilary B. Miller" wrote: I tend to mull, re-read, and fidget a lot. Here are a couple more questions for you: 1. We have retained the language in the body (now deleted from the abstract) that “… longer-term borrowers have better outcomes than consumers whose borrowing is restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days” came from originally, since there is no state with a 30-day limitation. Let’s delete that here, okay? Is there some better or more accurately descriptive way we can explain this? I like the generality, so I’m not pressing for more detail. 2. This raises a somewhat larger question, which is how we chose to classify states as “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft if you agree. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean. Do you concur? 4. I see that you added additional language about how VantageScores treat mortgages, but I’m not sure that this added language is relevant to this market (which you seem to point out in the parenthetical). 5. I’m going to draft a footnote regarding the metric you created, which is percentage of loans rolled over. I think this needs explanation and it seems like a very simple and powerful tool. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to subsets of 29K. What should we say about that? Will await your response and incorporate these matters in a redlined revised draft later today or tomorrow. Hilary All of those points are easy enough to address...I have a few kid things to do this morning...will get back to you this afternoon. Sent from my iPhone On Apr 5, 2014, at 10:12 AM, "Hilary B. Miller" wrote: I tend to mull, re-read, and fidget a lot. Here are a couple more questions for you: 1. We have retained the language in the body (now deleted from the abstract) that “… longer-term borrowers have better outcomes than consumers whose borrowing is restricted by law to 30 days or less ….” On reflection, I’m not sure where the “30 days” came from originally, since there is no state with a 30-day limitation. Let’s delete that here, okay? Is there some better or more accurately descriptive way we can explain this? I like the generality, so I’m not pressing for more detail. 2. This raises a somewhat larger question, which is how we chose to classify states as “permissive” or “restrictive,” and some states are arguably ambiguous in terms of how they should be classified. Maybe a paragraph on this topic is warranted. Happy to draft if you agree. 3. I think the “36 day” number from Mann is the median ex ante estimate, not the mean. Do you concur? 4. I see that you added additional language about how VantageScores treat mortgages, but I’m not sure that this added language is relevant to this market (which you seem to point out in the parenthetical). 5. I’m going to draft a footnote regarding the metric you created, which is percentage of loans rolled over. I think this needs explanation and it seems like a very simple and powerful tool. 6. I’m not sure we ever explain how we got from a total sample of 37K borrowers to subsets of 29K. What should we say about that? Will await your response and incorporate these matters in a redlined revised draft later today or tomorrow. Hilary Can I call you? What is a good number? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 4:24:49 PM Subject: Re: Rollover Paper That’s a difficult concept to grasp and one that is not used in this industry. Let’s say I  take one loan and roll it over 4 times. Is that .8? How about if I take out five loans and  roll four of them over once each (but don’t roll over the fifth one)? Is that .8? From: Jennifer Priestley Reply-To: Jennifer Priestley Date: Friday, April 4, 2014 at 4:01 PM To: Hilary Miller Subject: Re: Rollover Paper The computation is actually the average pct of loans rolled over per borrower...by state. So, if an individual took out 10 loans, and 8 were rolled over (under the definition of <=2 days), then their value here is .8. Each individual borrower (keyflag) in the datasets have a pct rolled over value. I needed to get this value per customer (as well as the number of loans rolled over per customer) because I used this later in the modeling process as a predictor. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 3:21:11 PM Subject: RE: Rollover Paper I’m reading the text description of Table 3. The “proportion of loans rolled over by state” – which indeed is represented here – is not a common metric. I think I was confused about it the last time we discussed it and I don’t remember what it means exactly. Is it the proportion of “new” loans (loans that are not rollovers) that are followed by a loan that is deemed (under your 2-day rule) to be a rollover? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 3:10 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary - I made the update to Table 3 - this now reflects the total number of loans across the two time periods. You may recall that Danielle had 431,430 loans reported in Table 8. I have 420,405 in the Jan2006 file and 432,202 in the Jan2008 file (perhaps the 431,430 was an avg), which are all unique loans used for most of the analysis. So, the total now reported in Table 3 reflects the total number of unique loans across the two files (852,607) which was used for a majority of the analysis. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 1:34:21 PM Subject: RE: Rollover Paper I don’t actually see that table in the “legacy” materials. A more useful table would be Table 8 from the “legacy” materials, but restated on the basis of your definition of “rollover” (i.e., 2 days rather than same day). I have made some changes to the paper (only major substantive change is material related to the CSO model and CFSA “best practices.” Here’s the revised draft. Please incorporate a correct table instead of Table 3. Thanks. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 1:27 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary - that was one of the "legacy" tables that I did not create (one of two in the paper - the other is Table 4) - I created the proportion of rollovers column - which I can recreate. For the total number of loans, I have closer to 852,000. Would you like for me to update it? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Friday, April 4, 2014 1:14:38 PM Subject: Rollover Paper Jen I?m looking at Table 3. How could 38,000 borrowers have 7,253,000 loans? Hilary The computation is actually the average pct of loans rolled over per borrower...by state. So, if an individual took out 10 loans, and 8 were rolled over (under the definition of <=2 days), then their value here is .8. Each individual borrower (keyflag) in the datasets have a pct rolled over value. I needed to get this value per customer (as well as the number of loans rolled over per customer) because I used this later in the modeling process as a predictor. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 3:21:11 PM Subject: RE: Rollover Paper I’m reading the text description of Table 3. The “proportion of loans rolled over by state” – which indeed is represented here – is not a common metric. I think I was confused about it the last time we discussed it and I don’t remember what it means exactly. Is it the proportion of “new” loans (loans that are not rollovers) that are followed by a loan that is deemed (under your 2-day rule) to be a rollover? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 3:10 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary I made the update to Table 3 - this now reflects the total number of loans across the two time periods. You may recall that Danielle had 431,430 loans reported in Table 8. I have 420,405 in the Jan2006 file and 432,202 in the Jan2008 file (perhaps the 431,430 was an avg), which are all unique loans used for most of the analysis. So, the total now reported in Table 3 reflects the total number of unique loans across the two files (852,607) which was used for a majority of the analysis. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 1:34:21 PM Subject: RE: Rollover Paper I don’t actually see that table in the “legacy” materials. A more useful table would be Table 8 from the “legacy” materials, but restated on the basis of your definition of “rollover” (i.e., 2 days rather than same day). I have made some changes to the paper (only major substantive change is material related to the CSO model and CFSA “best practices.” Here’s the revised draft. Please incorporate a correct table instead of Table 3. Thanks. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 1:27 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary - that was one of the "legacy" tables that I did not create (one of two in the paper - the other is Table 4) - I created the proportion of rollovers column - which I can recreate. For the total number of loans, I have closer to 852,000. Would you like for me to update it? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Friday, April 4, 2014 1:14:38 PM Subject: Rollover Paper Jen – I’m looking at Table 3. How could 38,000 borrowers have 7,253,000 loans? Hilary Hi Hilary I made the update to Table 3 - this now reflects the total number of loans across the two time periods. You may recall that Danielle had 431,430 loans reported in Table 8. I have 420,405 in the Jan2006 file and 432,202 in the Jan2008 file (perhaps the 431,430 was an avg), which are all unique loans used for most of the analysis. So, the total now reported in Table 3 reflects the total number of unique loans across the two files (852,607) which was used for a majority of the analysis. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 1:34:21 PM Subject: RE: Rollover Paper I don’t actually see that table in the “legacy” materials. A more useful table would be Table 8 from the “legacy” materials, but restated on the basis of your definition of “rollover” (i.e., 2 days rather than same day). I have made some changes to the paper (only major substantive change is material related to the CSO model and CFSA “best practices.” Here’s the revised draft. Please incorporate a correct table instead of Table 3. Thanks. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 1:27 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary - that was one of the "legacy" tables that I did not create (one of two in the paper - the other is Table 4) - I created the proportion of rollovers column - which I can recreate. For the total number of loans, I have closer to 852,000. Would you like for me to update it? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Friday, April 4, 2014 1:14:38 PM Subject: Rollover Paper Jen – I’m looking at Table 3. How could 38,000 borrowers have 7,253,000 loans? Hilary Will do. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, April 4, 2014 1:34:21 PM Subject: RE: Rollover Paper I don’t actually see that table in the “legacy” materials. A more useful table would be Table 8 from the “legacy” materials, but restated on the basis of your definition of “rollover” (i.e., 2 days rather than same day). I have made some changes to the paper (only major substantive change is material related to the CSO model and CFSA “best practices.” Here’s the revised draft. Please incorporate a correct table instead of Table 3. Thanks. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, April 04, 2014 1:27 PM To: Hilary B. Miller Subject: Re: Rollover Paper Hi Hilary - that was one of the "legacy" tables that I did not create (one of two in the paper - the other is Table 4) - I created the proportion of rollovers column - which I can recreate. For the total number of loans, I have closer to 852,000. Would you like for me to update it? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Friday, April 4, 2014 1:14:38 PM Subject: Rollover Paper Jen – I’m looking at Table 3. How could 38,000 borrowers have 7,253,000 loans? Hilary Hi Hilary I addressed the points that you had identified. I think the paper reads well. I have one last number that I need to verify - but the university server is down - so I can update the number tomorrow if needed - but the text is all fine. I will be around in the morning - and then client meeting at 11. Let me know if you want to catch up. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - no, none of the principle findings change at all. There were a few references to these states that I took out (but not all). I am working on a final review now. I will get the final paper back to you tonight. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, April 3, 2014 2:41:47 PM Subject: RE: one clarification Shapiro, R. (2011). The Consumer and Social Welfare Benefits and Costs of Payday Loans: A Review of the Evidence. Hispanic Institute (unpublished manuscript), available at http://www.sonecon.com/docs/studies/Report-Payday-Loans-Shapiro-Sonecon.pdf. Question for you: given the ambiguity of California as a “strict” or “loose” state, and the possible reclassification of Utah as a limited-rollover state, does the principal finding of the paper still hold? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, April 03, 2014 1:38 PM To: Hilary B. Miller Subject: one clarification Hi Hilary Almost done with the edits... There is a reference to a publication by "Shapiro 2011". I cant seem to determine the citation - too many Shapiros in SSRN. Can you provide more detail? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Almost done with the edits... There is a reference to a publication by "Shapiro 2011". I cant seem to determine the citation - too many Shapiros in SSRN. Can you provide more detail? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Got it. Thanks. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, April 2, 2014 8:04:37 AM Subject: RE: Rollover Paper One more: APA format requires 1” margins in all sides. From: Hilary B. Miller Sent: Wednesday, April 02, 2014 8:03 AM To: 'Jennifer Lewis Priestley' Subject: RE: Rollover Paper Sorry, there’s always one more comment: In the text where you discuss your choice of a 2-day lookback period for purposes of determining whether a “rollover” has occurred, perhaps you will rethink why this is a “conservative” assumption. It actually captures fewer transactions than the use of a longer period. From: Hilary B. Miller Sent: Wednesday, April 02, 2014 8:01 AM To: Jennifer Lewis Priestley Subject: RE: Rollover Paper Use this version to start, please – it contains the “lite” abstract. BTW, authors generally refer to themselves in the first person in economic literature such as this. It’s crazy but that’s the custom. From: Hilary B. Miller Sent: Wednesday, April 02, 2014 7:46 AM To: Jennifer Lewis Priestley Subject: Re: Rollover Paper Sorry, two more things: 1. I have decided that the abstract is too long. I am going to shrink it, mostly by shortening the introductory sentences. Let me mess with it, please. 2. In line with the discussion below of California being a fairly liberal (rather than strict) rollover state in economic reality, you may need to adjust the text, where you hold California out as an example of bad outcomes. 3. In Table 1, and possibly elsewhere, you will need to discuss that the applicable regulatory environment in Texas during the study period permitted operators to lend statewide under the so-called “credit services organization” or “CSO” model, which, despite the existence of other regulations that limit rates and fees in Texas, permitted unregulated rates and rollovers. HM From: Jennifer Priestley Date: Tuesday, April 1, 2014 at 7:49 PM To: Hilary Miller Subject: Re: Rollover Paper Thanks Hilary. I will get on this tomorrow. :) Sent from my iPhone On Apr 1, 2014, at 4:56 PM, "Hilary B. Miller" wrote: Jen – I have completed a preliminary round of editing your paper. I have spent quite a bit of time on it and have been as careful as possible. The principal changes I have made are organizational and editorial, while attempting to the greatest extent possible to leave your original substance intact. I think the paper is now more concise and less verbose, better organized and a bit more linear in how it reaches its conclusions. I have beefed up some portions of the paper with additional sources and explanations, while deleting a fair amount of the dated literature discussion. The changes are numerous and fairly extensive. This draft is not redlined. Please review it and feel free to make any further additional changes (or reversions) you feel strongly about. There are a couple of tasks left for you: 1. The references need to be double-checked against the text. Some of the articles (e.g., Bhutta) have been revised and republished in later-year editions. When possible, refer to the latest edition, which will usually be at SSRN rather than elsewhere. I will do this again myself, too. Please. 2. Table 1 provides a somewhat jumbled version of how rollover limitations work. (This paper is supposed to be about rollover limitations, BTW, not general payday-loan limitations; I have changed several references in the text on this.) Two things here: (a) California has a state law that prohibits rollovers but allows unlimited same-day transactions as long as they aren’t interest-only payments. In other words, a borrower can repay his loan and immediately re-borrow, and do so an unlimited number of times in succession. I don’t think your description in the text of California as a “strict” rollover state is correct with this in mind. (b) Utah is not an unregulatedrollover state, as your text indicated – see the new fn. 7, which I have added in the attachment. I think Table 1 would be more useful if it were modified to give more emphasis to interstate rollover-regulation variation and to downplay (or omit altogether) minimum and maximum loan terms, which are not “binding” in the economic sense and really do not operate as interstate differences. 3. I have highlighted in yellow a number of suggested areas for minor additional text. 4. Another item – and this is really big – is that you will need to test your results for robustness under a different definition of “rollover” that comports with the new CFPB paper (CFPB 2014) – i.e., 14 rather than 2 days. I leave to you just how much you need to do to persuade yourself that the results don’t really change. Once you are satisfied, you can update the footnote to state what procedures you followed and why you are persuaded. This is a terrific paper. When it is done, you are going to be famous and your phone will ring off the hook. We are actually talking about a “quiet” release to a few peer reviewers and including the CFPB in the review group. We want them to believe that the results are honest, verifiable and, most importantly, correct. Thanks so much for your help. Please try to finish this up quickly so that we can get it in peer review circulation. Regards, Hilary Thanks Hilary. I will get on this tomorrow. :) Sent from my iPhone On Apr 1, 2014, at 4:56 PM, "Hilary B. Miller" wrote: Jen – I have completed a preliminary round of editing your paper. I have spent quite a bit of time on it and have been as careful as possible. The principal changes I have made are organizational and editorial, while attempting to the greatest extent possible to leave your original substance intact. I think the paper is now more concise and less verbose, better organized and a bit more linear in how it reaches its conclusions. I have beefed up some portions of the paper with additional sources and explanations, while deleting a fair amount of the dated literature discussion. The changes are numerous and fairly extensive. This draft is not redlined. Please review it and feel free to make any further additional changes (or reversions) you feel strongly about. There are a couple of tasks left for you: 1. The references need to be double-checked against the text. Some of the articles (e.g., Bhutta) have been revised and republished in later-year editions. When possible, refer to the latest edition, which will usually be at SSRN rather than elsewhere. I will do this again myself, too. Please. 2. Table 1 provides a somewhat jumbled version of how rollover limitations work. (This paper is supposed to be about rollover limitations, BTW, not general payday-loan limitations; I have changed several references in the text on this.) Two things here: (a) California has a state law that prohibits rollovers but allows unlimited same-day transactions as long as they aren’t interest-only payments. In other words, a borrower can repay his loan and immediately re-borrow, and do so an unlimited number of times in succession. I don’t think your description in the text of California as a “strict” rollover state is correct with this in mind. (b) Utah is not an unregulatedrollover state, as your text indicated – see the new fn. 7, which I have added in the attachment. I think Table 1 would be more useful if it were modified to give more emphasis to interstate rollover-regulation variation and to downplay (or omit altogether) minimum and maximum loan terms, which are not “binding” in the economic sense and really do not operate as interstate differences. 3. I have highlighted in yellow a number of suggested areas for minor additional text. 4. Another item – and this is really big – is that you will need to test your results for robustness under a different definition of “rollover” that comports with the new CFPB paper (CFPB 2014) – i.e., 14 rather than 2 days. I leave to you just how much you need to do to persuade yourself that the results don’t really change. Once you are satisfied, you can update the footnote to state what procedures you followed and why you are persuaded. This is a terrific paper. When it is done, you are going to be famous and your phone will ring off the hook. We are actually talking about a “quiet” release to a few peer reviewers and including the CFPB in the review group. We want them to believe that the results are honest, verifiable and, most importantly, correct. Thanks so much for your help. Please try to finish this up quickly so that we can get it in peer review circulation. Regards, Hilary Ah. a kindred spirit. :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, March 26, 2014 10:59:46 AM Subject: RE: Data Point Response Heck, that’s nothing. We solve multicollinearity problems at the dinner table every night. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, March 26, 2014 10:42 AM To: Hilary B. Miller Subject: Re: Data Point Response Thanks Hilary. I actually think this paper really tees up our (potential) second paper on default rates for payday loan borrowers. BTW - I am embarrased to say that I had to look up "prolix". Now it is my word of the day. :) I am writing notes today for class about "heteroscedasticity". I dont expect to teach a lawyer any new words but that might be a fun one for you to incorporate into your conversations today. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, March 26, 2014 10:02:41 AM Subject: RE: Data Point Response Thanks for this. It’s more prolix than what I think is appropriate, but I’ll skinny it down and insert it in the paper. Recall that both Mann and Fusaro & Cirillo use a “window” much wider than yours to define a “rollover.” In doing so, they are capturing an economic, rather than literal, refinancing. The theory is that, if the consumer needs to re-incur the debt before reaching his or her next payday, the consumer lacked the means to repay the debt in full from recurring cash inflows. This is something of an effort to bend over backwards to accommodate our antagonists but nevertheless captures an issue that is important to policymakers. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, March 26, 2014 9:25 AM To: Hilary B. Miller Subject: Data Point Response Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks Hilary. I actually think this paper really tees up our (potential) second paper on default rates for payday loan borrowers. BTW - I am embarrased to say that I had to look up "prolix". Now it is my word of the day. :) I am writing notes today for class about "heteroscedasticity". I dont expect to teach a lawyer any new words but that might be a fun one for you to incorporate into your conversations today. Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, March 26, 2014 10:02:41 AM Subject: RE: Data Point Response Thanks for this. It’s more prolix than what I think is appropriate, but I’ll skinny it down and insert it in the paper. Recall that both Mann and Fusaro & Cirillo use a “window” much wider than yours to define a “rollover.” In doing so, they are capturing an economic, rather than literal, refinancing. The theory is that, if the consumer needs to re-incur the debt before reaching his or her next payday, the consumer lacked the means to repay the debt in full from recurring cash inflows. This is something of an effort to bend over backwards to accommodate our antagonists but nevertheless captures an issue that is important to policymakers. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, March 26, 2014 9:25 AM To: Hilary B. Miller Subject: Data Point Response Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - I read the paper...I am writing a response for you...I will send it over tomorrow. In the end, I thought it was really alot of nothing certainly compared to the far more rigorous papers that I have been reading. The CFPB paper was really just descriptive statistics...there was no prediction/explanation. There was no reference or response to the main question central to the discourse - "Does payday lending have a negative, neutral or positive impact on borrowers' financial wellbeing?" Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, March 25, 2014 7:23:17 AM Subject: RE: CFPB Finds Four Out Of Five Payday Loans Are Rolled Over Or Renewed Sorry — my full message got deleted. The CFPB is releasing this new paper — very noisily — today. We need to update your paper to refer to and dispense with it. Essentially, it is new lipstick on the same old pig: repeat usage without evidence of actual consumer detriment. In any event, it is sufficiently important that we need to say something about it. Would you please read it and write a couple of paragraphs? I’ll find the appropriate place to slot it in. I’m going to begin editing in earnest in the next day or so. Back now from London and focusing on this stuff. Thanks. Regards, Hilary This message, together with any attachments, is intended only for the use of the individual or entity to which it is addressed and may contain information that is legally privileged, confidential and exempt from disclosure. If you are not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this message, or any attachment, is strictly prohibited. If you have received this message in error, please notify the original sender immediately by telephone (203-399-1320) or by return e-mail and delete the message, along with any attachments, from your computer. IRS Circular 230 disclosure: Any tax advice contained in this communication (including any attachments) was not intended or written to be used, and cannot be used, for the purpose of (i) avoiding tax-related penalties under the Internal Revenue Code or (ii) promoting, marketing or recommending to another party any matters addressed herein. Thank you. -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Tuesday March 25, 2014, 07:15 AM To: Hilary B. Miller [hilary@miller.net] Subject: Re: CFPB Finds Four Out Of Five Payday Loans Are Rolled Over Or Renewed Would you like to catch up this afternoon? Sent from my iPhone On Mar 25, 2014, at 7:13 AM, "Hilary B. Miller"  wrote: Jen — -----Original Message----From: CFPB_Communications [CFPB_Communications@cfpb.gov] Received: Tuesday March 25, 2014, 12:08 AM To: CFPB_Communications [CFPB_Communications@cfpb.gov] Subject: CFPB Finds Four Out Of Five Payday Loans Are Rolled Over Or Renewed FOR IMMEDIATE RELEASE: March 25, 2014 CONTACT: Office of Communications Tel: (202) 435-7170 CONSUMER FINANCIAL PROTECTION BUREAU FINDS FOUR OUT OF FIVE PAYDAY LOANS ARE ROLLED OVER OR RENEWED Research Shows the Majority of Payday Loans Are Made to Borrowers Caught in a Revolving Door of Debt WASHINGTON, D.C. — Today the Consumer Financial Protection Bureau (CFPB) issued a report on payday lending finding that four out of five payday loans are rolled over or renewed within 14 days. The study also shows that the majority of all payday loans are made to borrowers who renew their loans so many times that they end up paying more in fees than the amount of money they originally borrowed. “We are concerned that too many borrowers slide into the debt traps that payday loans can become,” said CFPB Director Richard Cordray. “As we work to bring needed reforms to the payday market, we want to ensure consumers have access to small-dollar loans that help them get ahead, not push them farther behind.” The report is at: http://files.consumerfinance.gov/f/201403_cfpb_report_paydaylending.pdf Payday loans are typically described as a way to bridge a cash flow shortage between paychecks or other income. Also known as “cash advances” or “check loans,” they are usually expensive, small-dollar loans, of generally $500 or less. They can offer quick and easy accessibility, especially for consumers who may not qualify for other credit. Today’s report is based on data from a 12-month period with more than 12 million storefront payday loans. It is a continuation of the work in last year’s CFPB report on Payday Loans and Deposit Advance Products, one of the most comprehensive studies ever undertaken on the market. That report raised questions about the loose lending standards, high costs, and risky loan structures that may contribute to the sustained use of these products. Today’s report provides a deeper analysis of the data, focusing on repeated borrowing by consumers after they take out an initial payday loan. A primary driver of the cost of payday loans is that consumers may roll over the loans or engage in re-borrowing within a short window of time after repaying their first loan. Today’s study looks at not only the initial loans but also loans taken out within 14 days of paying off the old loans; it considers these subsequent loans to be renewals and part of the same “loan sequence.” Today’s study is the most in-depth analysis of this pattern to date. Key Findings: Many Payday Loans Become Revolving Doors of Debt By focusing on payday loan renewals, the study found that a large share of consumers end up in cycles of repeated borrowing and incur significant costs over time. Specifically, the study found:  Four out of five payday loans are rolled over or renewed: More than 80 percent of payday loans are rolled over or renewed within two weeks. The study found that when looking at 14-day windows in the states that have cooling-off periods that reduce the level of same-day renewals, the renewal rates are nearly identical to states without these limitations.  Three out of five payday loans are made to borrowers whose fee expenses exceed amount borrowed: Over 60 percent of loans are made to borrowers in the course of loan sequences lasting seven or more loans in a row. Roughly half of all loans are made to borrowers in the course of loan sequences lasting ten or more loans in a row.  One out of five new payday loans end up costing the borrower more than the amount borrowed: For 48 percent of all initial payday loans – those that are not taken out within 14 days of a prior loan – borrowers are able to repay the loan with no more than one renewal. But for 22 percent of new loans, borrowers end up renewing their loans six times or more. With a typical payday fee of 15 percent, consumers who take out an initial loan and six renewals will have paid more in fees than the original loan amount.  Four out of five payday borrowers either default or renew a payday loan over the course of a year: Only 15 percent of borrowers repay all of their payday debts when due without re-borrowing within 14 days; 20 percent default on a loan at some point; and 64 percent renew at least one loan one or more times. Defaulting on a payday loan may cause the consumer to incur bank fees. Renewing loans repeatedly can put consumers on a slippery slope toward a debt trap where they cannot get ahead of the money they owe.  Four out of five payday borrowers who renew end up borrowing the same amount or more: Specifically, more than 80 percent of borrowers who rolled over loans owed as much or more on the last loan in a loan sequence than the amount they borrowed initially. These consumers are having trouble getting ahead of the debt. The study also found that as the number of rollovers increases, so too does the percentage of borrowers who increase their borrowing.  One out of five payday borrowers on monthly benefits trapped in debt: The study also looked at payday borrowers who are paid on a monthly basis and found one out of five remained in debt the entire year of the CFPB study. Payday borrowers who fall into this category include elderly Americans or disability recipients receiving Supplemental Security Income and Social Security Disability. Today’s report will help educate regulators and the public about how the payday lending market works and about the behavior of borrowers in the market. The CFPB has authority to oversee the payday loan market. It began its supervision of payday lenders inJanuary 2012. In November 2013, the CFPB began accepting complaints from borrowers encountering problems with payday loans.   ### The Consumer Financial Protection Bureau is a 21st century agency that helps consumer finance markets work by making rules more effective, by consistently and fairly enforcing those rules, and by empowering consumers to take more control over their economic lives. For more information, visit consumerfinance.gov. Would you like to catch up this afternoon? Sent from my iPhone On Mar 25, 2014, at 7:13 AM, "Hilary B. Miller" wrote: Jen — -----Original Message----From: CFPB_Communications [CFPB_Communications@cfpb.gov] Received: Tuesday March 25, 2014, 12:08 AM To: CFPB_Communications [CFPB_Communications@cfpb.gov] Subject: CFPB Finds Four Out Of Five Payday Loans Are Rolled Over Or Renewed FOR IMMEDIATE RELEASE: March 25, 2014 CONTACT: Office of Communications Tel: (202) 435-7170 CONSUMER FINANCIAL PROTECTION BUREAU FINDS FOUR OUT OF FIVE PAYDAY LOANS ARE ROLLED OVER OR RENEWED Research Shows the Majority of Payday Loans Are Made to Borrowers Caught in a Revolving Door of Debt WASHINGTON, D.C. — Today the Consumer Financial Protection Bureau (CFPB) issued a report on payday lending finding that four out of five payday loans are rolled over or renewed within 14 days. The study also shows that the majority of all payday loans are made to borrowers who renew their loans so many times that they end up paying more in fees than the amount of money they originally borrowed. “We are concerned that too many borrowers slide into the debt traps that payday loans can become,” said CFPB Director Richard Cordray. “As we work to bring needed reforms to the payday market, we want to ensure consumers have access to small-dollar loans that help them get ahead, not push them farther behind.” The report is at: http://files.consumerfinance.gov/f/201403_cfpb_report_paydaylending.pdf Payday loans are typically described as a way to bridge a cash flow shortage between paychecks or other income. Also known as “cash advances” or “check loans,” they are usually expensive, small-dollar loans, of generally $500 or less. They can offer quick and easy accessibility, especially for consumers who may not qualify for other credit. Today’s report is based on data from a 12-month period with more than 12 million storefront payday loans. It is a continuation of the work in last year’s CFPB report on Payday Loans and Deposit Advance Products, one of the most comprehensive studies ever undertaken on the market. That report raised questions about the loose lending standards, high costs, and risky loan structures that may contribute to the sustained use of these products. Today’s report provides a deeper analysis of the data, focusing on repeated borrowing by consumers after they take out an initial payday loan. A primary driver of the cost of payday loans is that consumers may roll over the loans or engage in re-borrowing within a short window of time after repaying their first loan. Today’s study looks at not only the initial loans but also loans taken out within 14 days of paying off the old loans; it considers these subsequent loans to be renewals and part of the same “loan sequence.” Today’s study is the most in-depth analysis of this pattern to date. Key Findings: Many Payday Loans Become Revolving Doors of Debt By focusing on payday loan renewals, the study found that a large share of consumers end up in cycles of repeated borrowing and incur significant costs over time. Specifically, the study found:  Four out of five payday loans are rolled over or renewed: More than 80 percent of payday loans are rolled over or renewed within two weeks. The study found that when looking at 14-day windows in the states that have cooling-off periods that reduce the level of same-day renewals, the renewal rates are nearly identical to states without these limitations.  Three out of five payday loans are made to borrowers whose fee expenses exceed amount borrowed: Over 60 percent of loans are made to borrowers in the course of loan sequences lasting seven or more loans in a row. Roughly half of all loans are made to borrowers in the course of loan sequences lasting ten or more loans in a row.  One out of five new payday loans end up costing the borrower more than the amount borrowed: For 48 percent of all initial payday loans – those that are not taken out within 14 days of a prior loan – borrowers are able to repay the loan with no more than one renewal. But for 22 percent of new loans, borrowers end up renewing their loans six times or more. With a typical payday fee of 15 percent, consumers who take out an initial loan and six renewals will have paid more in fees than the original loan amount.  Four out of five payday borrowers either default or renew a payday loan over the course of a year: Only 15 percent of borrowers repay all of their payday debts when due without re-borrowing within 14 days; 20 percent default on a loan at some point; and 64 percent renew at least one loan one or more times. Defaulting on a payday loan may cause the consumer to incur bank fees. Renewing loans repeatedly can put consumers on a slippery slope toward a debt trap where they cannot get ahead of the money they owe.  Four out of five payday borrowers who renew end up borrowing the same amount or more: Specifically, more than 80 percent of borrowers who rolled over loans owed as much or more on the last loan in a loan sequence than the amount they borrowed initially. These consumers are having trouble getting ahead of the debt. The study also found that as the number of rollovers increases, so too does the percentage of borrowers who increase their borrowing.  One out of five payday borrowers on monthly benefits trapped in debt: The study also looked at payday borrowers who are paid on a monthly basis and found one out of five remained in debt the entire year of the CFPB study. Payday borrowers who fall into this category include elderly Americans or disability recipients receiving Supplemental Security Income and Social Security Disability. Today’s report will help educate regulators and the public about how the payday lending market works and about the behavior of borrowers in the market. The CFPB has authority to oversee the payday loan market. It began its supervision of payday lenders inJanuary 2012. In November 2013, the CFPB began accepting complaints from borrowers encountering problems with payday loans.   ### The Consumer Financial Protection Bureau is a 21st century agency that helps consumer finance markets work by making rules more effective, by consistently and fairly enforcing those rules, and by empowering consumers to take more control over their economic lives. For more information, visit consumerfinance.gov. All of that sounds fine. Look forward to speaking with you when you get back. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, March 22, 2014 9:38:03 AM Subject: RE: Checking In Got it. I am focusing now primarily on editorial efforts. I may have some questions for you. I am traveling and have not been able to block out enough time to do this, so it will await my return on Tuesday. Also, the CFPB is holding a public hearing on payday late next week, and I am looking to see if they have any new data to report. I’ll be back to you with suggested edits. Thanks. -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Saturday March 22, 2014, 01:36 PM To: Hilary B. Miller [hilary@miller.net] Subject: Checking In Hi Hilary - I trust you are well. I wanted to check in to ensure that you received the updated draft of the paper on Monday and to see if you wanted to catch up at any point. Jen :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I trust you are well. I wanted to check in to ensure that you received the updated draft of the paper on Monday and to see if you wanted to catch up at any point. Jen :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary. I worked on this over the weekend in Memphis - I think it reads well. I stripped out all of the default analysis - I agree that that analysis could form a paper on its own. Let me know your thoughts. Jen Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I am making all the edits to the paper - almost done. I would like to have the weekend to work with it - I have to go to Memphis today...I plan to do some of the editing on the way. I would expect to have an even more improved revision to you by Sunday. Jen :) Jennifer Lewis Priestley, MBA, Ph.D. Professor of Applied Statistics and Data Science Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks. I will get on this this week. :) Sent from my iPhone On Mar 9, 2014, at 11:50 AM, "Hilary B. Miller" wrote: Jen – Thanks so much for this. I agree, it is greatly improved. I have some further comments for you, and I think I can take it from there after you accommodate these issues: 1. Another 30,000’ observation, and this one about organization. From a structural standpoint, there is something amiss here, and I realize that it’s a monster of my own creation. The paper is supposed to be about payday rollovers. It has that title, and that is its indeed its thrust. But then it suddenly lurches into this “default” discussion, which is a non sequitur, largely unrelated, and a separate issue altogether. And it certainly confuses the main issue. So let’s break all that stuff out and make it a SEPARATE PAPER. And it’s close and doesn’t need a lot of additional work. We can discuss that later. In the meanwhile, we can devote our resources to prompt completion of a rollover-only paper. 2. You will hear a fairly consistent theme in many of the next several comments, which is that they reiterate comments given to you in the 2/19 memo which have not been reflected in this draft. To begin, the “Background” section of the paper, which includes your lit survey, still needs some work. Please go back and look at my previous memo of 2/19 regarding Fusaro and Cirillo (2011) and Mann (2013). These are the canonical works on rollovers, but you don’t even bother mentioning them in this section. The later references in the paper should be deleted. You should specifically discuss not only that they studied these matters, but what and how they studied, and what they concluded. Please look at my previous comments. At multiple points in the paper it feels as if you are citing Einstein for his cake recipe instead of for his general theory of relativity. 3. Likewise, you need to pick up my comments about Bhutta (2013) from my 2/19 memo. You have quoted from the paper on p. 3, but this quote completely misses the principal finding of the paper. Take another look at my memo. 4. In general, this section of the paper enumerates the papers, but doesn’t explain them well. For example, you say very little about Kaufman. This is an extremely important paper and it is, together with Bhutta, the work on which your paper naturally builds. You therefore need to explain both papers, and then explain the role of your paper plays in adding to science. 5. In general, we do not accept the notion that a “cycle of debt” even exists, and I would appreciate it if you would delete all references to this term, unless you are rebutting its existence. As a threshold matter, and you and I have discussed, the term “cycle of debt” is itself problematic. In common parlance, we do not use this term to refer to a credit card “revolver” who repays his balance over the longest possible time period, nor do we apply it to a mortgage borrower who gets a 10-year interest-only loan (which banks happily provide). It is a term that is nearly exclusively reserved for payday borrowers, so it must import something more than merely a borrower who remains indebted for “too long.” The “something more” is a feature of payday loans that is asserted by our antagonists to exist, but which does not, in fact, exist: a “debt trap.” The theory of this “trap” seems to be that borrowers devote so much of their free cash flow to paying interest on their payday loans that they cannot repay principal. Thus, according to the “trap” theory, borrowers are compelled to borrow ever-increasing amounts just to cover the interest, with no hope of repaying principal. The problem with this theory of a “trap” is that there is no non-anecdotal evidence to support its existence, and the numbers used by CRL to illustrate it are cooked – false, logically inconsistent, deliberately misleading. See http://ssrn.com/abstract=2029146 at page 9 (pages are unnumbered). And the science, as I have previously discussed with you, completely negates the concept of an interest-caused “trap.” See http://ssrn.com/abstract=1960776. Because of the lack of science, and the lack of any principled application of the term “cycle” to this kind of usage, we begin simply by denying the existence of a “cycle of debt” and, perhaps more importantly, by denying that extended use is per se harmful. As I frequently state publicly, the term “cycle of debt” is a political epithet (usually combined with terms like “trap,” “triple-digit” and “predatory”) which is both loaded and implies some kind of contrary norm. It is not a term of science, and the term is not accepted in peer-reviewed economics literature. I think even you fall into this fallacy in the paper (for example, it is impossible to compound interest on a payday loan, but you seem to imply otherwise in the paper). Perhaps you could rethink this and work some of the Stoesz material into your paper instead. Let’s just call this section “Background on Rollovers.” 6. The lit survey also needs a broader discussion of the CFPB “White Paper,” to which you allude but which you summarize only for its non-data-based findings. I can fix this in your next draft, but it would be easier for you to do it yourself. Again, state what they studied, how they studied it, and what the conclusions were that were supported by their data. You can then discuss separately the political conclusions included in the White Paper that were not supported by the data. I can help you with this if you want. 7. A new point: each of the states you studied in your work has a different regulatory scheme, and rollovers aren’t the only issue that is regulated differently between states. You don’t, for example, control for the differences in interest rates permitted in California (459% APR), Florida (260% APR) and Texas (unlimited). I think you need to explain how controlling for “state” rather than individual regulatory features is a good proxy for “rollovers.” The consumer market experience of interest rates, in practice, is that they are immaterially different. 8. The tables are still not self-explanatory. A non-professional reader should be able to look at any table and tell exactly what is being represented. This can be accomplished through footnotes or more detailed headers. We do not want people’s eyes to glaze over when they look at these tables or to be required to refer back to the text to see what is happening. See comment #12 in the 2/29 memo. 9. Please go back and re-read my comment #14 about the “two day” choice from the 2/29 memo. I don’t think you have addressed this election you made, which you made differently from, for example, Mann. Fusaro and Cirillo show outcomes under alternative definitions of what constitutes a “rollover.” You show one. You need to explain. 10. What happened to the tables from Sandler’s paper that I had asked you to include? They now seem to be missing. The “days to clearance” issue is important to be able to refer to prior art. 11. In your discussion of databases (FL and OK), you state that the rollover rates are low, which they must necessarily be. You then drop the discussion. There’s a “But …” (look at UT) and a greatly expanded further discussion warranted here. While the numbers are important, the reader needs to know what’s going on here. The presumption is that databases greatly enhance consumer welfare. Surprise! They don’t. Why? I have sent you an APA template that you can apply to the paper. Thanks for all your hard work. It is really coming along. I am hoping that I can start line-editing your next draft soon and that we can finish the paper this month. Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, March 04, 2014 10:53 PM To: Hilary B. Miller Subject: Next Round... Hi Hilary I have reworked the paper - I think it is improved. Looking forward to your feedback. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Thanks for the doc. Is the idea that you were pleased with the content and you just want me to drop it into this template? I am reworking the discussion section a bit more...but let me know what you thought about the rest of the document. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)" Sent: Sunday, March 9, 2014 10:55:04 AM Subject: Emailing: APA Format New.dotx Your message is ready to be sent with the following file or link attachments: APA Format New.dotx Note: To protect against computer viruses, e-mail programs may prevent sending or receiving certain types of file attachments. Check your e-mail security settings to determine how attachments are handled. Safe Travels. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, March 4, 2014 10:58:02 PM Subject: Re: Next Round... Starting a long plane ride later tonight. Will read. Thanks. _____________ Hilary B. Miller 500 West Putnam Avenue ­ Suite 400 Greenwich, CT 06830­6096 (203) 399­1320 voice (203) 517­6859 cell (914) 206­3727 fax hilary@miller.net (sent from iPhone) On Mar 4, 2014, at 6:52 PM, "Jennifer Lewis Priestley"  wrote: Hi Hilary I have reworked the paper - I think it is improved. Looking forward to your feedback. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I have reworked the paper - I think it is improved. Looking forward to your feedback. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Got it. (I really do have an aunt in Savannah TN who was a High School English teacher for 30 years). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, February 25, 2014 10:33:57 AM Subject: Re: Update Don't forget your maiden aunt. _____________ Hilary B. Miller Law offices of Hilary B. Miller 500 West Putnam Avenue ­ Suite 400 Greenwich, CT 06830­6096 (203) 399­1320 voice (203) 517­6859 cell (914) 206­3727 fax hilary@miller.net (sent from iPhone) On Feb 25, 2014, at 10:31 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary Jus a quick update. I read the papers that you sent to me - thank you for that. I will be working on a new draft over the next few days. I expect to have something back to you early next week. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Monday, February 17, 2014 7:44:13 PM Subject: Full Draft Hi Hilary Attached is my full draft of the paper. I anticipate a round or two of edits...but I think it generally makes a contribution to the current research on the topic. I will be around most of tomorrow if you want to get on the phone at any point. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Jus a quick update. I read the papers that you sent to me - thank you for that. I will be working on a new draft over the next few days. I expect to have something back to you early next week. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Monday, February 17, 2014 7:44:13 PM Subject: Full Draft Hi Hilary Attached is my full draft of the paper. I anticipate a round or two of edits...but I think it generally makes a contribution to the current research on the topic. I will be around most of tomorrow if you want to get on the phone at any point. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Could I trouble you to forward to me the papers that you make reference to? Or provide a full citation and I will have my grad students track them down? I took the GEE model out - in the end, it did not add anything incremental to the discussion...and most importantly, I found an error in the math - that I could have "buried" - but in the end I want the analytics to be bulletproof if challenged (I want to be able to replicate - and have others replicate - all of the results cleanly without issue). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 19, 2014 6:24:35 PM Subject: RE: Full Draft Jen – Again, sorry for the long feedback loop. I have had a couple of days from hell. Here are some comments on the paper: 1. On balance, the paper covers most of the key topics. The numbers seem to be plentiful. There are a few omissions, which I discuss below. 2. As a “30,000 foot” observation, the text seems light. I realize that the consumer-welfare aspects and the literature are relatively new to you, but the narrative seems somewhat superficial, without much discussion of what the underlying processes might be and how your findings dovetail with other literature. In the comments below, I will suggest some additional areas for development. 3. Punctuation and capitalization are somewhat random. You might want to have your maiden aunt who went to high school before 1960 read this. 4. The paper should start off with a discussion of payday loans, not with a discussion of DoddFrank or “abusive” practices. Actually, this material doesn’t seem to belong at all. What is a payday loan? Who uses them, how do they use them, and why is this population potentially vulnerable? Why do rollovers matter? What is the potential harm from rollovers? What previously unanswered questions does your paper answer, and why are your questions important to policymakers? You need to set the paper up better. I get to the third paragraph before you even start this discussion. It needs to start with a bang. It now starts with kind of a thud. This is blockbuster stuff. You put me to sleep before I got to the “good stuff.” 5. The paper needs a more comprehensive discussion of the consumer-welfare impacts of payday borrowing. This really means you need to flesh out your lit survey a bit. I would consider adding Morgan and Strain (2007); Adair Morse (2009); and Stoianovici and Maloney (2008), all to be inserted at around the existing discussion of Zinman (2010). It serves your purposes to suggest that the answers to the ultimate welfare questions are murky. These papers tend to counterbalance the CFPB’s white paper. 6. The quoted material from Desai and Elliehausen (2013) doesn’t need to be set forth as a quotation. Instead, here simply set forth the propositions they refer to and you can include the citation as a reference. We’ll give you some other citations for the same material. The quoted matter is old hat and the industry’s longstanding position, not original with Desai. 7. The literature survey needs to be more comprehensive with respect to the evidence on rollovers. You need to discuss Fusaro and Cirillo (2011) and Mann (2013). 8. Critically, although you cite Bhutta, Skiba and Tobacman (2013), for several possible adverse demographic findings, you do not cite the paper for its principal finding, which is that payday loans have a “precise zero” long-run effect on consumers’ financial well-being. This paper is and remains the “gold standard” for whether payday loans are harmful or helpful to consumers. The results found by these investigators fully take into account all of the sustained usage of payday loans criticized by the CFPB. The CFPB simply chooses to ignore it. There is no other academic research that relates sustained usage with consumer outcomes, and there is no economically demonstrated “need” to protect consumers either from multiple loans or longer usage terms. The Mann paper effectively destroys the notion that consumers are being misled, as alleged by Pew, into taking out a short-term product for long-term use. These relationships need to be developed in the text. 9. On page 3, I don’t think you have defined the dataset properly. The borrower histories relate to borrowers who incurred “new” loans in the first six months of 2006 and the first six months of 2008. In this regard, “new” means no loans in the 90 days prior to the first loan in the dataset (not necessarily “virgin”). My understanding was that borrowers from 2006 were followed in 2007 but not necessarily in any subsequent period, etc. Can you check this? 10. I would like you to add at least one or two extra paragraphs on the VantageScore in general. You should cover what the score is, what its principal components are, and how it works. This can be a relatively brief discussion, although it is an opportunity to introduce the weights applied to the components, which are relevant to our population. Then, you should explain – and this is really critical – why VantageScore is an appropriate outcome variable for this kind of study. You can borrow from Bhutta et al. if you need to do so here, but the key is to give a clear indication of the wisdom of selecting this outcome variable in preference to others than you might have chosen (such as, for example, simply using defaults in the style of Desai). You should anticipate and counter the argument made by Pew that these scores are either too uniformly low or irrelevant for this population. 11. Where you refer on pp. 3-4 to Kaufman, you should relocate these discussions back to the “literature survey” part of the paper, and leave this part to discuss your own findings. 12. In general, I find the tables are not self-explanatory. By that I mean that a reasonably skilled reader cannot turn to a table and immediate tell what is being represented, either because the column and row headings are omitted, too abbreviated, or too cryptic. This should be remedied, including by the addition of footnotes where necessary. Go overboard on explaining in the footnotes how to read the tables, giving express examples if necessary. A key audience for the paper will be highly educated but innumerate policymakers. 13. The material starting on page 4 is where some key “beefing up” of the text is required. Here, you need to explain not only what the tables say, but also what they mean. As a policymaker, what am I supposed to take away from this? 14. You rollover “definition” isn’t a universal definition, it’s just one you assumed for purposes of this paper. Others have made different assumptions. See Mann’s paper (using 14-day debt-free period to determine whether loan rolled over or not). You need to explain why you chose 2 days instead of some other period, why that’s an appropriate choice, and how your results might have varied if you had made a Mann-like choice, for example. 15. In general, you should refer to payday borrowers as “borrowers” rather than “customers.” Globally. 16. You have a tendency to launch right into the numbers (see, e.g., the first full paragraph after the Section 3.2 header), rather than state what you sought to study, why it matters, how to interpret the results, and what the implication is for policymakers. Put yourself in the shoes of the reader and take a more gradual start, then dump the number, and finally explain what they mean. 17. The astounding relationship between sustained use and outcome should be more fully developed. Why does this relationship exist? What theories from economics should have put us on notice of this result? Why is the market a better judge of who should obtain credit than a state legislature????? 17. I think the “default” discussion is somewhat confusing. Going back to the original purpose of this inquiry, opponents of payday lending hypothesize that defaults are harmful for consumers, although there seems to no data to support that hypothesis. We want to test this hypothesis and report the results of our testing. (At least one possible counterfactual is that defaults are actually welfare-enhancing because the borrower gets to keep the loan principal and collection efforts are largely ineffective. This may explain what is going on.) In any event, we once again launch directly into the numbers without explaining why we are making this inquiry and why anyone should care about it. We then don’t connect the results to the original question. 18. The second part of the “default” discussion is equally lost here, which is whether it is possible to identify ex ante the loan applicants who will have adverse outcomes from borrowing, and simply deny them credit through scoring or otherwise. For purpose of this discussion, an “adverse” outcome is one that harms the borrower, not the lender. This is part of a popular new theory of consumer protection that involves having the lender make sure that the loans are “safe” or “suitable” or “affordable” for the consumer, which is a different inquiry from whether the consumer is “creditworthy.” We wanted to test our ability to identify and segregate these consumers. The discussion is lost, and I fear that the data aren’t adequately explained. 19. In the conclusion, the “$64 question” is the rollover one. But we misstate (or don’t reach) the consumer-harm question related to suitability. 20. What happened to the general estimating equation? 21. I am out of time and I want to give you some feedback on a few of the tables. I will try to do that later tonight, tomorrow or Friday. As you say, you have some blockbuster material. The paper needs to be a bit more literary and to “sing.” I think this can be accomplished primarily by slowing down and taking the reader ponderously through each analysis, explanation and, importantly, meaning. It is going to be extremely good when done. I’ll be on my cell tomorrow most of the day if you want to chat, and I have email access. Thank you! Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, February 17, 2014 7:44 PM To: Hilary B. Miller Subject: Full Draft Hi Hilary Attached is my full draft of the paper. I anticipate a round or two of edits...but I think it generally makes a contribution to the current research on the topic. I will be around most of tomorrow if you want to get on the phone at any point. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks. I will get on this this weekend. :) Sent from my iPhone On Feb 19, 2014, at 6:24 PM, "Hilary B. Miller" wrote: Jen – Again, sorry for the long feedback loop. I have had a couple of days from hell. Here are some comments on the paper: 1. On balance, the paper covers most of the key topics. The numbers seem to be plentiful. There are a few omissions, which I discuss below. 2. As a “30,000 foot” observation, the text seems light. I realize that the consumer-welfare aspects and the literature are relatively new to you, but the narrative seems somewhat superficial, without much discussion of what the underlying processes might be and how your findings dovetail with other literature. In the comments below, I will suggest some additional areas for development. 3. Punctuation and capitalization are somewhat random. You might want to have your maiden aunt who went to high school before 1960 read this. 4. The paper should start off with a discussion of payday loans, not with a discussion of DoddFrank or “abusive” practices. Actually, this material doesn’t seem to belong at all. What is a payday loan? Who uses them, how do they use them, and why is this population potentially vulnerable? Why do rollovers matter? What is the potential harm from rollovers? What previously unanswered questions does your paper answer, and why are your questions important to policymakers? You need to set the paper up better. I get to the third paragraph before you even start this discussion. It needs to start with a bang. It now starts with kind of a thud. This is blockbuster stuff. You put me to sleep before I got to the “good stuff.” 5. The paper needs a more comprehensive discussion of the consumer-welfare impacts of payday borrowing. This really means you need to flesh out your lit survey a bit. I would consider adding Morgan and Strain (2007); Adair Morse (2009); and Stoianovici and Maloney (2008), all to be inserted at around the existing discussion of Zinman (2010). It serves your purposes to suggest that the answers to the ultimate welfare questions are murky. These papers tend to counterbalance the CFPB’s white paper. 6. The quoted material from Desai and Elliehausen (2013) doesn’t need to be set forth as a quotation. Instead, here simply set forth the propositions they refer to and you can include the citation as a reference. We’ll give you some other citations for the same material. The quoted matter is old hat and the industry’s longstanding position, not original with Desai. 7. The literature survey needs to be more comprehensive with respect to the evidence on rollovers. You need to discuss Fusaro and Cirillo (2011) and Mann (2013). 8. Critically, although you cite Bhutta, Skiba and Tobacman (2013), for several possible adverse demographic findings, you do not cite the paper for its principal finding, which is that payday loans have a “precise zero” long-run effect on consumers’ financial well-being. This paper is and remains the “gold standard” for whether payday loans are harmful or helpful to consumers. The results found by these investigators fully take into account all of the sustained usage of payday loans criticized by the CFPB. The CFPB simply chooses to ignore it. There is no other academic research that relates sustained usage with consumer outcomes, and there is no economically demonstrated “need” to protect consumers either from multiple loans or longer usage terms. The Mann paper effectively destroys the notion that consumers are being misled, as alleged by Pew, into taking out a short-term product for long-term use. These relationships need to be developed in the text. 9. On page 3, I don’t think you have defined the dataset properly. The borrower histories relate to borrowers who incurred “new” loans in the first six months of 2006 and the first six months of 2008. In this regard, “new” means no loans in the 90 days prior to the first loan in the dataset (not necessarily “virgin”). My understanding was that borrowers from 2006 were followed in 2007 but not necessarily in any subsequent period, etc. Can you check this? 10. I would like you to add at least one or two extra paragraphs on the VantageScore in general. You should cover what the score is, what its principal components are, and how it works. This can be a relatively brief discussion, although it is an opportunity to introduce the weights applied to the components, which are relevant to our population. Then, you should explain – and this is really critical – why VantageScore is an appropriate outcome variable for this kind of study. You can borrow from Bhutta et al. if you need to do so here, but the key is to give a clear indication of the wisdom of selecting this outcome variable in preference to others than you might have chosen (such as, for example, simply using defaults in the style of Desai). You should anticipate and counter the argument made by Pew that these scores are either too uniformly low or irrelevant for this population. 11. Where you refer on pp. 3-4 to Kaufman, you should relocate these discussions back to the “literature survey” part of the paper, and leave this part to discuss your own findings. 12. In general, I find the tables are not self-explanatory. By that I mean that a reasonably skilled reader cannot turn to a table and immediate tell what is being represented, either because the column and row headings are omitted, too abbreviated, or too cryptic. This should be remedied, including by the addition of footnotes where necessary. Go overboard on explaining in the footnotes how to read the tables, giving express examples if necessary. A key audience for the paper will be highly educated but innumerate policymakers. 13. The material starting on page 4 is where some key “beefing up” of the text is required. Here, you need to explain not only what the tables say, but also what they mean. As a policymaker, what am I supposed to take away from this? 14. You rollover “definition” isn’t a universal definition, it’s just one you assumed for purposes of this paper. Others have made different assumptions. See Mann’s paper (using 14-day debt-free period to determine whether loan rolled over or not). You need to explain why you chose 2 days instead of some other period, why that’s an appropriate choice, and how your results might have varied if you had made a Mann-like choice, for example. 15. In general, you should refer to payday borrowers as “borrowers” rather than “customers.” Globally. 16. You have a tendency to launch right into the numbers (see, e.g., the first full paragraph after the Section 3.2 header), rather than state what you sought to study, why it matters, how to interpret the results, and what the implication is for policymakers. Put yourself in the shoes of the reader and take a more gradual start, then dump the number, and finally explain what they mean. 17. The astounding relationship between sustained use and outcome should be more fully developed. Why does this relationship exist? What theories from economics should have put us on notice of this result? Why is the market a better judge of who should obtain credit than a state legislature????? 17. I think the “default” discussion is somewhat confusing. Going back to the original purpose of this inquiry, opponents of payday lending hypothesize that defaults are harmful for consumers, although there seems to no data to support that hypothesis. We want to test this hypothesis and report the results of our testing. (At least one possible counterfactual is that defaults are actually welfare-enhancing because the borrower gets to keep the loan principal and collection efforts are largely ineffective. This may explain what is going on.) In any event, we once again launch directly into the numbers without explaining why we are making this inquiry and why anyone should care about it. We then don’t connect the results to the original question. 18. The second part of the “default” discussion is equally lost here, which is whether it is possible to identify ex ante the loan applicants who will have adverse outcomes from borrowing, and simply deny them credit through scoring or otherwise. For purpose of this discussion, an “adverse” outcome is one that harms the borrower, not the lender. This is part of a popular new theory of consumer protection that involves having the lender make sure that the loans are “safe” or “suitable” or “affordable” for the consumer, which is a different inquiry from whether the consumer is “creditworthy.” We wanted to test our ability to identify and segregate these consumers. The discussion is lost, and I fear that the data aren’t adequately explained. 19. In the conclusion, the “$64 question” is the rollover one. But we misstate (or don’t reach) the consumer-harm question related to suitability. 20. What happened to the general estimating equation? 21. I am out of time and I want to give you some feedback on a few of the tables. I will try to do that later tonight, tomorrowor Friday. As you say, you have some blockbuster material. The paper needs to be a bit more literary and to “sing.” I think this can be accomplished primarily by slowing down and taking the reader ponderously through each analysis, explanation and, importantly, meaning. It is going to be extremely good when done. I’ll be on my cell tomorrow most of the day if you want to chat, and I have email access. Thank you! Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, February 17, 2014 7:44 PM To: Hilary B. Miller Subject: Full Draft Hi Hilary Attached is my full draft of the paper. I anticipate a round or two of edits...but I think it generally makes a contribution to the current research on the topic. I will be around most of tomorrow if you want to get on the phone at any point. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? No rush...just checking. Sent from my iPhone On Feb 19, 2014, at 4:28 PM, "Hilary B. Miller" wrote: Yes. Got it. I am going to write comments for you later today. Sorry for the delay. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 19, 2014 4:18 PM To: Hilary B. Miller Subject: Checking In Hi Hilary - just checking to make sure you received the draft and to see if you had made edits? Or if you would like to discuss. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - just checking to make sure you received the draft and to see if you had made edits? Or if you would like to discuss. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Attached is my full draft of the paper. I anticipate a round or two of edits...but I think it generally makes a contribution to the current research on the topic. I will be around most of tomorrow if you want to get on the phone at any point. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary A quick update. I finished the draft tonight. I would like to read through it again tomorrowbefore I send it to you. So, you can look for a final draft from me tomorrow afternoon. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Can you call me? Sent from my iPhone On Feb 12, 2014, at 11:42 AM, "Hilary B. Miller" wrote: Do I read this correctly — consumers who default have about the same change in their pre-/post- score, regardless of when the default occurs? Is the comparison the change from 2006-2007 or 2006-2008? I’m not sure I understand what is represented here. u. -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday February 12, 2014, 11:37 AM To: Hilary B. Miller [hilary@miller.net] Subject: Change in Credit Score by Loan of First Default Take a look at this...there is an out of trend increase at Loans 4 and 5...but its small. And, to be clear, this is not isolating rollovers...just the loan at which they defaulted. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:31:51 AM Subject: RE: This (I think) is really good news Yes, just for 0 through 5 rollovers. Are the differences large? Significant? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:31 AM To: Hilary B. Miller Subject: Re: This (I think) is really good news The answer is "yes" ...those who default earlier have a larger decrease in credit scores. How would you like to see this reported? The numbers start to get small as you move out - so I dont think BINS is the right way to go...would you like to see avg pct decrease by loan of first default? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:19:11 AM Subject: RE: This (I think) is really good news Yes, makes sense – people who never default (regardless of how they arrange not to do so) should have higher credit scores. What is not clear, however, from your previous run is whether people who default “sooner” have worse changes in their credit scores over the ensuing year than people who default “later” – I think you only look at default/no-default. (This is “question 2” below). From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:07 AM To: Hilary B. Miller Subject: This (I think) is really good news Hilary Look at the attached. There are (I think) two take aways here... 1. There is a difference in Credit Scores between those who default on Payday loans and those who do not... This is likely intuitive...but... 2. Those who engage in sustained usage...actually have credit scores that are closer to those who don't default. You will see that, among those who eventually have a payday loan default, those who default on Loan 1 have much lower credit scores than those whose first default is on loan 6. I think the critics would contend that those who even have a loan 6 would be in worse shape. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Take a look at this...there is an out of trend increase at Loans 4 and 5...but its small. And, to be clear, this is not isolating rollovers...just the loan at which they defaulted. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:31:51 AM Subject: RE: This (I think) is really good news Yes, just for 0 through 5 rollovers. Are the differences large? Significant? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:31 AM To: Hilary B. Miller Subject: Re: This (I think) is really good news The answer is "yes" ...those who default earlier have a larger decrease in credit scores. How would you like to see this reported? The numbers start to get small as you move out - so I dont think BINS is the right way to go...would you like to see avg pct decrease by loan of first default? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:19:11 AM Subject: RE: This (I think) is really good news Yes, makes sense – people who never default (regardless of how they arrange not to do so) should have higher credit scores. What is not clear, however, from your previous run is whether people who default “sooner” have worse changes in their credit scores over the ensuing year than people who default “later” – I think you only look at default/no-default. (This is “question 2” below). From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:07 AM To: Hilary B. Miller Subject: This (I think) is really good news Hilary Look at the attached. There are (I think) two take aways here... 1. There is a difference in Credit Scores between those who default on Payday loans and those who do not... This is likely intuitive...but... 2. Those who engage in sustained usage...actually have credit scores that are closer to those who don't default. You will see that, among those who eventually have a payday loan default, those who default on Loan 1 have much lower credit scores than those whose first default is on loan 6. I think the critics would contend that those who even have a loan 6 would be in worse shape. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Everything is "significant" statistically, because of the number of observations (be wary of people reporting p-values with 50,000 obs). The issue is one of "practical" significance. Let me create the graphic and send it to you. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:31:51 AM Subject: RE: This (I think) is really good news Yes, just for 0 through 5 rollovers. Are the differences large? Significant? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:31 AM To: Hilary B. Miller Subject: Re: This (I think) is really good news The answer is "yes" ...those who default earlier have a larger decrease in credit scores. How would you like to see this reported? The numbers start to get small as you move out - so I dont think BINS is the right way to go...would you like to see avg pct decrease by loan of first default? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:19:11 AM Subject: RE: This (I think) is really good news Yes, makes sense – people who never default (regardless of how they arrange not to do so) should have higher credit scores. What is not clear, however, from your previous run is whether people who default “sooner” have worse changes in their credit scores over the ensuing year than people who default “later” – I think you only look at default/no-default. (This is “question 2” below). From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:07 AM To: Hilary B. Miller Subject: This (I think) is really good news Hilary Look at the attached. There are (I think) two take aways here... 1. There is a difference in Credit Scores between those who default on Payday loans and those who do not... This is likely intuitive...but... 2. Those who engage in sustained usage...actually have credit scores that are closer to those who don't default. You will see that, among those who eventually have a payday loan default, those who default on Loan 1 have much lower credit scores than those whose first default is on loan 6. I think the critics would contend that those who even have a loan 6 would be in worse shape. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? The answer is "yes" ...those who default earlier have a larger decrease in credit scores. How would you like to see this reported? The numbers start to get small as you move out - so I dont think BINS is the right way to go...would you like to see avg pct decrease by loan of first default? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, February 12, 2014 10:19:11 AM Subject: RE: This (I think) is really good news Yes, makes sense – people who never default (regardless of how they arrange not to do so) should have higher credit scores. What is not clear, however, from your previous run is whether people who default “sooner” have worse changes in their credit scores over the ensuing year than people who default “later” – I think you only look at default/no-default. (This is “question 2” below). From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, February 12, 2014 10:07 AM To: Hilary B. Miller Subject: This (I think) is really good news Hilary Look at the attached. There are (I think) two take aways here... 1. There is a difference in Credit Scores between those who default on Payday loans and those who do not... This is likely intuitive...but... 2. Those who engage in sustained usage...actually have credit scores that are closer to those who don't default. You will see that, among those who eventually have a payday loan default, those who default on Loan 1 have much lower credit scores than those whose first default is on loan 6. I think the critics would contend that those who even have a loan 6 would be in worse shape. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary Look at the attached. There are (I think) two take aways here... 1. There is a difference in Credit Scores between those who default on Payday loans and those who do not... This is likely intuitive...but... 2. Those who engage in sustained usage...actually have credit scores that are closer to those who don't default. You will see that, among those who eventually have a payday loan default, those who default on Loan 1 have much lower credit scores than those whose first default is on loan 6. I think the critics would contend that those who even have a loan 6 would be in worse shape. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Sorry - for the delayed response... For 2006-2007, the results were: Default - decrease in Credit Score of almost 9 points. No Default - decrease in Credit Score of 5 points. For 2008 - 2009, the results were: Default - decrease in Credit Score of about .8 points. No Default - increase in Credit Score of about .5 points. This is consistent with other findings...little differences in 2008-2009 but more substantive differences in 2006-2007. Still working on the prediction model... Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Tuesday, February 11, 2014 11:06:08 AM Subject: RE: Talk today? Related question on #2 -- how do score changes for defaulters compare with score changes of non-defaulters? -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, February 11, 2014 10:48 AM To: Hilary B. Miller Subject: Re: Talk today? Hi Hilary I ran the first two of your questions...the third is taking some more time. I have attached the results of the first two. The results are effectively this... 1. The probability of default starts high (about 18%) and then drops precipitously as the number of rollovers increases. 2. There is effectively no difference on the change in credit scores based upon payday loan default (I ran this overall and by state). 3. - still working on this. Regarding your question about the number of rollovers being lower than expected...recall that there is a substantive proportion of the dataset that did not rollover a loan - even under our fairly liberal definition. So, there are ALOT of 0's in the file for number of rollovers. When these are averaged in with those who do rollover, the average may be lower than expected. If the 0's are suppressed, the mean will go up. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Monday, February 10, 2014 3:50:29 PM Subject: RE: Talk today? This looks like a good start. I think from copy-editing, but there's a long A few things aren't phrased the way I can give you comments on those issues it will benefit way to go on that. would do it, but I later. The rollover count -- in the range of 3 -- seems very low by the standards of other literature (Mann, CFPB White Paper). You will want to expand your literature survey and include the CFPB White Paper and Kaufman. Since this paper expressly builds on Bhutta et al. (2013) and Kaufman, you'll want to explain how. Good start. Thank you! HM -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 8:25 PM To: Hilary B. Miller Subject: Re: Talk today? Hi Hilary As we discussed, this is the current working draft of the final report...its not done, but you can get a good feel for where I am going... I am going to the incremental analysis tomorrow... Please provide edits/commentary on the paper as needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Sunday, February 9, 2014 4:43:38 PM Subject: RE: Talk today? Confirming our telephone call today: We would like to add additional analysis regarding outcomes for borrowers who take out an initial payday loan and then default. By definition, everyone in our dataset has not previously had a payday loan with this lender during the 60 days preceding his or her first loan (consistent with the definition used by Ronald Mann). 1. What percentage of borrowers default after n substantially consecutive (i.e., within 2 days) loans, where n=1,2...8? 2. What are the impacts of defaults after 1,2,...n loans on credit scores a year later? 3. Can we identify, through a scorecard method, those borrowers most likely to default after one loan? In prospective use of this model, what can we expect as the proportion of type I and type II errors (split the sample?)? -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 1:49 PM To: Hilary B. Miller Subject: Talk today? Sent from my iPhone Hi Hilary I ran the first two of your questions...the third is taking some more time. I have attached the results of the first two. The results are effectively this... 1. The probability of default starts high (about 18%) and then drops precipitously as the number of rollovers increases. 2. There is effectively no difference on the change in credit scores based upon payday loan default (I ran this overall and by state). 3. - still working on this. Regarding your question about the number of rollovers being lower than expected...recall that there is a substantive proportion of the dataset that did not rollover a loan - even under our fairly liberal definition. So, there are ALOT of 0's in the file for number of rollovers. When these are averaged in with those who do rollover, the average may be lower than expected. If the 0's are suppressed, the mean will go up. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Monday, February 10, 2014 3:50:29 PM Subject: RE: Talk today? This looks like a good start. I think from copy-editing, but there's a long A few things aren't phrased the way I can give you comments on those issues it will benefit way to go on that. would do it, but I later. The rollover count -- in the range of 3 -- seems very low by the standards of other literature (Mann, CFPB White Paper). You will want to expand your literature survey and include the CFPB White Paper and Kaufman. Since this paper expressly builds on Bhutta et al. (2013) and Kaufman, you'll want to explain how. Good start. Thank you! HM -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 8:25 PM To: Hilary B. Miller Subject: Re: Talk today? Hi Hilary As we discussed, this is the current working draft of the final report...its not done, but you can get a good feel for where I am going... I am going to the incremental analysis tomorrow... Please provide edits/commentary on the paper as needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Sunday, February 9, 2014 4:43:38 PM Subject: RE: Talk today? Confirming our telephone call today: We would like to add additional analysis regarding outcomes for borrowers who take out an initial payday loan and then default. By definition, everyone in our dataset has not previously had a payday loan with this lender during the 60 days preceding his or her first loan (consistent with the definition used by Ronald Mann). 1. What percentage of borrowers default after n substantially consecutive (i.e., within 2 days) loans, where n=1,2...8? 2. What are the impacts of defaults after 1,2,...n loans on credit scores a year later? 3. Can we identify, through a scorecard method, those borrowers most likely to default after one loan? In prospective use of this model, what can we expect as the proportion of type I and type II errors (split the sample?)? -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 1:49 PM To: Hilary B. Miller Subject: Talk today? Sent from my iPhone Hi Hilary As we discussed, this is the current working draft of the final report...its not done, but you can get a good feel for where I am going... I am going to the incremental analysis tomorrow... Please provide edits/commentary on the paper as needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Sunday, February 9, 2014 4:43:38 PM Subject: RE: Talk today? Confirming our telephone call today: We would like to add additional analysis regarding outcomes for borrowers who take out an initial payday loan and then default. By definition, everyone in our dataset has not previously had a payday loan with this lender during the 60 days preceding his or her first loan (consistent with the definition used by Ronald Mann). 1. What percentage of borrowers default after n substantially consecutive (i.e., within 2 days) loans, where n=1,2...8? 2. What are the impacts of defaults after 1,2,...n loans on credit scores a year later? 3. Can we identify, through a scorecard method, those borrowers most likely to default after one loan? In prospective use of this model, what can we expect as the proportion of type I and type II errors (split the sample?)? -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 1:49 PM To: Hilary B. Miller Subject: Talk today? Sent from my iPhone thanks for the summary. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Sunday, February 9, 2014 4:43:38 PM Subject: RE: Talk today? Confirming our telephone call today: We would like to add additional analysis regarding outcomes for borrowers who take out an initial payday loan and then default. By definition, everyone in our dataset has not previously had a payday loan with this lender during the 60 days preceding his or her first loan (consistent with the definition used by Ronald Mann). 1. What percentage of borrowers default after n substantially consecutive (i.e., within 2 days) loans, where n=1,2...8? 2. What are the impacts of defaults after 1,2,...n loans on credit scores a year later? 3. Can we identify, through a scorecard method, those borrowers most likely to default after one loan? In prospective use of this model, what can we expect as the proportion of type I and type II errors (split the sample?)? -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Sunday, February 09, 2014 1:49 PM To: Hilary B. Miller Subject: Talk today? Sent from my iPhone Sure. Let's talk tomorrow afternoon. Anytime between 1-4 works for me. Sent from my iPhone On Feb 8, 2014, at 12:22 PM, "Hilary B. Miller" wrote: Jennifer — I know you are busy writing. I have had a small epiphany on the “default” topic (which is ancillary and not addressed in the synopsis I proposed to you). I think that the answer to a significant policy question lies in slicing the same default baloney in a slightly different way from the way I had previously suggested. Might you have a couple of minutes to discuss this onMonday (or even over the weekend — I’m working)? Sorry to keep moving the ball on you. This is a small move. Regards, Hilary . Great - a template would be helpful. I was using the Desai paper as my guide...but I want to use whatever form is preferable. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, February 3, 2014 6:26:58 PM Subject: RE: Update I think steps less than a bodyguard (such as, for example, a guard dog or barbed wire at your residence) may suffice. Please do take the weekend. More important to get it right than fast. In terms of form, Sandler’s paper was just about right — standard APA works best. I have a template if you need one. HM u. -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Monday February 3, 2014, 06:24 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Update Sure. In that case, I may take the weekend. Should I hire a bodyguard? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, February 3, 2014 5:23:34 PM Subject: Re: Update Jennifer, thanks for this progress report. Please take a little extra time and make sure  you are going to be happy with both the form and substance of your report. The nature  of your findings suggests that you will be subject to intense scrutiny from opponents of  the industry. I want to make sure we have anticipated their criticisms. _____________ On Feb 3, 2014, at 2:25 PM, "Jennifer Lewis Priestley"  wrote: Hi Hilary I just wanted to give you an update...I have made good progress - I expect to have a completed draft to you before the end of the week (look for something aroundThursday). :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 30, 2014 11:11:17 AM Subject: RE: Abstract ­ First Pass ­ Subject to Further Thought and Your Input Abstract The discourse surrounding payday loans has recently focused sharply on consumers' propensity to “roll over” these loans, which are typically two-week, very-high-cost advances. The industry’s principal regulator has suggested that this sustained usage may be harmful to consumers. Exploiting interstate differences in rollover regulation, and using administrative data supplied by three lenders for 28,000 borrowers that have been matched to credit scores from a national credit reporting agency, I explore the effectiveness of various regulatory schemes in improving consumer outcomes in the years following initial payday borrowing. I also evaluate the effects of sustained payday-loan usage irrespective of regulatory scheme. I find that, while state regulation has a small effect on longer-term usage patterns, consumers whose borrowing is unrestricted by regulation fare better than consumers in the most restrictive states, after controlling for initial financial status. I also find that longer-term borrowers (three months or more) have better outcomes than consumers whose borrowing is concluded in one month or less. These findings raise significant policy questions and suggest the appropriateness of further study of actual consumer outcomes before the imposition of new regulation at the federal level. u. -----Original Message----From: Hilary B. Miller [hilary@miller.net] Received: Thursday January 30, 2014, 11:09 AM To: Priestley, Jennifer Lewis [jpriestl@kennesaw.edu] Subject: Abstract - First Pass - Subject to Further Thought and Your Input Abstract Sure. In that case, I may take the weekend. Should I hire a bodyguard? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, February 3, 2014 5:23:34 PM Subject: Re: Update Jennifer, thanks for this progress report. Please take a little extra time and make sure  you are going to be happy with both the form and substance of your report. The nature  of your findings suggests that you will be subject to intense scrutiny from opponents of  the industry. I want to make sure we have anticipated their criticisms. _____________ On Feb 3, 2014, at 2:25 PM, "Jennifer Lewis Priestley"  wrote: Hi Hilary - I just wanted to give you an update...I have made good progress - I expect to have a completed draft to you before the end of the week (look for something aroundThursday). :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 30, 2014 11:11:17 AM Subject: RE: Abstract ­ First Pass ­ Subject to Further Thought and Your Input Abstract The discourse surrounding payday loans has recently focused sharply on consumers' propensity to “roll over” these loans, which are typically two-week, very-high-cost advances. The industry’s principal regulator has suggested that this sustained usage may be harmful to consumers. Exploiting interstate differences in rollover regulation, and using administrative data supplied by three lenders for 28,000 borrowers that have been matched to credit scores from a national credit reporting agency, I explore the effectiveness of various regulatory schemes in improving consumer outcomes in the years following initial payday borrowing. I also evaluate the effects of sustained payday-loan usage irrespective of regulatory scheme. I find that, while state regulation has a small effect on longer-term usage patterns, consumers whose borrowing is unrestricted by regulation fare better than consumers in the most restrictive states, after controlling for initial financial status. I also find that longer-term borrowers (three months or more) have better outcomes than consumers whose borrowing is concluded in one month or less. These findings raise significant policy questions and suggest the appropriateness of further study of actual consumer outcomes before the imposition of new regulation at the federal level. u. -----Original Message----From: Hilary B. Miller [hilary@miller.net] Received: Thursday January 30, 2014, 11:09 AM To: Priestley, Jennifer Lewis [jpriestl@kennesaw.edu] Subject: Abstract - First Pass - Subject to Further Thought and Your Input Abstract Hi Hilary I just wanted to give you an update...I have made good progress - I expect to have a completed draft to you before the end of the week (look for something around Thursday). :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 30, 2014 11:11:17 AM Subject: RE: Abstract ­ First Pass ­ Subject to Further Thought and Your Input Abstract The discourse surrounding payday loans has recently focused sharply on consumers' propensity to “roll over” these loans, which are typically two-week, very-high-cost advances. The industry’s principal regulator has suggested that this sustained usage may be harmful to consumers. Exploiting interstate differences in rollover regulation, and using administrative data supplied by three lenders for 28,000 borrowers that have been matched to credit scores from a national credit reporting agency, I explore the effectiveness of various regulatory schemes in improving consumer outcomes in the years following initial payday borrowing. I also evaluate the effects of sustained payday-loan usage irrespective of regulatory scheme. I find that, while state regulation has a small effect on longer-term usage patterns, consumers whose borrowing is unrestricted by regulation fare better than consumers in the most restrictive states, after controlling for initial financial status. I also find that longer-term borrowers (three months or more) have better outcomes than consumers whose borrowing is concluded in one month or less. These findings raise significant policy questions and suggest the appropriateness of further study of actual consumer outcomes before the imposition of new regulation at the federal level. u. -----Original Message----From: Hilary B. Miller [hilary@miller.net] Received: Thursday January 30, 2014, 11:09 AM To: Priestley, Jennifer Lewis [jpriestl@kennesaw.edu] Subject: Abstract - First Pass - Subject to Further Thought and Your Input Abstract Excellent. Thank you. Sent from my iPhone On Jan 30, 2014, at 11:11 AM, "Hilary B. Miller" wrote: Abstract The discourse surrounding payday loans has recently focused sharply on consumers' propensity to “roll over” these loans, which are typically two-week, very-high-cost advances. The industry’s principal regulator has suggested that this sustained usage may be harmful to consumers. Exploiting interstate differences in rollover regulation, and using administrative data supplied by three lenders for 28,000 borrowers that have been matched to credit scores from a national credit reporting agency, I explore the effectiveness of various regulatory schemes in improving consumer outcomes in the years following initial payday borrowing. I also evaluate the effects of sustained payday-loan usage irrespective of regulatory scheme. I find that, while state regulation has a small effect on longer-term usage patterns, consumers whose borrowing is unrestricted by regulation fare better than consumers in the most restrictive states, after controlling for initial financial status. I also find that longer-term borrowers (three months or more) have better outcomes than consumers whose borrowing is concluded in one month or less. These findings raise significant policy questions and suggest the appropriateness of further study of actual consumer outcomes before the imposition of new regulation at the federal level. u. -----Original Message----From: Hilary B. Miller [hilary@miller.net] Received: Thursday January 30, 2014, 11:09 AM To: Priestley, Jennifer Lewis [jpriestl@kennesaw.edu] Subject: Abstract - First Pass - Subject to Further Thought and Your Input Abstract I think you hit send prematurely? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 30, 2014 11:09:16 AM Subject: Abstract ­ First Pass ­ Subject to Further Thought and Your Input Abstract Sure. Just call when you are ready. I have an appt at 1:30 but I am open before that. Sent from my iPhone On Jan 28, 2014, at 9:37 AM, "Hilary B. Miller" wrote: I don’t understand why this is so. I’ll give you a call in a while and we can discuss it. I’ll be in the car so it may be a bit weird. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Tuesday, January 28, 2014 9:34 AM To: Hilary B. Miller Subject: Re: Revised Results Sorry for the delay - trust the dentist went well. I have updated tables x.10 and x.11. Basically, when the modeling is done around the probability of NOT rolling over more than 90 days, the effects just change sign (from pos to neg). See attached. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, January 28, 2014 6:19:18 AM Subject: Re: Revised Results Did I miss this yesterday? On Jan 26, 2014, at 4:14 PM, "Jennifer Lewis Priestley"   wrote: Will do.  Look for it tomorrow. Sent from my iPhone On Jan 26, 2014, at 3:05 PM, "Hilary B. Miller"  wrote: Let’s take a look and see. Thanks. . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday January 26, 2014, 02:57 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Revised Results Thanks for the note. I wrote the code to develop the estimates around the people who DO default (event = 1). I can rerun the computation around the people who DO NOT default (event = 0). Same basic results, different interpretation. I can then put that in these tables. Would that work? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, January 25, 2014 5:40:34 PM Subject: RE: Revised Results Jennifer —   This looks as if it is as much, if not more, data than we can use. So we will certainly  have our choices of which tables to use to tell the story. With the addition of the new  tables, it appears that there are meaningful differences between states in patterns of  usage. Perhaps paradoxically, those states with the “longest” usage are not the states  with the worst outcomes, as antagonists of the industry posit. I suppose this finding is  consistent with our previous observations that borrowers who are empowered to use  credit as they need it == rather than being artificially constrained by regulation — do  best. That’s a fine message.   I do have one question about the material presented around tables x.10 and x.11. The  idea here should be to identify borrowers who (a) borrow for an excessively long time,  and who (b) do not default. I’m not sure the (b) part is being captured here.    Otherwise, looks good. Thanks.   Hilary . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 22, 2014, 11:29 PM To: Hilary B. Miller [hilary@miller.net] Subject: Revised Results Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion on Friday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Sorry for the delay - trust the dentist went well. I have updated tables x.10 and x.11. Basically, when the modeling is done around the probability of NOT rolling over more than 90 days, the effects just change sign (from pos to neg). See attached. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, January 28, 2014 6:19:18 AM Subject: Re: Revised Results Did I miss this yesterday? On Jan 26, 2014, at 4:14 PM, "Jennifer Lewis Priestley"   wrote: Will do.  Look for it tomorrow. Sent from my iPhone On Jan 26, 2014, at 3:05 PM, "Hilary B. Miller"  wrote: Let’s take a look and see. Thanks. . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday January 26, 2014, 02:57 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Revised Results Thanks for the note. I wrote the code to develop the estimates around the people who DO default (event = 1). I can rerun the computation around the people who DO NOT default (event = 0). Same basic results, different interpretation. I can then put that in these tables. Would that work? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, January 25, 2014 5:40:34 PM Subject: RE: Revised Results Jennifer — This looks as if it is as much, if not more, data than we can use. So we will certainly  have our choices of which tables to use to tell the story. With the addition of the new  tables, it appears that there are meaningful differences between states in patterns of  usage. Perhaps paradoxically, those states with the “longest” usage are not the states  with the worst outcomes, as antagonists of the industry posit. I suppose this finding is  consistent with our previous observations that borrowers who are empowered to use  credit as they need it == rather than being artificially constrained by regulation — do  best. That’s a fine message. I do have one question about the material presented around tables x.10 and x.11. The  idea here should be to identify borrowers who (a) borrow for an excessively long time,  and who (b) do not default. I’m not sure the (b) part is being captured here.  Otherwise, looks good. Thanks. Hilary . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 22, 2014, 11:29 PM To: Hilary B. Miller [hilary@miller.net] Subject: Revised Results Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion on Friday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? No. Sorry. Got tied up yesterday. Let me get the kids to school and then will get it off first thing Sent from my iPhone On Jan 28, 2014, at 6:19 AM, "Hilary B. Miller" wrote: Did I miss this yesterday? On Jan 26, 2014, at 4:14 PM, "Jennifer Lewis Priestley" wrote: Will do. Look for it tomorrow. Sent from my iPhone On Jan 26, 2014, at 3:05 PM, "Hilary B. Miller" wrote: Let’s take a look and see. Thanks. . -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday January 26, 2014, 02:57 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Revised Results Thanks for the note. I wrote the code to develop the estimates around the people who DO default (event = 1). I can rerun the computation around the people who DO NOT default (event = 0). Same basic results, different interpretation. I can then put that in these tables. Would that work? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, January 25, 2014 5:40:34 PM Subject: RE: Revised Results Jennifer — This looks as if it is as much, if not more, data than we can use. So we will certainly  have our choices of which tables to use to tell the story. With the addition of the new  tables, it appears that there are meaningful differences between states in patterns of  usage. Perhaps paradoxically, those states with the “longest” usage are not the states  with the worst outcomes, as antagonists of the industry posit. I suppose this finding is  consistent with our previous observations that borrowers who are empowered to use  credit as they need it == rather than being artificially constrained by regulation — do  best. That’s a fine message. I do have one question about the material presented around tables x.10 and x.11. The  idea here should be to identify borrowers who (a) borrow for an excessively long time,  and who (b) do not default. I’m not sure the (b) part is being captured here.  Otherwise, looks good. Thanks. Hilary . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 22, 2014, 11:29 PM To: Hilary B. Miller [hilary@miller.net] Subject: Revised Results Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion on Friday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Will do. Look for it tomorrow. Sent from my iPhone On Jan 26, 2014, at 3:05 PM, "Hilary B. Miller" wrote: Let’s take a look and see. Thanks. . -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday January 26, 2014, 02:57 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Revised Results Thanks for the note. I wrote the code to develop the estimates around the people who DO default (event = 1). I can rerun the computation around the people who DO NOT default (event = 0). Same basic results, different interpretation. I can then put that in these tables. Would that work? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, January 25, 2014 5:40:34 PM Subject: RE: Revised Results Jennifer — This looks as if it is as much, if not more, data than we can use. So we will certainly  have our choices of which tables to use to tell the story. With the addition of the new  tables, it appears that there are meaningful differences between states in patterns of  usage. Perhaps paradoxically, those states with the “longest” usage are not the states  with the worst outcomes, as antagonists of the industry posit. I suppose this finding is  consistent with our previous observations that borrowers who are empowered to use  credit as they need it == rather than being artificially constrained by regulation — do  best. That’s a fine message. I do have one question about the material presented around tables x.10 and x.11. The  idea here should be to identify borrowers who (a) borrow for an excessively long time,  and who (b) do not default. I’m not sure the (b) part is being captured here.  Otherwise, looks good. Thanks. Hilary . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 22, 2014, 11:29 PM To: Hilary B. Miller [hilary@miller.net] Subject: Revised Results Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion onFriday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks for the note. I wrote the code to develop the estimates around the people who DO default (event = 1). I can rerun the computation around the people who DO NOT default (event = 0). Same basic results, different interpretation. I can then put that in these tables. Would that work? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, January 25, 2014 5:40:34 PM Subject: RE: Revised Results Jennifer — This looks as if it is as much, if not more, data than we can use. So we will certainly  have our choices of which tables to use to tell the story. With the addition of the new  tables, it appears that there are meaningful differences between states in patterns of  usage. Perhaps paradoxically, those states with the “longest” usage are not the states  with the worst outcomes, as antagonists of the industry posit. I suppose this finding is  consistent with our previous observations that borrowers who are empowered to use  credit as they need it == rather than being artificially constrained by regulation — do  best. That’s a fine message. I do have one question about the material presented around tables x.10 and x.11. The  idea here should be to identify borrowers who (a) borrow for an excessively long time,  and who (b) do not default. I’m not sure the (b) part is being captured here.  Otherwise, looks good. Thanks. Hilary . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 22, 2014, 11:29 PM To: Hilary B. Miller [hilary@miller.net] Subject: Revised Results Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion on Friday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary Sorry this is a bit late. I have incorporated all of my notes from our discussion on Friday. At this stage - my focus was to get the tables (analysis) completed - then focus on the writing...take a look and let me know what analytical edits are still needed. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I am here - just call me when you are ready. 770-423-6107 Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, January 17, 2014 12:42:50 PM Subject: RE: Call I forget – are you calling me? From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, January 16, 2014 9:09 AM To: Hilary B. Miller Subject: Re: Call Thats fine. I have class at 11 - but otherwise I should be here at my desk. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 16, 2014 9:07:42 AM Subject: RE: Call Certainly — I’m in a meeting and will call you as soon as I finish, probably just before  10:00. . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Thursday January 16, 2014, 09:05 AM To: Hilary B. Miller [hilary@miller.net] Subject: Call Hilary - Could I trouble you to call me in my office? 770-423-6107 Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thats fine. I have class at 11 - but otherwise I should be here at my desk. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 16, 2014 9:07:42 AM Subject: RE: Call Certainly — I’m in a meeting and will call you as soon as I finish, probably just before  10:00. . ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Thursday January 16, 2014, 09:05 AM To: Hilary B. Miller [hilary@miller.net] Subject: Call Hilary - Could I trouble you to call me in my office? 770-423-6107 Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - Could I trouble you to call me in my office? 770-423-6107 Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks Hilary! I lived there for a few years - my favorite pub was the Grenadier on Wilton Crescent. :) Safe Travels. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  Cc: "Jennifer Lewis Priestley" , "Ronald J. Mann"  , "Victor O. Stango"  Sent: Thursday, January 9, 2014 10:10:45 AM Subject: Homonoff http://scholar.princeton.edu/jgoldin/files/Goldin_Homonoff_11_5_2013.pdf HM. Hi Hilary - safe Travels... The basic takeway is that yes, ><90 days duration for a loan can be predicted by the same factors that were used to predict default. If you think that this is a relevant finding, I can include this information in the results section. Can we plan to speak on the morning of the 16th? I am available from 9 - 10:45 (I have class at 11). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, January 8, 2014 9:25:33 AM Subject: RE: Brief Update I’m in London on business — will read this on the plane home tomorrow. I looked briefly at the <> 90 analysis and it’s not immediately obvious to me what that represents, so let’s discuss, possibly next week when I am back in the office (first full day, 1/15). . -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday January 8, 2014, 01:30 AM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Brief Update Hilary - sorry for the delay. Attached is the developing results narrative...I like the themes that are emerging...basically that sustained usage of rollovers (frequency and pct of total payday loans) leads to a stronger financial position. :) I have also attached the "Over 90 Days" analysis that you had requested (<90 days = 0, 90+ days = 1). Basically, the trends are the same - the model predicts well (see the ROC curve). Let me know if you want me to integrate these results. I will be at a client most of tomorrow...but back online on Thursday. Let me know if you want to discuss. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, January 7, 2014 5:48:56 AM Subject: Re: Brief Update Excellent. Did you get a chance to do the analyses suggested in my email to you of 1/2? On Jan 7, 2014, at 4:13 AM, "Jennifer Lewis Priestley"  wrote: Hilary I just wanted to let you know that I am working on the results narrative - I expect to have a substantive results section to you by tomorrow afternoon. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - sorry for the delay. Attached is the developing results narrative...I like the themes that are emerging...basically that sustained usage of rollovers (frequency and pct of total payday loans) leads to a stronger financial position. :) I have also attached the "Over 90 Days" analysis that you had requested (<90 days = 0, 90+ days = 1). Basically, the trends are the same - the model predicts well (see the ROC curve). Let me know if you want me to integrate these results. I will be at a client most of tomorrow...but back online on Thursday. Let me know if you want to discuss. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, January 7, 2014 5:48:56 AM Subject: Re: Brief Update Excellent. Did you get a chance to do the analyses suggested in my email to you of 1/2? On Jan 7, 2014, at 4:13 AM, "Jennifer Lewis Priestley"  wrote: Hilary I just wanted to let you know that I am working on the results narrative - I expect to have a substantive results section to you by tomorrow afternoon. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I did - I will send it as a separate attachment - but the patterns were generally the same. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, January 7, 2014 5:48:56 AM Subject: Re: Brief Update Excellent. Did you get a chance to do the analyses suggested in my email to you of 1/2? On Jan 7, 2014, at 4:13 AM, "Jennifer Lewis Priestley"  wrote: Hilary I just wanted to let you know that I am working on the results narrative - I expect to have a substantive results section to you by tomorrow afternoon. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary I just wanted to let you know that I am working on the results narrative - I expect to have a substantive results section to you by tomorrow afternoon. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Sorry - I hit send prematurely. I meant to add that they could be rolling it over UNTIL they are ready to pay for it. So, basically the rollover is buying them time until they have the resources to pay it off. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 2, 2014 2:38:46 PM Subject: RE: Document for call tomorrow That’s amazing and both logical and counterintuitive! ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Thursday January 2, 2014, 02:35 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Document for call tomorrow An initial check indicates the same pattern of significance. Repossessions, public records and collections balances are the strongest predictors. As I was checking to see how rollovers are related to default rates I found an interesting and very strong relationship between number of times a loan is rolled over and prob of default. Basically, as the number of times rolled over increases, the likelihood of default decreases - at a strong rate. See the attached (the value of 10 is actually 10+). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 2, 2014 12:16:32 PM Subject: RE: Document for call tomorrow Great. Do you think it is worth doing a similar multivariate analysis to determine how to identify those loan applicants most likely to stay in protracted debt? There are several possible formulations of this question: 1. If we define a negative outcome as rolling over for 90 or more days (using your present 2-day definition), what predicts that? 2. If we use a more inclusive definition of “rollover” that expands the time frame from 2 to 14 days (see, e.g., Mann, Ronald J., Assessing the Optimism of Payday Loan Borrowers (March 12, 2013). Columbia Law and Economics Working Paper No. 443. Available at SSRN:http://ssrn.com/abstract=2232954), and look at who then rolls over (as redefined) for > 90 days, what predicts that? 3. Under either scenario 1 or 2 above, in the “worst case” outcome, the borrower rolls over for > 90 days and then is charged off. This is the borrower who never had the means to repay in the first place but arguably just tried, unsuccessfully, to put off the day of reckoning. What predicts this behavior? I think these additional answers would be the end of the line unless turning over these rocks generates some huge surprises. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, January 02, 2014 12:02 PM To: Hilary B. Miller Subject: Document for call tomorrow Hilary - in preparation for our call tomorrow at 10, I have attached a document for discussion. Basically, I created the foundations of a "score card" for this segment - looking at the presence of different conditions that would contribute to default on a payday loan. The first part of the analysis is just a characteristic like aggregate balances in collections. For those with a balance <$500, the probability of default is 1.06%. For those with a balance >$2500, the probability of default is 2.6% (the overall average default rate is 1.98%). So, there is some separation there (2.45x more likely to default). On page 3, you will see that those with <3 public records have a default probability of 1.75% while those with >=3 public records, the probability of default is 2.8% (1.6x more likely to default). On page 5, you will see that those without a repossession in the last 24 months have a default probability of 1.8% while those with a repossession in the last 24 months have a default probability of 3.6% (2x more likely to default). This is meaningful because these pieces of information would be known pre-loan...and are simple enough to be self reported. Starting on page 7, you will see a logistic regression model - it performed fairly well predicting default (c=.63). The main take away from this output is the odds ratio estimates on page 8 which are a multivariate form of the results above, including the 95% confidence intervals - which are statistically significant. :) Talk to you tomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Actually, the more I think about it, the more it might make sense. You cant default on a loan that does not end. Right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 2, 2014 2:38:46 PM Subject: RE: Document for call tomorrow That’s amazing and both logical and counterintuitive! ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Thursday January 2, 2014, 02:35 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Document for call tomorrow An initial check indicates the same pattern of significance. Repossessions, public records and collections balances are the strongest predictors. As I was checking to see how rollovers are related to default rates I found an interesting and very strong relationship between number of times a loan is rolled over and prob of default. Basically, as the number of times rolled over increases, the likelihood of default decreases - at a strong rate. See the attached (the value of 10 is actually 10+). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 2, 2014 12:16:32 PM Subject: RE: Document for call tomorrow Great. Do you think it is worth doing a similar multivariate analysis to determine how to identify those loan applicants most likely to stay in protracted debt? There are several possible formulations of this question: 1. If we define a negative outcome as rolling over for 90 or more days (using your present 2-day definition), what predicts that? 2. If we use a more inclusive definition of “rollover” that expands the time frame from 2 to 14 days (see, e.g., Mann, Ronald J., Assessing the Optimism of Payday Loan Borrowers (March 12, 2013). Columbia Law and Economics Working Paper No. 443. Available at SSRN:http://ssrn.com/abstract=2232954), and look at who then rolls over (as redefined) for > 90 days, what predicts that? 3. Under either scenario 1 or 2 above, in the “worst case” outcome, the borrower rolls over for > 90 days and then is charged off. This is the borrower who never had the means to repay in the first place but arguably just tried, unsuccessfully, to put off the day of reckoning. What predicts this behavior? I think these additional answers would be the end of the line unless turning over these rocks generates some huge surprises. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, January 02, 2014 12:02 PM To: Hilary B. Miller Subject: Document for call tomorrow Hilary - in preparation for our call tomorrow at 10, I have attached a document for discussion. Basically, I created the foundations of a "score card" for this segment - looking at the presence of different conditions that would contribute to default on a payday loan. The first part of the analysis is just a characteristic like aggregate balances in collections. For those with a balance <$500, the probability of default is 1.06%. For those with a balance >$2500, the probability of default is 2.6% (the overall average default rate is 1.98%). So, there is some separation there (2.45x more likely to default). On page 3, you will see that those with <3 public records have a default probability of 1.75% while those with >=3 public records, the probability of default is 2.8% (1.6x more likely to default). On page 5, you will see that those without a repossession in the last 24 months have a default probability of 1.8% while those with a repossession in the last 24 months have a default probability of 3.6% (2x more likely to default). This is meaningful because these pieces of information would be known pre-loan...and are simple enough to be self reported. Starting on page 7, you will see a logistic regression model - it performed fairly well predicting default (c=.63). The main take away from this output is the odds ratio estimates on page 8 which are a multivariate form of the results above, including the 95% confidence intervals - which are statistically significant. :) Talk to you tomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? An initial check indicates the same pattern of significance. Repossessions, public records and collections balances are the strongest predictors. As I was checking to see how rollovers are related to default rates I found an interesting and very strong relationship between number of times a loan is rolled over and prob of default. Basically, as the number of times rolled over increases, the likelihood of default decreases - at a strong rate. See the attached (the value of 10 is actually 10+). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Thursday, January 2, 2014 12:16:32 PM Subject: RE: Document for call tomorrow Great. Do you think it is worth doing a similar multivariate analysis to determine how to identify those loan applicants most likely to stay in protracted debt? There are several possible formulations of this question: 1. If we define a negative outcome as rolling over for 90 or more days (using your present 2-day definition), what predicts that? 2. If we use a more inclusive definition of “rollover” that expands the time frame from 2 to 14 days (see, e.g., Mann, Ronald J., Assessing the Optimism of Payday Loan Borrowers (March 12, 2013). Columbia Law and Economics Working Paper No. 443. Available at SSRN:http://ssrn.com/abstract=2232954), and look at who then rolls over (as redefined) for > 90 days, what predicts that? 3. Under either scenario 1 or 2 above, in the “worst case” outcome, the borrower rolls over for > 90 days and then is charged off. This is the borrower who never had the means to repay in the first place but arguably just tried, unsuccessfully, to put off the day of reckoning. What predicts this behavior? I think these additional answers would be the end of the line unless turning over these rocks generates some huge surprises. H From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Thursday, January 02, 2014 12:02 PM To: Hilary B. Miller Subject: Document for call tomorrow Hilary - in preparation for our call tomorrow at 10, I have attached a document for discussion. Basically, I created the foundations of a "score card" for this segment - looking at the presence of different conditions that would contribute to default on a payday loan. The first part of the analysis is just a characteristic like aggregate balances in collections. For those with a balance <$500, the probability of default is 1.06%. For those with a balance >$2500, the probability of default is 2.6% (the overall average default rate is 1.98%). So, there is some separation there (2.45x more likely to default). On page 3, you will see that those with <3 public records have a default probability of 1.75% while those with >=3 public records, the probability of default is 2.8% (1.6x more likely to default). On page 5, you will see that those without a repossession in the last 24 months have a default probability of 1.8% while those with a repossession in the last 24 months have a default probability of 3.6% (2x more likely to default). This is meaningful because these pieces of information would be known pre-loan...and are simple enough to be self reported. Starting on page 7, you will see a logistic regression model - it performed fairly well predicting default (c=.63). The main take away from this output is the odds ratio estimates on page 8 which are a multivariate form of the results above, including the 95% confidence intervals - which are statistically significant. :) Talk to you tomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - in preparation for our call tomorrow at 10, I have attached a document for discussion. Basically, I created the foundations of a "score card" for this segment - looking at the presence of different conditions that would contribute to default on a payday loan. The first part of the analysis is just a characteristic like aggregate balances in collections. For those with a balance <$500, the probability of default is 1.06%. For those with a balance >$2500, the probability of default is 2.6% (the overall average default rate is 1.98%). So, there is some separation there (2.45x more likely to default). On page 3, you will see that those with <3 public records have a default probability of 1.75% while those with >=3 public records, the probability of default is 2.8% (1.6x more likely to default). On page 5, you will see that those without a repossession in the last 24 months have a default probability of 1.8% while those with a repossession in the last 24 months have a default probability of 3.6% (2x more likely to default). This is meaningful because these pieces of information would be known pre-loan...and are simple enough to be self reported. Starting on page 7, you will see a logistic regression model - it performed fairly well predicting default (c=.63). The main take away from this output is the odds ratio estimates on page 8 which are a multivariate form of the results above, including the 95% confidence intervals - which are statistically significant. :) Talk to you tomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks Hilary. All of that sounds fine. I completely agree with getting the analysis correct first. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, December 30, 2013 10:57:34 AM Subject: RE: Checking In 8:30 is fine. I agree on getting the numbers right and am not in a great rush to start writing. When we do, it will be fine to paraphrase liberally from Sandler’s paper and to use some of her additional work that’s not in the paper. I do want to focus on the analytics and make sure they are bulletproof conceptually. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, December 30, 2013 10:54 AM To: Hilary B. Miller Subject: Re: Checking In Excellent. Lets plan on 8:30? I will call you. My goal is to "finalize" at least conceptually what needs to be done analytically so that I can begin the writing process. One question that I will have - to what extent can we utilize/leverage Danielle's writing and provide her with second authorship. Or, should I start the writing from scratch? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, December 30, 2013 10:41:52 AM Subject: RE: Checking In Yes, back from vacation. Best time for me is tomorrow morning – anytime, you pick. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, December 30, 2013 10:41 AM To: Hilary B. Miller Subject: Checking In Hi Hilary - let me know if you can catch up this week. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Excellent. Lets plan on 8:30? I will call you. My goal is to "finalize" at least conceptually what needs to be done analytically so that I can begin the writing process. One question that I will have - to what extent can we utilize/leverage Danielle's writing and provide her with second authorship. Or, should I start the writing from scratch? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, December 30, 2013 10:41:52 AM Subject: RE: Checking In Yes, back from vacation. Best time for me is tomorrow morning – anytime, you pick. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, December 30, 2013 10:41 AM To: Hilary B. Miller Subject: Checking In Hi Hilary - let me know if you can catch up this week. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - let me know if you can catch up this week. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I trust you are winding down for a few days of Christmas rest. I am sending to you the "results narrative" that I had promised for today. I have taken what I believe are the most salient and relevant outputs from our iterative discussions and placed them into a "story line". I will be offline from today until the 26th. I could talk on the phone on Friday if you would like - just let me know. Mornings are always better for me. :) Best wishes for a very Merry Christmas. Jen Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I want to make sure that you received this for our call. I am really pleased with the results. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Monday, December 16, 2013 1:00:02 PM Subject: Output for Call Here is the output for our conversation tomorrow. A GEE is a Generalized Estimating Equation (basically an OLS that controls for specified variables and then reports the marginal effects). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I want to make sure that you received this for our call. I am really pleased with the results. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Monday, December 16, 2013 1:00:02 PM Subject: Output for Call Here is the output for our conversation tomorrow. A GEE is a Generalized Estimating Equation (basically an OLS that controls for specified variables and then reports the marginal effects). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? I have updated results. Would you be available for a call tomorrow or Wed AM? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, December 11, 2013 1:32:43 PM Subject: RE: Analysis Sure. From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, December 11, 2013 12:57 PM To: Hilary B. Miller Subject: Analysis Hilary I think I have an analytical solution to controlling for Pre-Borrowing Score - it involves Autoregressive Time Series Analysis. I am not an expert in this but I have a colleague who is. I would like to have permission to have him help me with some of the math/programming for this technique. His name is Herman (Gene) Ray - he is an assistant professor of Statistics here at the University. I plan to pull a sample of the data and have him help me build the program to do the analysis... I will then "operationalize" the code to the larger dataset (he wont be accessing the full file). Let me know if this is ok with you. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Thursday, December 5, 2013 10:25:46 AM Subject: FW Hilary I think I have an analytical solution to controlling for PreBorrowing Score - it involves Autoregressive Time Series Analysis. I am not an expert in this - but I have a colleague who is. I would like to have permission to have him help me with some of the math/programming for this technique. His name is Herman (Gene) Ray - he is an assistant professor of Statistics here at the University. I plan to pull a sample of the data and have him help me build the program to do the analysis... I will then "operationalize" the code to the larger dataset (he wont be accessing the full file). Let me know if this is ok with you. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - here is a resend of the analysis from last night - now using CHANGE07to06PCT (or 09 to 08) as the primary dependent variable. Note that PCT was created by taking the change from 07 to 06 and dividing this by the 06 number (same logic for 09 to 08). I think the most important piece of information for both is the Table of CHANGE07to06PCTBIN by STATE. The same general patterns hold - but some results are stronger. Take a look. I will send you an invite for tomorrow for 11. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Monday, December 9, 2013 9:04:05 PM Subject: RE: Latest Results Barring unforeseen travel issues, I should be back at my desk at 11:00. Do you want to  talk then? Do you see the general pattern I do where the Ss with the worst outcomes are those  with the highest starting scores? ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Monday December 9, 2013, 08:47 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Latest Results Ok.  Let's catch up Wednesday when you get back.  Just let me know a good time. Sent from my iPhone > On Dec 9, 2013, at 8:29 PM, "Hilary B. Miller"  wrote: >  > I'll be traveling Wednesday morning ­­ back in the office around 10:30. I'm still digesting this. Seeing this information suggests to me, again, that we need to find a way to control for the subjects' pre­test Vantage  score. >  > ­­­­­Original Message­­­­­ > From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu]  > Sent: Monday, December 09, 2013 8:15 PM > To: Hilary B. Miller > Subject: Latest Results >  > Hi Hilary ­  >  > I have attached two sets of results for you.  The first set includes the Pre­Post for 2006­2007 and the  second includes the Pre­Post results for 08­09.  There are some interesting results. >  > Here is a brief summarization of my observations: >  > From the 2006­2007 file ­  >  > 1. The VATNAGE score decreased about 6 points overall. > 2. However, the change is very different by state ­ while all states experienced a decrease, CA  experienced the greatest decrease (8.41 points) and TX experienced the least decrease (3.97 points). > 3. I categorized the array of differences by BIN: <­100, ­100 to ­50, ­49 to ­25, ­25 to 0 and then >0. > Using this framework, MO has the smallest proportion of customers with a positive difference (BIN5) at  42%...TX has the largest proportion of customers with a positive difference (BIN5) at 46%. > 4. While not a steadily linear pattern, the total rollovers and the percent of loans rolled over per  customer per year increases from BIN1 to BIN5. >  > From the 2008 ­ 2009 file ­  >  > 1. The VANTAGE score changes very little (effectively 0) overall ­ but demonstrates greater variation by state. > 2. UT had the largest pt drop (3.22) while KS had the largest improvement (3.44). > 3. Using the same BIN structure as above, KS had the largest percent of customer in BIN5 (56.66%)  and UT had the smallest (50.82). > 4. As above, the total rollovers rollovers and the percent of loans rolled over per customer per year  increases from BIN1 to BIN5. >  > I dont have much availability on Tuesday, but I could be available on Wednesday for a call to catch up.   Just let me know what works best for your schedule.  :) >  >  >  >  >  >  >  > Jennifer Lewis Priestley, MBA, Ph.D.  > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services  >  > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would  dagny do?  >  >  >  >  >  > ­­­­­ Original Message ­­­­­ > From: "Hilary B. Miller"  > To: "Jennifer Lewis Priestley"  > Sent: Monday, December 9, 2013 12:41:55 PM > Subject: RE: Call today at 10?  Does that work? >  > Doing this by percentage change makes the most sense to me. I can also see a rationale for bins for  ranges of absolute changes in credit score (i.e., 25 points, 50 points, etc., which roughly corresponds to  5%, 10%, etc.).  >  > ­­­­­Original Message­­­­­ > From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Monday, December 09, 2013 12:37 PM > To: Hilary B. Miller > Subject: Re: Call today at 10? Does that work? >  > Hilary ­  >  > Question ­ I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as  we discussed.  Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest  20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? >  >  >  >  >  >  >  > Jennifer Lewis Priestley, MBA, Ph.D.  > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services  >  > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would  dagny do?  >  >  >  >    >  > ­­­­­ Original Message ­­­­­ > From: "Hilary B. Miller"  > To: "Jennifer Lewis Priestley"  > Sent: Friday, December 6, 2013 4:06:46 PM > Subject: RE: Call today at 10?  Does that work? >  > Jennifer ­­ >  > On further reflection, I wonder if we should use a specification that controls for pre­treatment Vantage  score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer  outcomes ­­ assuming that there might be one ­­ that is attributable to the consumer's being in "bad  shape" to begin with. Second, this issue emerges in some literature in which there is speculation that  worse­off consumers don't benefit from payday loans; we could test this, too. So having the pre­borrowing Vantage score on the right and the post­borrowing "delta" on the left seems to make some sense. >  > What do you think of this? >  > Hilary >  > ­­­­­Original Message­­­­­ > From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Friday, December 06, 2013 8:21 AM > To: Hilary B. Miller > Subject: Call today at 10? Does that work? >  >  >  > Sent from my iPhone Yep. I will send you a revised doc tomorrow PM Sent from my iPhone On Dec 9, 2013, at 9:04 PM, "Hilary B. Miller" wrote: Barring unforeseen travel issues, I should be back at my desk at 11:00. Do you want to talk then? Do you see the general pattern I do where the Ss with the worst outcomes are those with the highest starting scores? -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Monday December 9, 2013, 08:47 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Latest Results Ok. Let's catch up Wednesday when you get back. Just let me know a good time. Sent from my iPhone > On Dec 9, 2013, at 8:29 PM, "Hilary B. Miller" wrote: > > I'll be traveling Wednesday morning -- back in the office around 10:30. I'm still digesting this. Seeing this information suggests to me, again, that we need to find a way to control for the subjects' pre-test Vantage score. > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Monday, December 09, 2013 8:15 PM > To: Hilary B. Miller > Subject: Latest Results > > Hi Hilary > > I have attached two sets of results for you. The first set includes the Pre-Post for 2006-2007 and the second includes the Pre-Post results for 08-09. There are some interesting results. > > Here is a brief summarization of my observations: > > From the 2006-2007 file > > 1. The VATNAGE score decreased about 6 points overall. > 2. However, the change is very different by state - while all states experienced a decrease, CA experienced the greatest decrease (8.41 points) and TX experienced the least decrease (3.97 points). > 3. I categorized the array of differences by BIN: <-100, -100 to -50, -49 to -25, -25 to 0 and then >0. > Using this framework, MO has the smallest proportion of customers with a positive difference (BIN5) at 42%...TX has the largest proportion of customers with a positive difference (BIN5) at 46%. > 4. While not a steadily linear pattern, the total rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. > > From the 2008 - 2009 file > > 1. The VANTAGE score changes very little (effectively 0) overall - but demonstrates greater variation by state. > 2. UT had the largest pt drop (3.22) while KS had the largest improvement (3.44). > 3. Using the same BIN structure as above, KS had the largest percent of customer in BIN5 (56.66%) and UT had the smallest (50.82). > 4. As above, the total rollovers rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. > > I dont have much availability on Tuesday, but I could be available on Wednesday for a call to catch up. Just let me know what works best for your schedule. :) > > > > > > > > Jennifer Lewis Priestley, MBA, Ph.D. > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services > > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? > > > > > > ----- Original Message ----> From: "Hilary B. Miller" > To: "Jennifer Lewis Priestley" > Sent: Monday, December 9, 2013 12:41:55 PM > Subject: RE: Call today at 10? Does that work? > > Doing this by percentage change makes the most sense to me. I can also see a rationale for bins for ranges of absolute changes in credit score (i.e., 25 points, 50 points, etc., which roughly corresponds to 5%, 10%, etc.). > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Monday, December 09, 2013 12:37 PM > To: Hilary B. Miller > Subject: Re: Call today at 10? Does that work? > > Hilary > > Question - I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as we discussed. Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest 20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? > > > > > > > > Jennifer Lewis Priestley, MBA, Ph.D. > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services > > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? > > > > > > ----- Original Message ----> From: "Hilary B. Miller" > To: "Jennifer Lewis Priestley" > Sent: Friday, December 6, 2013 4:06:46 PM > Subject: RE: Call today at 10? Does that work? > > Jennifer -> > On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -- assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. > > What do you think of this? > > Hilary > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Friday, December 06, 2013 8:21 AM > To: Hilary B. Miller > Subject: Call today at 10? Does that work? > > > > Sent from my iPhone Ok. Let's catch up Wednesday when you get back. let me know a good time. Just Sent from my iPhone > On Dec 9, 2013, at 8:29 PM, "Hilary B. Miller" wrote: > > I'll be traveling Wednesday morning -- back in the office around 10:30. I'm still digesting this. Seeing this information suggests to me, again, that we need to find a way to control for the subjects' pre-test Vantage score. > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Monday, December 09, 2013 8:15 PM > To: Hilary B. Miller > Subject: Latest Results > > Hi Hilary > > I have attached two sets of results for you. The first set includes the Pre-Post for 2006-2007 and the second includes the Pre-Post results for 08-09. There are some interesting results. > > Here is a brief summarization of my observations: > > From the 2006-2007 file > > 1. The VATNAGE score decreased about 6 points overall. > 2. However, the change is very different by state while all states experienced a decrease, CA experienced the greatest decrease (8.41 points) and TX experienced the least decrease (3.97 points). > 3. I categorized the array of differences by BIN: <100, -100 to -50, -49 to -25, -25 to 0 and then >0. > Using this framework, MO has the smallest proportion of customers with a positive difference (BIN5) at 42%...TX has the largest proportion of customers with a positive difference (BIN5) at 46%. > 4. While not a steadily linear pattern, the total rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. > > From the 2008 - 2009 file > > 1. The VANTAGE score changes very little (effectively 0) overall - but demonstrates greater variation by state. > 2. UT had the largest pt drop (3.22) while KS had the largest improvement (3.44). > 3. Using the same BIN structure as above, KS had the largest percent of customer in BIN5 (56.66%) and UT had the smallest (50.82). > 4. As above, the total rollovers rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. > > I dont have much availability on Tuesday, but I could be available on Wednesday for a call to catch up. Just let me know what works best for your schedule. :) > > > > > > > > Jennifer Lewis Priestley, MBA, Ph.D. > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services > > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? > > > > > > ----- Original Message ----> From: "Hilary B. Miller" > To: "Jennifer Lewis Priestley" > Sent: Monday, December 9, 2013 12:41:55 PM > Subject: RE: Call today at 10? Does that work? > > Doing this by percentage change makes the most sense to me. I can also see a rationale for bins for ranges of absolute changes in credit score (i.e., 25 points, 50 points, etc., which roughly corresponds to 5%, 10%, etc.). > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Monday, December 09, 2013 12:37 PM > To: Hilary B. Miller > Subject: Re: Call today at 10? Does that work? > > Hilary > > Question - I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as we discussed. Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest 20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? > > > > > > > > Jennifer Lewis Priestley, MBA, Ph.D. > Associate Professor of Statistics > Director, Center for Statistics and Analytical Services > > faculty page: http://www.science.kennesaw.edu/~jpriestl/ > department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? > > > > > > ----- Original Message ----> From: "Hilary B. Miller" > To: "Jennifer Lewis Priestley" > Sent: Friday, December 6, 2013 4:06:46 PM > Subject: RE: Call today at 10? Does that work? > > Jennifer -> > On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. > > What do you think of this? > > Hilary > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Friday, December 06, 2013 8:21 AM > To: Hilary B. Miller > Subject: Call today at 10? Does that work? > > > > Sent from my iPhone Hi Hilary I have attached two sets of results for you. The first set includes the Pre-Post for 2006-2007 and the second includes the Pre-Post results for 08-09. There are some interesting results. Here is a brief summarization of my observations: From the 2006-2007 file 1. The VATNAGE score decreased about 6 points overall. 2. However, the change is very different by state - while all states experienced a decrease, CA experienced the greatest decrease (8.41 points) and TX experienced the least decrease (3.97 points). 3. I categorized the array of differences by BIN: <-100, -100 to -50, -49 to -25, -25 to 0 and then >0. Using this framework, MO has the smallest proportion of customers with a positive difference (BIN5) at 42%...TX has the largest proportion of customers with a positive difference (BIN5) at 46%. 4. While not a steadily linear pattern, the total rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. From the 2008 - 2009 file 1. The VANTAGE score changes very little (effectively 0) overall - but demonstrates greater variation by state. 2. UT had the largest pt drop (3.22) while KS had the largest improvement (3.44). 3. Using the same BIN structure as above, KS had the largest percent of customer in BIN5 (56.66%) and UT had the smallest (50.82). 4. As above, the total rollovers rollovers and the percent of loans rolled over per customer per year increases from BIN1 to BIN5. I dont have much availability on Tuesday, but I could be available on Wednesday for a call to catch up. Just let me know what works best for your schedule. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Monday, December 9, 2013 12:41:55 PM Subject: RE: Call today at 10? Does that work? Doing this by percentage change makes the most sense to me. I can also see a rationale for bins for ranges of absolute changes in credit score (i.e., 25 points, 50 points, etc., which roughly corresponds to 5%, 10%, etc.). -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, December 09, 2013 12:37 PM To: Hilary B. Miller Subject: Re: Call today at 10? Does that work? Hilary Question - I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as we discussed. Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest 20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Friday, December 6, 2013 4:06:46 PM Subject: RE: Call today at 10? Does that work? Jennifer -On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. What do you think of this? Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, December 06, 2013 8:21 AM To: Hilary B. Miller Subject: Call today at 10? Does that work? Sent from my iPhone Will do. And, you are correct. 1std is about 50 points...which is also 10% from the baseline. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Monday, December 9, 2013 12:41:55 PM Subject: RE: Call today at 10? Does that work? Doing this by percentage change makes the most sense to me. I can also see a rationale for bins for ranges of absolute changes in credit score (i.e., 25 points, 50 points, etc., which roughly corresponds to 5%, 10%, etc.). -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Monday, December 09, 2013 12:37 PM To: Hilary B. Miller Subject: Re: Call today at 10? Does that work? Hilary Question - I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as we discussed. Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest 20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Friday, December 6, 2013 4:06:46 PM Subject: RE: Call today at 10? Does that work? Jennifer -On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. What do you think of this? Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, December 06, 2013 8:21 AM To: Hilary B. Miller Subject: Call today at 10? Does that work? Sent from my iPhone Hilary Question - I am "bucketing" the change in Vantage score to group those with an "adverse" outcome as we discussed. Would you rather see the bucketing logic based on percentiles (e.g., lowest 10%, lowest 20%, etc) or standard deviations (.5 std below the mean, 1 std below the mean, etc)? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Friday, December 6, 2013 4:06:46 PM Subject: RE: Call today at 10? Does that work? Jennifer -On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. What do you think of this? Hilary -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, December 06, 2013 8:21 AM To: Hilary B. Miller Subject: Call today at 10? Does that work? Sent from my iPhone Mathematically they would be highly correlated...since the delta is a function of the pre...but I will send you another round of analysis next week and include it. Sent from my iPhone > On Dec 6, 2013, at 4:06 PM, "Hilary B. Miller" wrote: > > Jennifer -> > On further reflection, I wonder if we should use a specification that controls for pre-treatment Vantage score. To my mind, this would serve two purposes: first, it would tease out the changes in consumer outcomes -assuming that there might be one -- that is attributable to the consumer's being in "bad shape" to begin with. Second, this issue emerges in some literature in which there is speculation that worse-off consumers don't benefit from payday loans; we could test this, too. So having the pre-borrowing Vantage score on the right and the post-borrowing "delta" on the left seems to make some sense. > > What do you think of this? > > Hilary > > -----Original Message----> From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] > Sent: Friday, December 06, 2013 8:21 AM > To: Hilary B. Miller > Subject: Call today at 10? Does that work? > > Sent from my iPhone Hi Hilary - take a look at the attached - this is just a quick check to ensure that I am on the righ track. Score06 is the 2006 Vantagescore. Score07 is the 2007 Vantagescore and Change06to07 is the Change from 2006 to 2007. You will see a complete table of descriptive stats by state and then an ANOVA - the bottom of the ANOVA output will provide the direct comparison of each state relative to TX. Those comparisons with *** are significant - those without are not. I will be available again tomorrow at 10 - let me know if that works for you. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Attached is a very small sample of the data that we can pull up in Excel for discussiontomorrow. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, December 4, 2013 7:58:46 PM Subject: RE: Checking In Okay — I see.  This is a useful first step and considerably more detail about the components of credit  scores than I was expecting. So, this is really “Part 1,” which is scores as the LHS variable and regulation on the  right. “Part 2” is scores on the left and actual rollover behavior on the right. When we speak tomorrow, I’d like to convince you that we need to explore changes in  credit scores for individual borrowers before and after borrowing — not within­state  changes in scores over time — as the dependent variable. If you have, in fact, done  this, it’s not clear, but I don’t think you have. The purpose of this method of data  collection was to set up before­and­after snapshots for individual borrowers. If you have, as I suspect, just tracked scores over time vs. regulatory scheme, then I see an issue  with possibly confounding variables between states, including the fact that some were  more deeply harmed by the recession. It’s really important to get the concept right  before we start the analysis, and for now I’d rather that we focus on the specifications  rather than trying to write the paper. I don’t want to waste your time but I think we have  to agree on the foundation. Please give this some thought and we’ll speak at 10:00. Thanks. Hilary ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday December 4, 2013, 05:32 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Checking In Hi Hilary I trust all is well. Attached is the discussion document for tomorrow - please view this as a work in progress to communicate the initial analytical findings. The primary emphasis will be on the results section and on the tables/figures. I look forward to speaking with you at 10 tomorrow. I will call you from my office. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, December 1, 2013 11:55:04 AM Subject: RE: Checking In That’s a good time slot as things now stand. Please send me an Outlook appointment  for your preferred time, and I’ll lock it down. Thanks. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday December 1, 2013, 11:48 AM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Checking In Hi Hilary - I have some interesting interim results that I would like to share with you - can we set up some time to review an interim (work-in-progress) document? Thursday works particularly well for me between 9 and 12 - let me know if that time works well for you. I will get the doc to you in advance of the call. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thats great feedback. I did start a pre-post analysis, but changed direction...so that is easy enough to fold back in. And, you don't have to "convince me" - I am here to serve. I just want to make sure that what I am doing analytically is reflecting your thinking. Lets use this document as a basis for the conversation and determine what needs to be added/changed. Talk to you at 10. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, December 4, 2013 7:58:46 PM Subject: RE: Checking In Okay — I see.  This is a useful first step and considerably more detail about the components of credit  scores than I was expecting. So, this is really “Part 1,” which is scores as the LHS variable and regulation on the  right. “Part 2” is scores on the left and actual rollover behavior on the right. When we speak tomorrow, I’d like to convince you that we need to explore changes in  credit scores for individual borrowers before and after borrowing — not within­state  changes in scores over time — as the dependent variable. If you have, in fact, done  this, it’s not clear, but I don’t think you have. The purpose of this method of data  collection was to set up before­and­after snapshots for individual borrowers. If you have, as I suspect, just tracked scores over time vs. regulatory scheme, then I see an issue  with possibly confounding variables between states, including the fact that some were  more deeply harmed by the recession. It’s really important to get the concept right  before we start the analysis, and for now I’d rather that we focus on the specifications  rather than trying to write the paper. I don’t want to waste your time but I think we have  to agree on the foundation. Please give this some thought and we’ll speak at 10:00. Thanks. Hilary ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Wednesday December 4, 2013, 05:32 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Checking In Hi Hilary I trust all is well. Attached is the discussion document for tomorrow - please view this as a work in progress to communicate the initial analytical findings. The primary emphasis will be on the results section and on the tables/figures. I look forward to speaking with you at 10 tomorrow. I will call you from my office. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, December 1, 2013 11:55:04 AM Subject: RE: Checking In That’s a good time slot as things now stand. Please send me an Outlook appointment  for your preferred time, and I’ll lock it down. Thanks. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday December 1, 2013, 11:48 AM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Checking In Hi Hilary - I have some interesting interim results that I would like to share with you - can we set up some time to review an interim (work-in-progress) document? Thursday works particularly well for me between 9 and 12 - let me know if that time works well for you. I will get the doc to you in advance of the call. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I trust all is well. Attached is the discussion document for tomorrow - please view this as a work in progress to communicate the initial analytical findings. The primary emphasis will be on the results section and on the tables/figures. I look forward to speaking with you at 10 tomorrow. I will call you from my office. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, December 1, 2013 11:55:04 AM Subject: RE: Checking In That’s a good time slot as things now stand. Please send me an Outlook appointment  for your preferred time, and I’ll lock it down. Thanks. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday December 1, 2013, 11:48 AM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Checking In Hi Hilary - I have some interesting interim results that I would like to share with you - can we set up some time to review an interim (work-in-progress) document? Thursday works particularly well for me between 9 and 12 - let me know if that time works well for you. I will get the doc to you in advance of the call. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - I have some interesting interim results that I would like to share with you - can we set up some time to review an interim (work-in-progress) document? Thursday works particularly well for me between 9 and 12 - let me know if that time works well for you. I will get the doc to you in advance of the call. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - can we move our call to tomorrow? I am running a training session today until 5:30 for The Southern Company. I can call you tomorrow anytime after 4:30. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Priestley"  Sent: Saturday, November 16, 2013 10:07:03 AM Subject: Re: Checking In Okay. Might be early next week. _______________________________ On Nov 16, 2013, at 11:06 AM, "Jennifer Priestley"  wrote: I meant to say...yes, please go ahead and send it... Sent from my iPhone On Nov 16, 2013, at 10:05 AM, "Hilary B. Miller"  wrote: Also, when you have a moment, please see if you have the TransUnion "codebook." If  not, I'll get it for you. On Nov 16, 2013, at 11:02 AM, "Jennifer Lewis Priestley"   wrote: Will do. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page:http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week except Thursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgivingto give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page:http://www.science.kennesaw.edu/~jpriestl/ department page:http://math.kennesaw.edu/ center page:http://www.kennesaw.edu/csas/ what would dagny do? I meant to say...yes, please go ahead and send it... Sent from my iPhone On Nov 16, 2013, at 10:05 AM, "Hilary B. Miller" wrote: Also, when you have a moment, please see if you have the TransUnion "codebook." If not, I'll get it for you. On Nov 16, 2013, at 11:02 AM, "Jennifer Lewis Priestley" wrote: Will do. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week except Thursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page:http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Ok. Sent from my iPhone On Nov 16, 2013, at 10:05 AM, "Hilary B. Miller" wrote: Also, when you have a moment, please see if you have the TransUnion "codebook." If not, I'll get it for you. On Nov 16, 2013, at 11:02 AM, "Jennifer Lewis Priestley" wrote: Will do. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week except Thursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page:http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? The following meeting has been modified: Subject: Call with JLP and HM Organizer: "Jennifer Priestley" Location: JLP to Call Monday, November 18, 2013, 4:30:00 PM - 5:00:00 PM GMT -05:00 US/Canada Time: Eastern [MODIFIED] Invitees: hilary@miller.net *~*~*~*~*~*~*~*~*~* Will do. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 10:01:51 AM Subject: Re: Checking In Please give me a call as close to 5 on Monday as you can. I'll be at my desk.   On Nov 16, 2013, at 10:53 AM, "Jennifer Lewis Priestley"   wrote: OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week exceptThursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week after Thanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Ah, I had this saved as "previous Ph.D. paper" - i see now that its the same paper. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley Ph. D." Sent: Saturday, November 16, 2013 9:57:50 AM Subject: Payday_Rollovers_82813.pdf Thought I had previously sent this. HM OK - I could speak after 5 on M,T,W...just let me know... And, I dont think I have the Sandler draft - could I trouble you for a copy? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Saturday, November 16, 2013 9:47:18 AM Subject: Re: Checking In We should probably talk about some definitional issues before you get too deeply into  even the univariate analysis ­­ such as why the four snapshots were chosen, what's a  "rollover" in economic (rather than literal) terms, etc. it would be worth rereading  Sandler's draft before this conversation. I'm around all next week except Thursday. On Nov 16, 2013, at 10:40 AM, "Jennifer Lewis Priestley"   wrote: Hi Hilary - I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week afterThanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I did not want my "radio silence" to be troubling...I wanted to let you know that I have been working on your project...and have formulated an analysis plan. I have your data loaded into SAS and should have the univariate analysis complete next week and the interim multivariate analysis complete shortly after. I would like to set up some time with you to speak on the phone the week afterThanksgiving to give you an update, and ensure that I am on the right track with the analytical plan. Generally, I am only 100% offline when I am in the classroom between 9 and 12 on M/W. Other than that, I can generally make myself available. It might be useful to have a data person on the phone with us - because I want to validate my computational/programming approach to the calculation of the rollover rates (alternatively, I can just send that to you in advance and you can confirm it prior to the call). Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Got it! :) Sent from my iPhone On Nov 10, 2013, at 10:49 PM, "Hilary B. Miller" wrote: Yes. The original lender files contained all of the personal identifying information — these fields, as shown in the spreadsheet: CUSTOMERNUM LOANNUM FIRSTNAME LASTNAME SSN Those files were sent to TransUnion to be matched with credit bureau data. We do not have a “permissible use” under the Fair Credit Reporting Act for the matched data, so TransUnion deleted all of the personal identifying information and replaced it with a random, unique customer and loan identifyer. These are the INPUT_SEQNUM Key\Flag fields from the spreadsheet. That way, we can work with the data, but we can’t identify the borrowers. Make sense? H -----Original Message----From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:44 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call Never mind - I figured it out - there is a field called "KEYFLAG" that appears to be a proxy for customer ID. OK I feel better. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Sunday, November 10, 2013 10:40:24 PM Subject: Re: Data call OK - then I have one more Q - I went into the original STATA files and the loan number and customer number values are set to missing in STATA (consistent with what I see in SAS). Does that sound right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:36:54 PM Subject: RE: Data call No bother at all. That’s what I’m in business to do. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:36 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call OK. Thanks. I will stop bothering you tonight. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:33:06 PM Subject: RE: Data call That sounds right. That’s how we selected the subjects — a number from each of those  states from each lender. Essentially, Texas is the “control” state (unregulated with  respect to rollovers), and each of the other states has some different treatment. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:30 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call Are those the only available states? CA FL KS MO OK TX UT Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Never mind - I figured it out - there is a field called "KEYFLAG" that appears to be a proxy for customer ID. OK I feel better. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Jennifer Lewis Priestley"  To: "Hilary B. Miller"  Sent: Sunday, November 10, 2013 10:40:24 PM Subject: Re: Data call OK - then I have one more Q - I went into the original STATA files and the loan number and customer number values are set to missing in STATA (consistent with what I see in SAS). Does that sound right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:36:54 PM Subject: RE: Data call No bother at all. That’s what I’m in business to do. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:36 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call OK. Thanks. I will stop bothering you tonight. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:33:06 PM Subject: RE: Data call That sounds right. That’s how we selected the subjects — a number from each of those  states from each lender. Essentially, Texas is the “control” state (unregulated with  respect to rollovers), and each of the other states has some different treatment. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:30 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call Are those the only available states? CA FL KS MO OK TX UT Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? OK - then I have one more Q - I went into the original STATA files and the loan number and customer number values are set to missing in STATA (consistent with what I see in SAS). Does that sound right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:36:54 PM Subject: RE: Data call No bother at all. That’s what I’m in business to do. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:36 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call OK. Thanks. I will stop bothering you tonight. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:33:06 PM Subject: RE: Data call That sounds right. That’s how we selected the subjects — a number from each of those  states from each lender. Essentially, Texas is the “control” state (unregulated with  respect to rollovers), and each of the other states has some different treatment. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:30 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call Are those the only available states? CA FL KS MO OK TX UT Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are - the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? OK. Thanks. I will stop bothering you tonight. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:33:06 PM Subject: RE: Data call That sounds right. That’s how we selected the subjects — a number from each of those  states from each lender. Essentially, Texas is the “control” state (unregulated with  respect to rollovers), and each of the other states has some different treatment. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:30 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call Are those the only available states? CA FL KS MO OK TX UT Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are - the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Are those the only available states? CA FL KS MO OK TX UT Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? OK - thanks. I will make another run at it. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:28:36 PM Subject: RE: Data call That’s my understanding. ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:27 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Data call So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are - the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? So, to clarify, the Jan2006, Jan2007, Jan2008 and Jan2009 files should all have layouts consistent with the attached - right? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Ok - this insight helps. Thanks. I will run the STATA programs again and see where the import error occurred. Would it be possible to have your guys drop it into the shared drive as SAS files? That would eliminate the interim step of file conversion. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Sunday, November 10, 2013 10:20:14 PM Subject: RE: Data call That doesn’t sound right. All of the raw data files should be in the same format. You  have the Excel spreadsheet with the file layout. Each record should have an  anonymized (but consistent) TransUnion­assigned customer number and related  transaction number so that a customer with multiple loans can be tracked. Is that what  the merged file looks like, too? In view of Danielle’s paper, this really makes no sense. I  think the “rollover” file is output, not raw data.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Sunday November 10, 2013, 10:14 PM To: Hilary B. Miller [hilary@miller.net] Subject: Data call Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary - we have "cracked open" the data. I have a few questions about it that I was hoping to clarify with you. Can we speak on Tuesday morning? The main questions are the files for Jan2006 and Jan2008 are similar (in terms of variables), but the customer number and loan number fields are not populated. The files for Jan2007 and Jan2009 are similar (but different from the first two) and do not appear to have a field for customer/loan number. As a result, these files cannot be merged - or used for longitudinal analysis. The Payday_rollover file is similar to 2007 and 2009 files, but again, no customer/loan identifier is present...and only CA FL KS MO OK TX UT are present. Is all of this consistent with your understanding? Ideally - to evaluate the effect of payday regulations on economic welfare (credit score), I would like to see data for all states - and then i could create a field to flag those states with and without payday regulations. In addition, some kind of identifier per customer and/or per loan is needed (if possible). I have read all of the papers, and have created an analytical plan - but I really need all states and loan/customer identifiers to do what needs to be done. If they are just not available, we can work with that - but it would be better (the results would have stronger validity) if they could be included. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Friday, November 8, 2013 12:36:55 PM Subject: RE: Files I don’t have them on my hard drive but will go hunting.  ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Friday November 8, 2013, 12:35 PM To: Hilary B. Miller [hilary@miller.net] Subject: Files Hi Hilary - I think we are missing four STATA program files - I was hoping you could send these to me directly: TransUnion2006.dct TransUnion2007.dct TransUnion2008.dct TransUnion2009.dct Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I think we are missing four STATA program files - I was hoping you could send these to me directly: TransUnion2006.dct TransUnion2007.dct TransUnion2008.dct TransUnion2009.dct Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Thanks. We are working on getting the data converted right now from STATA to SAS. :) Sent from my iPhone On Nov 7, 2013, at 5:38 PM, "Hilary B. Miller" wrote: Here is another, brand-new paper for you to take a look at. I just received it and am reading it this evening. From: Gregory Elliehausen [mailto:gregory.elliehausen@frb.gov] Sent: Thursday, November 07, 2013 1:41 PM To: Hilary B. Miller Subject: RE: The Paper Thank you for your comments and suggestions. I have attached another paper that may be of interest to you. From: Hilary B. Miller [mailto:hilary@miller.net] Sent: Thursday, November 07, 2013 09:14 To: Thomas Durkin Cc: Gregory Elliehausen Subject: RE: The Paper Excellent work, gents. HM From: Thomas Durkin [mailto:tdurkin33@yahoo.com] Sent: Thursday, November 07, 2013 8:25 AM To: Hilary B. Miller Cc: Gregory.Elliehausen@frb.gov Subject: The Paper Good morning, I fussed with the paper and fixed a couple of typos (run ons with parentheses, extra article, etc.). I added  one footnote and removed a duplicate reference but otherwise did not do anything with the paper. It looks like it is in pretty good shape. Thanks for your efforts with the editing, and for your patience. Greg is going to look into SSRN, and we will circulate it to various colleagues. We also will look into  publication outlets. Best, Tom   <2013-10-09 Effects of payday bans.docx> Just let me know when to expect the full dataset. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, November 6, 2013 4:25:52 PM Subject: NDA Re Consumer Credit Research Foundation ­ Revised (between Law  offices of Hilary B. Miller and Jennifer Lewis Priestley) is Signed and Filed! NDA Re Consumer Credit Research Foundation Revised (between Law offices of Hilary B. Miller and Jennifer Lewis Priestley) is Signed and Filed! From: Hilary Miller (Law offices of Hilary B. Miller) To: Jennifer Lewis Priestley Attached is a final copy of NDA Re Consumer Credit Research Foundation - Revised. Copies have been automatically sent to all parties to the agreement. You can view a copy in your EchoSign account. Why use EchoSign:  Exchange, Sign, and File Any Document. In Seconds!  Set-up Reminders. Instantly Share Copies with Others.  See All of Your Documents, Anytime, Anywhere. To ensure that you continue receiving our emails, please add echosign@echosign.com to your address book or safe list. Thanks. I will bring this to our legal dept attention and talk them back from the cliff. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Wednesday, November 6, 2013 12:25:49 PM Subject: RE: NDA Jennifer – I think you are covered by this under the separate Section 3(c). Section1 simply identifies what is “confidential.” Section 3 specifies the circumstances under which you may disclose. You are expressly permitted to comply with laws and subpoenas. Take a look at 3(c). Hilary From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Wednesday, November 06, 2013 12:20 PM To: Hilary B. Miller Subject: NDA Hi Hilary - I am writing in reference to the NDA - while I dont have any issues - I am a professor in a public university...which has one (I think small) implication for the NDA... The last sentences of section 1 are a problem, because I am subject to the open records act. This has never been an issue none of my work has ever been supoenaed - but there is a nonzero probability, so I need it taken out. Let me know if this is an issue - otherwise, its fine. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hi Hilary I am writing in reference to the NDA - while I dont have any issues - I am a professor in a public university...which has one (I think small) implication for the NDA... The last sentences of section 1 are a problem, because I am subject to the open records act. This has never been an issue none of my work has ever been supoenaed - but there is a nonzero probability, so I need it taken out. Let me know if this is an issue - otherwise, its fine. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? Hilary - this additional perspective is helpful - thank you. Regarding the consulting contract - should I identify you personally as the client (in CT)? Or the Association (in DC)? Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley"  Sent: Tuesday, November 5, 2013 8:43:09 AM Subject: RE: Payday File Layout Jennifer — I think this is pretty close as an overview. I don’t think you need to modify your  summary, but I want to give you some additional thoughts that should inform your  design and methodology: To understand the importance of this paper and what I visualize as its contribution to  science, it’s worth viewing Bhutta et al. (and nearly all of its predecessor science) as  looking at mean effects; that is, Bhutta determines that users as a whole aren’t any  better or worse off as a result of using payday loans. This makes perfect sense, as he  explains, because the loans are small and the borrowers were in pretty bad financial  shape to begin with. But we know, or at least suspect, that there is a distribution of  outcomes. Okay, we know it. For most borrowers, having a payday loan either makes  little difference or is welfare­enhancing. But there is that pesky left tail. Policymakers are appropriately focused on the left tail. They want to know: what do those people look like, and how did they get there? So, we’d like to take a deeper dive into how heavy users  differ from others, both in terms of their welfare outcomes, as well as whether it is  possible to identify loan applicants at the pre­loan stage who have a propensity to  “stuck” in the product for a long time.  (Interestingly, this latter discrimination function is not a traditional role of credit scoring,  but it is perceived by policymakers as a function of “underwriting.” We’ll talk more about  that issue at the paper­writing stage.) These are basically subsidiary questions that need to be addressed regarding issue #2,  not entirely separate ones. Please do go ahead and send me the contract. We’ll send you an NDA, which I will try  to get out later this morning. I have all the data here and can transmit it to you  immediately upon NDA execution. Hilary ­­­­­Original Message­­­­­ From: Jennifer Lewis Priestley [jpriestl@kennesaw.edu] Received: Monday November 4, 2013, 08:38 PM To: Hilary B. Miller [hilary@miller.net] Subject: Re: Payday File Layout Hi Hilary ­  I took a look at the file layout from Transunion ­ its fairly "typical" consumer credit data ­ we have several  of these from Equifax that we use in the classroom to teach credit scoring.   I have gone through my notes and our correspondence ­ I have summarized my understanding of the  project in the attached work plan.  Take a look and let me know what you think.  If you agree, then I will  get a consulting contract out to you tomorrow.  Let me know if you need for me (and Joe Dolan ­ my  graduate student assistant who will be working with me on this project) to sign an NDA.   I look forward to working with you.  :) Jennifer Lewis Priestley, MBA, Ph.D.  Associate Professor of Statistics  Director, Center for Statistics and Analytical Services  faculty page: http://www.science.kennesaw.edu/~jpriestl/  department page: http://math.kennesaw.edu/  center page: http://www.kennesaw.edu/csas/  what would dagny do?  ­­­­­ Original Message ­­­­­ From: "Hilary B. Miller"  To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)"  Sent: Sunday, November 3, 2013 3:07:09 PM Subject: Payday File Layout Jennifer ­­ Here is the file layout you requested. The raw files are fixed­field files which were imported into Stata.  Each discrete record represents a single loan transaction (there may be multiple records per individual).  The customer data are anonymized (TransUnion replaces SSNs with unique individual identifiers).  Hilary Hilary B. Miller 500 West Putnam Avenue ­ Suite 400 Greenwich, CT 06830­6096 (203) 399­1320 voice (203) 517­6859 cell (914) 206­3727 fax hilary@miller.net (sent from laptop)   Hi Hilary I took a look at the file layout from Transunion - its fairly "typical" consumer credit data - we have several of these from Equifax that we use in the classroom to teach credit scoring. I have gone through my notes and our correspondence - I have summarized my understanding of the project in the attached work plan. Take a look and let me know what you think. If you agree, then I will get a consulting contract out to youtomorrow. Let me know if you need for me (and Joe Dolan - my graduate student assistant who will be working with me on this project) to sign an NDA. I look forward to working with you. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley, Ph.D. (jpriestl@kennesaw.edu)" Sent: Sunday, November 3, 2013 3:07:09 PM Subject: Payday File Layout Jennifer -Here is the file layout you requested. The raw files are fixed-field files which were imported into Stata. Each discrete record represents a single loan transaction (there may be multiple records per individual). The customer data are anonymized (TransUnion replaces SSNs with unique individual identifiers). Hilary Hilary B. Miller 500 West Putnam Avenue - Suite 400 Greenwich, CT 06830-6096 (203) 399-1320 voice (203) 517-6859 cell (914) 206-3727 fax hilary@miller.net (sent from laptop) Great. Thanks. I will send a project plan tomorrow. Sent from my iPhone > On Nov 3, 2013, at 3:07 PM, "Hilary B. Miller" wrote: > > Jennifer -> > Here is the file layout you requested. The raw files are fixed-field files which were imported into Stata. Each discrete record represents a single loan transaction (there may be multiple records per individual). The customer data are anonymized (TransUnion replaces SSNs with unique individual identifiers). > > Hilary > > Hilary B. Miller > 500 West Putnam Avenue - Suite 400 > Greenwich, CT 06830-6096 > (203) 399-1320 voice > (203) 517-6859 cell > (914) 206-3727 fax > hilary@miller.net > (sent from laptop) > > > Hi Hilary I think I understand. Lets do plan to speak in more detail. Friday is good for me - I am not teaching that day. I am available from 10 - 1. Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: "Jennifer Lewis Priestley" Sent: Tuesday, October 29, 2013 4:52:37 PM Subject: RE: New Project(s) Jennifer -As we discussed, we are interested in answering some of the "$64 questions" about payday lending, namely whether: (a) variation in state rollover regulation affects borrower welfare outcomes, and (b) variation in rollover usage affects borrower welfare outcomes. To answer these questions, I have a large file of administrative data (38,000 borrower histories) from three storefront lenders that has been matched with 300 variables from TransUnion. These data have been in the hands of a junior investigator for several months, and I have pulled the plug on this project because it was not progressing satisfactorily and it was impossible to get even professionals (let alone policymakers) to make sense out of her choice of methodology. I would like to start from scratch on the analysis and produce two papers (or possibly one consolidated paper) of academic quality, peer-reviewable, that would respond to these issues. In my model, you would be the PI and would publish the paper under your name. We are here to help but want the paper to be yours. I can give you the work product of the now-terminated investigator and you can use it as a starting point. We are on a relatively short timetable and need to have a finished paper by the end of February, no fooling. I have a budget that can support a decent stipend and defray any expenses. If you want to look at some related work, you might look at Kaufman (2013),http://www.federalreserve.gov/pubs/feds/2013/20136 2/201362pap.pdf, and Bhutta et al. (2012),http://ssrn.com/abstract=2160947. We modeled the data collection for this project on Bhutta. Kaufman is hot off the presses. Please give this some thought, and then let's speak later in the week about your interest and what you would need to make this worth your while. I should be in my office each day. Regards, Hilary (contact info below) -----Original Message----From: Jennifer Lewis Priestley [mailto:jpriestl@kennesaw.edu] Sent: Friday, October 25, 2013 11:46 AM To: Hilary B. Miller Subject: Re: New Project(s) Hi Hilary - pleasure to speak with you. My cell number is 404-229-3216 My office number is 770-423-6107 Look forward to hearing from you next week. :) Jennifer Lewis Priestley, MBA, Ph.D. Associate Professor of Statistics Director, Center for Statistics and Analytical Services faculty page: http://www.science.kennesaw.edu/~jpriestl/ department page: http://math.kennesaw.edu/ center page: http://www.kennesaw.edu/csas/ what would dagny do? ----- Original Message ----From: "Hilary B. Miller" To: jpriestl@kennesaw.edu Sent: Friday, October 25, 2013 11:38:26 AM Subject: New Project(s) Dear Prof. Priestley: I understand that you have been working with Michael Flores on an online-lending project. If these kinds of projects interest you, we have more projects, access to large administrative datasets, and a budget. Would you please give me a call? Hilary Miller -Hilary B. Miller Chairman of the Board Consumer Credit Research Foundation 500 West Putnam Avenue - Suite 400 Greenwich, Connecticut 06830-6096 (203) 399-1320 (voice) (203) 517-6859 (cell) (914) 206-3727 (fax) info@creditresearch.org