PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. (21 JULY 2017) This document has been edited and peer-reviewed, but not proofread. A RRO YO CENT ER The Military Accessions Vital to the National Interest (MAVNI) Program Army Performance, Cost, Security Risk, and Potential Pool for Recruiting Beth J. Asch, Jennie W. Wenger, Troy D. Smith Prepared for the United States Army TIWARI_PROD_00002357 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Published by the RAND Corporation, Santa Monica, Calif. © Copyright 2017 RAND Corporation R® is a registered trademark. Limited Print and Electronic Distribution Rights This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited. Permission is given to duplicate this document for personal use only, as long as it is unaltered and complete. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial use. For information on reprint and linking permissions, please visit www.rand.org/pubs/permissions. The RAND Corporation is a research organization that develops solutions to public policy challenges to help make communities throughout the world safer and more secure, healthier and more prosperous. RAND is nonprofit, nonpartisan, and committed to the public interest. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors. Support RAND Make a tax-deductible charitable contribution at www.rand.org/giving/contribute www.rand.org TIWARI_PROD_00002358 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Preface This document reports the results of a study entitled “Assessing the Potential of the Military Accessions Vital to the National Interest (MAVNI) Program.” MAVNI is a U.S. Department of Defense program that allows the armed services to recruit legal aliens in certain immigration categories, such as foreign students, or undocumented immigrants who have been granted deferred action by the Department of Homeland Security who have skills deemed to be vital to the national interest, particularly language or health professional skills. The purpose of this study was to provide information to the U.S. Army on the performance and cost of Army MAVNI versus nonMAVNI recruits, an estimate of the number of potential MAVNI enlistees, and an assessment of the potential security risks associated with the program. This document presents our findings in each of these areas. The findings should be of interest to those involved in recruiting and retaining soldiers, especially those with language and cultural capabilities. This research was sponsored by the Assistant Secretary of the Army for Manpower and Reserve Affairs and conducted within the RAND Arroyo Center’s Personnel, Training, and Health Program. RAND Arroyo Center, part of the RAND Corporation, is a federally funded research and development center sponsored by the United States Army. The Project Unique Identification Code (PUIC) for the project that produced this document is HQD157604. iii TIWARI_PROD_00002359 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Figures and Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv CHAPTER ONE Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Past Research on the Performance of MAVNI Enlistees and Other Noncitizens . . . . . . . . . . . . . . 3 Policy Questions Motivating This Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Overview of Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Layout of the Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 CHAPTER TWO The MAVNI Program and the Enlistment Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Eligibility Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 The Enlistment Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Learning About the MAVNI Enlistment Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Descriptive Statistics of the MAVNI Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 CHAPTER THREE Relative Performance of MAVNI Recruits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Definition of Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Measurements of Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Tabular Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Decomposition of Selected Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 v TIWARI_PROD_00002361 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. vi The Military Accessions Vital to the National Interest (MAVNI) Program CHAPTER FOUR Cost to the Army . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Cost Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Marginal Cost Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 CHAPTER FIVE Estimating the Number of Potential MAVNI Enlistees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Estimating the Size of the Pool from ACS Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Estimates Based on SEVIS Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Estimates Based on Admissions Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 CHAPTER SIX Estimating Security Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Data and Methodological Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 Overview of Psychological and Sociological Factors Related to Terrorism and Extremism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Overview of Demographic Factors Related to Terrorism and Extremism . . . . . . . . . . . . . . . . . . . . . 67 Literature on Observed Behavioral Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Overview of Factors Related to Espionage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 Discussion and Implications for the MAVNI Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 Mitigation Efforts: Screening and Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 CHAPTER SEVEN Final Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 APPENDIXES A. MAVNI Languages and Nonimmigrant Legal Status Categories . . . . . . . . . . . . . . . . . . . . 87 B. Blinder-Oaxaca Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 TIWARI_PROD_00002362 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Figures and Tables Figures 2.1. 2.2. 2.3. 2.4. 3.1. 3.2. 3.3. 3.4. 3.5. 3.6. 3.7. 3.8. 5.1. 5.2. 5.3. 6.1. 6.2. 6.3. MAVNI Enlistment Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Google Trends: Searches for “MAVNI” Over Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Google Trends: MAVNI-Relevant Searches Over Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 New Users and Sessions on GoArmy.com, MAVNI Versus Others, by DMA. . . 18 Distribution of Contracts Across Enlistment Term. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Percentage of Contracts That Do Not Access by FY, Conditional on Four-Year Enlistment Term, MAVNI Recruits and Comparison Groups . . . . . . . . . . . . . . . . . . . . 33 Cumulative Percentage Retention of Wave 1 MAVNI Accessions, Conditional on Four-Year Enlistment Term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Cumulative Percentage Retention (Non–Failure to Adapt) of Wave 1 MAVNI and Non-MAVNI Accessions, Conditional on Four-Year Enlistment Term . . . . . 35 Reason for Separation Among Wave 1 MAVNI and Non-MAVNI Accessions Who Separated, Conditional on Four-Year Enlistment Term . . . . . . . . . . . . . . . . . . . . . 36 Reenlistment or Officer/RC Transfer Among Wave 1 Accessions Who Complete First Term, Conditional on Four-Year Enlistment Term . . . . . . . . . . . . . . . 37 Promotion to E-4 Among Wave 1 Accessions Who Complete First Term, Conditional on Four-Year Enlistment Term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Promotion to E-5 Among Wave 1 Accessions Who Complete First Term, Conditional on Four-Year Enlistment Term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 Number of F-1 and M-1 Students by MAVNI-Language-Speaking Country of Citizenship, April 2014–July 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Number of Admissions for MAVNI Visa Types from Countries That Speak a MAVNI Language, 18- to 34-Year-Olds, 2010–2015 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Number of F and M Visa Admissions for Countries That Speak a MAVNI Language, by Language, 18- to 34-Year-Olds, 2010–2015 . . . . . . . . . . . . . . . . . . . . . . . . 60 Percentage of Accessions with Incomplete Security Investigation Before Ship Date to Basic Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Percentage of Accessions with Incomplete Security Investigation Before Start of Advanced Individual Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Mean Winsorized Days to Complete Security Investigation . . . . . . . . . . . . . . . . . . . . . . . 79 vii TIWARI_PROD_00002363 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. viii The Military Accessions Vital to the National Interest (MAVNI) Program 6.4. A.1. Army Asymmetric Warfare Group Tactical Reference Guide for Risk Mitigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 MAVNI Languages in FY 2016 and Changes for FY 2017 . . . . . . . . . . . . . . . . . . . . . . . 88 Tables 1.1. 2.1. 2.2. 2.3. 2.4. 2.5. 2.6. 2.7. 3.1. 3.2. 3.3. 4.1. 4.2. 4.3. 4.4. 6.1. A.1. B.1. MAVNI Enlistment Cap by Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 MAVNI Enlistment Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Measures of Engagement, GoArmy.com . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 MAVNI HCP Contracts as Percentage of All MAVNI Contracts . . . . . . . . . . . . . . . . 19 Visa Category Breakdown of MAVNI CFL Contracts in FY 2014 and 2015 . . . 20 Percentage of MAVNI CFL Contracts That Are Regular Army and DACA, by FY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Percentage Distribution of Language Capability Among MAVNI CFL Contracts, by Fiscal Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Percentage Distribution of Army Career Management Field Among MAVNI CFL Contracts, by FY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Demographic and Job Characteristics of Army Accessions with Four-Year Enlistment Term of Service, Wave 1 and Wave 2, MAVNI Versus Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Army DEP and Accession Waivers Among Those with Four-Year Enlistment Term of Service, Wave 1 and Wave 2, MAVNI Contracts and Accessions Versus Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Decomposition of the Difference in Mean Performance Metrics into Percentages Due to Differences in Characteristics and to Differences in Other Factors, MAVNI Recruits Versus All High-Quality Non-MAVNI Recruits . . . . 39 Incremental Budget Cost per Contract of MAVNI Recruits Versus Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Computation of First-Term Person-Years, Given Contract for MAVNI and Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Incremental Budget Cost per First-Term Person-Year of MAVNI Versus Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Incremental Budget Cost per Contract and First-Term Person-Year of MAVNI Versus Three Comparison Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Risk Factors Identified in the Psychology and Sociology Literature . . . . . . . . . . . . . . . 67 Nonimmigrant Legal Status Categories Eligible for MAVNI . . . . . . . . . . . . . . . . . . . . . . 89 MAVNI and High-Quality Non-MAVNI Regression Estimates (Linear Probability Model). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 TIWARI_PROD_00002364 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Summary The Military Accessions Vital to the National Interest (MAVNI) program is a U.S. Department of Defense (DoD) recruiting pilot program that was initially authorized in fiscal year (FY) 2009 for the purpose of enlisting individuals with critical skills vital to national security. This program, which has been reauthorized in most years since FY 2009, allows the armed services to recruit qualified, legal nonimmigrants who possess skills deemed vital to the national interest. Most MAVNI enlistees possess language skills deemed critical, but some possess critical health skills. The program allows certain categories of noncitizens to enlist; most enlistees to date have entered the United States on student visas. Those on some work visas, spouses of those on some student or work visas, refugees, and asylees are eligible to enlist under this program, as are individuals whose status is as a Deferred Action for Childhood Arrival. DoD has placed annual caps on the MAVNI program; the Army has always taken the largest share of the cap. In this project, we provide information on several aspects of MAVNI enlistees and the program. 1. 2. 3. 4. 5. We characterize the performance of MAVNI enlistees to date. We estimate the relative costs of MAVNI and non-MAVNI enlistees. We estimate the number of potential MAVNI enlistees. We review the MAVNI enlistment process. We provide some information on the potential security risks associated with the MAVNI program. Our research, which focused primarily on MAVNI enlistments with critical language skills, indicates that MAVNI has been a cost-effective recruiting program for the Army. The cost-effectiveness of MAVNI recruits results from their relatively low attrition rates. Their better performance stems from their faster promotion rates and their higher likelihood of having characteristics associated with better performance, such as more education and higher aptitude scores. The program taps a potential recruiting pool that could perhaps sustain a modest increase, though we were unable to estimate the size of the potential pool precisely or the number of additional accessions that ix TIWARI_PROD_00002365 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. x The Military Accessions Vital to the National Interest (MAVNI) Program would be possible in the absence of the cap. We were also unable to estimate the specific security risk associated with MAVNI recruits. Our review of policy memoranda and the literature indicates that MAVNI recruits undergo extensive screening relative to the typical Army recruit for factors that include family connections, education, foreign ties, mental health issues, criminal activity, drug and alcohol abuse, credit problems, and other external factors. Many of these factors are identified in the literature as being associated with espionage. Although this is the screening that is used to determine eligibility for clearances, MAVNI recruits do not receive a security clearance. When we characterized performance, we compared MAVNI recruits with all non-MAVNI recruits, with all high-quality non-MAVNI recruits, and with all noncitizen non-MAVNI (e.g., green card holder) recruits. The last group includes individuals who enlist as immigrants, though not necessarily with the same language skills and cultural capabilities as MAVNI recruits. We found that MAVNI recruits perform better in terms of lower Delayed Entry Program (DEP) attrition and first-term attrition and faster promotion times to E-4 and E-5.1 They also have a number of characteristics associated with better performance, including age (recruits who are no longer teenagers perform better), higher levels of education, higher Armed Forces Qualification Test (AFQT) scores, and lower incidence of DEP waivers. On the other hand, they are less likely to reenlist than non-MAVNI recruits or to enter occupations deemed critical. The lower reenlistment rate of MAVNI recruits is primarily attributable to factors other than observed entry characteristics. For example, the opportunity to gain citizenship under MAVNI may draw into service individuals with a lower taste for military service and who, once their service obligation is complete, are more likely to leave. MAVNI recruits are not permitted to enter many occupations (including most of those deemed “critical”) because they lack a security clearance. We also find that, on a per-contract basis, MAVNI recruits are more costly from a budget standpoint than non-MAVNI recruits, but, because they have lower first-term attrition, they provide more first-term person years than non-MAVNI recruits. On a person-year basis, then, MAVNI personnel have been less costly than all non-MAVNI recruits and high-quality non-MAVNI recruits. MAVNI recruits have been about as costly per person-year as noncitizen non-MAVNI recruits. Because MAVNI recruiting is demand-constrained, meaning that more individuals would like to enlist than the cap allows, the incremental cost of expanding enlistment supply through the MAVNI program equals the average cost of MAVNI, about $41,300 per contract. In contrast, past studies estimate an incremental cost of a high-quality (non-MAVNI) recruit ranging from $66,300 to $118,000 per contract. Thus, at the margin, increasing MAVNI recruits has been a less costly approach to expanding enlistment supply, given the cur1 DEP is formally called the Future Soldier Program by the Army. We use the term DEP rather than Future Soldier Program to refer to the period between when the applicant signs a contract and when the applicant actually enters the Army and ships to basic training. TIWARI_PROD_00002366 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Summary xi rent level of caps on MAVNI, than increasing non-MAVNI high-quality recruits. We note, however, that additional security and screening requirements were placed on the program beginning in FY 2017; our data are drawn from earlier years. These requirements will increase the cost of MAVNI contracts and recruits. Although we have no information about the specific cost increases, these costs have the potential to change the comparison of MAVNI and non-MAVNI recruits. We attempted to estimate how large the potential pool of MAVNI recruits is, but we found that none of the existing data sources provided sufficient information on all of the key qualification characteristics of MAVNI recruits. Student and Exchange Visitor Information System data from the Department of Homeland Security suggest that, at any point in time, there are over 900,000 F-1 and M-1 visa students from countries where a MAVNI language is spoken, but this figure does not account for residency requirements for MAVNI recruits or restrictions related to age, AFQT, waivers, health status, or weight. That said, these data do suggest that the pool is growing and that the pool is sufficient to meet the caps set on the MAVNI program in recent years, with perhaps some increase in the cap. We are unable to say how many more accessions would be possible, however, not only because of uncertainty about the size of the potential pool and the number of MAVNI who would be eligible and positively inclined to join the Army, but also because the Army’s current capacity to process MAVNI recruits is limited. We also investigated what is known about the security risk associated with MAVNI and specifically reviewed the literature on whether foreign-born status is a relevant factor considered to be associated with higher risk. There are no publicly available reports of MAVNI recruits engaging in terror-, sabotage-, or espionage-related activities. More broadly, available data sources indicate that such acts are extremely rare among military personnel and among Americans in general. (To ensure that our analyses are available to the broadest possible audience, we utilize only unclassified information and data.) Much of the past literature focuses on factors thought to be associated with extremist activity and espionage and provides guidance on relevant factors to screen and monitor to prevent extremist activity and espionage. In general, we found that the literature on security risk is still evolving, with much of it challenged by data and methodological issues. While the literature provides no real guidance on how to estimate security risk for specific groups, MAVNI included, it is clear that MAVNI recruits undergo extensive security screening. That said, the screening they receive is informed more by the factors identified in the literature as being associated with espionage risk than by those factors specifically associated with terrorism. MAVNI recruits do have links to foreign countries; research on espionage finds that foreign ties and loyalties are prevalent among those who have engaged in espionage. However, we find mixed evidence on the role of immigrant status. Past research finds that immigration from terrorist-prone countries is a transmission mechanism of TIWARI_PROD_00002367 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. xii The Military Accessions Vital to the National Interest (MAVNI) Program terrorist activity, but that countries with more migrant flows actually experience less terrorism. Data on terrorist acts in the United States since 2001 show that refugees or other nonimmigrants do not perpetrate the majority of such acts. The Office of the Secretary of Defense has made significant changes to the MAVNI program in FY 2017, as mentioned above. These include a decrease in the total cap and changes to the critical language list. Overall, these changes could be described as increasing emphasis on the critical language capacity of the program (versus the capacity to access additional high-quality recruits). In the future, languages spoken by MAVNI enlistees will not include several of the languages that made up a large proportion of past MAVNI enlistees; also, languages will be capped in the sense that no more than 10 percent of MAVNI accessions will speak a single language. At the same time, the screening requirements of the program have been enhanced. These changes will increase the relative cost of MAVNI enlistments and will also decrease the pool of potential MAVNI enlistees. To date, MAVNI recruiting has been quite localized in nature; many enter the Army from one of a few metropolitan areas. Ensuring a wider breadth of language skills may involve some expansion in recruiting resources. Nonetheless, our analyses also suggest continued widespread interest in the program; while our attempts to estimate the total pool of potential MAVNI enlistees were hampered by a lack of data, most of the information we examined suggests that interest in the MAVNI program is likely to continue to exceed the demand, especially at the new lower total cap. Increased screening will increase the relative costs of MAVNI enlistees by an unknown amount. That said, the higher cost associated with additional screening would have to be dramatic for MAVNI recruits to be less cost-effective than other high-quality recruits. Finally, increasing the initial obligation offers one option to further increase the cost-effectiveness of MAVNI recruits. TIWARI_PROD_00002368 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Acknowledgments We are grateful to Linden St Clair, assistant deputy for recruiting in the Office of the Assistant Secretary of the Army for Manpower and Reserve Affairs, who supported this research and provided guidance and input throughout this project. We would like to thank Karl Kalb for information related to Army personnel costing, as well as Marsha Popp and the staff who maintain the Army Military-Civilian Cost System for providing detailed explanations of the costing methodology. Our project benefited substantially from the input of the subject matter experts with whom we spoke at the U.S. Army Recruiting Command and U.S. Army Human Resources Command, as well as the recruiters with whom we spoke about MAVNI recruiting. At RAND, Tina Panis provided terrific assistance with the Army personnel data used in this analysis, and we would also like to thank Kristine Brown, Angela Clague, Steven Davenport, Etienne Rosas, and Gabriel Weinberger for their contributions to our project. We also gratefully acknowledge the very helpful comments on our research by Michael Hansen at RAND and from three reviewers, Heather Krull and John Hollywood from RAND and Curtis Gilroy, former director of accession policy with the Office of the Secretary of Defense (Personnel and Readiness). Our report greatly benefited from the input. xiii TIWARI_PROD_00002369 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Abbreviations ACES Automated Continuous Evaluation System ACS American Community Survey AFQT Armed Forces Qualification Test AMCOS Army Military-Civilian Cost System ASVAB Armed Services Vocational Aptitude Battery BCT basic combat training BMI body mass index CFL critical foreign language CI counterintelligence CMF career management field DACA Deferred Action for Childhood Arrival DEP Delayed Entry Program DHS Department of Homeland Security DoD U.S. Department of Defense DMA designated marketing area FY fiscal year GED General Educational Development HCP health care professional ISIL the Islamic State of Iraq and the Levant xv TIWARI_PROD_00002371 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. xvi The Military Accessions Vital to the National Interest (MAVNI) Program MAVNI Military Accession Vital to the National Interest MEPS Military Entrance Processing Station MOS Military Occupational Specialty MPA Military Personnel Appropriation NACLC National Agency Check with Law and Credit Check NIAC National Intelligence Agency Check OMA Operations and Maintenance Appropriation OPI oral proficiency interview OSD Office of the Secretary of Defense PERSEREC Defense Personnel Security Research Center PSI Personnel Security Investigation RA Regular Army RC reserve component SAVE Systematic Alien Verification for Entitlements SEVIS Student and Exchange Visitor Information System SEVP Student and Exchange Visitor Program SME subject-matter expert SSBI Single-Scope Background Investigation TAPDB Total Army Personnel Database USAR U.S. Army Reserve USAREC U.S. Army Recruiting Command TIWARI_PROD_00002372 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER ONE Introduction The Military Accession Vital to the National Interest (MAVNI) program is a U.S. Department of Defense (DoD) military recruiting pilot program that was authorized on November 24, 2008.1 Though initially a one-year pilot program, it has been reauthorized in most years since then. The program allows the armed services to recruit qualified, legal nonimmigrants who possess skills deemed vital to the national interest. These skills are either critical foreign language skills or health professional skills. Nonimmigrants refers to specific categories of noncitizens, as described in more detail in Chapter Two. For example, in the case of MAVNI, these include individuals in the United States on student visas (e.g., F-1 visas) or on work visas in specialized occupations or professions (e.g., H-1B visas) and their spouses, as well as refugees and asylees. MAVNI accessions can join either the active component or a reserve component (RC) (though not the National Guard), but DoD puts an annual cap on how many MAVNI recruits can enter the military, with the Army allocated the largest share of the total. When the program was initiated in fiscal year (FY) 2009, the cap was 1,500; in FY 2016, the overall cap was 5,200. In response to a comprehensive review of the program, particularly security requirements, DoD made a number of changes for FY 2017, including reducing the cap to 1,400 overall and 1,200 for the Army (DoD, 2016). Table 1.1 shows the overall DoD cap and Army cap in each year of the program. No values are shown for FY 2011 and FY 2012 because no MAVNI enlistments were accepted, as DoD sought to develop guidance to strengthen the program to mitigate security concerns. The program was reinstated in FY 2013. To date, the Army has had little trouble filling the MAVNI cap, often doing so prior to the end of the FY. In addition to the usual enlistment qualification criteria that all recruits must meet, MAVNI recruits must meet several additional qualification criteria, described in more detail in Chapter Two. First, all MAVNI accessions must be high-quality recruits, where high-quality is defined as having at least a high school diploma and a score of at least 50 on the Armed Forces Qualification Test (AFQT). Furthermore, MAVNI recruits cannot receive an enlistment waiver for any kind of misconduct, 1 See Secretary of the Army (2008), as well as DoD (2010) and DiFazio et al. (2013). 1 TIWARI_PROD_00002373 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 2 The Military Accessions Vital to the National Interest (MAVNI) Program Table 1.1 MAVNI Enlistment Cap by Year Fiscal Year DoD Cap Army Cap 2009 1,000 890 2010 1,000 890 2011 N/A N/A 2012 N/A N/A 2013 1,500 1,200 2014 1,500 1,300 2015 3,200 3,000 2016 5,200 5,000 2017 1,400 1,200 SOURCES: Various Army and DoD instructions and memoranda; see Department of the Army (2014) and Office of the Under Secretary of Defense (2010). though they can receive medical waivers. Thus, the education, aptitude, and waiver requirements are more stringent than they are for the typical recruit. In addition, MAVNI recruits must satisfy a residency requirement: They must have continuously resided in the United States for at least two years to be eligible to enlist. We describe these criteria in more detail in Chapter Two. Aspects of the MAVNI program have changed over time, in addition to the overall cap. The list of critical languages and medical specialties has changed from year to year. For example, general nursing was included among the list of specialties under MAVNI in FY 2009 when the program started,2 but is no longer included. Security investigation requirements have also become more rigorous over time. Since FY 2013, MAVNI recruits have been required to undergo a more rigorous security investigation prior to entry than the typical recruit; this includes a Single Scope Background Investigation (SSBI) and counterintelligence (CI) security review. In FY 2014, the Office of the Secretary of Defense (OSD) allowed the military to recruit individuals with Deferred Action for Childhood Arrival (DACA) status. For FY 2017, additional security-related changes have been made, along with new limits on the percentage of accessions who speak a single language. The enlistment process of MAVNI recruits is discussed in more detail in Chapter Two. A unique aspect of the MAVNI program is that MAVNI recruits receive U.S. citizenship shortly after accession; however, citizenship may be revoked if MAVNI acces2 See Headquarters USAREC (2015). TIWARI_PROD_00002374 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Introduction 3 sions fail to satisfactorily complete their initial service obligation. Thus, the program provides an enlistment incentive and embeds an incentive for MAVNI recruits to complete their initial service obligation.3 Past research on the retention effects of incentives intended to enhance high-quality enlistments, such as enlistment bonuses and educational benefits, show that these incentives can have a small but discernable selection effect, whereby they attract youth who are less attached to the military and thus more likely to separate after completion of their obligation (Simon, Negrusa, and Warner, 2010; Asch and Dertouzos, 1994). This past research raises the possibility that the MAVNI program also draws into service individuals with less attachment to the military. Thus, it is possible that MAVNI recruits may be less likely than other recruits to leave during their first enlistment term, but they may be more likely to leave once they complete the first term. Past Research on the Performance of MAVNI Enlistees and Other Noncitizens The MAVNI program is quite new, and existing research on any aspect of MAVNI enlistees’ performance is limited. Indeed, the only published study to focus on firstterm performance of MAVNI enlistees tracks performance of an early cohort that enlisted in 2009 through their first 36 months of service (DiFazio et al., 2013). This research compares the performance of the 840 MAVNI recruits who contracted to enter the Army prior to FY 2012 with the performance of the 90,393 non-MAVNI recruits who contracted to enter the Army in FY 2009. The study found that MAVNI recruits were better qualified and performed better in terms of having lower loss rates from the Delayed Entry Program (DEP); lower early and first-term attrition rates; and better reported performance, according to their supervisors. It also found that MAVNI soldiers’ strong performance was linked to such personal characteristics as age, education, and aptitude scores. In other words, MAVNI recruits possess the characteristics that are associated with low levels of attrition. However, after controlling for these characteristics, being a part of the MAVNI program did not have a statistically significant effect on attrition. The study also conducted a series of surveys; MAVNI soldiers’ supervisors reported that MAVNI soldiers’ levels of performance were above average and that MAVNI soldiers possessed good communication skills. That said, 79 percent of surveyed MAVNI soldiers reported at least some degree of dissatisfaction with the MAVNI program, with many indicating that their skills were underutilized. This level 3 Immigrant noncitizens—i.e., legal permanent residents or “green card” holders—are another group of noncitizens who can enter the military and achieve citizenship through an accelerated process. That is, military service exempts these individuals from residency requirements. These individuals will have their citizenship revoked if they separate from the military under “other than honorable conditions” and before completing five years of honorable service (USCIS, 2016). TIWARI_PROD_00002375 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 4 The Military Accessions Vital to the National Interest (MAVNI) Program of dissatisfaction could also be a factor in reenlistment decisions, though the study did not consider reenlistment. There is also past research related to immigrants in the military who are not MAVNI recruits, though it too is quite sparse. These noncitizens are “green card” holders who are permanent residents, unlike MAVNI recruits who are noncitizens in the United States temporarily or are refugees or asylees. About 10 percent of the youth population are noncitizens (Acosta, Larsen, and Grieco, 2014), while about 5 percent of enlisted Army accessions are noncitizens. While the process of becoming a citizen in the United States has evolved considerably over the past century, military service has typically provided exceptions leading to an accelerated path to citizenship, a process that would otherwise involve a long waiting period. Many immigrants have served in the U.S. armed forces, especially during past periods of extensive immigration. For example, nearly half of Army recruits in the 1840s were immigrants.4 During the Civil War, Congressional legislation allowed immigrants who served in the Union forces to become citizens after one year (Sohani and Vafa, 2010). Noncitizens comprised about 18 percent of the Union Army during this period; during World Wars I and II, military service remained a fast track to naturalization (Barry, 2013). In recent years, opportunities for noncitizens to serve in the armed forces and receive expedited citizenship have continued; on July 3, 2002, President Bush signed Executive Order 13269, accelerating immigrant soldiers’ naturalization period from three years to one day of service; this policy was designed for permanent residents already holding a green card.5 In 2013, the Obama administration instituted a new policy permitting the undocumented family members of active-duty personnel in the military, reservists, and veterans to stay in the United States (Veterans for Immigration Reform, 2014). Past research on the performance of noncitizen recruits suggests that noncitizens in the military are more likely to complete their initial terms of service than similar citizens (Hattiangadi et al., 2005; McIntosh, Sayala, and Gregory, 2011). In each case, the results hold after controlling for other observable characteristics at entry, such as aptitude scores and educational credentials. The estimates of Hattiangadi et al. (2005) also indicated that noncitizens have the potential to expand the size of the recruiting pool. Policy Questions Motivating This Study Given that MAVNI recruits must meet more-stringent eligibility criteria and have a citizenship incentive that could affect their decision to enter and stay in the military 4 Weigley (1967), cited in Hattiangadi et al. (2005). 5 Executive Order 13269, 2002. TIWARI_PROD_00002376 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Introduction 5 and that the program allows accessions from a new population source, it would seem that the MAVNI program offers the armed services a source of high-quality recruits with critical skills who might initially perform better than non-MAVNI recruits. Indeed, the available past research discussed above suggests that MAVNI recruits do perform better, as do noncitizen recruits in general. The Army requested that the RAND Arroyo Center provide an updated analysis of the program with more recent data, including data after the hiatus in FY 2011 and FY 2012. The Army requested that the updated study include a cost analysis that provided information on both the relative performance and the relative cost of MAVNI and non-MAVNI recruits. It also requested an estimate of the size of the pool of potential MAVNI enlistments to better understand how many MAVNI accessions are feasible and the extent to which MAVNI might expand the pool of potential high-quality enlistees. Finally, the Army requested information related to security risk, given potential concerns with expanding the recruiting pool to foreign-born nonimmigrants in an age of heightened awareness and concerns about terrorism, as well as the increasing investigation-related requirements over the history of the MAVNI program. The current study focuses on the subset of noncitizens in the Army who are also nonimmigrants—i.e., those who are not in the United States permanently. Our work expands on previous work on the MAVNI program by tracking performance over a longer period of time, including additional comparison groups, and estimating costs associated with recruiting MAVNI soldiers. We also examine the issue of security risk associated with MAVNI recruiting and attempt to estimate the size of the potential pool of MAVNI recruits in the U.S. population. Overview of Approach The research summarized in this report is divided into five main tasks, based on the Army’s request. 1. The study describes the MAVNI enlistment process. 2. The study uses Army recruiting and personnel data on MAVNI and nonMAVNI recruits to characterize the performance of MAVNI enlistments to date. Performance is measured in terms of DEP attrition, early and first-term attrition, first-term reenlistment, and promotion speed. 3. The study develops estimates of the average and marginal cost of MAVNI and non-MAVNI recruits. These estimates provide information on the additional budgetary cost of MAVNI accessions on a per-recruit basis and on a per-personyear basis. They also provide information on the incremental or marginal cost of expanding high-quality enlistments by expanding MAVNI versus by other means, such as increasing pay. Both the analysis of performance and cost use TIWARI_PROD_00002377 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 6 The Military Accessions Vital to the National Interest (MAVNI) Program data through September 2015 on Regular Army (RA) MAVNI accessions in critical languages.6 4. The study assesses the usefulness of available data sources and past studies for estimating the number of potential MAVNI enlistments and provides a rough upper-bound estimate of the size of the eligible pool of MAVNI recruits, as well as information on how that pool has been growing in recent years. These data sources include information from the Department of Homeland Security (DHS) on the number of student visas and admissions to the United States of individuals with different visa categories, as well as census data from the American Community Surveys (ACS). 5. Finally, the study reviews a variety of information in the unclassified realm, including the available academic and policy literature and policy memoranda from DoD and the Army to describe what is known about security risks associated with military recruits and what is known about the factors that seem to be related to acts of terrorism and espionage and the efforts on the part of DoD and the Army to mitigate these risks. Layout of the Report The remainder of the report describes our methods and findings. Chapter Two provides an overview of the MAVNI program and the MAVNI enlistment process and reviews the limited literature on MAVNI enlistees’ performance, as well as the literature on other noncitizens. Chapter Three shows the results of our analysis of performance of MAVNI recruits relative to comparison groups; it discusses the choice of comparison groups, describes the data and our metrics of performance, and presents the results. In Chapter Four, we present our cost analysis. Chapter Five discusses the estimation of the pool of potential MAVNI accessions and the drawbacks of available data sources. In Chapter Six, we present the literature review and review of policies related to security risk. Chapter Seven gives our concluding thoughts. 6 Our data analysis does not include MAVNI accessions in the health care professions because they comprise a relatively small group. Furthermore, their accession process and early career is not directly comparable to nonMAVNI health professions, so identifying a non-MAVNI comparison group is difficult, given the different ways health professionals can enter the military. We discuss these topics in more detail in Chapter Two. TIWARI_PROD_00002378 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER TWO The MAVNI Program and the Enlistment Process We begin this chapter with an overview of the eligibility requirements that MAVNI recruits must meet and then discuss the enlistment process. As part of the latter discussion, we also provide information we gathered from interviews with subject-matter experts (SMEs) about how the process works, as well as information about how potential MAVNI applicants learn about the MAVNI program, especially through social media. We conclude the chapter with some descriptive statistics of the MAVNI program. Eligibility Requirements All enlistees must meet several enlistment standards. These standards include required education credentials and test scores and the capacity to meet or exceed health- and fitness-related standards and certain “moral” standards. For example, the latter require that enlistees not have outstanding legal issues or convictions and that enlistees pass tests for illegal substances. Some of the standards, most often the required test scores, vary based on the Military Occupational Specialty (MOS) or occupation the enlistees wish to enter, but the vast majority of new enlistees possess a high school diploma or an equivalent credential (a college diploma in the case of officers). Among enlisted personnel, key test scores are based on the Armed Services Vocational Aptitude Battery (ASVAB)/AFQT.1 Health- and fitness-related standards have been problematic for many enlistees in recent years; in particular, enlistment requires that personnel meet weight-for-height and body fat standards. As a group, these standards disqualify many young people from enlistment. As we discuss in Chapter Five, only 26 percent of young adults meet qualification standards (Seifert, Hogan, and Moore, 2007). Those wishing to enlist as part of the MAVNI program are also required to meet additional, more-stringent standards. These standards are listed in Table 2.1. MAVNI recruits are required to possess a high school diploma, score at or above the 50th percentile on the AFQT, demonstrate an understanding of conversational English, and typically be younger than 35. The high school credential and test score requirements 1 The AFQT is based on four of the ten subtests that make up the ASVAB. 7 TIWARI_PROD_00002379 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 8 The Military Accessions Vital to the National Interest (MAVNI) Program Table 2.1 MAVNI Enlistment Requirements Characteristic Qualification Immigrant status DACA,a asylee, refugee, or Temporary Protected Status (TPS)b in an enumerated nonimmigrant category (E, F, H, I, J, K, L, M, O, P, Q, R, S, T, TC, TD, TN, U, or V); foreign tourists with legal, nonpermanent visas; and students with legal nonpermanent visasc Length in status Asylees, refugees, TPS, and those in an enumerated nonimmigrant category must have their status in one or more of the approved statuses at least two years prior to enlistment. Nonimmigrants must not have a single absence from the country for more than 90 days during the two-year eligibility period. Age All MAVNI enlistees must be at least 17 years old. Critical foreign language (CFL) MAVNI enlistees must be under the age of 35 at the time of shipping and training. Health care professional (HCP) MAVNI enlistees must be under the age of 40 at the time of shipping. Education requirements High school diploma AFQT score Minimum of 50 Foreign language proficiency CFL MAVNI enlistees must score at least 2+/2+ in their foreign language, scores that mean “Limited Working Proficiency, Plus.” English proficiency Health care professional applicants must pass the oral proficiency interview (OPI), demonstrating an understanding of “conversational” English. Waivers Applicants cannot require an enlistment waiver for any kind of previous misconduct. They must meet physical fitness standards; however, medical waivers are authorized. Other Applicants cannot have a current or remaining affiliation with another country’s armed forces, which includes reserve or guard commitments. a DACA is an immigration policy begun in 2012 that allows eligible undocumented immigrants who enter the United States as minors to receive a renewable two-year period of deferred action from deportation and eligibility for a work permit. b Those with TPS are foreign nationals in the United States because conditions in the foreign country temporarily prevent them from returning safely. c Appendix A includes brief descriptions of each relevant immigrant status. SOURCES: Various Army and DoD Instructions and memoranda. See, in particular, DoD (2010). mean that MAVNI enlistees meet the traditional definition of “high quality”; such enlistees have been found to complete their initial terms of service at higher rates than others (Buddin, 2005). While the majority of non-MAVNI enlistees meet the “highquality” standard, some do not. Therefore, MAVNI enlistees may have better performance than others because of their educational and test score credentials. Prior to the establishment of the MAVNI program, those holding immigrant visas or “green cards” were eligible to enlist, but those holding nonimmigrant (temporary) visas were not eligible to serve in the U.S. armed forces. The MAVNI program served to considerably expand the number and type of visa holders who were eligible TIWARI_PROD_00002380 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 9 to enlist, as shown in Table 2.1. Specifically, it allowed the enlistment of those holding student visas, as well as those holding visas related to specific occupations; victims of trafficking; and refugees, asylees, and those with temporary protected status. In FY 2014, OSD further extended MAVNI eligibility to include those with DACA status. To date, the vast majority of MAVNI recruits have been in the United States on student visas, and student visas make up the largest group of potential MAVNI enlistees, but other groups are also eligible to enlist under the program. We discuss this in more detail in Chapter Five. The MAVNI program provides a fast track to citizenship. Indeed, those holding student visas are not otherwise eligible for citizenship; in most cases, students must first convert to a different visa category, then apply for a green card (permanent status), and then begin the application for citizenship. The waiting period between receiving a green card and receiving citizenship is long.2 MAVNI enlistees must undergo considerable screening prior to accession, but they generally receive citizenship within a few months of beginning basic combat training (BCT). Therefore, enlistment under the MAVNI program essentially allows the service member to skip the green card process. Green card holders also are eligible to enlist, and they also are eligible for fasttracked processing of citizenship. The difference between MAVNI recruits and these green card holders is that MAVNI recruits are guaranteed citizenship at entry to the military or shortly thereafter. In contrast, green card holders who join the military are simply given access to an accelerated process; military service gives them a residency exemption so that the process goes faster, but they may not receive citizenship as quickly as MAVNI recruits. The MAVNI program has two parts: CFL-MAVNI, for critical foreign language experts, and HCP-MAVNI, for health care professionals. CFL-MAVNI recruits provide language and cultural skills considered critical. The list varies over time, as we discuss later in this chapter. The Enlistment Process MAVNI personnel serve in a variety of capacities. As we show later in this chapter and in the next one, a substantial fraction joins the RC, and some MAVNI personnel become officers. Here, we describe the enlistment process mostly from the perspective of a recruit enlisting in the active component (RA) and, in the case of MAVNI, with CFL skills. We note pertinent differences as relevant. We use two sources of information. The first group of sources includes various Army and DoD instructions and memoranda. The second group of sources includes 2 An applicant must first apply for a green card; after holding the green card for at least five years, the applicant may begin the process of applying for citizenship. Refugees and asylees go through a similar process; they are eligible to apply for a green card one year after receiving refugee or asylee status. See USCIS (2013). TIWARI_PROD_00002381 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 10 The Military Accessions Vital to the National Interest (MAVNI) Program selected interviews with SMEs who work with MAVNI applicants and recruits at Fort Knox and also with three recruiters who have substantial experience recruiting MAVNI soldiers. The purpose of the interviews was to identify processing issues specific to the MAVNI program. We first discuss the process, as summarized in Figure 2.1, and then the broad themes that emerged from the interviews. The MAVNI enlistment process is an extended version of the process that applies to the typical recruit and is shown in Figure 2.1. The typical applicant is first interviewed by a recruiter who performs an initial screening. As part of this process, the recruiter will administer an enlistment screening test and a test to predict the applicant’s AFQT score. The typical applicant will then proceed to the closest Military Entrance Processing Station (MEPS) for further testing. MEPS testing includes physical tests, aptitude tests, and questions related to moral standards. Applicants who meet enlistment standards can then enlist by signing an enlistment contract. Those who do not immediately ship to basic training become part of DEP. During DEP, enlistees entering occupations requiring a security clearance undergo additional security screening. For example, those requiring a secret clearance will receive the National Agency Check with Law and Credit Check (NACLC) or the National Intelligence Agency Check (NIAC), while those requiring a top-secret clearance will receive a Single-Scope Background Investigation (SSBI). MAVNI recruits regularly undergo additional screening steps, as shown in Figure 2.1, because they are subject to additional eligibility requirements and because they receive citizenship at entry. During the applicant stage, before they go to the MEPS, their immigration forms are submitted, their U.S. residency is checked, and they undergo an initial security screening. Their immigration information is submitted to the DHS SAVE system for processing, and their paperwork is submitted to the Social Security Administration to receive a Social Security number. During this stage, they also undergo language proficiency testing—the OPI—by the Defense Language Institute. Applicants must receive a 2+/2+ score in their foreign language, scores that mean “Limited Working Proficiency, Plus.” Like other recruits, MAVNI applicants are further tested at the MEPS. Once in the DEP, MAVNI enlistees must pass two PSIs. First, they undergo a NACLC or NIAC screen; this involves basic national agency checks (Security/ Suitability Investigations Index, Defense Central Index of Investigations, fingerprint classification, and Federal Bureau of Investigation [FBI] index) and a credit search covering all residences, employment, and education locations. In the case of other enlistees, these screens are required prior to granting secret or confidential clearances. Second, MAVNI enlistees also undergo an SSBI. This investigation, designed to screen personnel who require a top secret clearance, includes local law enforcement checks; contacting references; checks of education, employment, and Selective Service records; and an in-person interview. Note that MAVNI recruits do not receive a security clearance when they successfully complete the SSBI and NIAC. TIWARI_PROD_00002382 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 11 Figure 2.1 MAVNI Enlistment Process MAVNI Recruit Enlistment Procedure Recruiter preliminary interview MAVNI applicant EST (Enlistment Screening Test) administered 2+/2+ result waiverable to 2/2 with PSI– approval Corresponding OPI branch gives testing approval EST scores YES project qualifying AFQT SAVE eligible YES NO NO YES Immigration documents submitted Processing discontinued SAVE docs submitted to DHS YES Verified minimum 2-year residency (G845 form) NO YES NO Preliminary screen with respective security org YES ASVAB qualified <2/2 result NO MEPS scheduled NO Processing discontinued NO NO YES Physically qualified/ IFA * Previously, enlistees could ship to BCT upon passing NAIC if granted an ETP. Currently, most cases suggest that this is no longer an option, and enlistees must wait for favorable adjudication of SSBI before shipping to BCT. NO YES Exception to Policy* (ETP) NO NO YES Delayed Entry Program NACLC/ NAIC YES SSBI YES Ship to BCT SOURCES: Figure derived from a combination of sources, including discussions with SMEs, program documentation, and online sources; see Jointhemilitary.org (2016), United States Navy Recruiting Command (2015), and United States Department of Defense (undated) for additional details. NOTES: Green rectangles denote major enlistment phases, red rectangles denote discontinuation of the process, and blue and orange diamonds denote a processing step. The orange diamonds represent additional steps that are specific to MAVNI that non-MAVNI recruits typically do not undergo. PSI = Personnel Security Investigation; SAVE = Systematic Alien Verification for Entitlements. RAND RR1976-2.1 The process changed somewhat in FY 2017. There are no longer exceptions to policy, so MAVNI recruits are required to ship to BCT with an SSBI without exception, and all MAVNI recruits must also undergo and pass a counterintelligence (CI) interview before shipping to BCT. (Note that BCT is the first component of basic training; the other is advanced individual training). Finally, to accommodate the amount of time needed to meet these requirements and the backlog in SSBI investigations, MAVNI recruits are permitted to be in DEP for up to 730 days. The maximum time in DEP is otherwise 365 days. Figure 2.1 describes the process prior to these changes. The process shown for MAVNI recruits in Figure 2.1 is for those with CFL skills. HCPs undergo additional, or in some cases different, steps. For example, they must be screened by a board of medical peers that evaluates their medical skills, and, because TIWARI_PROD_00002383 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 12 The Military Accessions Vital to the National Interest (MAVNI) Program officers must be U.S. citizens, HCP MAVNI accessions enter the U.S. Army Reserve (USAR) as enlisted drilling reservists and are not allowed to practice medicine until their citizenship paperwork is cleared. Furthermore, MAVNI HCP accessions must also get malpractice insurance while their citizenship process is occurring. MAVNI HCP accessions have a six-year obligation if they join the USAR but a three-year obligation if they enter the RA. By way of contrast, MAVNI accessions with CFL skills have at least a four-year obligation. Unlike CFI-MAVNI accessions, HCP-MAVNI accessions are not required to be tested in a foreign language. Instead, they are tested for English-language proficiency. The entry process for MAVNI HCPs is markedly different than the process for most non-MAVNI officers in the health professions. HCPs entering as a MAVNI recruit must already be fully trained. In contrast, most non-MAVNI HCPs enter military service without having completed their training; indeed, they often enter the military to cover the cost of their medical training—medical school and/or residency (Hosek et al., 2016). There are a multitude of ways that a non-MAVNI HCP can enter the military, depending on the source and type of financial assistance they receive for their medical education. For example, many participate in the Health Professional Scholarship Program, and individuals can choose this program at entry into or during medical school. Others may participate in the Financial Assistance Program or attend the Uniformed Services University of the Health Sciences. The Financial Assistance Program is for those who have completed medical school and are in a civilian residency or plan to enter one. Each path for entry has different service obligations and may require service in the inactive reserves. This variation means that there is no obvious non-MAVNI HCP control group against which to compare MAVNI HCPs. Broad Themes Emerging from Interviews To add to our understanding of the enlistment process, we also spoke with SMEs, some of whom work with MAVNI recruits at Fort Knox. We also spoke with three recruiters with substantial experience recruiting MAVNI soldiers. Several broad themes emerged from these conversations. Those at Fort Knox characterized MAVNI recruits as high quality but requiring considerable attention. Non-MAVNI recruits also require considerable attention, but the SMEs indicated that this occurs within the recruiting battalions and that many of the steps are automated. In contrast, many of the additional steps required to process MAVNI recruits also require proactive attention by personnel at the U.S. Army Recruiting Command (USAREC). They also noted the increased screening costs associated with MAVNI recruits, caused by the requirement of an SSBI. SMEs expressed concern that the amount of time required for MAVNI recruits to enlist can create problems for the potential soldiers because their student visas can expire while they are in the DEP waiting to complete the enlistment process. However, SMEs also described MAVNI applicants as organized, able to keep track of forms and paperwork, and savvy TIWARI_PROD_00002384 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 13 users of social media to learn about the MAVNI program and to communicate with other potential recruits. Our conversations with recruiters touched on similar themes. The recruiters with whom we spoke had built up considerable expertise on the MAVNI program. However, this expertise required a substantial time commitment on their parts. They described their efforts as worthwhile because, after they gained an understanding of the paperwork and process, they found ways to locate and recruit many MAVNI soldiers—far more recruits than they would have enlisted if they had focused their efforts on typical recruits. The recruiters also utilized social media as a way to communicate with potential MAVNI recruits and to share the experiences of new MAVNI recruits with other potential recruits. Recruiters indicated that joining the reserves was especially attractive to many potential MAVNI recruits because it allowed them to continue their education while in service and, once they obtained citizenship, to pursue civilian employment. Finally, recruiters noted that not all MAVNI applicants qualified for service. Some exceeded the weight standards, some failed to qualify based on AFQT scores, and a few even failed to qualify in their language. But recruiters expressed a sense that most MAVNI enlistees were qualified. The recruiters with whom we spoke had developed systems to help keep track of all required paperwork and processes. They emphasized the need to pay close attention to all of the details during the long recruitment process. The recruiters indicated that they have helped other recruiters get up to speed on MAVNI recruiting and that they have worked with MAVNI recruits who live far from their recruiting station but wished to enlist under the program and wished to work with an experienced recruiter. Recruiters acknowledged issues with determining which MAVNI recruits were cleared to ship. The system involves setting a BCT ship date and then waiting to see whether all processes are completed by the ship date. The existing systems seem to assume that the recruit is ready to ship as scheduled, unless updated information (for example, a SSBI completion delay) is entered. Recruiters mentioned the SSBI backlog in particular, and they were not notified about whether and when the SSBI had been successfully adjudicated. Consequently, recruiters were unable to keep applicants up to date about whether the ship date was still valid or needed to be postponed until the SSBI had been completed. They also noted that, in some cases, visa-related issues arose with family members of MAVNI enlistees in the DEP and in BCT if the enlistee was not in school but had not yet received citizenship.3 Finally, across the board, one theme in the course of our conversations with recruiters and other SMEs was whether and how MAVNI skills were used. They expressed concerns that the MAVNI enlistees did not use their language skills and/or that the program did not enlist the optimal mix of languages. Recruiters also mentioned an issue with MOS or job match; many MAVNI recruits have considerable civilian educa3 See U.S Army (2012). TIWARI_PROD_00002385 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 14 The Military Accessions Vital to the National Interest (MAVNI) Program tion, often with advanced degrees, but they are not assigned to jobs using that education, sometimes because of restrictions on receiving a security clearance. Learning About the MAVNI Enlistment Process One theme emerging from the interviews was that potential MAVNI enlistees were sophisticated users of social media, both to learn about the program and to communicate with other potential recruits. We therefore examined relevant search data to learn more about how potential MAVNI recruits might learn about the program, their level of interest in the program, and the locations of potential MAVNI enlistees. We show results for two types of data: 1. Google analytics data to determine the general level of interest in the MAVNI program among all who search for information on the Internet 2. information from page views from the GoArmy.com website to find out more about the search patterns and geographic locations of those who are steered to the Army’s main recruiting website. There are other social media sources of information about MAVNI; for example, the largest MAVNI Facebook group is a closed group that had nearly 17,000 members as of November 3, 2016.4 Google Search Data on MAVNI-Related Topics Google analytics data are a form of big data, generally defined as very large and unstructured data sets usually accessed from the Internet. Search data differ somewhat from other types of big data; search engines typically gather, anonymize, and aggregate the data. Search data have been found to be correlated with outcomes of interest in a variety of situations, including flu outbreaks and unemployment; in the military context, they have been found to be correlated with the proportion of Army enlistees who are of high quality (Ginsberg et al., 2009; Ettredge et al., 2005; Jahedi et al., 2016). Because search data can be gathered nearly in real time, they have the potential to relay key information more quickly than surveys or official statistics. Because individuals search for information in which they have interest but do not possess, search data have the potential to provide insights into preferences and interests that otherwise might be difficult or expensive to obtain. Of course, these data have drawbacks. Not everyone utilizes the Internet to find information, and Google is not the only search platform. 4 Other websites relevant to Army recruiting have larger audiences. For example, the main USAREC Facebook page had over 175,000 “likes” on the same date. These two metrics are not exactly comparable because joining a group requires entering basic information and receiving approval from a group administrator, while “liking” requires only a click. This suggests considerable interest in the MAVNI program. TIWARI_PROD_00002386 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 15 However, for our population of interest, Google data appear quite relevant: 98 percent of those 18 to 29 years old search for information on the Internet, and 83 percent of U.S. searchers most often use Google (Pew Research Center, 2013; Purcell et al., 2012). Google provides several tools to examine aggregated search data. We used the Google Trends tool. Figure 2.2 shows the overall level of searches over time for the term “MAVNI.” Google Trends data are rescaled so that the maximum level of relative searches is set to 100. A downward trend would imply that MAVNI searches made up a smaller proportion of total searches over time. Here, we see a generally upward trend in MAVNI-related searches, indicating that the relative level (and the absolute level) of searches was increasing over the time period included; overall Google searches were also increasing during this time period. MAVNI-related searches peaked at two different times—once as the program reopened after the two-year hiatus and again as the program prepared to close in 2016. Figure 2.3 compares the prevalence of MAVNI searches with the prevalence of two related searches: the phrase “Army linguist,” which might be viewed as a substitute search for noncitizens seeking to enlist, and the generic phrase “join Army.” The overall level of “join Army” searches is, as would be expected, much higher than the level of MAVNI searches. However, MAVNI searches are far more common than searches related to Army linguists. This suggests that there is a substantial interest in, or awareness of, the MAVNI program, and that Google-based searches may provide important information to potential MAVNI enlistees. Figure 2.2 Google Trends: Searches for “MAVNI” Over Time 100 90 “MAVNI” Program reopened Program closed for FY Relative search intensity 80 70 60 50 40 30 20 10 9/ 10 / 12 201 1 /1 0/ 2 3/ 01 10 1 /2 6/ 01 10 2 /2 9/ 01 10 2 / 12 201 2 /1 0/ 2 3/ 01 10 2 /2 6/ 01 10 3 /2 9/ 01 10 3 / 12 201 3 /1 0/ 2 0 3/ 10 13 /2 6/ 01 10 4 /2 9/ 01 10 4 / 12 201 4 /1 0/ 3/ 201 10 4 /2 6/ 01 10 5 /2 9/ 01 10 5 / 12 201 5 /1 0/ 3/ 201 10 5 /2 6/ 01 10 6 /2 01 6 0 SOURCE: Authors’ calculations based on Google Trends data (www.google.com/trends); searches were carried out within the United States. RAND RR1976-2.2 TIWARI_PROD_00002387 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 16 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 2.3 Google Trends: MAVNI-Relevant Searches Over Time 100 Relative search intensity 90 80 70 60 50 40 “MAVNI” “Army linguist” “Join Army” 30 20 10 0 9/17/2011 9/17/2012 9/17/2013 9/17/2014 9/17/2015 SOURCE: Authors’ calculations based on Google Trends data; searches were carried out within the United States. RAND RR1976-2.3 GoArmy.com MAVNI-Relevant Page Views Google-based searches for MAVNI-related information generally steer the searches to GoArmy.com, a major source of information for potential (and current) soldiers on various aspects of Army service. The site includes several MAVNI-specific pages, and these are among the most prominent results of a Google search on the term “MAVNI.” However, visitors to the GoArmy website initially seeking other information may also access the MAVNI-specific pages. GoArmy.com produces a substantial amount of analytical information that allows us to determine how many people view the MAVNI pages on the GoArmy website, how long they spend on those pages, and their geographic locations.5 We found that MAVNI-related visits to GoArmy.com made up a small fraction of all visits during the time period covered by our data; somewhat less than 1 percent of all visits included a MAVNI-related page. Because both current and prospective soldiers use GoArmy. com, we cannot determine which visits concerned enlistment questions and so cannot estimate the proportion of enlistment-related visits that included a MAVNI-related page. In general, the data indicate a substantial level of engagement on the part of those searching for MAVNI-relevant information. Table 2.2 lists three measures of engage5 We gratefully acknowledge the assistance of the Army Marketing and Research Group in accessing the limited subset of the GoArmy.com information used in our analyses. TIWARI_PROD_00002388 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 17 Table 2.2 Measures of Engagement, GoArmy.com MAVNI-Related Sessions All Sessions 6.7 3.7 Bounce rate (proportion of users who arrive at the website but leave without interacting or moving to any other page on the website) 46% 53% Percentage of sessions resulting in Thank You page (providing contact information to a recruiter) 2% 0.5% Measure of Engagement Average session duration, in minutes NOTE: RAND Arroyo Center analysis based on data from GoArmy.com. Data include August 2013 through July 2016. During this period, there were more than 50 million sessions on GoArmy.com; over 350,000 of them included a visit to a MAVNI page. ment: average session duration, bounce rate, and percentage of sessions resulting in a Thank You page. Bounce rate refers to the proportion of users who arrive at the website but leave without interacting or moving to any other page on the website, and ending a session with a Thank You page means that the individual was interested enough in the Army to leave contact information for a recruiter. The table compares these metrics for MAVNI-related sessions with the metrics for all sessions. In each case, those who viewed at least one MAVNI-related page could be viewed as being more engaged than the average website visitor. In particular, visitors who view MAVNI-related pages are about four times more likely than the average visitor to provide information so that a recruiter may contact them. We can also make use of website analytics to characterize the proportion of visits from various areas of the country, thereby obtaining the general geographic location of the searcher. The data are reported by designated market area (DMA); the United States is divided into 210 unique DMAs, a geographic unit used for media marketing purposes. Most DMAs include a large urban area, but some do not. In general, large metropolitan areas have more visitors to GoArmy.com (and other websites) than areas with sparser populations. In the case of MAVNI-related pages, visitors are quite concentrated in a few metropolitan areas, with many coming from large urban areas. Figure 2.4 indicates the geographic area of new users (those who have never visited the site before), and of total number of sessions, by DMA. The figure presents separate tallies for MAVNI-related pages and other pages. The figure includes only the 20 DMAs with the highest numbers of sessions, and the DMAs are sorted by total number of sessions. Figure 2.4 first demonstrates that new users and sessions are distributed in a similar manner, suggesting that these two measures include similar information. The 20 DMAs in Figure 2.4 include about 45 percent of the U.S. population; about half of all GoArmy.com sessions originated from one of these DMAs. Thus, searches (and TIWARI_PROD_00002389 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 18 The Military Accessions Vital to the National Interest (MAVNI) Program 14 Percentage of MAVNI new users Percentage of MAVNI sessions Percentage of new users Percentage of sessions 12 10 8 6 4 2 Y W Lo as s A ork hi ng nge to le n, s D . Ch C. i ca D go al A la s/F tla Sa or nt n tW a Fr an o ci H rth sc ou o/ O Phi sto ak la n d la nd ep M /S hia ia an m Jo i/F se or t L Bo st a O o rla Se ude n rd nd at al o/ tle e D ay /Tac to om na a Be Ch ach ar lo t Ph te oe ni Ta m D x pa et ro Sa M /Sai cr i nt De it am nn n P en eap ete ver to ol rs is b /S to /Sa urg ck in to t P n/ a M ul od es to 0 N ew Percentage of users or sessions in DMA Figure 2.4 New Users and Sessions on GoArmy.com, MAVNI Versus Others, by DMA DMA SOURCE: RAND Arroyo Center analysis based on data from GoArmy.com. Data include August 2013 through July 2016. During this period, there were more than 50 million sessions on GoArmy.com; more than 350,000 of them included a visit to a MAVNI page. RAND RR1976-2.4 population) are concentrated, with 10 percent of DMAs responsible for about half of all visits to GoArmy.com. Figure 2.4 also indicates that MAVNI-related searches are even more concentrated by DMA than others—about 64 percent of all GoArmy.com MAVNI-related sessions originated from these 20 DMAs. Indeed, over one-third of all MAVNI-related sessions originated from New York; Los Angeles; Washington, D.C.; Dallas/Fort Worth; or San Francisco/Oakland/San Jose. Consistent with these results, nearly 40 percent of MAVNI contracts originated from one of a handful of DMAs. In general, these results match the pattern of overall immigration; immigrants tend to be concentrated in large urban areas and also, to some extent, in the Southwest (Pew Research Center 2015). This suggests that potential MAVNI recruits may also learn about the program from local or in-person networks, in addition to social media. The geographic distribution of MAVNI recruits could also reflect the locations of recruiters who specialize in MAVNI enlistments. Of course, recruiters who find it productive to focus on MAVNI recruits are also likely to be assigned to locations with larger enclaves of immigrants and noncitizens. TIWARI_PROD_00002390 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 19 In summary, the information we examined from Internet search data, as well as the page views from the internal metrics associated with the GoArmy.com webpage, indicate a substantial level of online interest and engagement among potential MAVNI enlistees. Also, potential MAVNI enlistees appear to be highly concentrated in a few large urban areas. These results are consistent with the information provided by recruiters and other SMEs. Descriptive Statistics of the MAVNI Program We conclude the chapter with some descriptive statistics of Army MAVNI contracts since FY 2013, shown in Tables 2.3–2.7. Table 2.3 shows tabulations on the size of the HCP-MAVNI program, while Tables 2.4–2.7 show tabulations for the CFL-MAVNI program. Descriptive statistics of the demographic characteristics of MAVNI accessions are shown in Chapter Three. Table 2.3 shows the percentage of MAVNI recruits in the HCP-MAVNI program in FY 2014 and 2015. In 2014, about 10 percent of MAVNI contracts were HCPs, falling to just over 1 percent of contracts in 2015. The drop could be due to changes in the health professions permitted to enter under MAVNI. One theme that emerged from the interviews we conducted is that the training requirements and MAVNI eligibility requirements are better suited for the licensing requirements of some health professions, such as nursing and dentistry. Since 2015, the HCP-MAVNI program continues to be quite small (not shown). Table 2.4 shows the breakdown of MAVNI CFL contracts by visa category (we have aggregated across categories to summarize the difference in student-related and work-related visas). In each year, the vast majority of those signing MAVNI contracts entered the United States on a student visa. Among the student visas held by MAVNI enlistees, F-1 is by far the most common visa. F-1 student visas are granted to international students attending an academic program at a U.S. college or university.6 Table 2.3 MAVNI HCP Contracts as Percentage of All MAVNI Contracts FY HCP Percentage FY 2014 9.9% FY 2015 1.2% SOURCE: Authors’ tabulations using MAVNI data provided by sponsor. 6 F-2 visas are granted to spouses and children of those holding F-1 visas; a small number of MAVNI enlistees hold F-2 visas. We include these visa holders in the “student” category in Table 2.4. MAVNI status codes are defined in Table A.1 in Appendix A. TIWARI_PROD_00002391 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 20 The Military Accessions Vital to the National Interest (MAVNI) Program Table 2.4 Visa Category Breakdown of MAVNI CFL Contracts in FY 2014 and 2015 Visa Category Percentage of MAVNI CFL Contracts Student 82.0% Worker 4.8% Other 13.1% SOURCE: Authors’ tabulations using MAVNI data provided by sponsor. NOTES: “Student” visas include all F visas; “worker” visas include H visas; Other visas include A, E, J, O, and P visas, as well as asylees, refugees, those with temporary protected status, and DACAs. Data provided by sponsor, based on MAVNI contracts for FY 2014–FY 2015. Visa status is not available in Total Army Personnel Database (TAPDB) or RA Analyst file. See Appendix A for more information on visa categories. Percentages do not total 100 because of rounding. Table 2.5 shows that the percentage of contracts that enter the RA (versus the USAR) declined over time, from 100 percent among FY 2013 contracts to 41.6 percent among FY 2016 contracts. The MAVNI program was only expanded to include the RCs in FY 2014. As mentioned earlier, OSD also extended the MAVNI program in FY 2014 to include those with DACA status. As Table 2.4 also shows, beginning in FY 2015, about one in eight MAVNI contracts are DACA contracts. Since FY 2013, MAVNI language capability has been concentrated in about seven languages, as shown in Table 2.6, though the percentage of contracts attributable to a single language has changed over time. Korean is the most common language, followed by Mandarin, and both have grown over time. Nepalese and Swahili are also common languages, though they have declined over time, as have French and Tagalog. The percentage with capability in Hindi has varied over time but was just over 10 percent of contracts in FY 2016. As shown in Figure A.1 in Appendix A, the language list changed dramatically in 2017. Notably, Korean and Mandarin were removed from the list. Table 2.7 shows the distribution of MAVNI contracts across Army career management fields (CMFs), though the first row shows the percentage of contracts placed in an undesignated occupational code, 09U. In 2016, over a third of contracts had an undesignated occupation. Because the SSBI must be adjudicated favorably and completion of the investigation process can be longer than expected, some MAVNI recruits TIWARI_PROD_00002392 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. The MAVNI Program and the Enlistment Process 21 Table 2.5 Percentage of MAVNI CFL Contracts That Are Regular Army and DACA, by FY FY Regular Army 2013 100 DACA 0 2014 91.5 0 2015 62.6 12.9 2016 41.6 11.2 SOURCE: Authors’ tabulations using MAVNI data provided by sponsor. Table 2.6 Percentage Distribution of Language Capability Among MAVNI CFL Contracts, by Fiscal Year Language 2013 2014 2015 2016 Korean 13.9 14.8 28.8 27.6 Mandarin 12.0 15.7 20.4 24.1 Hindi 11.4 13.1 5.9 10.9 Nepalese 13.7 12.3 14.9 8.4 Swahili 13.0 9.1 3.6 3.2 French 5.9 5.7 4.5 3.7 Tagalog 4.9 7.6 4.0 2.3 25.2 21.7 17.9 19.8 Other SOURCE: Authors’ tabulations using MAVNI data provided by sponsor. have experienced delays in their shipping date—e.g., their investigation was not completed before they were required to ship to BCT. This issue caused empty training seats in the occupations to which these MAVNI recruits were assigned. To address this issue, since 2016, affected MAVNI recruits have been given the undesignated occupation 09U. These contracts will have an occupation assigned at a later date. Between FY 2013 and FY 2015, CMFs 68 and 92, medical and supply and services, respectively, comprised the majority of CFL MAVNI contracts. As the table shows, other common career management fields in these years were mechanical maintenance, engineering, aviation, transportation, and financial management. TIWARI_PROD_00002393 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 22 The Military Accessions Vital to the National Interest (MAVNI) Program Table 2.7 Percentage Distribution of Army Career Management Field Among MAVNI CFL Contracts, by FY Army Career Management Field 2013 2014 2015 2016 0.2 0.4 1.5 36.3 CMF 68: Medical 44.3 35.5 31.6 13.0 CMF 92: Supply and Services 20.4 32.5 25.4 18.8 CMF 91: Mechanical Maintenance 6.8 11.2 11.4 9.2 CMF 12: Engineer Corps 3.7 2.6 5.2 6.2 CMF 15: Aviation 8.1 3.0 4.8 1.0 CMF 88: Transportation 1.3 1.8 5.4 9.9 CMF 36: Financial Management 8.3 6.4 6.4 1.1 Other CMF 6.9 6.6 8.3 4.5 MOS 09U: Undesignated SOURCE: Authors’ tabulations using MAVNI data provided by sponsor. Summary MAVNI enlistees must meet additional requirements not faced by other enlistees; some of these requirements, such as relatively high AFQT scores, ensure that MAVNI enlistees possess characteristics that are associated with high levels of performance. The MAVNI enlistment process is longer and more complex than that of other recruits; in particular, all MAVNI recruits must pass a series of security-related requirements. Nonetheless, data from Internet searches and from GoArmy.com internal analytics suggest that there is substantial interest in the program; searchers who visit MAVNIrelated pages on the GoArmy website appear quite engaged in the material. Also, many MAVNI searchers come from a few large urban areas. This suggests that MAVNI enlistees may also learn about the program from local or in-person networks (rather than solely or primarily from online sources). TIWARI_PROD_00002394 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER THREE Relative Performance of MAVNI Recruits This chapter compares the performance of MAVNI recruits and non-MAVNI recruits using administrative Army recruiting and personnel data. As we discuss in more detail later in the chapter, we measure performance in terms of DEP attrition, early and firstterm attrition, first-term reenlistment, and promotion speed. Our analysis of performance focuses on MAVNI recruits with CFL skills rather than HCPs. The HCP portion of the MAVNI program is relatively small (Table 2.4), and it is difficult to define a comparison group because the entry process and the associated initial service obligation of MAVNI health professionals differ from those of most non-MAVNI health professionals. Performance might differ between MAVNI and non-MAVNI recruits for two reasons. First, MAVNI recruits have different entry characteristics, reflecting the different screening and qualification requirements under the MAVNI program. These different characteristics can lead to different performance outcomes insofar as those characteristics are correlated with performance. For example, Buddin (2005) shows that first-term attrition is higher among recruits who have a General Education Development (GED) certificate and who enter the Army in poor physical health. Because a MAVNI recruit must have at least a high-school diploma, we would expect MAVNI recruits to exhibit lower attrition rates. The characteristics of MAVNI recruits may also differ because their propensity to join the military and the characteristics that predict propensity to join the military may differ from non-MAVNI recruits. Second, MAVNI recruits may have a stronger incentive to perform, given that MAVNI recruits who fail to complete their service obligation will have their U.S. citizenship revoked. Since it can typically take many years to become a U.S. citizen through standard channels, the immediate citizenship provided to MAVNI recruits can be a strong incentive to complete the initial service obligation. Furthermore, the citizenship incentive could induce individuals who qualify for MAVNI, but who otherwise have a relatively low taste for military service, to join the Army. Consequently, once they finish the service obligation associated with MAVNI, they could be more likely to leave and not reenlist. Other non-MAVNI recruits also face incentives to complete their service obligation. For example, enlistees who qualify for an enlistment bonus receive it in install- 23 TIWARI_PROD_00002395 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 24 The Military Accessions Vital to the National Interest (MAVNI) Program ments over their first term, if the bonus exceeds a minimum threshold. Those who terminate their military service before the end of their term are ineligible to receive outstanding bonus payments and may need to repay the bonus they have received, on a prorated basis. Those who enlist with a bonus may have lower taste for service than other recruits but are motivated to join the Army because of the bonus. Thus, they may also be more likely to separate after completion of their service obligation. Asch et al. (2010) investigated this issue and examined the effect of enlistment bonuses on attrition after correcting for the selection effect on tastes provided by the bonus program in the Army. That report found that enlistment bonuses reduced first-term Army attrition, but the effect was modest—only a 5-percent reduction in the attrition rate. This suggests that the enlistment bonus effect on attrition and, potentially, on reenlistment for non-MAVNI recruits is not particularly strong. We begin the chapter with an overview of the data we used to compare performance of MAVNI and non-MAVNI recruits. We then define the non-MAVNI comparison groups and the metrics of performance we use from the administrative data and describe our methodology. We then present our results. Data Two Army administrative personnel files make up our primary sources of performance data. TAPDB is a monthly file that includes a variety of variables describing the trajectory of each soldier’s career; examples include MOS, months of service, reason for departing the Army, and indications of reenlistment. We also use information from the RA Analyst file. RA Analyst includes many measures at the point of accession; examples include age, marital status, test scores at accession, bonus amounts, and indications of waivers. We include all MAVNI enlistees (and all non-MAVNI enlistees) who entered the Army during the period from October 2008 through December 2015. Beginning in October 2008 allowed us to capture the earliest enlistees into the MAVNI program, and December 2015 data were the most recent available when this project began. For our performance measures that require 48 months of observation, we focus on a subset made up of those MAVNI (and non-MAVNI) enlistees who signed 48-month contracts between October 2008 and September 2010. Definition of Comparison Groups To compare the performance of MAVNI recruits, we need to identify one or more relevant comparison groups. The choice of group depends on how the Army would meet the requirement that the MAVNI program fulfills in the absence of the program. We defined three alternative comparison groups, based on three different ways that TIWARI_PROD_00002396 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 25 MAVNI contributes to Army recruiting and, consequently, three different ways the Army could replace MAVNI recruits in the absence of the MAVNI program. At the most basic level, the MAVNI CFL program helps to fulfill the Army’s foreign language and cultural needs. In the absence of MAVNI, the Army would be required to enlist non-MAVNI recruits with these skills. Consequently, the relevant comparison group is the pool of non-MAVNI recruits with foreign language and cultural skills. Because the number of recruits who have MAVNI language skills is limited and identifying them in our data is challenging,1 we approximated this pool as the group of recruits who are noncitizens and not in MAVNI. Thus, for our first definition, we use noncitizen non-MAVNI recruits to compare the performance of MAVNI recruits. Enlistment also decreases the time to citizenship for this group; therefore, they may face some of the same incentives as MAVNI recruits. We defined the second comparison group as the pool of high-quality non-MAVNI recruits. This definition recognizes that in addition to providing language and cultural capability, the MAVNI program also provides a supply of high-quality recruits. In the absence of MAVNI, the Army would need to recruit non-MAVNI high-quality recruits as replacements. Our third group is defined as all non-MAVNI recruits. This definition recognizes that the MAVNI program assists the Army in meeting its annual contract and accession mission. In the absence of the MAVNI program, the Army would need to enlist non-MAVNI recruits to replace MAVNI recruits and meet its overall recruiting mission. We recognize that other definitions could be considered. However, we found that the results were generally consistent across the three definitions we used. This suggests that our results are robust to alternative definitions of the comparison group. Measurements of Performance We measure performance in terms of attrition behavior and promotion speed and in terms of characteristics that are thought to be or have been shown to be correlated with performance, such as aptitude scores. Attrition captures individuals’ decisions to participate and stay in the Army, though not their performance, given that they choose to stay. In contrast, promotion speed captures performance while in service, insofar as those who are promoted more quickly are individuals who have demonstrated that they are better matched to the military and perform better. Past studies have used informa1 Few Army occupations require language skills. Exceptions are cryptologic linguists (35P) and (until recently) interpreter/translator (09L). Others in the Army may possess language skills, but the only reliable measure of this is receipt of the Foreign Language Proficiency Bonus. Only soldiers who have taken and passed a language exam have this pay. Many soldiers do not take this test immediately upon accession, so it will not show in the accession data. TIWARI_PROD_00002397 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 26 The Military Accessions Vital to the National Interest (MAVNI) Program tion on promotion speed together with data on education and aptitude to develop a measure of personnel quality that reflects performance in service (Hosek and Mattock, 2003; Asch, Romley, and Totten, 2005). With respect to attrition, we specifically consider DEP attrition and attrition during the first enlistment term, given accession. DEP attrition is measured as the percentage of individuals who sign an enlistment contract but fail to actually access and go to basic training. We measure attrition, given accession, at six-month intervals. Because non-MAVNI enlistees can choose enlistment terms that are less than 48 months, while MAVNI are required to enlist for at least 48 months, comparing 48-month attrition rates between MAVNI enlistees and all non-MAVNI enlistees could lead to confusion between successful completion of a shorter contract and attrition. To avoid this problem, our analysis of first-term attrition is conducted for only those who signed a fouryear enlistment contract. As shown in Figure 3.1, 88 percent of MAVNI contracts are for a 48-month obligation, while 12.4 percent of signed contracts are for either a 60- or 72-month obligation. For all non-MAVNI contracts, 43 percent had an obligation of 48 months or more. In addition, to ensure that enlistees have sufficient time to complete DEP as well as a 48-month obligation, our analysis of attrition, promotion, and reenlistment only uses data on contracts signed between October 2008 and September 2010 (the beginning of FY 2011), given that our data end in December 2015. We denote those who enlisted before FY 2011 as “wave 1” enlistments and those who enlisted after that as “wave 2” enlistments. Thus, our tabulations and analysis of DEP and accession, which Figure 3.1 Distribution of Contracts Across Enlistment Term 100 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality Percentage of contracts 80 60 40 20 0 24 36 48 60 72 Enlistment term (months) SOURCE: Authors’ tabulations using Army RA Analyst data. RAND RR1976-3.1 TIWARI_PROD_00002398 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 27 require data on the DEP period and the initial time in service, include data from both wave 1 and wave 2. However, our analyses of promotion, attrition, and reenlistment analysis only use wave 1 data so that we can observe the period in DEP as well as the full 48-month contract period, thus allowing 48 months for personnel to attrite, receive promotions, and make reenlistment decisions. Also, in measuring attrition, we define attrition strictly as “failure to adapt” attrition; other reasons for attrition include disability separations and enlisted conversions to the officer corps or to the RC. Army personnel data allow us to distinguish between these different forms of attrition. We measure promotion speed to E-4 and to E-5. In the case of E-4, we measure the percentage of accessions that did not reach E-4 by 48 months and the percentage that reached E-4 by 19 months, by 24 months, and by 48 months. In the case of E-5, we measure the percentage of accessions that did not reach E-4 by 60 months and the percentage that reached E-5 by 24 months, by 36 months, by 48 months, and by 60 months. In addition to promotion speed and attrition, we also compare the characteristics of MAVNI accessions with non-MAVNI accessions. We consider demographic characteristics, aptitude, and aspects of the enlistment contract, including whether the recruit received an accession waiver. While all recruits are required to meet enlistment standards, recruits may receive a waiver in some cases, with approval of the chain of command.2 The evidence on the relationship between performance and waivers suggests that Army recruits receiving an enlistment waiver are more likely to attrite in their first term (Asch et al., 2010) but are not the most risky recruits in terms of performance. For example, Malone et al. (2011) finds that those receiving waivers do not have higher attrition rates than those without a high-school diploma. For demographic characteristics, we present comparisons of MAVNI and non-MAVNI recruits in terms of age, gender, aptitude scores, education, race, ethnicity, marital status, and dependents. In part, these comparisons provide information on how MAVNI and nonMAVNI recruits differ. Insofar as prior research shows that demographic characteristics are associated with attrition behavior, they also provide information on factors correlated with performance. For example, Buddin (2005) finds that attrition rates are higher among female recruits, those who enter with education less than a high-school diploma, those who are younger at entry or overweight, those who are married or who are single with children, and those who are non-Hispanic white. 2 See Chapter Two for more information on enlistment standards. TIWARI_PROD_00002399 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 28 The Military Accessions Vital to the National Interest (MAVNI) Program Approach The approach we take is first to tabulate the mean values of the performance metrics for MAVNI and non-MAVNI recruits, without controlling for observed factors (other than MAVNI program participation) that could explain differences in the performance metrics. We then decompose several of the metrics into the portion caused by observed factors and the portion caused by other factors. More specifically, observed differences in performance measures between MAVNI and non-MAVNI recruits can be attributed to • differences in the job and individual characteristics that influence the performance metric • differences in other factors that affect performance, holding characteristics constant. As mentioned, unlike non-MAVNI recruits, citizenship is revoked for MAVNI recruits who fail to complete their service obligation, giving a strong incentive to satisfactorily complete service. Thus, other factors could include the incentive provided by accelerated citizenship, as well as selection on tastes for service induced by the citizenship opportunity. Oaxaca (1973) and Blinder (1973) developed a method to decompose differences in outcomes between their explainable and structural components. The method has been widely used in the literature—for example, to decompose differences in male and female pay into explainable and structural components—and the method has been refined over time.3 We use the Blinder-Oaxaca method to decompose the observed differences in performance metrics for MAVNI and non-MAVNI recruits.4 The Blinder-Oaxaca method requires a regression analysis that provides estimates of the relationship between each observed characteristic and the performance metric of interest. In our analysis, we considered three performance metrics: accession among contracts; cumulative retention (non-attrition) to 48 months; and reenlistment, given completion of term of service. We then used these regressions to decompose the observed difference in the mean performance metric between MAVNI and non-MAVNI recruits into two parts: 1. the part attributable to mean differences in observed characteristics 2. the part attributable to factors other than observed characteristics. 3 This literature uses different terminology for the two components. Some studies refer to them as the explainable and unexplainable components, while others refer to them as the observable and structural components. 4 Appendix B presents additional detail on the methodology we used in this analysis, drawing on Jann (2008) and Fortin, Lemieux, and Firpo (2011). TIWARI_PROD_00002400 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 29 The second part shows the difference in metrics among recruits with the same observed characteristics. In performing the analysis, we note that we are unable to give a causal interpretation of the effect of MAVNI status on performance because of the potential influence of self-selection and endogeneity. Individuals at a given career milestone are a self-selected group—i.e., those who are eligible to reenlist are a group who self-selected to complete their first term. This selectivity effect could differ for MAVNI and nonMAVNI recruits because of unobserved characteristics (such as motivation). Such differences result in biased regression coefficient estimates of the causal effect of MAVNI status on a specific performance metric. Finally, characteristics that we do not observe might be correlated with observed characteristics, and the correlation might differ for MAVNI and non-MAVNI recruits. This could also bias the coefficient estimates of the causal effects of observed characteristics on an outcome. Tabular Results Demographic and Job Characteristics Table 3.1 shows tabulations of the demographic characteristics of accessions under the MAVNI program and the three comparison groups for those who signed a four-year enlistment contract. We tested whether differences in characteristics between MAVNI and the comparison groups were statistically significant, and we found that they were significant at the 1-percent level for every variable shown, relative to all three groups. That said, the magnitude of the differences, though statistically significant, is not necessarily large in every case. In general, the tabulations show that MAVNI recruits are more likely to have characteristics associated with better performance—and, specifically, lower attrition— than the comparison groups. MAVNI recruits are older on average, 26.1 years of age versus about 21 years for all non-MAVNI recruits or other high-quality non-MAVNI recruits. Older recruits are less likely to attrite (Buddin, 2005). We find that nearly 60 percent of MAVNI recruits are college graduates at entry, compared with 7.8 percent of high-quality non-MAVNI recruits and 24.4 percent of noncitizen non-MAVNI recruits. MAVNI recruits also have higher AFQT percentile scores on average, 74.2 versus 71.9 for high-quality non-MAVNI recruits and 63.8 and 58.2 for all nonMAVNI recruits and noncitizen non-MAVNI recruits, respectively. Not surprisingly, then, MAVNI recruits have more years of education on average (15.4), and all MAVNI recruits are high quality. MAVNI recruits are also less likely to be considered overweight (i.e., have a BMI over 25) than any of the comparison groups. On the other hand, MAVNI recruits are more likely to be married and have dependents, and married recruits also are more likely to attrite. We also find that MAVNI accessions spent TIWARI_PROD_00002401 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 30 The Military Accessions Vital to the National Interest (MAVNI) Program Table 3.1 Demographic and Job Characteristics of Army Accessions with Four-Year Enlistment Term of Service, Wave 1 and Wave 2, MAVNI Versus Three Comparison Groups MAVNI Accessions All Non-MAVNI Accessions Non-MAVNI Noncitizen Accessions Non-MAVNI High-Quality Accessions 26.1 21.0 25.7 21.1 Male 80.5% 78.1% 71.4% 82.3% Married 20.0% 14.8% 42.3% 14.8% Has dependents 11.1% 7.4% 20.5% 7.3% White, non-Hispanic 6.8% 58.1% 14.9% 66.5% Black, non-Hispanic 19.8% 23.8% 32.9% 17.0% Hispanic 0.9% 13.0% 22.0% 11.6% Asian 72.0% 4.9% 29.8% 4.7% College graduate 57.4% 6.4% 24.4% 7.8% Years of education 15.4 12.4 13.4 12.6 AFQT score 74.2 63.8 58.2 71.9 High quality 100% 71.3% 65% 100% 24.1 24.6 24.6 24.8 BMI above 25 (overweight) 38.4% 44.0% 43.7% 45.8% Enlist in critical MOS 15.7% 36.0% 26.6% 43.4% Received an enlistment bonus 0.4% 25.3% 11.8% 30.5% Wave 1 22.5% 52.0% 55.2% 51.9% Months in DEP, mean 4.71 3.92 3.85 3.45 Months in DEP, standard deviation 1.62 2.73 2.70 2.05 Characteristic Age (number of years)a Body mass index (BMI) a Because data on age are only available for those who accessed, this variable is average age at accession, given that an accession occurred. SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. more time in DEP—an average of 4.71 months, compared with 3.92 months for all non-MAVNI accessions. While CFL-MAVNI recruits are required to have a critical language skill, we find that they are less likely to enlist in an occupation that is deemed critical by the TIWARI_PROD_00002402 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 31 Army than non-MAVNI recruits.5 Specifically, we find that 15.7 percent of MAVNI accessions enter a critical occupation, compared with 43.3 percent of high-quality nonMAVNI recruits, 36 percent of all non-MAVNI recruits, and 26.6 percent of noncitizen non-MAVNI recruits. MAVNI recruits are not allowed to enter occupations requiring a security clearance, and this restriction can serve to lower the percentage of MAVNI recruits that enter critical occupations, since critical occupations are more likely to require a security clearance. Among those RA occupations requiring a clearance is the cryptologic linguist occupation, the only occupation that currently requires language skills to be considered MOS-qualified. The results suggest that MAVNI recruits meet general Army occupation requirements but not requirements in critical occupations, including the linguist occupation. Reinforcing this conclusion is the finding that virtually no MAVNI recruit received an enlistment bonus, while 30.5 percent of high-quality non-MAVNI recruits did so. Enlistment bonuses are generally used to channel recruits into hard-to-fill and critical occupations. Table 3.2 compares the likelihood that a contract received a DEP waiver and the likelihood that an accession received an accession waiver for MAVNI recruits and for the three comparison groups. MAVNI recruits are not authorized to receive a drug or alcohol waiver or a conduct waiver. Not surprisingly, then, we find that no MAVNI recruit received these categories of waiver. A small percentage of non-MAVNI contracts received a DEP waiver, and a small percentage of non-MAVNI accessions received accession waivers in these categories. MAVNI contracts are also less likely to receive a DEP medical waiver—7.5 percent, compared with 9.1 percent for all non-MAVNI contracts.6 While accession waivers are less common among accessions than DEP waivers are among contracts, the pattern across accession categories is generally similar to that of DEP waivers. Again, medical waivers are the most common. Overall, MAVNI contracts were less likely to require a DEP waiver than the comparison groups. At accession, MAVNI accessions were slightly less likely than all non-MAVNI accessions and high-quality non-MANVI accessions to receive a waiver and were slightly more likely than noncitizen non-MAVNI accessions to receive a waiver. DEP and First-Term Attrition We next consider attrition, both during DEP and during the first enlistment term. To compute DEP attrition, we limit the data to contracts signed at least 18 months before our data end in December 2015. That is, we only consider contracts signed until June 2014, to ensure that all contracts had at least 18 months to access before the data end. For both our DEP attrition analysis and first-term attrition analysis, we include 5 The designation of critical MOS is unrelated to the designation of critical language and is related to the Army’s assessment of what skills are undermanned over time. 6 For all groups, medical waivers make up the majority of DEP waivers, but for MAVNI contracts, medical waivers make up nearly 95 percent of all waivers. TIWARI_PROD_00002403 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 32 The Military Accessions Vital to the National Interest (MAVNI) Program Table 3.2 Army DEP and Accession Waivers Among Those with Four-Year Enlistment Term of Service, Wave 1 and Wave 2, MAVNI Contracts and Accessions Versus Three Comparison Groups Waiver Type MAVNI All Non-MAVNI Non-MAVNI Noncitizen Non-MAVNI High Quality DEP waivera Misdemeanor or felony waiver 0.0% 2.8% 1.1% 3.0% Drug or alcohol waiver 0.0% 0.1% 0.0% 0.1% Medical waiver 7.5% 9.1% 8.8% 9.7% Family waiver 0.3% 1.3% 2.7% 1.3% Administrative waiver (other) 0.1% 0.2% 0.2% 0.2% Any waiver 7.9% 13.5% 12.8% 14.2% Misdemeanor or felony waiver 0.0% 0.3% 0.1% 0.3% Drug or alcohol waiver 0.0% 0.0% 0.0% 0.0% Medical waiver 3.0% 3.0% 1.9% 3.3% Family waiver 0.1% 0.2% 0.3% 0.2% Administrative waiver (other) 0.1% 0.1% 0.2% 0.1% Any waiver 3.2% 3.5% 2.6% 3.8% Accession waiverb a Tabulations shown are the percentage of MAVNI contracts who receive each type of DEP waiver versus the percentage of contracts of the three comparison groups. Note that because of DEP attrition, not all contracts become an accession. b Tabulations shown are the percentage of MAVNI accessions who receive each type of accession waiver versus the percentage of accessions of the three comparison groups. SOURCE: Authors’ tabulations using Army RA Analyst data. only those who enlisted for a four-year term. For our analysis of attrition, we only used data on wave 1 accessions to ensure that they had at least four years to complete their enlistment and possibly reenlist. With respect to DEP attrition, we would expect MAVNI DEP attrition rates to be higher because MAVNI contracts spend more time in DEP (Table 3.1) and they have more time to be “at risk” of failing to access. In addition, since 2012, MAVNI contracts are required to be more extensively screened as part of the enlistment process than the typical recruit, and the additional screening may disproportionately weed out more MAVNI contracts, resulting in higher DEP attrition. On the other hand, MAVNI recruits have characteristics, such as higher educational achievement, that are associated with lower DEP attrition. And while the more-extensive screening might weed out more MAVNI recruits, only those willing to undergo the more-extensive TIWARI_PROD_00002404 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 33 screening process and who expect to pass the screen are likely to contract in the first place. Consequently, the additional screening might attract recruits who are more likely to access. Thus, the additional screening could increase or decrease DEP attrition. Figure 3.2 shows the percentage of contracts that fail to access, by FY, for MAVNI and non-MAVNI contracts. DEP attrition among MAVNI recruits increased between 2009 and 2014, especially after 2012, when new screening requirements for MAVNI recruits were introduced (recall that there were no MAVNI contracts in 2011 or 2012). Before 2012, MAVNI DEP attrition rates were below those for non-MAVNI recruits, but, as a result of rising DEP attrition among MAVNI recruits and declining rates among non-MAVNI recruits, DEP attrition in 2014 for MAVNI recruits was just below the rates for all non-MAVNI recruits and high-quality non-MAVNI recruits and about equal to the rate for noncitizen non-MAVNI recruits. Thus, we observe no large differences in DEP attrition among more-recent contracts but lower DEP attrition for MAVNI contracts among those who signed contracts in 2009, 2010, and 2013. Turning next to first-term attrition, Figure 3.3 shows the cumulative percentage of MAVNI recruits with a four-year enlistment term that are retained by month of service. By 48 months, 67.5 percent of MAVNI wave 1 accessions are still part of the Army enlisted force. Only some of those who leave in the first 48 months do so for “failure to adapt” reasons. Separation also occurs because some individuals become an officer, join the RC, or leave service for a medical-related reason. In this analysis, we do not consider such separations to constitute attrition. When we only consider “failure Figure 3.2 Percentage of Contracts That Do Not Access by FY, Conditional on Four-Year Enlistment Term, MAVNI Recruits and Comparison Groups 12.00 Percentage of contracts 10.00 8.0 6.0 4.0 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality 2.0 0 2009 2010 2013 2014 Fiscal year SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. RAND RR1976-3.2 TIWARI_PROD_00002405 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 34 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 3.3 Cumulative Percentage Retention of Wave 1 MAVNI Accessions, Conditional on Four-Year Enlistment Term 100 90 80 Percentage 70 60 50 40 30 20 MAVNI (failure-to-adapt attrition) MAVNI (all attrition) 10 0 0 6 12 18 24 30 36 42 48 Months of service SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. NOTES: “All attrition” is defined as a separation from the Army for any reason. Soldiers who become an officer or join the RC are not considered attrition in this figure, even though they are no longer part of the Army enlisted force. RAND RR1976-3.3 to adapt” attrition, by 48 months, 82.7 percent of MAVNI wave 1 accessions are still part of the enlisted force. Attrition due to failure to adapt is lower for MAVNI accessions than for nonMAVNI accessions, as seen in Figure 3.4. As in Figure 3.3, the figure shows the cumulative percentage of accessions that are retained through 48 months where attrition is due to failure to adapt. By 48 months, only 75 percent of noncitizen non-MAVNI accessions, 66 percent of high-quality non-MAVNI accessions, and 63 percent of all non-MAVNI accessions are still in the enlisted force. In short, consistent with the DiFazio et al. (2013) study, we find that MAVNI recruits perform better than other soldiers in terms of first-term attrition. Figure 3.5 shows the distribution of reasons for separation among MAVNI and non-MAVNI recruits, conditional on leaving. It shows that, among MAVNI recruits who enlist for four years and separate, 63.4 percent completed their enlistment term.7 7 Figure 3.5 shows a lower completion rate than the 67.5-percent retention rate shown in Figure 3.3 because the tabulations in Figure 3.5 are based on the final reason for separation. Thus, if a member completes his or her enlistment term and then subsequently becomes an officer or joins the RC, the final reason is no longer “completed term.” Because of this, the tabulations in Figure 3.5 will understate the percentage of accessions that have completed their term. TIWARI_PROD_00002406 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 35 Figure 3.4 Cumulative Percentage Retention (Non–Failure to Adapt) of Wave 1 MAVNI and Non-MAVNI Accessions, Conditional on Four-Year Enlistment Term 100 90 80 Retention rate 70 60 50 40 MAVNI Non-MAVNI, noncitizen Non-MAVNI, high quality Non-MAVNI, all 30 20 10 0 0 6 12 18 24 30 36 42 48 Months of service SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. RAND RR1976-3.4 We find that 13.7 percent of MAVNI recruits who separate attrited during their first 48 months because of failure to adapt, while 10.2 percent became officers or joined the RC. Again, more than 60 percent finish their term of service; the remainder leave because of a disability or because they are considered missing in action or killed in action.8 Reenlistment and Promotion Figure 3.6 shows that, although MAVNI recruits who complete their first term are more likely to become officers or join the RC than non-MAVNI recruits who complete their first term, they are less likely to reenlist. Specifically, the figure shows the percentage of those who reenlist, become an officer, or join the RC among soldiers who completed their first term. We find that 31.4 percent of MAVNI recruits who complete their first term reenlist for a second term, compared with 57.9 percent of non-MAVNI noncitizens and 48.1 percent of non-MAVNI high-quality recruits. The higher likelihood of becoming an officer or joining the RC is not sufficient to offset the lower reenlistment rate of MAVNI recruits. Overall, 41.3 percent of MAVNI recruits 8 Though we group disability, missing in action, and killed in action, the majority of those in this category are separated for disability. For example, none of the MAVNI recruits who separated did so because they were missing in action or killed in action, and only 0.4 percent of all non-MAVNI recruits who separated did so because they were missing in action or killed in action. TIWARI_PROD_00002407 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 36 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 3.5 Reason for Separation Among Wave 1 MAVNI and Non-MAVNI Accessions Who Separated, Conditional on Four-Year Enlistment Term 100 90 Percentage separating 80 70 To officer or RC 60 50 Missing in action, killed in action, or disabled 40 Failure to adapt Completed term 30 20 10 0 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. NOTES: The tabulations shown are based on the final reason for separation. Thus, if an individual completed a term, reenlisted, and became an officer, the separation reason would be “to officer or RC.” Consequently, the category “completed term” is undercounted. RAND RR1976-3.5 who complete their first term remain in service (including as an officer or in the RC), compared with 59.7 percent of noncitizen non-MAVNI recruits, 52.7 percent of all non-MAVNI recruits, and 50.1 percent of high-quality non-MAVNI recruits. Despite their lower likelihood of staying beyond the first term, MAVNI recruits are promoted faster than non-MAVNI recruits. Figure 3.7 shows the percentage of accessions who complete their first term who are promoted to E-4 by 19 months, 24  months, and 48 months, as well as the percentage who fail to become an E-4. MAVNI recruits who complete their first term are far more likely to become an E-4 by 19 months; 72.6 percent of MAVNI accessions reach E-4 by 19 months, compared with 42.2 percent of noncitizen non-MAVNI recruits, 24.4 percent of high-quality non-MAVNI recruits, and 23.6 percent of all non-MAVNI recruits. To some extent, this difference reflects the high levels of education of MAVNI accessions; because of their qualifications, MAVNI recruits are likely to be able to enter the Army above E-1. By 48 months, virtually all accessions, for both MAVNI and non-MAVNI groups, become an E-4. Promotions to E-5 are more competitive than are promotions to E-4, and, as shown in Figure 3.8, only 40.2 percent of MAVNI accessions who complete their first term achieve E-5 by 48 months. On the other hand, even fewer non-MAVNI accesTIWARI_PROD_00002408 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 37 Percentage completing first term who reenlist or transfer to officer or reserve component Figure 3.6 Reenlistment or Officer/RC Transfer Among Wave 1 Accessions Who Complete First Term, Conditional on Four-Year Enlistment Term 70 60 50 40 To officer or RC Reenlisted 30 20 10 0 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. RAND RR1976-3.6 Figure 3.7 Promotion to E-4 Among Wave 1 Accessions Who Complete First Term, Conditional on FourYear Enlistment Term 100 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality Percentage of E-4 promotions 90 80 70 60 50 40 30 20 10 0 No E-4 by 48 months E-4 by 19 months E-4 by 24 months E-4 by 48 months SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. RAND RR1976-3.7 TIWARI_PROD_00002409 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 38 The Military Accessions Vital to the National Interest (MAVNI) Program sions reach E-5 by 48 months; 32.0 percent of noncitizen non-MAVNI accessions, 36.1  percent of high-quality non-MAVNI accessions, and 26.8 percent of all nonMAVNI accessions reach E-5 by that point. MAVNI accessions are also more likely to reach E-5 by 60 months, conditional on completing their first term, as shown in Figure 3.8. Decomposition of Selected Performance Metrics The above results indicate that MAVNI recruits generally perform better in terms of having lower first-term attrition; faster promotion to E-4 and E-5; and a higher likelihood of having many of the characteristics found to be associated with better performance, such as more education, being older, and having higher AFQT scores. On the other hand, MAVNI recruits are less likely to stay beyond the first term even though they are more likely than non-MAVNI recruits to become officers or join the RC. Because MAVNI recruits face more-stringent enlistment standards and have a number of characteristics associated with better performance, a question of interest is whether their lower attrition and lower reenlistment rates relative to non-MAVNI recruits are due to differences in their characteristics or to differences in other factors. MAVNI recruits are provided citizenship upon entry to the military, but that citizenship is revoked if they fail to complete their initial service obligation. CitizenFigure 3.8 Promotion to E-5 Among Wave 1 Accessions Who Complete First Term, Conditional on FourYear Enlistment Term 100 MAVNI Non-MAVNI, all Non-MAVNI, noncitizen Non-MAVNI, high quality Percentage of E-5 promotions 90 80 70 60 50 40 30 20 10 0 No E-5 by 48 months E-5 by 24 months E-5 by 36 months E-5 by 48 months E-5 by 60 months SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. RAND RR1976-3.8 TIWARI_PROD_00002410 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 39 ship gives MAVNI recruits an incentive to perform sufficiently well to complete the first term. On the other hand, the ability to become a citizen in an expedited manner could induce individuals who qualify for MAVNI but who otherwise have a low taste for service to join the military. Once their service obligation is completed, their lower taste means that they are more likely to leave. They may also be more likely to leave at the completion of their obligation because of lower job satisfaction. The SME we interviewed expressed concerns that MAVNI enlistees do not use their language skills and many have considerable civilian education that is not matched to Army jobs that use that education. We use the Blinder-Oaxaca decomposition methodology to ascertain the extent to which MAVNI recruits have differential attrition and reenlistment outcomes from non-MAVNI recruits after controlling for differences in characteristics. Any remaining differences in attrition and reenlistment after controlling for observed characteristics are due to differences in unobserved characteristics or to structural differences, such as differences in incentives. Specifically, we can use the Blinder-Oaxaca methodology to decompose differences in DEP attrition, first-term attrition, and first-term reenlistment into the percentages that are due to differences in observed characteristics and the percentages that are due to other factors, including unobserved characteristics and incentives.9 Table 3.3 shows the decomposition results. The first row shows the decomposition of the difference in DEP completion (i.e., accessions among contracts) between MAVNI and high-quality non-MAVNI contracts. The second row shows the decomTable 3.3 Decomposition of the Difference in Mean Performance Metrics into Percentages Due to Differences in Characteristics and to Differences in Other Factors, MAVNI Recruits Versus All High-Quality Non-MAVNI Recruits Difference (MAVNI Part of Part of recruits Difference Difference Percentage Percentage minus Due to Due to Due to Due to nonDifferences Differences Differences Differences MAVNI in in Other in in Other recruits) Observables Factors Observables Factors MAVNI Recruits HighQuality NonMAVNI Recruits DEP completion 95.1% 91.0% 4.1% 3.7% 0.4% 90.1% 9.9% Cumulative 48-month retention rate 61.4% 47.6% 13.7% 8.9% 4.9% 64.4% 35.6% Reenlistment 31.7% 48.0% –16.3% –1.4% –14.8% 8.8% 91.2% Metric SOURCE: Authors’ tabulations using Army RA Analyst data and TAPDB data. 9 The decomposition is performed using STATA, and the regression results that are used to decompose differences in attrition and reenlistment outcomes are given in Appendix B. TIWARI_PROD_00002411 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 40 The Military Accessions Vital to the National Interest (MAVNI) Program position in the 48-month cumulative retention rate between MAVNI and high-quality non-MAVNI accessions. This is the retention rate that is based on all reasons for separation, not just failure to adapt. The last row shows the decomposition in the reenlistment rate (not including officer and RC transfers) between MAVNI and high-quality non-MAVNI recruits who have completed their first term. The analysis is conducted using data on wave 1 recruits who signed a four-year enlistment contract. Figure 3.3 showed that MAVNI recruits who contracted in 2009 and 2010 had lower DEP attrition and were more likely to access. As shown in the first row of Table 3.3, our regression analysis predicts that 95.1 percent of MAVNI contracts and 91 percent of high-quality non-MAVNI contracts accessed, a difference of 4.1 percentage points. We find that 3.7 percentage points of the 4.1-percentage-point difference are attributable to differences in observed characteristics. Thus, 90.1 percent of the difference in average DEP completion between MAVNI and non-MAVNI recruits is estimated to be due to the fact that MAVNI recruits have characteristics that are associated with higher DEP completion. Only 9.9 percent of the difference (0.4 percentage point) is attributable to other factors, such as unobserved characteristics or differing incentives of MAVNI and non-MAVNI recruits. MAVNI recruits are also more likely to complete 48 months of service than high-quality non-MAVNI recruits. The regression analysis predicts a mean 48-month retention rate for MAVNI recruits of 61.4 percent versus 47.6 percent for high-quality non-MAVNI recruits, a difference of 13.7 percentage points. Of this difference, 64.4 percent (or 8.9 percentage points) is attributable to the fact that MAVNI recruits are more likely to have characteristics that are associated with lower attrition than are other high-quality recruits who are not in the MAVNI program. Thus, slightly less than two-thirds of the difference in attrition is attributable to differences in observed characteristics. However, we find that, even holding observed characteristics constant, MAVNI recruits still have higher predicted retention, by 4.9 percentage points. Consequently, 35.6 percent of the predicted difference in the mean retention rate is attributable to other factors, including differences in the incentive to complete the first term. The predicted reenlistment rate of MAVNI recruits is lower than for non-MAVNI high-quality recruits, by 16.3 percentage points. Most of this difference—91.2 percent— is attributable to factors other than observed differences in MAVNI and non-MAVNI recruits who complete their first term. Only 8.8 percent of the difference is due to differences in observed characteristics. While we cannot attribute a causal effect of citizenship on reenlistment, the results are consistent with MAVNI recruits having a lower taste for service and separating at a higher rate once they complete service. This explanation is predicated on a key assumption of our decomposition analysis: that tastes for service and observed characteristics are uncorrelated.10 Another explanation 10 If tastes and observed characteristics are correlated, then the 91.2 percent could be understating or overstating the amount of the difference attributable to tastes rather than to observed factors. TIWARI_PROD_00002412 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Relative Performance of MAVNI Recruits 41 is that MAVNI recruits are less satisfied at the completion of their enlistment term because they are unlikely to use their language skills or be well matched to their Army job, given their civilian education. TIWARI_PROD_00002413 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER FOUR Cost to the Army The previous chapter considered the relative performance of MAVNI and non-MAVNI recruits. Also of interest to the Army is the relative cost. In this chapter, we consider two related questions on cost: 1. What is the incremental effect of a MAVNI recruit on the Army’s budget? 2. What is the effect, at the margin, of increasing accessions with MAVNI versus other high-quality recruits? The first question focuses on average budgetary cost to the Army of the existing MAVNI program and addresses whether MAVNI contracts are, on average, more or less costly in terms of the Army budget than non-MAVNI contracts. The second question focuses on whether, at the margin, an additional MAVNI recruit is less costly than a typical high-quality recruit. This second question addresses the issue of whether expanding the MAVNI program as a means of accessing more high-quality recruits is cost-effective relative to expanding high-quality enlistments in general. In this analysis, we measure average budget cost on a per-contract basis and on a per-person-year basis, where person years are measured in terms of service provided during the first 48 months of service. We measure costs this way because analysis in the previous chapter showed that MAVNI contracts are more likely to access—i.e., they are less likely to drop out of DEP. By measuring costs on a per-contract basis, we control for differences in attrition between contract and accession. The person-year metric is also motivated by the results in the previous chapter that showed that MAVNI recruits supply more first-term person years, thereby providing more service than nonMAVNI recruits. By showing results on a per-year basis, we are able to account for the differences in the number of years provided by MAVNI recruits versus non-MAVNI recruits. Ideally, we would measure the number of years provided over an entire Army career, but, even with the earliest contracts in our data (some of whom did not access until 12 months later), we lack data for some recruits for years provided beyond the first term, and, for those for which we do have data beyond the first term, we have no data through the completion of the second term. 43 TIWARI_PROD_00002415 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 44 The Military Accessions Vital to the National Interest (MAVNI) Program We use the TABDB and RA Analyst data, described in Chapter Three, to compute person-years. Because we need to consider contracts with sufficient time to complete at least a first term of enlistment, we only use data on the 2009–2010 entry cohorts. Furthermore, to ensure that the differences in person years are not attributable to differences in contract length, we only consider recruits who signed a four-year enlistment contract. In addition, in our analysis of budget costs, we use the three alternative definitions of the non-MAVNI group used in Chapter Two: all non-MAVNI recruits, non-MAVNI high-quality recruits, and non-MAVNI noncitizen recruits. Cost Methodology The incremental budget cost of MAVNI recruits is computed as the difference in the average cost of recruiting and training a MAVNI recruit versus a non-MAVNI recruit. We consider three comparison groups to define non-MAVNI recruits: non-MAVNI high-quality recruits, all non-MAVNI recruits, and non-MAVNI noncitizen recruits. Average cost is computed on a per-contract basis and a per-person-year basis over the first term.1 Data Sources and Cost Elements The source of data we use is the Army Military-Civilian Cost System (AMCOS), an Army web-based automated tool that provides cost factors for military personnel by grade, occupational grouping, and appropriation type, such as Military Personnel Appropriation (MPA) and Operations and Maintenance Appropriation (OMA).2 All AMCOS data we use are in 2016 dollars. We use the cost factors for an E-1 for all Army occupations. Because AMCOS recruiting costs exclude the cost of a security investigation, we also used information on the FY 2016 reimbursable billing rates for a background security investigation, available from the Office of Personnel Management (U.S. Office of Personnel Management, 2015). In FY 2016, the Army was charged $5,188 per recruit for a standard SSBI and was charged $595 per recruit for a NACLC. AMCOS recruiting costs include a number of elements. MPA costs include the effort or time cost of recruiters, moving costs for accessions, and, in the case of high1 An alternative measure is cost per applicant rather than cost per contract. As shown in Figure 2.1, MAVNI recruits undergo several screening steps prior to entry into DEP (e.g., when they are applicants and before they sign a contract), including DHS screening related to their immigration status and language testing. A metric based on cost per applicant would incorporate any applicant attrition that occurred during this stage of the enlistment process. Unfortunately, we were unable to measure attrition of applicants prior to DEP entry with the data we used, and so we are unable to measure cost on a per-applicant basis. That said, given that MAVNI recruits have lower DEP attrition than other recruits, despite their additional screening steps, it is unclear, a priori, whether MAVNI applicants would have higher or lower applicant attrition. 2 AMCOS is a tool maintained by the Office of the Assistant Secretary of the Army for Financial Management, and access requires Army authorization. TIWARI_PROD_00002416 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Cost to the Army 45 quality accessions, enlistment bonus costs. The AMCOS methodology assumes that the time cost of recruiting a high-quality recruit is 2.25 times the effort and cost of recruiting a low-quality one. OMA costs include examination, recruit travel, USAREC operations costs, and, in the case of high-quality recruits, advertising costs. Advertising costs are only included for high-quality recruits, presumably because the advertising is directed toward these more-difficult-to-enlist individuals. MPA and OMA costs for high- and low-quality recruits are each divided by the first-year retention rate of high- and low-quality recruits, respectively. The average recruiting cost over all enlisted recruits is computed by taking a weighted average over the high- and low-quality MPA cost factors and over OMA cost factors, where the weights reflect the relative number of high- versus low-quality accessions (with no adjustment for attrition). AMCOS training costs are also divided into MPA and OMA costs and include basic training costs, initial skills training (or one-station unit training when relevant), and career training. Unlike recruiting costs, AMCOS only provides training costs for all recruits and not specifically for high-quality recruits. The cost information originates from the installations where the training takes place and is compiled by the U.S. Army Training and Doctrine Command. AMCOS does not provide extensive documentation on the elements of cost included in training costs or the algorithms used to distribute fixed costs across courses, but the documentation states that both direct and indirect costs are included. Beginning in FY 2017, additional screening and monitoring requirements were levied on the MAVNI program. We discuss these requirements in more detail in Chapter Six; here we note that our analyses are based on cost data collected prior to these additional requirements. Estimating Costs for MAVNI, Non-MAVNI, and Non-MAVNI Noncitizen Recruits Unfortunately, AMCOS does not include cost factors specifically for MAVNI recruits or noncitizen recruits. It provides cost factors for all recruits and, in selected cases, for high-quality recruits. To address this issue for training costs, we assume average E-1 training costs for all of the groups. Consequently, training costs per accession are assumed to be the same, though costs per contract will differ between MAVNI and the comparison groups because MAVNI contracts are more likely to access, as will be shown in Table 4.2. To estimate recruiting cost factors for MAVNI recruits, we start with the cost factors for high-quality recruits provided by AMCOS but make several adjustments to better reflect aspects of the MAVNI program. 1. We do not include the cost of enlistment bonuses because virtually no MAVNI recruit received an enlistment bonus, as shown in Table 3.1. 2. We assume that advertising costs for MAVNI recruits are zero because our interviews and our analyses of data from GoArmy.com (see Chapter Two) indiTIWARI_PROD_00002417 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 46 The Military Accessions Vital to the National Interest (MAVNI) Program cated that MAVNI applicants find out about the program through word of mouth and social media. 3. We adjust recruiting costs with the first-year retention rate of 94.8 percent that we observe in the data for MAVNI recruits rather than the 88.3 percent rate used by AMCOS for all high-quality recruits. 4. We assume that MAVNI recruits take 2.25 times more effort to recruit than low-quality recruits; 2.25 is the same factor AMCOS uses for computing the time of high-quality recruiting. While our interviews indicated that MAVNI recruits require less time prospecting and finding, their applications can take far more time to process. Of course, it is possible that the true adjustment factor is more or less than 2.25. Consequently, our MAVNI recruiting cost estimate should be considered somewhat less precise than for other recruits. For recruiting costs for the three comparison groups, we use the AMCOS recruiting cost factors but adjust costs using the first-year retention rate observed in the TAPBD for the 2009–2010 contract cohort. For non-MAVNI high-quality recruits, we use AMCOS data for all high-quality recruits but use the first-year retention rate observed in TAPDB for this group. Similarly, for all non-MAVNI recruits, we use AMCOS data for all recruits but use the first-year retention rate for non-MAVNI recruits.3 We approximate the cost factors for noncitizen recruits as the cost factors for all non-MAVNI recruits. Because MAVNI recruits are only eligible to enlist in occupations for which a security clearance is not required, their occupational distribution differs from that of non-MAVNI recruits. Indeed, we find that among MAVNI recruits enlisting between 2013 and 2016, one of the most common occupational groups is CMF 92, supply and services, fields that do not require a clearance. (Table 2.7 shows tabulations of occupational fields.) Importantly, training costs are lower for recruits in CMF 92 in AMCOS. Therefore, as sensitivity analysis using training costs for an occupation that a MAVNI recruit is more likely to enter, we also computed the total cost of MAVNI contracts using CMF 92 training cost factors. For non-MAVNI contracts, we continued to use the cost factors relevant to all enlisted recruits. Finally, we mentioned earlier in the chapter that we use OPM cost factors for security background investigation costs. To compute average background investigation costs per contract for MAVNI recruits and the three comparison groups, we used the TAPDB to tabulate the percentage of recruits in each of the comparison groups receiving different types of background investigations. Virtually all of the recruits who 3 Using the TAPDB, we find an 88.3-percent first-year retention rate among high-quality non-MAVNI recruits who signed contracts in 2009–2010 (see Figure 3.5). AMCOS uses a low-quality first-year retention rate of 90.9 percent, implying a weighted average of 89.4 percent overall. Using the TAPDB, we use a rate of 86.9 percent for low-quality non-MAVNI recruits, implying a weighted average of 87.7 percent using TAPDB figures rather than 89.4 percent using AMCOS figures. TIWARI_PROD_00002418 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Cost to the Army 47 did not receive an SSBI received a NACLC. We found that nearly 10.4 percent of high-quality non-MAVNI accessions received an SSBI versus 6.5 percent of all nonMAVNI accessions and versus 0 percent of non-MAVNI noncitizen accessions. We assumed, as is current policy, that all MAVNI recruits receive an SSBI. Using this information together with information on DEP attrition rates, we estimated the average background investigation costs per contract for each group. In sum, we compare the recruiting and training costs of MAVNI recruits, including security investigation costs, to the costs of the three comparison groups. Furthermore, by way of sensitivity analysis to control for differences in occupations, we also perform these comparisons using the cost factors specific to CMF 92 for MAVNI recruits. Together, these comparisons provide a general picture of whether MAVNI recruiting is incrementally more or less costly than non-MAVNI recruiting. Results Incremental Cost Per Contract Table 4.1 shows the computation of the incremental cost per contract of a MAVNI recruit versus the three comparison groups, with the top part of the table showing recruiting costs, the middle showing training costs, and the bottom showing total costs per contract. We find that recruiting costs per contract, excluding security investigation costs, are somewhat higher for MAVNI recruits than for all non-MAVNI recruits—$23,104 Table 4.1 Incremental Budget Cost per Contract of MAVNI Recruits Versus Three Comparison Groups Cost Type MAVNI Recruits All NonMAVNI Recruits Non-MAVNI Non-MAVNI High-Quality Noncitizen Recruits Recruits $23,104 $21,630 $27,013 $22,090 $4,985 $794 $955 $543 $28,089 $22,424 $27,968 $22,633 $5,665 $121 $5,456 $11,195 $11,233 $11,433 $1,015 $978 $777 $33,619 $39,201 $34,066 $6,680 $1,099 $6,233 Recruiting costs per contract Recruiting excluding security investigation Security investigation Total Difference, MAVNI minus non-MAVNI Training costs per contract $12,210 Difference, MAVNI minus non-MAVNI Total cost per contract Difference in total cost per contract $40,299 SOURCE: Authors’ tabulations using AMCOS data and Army TAPDB and RA Analyst data on four-year enlistment Army contracts signed in 2009 and 2010. TIWARI_PROD_00002419 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 48 The Military Accessions Vital to the National Interest (MAVNI) Program versus $21,630—but lower than for high-quality non-MAVNI recruits. The latter result is not surprising, given that recruiting costs for all high-quality recruits include enlistment bonus and advertising costs where we excluded these costs for MAVNI recruits. We also find that the cost per contract of security background investigations is higher for MAVNI recruits because all MAVNI contracts are required to receive the highercost SSBI investigation, while non-MAVNI recruits are far more likely to receive the less-costly NACLC investigation. Overall, recruiting costs per contract are higher for MAVNI recruits than for all non-MAVNI recruits or non-MAVNI noncitizens. On the other hand, relative to high-quality recruits, the higher security background costs of MAVNI recruits roughly offset the lower recruiting costs, other than background checks. Consequently, we estimate that budgetary recruiting costs per contract are $121 higher for MAVNI recruits than for non-MAVNI high-quality recruits. As mentioned earlier in the chapter, we assume that training costs per accession are the same for MAVNI recruits and for each of the three comparison groups we consider. However, because MAVNI contracts are more likely to access, training costs per contract are higher for MAVNI recruits by $1,015 relative to all non-MAVNI recruits. Given the higher recruiting and training costs for MAVNI recruits, it is not surprising that we find that the cost per contract for MAVNI recruits is higher than for the comparison groups. The last line of the table shows the change in cost— representing the incremental budget cost per contract of MAVNI recruits relative to the other groups. The incremental cost per contract is roughly $6,700 relative to all non-MAVNI recruits and all non-MAVNI noncitizen recruits but about $1,100 relative to all high-quality non-MAVNI recruits. Incremental Cost per First-Term Person-Year We next consider the incremental cost of MAVNI per person-year. Table 4.2 shows the computation of first-term person-years for four-year Army contracts signed in 2009 and 2010. The 12- to 48-month cumulative continuation rates do not distinguish service as an active component enlisted or officer member or as a member of the selected reserve. That is, if an individual who signed a four-year contract in 2009 or 2010 later became an officer or transferred to the selected reserve during the 48 months, we include that individual as someone who continued in service. Thus, separation only means separation from the Army and not just from enlisted service. It is not surprising, given the results shown in Chapter Three, that we find that person-years per contract (shown in the last line of Table 4.2) are higher for MAVNI recruits than for non-MAVNI recruits. Indeed, person-years per contract for MAVNI are substantially higher than any of the three comparison groups, about 25 percent higher than other high-quality contracts, and 28 percent higher relative to all non-MAVNI contracts. Table 4.3 shows our results on the incremental budget cost of MAVNI relative to non-MAVNI recruits when costs are expressed per person-year. The earlier results showed that MAVNI recruits are more costly per contract. However, the fact that TIWARI_PROD_00002420 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Cost to the Army 49 Table 4.2 Computation of First-Term Person-Years, Given Contract for MAVNI and Three Comparison Groups Metric Contract rate All Non-MAVNI MAVNI Recruits Recruits 100% Accession rate 100% Non-MAVNI High-Quality Recruits Non-MAVNI Noncitizen Recruits 100% 100% 97.4% 89.3% 89.6% 91.2% 12 months, given accession 94.8% 87.7% 88.3% 90.2% 24 months, given accession 89.8% 81.6% 82.8% 86.8% 36 months, given accession 82.1% 69.2% 71.3% 75.3% 48 months, given accession 73.5% 57.9% 60.7% 69.1% Person-years, given accession 3.11 2.63 2.69 2.82 Person-years, given contract 3.02 2.35 2.41 2.58 Continuation rate SOURCE: Authors’ tabulations using Army TAPDB and RA Analyst data on four-year enlistment Army contracts signed in 2009 and 2010. NOTE: Computation of continuation rate includes service as an officer or selected reservist. Table 4.3 Incremental Budget Cost per First-Term Person-Year of MAVNI Versus Three Comparison Groups Cost Recruiting costs per person-year MAVNI Recruits $9,288 Difference, MAVNI minus non-MAVNI Training costs per person-year $4,037 Difference, MAVNI minus non-MAVNI Total cost per person-year Difference in total cost per person-year $13,325 Non-MAVNI All Non-MAVNI High-Quality Recruits Recruits Non-MAVNI Noncitizen Recruits $9,538 $11,603 $8,786 –$250 –$2,315 $502 $4,762 $4,660 $4,438 –$724 –$622 –$401 $14,300 $16,263 $13,225 –$974 –$2,937 $101 SOURCE: Authors’ tabulations using AMCOS data and Army TAPDB and RA Analyst data on four-year enlistment Army contracts signed in 2009 and 2010. NOTE: Computation of continuation rate includes service as an officer or selected reservist. TIWARI_PROD_00002421 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 50 The Military Accessions Vital to the National Interest (MAVNI) Program MAVNI recruits supply more years of service over the first 48 months of their service means that the fixed costs of recruiting and training can be spread over more time. In total, we find that MAVNI contracts reduce the Army budget per person-year by $974 relative to all non-MAVNI recruits and by $2,937 relative to non-MAVNI high-quality recruits. Relative to other noncitizen contracts, the cost per person-year is roughly the same for MAVNI contracts, with MAVNI costs exceeding those of other noncitizen recruits by about $101 per person-year. Noncitizens provide more personyears during the first term than other non-MAVNI groups. In summary, we find that the much higher person-years supplied by MAVNI recruits are enough to offset or to just about equalize the higher per-contract recruiting and training costs of MAVNI recruits, depending on which comparison group we consider. The similarity between cost per person-year of MAVNI and noncitizen non-MAVNI recruits suggests that both types of noncitizens provide a similar return to the Army in terms of cost per first-term person years. However, they may not provide similar language and cultural capabilities and, indeed, noncitizens who are permanent residents are not tested for their foreign language capability at entry, as MAVNI recruits are. Sensitivity Analysis We recomputed the results in Tables 4.1 and 4.3 for MAVNI contracts using AMCOS training cost factors for CMF 92, one of the top occupational fields in which MAVNI recruits with CFL skills enter. As mentioned earlier in the chapter, training costs are lower for CMF 92 than for many other occupational fields. The results are shown in Table 4.4. Relative to all non-MAVNI recruits and nonMAVNI noncitizen recruits, we continue to find that the incremental cost of MAVNI recruits per contract is higher, despite the lower training costs for MAVNI recruits. Table 4.4 Incremental Budget Cost per Contract and First-Term Person-Year of MAVNI Versus Three Comparison Groups All Non-MAVNI Recruits Non-MAVNI High-Quality Recruits Non-MAVNI Noncitizen Recruits Based on data for all high-quality recruits $6,680 $1,099 $6,233 Based on data for high-quality recruits in CMF 92 $4,260 –$1,321 $3,813 –$974 –$2,937 $101 –$1,774 –$3,738 –$700 Difference in Cost Difference in total cost per contract Difference in total cost per person-year Based on data for all high-quality recruits Based on data for high-quality recruits in CMF 92 SOURCE: Authors’ tabulations using AMCOS data and Army TAPDB and RA Analyst data on four-year enlistment Army contracts signed in 2009 and 2010. NOTE: Computation of continuation rate includes service as an officer or selected reservist. TIWARI_PROD_00002422 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Cost to the Army 51 However, the difference is now smaller than before. For example, the difference is $4,260 rather than $6,680 compared with all non-MAVNI recruits (see Tables 4.4 and 4.1, respectively). Now, relative to other high-quality recruits, the lower training costs of MAVNI recruits are enough to offset the higher recruiting costs, so MAVNI recruits cost the Army $1,321 less per contract than other high-quality recruits. In terms of person-years, we find that MAVNI contracts cost the Army less per personyear than non-MAVNI contracts, even noncitizen contracts. These results broadly reinforce the findings when we use training costs for all enlisted occupations for MAVNI recruits. We generally find that MAVNI recruits increase the Army budget requirement, per contract, relative to other recruits. However, on a first-term person-year basis, when the higher fixed costs of recruiting and training MAVNI recruits can be spread over more months of service, MAVNI recruits are less costly. Marginal Cost Estimates If the MAVNI program were reduced in size while the number of total accessions remained constant, the reduction in MAVNI accessions would need to be compensated for with an increase in other high-quality enlistments. A relevant question is whether replacing a MAVNI recruit with a generic high-quality recruit would, at the margin, increase Army costs. That is, is the marginal cost of a MAVNI recruit more or less than the marginal cost of a typical high-quality recruit? To address this question, we require marginal cost estimates of a typical high-quality recruit and a MAVNI recruit. Past studies consistently show that expanding the enlistment supply of highquality recruits increases Army costs at the margin because the Army must also expand recruiting resources, such as recruiters, advertising, and bonuses, to achieve more highquality enlistments (Simon and Warner, 2005; Asch et al., 2010). Because expanding the supply of high-quality enlistments requires additional resources, the marginal cost exceeds the average cost of a high-quality recruit.4 Estimates of the marginal cost of an Army recruit vary. Orvis et al. (2016) provides an Army estimate of $118,000. The marginal cost estimate of expanding Army high-quality enlistments using military pay from Asch et al. (2010) in 2016 dollars is $66,300, about half of the estimate in the Orvis et al. study. Using the elasticity estimate of 1.05 for Army high-quality enlistments in Simon and Warner (2005) and using a 2016 average cost of regular military compensation for an E-1 of $20,574, taken from AMCOS, gives a marginal cost 4 This is shown in Stigler (1966), which derives the formula MC = AC (1 + 1/e), where MC is marginal cost, AC is average cost, and e is the elasticity of high-quality supply with respect to military pay. TIWARI_PROD_00002423 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 52 The Military Accessions Vital to the National Interest (MAVNI) Program estimate of $80,400. Thus, recent studies provide a 2016 marginal cost estimate of expanding high-quality enlistments of between $66,300 and $118,000. To estimate the marginal cost of a MAVNI recruit, it is useful to recognize that the scale of the MAVNI program has been quite small since it began in 2009, given the caps on the program. This is true even for the Army, the service that takes the bulk of the cap. Therefore, the MAVNI program is arguably demand-constrained, meaning that more individuals would like to enlist in the Army’s MAVNI program than are permitted to do so, given the constrained cap. While we do not have direct evidence that the cap constrains MAVNI enlistments, several indicators suggest that this is the case. First, both the interviews we conducted and our review of social media, discussed in Chapter Two, suggested that there is substantial interest among potential MAVNI applicants. Second, in FY 2016, the Army cap for MAVNI enlistments was the largest of any year since the program began, yet the cap was reached in late June, suggesting that additional MAVNI enlistments would have been possible before the end of the fiscal year in September, had the cap not been place.5 In a demand-constrained environment, expanded recruiting resources are not required to also expand MAVNI enlistments. In this case, the marginal cost of a MAVNI recruit equals average cost. As we show in Table 4.1, the average cost of a MAVNI contract is about $40,300, implying that in the current demand-constrained environment, the marginal cost of a MAVNI contract is about $40,300. Comparing estimates, we find that the marginal cost of a MAVNI recruit of $40,300 falls below the range of $66,300 to $118,000 for a high-quality recruit. The implication is that a MAVNI recruit is less costly at the margin than a generic high-quality recruit. Put differently, if MAVNI enlistments are demand-constrained, swapping a high-quality recruit for a MAVNI recruit would increase Army costs at the margin because increasing high-quality recruits requires expansion of recruiting resources while increasing MAVNI enlistments would not. As we argue in the next chapter, we think a modest increase in the cap could be sustained with relatively small or no incremental cost in terms of recruiting resources, but we are unable to estimate MAVNI enlistment supply or cost beyond that, given uncertainty about the size of the potential applicant pool and Army capacity to process MAVNI recruits. As noted above, additional screening and monitoring requirements were put in place at the beginning of FY 2017. These requirements will increase the cost of MAVNI contracts and person-years, though we do not know by how much. Incorporating these additional costs could change the ranking of costs indicated in Tables 4.1–4.4 and the marginal cost comparison. 5 At that point in time, USAREC retained a limited number of slots for the continued recruitment and processing of HCPs (personal communication, ASA, M&RA, July 20, 2016). TIWARI_PROD_00002424 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER FIVE Estimating the Number of Potential MAVNI Enlistees The previous chapter showed evidence that MAVNI recruits are more cost-effective, measured in terms of per-person years during the first term, relative to non-MAVNI recruits. An additional question concerns the size of the pool of potential MAVNI recruits: Are there sufficient numbers of potential recruits to meet the DoD cap or to support an increase in the cap? We use available data to address this question in this chapter and estimate the number of individuals in the United States who could meet MAVNI eligibility requirements. A key conclusion is that the data necessary to make this estimate are lacking, so we can only provide a rough estimate to answer these questions. Past research has estimated the size of the young adult population in the United States that would be eligible for military enlistment. For example, Bicksler and Nolan (2009), citing work by Seifert, Hogan, and Moore (2007), report that the pool that was eligible for service was about 8 million in 2007, or about 26 percent of the 17- to 24-year-old population, where eligibility is defined in terms of ability to meet standards related to medical or physical conditions, drug and alcohol screens, dependent status, aptitude, and prior criminal record. Past studies estimate the percentage eligible for service using several data sources, including the 1997 Profile of American Youth Survey (part of the National Longitudinal Survey of Youth), the National Health and Nutrition Examination Survey, and the National Survey on Drug Use and Health. Once the percentage is estimated, it is applied to an estimate of the size of the overall youth population taken from U.S. Census Bureau data. Unfortunately, these data sources are not well suited for estimating the potential pool of young adults who would meet the MAVNI eligibility requirements because data on the additional requirements for MAVNI recruits are lacking. In addition to the requirements for all recruits, MAVNI recruits must have specific visa or immigrant statuses, must have resided legally in the United States for at least two years with a temporary visa, must speak a language judged to be critical for DoD’s capabilities (or possess medical skills and credentials in a critical specialty), must speak English proficiently, and must successfully complete an SSBI. Data sources used in past research do not include information on specific visa categories, and, even in cases where information on language proficiency or other key variables are included, they do not include 53 TIWARI_PROD_00002425 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 54 The Military Accessions Vital to the National Interest (MAVNI) Program a sufficient number of observations on the MAVNI languages to be useful. None of these data sources provides information related to the ability to successfully complete an SSBI. Another complicating factor is that these data sets generally include country of origin but often exclude language. This is problematic because MAVNI enlistment requires proficiency in one of a set of languages rather than origination from one of a set of specific countries.1 Ultimately, to answer the question about the size of the potential MAVNI pool of recruits, DoD will need to design a survey specifically targeted to address this topic. We return to this topic when we discuss areas for future research. With available data, at best, we can provide a rough upper-bound estimate of the number of individuals who might meet MAVNI eligibility requirements and assess whether this upper bound exceeds recent caps on the MAVNI program. To estimate an upper bound, we first use the American Community Survey (ACS), an ongoing survey of U.S. residents conducted by the Census Bureau. The ACS is representative of the entire U.S. population and includes information on citizenship, age, education credentials, language spoken, and English proficiency. However, it lacks information on other enlistment qualification characteristics, such as health status, aptitude, and drug use, and other qualification characteristics specific to MAVNI, such as visa category and ability to pass an SSBI. As we show below, it is likely that the ACS undercounts highly mobile populations, such as individuals holding student visas.2 Thus, we also make use of a second data set, Student and Exchange Visitor Information System (SEVIS) data, from I-20 forms generated by U.S. Immigration and Customs Enforcement.3 SEVIS has published quarterly reports since at least 2009; detailed data including country of origin have been publicly available since 2014. SEVIS data are particularly useful because they include information on the stock of individuals holding student visas, and the majority of MAVNI recruits are student visa holders (see Table 2.3). On the other hand, SEVIS data lack information on other types of visa holders and have no information on such recruit qualification characteristics as health status, aptitude, drug use, and length of U.S. residency. 1 In some cases (such as Saudi Arabia, where nearly all speak Arabic), language and country of origin correspond quite closely, but in other cases (such as India), many languages may be spoken, while only some of those languages qualify potential enlistees for the MAVNI program. 2 This appears to be the case, even though we use the sample designed to include those living in group quarters (which includes dormitories). For more information, see U.S. Census Bureau, 2011. 3 On behalf of DHS, the Student and Exchange Visitor Program (SEVP) manages schools and nonimmigrant students in the F and M visa classifications as well as their dependents. If accepted by a SEVP-certified school, foreign students will be issued an I-20 and then may be admitted to the United States with the appropriate F or M nonimmigrant status. While the Department of State manages exchange visitor programs (J visa classifications), both SEVP and the Department of State use SEVIS to track and monitor F, M, and J nonimmigrants while they visit the United States and participate in the U.S. education system. The SEVIS data include the students’ country of citizenship, the U.S. school they attend, and the field of study. TIWARI_PROD_00002426 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating the Number of Potential MAVNI Enlistees 55 We also attempt to estimate whether the potential pool is growing or contracting, making use of DHS data on nonimmigrant admissions to the United States, both those on student visas and those on other visas that qualify for the MAVNI program. These data provide information on the flow of nonimmigrants to the United States, thereby indicating trends over time. Specifically, students (and others) receive an I-94  form each time they enter the United States; this documents their legal status during their stay; therefore, these data also provide information on those in nonstudent visa categories who may be eligible for the MAVNI program. These data include total counts of admissions of nonimmigrants required to complete form I-94 by country of citizenship, age group, visa type, gender, and fiscal year for 2010–2015 from DHS.4 An I-94 is completed during each admission; thus, the data refer to admissions, not individuals, and the same person may be admitted multiple times. Because the data include no identifying information, we cannot determine the number of times each person was admitted. Therefore, the data give an overestimate of the number of individuals and so do not constitute a true measure of flow into (and out of) the United States. However, we can use instances where we have information on both the number of admissions from the DHS data and the number of individual students from SEVIS data to estimate the average number of times that an individual enters the United States each year. We then use this number to estimate the number of individuals that the given number of admissions represents. Estimating the Size of the Pool from ACS Data We begin with the 2009–2014 ACS data and roughly estimate the number of potential MAVNI enlistees as the number of individuals who are noncitizens, speak a language on the MAVNI list, speak English well, have completed at least high school, and are in the target age range (17–34). We find that about 106,000 people meet these criteria. Of course, this estimate includes no information on health status, aptitude, drug use, or misdemeanors. It also includes noncitizens who are in visa categories that do not qualify for MAVNI and does not account for ability to successfully complete an SSBI. The estimate produced by ACS is much lower than the estimate we get using alternative approaches. Based on this ACS estimate, a substantial proportion of those eligible to enlist in the MAVNI program may have already done so. We suspect that ACS simply misses many foreign students in the visa categories of interest, perhaps 4 For all nonimmigrant data, the FY spans October 1 through September 30 (e.g., FY 2015 spans October 1, 2014, through September 30, 2015). Information collected on Form I-94 includes arrival date, departure date, port of entry, class of admission, country of citizenship, state of destination, age, and sex. The arrival and departure sections of Form I-94 are submitted separately, and we have no data on departures. Cells with fewer than three individuals in the data were suppressed. We gratefully acknowledge the assistance of our colleague at RAND, Henry Willis, and of James Lee of the Office of Immigration Statistics within DHS. TIWARI_PROD_00002427 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 56 The Military Accessions Vital to the National Interest (MAVNI) Program because they are mobile, travel home frequently, are hesitant to respond to surveys, and/or live in on-campus apartments or similar housing.5 For any of these reasons, such students could be missed by ACS. Estimates Based on SEVIS Data Next, to obtain an alternate estimate of the size of the pool of potential MAVNI recruits, we utilize SEVIS data rather than ACS data (recall that SEVIS data include those admitted on student visas and these data are generated by U.S. Immigration and Customs Enforcement from I-20 forms). Figure 5.1 uses the SEVIS data and the 2016  MAVNI language list to show the number of students with F-1 or M-1 visas studying in the United States at different points in time between April 2014 and July 2016. The total number of F-1 and M-1 students is divided into those who come from countries where a MAVNI language is spoken and those who come from other countries. About 80 percent of M-1 and F-1 students throughout this time period come from countries where a MAVNI language is spoken. This high percentage is driven by the large number of students from Asian countries, especially China, India, and Korea.6 Over this time period, the average number of students in the United States on F-1 or M-1 visas from countries where a MAVNI language is spoken was nearly 900,000. This figure is consistent with data from a recent survey of U.S. institutions of higher education indicating that roughly 1,000,000 foreign students were enrolled during the 2015–2016 school year.7 The range of estimates between the ACS and the SEVIS data is striking. From ACS, we estimate a pool of just over 100,000. From the SEVIS data, looking only at F-1 and M-1 visa holders, we estimate a pool of about 900,000. There are a number of 5 ACS is designed to be a representative sample, and it includes persons living in a variety of group quarters. Group quarters include college residence halls and other student housing but also military barracks, emergency shelters, residential treatment facilities, religious group quarters, nursing facilities, and correctional facilities. There is no detail provided to determine how on-campus apartments and similar residences are sampled. The sample appears to be roughly representative in terms of (all) students, but our analyses suggest that many foreign students are not included. 6 In fact, in July 2016, seven of the top ten countries of citizenship for students studying in the United States were countries where a MAVNI language is spoken. The top ten in order were China, India, South Korea, Saudi Arabia, Canada, Vietnam, Japan, Taiwan, Brazil, and Mexico. However, the FY 2017 language list is more restrictive than earlier lists (see Appendix A). In particular, Chinese and Korean are not included, although Arabic, Hindi, and Tamil remain MAVNI languages. Arabic is a dominant language in Saudi Arabia; Hindi and Tamil are spoken by large numbers of residents in India. Note that the policy documents reference “Chinese”; the Army has recruited speakers of Mandarin and of Cantonese. We discuss the implications of these changes in Chapter Seven. 7 See Institute of International Education (2016). TIWARI_PROD_00002428 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating the Number of Potential MAVNI Enlistees 57 Figure 5.1 Number of F-1 and M-1 Students by MAVNI-Language-Speaking Country of Citizenship, April 2014–July 2016 Number of F-1 and M-1 students 1,400,000 1,200,000 1,000,000 MAVNI Non-MAVNI 1,015,138 1,112,573 1,132,847 1,194,775 1,184,268 1,112,549 1,052,126 961,959 800,000 600,000 400,000 200,000 0 Percentage of MAVNI students Apr 2014 Jul 2014 Oct 2014 Feb 2015 Sep 2015 79.06% 79.10% 79.16% 79.12% 79.75% Total 1,015,138 961,959 Dec 2015 Mar 2016 80.21% 80.52% Jul 2016 80.58% 1,112,573 1,132,847 1,052,126 1,194,775 1,184,268 1,112,549 MAVNI 802,530 760,883 880,666 896,328 839,105 958,351 953,572 896,446 Non-MAVNI 212,608 201,076 231,907 236,519 213,021 236,424 230,696 216,103 SOURCE: Study in the States, undated. RAND RR1976-5.1 possible reasons for the large difference. The SEVIS data do not include any indication of English proficiency, native language, education level, age, or length of residency in the United States, as ACS does. The native language issue in particular means that all students from a country where a MAVNI language is spoken are considered to speak the MAVNI language. This is surely not true in countries with several dominant languages. But there are also other factors that likely contribute to the difference. First, the SEVIS data lack other details; although most students are likely to be in the age range required for MAVNI enlistees and likely to hold the required education, applying these additional criteria to the SEVIS data would likely reduce the SEVIS-based estimate.8 Similar to the ACS data, the SEVIS data also lack detail on health status, weight, and aptitude scores. On the other hand, the ACS data we use cover the years 8 For example, Baker (2012, p. 7) explains, “The active student estimate is expected to exceed the size of the residential population in general because students may retain active status while abroad between semesters, may TIWARI_PROD_00002429 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 58 The Military Accessions Vital to the National Interest (MAVNI) Program 2009 to 2014, while the SEVIS data are for 2014 to 2016, and the number of foreign students may have increased in recent years.9 But it appears that the ACS data understate the total number of foreign students in any case, falling far short of the 900,000–1,000,000 estimate of foreign students from other sources before applying additional criteria. It seems unlikely that applying language, English proficiency, and education-level filters to a student population would result in screening out nearly 90 percent of the population. Thus, while we suspect that the SEVIS data overstate the number of potential MAVNI recruits because of the native language issue, the SEVIS data are probably still more accurate than the much smaller estimate from the ACS data, implying that the upper part of the 100,000 to 900,000 range is more plausible as an estimate of the size of the potential MAVNI pool of recruits (before applying additional criteria, such as length of residence in the United States). Estimates Based on Admissions Data Next, we examine data provided by DHS to understand the movements (“flow”) of potential MAVNI recruits into and out of the United States and to learn more about other MAVNI-eligible visa holders. Recall that the SEVIS data include only F-1 and M-1 visa holders. While most MAVNI enlistees to date hold an F-1 or M-1 visa, other visa types are eligible, as shown in Chapter Two. The DHS data, in contrast, include all visa categories, but the DHS data refer to admissions versus individuals; people may enter the United States more than once in a given time period. However, we can combine subsets of the SEVIS and DHS data to develop an estimate of the number of admissions per visa holder. We begin by presenting basic counts of admissions from the DHS data. Over the time period included in our data, specifically 2010 to 2015, admissions from countries that speak a MAVNI language have increased markedly, both among those holding all MAVNI-eligible visa types and among those holding only F and M visas, as shown in Figure 5.2. Our tabulations are for individuals who are ages 18 to 34, holding MAVNI-eligible visas. We find that in 2015, there were more than 1.7 million admissions for all MAVNI visa types, of which just over 1 million, or about 60 percent, were for F and M student visas. In 2010, the latter figure was about 587,000, while the former was about 1.16 million, indicating that about 51 percent were for F and M student visas. When we consider admissions from specific countries (Figure 5.3), we observe a sharp increase in admissions from China and Arabic-speaking countries but be in the United States for a period of study lasting only a few weeks, or may travel abroad too often to be considered residents.” 9 Based on analyses of past SEVIS data, it appears that the number of foreign students studying on F-1 or M-1 visas increased somewhat in recent years (see, e.g., Ruiz, 2015). Based on our analysis of the SEVIS data, we estimate an annual growth rate of between 8 and 10 percent per year, on average, since 2009. TIWARI_PROD_00002430 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating the Number of Potential MAVNI Enlistees 59 Figure 5.2 Number of Admissions for MAVNI Visa Types from Countries That Speak a MAVNI Language, 18- to 34-Year-Olds, 2010–2015 2,000,000 1,800,000 All MAVNI visa types Number of admissions 1,600,000 1,400,000 1,200,000 1,000,000 800,000 F and M visa types 600,000 400,000 200,000 0 2010 2011 2012 2013 2014 2015 1,157,569 1,209,963 1,260,640 1,383,299 1,583,007 1,743,342 F and M visa types 587,269 645,129 704,812 799,769 934,643 1,051,971 Percentage of F and M visa types 50.73% 53.32% 55.91% 57.82% 59.04% 60.34% All Fiscal year SOURCE: U.S. Department of Homeland Security. NOTES: Admissions represent counts of events—i.e., arrivals, not unique individuals. Multiple entries of an individual on the same day are counted as one admission. RAND RR1976-5.2 little change from Korea. By 2015, there were more than 700,000 admissions among F and M visa types from those three countries. The data suggest substantial growth in admissions over time. We can use the foreign student admissions data from DHS together with the SEVIS data, including those on foreign student visas, to develop an alternative estimate of the potential MAVNI pool. This method somewhat resembles a “residual” method that has been used to estimate emigration of various groups as well as the population of undocumented immigrants (Jensen, 2013; Passel and D’Vera Cohn, 2011). Here, we compare the totals to assess the number of admissions per person. Using TIWARI_PROD_00002431 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 60 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 5.3 Number of F and M Visa Admissions for Countries That Speak a MAVNI Language, by Language, 18- to 34-Year-Olds, 2010–2015 500,000 450,000 Number of admissions 400,000 350,000 300,000 Arabic Chinese French (African) Korean Portuguese Serbo-Croatian Other 250,000 200,000 150,000 100,000 50,000 0 2010 2011 2012 2013 2014 2015 Arabic 58,320 85,326 114,172 136,395 164,729 179,894 Chinese 185,178 228,218 271,591 325,744 382,636 434,429 French (African) 4,649 4,439 4,665 5,382 6,529 6,921 Korean 106,387 100,217 91,034 87,609 90,522 89,650 Portuguese 21,956 25,436 26,852 27,052 28,865 30,729 SerboCroatian 1,325 1,246 1,170 1,267 1,628 3,266 209,454 200,247 195,328 216,320 259,734 307,082 Other Fiscal year SOURCE: U.S. Department of Homeland Security. NOTES: Admissions represent counts of events—i.e., arrivals, not unique individuals. Multiple entries of an individual on the same day are counted as one admission. RAND RR1976-5.3 data on student visa admissions between 2009 to 2014, we estimate that there were approximately 1.75 F-1 and M-1 students in the SEVIS data for each DHS admission. When we focus instead on only the countries where MAVNI languages are spoken, the ratio is roughly 1.2, suggesting that students from countries currently on the MAVNI list enter and exit the United States less frequently than others. This is likely due to TIWARI_PROD_00002432 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating the Number of Potential MAVNI Enlistees 61 the large number of students from countries, such as China, that are distant from the United States. Students from these countries are likely to return home less frequently than students from less-distant countries, such as Canada or Mexico. Taking this information together and assuming that the number of admissions remains at a relatively high level of roughly 1,740,000 for all MAVNI visa categories (the average value over the last few years) and then deflating by 1.2 suggests that there are roughly 1,450,000 individuals in MAVNI visa categories. That is, we generate an estimate of the size of the MAVNI pool as 1,450,000. If, instead, we deflate by the more-conservative estimate of 1.75, we get an estimate of 1,000,000 individuals in the potential pool. These alternative estimates are much closer to the SEVIS estimate of 900,000 than the ACS estimate of just over 100,000, lending more credence to the SEVIS estimate over the ACS estimate. The alternative estimates are larger than the estimate based on the SEVIS data because the alternative estimates include all who hold a MAVNI-eligible visa, while the SEVIS data include only those on student visas. Data from recent years also suggest that the number of immigrant students, the number from MAVNI countries, and the number of admissions are all growing; the growth rate appears to be in the range of 8 to 10 percent per year on average. This estimate of the growth rate is identical to the estimate based on SEVIS data alone. To summarize, the ACS estimates likely undercount foreign students. The SEVIS estimate of 900,000 is closer to estimates from other sources, including those from a survey of universities and estimates that we generated using a form of the residual method found in the literature. While we believe that the true estimate likely lies closer to 900,000 than to 100,000, we stress that the 900,000 estimate does not exclude those who fail to meet key qualifications (health status, test scores, language, and capacity to pass an SSBI), and there is currently no information available to estimate this proportion. Among the U.S. population, estimates suggest that roughly three-quarters of young people fail to meet at least one of the key qualifications, as discussed at the beginning of this chapter. Many potential MAVNI recruits likely are ineligible as well, but no data exist to estimate the proportion that are ineligible. Moreover, none of these estimates include any information about propensity; there are no data currently available to estimate how propensity differs between potential MAVNI recruits and others. Thus, while we believe that the true estimate of the population lies closer to 900,000 than to 100,000, we suspect that the true estimate of the eligible population is well below 900,000, but we lack the evidence necessary to offer a more-precise estimate. Implications It would be possible to produce an improved estimate of the potential pool of MAVNI enlistees by surveying the population. Such a survey could include questions on English TIWARI_PROD_00002433 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 62 The Military Accessions Vital to the National Interest (MAVNI) Program proficiency, health status, time spent in the United States, and the types of questions that come up in the SSBI, as well as questions designed to discern interest in military service. Without additional information on qualification characteristics, we are sharply limited in the types of definitive statements we can make about this population. While we cannot provide a robust estimate, we note that, to date, the MAVNI program has filled each cap before the end of the fiscal year. For example, during FY 2016, DoD enlistees reached the 5,000 MAVNI cap only nine months into the FY, at which point the program closed. This suggests that, even with modest increases in the cap or deletions from the list of languages considered critical, the MAVNI program likely would continue to have sufficient numbers of applicants. That said, we are unable to precisely estimate how many more additional accessions would be possible with a higher cap, other than to say that the number is positive. Aside from uncertainty about the size of the potential pool and the number of MAVNI recruits who would be eligible and positively inclined to join the Army, Army capacity to process MAVNI recruits is limited. The new FY 2017 guidelines include changes in the language list, with a shift away from some of the most-common languages; these changes have implications for the size of the potential pool. We discuss the implications of these changes in more detail in Chapter Seven. TIWARI_PROD_00002434 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER SIX Estimating Security Risk This chapter provides information on what is known about the potential security risk associated with the MAVNI program. Security risk is actually not well defined in policy debates; this is especially the case with terrorism and violent extremism, since there are no universally accepted definitions of these terms. Because of our focus on military enlistment, we rely on the definition used by the Defense Personnel Security Research Center (PERSEREC) in its interviews and meetings with personnel from the U.S. Military Entrance Processing Command, as discussed in Buck et al. (2005). This definition is anyone who was sympathetic to, or a member of, a group that could be characterized as both disloyal and hostile toward the U.S. government. Effectively, anyone who is antagonistic toward the U.S. government and who would be willing in any way to support the efforts of a specific group in working against the U.S. government, its citizens, and entities would qualify as someone whom we are interested in detecting and excluding from military service and from possible access to sensitive information and facilities. Our discussion focuses on the security risk posed by individuals already in the United States. One measure of security risk is the number of plots and acts of terrorism or espionage committed by those in the United States relative to the population that is at risk of committing or planning such plots or acts. By this metric, the security risk of MAVNI recruits is zero because there are no publicly available reports of such acts or plots among MAVNI recruits. More broadly, available data sources discussed later in this chapter indicate that such acts are extremely rare among military personnel and among Americans in general. For example, Buck et al. (2005) reports that the number of current and former enlisted members with associations with terrorist and extremist groups is extremely small, and the prevalence of such individuals is “nearly infinitesimal.” PERSEREC maintains a database of espionage acts by Americans, containing 63 TIWARI_PROD_00002435 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 64 The Military Accessions Vital to the National Interest (MAVNI) Program only 173 individuals between 1947 and 2007. Again, this figure is extremely small relative to the number of individuals who have or have had security clearances since 1947.1 On the other hand, because the human and financial consequences of terrorism and espionage can be so great, estimation of security risk in the available literature has focused less on estimates based on past incidences and more on the factors that predict the future perpetration and planning of such acts. As we will discuss, terrorism, extremism, and espionage reflect the culmination of a complex interaction between numerous factors that include those related to individuals’ underlying propensity toward disloyal and hostile actions toward the United States and their ability to translate that propensity into concrete actions—what are known as the processes of radicalization and mobilization in the terrorism literature—and the success of methods used by the United States to mitigate these factors. Under this approach toward estimating risk, the focus is on understanding predictors and using that information to develop effective methods of deterrence. We follow this broader literature and review the factors identified as predicting terrorism and espionage and then discuss their relevance to the MAVNI program. We then describe the Army mitigation efforts, including the steps used to screen MAVNI recruits and broader efforts to monitor personnel, both MAVNI and non-MAVNI, while in service. The chapter is organized according to this approach. However, before presenting our review of the factors identified in the literature, we first discuss methodological and data issues that arose in past studies. Data and Methodological Challenges Empirical research on terrorism and espionage is challenging. Terrorist events and incidents of espionage are rare and difficult to predict, those who plot or engage in terrorism and espionage seek to evade detection, and the factors that lead individuals to pursue these acts are difficult to detect. Furthermore, empirical research in this area is relatively new, and so it is still evolving in terms of data and methods. In light of these challenges, concerns about the rigor of past research in this area are not surprising. In 2016, the journal Studies in Conflict and Terrorism published a special issue devoted to these methodological issues (Freilich and LaFree, 2016), though concerns had been raised earlier (Victoroff, 2005). A key conclusion was that studies of terrorism have lagged behind related fields of study, such as criminology and forensic psychology, in the use and analysis of data and in the adoption of sophisticated research methods. Thus, inferences for the purpose of setting policy need to be done with appropriate 1 Even after recent decreases in the number of clearances, about 3 million people held a current security clearance and had access to classified information as of FY 2014 (Office of the Directorate of National Intelligence, 2015). TIWARI_PROD_00002436 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 65 caution and care about the relevancy and importance of these methodological and data issues. Four key areas of concern are identified. First, historically, researchers focusing on terrorism or espionage rarely used statistical methods. Reviews of published research between 1971 and 2003 found that only about 3 percent of studies employed statistical analysis, far lower than in related fields (Lum, Kennedy, and Sherley, 2006), with the key reason being the lack of data or sparse data. Second, terrorism scholars often use interviews and self-reports of terrorist plots and acts, but, in the context of terrorism, such methods potentially suffer from a number of limitations, including obstacles in gaining access to terrorists (raising the issue of selection bias), interviewer bias, generalizability, interviewer effects, lack of coding reliability, and issues related to retrospection (e.g., individuals are only interviewed retrospectively after having committed a terrorist act). Most terrorism studies that employ interview methods fail to note these methodological challenges. Third, studies often rely on open-source databases that are a compilation of terrorist acts or plots from online sources, including media reports, books, court records, watch-group materials, government sources, and even information produced by extremist or terrorist groups. However, these data sources may not be consistently reliable, and different sources may give conflicting information. For example, Behlendorf, Belur, and Kumar (2016) found that media-based data sets captured geographic prevalence of terrorism but severely underestimated the frequency of attacks, biasing toward lethal bombings and discounting nonlethal violent activity. Consequently, selected versions of violence in some countries are either consistently underrepresented or overrepresented in media reports, and media reports are the most-commonly-relied-upon source of data for open-source databases. Fourth, few statistical analyses that use these data sources consider a control group (e.g., nonterrorist groups), attempt to use causal methods, or use statistical methods to address selectivity bias and other sources of bias. According to the studies included in the special issue of Studies in Conflict and Terrorism, the use of more-systematic methods in the terrorism literature is growing and improving. One area in particular where the literature has grown has been in the development of terrorist risk-assessment tools based on observable behavioral indicators, though this literature is also evolving in the face of conceptual and methodological challenges (Monahan, 2012; Meloy and Gill, 2016). We discuss this literature below. Overview of Psychological and Sociological Factors Related to Terrorism and Extremism As mentioned, violent extremism is broadly seen as the culmination of a two-stage process involving radicalization and mobilization. Radicalization is the mental or emotional process by which individuals come to hold beliefs that become the core of one’s TIWARI_PROD_00002437 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 66 The Military Accessions Vital to the National Interest (MAVNI) Program identity and worldview and by which they come to believe that engagement in violence is necessary and justified. Mobilization is the conscious decision to act on those beliefs. A substantial literature exists on the dynamics of these two processes and the factors that contribute to each. Our review focuses on individual-level factors that are thought to be predictors, rather than on factors related to terrorism at the national level (such as gross domestic product, for example). The findings from the literature indicate that psychological and values-based variables are important individual-level predictors, though, again, it is important to recognize that much of the literature is theoretical, and the empirical analyses in this literature are marked by incomplete and potentially biased data, as well as methodological issues. Atran et al. (2014) reports the initial results of studies of Western volunteers for the Islamic State of Iraq and the Levant (ISIL), based on interviews of members of different groups. Their results suggest that volunteers to these forces are generally young males in transitional phases of their lives, such as immigrants, students, those between jobs or girlfriends, and those having left their native homes, who are looking for a new family of friends and who have no traditional religious education but seem to be “born again” into religious radicalism through the appeal of militant jihad. Kruglanski et al. (2014) argue that no single factor or personality alone accounts for radicalization but that the radicalization process reflects the interaction of several factors, including holding sacred values, having a personal grievance and/or a sense of cultural persecution, moral outrage at injustice, and access to a politically active network. LaFree and Ackerman (2009) review empirical studies of the factors associated with terrorism at the individual, group, and national levels. Here we summarize their findings related to individual factors. They divide the factors into psychological and demographic factors. They argue that the former can be encapsulated under the rubric “quest for personal significance,” meaning that such factors as alienation, deprivation, and perceived injustice lead individuals to seek to restore their personal significance and prevent further loss by engaging with terrorism. Furthermore, people who are not in a negative psychological state may still seek the gains in personal significance that joining a terrorist group and engaging in thrilling acts of daring can provide. McCauley and Moskalenko (2008) also identify personal victimization and political grievance as elements related to individual radicalization. Crossett and Spitaletta (2010) provide an extensive review of the main ideas and trends in the psychological and sociological literatures as they relate to the process of radicalization. The study was prepared for the U.S. Army Asymmetric Warfare Group and was intended as a reference guide for the military. The study reviews 16 theories that have been proposed to explain the causes of radicalization and argues that no single theory can explain the process. Based on the array of theories, the study distills the key risk factors from the literature, many of which are in common with those identified in other studies in this emerging literature. The main factors are listed in Table 6.1. TIWARI_PROD_00002438 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 67 Table 6.1 Risk Factors Identified in the Psychology and Sociology Literature Risk Factors Emotional vulnerability, such as the death of a loved one Personal grievance or humiliation Quest for significance Alienation Positive view of violence and/or the benefit of political violence Perceived injustice or perceived threat Dissatisfaction with the status quo of political activism Social ties to those experiencing similar issues Access to resources or external support for radicalization and extreme violence Overview of Demographic Factors Related to Terrorism and Extremism A different approach is to consider the eligibility characteristics for the MAVNI program, particularly demographic characteristics, and review the available literature on whether these characteristics are associated with participation in terrorism. However, past studies, including those cited earlier in this chapter, almost universally conclude that demographic characteristics alone have very poor power in predicting terrorist activity and extremist violence. Indeed, these studies caution against using demographic characteristics, rather than behavior, to predict terrorism. Nonetheless, many studies do consider whether demographic characteristics are associated with terrorism, and we review the key findings here, noting that analyses of these factors are affected by incomplete and potentially biased data, as well as methodological challenges. Chapter Two described the MAVNI eligibility criteria. In short, these criteria include having a critical language or health professional skill, being foreign born and in the United States as a nonimmigrant—usually as a student, being at least a high school graduate; having an AFQT score of 50 or higher, and being between the ages of 17 and 34. While both men and women are permitted to enlist under the MAVNI program, MAVNI recruits are more likely to be male relative to all non-MAVNI recruits (see Table 3.1). Of course, the Army enlists many non-MAVNI recruits who are also male high-school graduates with above-average AFQT scores and between the ages of 17 and 34. LaFree and Ackerman (2009) consider four categories of variables—gender and age, employment and education, marital status and parenthood, and military service— as they relate to the probability of taking part in terrorist acts. The study notes that TIWARI_PROD_00002439 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 68 The Military Accessions Vital to the National Interest (MAVNI) Program there have been relatively few systematic individual-level demographic studies of terrorists. The research that exists seems to support the conclusion that terrorists are very likely to be young and fairly likely to be male. In contrast, marriage, parenthood, education, employment status, and military service do not seem to insulate individuals from participating in terrorism. Available evidence suggests no connection between poverty and terrorism (Krueger and Maleckova, 2003). In fact, much of the prior research shows that those who participate in terrorist acts are somewhat better educated and more prosperous than the populations from which they are drawn (Berrebi, 2007). In studies of jihadi-related terrorism, the prototype is a young male from a middle-upper-class background with some university education. Kepel (2005, p. 112) concludes that contemporary jihadi terrorists are not so much drawn from the “downtrodden and ill-educated” as “they are the privileged children of an unlikely marriage between Wahabism and Silicon Valley.” Research on non-jihadi terrorism provides similar results. Finally, Monahan (2012, 2017) provides a review of the literature on the evidence in support of individual risk factors for terrorism, including demographic, psychological, and sociological factors. Specifically, he reviews the evidence on the role of age, gender, marital status, social class, major mental illness, prior criminal record, suicidality, substance abuse, personality disorder, and personality. As in the studies mentioned previously, Monahan finds that the evidence supports age, gender, and possibly marital status as risk factors, but he states, Indeed, the strongest empirical findings are entirely negative: terrorists in general tend not be impoverished or mentally ill or substance abusers or psychopaths or otherwise criminal; suicidal terrorists tend not to be clinically suicidal. In no society studied to date have personality traits been found to distinguish those who engage in terrorism from those who refrain from it (Monahan, 2012, p. 14). With regard to criminality, the studies that Monahan reviewed found no evidence of major criminal behavior, such as murder, and police records indicate that violent radicals are generally free of a record of past crime. He also reviewed analyses of the post-release rate of actual or suspected terrorism among detainees released from Guantanamo and found that the rate varied from 6 to 25 percent, though he notes that because not all of those released from Guantanamo were ever adjudicated as having committed terrorism, it is unclear what these rates actually mean. Immigration and Foreign-National Status As noncitizens who entered the United States on student visas, MAVNI enlistees may constitute a security threat. Recent changes in the program by OSD reflect this concern (Office of the Under Secretary of Defense for Personnel and Readiness, 2016). Studies suggest that the relationship between nationality, immigration, and terrorism is a complicated one. TIWARI_PROD_00002440 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 69 On the one hand, tabulations of terrorist acts since the September 11, 2001, attacks, such as those based on the Global Terrorism Database developed by the University of Maryland Study of Terrorism and Responses to Terrorism (Rivinius, 2014), find that most terrorist attacks occur in just a few countries, with more than half in 2013 carried out in Iraq, Pakistan, and Afghanistan. Tabulations by the Institute for National Security Studies (Tel Aviv University) found that virtually all of the 452 suicide attacks in 2015 worldwide were attributable to Muslim extremists, such as ISIL, in the Middle East, Africa, and Southeast Asia (Schitrit, Yogev, and Schweitzer, 2016). Such tabulations raise the question of whether immigration from these countries or regions could be a means of diffusing terrorism to the United States. Bove and Bohmelt (2016) explicitly consider whether migration is a vehicle for transmitting and diffusing terrorism from one country to another, using data collected for 145 countries covering 1970 to 2000. They estimate a series of spatial temporal autoregressive models that allow them to test whether a country’s level of terrorist attacks is affected by other countries’ levels of terrorism and whether migration is the mechanism by which this diffusion process occurs. Their study focuses on immigrants (i.e., permanent residents) rather than individuals eligible for MAVNI (i.e., nonimmigrants, such as refugees and those on student visas). They find empirical support for the hypothesis that immigration is an important vehicle for the diffusion of terrorism from one country to another, especially immigration from terrorist-prone states. However, they do not find that migrant flows per se positively affect terrorism in a country. In fact, they find the opposite; terrorism is lower in countries that receive more immigrants. From these results, they conclude that it may be counterproductive to implement overly restrictive immigration policies that are broadly applied and do not discriminate between migrants that come from terrorist-prone countries. But they also argue that discriminating among migrants and using overly restrictive entry criteria can run counter to legal and moral structures of Western countries and can prevent a country from reaping the positive benefits of immigration. On the other hand, the relevancy of incidence of terrorism overseas may be limited, given that the MAVNI program enlists foreign-born nonimmigrants who are already in the United States. The distinction of whether terrorist acts are perpetrated abroad or in the United States is relevant because, to be in the United States, an individual must have undergone a set of screenings, including DHS screens and military screening, in the case of military accessions. Tabulations from databases of terrorist cases among those in the United States indicate that the number of such incidents is quite small overall, and, as a fraction of total incidents, the number perpetrated by nonimmigrants and refugees is also quite small. The FBI keeps a list of incidents of terrorism and extremist violence in the United States between 1980 and 2005. While the immigrant statuses of perpetrators are not indicated, tabulations of these incidents indicate that nearly all of them were associated with non-Muslim groups, and virtually all incidents involved “special-interest terrorTIWARI_PROD_00002441 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 70 The Military Accessions Vital to the National Interest (MAVNI) Program ists” that included those active in animal rights and environmental movements and not those obviously associated with immigrants (U.S. Department of Justice, 2005). Leikin and Brooke (2006) have similar findings. They examined the biographies of 373 terrorists charged and convicted and/or killed in Western Europe and the United States between 1993 and 2004 and found that more than 40 percent of terrorists were Western nationals, while fewer than half were born in the Middle East.2 These tabulations capture the early years after the 9/11 attacks but not incidents since 2005. New America maintains a database of homegrown terrorism since the 9/11 attacks that consists of individuals accused of jihadist terrorism–related crimes, and the database extends beyond the FBI data, through 2016. The number of jihadist terrorist cases between 2001 and 2016 is quite small: 388 cases, according to New America (2016), with only 6.7 percent of these cases attributable to visa categories—refugees and nonimmigrants—from which MAVNI recruits are drawn, while 81 percent were carried out by U.S.-born citizens. That said, the citizenship status of about 10 percent of terrorists in the database is unknown, so the proportion of incidents carried out by refugees and nonimmigrants could be higher than 6.7 percent. In general, these tabulations suggest that the proportion of homegrown jihadi terrorist cases attributable to visa categories recruited under the MAVNI program is quite small. The General Accountability Office (2011) reports somewhat higher percentages of noncitizens among terrorist incidents, though still less than 50 percent. It provides tabulations by the Department of Justice of the number of individuals convicted as a result of an international terrorism investigation since the 9/11 attacks. The tabulations include 399 convictions, of which 43 percent were attributed to aliens with or without legal status, while the remaining 57 percent were attributed to U.S. citizens, naturalized citizens, aliens brought to the United States for prosecution, or persons of unknown status. Nowrasteh (2016) focuses on convictions of foreign-born terrorists and not on convictions of U.S. citizens. Still, his is the only study that shows results by visa category; specifically, he provides tabulations of foreign-born terrorists between 1975 and 2015, by visa category. Of the 154 foreign-born convicted terrorists in his database, he finds that 27.8 percent were in the United States as a student, refugee, or asylee (i.e., MAVNI categories). He found that 35.1 percent were lawful permanent residents (e.g., green card holders), and the remaining 37.1 percent were a mix of tourists, undocumented aliens, and other categories. While useful, these tabulations provide no information on the relative risk of someone in the MAVNI population being involved with terror. A tabulation of risk would provide information on the underlying propensity of an individual in the non2 The lower incidence of terrorists from the Middle East in these tabulations using data on terrorists in Western Europe and the United States from between 1993 and 2004 contrasts with the high fraction of suicide attacks in 2015 worldwide attributable to Muslim extremists in the Middle East, Africa, and Southeast Asia. The difference reflects not only differences in dates but also geography and data methods. TIWARI_PROD_00002442 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 71 immigrant population to engage in terrorist acts compared with the underlying propensity among citizens. For example, although the number of nonimmigrant terrorists is relatively small, the size of the nonimmigrant population is also relatively small. Thus, the terror risk could be higher among nonimmigrants than among citizens. Unfortunately, as we discussed in Chapter Five, we do not have robust estimates of the size of the nonimmigrant population that would enable us to estimate relative risk. Furthermore, the number of terrorist cases is so small that differences in risk estimates will also be extremely small, calling into question the usefulness of measuring such differences.3 In sum, while available data indicate that those who engage in terror are less likely to be a nonimmigrant, these data do not allow us to assess whether nonimmigrants are at a lower or higher risk of engaging in terror. The literature on terrorism has grown in recent years to include the development of risk assessment tools based on observed behavioral indicators. The next subsection briefly reviews this literature. Literature on Observed Behavioral Indicators Behavioral indicators are metrics of risk factors that permit an assessment of the likelihood that an adverse outcome, such as engaging in terror, will occur in a population. These indicators are the “data” for the appraisal of risk (Borum, 2015) and should meet two key criteria: (1) They are associated with the likelihood that the outcome will occur in a population, and (2) They have “potency,” meaning that they discriminate well between those who have a higher and lower probability of an adverse outcome (Borum, 2015; Monahan, 2017; Kraemer et al., 1997). Risk assessment tools based on behavioral indicators have been developed over the years by criminologists and forensic psychologists to assist law enforcement agencies in assessing the risk of violence and crime when they make decisions about the termination of detention of criminals. A question debated among terrorism researchers is whether assessment tools of “common” violence can be generalized to the risk assessment of terrorism. As noted by Monahan (2012), whether it is possible to generalize depends on whether there are valid individual risk factors of terrorism; he concludes that none of the main dimensions of risk of common violence—criminal history, an irresponsible lifestyle, psychopathic and criminal attitudes, and substance abuse—are risk factors for committing terrorist acts. Reviewing much of the same literature as discussed above, he also concludes that there is little empirical evidence, except for age, gender, and possibly marital status, for other commonly hypothesized risk factors of terrorism, such as mental illness and poverty. Monahan (2012, 2017) articulated a con3 Nowrasteh (2016) provides an estimate of risk among different visa categories and finds that the risk of dying in an attack by a foreign-born terrorist between 1975 and 2015 was 1 in 3.6 million. However, he does not report the risk for citizens. TIWARI_PROD_00002443 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 72 The Military Accessions Vital to the National Interest (MAVNI) Program ceptual framework, drawing on the literature and risk factors summarized in Table 6.1, for developing risk assessment tools of individual terrorism. He enumerates these into five domains: ideology, affiliation and social networks, grievances, moral emotions and sacred values, and identities and “oneness” with a group. He notes that an important area for future research is to move away from a set of “promising” risk factors to ones that are validated by empirical research, in the same manner that risk assessment tools have been doing for common violence. Researchers have also developed more-complex models of risk factors that show multiple pathways to terrorist acts and the interaction among risk factors. For example, research summarized in Davis and Cragin (2009) shows decision trees that identify the combination of risk factors that increase the likelihood of terrorism. In recent years, researchers have begun developing instruments to code behaviors for terrorism risk based on these risk factors. These include the Violent Extremist Risk Assessment, the Multi-Level Guide for the Assessment and Management of GroupBased Violence, the Extremist Risk Guidelines, and the Terrorist Radicalization Assessment Protocol.4 Other researchers have developed computation models (Davis and O’Mahony, 2013) and frameworks (Bulling et al., 2008) that enable analysis of how factors combine and interact to result in terrorism. DoD and the Army have also begun to develop a set of behavioral indicators for preventing terrorism and insider threats. We discuss the Army and DoD efforts later in this chapter. Researchers have also considered observable behavioral indicators related to suspicious activity, and not just individual risk factors related to personality, networks, or beliefs. Strom et al. (2017) analyzed data on 150 completed and foiled terrorist plots against the United States between 1995 and 2012 and found that more than 80 percent of foiled plots were due to observations by law enforcement or the general public. Their findings suggest that prevention efforts should focus on reports of behavior that is genuinely threatening, is atypical (i.e., not appropriate to the circumstance, such as wearing a heavy coat on a warm day), reflects a significant commitment toward carrying out the plot, comes from a credible source, and contains specific details. Overview of Factors Related to Espionage While the focus of concerns about security risk since the 9/11 attacks has been primarily on terrorism, security risk also encompasses espionage. Espionage refers to the betrayal of trust, by individuals who have been formally vetted and are placed in positions of trust, by making classified information available to an unauthorized person or using it in a way that hurts the interests of the United States or that benefits a foreign 4 See Meloy and Gill (2016) for a discussion of these tools, particularly the Terrorist Radicalization Assessment Protocol. TIWARI_PROD_00002444 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 73 government (Charney and Irvin, 2014). In 1985, the Department of Defense Security Review Commission, also known as the Stillwell Commission (after the chairman, Army General Richard Stillwell) reviewed the effectiveness of security clearance procedures and found that clearance decisions were often subjective (Fischer, 2000). The commission recommended that personnel security clearance policies should be grounded in data and analysis using scientific methods. This recommendation led to the creation of two organizations: PERSEREC and the Community Research Center; the Community Research Center oversees a research effort known as Project Slammer. An important contribution of PERSEREC was the development of a database on all Americans involved in espionage since the end of World War II. The database permits analysis of the demographic characteristics of spies and their motivations. Project Slammer focuses on conducting interviews with incarcerated spies. Together, these sources provide information on the factors related to espionage by Americans, conditional on being in the database. Though recent critiques of the data and methods used in the terrorism literature, discussed above, did not directly address these issues in the espionage literature, similar issues are likely to arise in that literature as well. The factors related to espionage have commonalities with those related to terrorism, but there are also significant differences. Like terrorism, acts of espionage by Americans (against the United States) are relatively few; the PERSEREC database includes only 173 instances overall, with 37 cases since 1990 and 11 cases since 2000 (Herbig, 2008). Consider first the demographic and job characteristics of those who have engaged in espionage. Like terrorists and those who have engaged in extreme violence, spies tend to be male and better educated (at least 12 years of education), though the proportion that is female has been increasing in recent years (Herbig, 2008). On the other hand, the PERSEREC espionage database indicates that spies tend to be older, with 83 percent over age 30 among cases between 1990 and 2007. The database also provides information on the percentage of cases that involved those employed by the military during the espionage. Among the 37 cases occurring since 1990, 30 percent were military, 34 percent were civil service, 9 percent were contractors, and 26 percent had a job unrelated to national security (percentages do not total 100 because of rounding). Among the military cases, the overwhelming number were enlisted personnel rather than officers, with enlisted individuals being most likely in grades E-4 to E-6. American spies are more likely to be citizens born in the United States than naturalized citizens, though the percentage who are U.S.-born citizens has decreased from 84 percent between 1980 and 1989 to 65 percent between 1990 and 2007 (Herbig, 2008). The high percentage of U.S.-born perpetrators is similar to the New America statistic of 81 percent cited above for jihadi terrorists. That said, the PERSEREC database indicates that, since 1990, the percentage of spies with foreign attachments (relatives or close friends) was 58 percent, and those with foreign business or cultural ties was about 50 percent, up from less than 10 percent before 1990. When asked TIWARI_PROD_00002445 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 74 The Military Accessions Vital to the National Interest (MAVNI) Program about their motivation, the top reason among cases after 1990 was “divided loyalties,” defined as holding and acting on an allegiance to a foreign country or cause in addition to allegiance to the United States. Finally, a significant fraction of espionage cases was carried out by those who did not hold a security clearance. Between 1990 and 2007, 37 percent of spies did not hold a security clearance and did not have authorized access to classified information when they committed espionage-related offenses. In terms of motivations for espionage, Herbig (2008) provides tabulations from the PERSEREC database, based on subjective assessments of researchers who developed the database using open sources. Before 1990, the data suggest that spying for money was more common, but after 1990, money was rarely the sole motivation. Furthermore, the PERSEREC data indicate that in recent years, spying has not been particularly lucrative, with 81 percent receiving no payment for spying among cases since 1990. As mentioned, divided loyalties, foreign influences, and allegiance to foreign countries became the most important sole motivations among cases of espionage after 1990—57 percent of cases indicated a sole motivation (rather than multiple motivations). After divided loyalties and money, the third-most-common motivation is disgruntlement, usually caused by the person’s relationship or treatment in the workplace and the associated desire to take revenge. Finally, the PERSEREC data also indicate the importance of life events that trigger a personal disruption or crisis as a motivation for spying. These events include a death of a family member or close friend, divorce, separation or marital discord, and physical relocation. Of the 173 cases in the PERSEREC database, 57 percent experienced one or more trigger events in the months before attempting espionage (Herbig, 2008). Several of the motivators of espionage are similar to those identified in the terrorism literature. Having a personal grievance or disgruntlement was also identified as a motivator, though workplace disgruntlement was not specifically discussed. Similarly, life triggers or being in a transitional phase of one’s life has also been identified as a motivator in the terrorism literature. Monetary gain is not typically mentioned as a motivator of terrorist or extremist violence, consistent with money being less of a motivator among more-recent cases of espionage. Discussion and Implications for the MAVNI Program Two findings from the literature suggest a potential area of risk for the MAVNI program, or at least an area for greater scrutiny. First, Bove and Bohmelt (2016) find that, although countries like the United States that experience more immigration also experience less terrorism, immigration from high-risk countries can be a source of transmission of terrorism in general (not just to the United States). Second, tabulations of espionage among Americans by Herbig (2008) indicate that divided loyalties, foreign influences, and allegiance to foreign countries were the most important sole motivaTIWARI_PROD_00002446 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 75 tions among cases of espionage in more recent years. These two findings, by themselves, suggest that immigrants, especially from terrorist-prone countries and/or with extensive foreign influences, could be a source of security risk. Unfortunately, four issues hamper our ability to draw definitive conclusions from these specific studies or, more broadly, to definitively address the security risk question for the MAVNI program posed above. First, the literature is still evolving, especially in terms of data and methods, so the evidence supporting any conclusions is not always solid. Second, the risk factors identified by the literature are generally psychological, as shown in Table 6.1. These factors are difficult to measure and, even if they were measurable, without information about their incidence and importance among potential MAVNI recruits specifically, they provide general guidance but little information about the security risk of the MAVNI program. Third, the literature universally cautions against using such demographic factors as immigration status to predict radicalization and mobilization, favoring instead specific behaviors related to the psychological factors listed in Table 6.1 and observed behavioral indicators. In addition, the literature argues that terrorist organizations will respond strategically by changing the demographics of those who perpetrate terrorism if they believe that demographic factors are used to screen potential terrorists. Finally, the demographic characteristics that do seem to be correlated with terrorist activity, specifically being male and somewhat better educated, are characteristics that MAVNI recruits have in common with most other Army recruits. The characteristic that is specific to MAVNI—being a nonimmigrant—is not particularly prevalent in terrorism databases; New America estimated that only 6.7 percent of accused terrorists were nonimmigrants, while the General Accountability Office reports that 43 percent of accused terrorists were foreign nationals. However, as mentioned earlier, these data do not allow us to assess whether nonimmigrants are at a lower or higher risk of engaging in terror. In many ways, the terrorism and espionage literature provides no real helpful information to allow prediction about whether a specific group, such as MAVNI recruits, is at heightened risk of perpetrating terrorist acts or espionage. However, the literature has been quite influential in helping the policy community develop criteria for screening and monitoring personnel to mitigate security risk. Indeed, many of the findings in the espionage literature discussed inform the questions included in security clearance screening, such as the NIAC and SSBI. For example, the SSBI requires completion of form SF-86, a form that asks individuals questions related to their foreign ties and potential financial difficulties. Unfortunately, the evidence on whether factors related to espionage are also relevant factors related to terrorist activity is lacking. TIWARI_PROD_00002447 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 76 The Military Accessions Vital to the National Interest (MAVNI) Program Mitigation Efforts: Screening and Monitoring We turn next to military screening and monitoring efforts in the context of the MAVNI program. Substantial effort is devoted to screening military recruits, and, as mentioned, MAVNI recruits undergo even more-rigorous screening than non-MAVNI recruits. Efforts are also under way in the Army to regularly monitor military personnel to identify and prevent insider threats. We review these efforts in this subsection. An important issue in this field is that research on the effectiveness of these mitigation efforts in preventing attacks and/or plots is lacking. Instead, much of the research on effectiveness has focused on how well new and existing processes are implemented— e.g., whether all of the screening steps are being followed (General Accountability Office, 2014; Youpa, Marshall-Mies, and Carney, 2004). Consequently, what is known about the effects of screening and monitoring on risk mitigation is sparse. Screening The Army and DoD use several screening processes to determine eligibility for enlistment or clearance. As discussed in Chapter Two, MAVNI recruits undergo additional screening requirements related to their immigration status, language skills, and security risk. With respect to the final requirement, MAVNI undergo two PSIs once they enter DEP: the NACLC or NIAC screen and the SSBI. They must also complete a CI-focused security review and/or issue-oriented polygraph, as needed to resolve any foreign influence or foreign preference concerns (Department of the Army, 2014).5 Other recruits also undergo these screens, but only if they are to receive a security clearance. That is, these screens are used for non-MAVNI recruits to determine whether they can be put in positions of trust and given access to sensitive material. Unlike these recruits, MAVNI recruits who undergo these screens do not receive a security clearance. The NACLC and NIAC are background checks with federal agencies (such as the FBI and the National Crime Information Center), local law enforcement, records of foreign travel, and credit agencies. The SSBI involves an investigation into an individual’s family, education, residences, employment, mental health, military service, criminal activity, drug and alcohol abuse, credit, and external activities. The investigation is intended to focus on the concerns in the 13 Uniform Adjudicative Guidelines, adopted in 2005 (Hadley, 2005). These guidelines require consideration of foreign influence and ties, personal conduct, financial distress, alcohol consumption, drug involvement, personality disorders or mental illness, divorce or family court issues, criminal conduct, allegiance to the United States, and outside activities. 5 Beginning in FY 2017, all MAVNI recruits must undergo a CI security review based on the findings of the NIAC and form SF-86, an input to the SSBI. TIWARI_PROD_00002448 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 77 While we have no information on the effectiveness of screening in preventing terrorist activity or espionage, a minimum criterion for screening effectiveness is that screening actually occurs. With this in mind, we examined the extent to which MAVNI recruits complete their security screening before shipping to basic training using Army data through September 2015. As a benchmark, we compared completion rates with the rates for a variety of non-MAVNI recruits, including (1) all nonMAVNI recruits, (2) non-MAVNI high-quality recruits, (3) non-MAVNI noncitizen recruits, (4) non-MAVNI recruits who receive an SSBI for security clearance purposes, and (5) non-MAVNI high-quality recruits who have initial occupation codes that are among the 11 most common initial occupation codes for MAVNI recruits.6 We label the last group “HQ Top Occ Non-MAVNI.” Figure 6.1 shows the percentage of accessions that shipped to basic training without a completed security investigation in 2013 and 2014. Only 1 or 2 percent of MAVNI accessions shipped to basic training without a completed investigation. The percentage of non-MAVNI recruits that shipped without a completed investigation is higher for all of the groups we considered, including non-MAVNI recruits receiving an SSBI, a group that also undergoes the longer SSBI process. But even for the nonMAVNI groups, the percentage who shipped without completing their investigation is quite small. Figure 6.2 shows that virtually all recruits, both MAVNI and nonMAVNI, have a completed investigation by the time they complete basic training and before they start advanced individual training. In conversations, the SMEs expressed concern about the length of time required to complete the screening of MAVNI enlistees. Longer screening can mean a longer time in DEP and more opportunity for a MAVNI recruit’s visa to expire before entering service. We found that the data on length of time between when an investigation is initiated and when it is complete is rather “noisy,” producing some irregularities. For example, we found small numbers of cases of negative values—i.e., the initiation date occurred after the completion date—as well as cases where the value exceeded 1,000 days (over three years). We dealt with these outliers by winsorizing the data on completion time at the 5th and 95th percentile levels.7 6 These 11 codes are 68W (health care specialist), 92Y (unit supply specialist), 92A (automated logistical specialist), 36B (financial management technician), 68X (mental health specialist), 91B (wheeled vehicle mechanic), 68E (dental specialist), 74D (chemical, biological, radiological, and nuclear specialist), 91D (power-generation equipment repairer), 68J (medical logistics specialist), and 92G (culinary specialist). Approximately 53 percent of MAVNI recruits have one of these 11 codes as the initial code, though 20 percent of MAVNI recruits are missing an initial code. These tabulations are based on RAND Arroyo Center computations using a merged data set of RA Analyst and TAPDB data. The tabulations of career management field in Chapter Two were based on Armyprovided data on the MAVNI program. 7 Winsorizing involves replacing values above the 95th percentile and below the 5th percentile with the values of the distribution of investigation times at these cutoff points. Doing so prevents outliers from having a disproportionate impact on the calculated mean. TIWARI_PROD_00002449 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 78 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 6.1 Percentage of Accessions with Incomplete Security Investigation Before Ship Date to Basic Training Percentage of accessions with incomplete security investigation 10 9 8 2013 2014 7 6 5 4 3 2 1 0 MAVNI Non-MAVNI High-quality non-MAVNI Noncitizen non-MAVNI SSBI non-MAVNI High-quality most common occupations, non-MAVNI SOURCE: Authors’ calculations using Army RA Analyst and TAPDB data. RAND RR1976-6.1 Figure 6.2 Percentage of Accessions with Incomplete Security Investigation Before Start of Advanced Individual Training Percentage of accessions with incomplete security investigation 10 9 8 2013 2014 7 6 5 4 3 2 1 0 MAVNI Non-MAVNI High-quality non-MAVNI Noncitizen non-MAVNI SSBI non-MAVNI High-quality most common occupations, non-MAVNI SOURCE: Authors’ calculations using Army RA Analyst and TAPDB data. RAND RR1976-6.2 TIWARI_PROD_00002450 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 79 Figure 6.3 shows the mean (winsorized) number of days required to complete a security investigation. Because MAVNI recruits undergo more screening, including the longer SSBI, we would expect the amount of time to complete an investigation to be longer on average. The results in Figure 6.3 support this hypothesis. On average, MAVNI recruits required 63 days to complete the investigation process in 2013 and 48 days to complete the process in 2014. Completion of the investigation required many fewer days for non-MAVNI recruits, even those requiring an SSBI. This finding is consistent with the longer months in DEP for MAVNI versus non-MAVNI recruits shown in Table 3.1. These results suggest that, at a minimum, MAVNI recruits have nearly all completed the security investigation process before they ship to BCT, though the process takes longer than for non-MAVNI recruits. That said, the security investigation process that MAVNI recruits undergo, particularly the NIAC and SSBI, are geared toward screening out individuals who should not have access to sensitive information. Many of the issues considered in the NIAC and SSBI are factors that the literature has identified as related to espionage, such as foreign loyalties, criminal behavior, and financial distress. These screens are not specifically oriented toward factors identified in the literature as being related to radicalization, mobilization, and extremist behavior. MAVNI recruits are required to undergo a CI security interview, but the interview is based on the findings of the NIAC and the form SF-86 that are input to the SSBI, both designed Figure 6.3 Mean Winsorized Days to Complete Security Investigation 80 2013 2014 70 Number of days 60 50 40 30 20 10 0 MAVNI Non-MAVNI High-quality non-MAVNI Noncitizen non-MAVNI SSBI non-MAVNI High-quality most common occupations, non-MAVNI SOURCE: Authors’ calculations using Army RA Analyst and TAPDB data. RAND RR1976-6.3 TIWARI_PROD_00002451 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 80 The Military Accessions Vital to the National Interest (MAVNI) Program to focus on factors found to be related to espionage but not necessarily radicalization and mobilization. As mentioned earlier in the chapter, past research finds little or no support for mental illness, substance abuse, personality disorders, or criminal conduct as risk factors for terrorism. More broadly, these background checks generally do not include the array of psychological, sociological, and observed behavioral indicators that the literature does suggest are associated with terrorism (listed in Table 6.1). It is important to recognize that our tabulations do not include the time period after the data breach in the spring of 2015 of the Office of Personnel Management, the government organization that oversees the SSBI process. Anecdotally, SSBI processing time has increased as a result of processes put in place after the breach, so our tabulations may understate processing time for more recent periods. Monitoring A study of terrorist activity in the military argued that radicalization of individuals after they enlist is likely to pose a significant threat (Buck et al., 2005).8 Such radicalization may occur among MAVNI personnel or non-MAVNI personnel. After entering the Army, all soldiers are subject to continuous evaluation and monitoring programs. With respect to MAVNI, Department of the Army policy (2012) requires that, at a minimum, annual checks of MAVNI recruits are made through the Automated Continuous Evaluation System (ACES). In cases of questionable results of any of the ACES checks, a follow-up CI security review is also required. ACES is an automated computer system that collects data from more than 40 government and commercial databases and flags issues of potential security concern (Herbig, Zimmerman, and Chandler, 2013). ACES checks information about finances, criminal activity, mental stability, and other information, in accordance to the 13 Uniform Adjudicative Guidelines (Hadley, 2005), similar to the initial screening process for MAVNI. As with the SSBI process for initial screening, ACES focuses on factors that have been identified in the literature as being relevant for the screening and detection of espionage, rather than for terrorism. Though foreign ties and loyalties are part of the screening, ACES generally does not focus on the behavioral indicators and risk factors associated with terrorism, such as emotional vulnerability, sacred values, alienation, quest for significance, and a positive view of political violence. It is possible these issues can and do emerge in the CI security review, but only if they are follow-up or flagged issues in the SSBI or ACES process. In other words, the CI review does not necessarily target the risk factors of terrorism. 8 The same authors found evidence of some radicalized individuals within the military, although they found no evidence that any of these individuals had enlisted for the purpose of carrying out acts against the United States (Buck et al., 2005). Some of these individuals had ties to foreign governments or sources of influence; others had ties to white supremacist or white nationalist groups or to domestic militias. TIWARI_PROD_00002452 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Estimating Security Risk 81 Since 2010, and especially since the November 2009 Fort Hood shooting, numerous efforts throughout the government, including the Army, have been made to identify and prevent insider threats, defined as threats that an insider will use his or her authorized access, wittingly or unwittingly, to do harm to the security of the United States (Presidential Memorandum, 2012). The Army established the Insider Threat Task Force in 2010, resulting in Army Directive 2011-14 and AR-525-2 that developed a management framework for protection policies. The Army has since developed an Insider Threat Program as a result of Army Directive 2013-18, following the unauthorized disclosure of classified information by Chelsea Manning and a former National Security Agency contractor. The program requires annual threat awareness training of all Army personnel and provides a set of guidelines and products to identify insider threat risk. As an example of Army guidelines, Figure 6.4 presents a tactical guide developed by the Army Asymmetric Warfare Group to assist soldiers and commanders in recognizing behavior that is associated with radicalization and in determining how to report and/or counsel in such cases. This tactical reference guide accords well with the types of behaviors that are associated with radicalization in the literature (listed in Table 6.1). While we do not discuss them in this context, there are additional programs designed to recognize radical and extremist behavior. For example, the Army iWatch program was developed to promote reporting at the local level. As is the case with initial screening, there is little evidence available on whether these programs actually reduce risk. These additional programs are described in materials prepared by the Office of Provost Marshal General.9 Summary There are multiple programs and steps in place with the capacity to screen and monitor individuals. In the specific case of MAVNI personnel, required security screens were nearly always carried out prior to the shipping date to basic training. As we might expect, the time to complete investigations on MAVNI enlistees is substantial; SMEs indicate that this can be an issue for potential enlistees who may face challenges to remain in legal status throughout the screening process. In the case of the Army’s current monitoring programs, the Insider Threat Program uses an approach that is consistent with the literature, in the sense that monitoring focuses on detecting behaviors that have been found to be predictive of radicalization and mobilization. However, the entry screening and automated monitoring programs used to screen and monitor MAVNI tend to use tools that are relevant for security clearance and espionage. Whether they are also relevant for terrorism is an open question. 9 See, for example, the Army Antiterrorism Officer Ready Reference (U.S. Army, undated). TIWARI_PROD_00002453 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 82 The Military Accessions Vital to the National Interest (MAVNI) Program Figure 6.4 Army Asymmetric Warfare Group Tactical Reference Guide for Risk Mitigation Radicalization into Violent Extremism OBSERVE Personal Issues Indicators that should be observed for the subject’s predisposition to radicalization and violence FLAG Possible Radicalization ACTION Prior to Violent Activity The subject’s actions that should encourage leaders to investigate and bring to the attention of seniorranking personnel Actions conducted by the subject that would indicate violent or terroristic planning activities that warrant investigation ■ Complains about bias ■ Is sympathetic to radical groups ■ ■ Advocates violence beyond what is “normal” ■ Visits extremist websites/blogs ■ ■ Exhibits abrupt behavioral shifts Establishes website/blog to display extremist views ■ ■ ■ Needs empowerment ■ Speaks about seeking revenge ■ ■ Attends rallies for extremist causes ■ ■ Associates with known radicals ■ Is frustrated with mainstream ideologies Exhibits extreme religious interolance ■ ■ Experiences personal crisis and does not properly recover Is personally connected with a grievance ■ Cuts ties with family and friends ■ Demonizes others freely ■ Isolates self from unit members ■ Rhetoric is more intense and concentrated ■ ■ Attempts to recruit others to extremist causes ■ ■ Is socially withdrawn ■ Believes in government conspiracies to the point of paranoia ■ ■ ■ Lacks positive identity with country, unit, family, or friends Exhibits sudden reclusiveness OBSERVE – Notify chain of command of ideologies that could threaten unit cohesion and morale – Counsel soldier in an effort to help and assist with personal needs IDEATION ■ ■ ■ Suddenly acquires weapons Organizes protests inspired by extremist ideology Takes part in criminal activity or has trouble with law enforcement Advocates violence as a viable option for various situations Shows a sudden visual shift from radical to “normal“ behavior to conceal radical behavior Takes suspicious or unreported travel (inside or outside of the continental United States) Stores or collects mass weapons or hazardous materials Verbally indicates hatred for the United States and/or the Constitution Exhibits new interests in public or government facilities Inquires about weapons of mass effects FLAG ACTION – Notify chain of command of possible radical or extremist activity – Counsel soldier on dangers of activities ACTION – Immediately notify CID and/ or federal law enforcement – If the situation is critical, confine subject to barracks under guard and remove all items that could harm others or self SOURCE: Asymmetric Warfare Group, 2011. NOTE: CID = Criminal Investigation Division. RAND RR1976-6.4 TIWARI_PROD_00002454 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. CHAPTER SEVEN Final Thoughts An important policy question is whether MAVNI is a cost-effective program that does not pose additional risk. Our research, which focused primarily on MAVNI enlistments with CFL skills, indicated that MAVNI has been a cost-effective recruiting program for the Army. The cost-effectiveness results from the relatively low attrition rates of MAVNI recruits. Their better performance stems from their faster promotion rates and their higher likelihood of having characteristics associated with performance, such as more education and higher aptitude scores. The program taps a potential recruiting pool that could probably sustain a modest increase, though we were unable to estimate the size of the potential pool precisely or the number of additional MAVNI recruits that would be possible in the absence of the cap. Our research was also unable to estimate the specific security risk associated with MAVNI recruits, but our review of the literature indicates that MAVNI recruits undergo extensive screening relative to the typical Army recruit. The risk factors that are the focus of the screening are ones generally found to be associated with espionage rather than terrorism, per se. In particular, past research finds little or no support for substance abuse, poverty, personality disorders, or criminal conduct (the focus of much of the screening process) as risk factors for terrorism, while, at the same time, the background checks generally do not include the psychological, sociological, and observed behavioral indicators that the literature does suggest are associated with terrorism, such as belief systems and networks that condone the use of violence. That said, foreign ties and loyalties are risk factors that are included in the screening. While MAVNI recruits undergo the screening required for a security clearance, they do not actually receive a security clearance at entry. Our research includes information from a broad range of sources, but we used unclassified data exclusively, with the goal of making the analyses available to the broadest possible audience. Our research indicates that the Army could modestly expand the MAVNI program to cost-effectively increase high-quality enlistment supply to the Army. The analysis of the performance and cost of the MAVNI program was based on Army data from 2009 to 2015. There are several indications to suggest that, over this period, the Army’s implementation of the MAVNI program primarily focused on expanding 83 TIWARI_PROD_00002455 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 84 The Military Accessions Vital to the National Interest (MAVNI) Program high-quality enlistments to the Army rather than expanding critical language capability in the RA. First, the share of MAVNI accessions entering the U.S. Army Reserve rather than the RA has been quite large in recent years (Table 2.3), raising questions about whether MAVNI is meeting a requirement in the RA. Second, the top languages for MAVNI recruits include some languages, such as Nepalese and Tagalog, for which the need for a sizable number of recruits is not obvious (Table 2.4). While MAVNI recruits provide cultural knowledge and diversity in addition to language capability, and MAVNI recruits can be assigned to special operation units, MAVNI recruits were typically in occupations that do not require language capability, such as supply and logistics. OSD made significant changes to the MAVNI program when it extended the program through 2017 and lowered the MAVNI cap. These changes require that the Army and the other services implement the program in a way that focuses more on meeting critical skill requirements and less on meeting overall high-quality accession requirements. For example, OSD now requires that at most 10 percent of annual MAVNI CFL accessions are reserve accessions and at most 10 percent of annual MAVNI CFL accessions are for a single language. Thus, a much larger share of MAVNI CFL accessions must be active component accessions than in the recent past, and MAVNI recruits as a group must have a wider breadth of language skills. Additionally, the new guidance removes some languages from the critical language list altogether and sorts the remaining languages into two different categories, requiring more extensive screening for those who enter with some languages. It is noteworthy that several of the most common languages among previous MAVNI recruits (in particular, Chinese and Korean) no longer appear on the list of critical languages.1 Indeed, speakers of these languages made up about half of MAVNI recruits and about one third of admissions and of foreign students enrolled in U.S. universities in recent years.2 Therefore, the new guidelines likely reduce the pool of potential MAVNI recruits substantially. However, the overall level of interest in the program (as judged by both enlistments and online inquiries) suggests that there will likely continue to be sufficient numbers of potential enlistees under the new guidance, especially given the new lower 2017 cap of 1,400. The new FY 2017 guidance also requires additional security review procedures for MAVNI recruits. The additional OSD requirements will make MAVNI recruiting more costly than we estimate because of the added cost of the additional security review. Furthermore, the additional requirements could make MAVNI recruiting more challenging than in the past, even with a lower MAVNI cap. First, during the past several years, a large proportion of MAVNI enlistees have opted to serve in the U.S. Army Reserve 1 DoD, 2016. While the document refers to “Chinese,” the Army has recruited speakers of Mandarin and of Cantonese. 2 For admissions by country, see Baker (2014); for students, see Ruiz (2014). TIWARI_PROD_00002456 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Final Thoughts 85 (Table 2.5). As mentioned, going forward, the proportion of MAVNI enlistees with CFL skills entering the U.S. Army Reserve will be capped at 10 percent.3 This is likely to cause a contraction in the supply. Also, MAVNI recruiting has, so far, been highly localized, occurring in a handful of locations in the United States that are populated by enclaves of noncitizens with specific language capabilities. The Army may need to expand MAVNI recruiting to additional locations to ensure a wider breadth of languages. This could require training and actively managing a cadre of recruiters who focus on MAVNI recruiting to ensure a more diverse placement of such recruiters and a wider array of languages.4 That said, although our analysis of the potential pool of MAVNI recruits was hampered by lack of data, it suggested that the pool at the national level exceeds the cap, and the Google Trends data, GoArmy.com Google Analytics data, and our interviews with recruiters all suggested substantial engagement by MAVNI searchers and substantial interest in the MAVNI program, especially its offer of expedited citizenship. Also, the cost advantage of MAVNI relative to non-MAVNI high-quality recruits, on a per person-year basis, was significant, and the marginal cost of a MAVNI recruit relative to a high-quality Army recruit is substantially lower. This suggests that the higher cost associated with the new OSD changes would have to be dramatic for MAVNI recruits to be less cost-effective than other high-quality recruits. One approach the Army might consider to further increase the cost-effectiveness of MAVNI recruiting, especially in light of the additional OSD requirements, is to increase the required initial obligation associated with MAVNI. Currently, MAVNI recruits must meet a four-year obligation. Although longer obligations are associated with greater first-term attrition, the differences are fairly modest (Buddin, 2005), and potential recruits do not appear to be dissuaded by longer contracts (Huff et al., 2013). The attractiveness of the citizenship incentive, together with the higher qualification standards of MAVNI recruits, will likely mean that many MAVNI recruits will continue to enlist and that they will continue to attrite at a lower rate than non-MAVNI recruits, even with a longer obligation. The Army’s Insider Threat Program has evolved over time, and guidelines to Army personnel for detecting and reporting insider threats are informed by behavioral indicators and terrorism risk factors identified in the literature. Still, the Army and the other services rely on the SSBI process and CI security review for initial screening of MAVNI recruits, as well as ACES to continuously monitor MAVNI recruits while in the services. These initial screening and automated monitoring approaches could be 3 DoD, 2016. 4 Estimates of the location of noncitizens or immigrants by country of origin could be helpful to the planning process. Such estimates could be based on ACS data. While ACS data undercount students, they may be helpful in locating pools of noncitizens. Conversely, estimates of the number of students in key visa categories, by country of origin, could also assist in the placement of recruiters. TIWARI_PROD_00002457 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 86 The Military Accessions Vital to the National Interest (MAVNI) Program enhanced to better reflect terrorism risk factors and behavioral indicators, such as those identified by the Insider Threat Program. Our interviews with a limited number of SMEs indicated several ways in which the processing of MAVNI recruits might be improved, and the Army should further investigate areas where efficiency could be enhanced. Issues raised include finding ways to reduce the length of the SSBI and to reduce the backlog of SSBIs of MAVNI recruits. They also include giving more visibility to recruiters about MAVNI enlistees’ progress through the SSBI investigation process, and importantly, providing explicit alerts to the recruiter when the process is complete. A related issue is to find ways to reduce the amount of time required to arrange and complete language testing. Another issue raised is the need for additional staff to process MAVNI accessions at USAREC, given the amount of “hands-on” processing steps required, relative to the typical recruit. Of course, with the lower cap, current staffing may be sufficient to process MAVNI recruits. Finally, several topics are promising areas for further research. First, more-precise estimates of the potential pool of MAVNI recruits require the collection of better data. One approach would be a survey of potential MAVNI recruits, such as individuals on student and work visas. Such a survey could also permit better information on the extent to which these individuals would meet qualification requirements, such as two-year residency and health status, and on their propensity to enlist. Second, more research is required on how MAVNI recruits are assigned and used by the Army— specifically, how and when the language and cultural skills of MAVNI recruits are used and in what capacity. This research is likely to involve qualitative research methods, such as interviews and focus groups. Such information could be very helpful in determining how best to design the list of critical languages. Finally, although quite small, the MAVNI program for recruits with HCP skills is a key element of the MAVNI program, and additional research is required to better understand their potential for the Army. TIWARI_PROD_00002458 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. APPENDIX A MAVNI Languages and Nonimmigrant Legal Status Categories MAVNI Languages Figure A.1 provides a list of languages required for enlistment as of FY 2017, along with the languages that were removed from and added to the program in FY 2017. FY 2017 guidance divided languages into two categories, with additional screening required of enlistees who speak a Category 2 language. Nonimmigrant Legal Status Categories Table A.1 provides a list of the nonimmigrant legal status categories eligible for MAVNI and their definitions. 87 TIWARI_PROD_00002459 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 88 The Military Accessions Vital to the National Interest (MAVNI) Program Figure A.1 MAVNI Languages in FY 2016 and Changes for FY 2017 Albanian Amharic Arabic Azerbaijani Balochi Bengali Bulgarian Burmese Cebuano Cambodian-Khmer Chinese Czech Dhivehi (Maldives) French (with citizenship from an African country) Georgian Haitian Creole Hausa Hindi Hungarian Igbo Indonesian Japanese Korean Kashmiri Kurdish Lao Malay Malayalam Moro Nepalese Pahari Persian Dari Persian Farsi Polish Portuguese Punjabi Pushto (also known as Pashto) Romanian Russian Serbo-Croatian Sindhi Sinhalese Somali Swahili Tagalog Tajik Tamil Thai Turkish Turkmen Ukrainian Urdu (with citizenship or birth certificate from Pakistan or Afghanistan) Uzbek Yoruba Key: FY 2016 and FY 2017 Requires additional screening for FY 2017 New for FY 2017 Removed for FY 2017 SOURCE: Office of the Under Secretary of Defense, Personnel and Readiness, 2016. NOTES: Gray text: Language removed from list in FY 2017. Underlined text: Language is new in FY 2017. In FY 2017, languages are categorized as either Category 1 or Category 2. In this figure, bold text indicates Category 2. Enlistees speaking Category 2 languages require additional screening. The Category 1/Category 2 distinction was introduced in FY 2017. The policy documents reference “Chinese,” but the Army has recruited speakers of Mandarin and of Cantonese over the course of the program. RAND RR1976-A.1 TIWARI_PROD_00002460 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. MAVNI Languages and Nonimmigrant Legal Status Categories 89 Table A.1 Nonimmigrant Legal Status Categories Eligible for MAVNI Status Code Definition E-1 Individual in the United States to conduct trade under a treaty E-2 Individual in the United States to develop and direct an organization in which the individual has invested substantial capital under a treaty E-3 Australian Free Trade agreement working in specialty occupations F-1 Academic- or language-training student F-2 Dependent of an F-1 student F-3 Part-time student commuting from Canada or Mexico H-1B Individual in the United States to perform professional services for a sponsoring employer for a specified period of time H-1C Individual in the United States to perform professional nursing for the sponsoring organization for a specific period of time H-2A Individual in the United States to perform temporary agricultural work H-2B Individual in the United States in a temporary position for which a shortage exists in the United States H-3 Individual in the United States to participate in a training program provided by the sponsoring employer H-4 Dependent of an H principal legal status holder I Legal status for foreign media representative J-1 Student admitted as a student for duration of stay under exchange visitor application J-2 Dependent of J-1 legal status holder K-1 Fiancée or fiancé of a U.S. citizen K-2 Minor child of K-1 K-3 Spouse of a U.S. citizen K-4 Child of K-3 L-1 Individual transferred to the United States from an affiliated office or entity of the U.S. employer to work as an executive or specialist L-2 Dependent of L-1 transferee M-1 Vocational student admitted for duration of status M-2 Dependent of M-1 principal legal status holder M-3 Part-time student commuting from Canada or Mexico O-1 Individual of extraordinary ability in the sciences, arts, business, education, or athletics O-2 Personnel accompanying an O-1 legal status holder O-3 Dependent of an O-1 or O-2 legal status holder TIWARI_PROD_00002461 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 90 The Military Accessions Vital to the National Interest (MAVNI) Program Table A.1—continued Status Code Definition P-1 Internationally recognized entertainer or athlete and essential personnel P-2 Artist or entertainer in the United States under a reciprocal exchange program P-3 Individual in the United States in a culturally unique program to perform, coach, teach, or train P-4 Dependent of a principal P legal status holder Q-1 Participant in an international cultural exchange program Q-2 Participant in the Irish peace process cultural and training program Q-3 Dependent of Q-2 legal status holders R-1 Individual in the United States as a bona fide religious organization carrying out religious activities R-2 Dependent of R-1 legal status holder S-5 Informant of criminal organization information S-6 Informant of terrorism information S-7 Spouse, parent, or child of S-5 or S-6 T-1 Victim of a severe form of trafficking in persons T-2 Spouse of a victim of severe form of trafficking in persons T-3 Child of a severe form of trafficking in persons T-4 Parent of victim of severe form of trafficking in persons (if T-1 victim is under 21 years of age) TC Canadian physician authorized to engage in direct patient care TD Dependent of a TN visa holder TN A citizen of Canada or Mexico in the United States under NAFTA to perform specific services for a sponsoring employer U-1 Victim of certain criminal activity U-2 Spouse of U-1 U-3 Child of U-1 U-4 Parent of U-1, if U-1 is under 21 years of age V-1 Spouse of a U.S. lawful permanent resident V-2 Child of a U.S. lawful permanent resident V-3 Child of a V-1 or V-2 status individual SOURCE: Department of the Army, G-1, 2014. NOTE: NAFTA = North American Free Trade Agreement. TIWARI_PROD_00002462 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. APPENDIX B Blinder-Oaxaca Methodology1 The Blinder-Oaxaca method begins with two groups (in our case, MAVNI and nonMAVNI recruits) and a performance metric for each group (in our case, an attrition or reenlistment outcome). We denote group with the subscript g (where g equals either MAVNI or non-MAVNI recruits), and we denote the performance metric for each group as Yg .2 The metric for each group is assumed to be linearly related to a set of observed characteristics, Xg , where βg is the slope parameters, and unobserved characteristics are vg : Yg = Xg βg + vg, for g = MAVNI, non-MAVNI. We assume that the unobserved and observed characteristics are independent— E(vg Xg) = 0 where E() is the expectation operator—so that we can decompose the difference in mean performance between groups into the explained and structural components.3 We also assume that E(vg ) = 0. The difference in the mean performance metric between groups, R, is given by R = _E(Yn_) − E(Ym) = Yn − Ym = E(X _ n)βn _− E(Xm)βm = Xn βn − Xm βm where n refers to non-MAVNI recruits and m refers to MAVNI recruits. We simplified the notation and replaced expected values with sample averages—e.g., we replaced 1 Some of the material in this appendix is drawn from Asch et al. (2016). 2 We have three metrics we consider—DEP completion, 48-month cumulative retention, and reenlistment— so we should also subscript each metric by i, where i varies from one to three. However, to reduce the clutter of notation, we drop the i subscript here and refer to a single metric, though it should be recalled throughout that we have three metrics to consider. 3 Later in this appendix, we discuss what happens when this assumption is violated, as is likely the case in our analysis, and the implications for interpretation of the results. 91 TIWARI_PROD_00002463 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 92 The Military Accessions Vital to the National Interest (MAVNI) Program _ E(Ym) withYm. Through simple algebraic manipulation, adding and subtracting terms, R can be expressed as _ _ R = X_nβn −_Xm βm _ = (Xn − Xm )βm + Xn(βn − βm) = E + S. _ _ The first term, (Xn − Xm )βm, is the composition portion, or the explained portion, of the difference in the mean performance metric because it represents the contribution of group differences in characteristics to the overall difference in the mean performance metric. We denote this term as E. The second term is the contribution of the differences _ in the coefficients (including differences in the intercept), Xn(βn − βm). Insofar as the coefficients capture structural factors in the performance decision process, this term is often termed the structural portion of the difference in mean performance.4 The expression for S presupposes that the appropriate benchmark for the MAVNI coefficients is the high-quality non-MAVNI coefficients. However, this choice is somewhat arbitrary, and alternatives have been proposed in the literature (Neumark [1988] and Fortin [2007]) including using regression coefficients from a pooled regression. In our decomposition analysis for MAVNI, we use the non-MAVNI coefficients as the benchmark, but we conducted sensitivity analysis to see how the results would change if we chose to use pooled regression results. We find the results unchanged qualitatively and do not report the pooled results. In Chapter Three, we presented estimates of E and S for the three metrics we consider. These estimates provided information on the overall explained and structural component of differences in mean outcome between MAVNI and high-quality nonMAVNI recruits. As we discuss below, we interpret S as the component attributable to other factors rather than the explained component. The Problem of Self-Selection and Endogeneity We face two issues that result in the violation of the independence assumption (E(v X) = 0). The first issue is self-selection and the problem that our metrics are actually early career milestones—e.g., DEP completion, first-term retention given DEP completion, and reenlistment given first-term retention. The problem arises because at any given career milestone, the group of individuals eligible to proceed to the next milestone (i.e., the group that participates in a particular milestone) is a self-selected group of person_ _ _ The decomposition for R is not unique. An alternative expression for R is R = (Xn − Xm)βn + Xm(βn − βm). Here, the differences in characteristics are weighted by the coefficients for non-MAVNI recruits, rather than for MAVNI recruits, while the differences in coefficients are weighted by the characteristics for MAVNI recruits, rather than for non-MAVNI recruits. The choice of which to use is arbitrary. 4 TIWARI_PROD_00002464 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Blinder-Oaxaca Methodology 93 nel. The second issue is that our data do not include unobserved factors (by definition), and these factors, as well as observed factors, could be correlated with the factors that we do observe. We discuss each issue in turn. Self-selection is particularly a problem when the self-selection process differs for MAVNI and non-MAVNI recruits and might be partially based on factors that are unobservable in our data. For example, those eligible for reenlistment are those who completed their first term. The eligible group is not a random group of individuals but consists of those who were self-selected to stay until completion. Importantly, this self-selection might differ for MAVNI and non-MAVNI recruits and is likely to be partially based on factors that are not observable in our data, such as taste for service and different performance incentives. Indeed, we hypothesize that the opportunity for accelerated citizenship gives MAVNI recruits a stronger incentive to complete their initial service obligation. When participation in the eligibility pool for a particular milestone _is based on unobserved factors, the expected value of a given outcome is no longer Xg βg but is instead given by _ E(Y ) = X β + E(v eligibility) This expression can be rewritten as _ _ E(Y ) = X β + λσ, _ where λ is the mean of the usual inverse Mills ratio term that is obtained by estimating a probit model of the likelihood of eligibility, and σ is the coefficient on the Mills ratio term in the regression, related to the variance of the error term of the selection eligibility process and the covariance between that error and v. Performing stan_ dard ordinary least squares regression without accounting for the term λσ will lead to biased estimates of β if the eligibility selection process is correlated with the outcome of interest. This is likely to be the case in our analysis. Selection bias will also affect the decomposition. The Blinder-Oaxaca decomposition becomes _ _ _ _ _ _ R = (Xn − Xm)βm + Xn(βn − βm) + ( λn − λm)σm + λn( σn − σm) = E + S'. E is defined as before, but the structural portion now includes not only _ Xn(βn − βm) but also terms related to differences attributable to unobserved _ additional _ characteristics, ( λn − λm)σm, as well as differences_ in the effects of those unobservables (e.g., the coefficients on the unobservables), λn( σn − σm). If the self-selection processes differ for MAVNI and non-MAVNI recruits, the additional terms in S' do not disappear. TIWARI_PROD_00002465 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 94 The Military Accessions Vital to the National Interest (MAVNI) Program One approach to addressing this issue is to use a correction procedure, such as the one recommended by Heckman (1976), though Neuman and Oaxaca (2004) argued that the Heckman approach introduces fundamental ambiguities about how to interpret the structural component of the decomposition. However, even if valid, these procedures require an instrumental variable that predicts eligibility but not the outcome of interest. That is, there is an exclusion criterion. Unfortunately, we lack such an instrument, given the state of the data we have available. Consequently, we cannot give a causal interpretation to our results. The explained portion and the structural portion of the decomposition of the differences in performance metrics between MAVNI and non-MAVNI recruits will reflect factors that affected the likelihood of a given milestone and the selection factors that resulted in a recruit being eligible to reach that milestone. Furthermore, because S' also includes differences attributable to unobserved characteristics and differences in the selection processes, we can no longer interpret this part as structural. Instead, we call this the part of the decomposition due to “other” factors because it includes both the structural part—how observed characteristics affects the difference in a milestone for MAVNI and non-MAVNI recruits and the likelihood of reaching a given milestone—and the differential selection process that leads MAVNI and non-MAVNI recruits to be eligible to reach that milestone. The second issue we must confront is endogeneity of the characteristics. With respect to the unobserved characteristics, we assume that the observed and unobserved characteristics are independent, but it is possible that they are correlated. For example, having an advanced degree (observed) might be correlated with taste for the military or with ability (unobserved). If so, and we did not account for this correlation, the coefficient estimates would be biased. Given that we have no means to correct for this issue, we must use caution in interpreting our results as causal effects; acknowledge that the decompositions we compute can reflect correlations between observed and unobserved characteristics; and only argue that our analysis shows the contributions of observed factors and unobserved factors, where that contribution reflects both causal relationships and possible correlation with observed and unobserved factors. TIWARI_PROD_00002466 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. Blinder-Oaxaca Methodology 95 Table B.1 MAVNI and High-Quality Non-MAVNI Regression Estimates (Linear Probability Model) DEP Completion 48-Month Cumulative Retention Reenlistment MAVNI NonMAVNI MAVNI NonMAVNI 0.0001 0.00196 c 0.23856 c 0.12773c Age at contract –0.00077a 0.00008 0.00486 0.00333c Married –0.00420 0.00034 –0.04733 0.00286 –0.02281 0.07988 c Have dependents 0.00140 –0.00015 –0.01606 –0.01432b 0.01749 0.03912c Received DEP waiver 0.00907 –0.00072 0.08270 –0.04648 c 0.12153 0.01123 –0.02388 –0.04497c –0.00637 –0.01443a 0.00719 0.01812 –0.03569c 0.01330a 0.41549a 0.01212 Male Received accession waiver NonMAVNI –0.16467b –0.02723b 0.02007c 0.00036 BMI = 25 or above 0.00346 Enlist in critical skill 0.00520 0.00324 c –0.02176 –0.00096a –0.10866 Got loan repayment 0.00778 –0.00248b 0.09046 –0.00410 –0.09337 0.04782c Entered at advanced rank 0.01129 0.00317c –0.09032 –0.11381c –0.02015 –0.01443c Got enlistment bonus –0.00044 MAVNI 0.08040 b FY 2010 –0.02067c –0.00160 c FY 2013 –0.01904 c FY 2014 –0.02033c –0.00126a Black non-Hispanic –0.00021 0.00062 0.21534 c 0.01133 –0.00452 –0.03569c Hispanic 0.00296 0.00024 0.49403 0.03915c 0.41637 0.01212c Asian/Pacific Islander 0.00026 –0.00036 0.14544 b 0.03795b –0.15298a 0.04782 0.17044 0.17306 c –0.13401 0.25167c 0.00038 Other race/ethnicity –0.87967c –0.97423c AFQT Cat I –0.00626 –0.00031 –0.09426 0.05599c –0.08190 –0.05209c AFQT Cat II –0.00309 –0.00054 –0.08081a 0.01781c 0.01841 –0.02244 c Some college –0.00136 –0.00053 –0.09714 0.02869b 0.01990 –0.00689 College graduate (or more) –0.00276 –0.00007 –0.02392 0.06786a –0.21293 –0.14330 c Region: Midwest 0.00199 –0.00074 –0.00568 –0.01588a –0.19787c –0.00289 Region: South 0.00717a –0.00034 0.06350 –0.03730 c –0.12588 b –0.12617b –0.01158 Region: Northeast Region: U.S. territories Quarter January–March 0.03939c –0.00203 0.00059 0.01895 –0.01619 0.00000 –0.00175 0.00000 –0.02696 0.00000 0.05986 –0.01920a 0.07888 –0.01243 0.00258 0.00019 –0.02782 TIWARI_PROD_00002467 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 96 The Military Accessions Vital to the National Interest (MAVNI) Program Table B.1—continued DEP Completion Quarter April–June 48-Month Cumulative Retention Reenlistment MAVNI NonMAVNI MAVNI NonMAVNI MAVNI NonMAVNI –0.00517 –0.00068 –0.02232 –0.07684 c 0.05810 –0.01555 –0.00317 –0.03976 c Quarter July–September –0.00143 –0.00045 –0.07052 –0.09717c Combat Support MOS –0.00269 0.00115a –0.03644 –0.01730 b –0.00837 0.07703c Combat Service Support MOS –0.00547 0.00048 –0.04752 –0.02863c –0.13403b 0.07211c Entry grade: E-2 0.18492 0.04173c –0.19064 0.03886 c Entry grade: E-3 0.14535 0.04726 c –0.11538 0.06981c Entry grade: E-4 0.00119 0.01521 –0.09810 0.11400 b Constant 1.03099c 0.99481c 0.27710 0.39859c 0.44393a 0.41012c R-squared 0.880 0.973 0.047 0.038 0.089 0.048 Total observations 2,701 55,211 751 29,965 470 17,428 a Statistically significant from zero at the 10-percent level. b Statistically significant from zero at the 5-percent level. c Statistically significant from zero at the 1-percent level. TIWARI_PROD_00002468 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. References Acosta, Yesenia D., Luke J. Larsen, and Elizabeth M. Grieco, “Noncitizens Under Age 35: 2010– 2012,” U.S. Census Bureau American Community Survey Briefs, 2014. Adamson, Fiona, “Crossing Borders: International Migration and National Security,” International Security, Vol. 31, No. 1, 2006, pp. 165–199. Asch, Beth J., and James N. Dertouzos, Educational Benefits Versus Enlistment Bonuses: A Comparison of Recruiting Options, Santa Monica, Calif.: RAND Corporation, MR-302-OSD, 1994. As of May 19, 2017: https://www.rand.org/pubs/monograph_reports/MR302.html Asch, Beth J., John Romley, and Mark Totten, The Quality of Personnel in the Enlisted Ranks, Santa Monica, Calif.: RAND Corporation, MG-324-OSD, 2005. As of May 19, 2017: https://www.rand.org/pubs/monographs/MG324.html Asch, Beth J., Christopher Buck, Jacob Alex Klerman, Meredith Kleykamp, and David S. Loughran, Military Enlistment of Hispanic Youth: Obstacles and Opportunities, Santa Monica, Calif.: RAND Corporation, MG-773-OSD, 2009. As of May 19, 2017: https://www.rand.org/pubs/monographs/MG773.html Asch, Beth J., Paul Heaton, James Hosek, Francisco Martorell, Curtis Simon, and John Warner, Cash Incentives, Military Enlistment, Attrition, and Reenlistment, Santa Monica, Calif.: RAND Corporation, MG-950-OSD, 2010. As of May 19, 2017: https://www.rand.org/pubs/monographs/MG950.html Asch, Beth J., Trey Miller, and Gabriel Weinberger, Can We Explain Gender Differences in Officer Career Progression? Santa Monica, Calif.: RAND Corporation, RR-1288-OSD, 2016. As of May 19, 2017: https://www.rand.org/pubs/research_reports/RR1288.html Asymmetric Warfare Group, “Insider Threats in Partnering Environments,” August 2011. As of May 22, 2017: http://www.kansastag.gov/AdvHTML_Upload/files/ (U)%20Radicalization%20into%20Violent%20Extremism.pdf Atran, Scott, Lydia Wilson, Richard Davis, and Hammad Sheik, “The Devoted Actor, Sacred Values, and Willingness to Fight: Preliminary Studies with ISIL Volunteers and Kurdish Frontline Fighters,” unpublished manuscript presented to the Strategic Multilayer Assessment of ISIL in Support of SOCCNET, Washington, D.C., November 2014. As of August 10, 2016: http://artisresearch.com/wp-content/uploads/2015/05/Atran-Soccnet-MINERVA-ISIS.pdf 97 TIWARI_PROD_00002469 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 98 The Military Accessions Vital to the National Interest (MAVNI) Program Baker, Bryan, “Estimates of the Size and Characteristics of the Resident Nonimmigrant Population in the United States: January 2012,” Office of Immigration Statistics, DHS, February 2014. As of May 19, 2017: https://www.dhs.gov/sites/default/files/publications/ois_ni_pe_2012.pdf Barry, Catherine N., New Americans in Our Nation’s Military: A Proud Tradition and Hopeful Future, Center for American Progress, 2013. Behlendorf, Brandon, Jyoti Belur, and Sumit Kumar, “Peering Through the Kaleidoscope: Variation and Validity in Data Collection on Terrorist Attacks,” Studies in Conflict and Terrorism, Vol. 39, No. 7–8, 2016, pp. 641–667. Berrebi, Claude, “Evidence About the Link Between Education, Poverty, and Terrorism Among Palestinians,” Peace Economics, Peace Science, and Public Policy, Vol. 13, No. 1, 2007, pp. 1–36. Bicksler, Barbara, and Lisa Nolan, Recruiting an All-Volunteer Force: The Need for Sustained Investment in Recruiting Resources—An Update, Strategic Analysis Inc., December 2009. As of January 18, 2017: http://www.people.mil/Portals/56/Documents/MPP/AP/Bicksler_Recruiting_Paper_2009.pdf Blinder, Alan, “Wage Discrimination: Reduced Form and Structural Estimates,” Journal of Human Resources, Vol. 8, 1973, pp. 436–455. Borum, Randy, “Assessing Risk for Terrorism Involvement,” Journal of Threat Assessment and Management, Vol. 2, No. 2, 2015, pp. 63–87. Bove, Vincenzo, and Tobias Bohmelt, “Does Immigration Induce Terrorism?” The Journal of Politics, Vol. 78, No. 2, 2016, pp. 572–588. Buck, Kelly R., Andree E. Rose, Martin F. Wiskoff, and Kahlila M. Liverpool, “Screening for Potential Terrorists in the Enlisted Military Accession Process,” Defense Personnel Security Research Center, Technical Report 05-08, April 2005. As of August 10, 2016: http://www.dhra.mil/perserec/reports/tr05-8.pdf Buddin, Richard J., Success of First-Term Soldiers: The Effects of Recruiting Practices and Recruit Characteristics, Santa Monica, Calif.: RAND Corporation, MG-262-A, 2005. As of October 5, 2016: http://www.rand.org/pubs/monographs/MG262.html Bulling, Denise, Mario Scalora, Randy Borum, Jill Panuzio, and Andrew Donica, “Behavioral Science Guidelines for Assessing Insider Risk,” University of Nebraska Public Policy Center, Paper 37, 2008. Charney, David, and John Irvin, “A Guide to the Psychology of Espionage,” Association of Former Intelligence Officers, 2014. As of November 9, 2016: https://www.afio.com/publications/ CHARNEY_The_Psychology_of_Espionage_DRAFT_2014Aug28.pdf Crosset, Chuck, and Jason Spitaletta, “Radicalization: Relevant Psychological and Sociological Concepts,” Johns Hopkins University Applied Physics Laboratory, 2010. Davis, Paul K., and Kim Cragin, eds., Social Science for Counterterrorism: Putting the Pieces Together, Santa Monica, Calif.: RAND Corporation, MG-849-OSD, 2009. As of May 19, 2017: https://www.rand.org/pubs/monographs/MG849.html Davis, Paul K., and Angela O’Mahony, A Computational Model of Public Support for Insurgency and Terrorism: A Prototype Model for More-General Social-Science Modeling, Santa Monica, Calif.: RAND Corporation, TR-1220-OSD, 2013. As of May 19, 2017: https://www.rand.org/pubs/technical_reports/TR1220.html TIWARI_PROD_00002470 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. References 99 Department of the Army, “Two-Year Extension of Military Accessions Vital to the National Interest (MAVNI) Pilot Program Implementation Policy,” Office of the Deputy Chief of Staff G-1, December 19, 2014. DiFazio, Ani, Rodney McCloy, Matthew Reeder, and Matthew Fleisher, Military Accessions Vital to the National Interest (MAVNI): Final Program Evaluation, HumRRO: Alexandria, Va., FR-12-67, 2013. DoD—See U.S. Department of Defense. Ettredge, Michael, John Gerdes, and Gilbert Karuga, “Using Web-Based Search Data to Predict Macroeconomic Statistics,” Communications of the ACM, Vol. 48, No. 11, November 2005, pp. 87–92. Executive Order 13269, “Expedited Naturalization of Aliens and Noncitizen Nationals Serving in an Active-Duty Status During the War on Terrorism,” Code of Federal Regulations, July 8, 2002, p. 1136. As of January 22, 2017: https://www.justice.gov/sites/default/files/eoir/legacy/2003/07/09/pd08jy02_txt-26.pdf Fischer, Lynn, “Espionage: Why Does It Happen?” DoD Security Institute, 2000. As of November 9, 2016: http://www.au.af.mil/au/awc/awcgate/dod/espionage_whyhappens.pdf Fortin, Nicole, “The Gender Wage Gap Among Young Adults in the United States: The Importance of Money Versus People,” Journal of Human Resources, Vol. 43, No. 4, 2007, pp. 884–918. Fortin, Nicole, Thomas Lemieux, and Sergio Firpo, “Decomposition Methods in Economics,” in Orley Ashenfelter and David Card, eds., Handbook of Labor Economics, Vol. 4, Part A, Philadelphia: Elsevier B.V., 2011, pp. 1–102. Freilich, Joshua, and Gary LaFree, “Measurement Issues in the Study of Terrorism: Introducing the Special Issue,” Studies in Conflict and Terrorism, Vol. 39, No. 7–8, 2016, pp. 569–579. General Accountability Office, “Personnel Security Clearances: Actions Needed to Ensure Quality of Background Investigations and Resulting Decisions,” GAO-14-138T, 2014. As of November 10, 2016: http://www.gao.gov/products/GAO-14-138T General Accountability Office, Criminal Alien Statistics: Information on Incarcerations, Arrests, and Costs, GAO-11-187, March 2011. As of March 20, 2017: http://www.gao.gov/assets/320/316959.pdf Geren, Pete, Authority to Implement Military Accessions Vital to National Interest (MAVNI), Washington, D.C., Secretary of the Army, U.S. Department of Defense, 2008. Ginsberg, Jeremy, Matthew H. Mohebbi, Rajan S. Patel, Lynnette Brammer, Mark S. Smolinski, and Larry Brilliant, “Detecting Influenza Epidemics Using Search Engine Query Data,” Nature, Vol. 457, No. 7232, February 19, 2009, pp. 1012–1014. Hadley, Stephen, “Adjudicative Guidelines for Determining Eligibility for Access to Classified Information,” The White House, December 29, 2005. As of March 20, 2017: https://fas.org/sgp/isoo/guidelines.pdf Hattiangadi, Anita, Aline Quester, Gary Lee, and Diana Lien, Non-Citizens in Today’s Military: Final Report, Alexandria, Va.: Center for Naval Analyses, CNA Research Memorandum D0011092.A2/Final, 2005. Headquarters USAREC, “Military Accessions Vital to the National Interest (MAVNI) Recruitment Pilot for Health Care Providers and Chaplains,” USAREC Message 15-016, January 2015. TIWARI_PROD_00002471 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 100 The Military Accessions Vital to the National Interest (MAVNI) Program Heckman, James, “The Common Structure of Statistical Models of Truncation, Sample Selection, and Limited Dependent Variables and a Simple Estimator For Such Models,” Annals of Economic and Social Measurement, Vol. 5, 1976, pp. 475–492. Herbig, Katherine, Ray Zimmerman, and Callie Chandler, “The Evolution of the Automated Continuous Evaluation System (ACES) for Personnel Security,” Defense Personnel Security Research Center, Technical Report 13-06, 2013. As of March 20, 2017: http://www.dtic.mil/docs/citations/ADA626819 Herbig, Katherine, “Changes in Espionage by Americans: 1947–2007,” Defense Personnel Security Research Center, Technical Report 08-05, 2008. As of November 9, 2016: https://www.fas.org/sgp/library/changes.pdf Hosek, James R., Shanthi Nataraj, Michael G. Mattock, and Beth J. Asch, The Role of Special and Incentive Pays in Retaining Military Mental Health Care Officers, Santa Monica, Calif.: RAND Corporation, RR-1425-OSD, 2017. As of May 19, 2017: https://www.rand.org/pubs/research_reports/RR1425.html Hosek, James R., and Michael G. Mattock, Learning About Quality: How the Quality of Military Personnel Is Revealed Over Time, Santa Monica, Calif.: RAND Corporation, MR-1593-OSD, 2003. As of May 19, 2017: https://www.rand.org/pubs/monograph_reports/MR1593.html Huff, Jared M., Yevgeniya K. Pinelis, and Jennie W. Wenger, Adjusting First-Term Contract Lengths in the Navy: Implications and Recommendations, Alexandria, Va.: Center for Naval Analyses, Research Report DRM-2013-U-004794, 2013. Institute of International Education, “International Students in U.S. Top One Million for the First Time; More U.S. Students Study Abroad and Pursue Experiential Learning Overseas,” press release, November 16, 2016. Jahedi, Salar, Jennie W. Wenger, and Douglas Yeung, Searching for Information Online: Using Big Data to Identify Concerns of Potential Army Recruits, Santa Monica, Calif.: RAND Corporation, RR-1197-A, 2016. As of May 19, 2017: https://www.rand.org/pubs/research_reports/RR1197.html Jann, Ben, “The Oaxaca-Blinder Decomposition for Linear Regression Models,” Stata Journal, Vol. 8, 2008, pp. 435–479. Jensen, Eric B., “A Review of Methods for Estimating Emigration,” Population Division, U.S. Census Bureau, Working Paper No. 101, September 2013. As of November 4, 2016: https://www.census.gov/population/www/documentation/twps0101/SWG_emigration_jensen.pdf Jointhemilitary.org, “MAVNI Recruiter Checklist & Processing Flowchart,” 2016. As of January 22, 2017: http://jointhemilitary.org/mavni-recruiter-checklist-processing-flowchart Kepel, Gilles, The Roots of Radical Islam, Saqi Books, 2005. Kleycamp, Meredith A., “College, Jobs, or the Military? Enlistment During a Time of War,” Social Science Quarterly, Vol. 87, No. 2, 2006, pp. 272–290. Kraemer, Helena, Alan Kazdin, David Offord, Ronald Kessler, Peter Jensen, and David Kupfer, “Coming to Terms with the Terms of Risk,” Archives of General Psychiatry, Vol. 54, 1997, pp. 337–343. Krueger, Alan, and Jitka Maleckova, “Education, Poverty, and Terrorism: Is There a Causal Connection?” Journal of Economic Perspectives, Vol. 17, No. 4, 2003, pp. 119–144. TIWARI_PROD_00002472 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. References 101 Kruglanski, Arie, Michele Gelfand, Jocelyn Belanger, Anna Sheveland, Malkanthi Hetiarchi, and Rohan Gunaratna, “The Psychology of Radicalization and Deradicalization: How Significance Quest Impacts Violent Extremism,” Advances in Political Psychology, Vol. 35, Supplement 1, 2014, pp. 69–93. LaFree, Gary, and Gary Ackerman, “The Empirical Study of Terrorism: Social and Legal Research,” National Consortium for the Study of Terrorism, University of Maryland, 2009. Leiken, Robert, and Steven Brooke, “The Quantitative Analysis of Terrorism and Immigration: An Initial Exploration,” Terrorism and Political Violence, Vol. 18, No. 4, 2006, pp. 503–521. Lum, Cynthia, Leslie Kennedy, and Alison Sherley, “Are Counterterrorism Strategies Effective: The Results of the Campbell Systematic Review on Counterterrorism Evaluation Research,” Journal of Experimental Criminology, Vol. 2, 2006, pp. 489–516. Malone, Lauren, Neil Carey, Yevgeniya Pinelis, and Dave Gregory, Waivered Recruits: An Evaluation of Their Performance and Attrition Risk, Center for Naval Analyses: Alexandria, Va., CNA Research Memorandum D0023955.A4/Final, 2011. McCauley, Clark, and Sophia Moskalenko, “Mechanisms of Political Radicalization: Pathways Toward Terrorism,” Terrorism and Political Violence, Vol. 20, No. 3, 2008, pp. 415–433. McIntosh, Molly, and Seema Sayala, Non-Citizens in Enlisted U.S. Military, Alexandria, Va.: Center for Naval Analyses, CNA Research Memorandum D0025768.A2/Final, 2011. Meloy, J. Reid, and Paul Gill, “The Lone-Actor Terrorist and the TRAP-18,” Journal of Threat Assessment and Management, Vol. 3, No. 1, 2016, pp. 37–52. Monahan, John, “The Individual Risk Assessment of Terrorism,” Psychology, Public Policy and Law, Vol. 18, 2012, pp. 167–204. Monahan, John, “The Individual Risk Assessment of Terrorism: Recent Developments,” in Gary LaFree and Joshua Freilich, eds., The Handbook of Criminology of Terrorism, Hoboken, N.J.: WileyBlackwell, 2017, pp. 520–534. National Defense Authorization Act 2017, H.R. 4909, 2016. As of May 19, 2017: https://armedservices.house.gov/hearings-and-legislation/ndaa-national-defense-authorization-act Neuman, Shoshana, and Ronald Oaxaca, “Wage Decompositions with Selectivity-Corrected Wage Equations: A Methodological Note,” Journal of Economic Inequality, Vol. 2, 2004, pp. 3–10. Neumark, David, “Employers’ Discriminatory Behavior and the Estimation of Wage Discrimination,” Journal of Human Resources, Vol. 23, No. 3, 1988, pp. 279–295. New America, “Terrorism in America After 9/11,” undated. As of November 10, 2016: https://www.newamerica.org/in-depth/terrorism-in-america Nowrasteh, Alex, “Terrorism and Immigration: A Risk Analysis,” Policy Analysis, No. 798, Cato Institute, September 13, 2016. As of March 20, 2017: https://www.cato.org/publications/policy-analysis/terrorism-immigration-risk-analysis Oaxaca, Ronald, “Male-Female Wage Differentials in the Urban Labor Market,” International Economic Review, Vol. 14, 1973, pp. 693–709. Office of the Directorate of National Intelligence, “2014 Report on Security Clearance Determinations,” April 2015. As of December 13, 2016: https://fas.org/sgp/othergov/intel/clear-2014.pdf Office of the Under Secretary of Defense, “Two-Year Extension of Military Accessions Vital to the National Interest (MAVNI) Pilot Program,” August 17, 2010. TIWARI_PROD_00002473 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 102 The Military Accessions Vital to the National Interest (MAVNI) Program Office of the Under Secretary of Defense, Personnel and Readiness, “Military Accessions Vital to the National Interest Pilot Program Extension,” September 30, 2016. Orvis, Bruce R., Steven Garber, Philip Hall-Partyka, Christopher Maerzluft, and Tiffany Tsai, Recruiting Strategies to Support the Army’s All-Volunteer Force, Santa Monica, Calif.: RAND Corporation, RR-1211-A, 2016. As of May 19, 2017: https://www.rand.org/pubs/research_reports/RR1211.html Passell, Jeffrey S., and D’Vera Cohn, Unauthorized Immigrant Population: National and State Trends, 2010, Pew Hispanic Center, 2011. As of November 4, 2016: http://pewhispanic.org/files/reports/133.pdf Pew Research Center, “How Americans Go Online,” infographic, September 25, 2013. As of March 30, 2015: http://www.pewinternet.org/2013/09/25/how-americans-go-online Pew Research Center, “Modern Immigration Wave Brings 59 Million to U.S., Driving Population Growth and Change Through 2065: Views of Immigration’s Impact on U.S. Society Mixed,” Washington, D.C., September 2015. Presidential Memorandum, National Insider Threat Policy and Minimum Standards for Executive Branch Insider Threat Programs,” November 2012. As of June 13, 2017: https://obamawhitehouse.archives.gov/the-press-office/2012/11/21/ presidential-memorandum-national-insider-threat-policy-and-minimum-stand Purcell, Kristen, Joanna Brenner, and Lee Rainie, Search Engine Use 2012, Pew Internet and American Life Project, Pew Research Center, March 9, 2012. Rivinius, Jessica, “Majority of 2013 Terrorist Attacks Occurred in Just a Few Countries,” National Consortium for the Study of Terrorism and Responses to Terrorism, University of Maryland, 2014. As of November 9, 2016: http://www.start.umd.edu/news/majority-2013-terrorist-attacks-occurred-just-few-countries Ruiz, Neil G., “The Geography of Foreign Students in U.S. Higher Education: Origins and Destinations,” Global Cities Initiative, A Joint Project of Brookings and JPMorgan Chase, August 2015. Schitrit, Daria, Einav Yogev, and Yoram Schweitzer, “Suicide Attacks in 2015,” Institute for National Security Studies, INSS Insight No. 789, January 2016. As of November 9, 2016: http://www.inss.org.il/publication/suicide-attacks-in-2015 Secretary of the Army, “Authority to Implement Military Accessions Vital to the National Interest (MAVNI),” December 23, 2008. Seifert, Rita Furst, Paul F. Hogan, and Carol Moore, Estimating the Qualified Military Available, Falls Church, Va.: The Lewin Group Inc., 2007. Simon, Curtis, and John Warner, “Managing the All-Volunteer Force in a Time of War,” Economics of Peace and Security Journal, Vol. 2, No. 1, 2005, pp. 20–29. Simon, Curtis, Sebastian Negrusa, and John Warner, “Educational Benefits and Military Service: An Analysis of Enlistment, Reenlistment, and Veterans’ Benefit Usage 1991–2005,” Economic Inquiry, Vol. 48, No. 4, 2010, pp. 1008–1031. Sohoni, Deenesh, and Amin Vafa, “The Fight to Be American: Military Naturalization and Asian Citizenship,” Asian American Law Journal, Vol. 17, No. 4, 2010, pp. 119–151. Stigler, George, The Theory of Price, 3rd edition, New York: MacMillan Publishers, 1966. TIWARI_PROD_00002474 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. References 103 Strom, Kevin J., John S. Hollywood, and Mark W. Pope, “Terrorist Plots in the United States: What We Have Really Faced, and How We Might Best Defend Against It,” in Gary LaFree and Joshua Freilich, eds., The Handbook of Criminology of Terrorism, Hoboken, N.J.: Wiley-Blackwell, 2017, pp. 468–481. Study in the States, “SEVIS by the Numbers Data,” undated. As of May 22, 2017: https://studyinthestates.dhs.gov/sevis-by-the-numbers/sevis-by-the-numbers-data USCIS—See United States Customs and Immigration Service. U.S. Army, Antiterrorism Officer Ready Reference, undated. As of May 22, 2017: http://www.garrison.redstone.army.mil/uploads/ATO_ReadyReference_web.pdf U.S. Army, “Military Accessions Vital to the National Interest (MAVNI) Information Paper,” 2012. U.S. Census Bureau, “Vacancy Rate Fact Sheet,” December 9, 2011. As of May 22, 2017: http://www.census.gov/housing/vacanciesfactsheet.html U.S. Customs and Immigration Service, “Path to U.S. Citizenship,” 2013. As of November 15, 2016: https://www.uscis.gov/us-citizenship/citizenship-through-naturalization/path-us-citizenship U.S. Customs and Immigration Service, “Naturalization Through Military Service Fact Sheet,” 2016. As of January 17, 2017: https://www.uscis.gov/news/fact-sheets/naturalization-through-military-service-fact-sheet U.S. Department of Defense, “Military Accessions Vital to the National Interest (MAVNI) Recruitment Pilot Program,” undated. As of January 22, 2017: https://www.defense.gov/news/mavni-fact-sheet.pdf U.S. Department of Defense, “Memorandum for Secretaries of the Military Departments: Two-Year Extension of Military Accessions Vital to National Interest (MAVNI) Pilot Program,” August 17, 2010. U.S. Department of Defense, “Military Accessions Vital to the National Interest Pilot Program Extension,” Office of the Under Secretary of Defense for Personnel and Readiness, September 30, 2016. U.S. Department of Justice, “Terrorism 2002–2005,” Federal Bureau of Investigation, 2005. As of November 9, 2016: https://www.fbi.gov/stats-services/publications/terrorism-2002-2005 U.S. Navy Recruiting Command, MAVNI Pilot Program, Millington, Tenn.: 2015. U.S. Office of Personnel Management, “Investigations Reimbursable Billing Rates Effective October 1, 2015” Federal Investigations Notice No. 16-01, Federal Investigation Services, October 2015. As of November 14, 2016: https://www.opm.gov/investigations/background-investigations/federal-investigations-notices/2016/ fin-16-01.pdf Veterans for Immigration Reform, On the Front Line: The Impact of Immigrants on Military Force Readiness, 2014. As of January 17, 2016: http://www.vets4reform.org/wp-content/uploads/2014/06/OnTheFrontLine.pdf Victoroff, Jeff, “The Mind of the Terrorist: A Review and Critique of Psychological Approaches,” Journal of Conflict Resolution, Vol. 49, No. 1, 2005, pp. 3–42. Weigley, Russell, History of the United States Army, New York: MacMillan Publishers, 1967. Youpa, Daniel, Joanne Marshall-Mies, and Ralph Carney, “Development and Evaluation of a Single Scope Background Investigation Quality Rating Form,” Defense Personnel Security Research Center, Technical Report 05-2, November 2004. TIWARI_PROD_00002475 PREPUBLICATION COPY. NOT CLEARED FOR PUBLIC RELEASE. 104 The Military Accessions Vital to the National Interest (MAVNI) Program Zeadally, Sherali, Byunggu Yu, Dong Hyun Jeong, and Lily Liang, “Detecting Insider Threats: Solutions and Trends,” Information Security Journal: A Global Perspective, Vol. 21, 2012, pp. 183–192. TIWARI_PROD_00002476