MEMO To: Commissioner Chloe Eudaly Cc: Jamey Duhamel, Policy Director From: Lisa K. Bates, Ph.D. Re: ‘Beta test’ of FAIR tenant screening policy Executive Summary The proposed FAIR tenant screening policy creates a standardized set of criteria for screening applicants for rental housing; landlords adopting the ‘fast track’ screening criteria adhere to a prescribed set of conditions under which rental housing can be denied. The ‘beta test’ analyzed how this policy would affect access to rental housing in terms of applicants’ ability to pass the standards applied by comparing the FAIR standards to three current practices (affordable housing; the industry standard operating procedure; and a strict market policy). Based on an assessment using the dataset of over 5,800 individuals who submitted information to the OneApp platform, the FAIR policy will significantly increase the number of renter applicants who are accepted into rental housing. The FAIR policy outcomes most closely resemble those of affordable housing provider screening. For market rate screening procedures, the shift to an acceptance affects from one-third to one-half of all renter applicants, depending on the comparison policy. The FAIR policy also substantially increases the acceptance rate for people of color; low-income applicants; renters with Housing Choice (‘Section 8’) Vouchers; and people with a history of criminal justice system contact. This memo describes the renter applicants and then provides the analysis of outcomes expected for their applications under the four comparison screening policies, including an overall acceptance rate and the change in access for individual applicants, and a more focused analysis for identified groups of concern. Analysis of tenant screening policies using the OneApp data Who are the renter applicants in the analysis? The provided OneApp database has some demographic information that describes the group of renters about which this analysis draws conclusions. This dataset is not a statistically representative sample of renter households in Portland and cannot be generalized as such; however the variety of OneApp users allows for the beta test and partnership with the City for data access provides information that would not otherwise be easily obtained. The OneApp renter applicant database analyzed includes 5,854 individuals. The following is a snapshot of demographic information:  50% are female identified (19% missing)  16% identify as people of color (48.8% missing)  Median age is 35   Median monthly income is $2,500; Mean monthly income is $3,1691 4.8% of applicants report they will use a Housing Choice (‘Section 8’) voucher There are individuals in the dataset with histories that present challenges in rental application screening due to restrictive policies. These include:  13.45% have an eviction history  9% have a history of criminal justice system contact (conviction)  19% have no rental history  25% lack a credit history  For those with a credit score, the median FICO score is 591 The analysis of the FAIR policy proposal demonstrates the impact of these factors on rental application success. For the market rate screening policies tested here, accepted applicants must have from 1 to 3 years of continuous rental history and clean credit reports; 5 years without an eviction; and 3 to 7 years without criminal convictions—depending on the offense. Policies also suggest there are some circumstances under which applicants with these issues in their histories may be considered for acceptance with additional conditions (extra security deposits, co-signers, etc) but it is unclear how those exceptions would be applied. Screening beta test: outcomes for renter applicants across four policies The FAIR screening policy results in more applicants’ being approved for rental access than currently used private market screening policies. Out of 5,854 individual applicants, 4,119 would be approved under the FAIR screening policy. Approximately 900 applicants would still be denied due to not meeting one or more criteria. The remaining applicants’ status is indeterminate given the data available and the policy’s details. 1 The dataset does not include household size, so calculating AMI is not possible; however, he data median income ($2,500) would be approximately 55% AMI for a one person household, and the mean income ($3,100) would be about 70% AMI for a one person household and at 50% AMI for a three person household. Comparing screening outcomes across industry standard operating procedures (5854 individuals) 4500 4000 3500 FAIR policy CCC MFNW Strict mkt 3000 2500 2000 1500 1000 500 0 No Indeterminate Yes In comparison, the industry standard Multi-Family Northwest policy denies approximately 45% of applicants; while the strict market standard provided denies two-thirds of applicants in the dataset. Comparing screening outcomes: FAIR policy and MFNW standard screening (5854 individuals) 80 70 60 50 FAIR policy MFNW 40 30 20 10 0 No Indeterminate Yes Comparing screening outcomes: FAIR policy and strict market screening (5854 individuals) 80 70 60 50 FAIR policy Strict mkt 40 30 20 10 0 No Indeterminate Yes The FAIR policy outcomes most resemble the outcomes for the policy provided by Central City Concern. Comparing screening outcomes: FAIR policy and affordable housing screening (5854 individuals) 80 70 60 50 FAIR policy CCC 40 30 20 10 0 No Indeterminate Yes These analyses show the aggregate outcomes for the entire pool of applicants. A crosstabulation analysis shows how individual applicants fare under the FAIR policy compared with other policies, allowing us to see how many applicants’ outcomes shift from No into a yes condition. Compared to the two private market screening policies, the FAIR screening shifts over 60 percent of those who would have been denied housing into an acceptance. Taking into account both shifts from ‘indeterminate’ to acceptance and denial to acceptance, the FAIR policy provides access for between one third and one half of renter applicants compared to the private market policies. Additional applicants who pass screening under FAIR policy compared to industry standard operating procedures 3500 3000 2500 New acceptance 2000 1500 1000 500 0 CCC MFNW Strict mkt Screening beta test: outcomes for renter applicants with characteristics of concern. The FAIR policy proposal would apply to all households; taking an equity lens to ask whether it has an impact for groups of particular concern, we analyze some subsets of data. The policy is assessed for applicants with low to moderate incomes; for those reporting they will use Housing Choice (‘Section 8’) vouchers; for people of color; and for people with a criminal justice contact history. To consider applicants with low to moderate incomes, we set a monthly income limit at $3,800 per month. This income represents 80% of Area Median Income for a one person household— considered ‘moderate income.’ It would be below 60% of AMI for a four person household. Therefore this figure fits into a low to moderate income level for most household sizes. At this income level, the FAIR policy and the CCC affordable housing screener are again very similar in outcomes. However, the difference between the FAIR policy and market rate screening procedures is sizeable. More than half of the low-moderate income applicants would be denied under MFNW’s policy; and the stricter market policy denies over two thirds of these individuals. Screening outcomes for low-mod incomes (4500 individuals with incomes below $3800/mo) 3500 3000 FAIR MFNW Strict mkt 2500 2000 1500 1000 500 0 No Indeterminate Yes There are 282 applicants who indicate they will use a voucher to pay for part of their rent. While using a voucher as an income source is a protected status under Oregon’s fair housing law, many voucher holders do not pass screening criteria for other reasons such as credit and rental history. About half of voucher holders are accepted outright under both the CCC and FAIR policies, with a quarter needing further review; whereas the reverse is true for the market providers—half are rejected, with a smaller number to be determined with additional review. Finally, we analyzed the screening outcomes for self-identified people of color. It is important to note that approximately half of OneApp users do not volunteer racial/ethnic identity data, so this analysis is not definitive. Within this subset, denial rates are high for market rate providers, with rental history a significant factor in denial. The FAIR policy increases acceptance for these. people of color through its criteria for rental history and credit history. Screening outcomes for self-identified POC (887 individuals) 700 600 500 FAIR MFNW Strict mkt 400 300 200 100 0 No Indeterminate Yes The FAIR policy also provides more access for applicants with a history of criminal justice system contact. A number of applicants’ outcomes could not be determined due to a lack of information about the charge (felony or misdemeanor); these applicants with conviction dates between 3 and 7 years ago are considered indeterminate outcomes. However, many applicants with a convictions history can be accepted because they pass all other screening requirements. It should be noted that the policy for Multifamily Northwest does allow for individualized screening of applicants’ criminal justice history and there may be discretion to allow for additional access; for the purposes of this analysis, we applied a strict reading of the screening criteria, finding most applicants with CJ history could be denied due to a combination of this and other factors. Screening outcomes for applicants with criminal justice history (530 individuals) 450 400 350 300 250 200 150 100 50 0 FAIR MFNW Strict mkt No Indeterminate Yes Procedures As part of crafting the FAIR policy proposal, a workgroup was convened to develop and consider a tenant screening policy that creates a standard set of conditions under which rental housing can be denied an applicant. As part of that deliberation, a study of the potential outcomes of the policy was devised; this analysis was discussed and presented to the workgroup in February 2019. The study simulates tenant screening for four comparison policies: the FAIR proposal; an affordable housing provider, using Central City Concern’s policy; an industry standard operating procedure, the Multi-Family Northwest policy template; and a stricter market policy shared by a market rate management company. Tenant applicant data are provided by OneApp, a technology platform that allows a prospective renter to submit application information for many housing units at once. These data are for 5,854 applicants who submitted complete data and had a completed background check through January 2019. Based on the applicant-supplied information and the screening criteria, each prospective renter is assigned an outcome of yes, no, or indeterminate as an overall response to an application under the policy’s rules. There are important methodological limits to this analysis: first, it does not include income to rent ratios as a factor, as those will vary depending on the unit and makes no conclusions on access to rental housing based on income requirements; second, the vast majority of applicants’ prior landlords have not submitted complete rental references, leaving some information about lease violations impossible to analyze; and third, there are indeterminate outcomes for some renters due to policies that require more information than is contained in the database--for instance, more specificity about the outcome of evictions or criminal justice system contact, or mitigating factors for those circumstances. The OneApp data renter applicants are not perfectly representative of all renter households in Portland. OneApp users have access to technology and a desire to use an app platform to conduct their housing search. Since all housing units are not available via the app, some renters may bypass it in favor of direct contact with the property of their choice. Nonetheless, the renter applicants in this dataset have demographic and income variety and the dataset does include individuals with barriers to housing access, making it viable for this test. It is important to note that these results are not statistically generalizable to all renters in Portland, particularly for subgroups of people, because the dataset was not created with random stratified sampling techniques to represent the renter population.