THE FINANCIAL WELFARE OF MILITARY HOUSEHOLDS: DESCRIPTIVE EVIDENCE FROM RECENT SURVEYS. William Skimmyhorn. November 10, 2014

Transcription

1 THE FINANCIAL WELFARE OF MILITARY HOUSEHOLDS: DESCRIPTIVE EVIDENCE FROM RECENT SURVEYS William Skimmyhorn November 10, 2014 Abstract Using survey data from large national samples of military and civilian households, this analysis provides evidence on the differences in financial decision-making and outcomes between comparable civilian and military members, between members of the military Services (Army, Navy, Air Force and Marines) and between members of the Service components (Active Duty, Reserves and National Guard). I find that military members have better solvency and savings outcomes but worse credit card outcomes relative to comparable civilians. Within the military, I find few differences in the financial outcomes between members of the Services and similarly few differences between members of the different components. I briefly discuss directions for future research and some policy implications. JEL Codes: D12, D14, D18, G28, H65 William Skimmyhorn Office of Economic and Manpower Analysis, Department of Social Sciences United States Military Academy West Point, NY william.skimmyhorn@usma.edu (845) Note: Thanks to Susan Carter, Gary Mottola, John Smith, Christine Kieffer and Bud Schneeweis for helpful comments. This paper was prepared with financial support from the FINRA Investor Education Foundation. The opinions provided herein are those of the author and do not necessarily reflect the views of the Financial Industry Regulatory Authority, the FINRA Foundation, the U.S. Military Academy, the Department of the Army or the Department of Defense. 1

2 1. Introduction In the United States, a decade of war and a struggling economy have generated substantial interest in the welfare of military families. Annual military compensation increases, new Veteran s benefits (e.g., the Post 9/11 GI Bill), public and private sector commitments to reduce Veteran unemployment and resistance to military pension reform all suggest a strong commitment by the government and the public to military servicemembers and their families. Yet despite significant media and government attention to these topics, to date there has been little scientific research on the household finances of military families. Some recent survey evidence suggests that younger military members are more financially secure than their civilian peers (WSJ 2014, FINRA Foundation 2013), though these reports fail to account for the many other potential differences between the groups. Academic researchers have evaluated more specific financial situations for military members including significant attention for the effects of payday lending (Tanik 2005, Carrel & Zinman 2014, Carter & Skimmyhorn 2014) and the use of alternative financial services more generally (Fox 2012). Economists and policy researchers have also been very interested in the role of financial education and there has been some academic work evaluating its effectiveness for military members (Bell, Gorin & Hogarth 2008 and Skimmyhorn 2014). Finally, there exists some limited evidence on the effects of military deployments on household financial decisions (Bell 2013). More importantly, to my knowledge there has been no comparative analyses between military members and their civilian counterparts, between members of the different military Services or between members of the different military components. This paper provides initial descriptive evidence on military members household finances of using two large national surveys of military members and their civilian counterparts. In particular, this research describes the financial decision-making of military families and how 2

3 these decisions compare with civilians; how they differ by services (Army, Navy, Air Force and Marines); and how they differ by components (Active Duty, Reserve and National Guard). The paper proceeds as follows. In Section 2, I describe the National Financial Capability Studies (NFCS) and data. In Section 3, I present the results. In Section 4, I discuss the results and conclude. 2. The National Financial Capability Study Data This analysis utilizes data from the 2009 and 2012 NFCS sponsored by the Financial Industry Regulatory Authority (FINRA) Investor Education Foundation. The FINRA Foundation developed the surveys in consultation with the U.S. Department of Treasury and other federal agencies, leading academics and researchers, and the President s Advisory Council on Financial Capability. My sample combines data from the State-by-State (hereafter State) and Military Surveys. 1 Applied Research and Consulting, on behalf of the FINRA foundation, generated the samples using quota sampling from an existing online panel of millions of individuals that includes but is not limited to military personnel. 2 The surveys were completed online in 2009 and 2012 and respondents were compensated for their participation. As a result, while the surveys are national in scope and their methodologies were carefully implemented, there may exist important sources of error that include: selection into participation, non-response bias among participants and measurement 1 The NFCS surveys and data are publicly available at: The surveys are nearly identical with a few exceptions: screening questions were ordered differently, military members were asked some additional military-specific questions and a few questions were reworded to apply to military settings (e.g., retirement accounts). 2 For more information on the Military and State-by-State survey methodologies, see: and respectively. 3

4 error. 3 In neither survey is the sample necessarily representative of the full U.S. or military populations, though I restrict the sample to mitigate some of these differences. 4 In order to focus on the population of the most policy interest (i.e., enlisted servicemembers and their low to moderate income and education counterparts), I restrict the sample to individuals aged years old with less than a college degree and less than $75K in annual income. 5 These restrictions generate a combined sample size of n=13,446 civilians from the 2009 and 2012 State Surveys and n=606 military servicemembers from the 2009 and 2012 State and Military Surveys. 6 Given these caveats, the analysis below should provide initial descriptive evidence on important questions related to military members financial conditions. While the NFCS contains a large number of financial outcomes of interest to economists and policy-makers, I analyze a set of outcomes that highlight the most basic financial conditions and behaviors and that facilitate meaningful comparisons between military and civilian groups. For example, since military members have access to a defined benefit pension but no access to a traditional 401k with employer matching like many civilians, 7 I omit retirement savings outcomes. I analyze ten outcomes that include recent conditions (e.g., spending greater than income in the past 12 months), routine activities (e.g., credit card behavior) and use of alternative 3 The NFCS relies on self-reported demographic and financial outcome variables and accordingly, may suffer from measurement error. Individuals might overreport behaviors perceived as desirable and underreport behaviors perceived as undesirable. However, since the paper relies on comparisons across groups (e.g., military members vs. civilians or military services vs. eachother), measurement errors are less concerning if the self-reports are biased equally across groups. Still, given the military s concern for servicemembers financial affairs and its legal authority over many aspects of their lives, these self-report concerns might be more likely among servicemembers. 4 The 2009 NFCS State and 2012 NFCS State Surveys include sample weights to enable approximations of the national population and the 2012 NFCS Military Survey includes sample weights to enable an approximation of the uniformed Department of Defense (DOD). But since the 2009 NFCS Military Survey lacks weights, I use unweighted data for this analysis and instead restrict the samples based on observable characteristics. 5 The education restriction (less than a college degree) removes a large number of individuals from the military and civilian samples, but I proceed with this smaller sample since individuals with less than a college education are of primary policy interest for financial decision-making. College graduates earn more (BLS 2014) and typically have higher levels of financial literacy (Lusardi and Mitchell 2011). For robustness checks, see fn14. 6 I further restrict the military sample to enlisted members of the Army, Navy, Air Force and Marines. 7 Military members do have access to the Thrift Savings Plan (TSP), a tax advantaged (deferred or Roth) plan without any matching. 4

5 financial services (e.g., use of payday loans). The alternative financial services product use is especially interesting in light of the 2007 Military Lending Act and recent Consumer Financial Protection Bureau (CFPB) efforts to regulate these products. I describe each outcome below: 8 General Conditions 1. Spending greater that monthly income (question J3) takes on a value of 1 for those who indicated that over the past year their household spending was more than their household income; and=0 otherwise. 2. Difficulty covering expenses (question J4) takes on a value of 1 for those who indicated that in a typical month it is very or somewhat difficult to cover their bills; and=0 otherwise. 3. Declared bankruptcy (question G4) takes on a value of 1 for those who indicated that they had declared bankruptcy in the last two years; and=0 otherwise. Routine Activities 4. Number of credit cards (question F1) takes on the midpoint value of the answer bin selected (i.e., =0 for bin No Credit Cards; =1 for bin 1; =2.5 for bin 2-3; =6 for bin 4-8; =10.5 for bin 9-12; =16.5 for bin 13-20; =21 for bin More than 20). 5. Poor credit card behavior index (question F2) takes on a value of 0-6 for where 0 represents no bad credit card behaviors and 6 represents all of the bad credit card behaviors. I first create an indicator for each of the following credit behaviors: not always paying in full, carrying over a balance and being charged interest, paying only the minimum payment in some months, being charged a late fee, being charged an over the limit fee, or using a credit card for a cash advance. Then I create the overall index by summing the six indicators. 6. Has an emergency fund for 3 months of expenses (question J5) takes on a value of 1 for those who indicated that they have set aside funds to cover expenses for 3 months in the case of sickness, job loss, economic downturn or other emergencies; and=0 otherwise. 7. Has investments other than for retirement (question B14) takes on a value of 1 for those who indicated that they have investments in stocks, bonds, mutual funds, or other securities not including their retirement accounts; and =0 otherwise. Use of Alternative Financial Services 8. Taken a payday loan (question G5_2 and G25_2 for the State and Military Surveys respectively) takes on a value of 1 for those who indicated that over the past 5 years they have taken out a payday loan; and=0 otherwise. 9. Taken an auto title loan (question G5_1 and G25_1 for the State and Military Surveys respectively) takes on a value of 1 for those who indicated that over the past 5 years they have taken out an auto title loan; and=0 otherwise. The question explains what an auto title loan is and explicitly differentiates them from auto purchase loans. 10. Taken a refund anticipation loan (question G5_3 and G25_3 for the State and Military Surveys respectively) takes on a value of 1 for those who indicated that over the past 5 8 For each outcome, I omit individuals who responded Don t know or Prefer not to say. As a result, observation counts may vary slightly for different outcomes. 5

6 years they have gotten an advance on their tax refund; and=0 otherwise. The question provides alternate names for these services (i.e., refund anticipation check and rapid refund ) and explicitly differentiates these loans from e-filing for taxes. The NFCS surveys also contain a number of individual characteristics that I use in my analysis. These include: gender, age bins (18-24, 25-34, 35-44) marital status (single, currently married, previously married), an indicator for having children, an indicator for minority status, education levels (less than high school, high school graduate, some college) and annual income bins (<$15K, $15-25K, $25-35K, $35-50K, $50-75K). The NFCS also includes measures of financial literacy and confidence. The financial literacy questions, developed by Lusardi and Mitchell (2011) are widely used and I create an index (from 0-5) that reflects the total number of questions (interest rates, inflation, bond prices, diversification and mortgages) 9 answered correctly. 10 For financial confidence I create an index (from 0-18) that reflects the sum of three self-reported measures (good at dealing with day-to-day financial matters, good at math and overall financial knowledge). 9 The five financial literacy questions and potential answers (correct answers italicized) are: 1. Interest rates (M6). Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow? (More than $102, Exactly $102, Less than $102, Don t Know[DK], Prefer not to say[na]) 2. Inflation (M7). Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, how much would you be able to buy with the money in this account? (More than today, Exactly the same, Less than today, DK, NA) 3. Bond Prices (M8). If interest rates rise, what will typically happen to bond prices? (They will rise, They will fall, They will stay the same, There is no relationship between bond prices and the interest rate, DK, NA) 4. Mortgages (M9). A 15-year mortgage typically requires higher monthly payments than a 30-year mortgage, but the total interest paid over the life of the loan will be less. (True, False, DK, NA) 5. Diversification (M10). Buying a single company s stock usually provides a safer return than a stock mutual fund. (True, False, DK, NA) 10 I score individuals who answered Don t know or Prefer not to say as incorrect (=0) for each question. 11 The three financial confidence questions and answer scales are: 1. Day-to-day (M1_1). I am good at dealing with day-to-day financial matters, such as checking accounts, credit and debit cards and tracking expenses. (Strongly Disagree=1 to 7=Strongly Agree, Don t Know[DK], Prefer not to say[na]) 2. Math (M1_2). I am pretty good at math. (Strongly Disagree=1 to 7=Strongly Agree, DK, NA) 3. Overall (M4). On a scale from 1 to 7, where 1 means very low and 7 means very high, how would you assess your overall financial knowledge? (Very Low=1 to 7=Very High, DK, NA). For all three questions I convert the original answers (1 to 7) to the new scale (0 to 6). 12 I score individuals who answered Don t know or Prefer not to say as very low (=0) for each question. 6

7 3. Empirical Results In this section I complete descriptive analyses for three important questions. First I provide evidence on how military and civilian financial outcomes differ. Second I provide evidence on how individuals financial outcomes differ by military Service (Army, Navy, Air Force and Marines). Finally, I provide evidence on how individuals financial outcomes differ by their military service component (Active Duty, Reserves and National Guard). The analyses are only descriptive given that there is no experimental assignment of individuals to the different groups. Said differently, none of the results in this paper should be taken as causal evidence on the effects of military service, serving in a specific Service, or serving in a specific Component. For each analysis, I provide some background and briefly motivate the question, present summary statistics and complete multivariate regression analysis. 3.1 How do financial outcomes and behaviors differ between military and civilian households? Given the voluntary nature of military service and the significant differences in military and civilian life, we expect differences in the characteristics of those who serve and those who do not. To provide initial insight into some of these observable differences, in Table 1A, I present summary statistics that describe how members of the military and civilian samples differ by observable characteristics (Panel A) and in their financial outcomes (Panel B). The results in Panel A suggest that on average, military members are younger, less likely to be female, more likely to be married, about equally likely to be divorced, about equally likely to have children, slightly less likely to be a minority, slightly more educated (among this non-college degree sample), and to have higher income. They also have higher financially literacy scores and more financial confidence relative to their civilian counterparts. The Panel B results suggest that military members have, on average, a lower probability of their spending exceeding their income, much lower difficulty covering expenses and a higher 7

8 probability of declaring bankruptcy. On average, military members have more credit cards, exhibit poorer credit behavior, are more likely to have a 3 month emergency fund and are more likely to have non-retirement savings. Finally, military members are slightly less likely to have used a tax refund anticipation loan, slightly less likely to have used a payday loan, and more likely to have used an auto title loan. [Insert Table 1A about here] Importantly, while these results provide some evidence as to the differing financial situations and behaviors for Americans, it does not account for the observable differences in the populations (Panel A). As a result, I complete a more meaningful analysis using a multivariate regression framework to more precisely estimate the differences in the two groups. This framework allows me to identify the differences in military and civilians conditional on other observable differences (e.g., age, education, financial literacy, etc.). For example, I can estimate the difference in financial outcomes between two individuals who are GED holders or two individuals who are both divorced with children. Without these controls, the differences we observe in military members and civilians might simply be explained by differences in their characteristics (e.g., their age, education or family situations). In Equation 1 I present my primary regression specification for this analysis: = (1) is the financial outcome of interest for individual. is an indicator that takes on the value of 1 if an individual is in the military and a value of 0 for civilians. is a vector of individual characteristics including gender, age, marital status, dependents, minority status, education level, income level, financial literacy index score and financial confidence index score. represents state fixed effects for the individual s state of residence. is an indicator that 8

9 equals 1 for 2012 NFCS respondents and equals 0 for 2009 NFCS respondents. is the parameter of interest and it reflects the average difference in each outcome between military and civilian members. In all specifications I cluster the standard errors at the state level to account for correlations in the error terms between individuals. Since military service is not randomly assigned, the coefficient cannot be interpreted causally. Instead, it represents a combination of all factors that make military members and civilians differ for each financial outcome. These differences include but are not limited to differences in cognitive abilities (e.g., mathematical abilities, actual financial knowledge), noncognitive abilities (e.g., propensity to plan, self-control), time preferences (i.e., future orientation), risk preferences (e.g., willingness to participate in the stock market) and any institutional/employer factors that differ on average between military and civilian members in this sample (e.g., financial education, counseling, choice architectures). While the individual characteristics can account for some of these differences (e.g, education and the financial literacy index may account for some of the differences in cognitive abilities), the typical concerns with omitted variable bias persist here and the estimates presented only provide descriptive evidence on the differences in military servicemembers and their civilian counterparts. In Table 1B, I present ordinary least squares regression results for Equation 1 for the ten outcomes of interest and I restrict the reported coefficients to the parameter of interest ( ) for clarity. 13 [Insert Table 1B about here] I focus my discussion on the coefficient estimate for, its economic magnitude (relative to the sample mean for civilians, which I include at the top of each column in 13 As a robustness check I complete logit regressions instead of the OLS linear probability model. The logit marginal effects estimates are slightly smaller than the OLS coefficients, but the statistical significance holds for all outcomes. 9

10 each panel) and its statistical significance (indicated by the stars for each coefficient, where ***, ** and * reflect statistical significance at the 1%, 5% and 10% levels respectively). I provide the standard errors (in parentheses) in the table and the associated p-values in the text for clarity. First, the regression results for the financial outcomes in Table 1B Panel A suggest that on average, members of the military do not have statistically significant differences in their likelihood of having monthly spending greater than monthly income (Col. 1., p=0.949) or in their likelihood of having declared bankruptcy in the past two years (Col. 3, p=0.178). However, military members report being percentage points (29% based on a mean outcome of 77% in this sample) less likely to have difficulty covering their monthly expenses and the difference is statistically significant (p<0.01). One potential explanation is that military members likely enjoy greater income stability than their civilian counterparts, though military spouses may enjoy less stability given the challenges of military relocations and increased parental responsibilities when service members are deployed. Another explanation might be that military members simply face fewer expenses than their civilian counterparts, either in lower food and transportation costs when on their military installations, or in total costs due to relatively frequent deployments. Overall, this outcome reflects a substantial difference between the two groups. The results for the differences in Table 1B Panel B suggest important differences between military and civilian households with respect to more routine financial behaviors. On average, military members have 0.82 (49%) more credit cards (Col. 1, p<0.01) and 0.50 (37%) more poor credit card behaviors (Col. 2, late payments, exceeding limits, etc., p<0.01). Military members are percentage points (65%) more likely to have a 3 month emergency fund (Col. 3, p<0.01) and percentage points (110%) more likely to have non-retirement investments (Col. 4, p<0.01). The Panel B results suggest that military members appear to be more at risk 10

11 with credit cards and less at risk in terms savings and investments. The degree of overlap between the emergency fund and non-retirement investment outcomes is unknown. If military households are reporting money in stocks, bonds, mutual funds, etc. for both outcomes, then they might be most usefully interpreted together. Even in this case, they suggest that military households are, on average, more financially prepared in terms of accessible savings. Interestingly, while emergency funds are generally considered an important measure of insurance for families in the event of unplanned expenses, the relative stability of military employment might suggest, all else equal, a lower need for such savings for military households. The higher levels of savings here among military members might be better used in other investments or to reduce credit card financing, especially if credit cards are being used to finance short term savings. Overall, the Panel B results provide mixed evidence on military servicemembers household financial decisions: they appear to save and invest more, but they also have more credit cards and poorer credit card behaviors. The results in Table 1B Panel C suggest no meaningful differences in the use of alternative financial services (AFS) by military members relative to civilians. The estimated effects for the likelihood of payday loan / auto title loan / refund anticipation loan use are all statistically insignificant (p=0.201 / p=0.201 / p=0.356) in this sample. 14 These findings stand in 14 Given my sample restriction to those with less than a college degree, I complete some robustness checks to determine if my results are driven by excluding the large population of college graduates. I complete a similar multivariate regression analysis for military vs. civilian members for a larger sample (approximately N=18,000) that includes college graduates and present the results in the Appendix Table 1D. The results are similar to my main results for nearly all ten outcomes. Two notable exceptions are: a. Military members are 1.97 percentage points (60%) more likely to have declared bankruptcy in the past 5 years, a marginally significant result (p=0.086). b. Military members are 3.47 percentage points (33%) more likely to have used an auto title loan in the past 5 years, a marginally significant result (p=0.063). I rule out the exclusion of officers as an explanation for these findings by including them in Appendix Table 1E; the results hold and in fact are slightly stronger. One possible explanation for these findings could be that individuals who have declared bankruptcy or used an auto title loan might subsequently select into the military. Both the bankruptcy and auto title loan questions are retrospective (2 and 5 years respectively) but the military service 11

12 contrast to many previous media and other reports (Tanik 2005) of concerning levels of AFS use by military members that prompted Congressional passage of the 2007 Military Lending Act (MLA) and continued attention from federal agencies (e.g., the CFPB). I discuss this finding more in Section 4. The findings here do not refute the fact that military members, on average, use these products at higher rates than the full civilian population. They do suggest that when compared to civilian individuals similar in age, education, income and the other individual characteristics used here, that military members are no more or less likely to use these services. But the timing of the NFCS surveys and the wording of these questions complicate their interpretation. First, these data were collected in 2009 and 2012, after the MLA was implemented. So the lack of differences in AFS use could reflect the special protections afforded servicemembers and their spouses under the law. But researchers remain divided on the effectiveness of the original MLA. 15 If it was ineffective (e.g., from lenders ignoring the law, altering their loans or services, or individuals going online for e.g., payday loans), then the comparable levels of use might suggest there is no need for additional special protections. An additional complicating factor is that the NFCS questions ask about AFS product use in the past 5 years, making it possible that individuals reported AFS use from before the MLA. This might also suggest no meaningful differences in AFS use by military status since military members and civilians had similar financial options before the MLA. 3.2 How do financial outcomes and behaviors differ between military members by Service? response is contemporaneous. Another explanation is that there could be positive selection into the military for enlisted members and negative selection into the military for officers. 15 Silver-Greenberg and Eavis (2013) find few effects from the MLA while Fox (2012) suggests that the MLA may have been effective in some states. The CFPB sought and secured increased enforcement authority for the MLA in (Kaplinsky 2012), suggesting their concern that the original law was inadequate. 12

13 The U.S. Military Services (i.e., the Army, Navy, Air Force and Marine Corps), often called branches, differ significantly on a number of dimensions that might affect financial decisionmaking or financial outcomes. I describe these briefly to provide some insight into why individuals might choose to join a particular service as well as to highlight why the financial behaviors and outcomes for members might vary by Service. I focus my attention on three important inter-service differences: the Service mission and associated job opportunities, the probability of overseas service (more likely for Army and Marine Corps members during the sample period) and the differences in assignment locations within the U.S. Since the Army is the largest (38% of Active Duty Forces, DMDC 2012) and perhaps the most typical military Service, I focus my analysis on the differences between each Service and the Army, as opposed to all of the potential inter-service differences. The Services each have unique operational responsibilities with similarly unique job opportunities. These opportunities will attract different types of individuals and these differences may correlate with differences in financial behaviors too. The differences in the probabilities of overseas unaccompanied service or deployment to combat locations have significant ramifications for individuals financial outcomes that differ by family considerations (i.e., the additional compensation and reduced expenditures may benefit single soldiers without dependents while the increased stress and marginal reductions in spending may harm families or individuals with children). The differences in assignment locations mean that military members and their families experience different economic conditions (e.g., different labor markets for spouses or military members seeking outside employment) and differential exposure to financial products (e.g., payday lending) and other laws with financial implications (e.g., no-fault auto insurance). These differences will also influence what type of individual joins each service and 13

14 so I again expect that individuals in different services will differ on observable and unobservable characteristics. In Table 2A, I present summary statistics that describe how military members in the different services differ on observable characteristics (Panel A) and in their financial outcomes (Panel B). [Insert Table 2A about here] Relative to the sample soldiers in the Army (Col. 2), Sailors in the Navy (Col. 5) are, on average, older, more often female, more likely to be married, more likely to be divorced, less likely to have children, less likely to be a minority, slightly less educated, and to have higher income. They also have higher financial literacy scores and lower financial confidence index scores relative to their Army counterparts. The outcomes for the members of the Navy (Panel B) reveal that they are less likely to report spending greater than income, less likely to report difficulty covering expenses and less likely to report having declared bankruptcy. They have fewer credit cards, comparable poor credit behaviors, are less likely to have a 3 month emergency fund and are equally likely to have non-retirement investments. Finally, with respect to AFS use, sailors are less likely to use payday loans, auto title loans and refund anticipation loans than Army members. Relative to Army members (Col. 2), Airmen in the Air Force (Col. 8) are, on average, older, less often female, more likely to be married, less likely to be divorced, less likely to have children, less likely to be a minority, more educated and to have higher income. Relative to Army members, they have higher financial literacy scores and financial confidence index scores. In their economic outcomes (Panel B), Air Force members, relative to their Army counterparts, are equally likely to report spending greater than income, less likely to report difficulty covering expenses and less likely to report having declared bankruptcy. They have fewer credit cards, 14

15 better credit behaviors, are more likely to have a 3 month emergency fund and are equally likely to have non-retirement investments. Finally, with respect to AFS use, Air Force members are less likely to use payday loans, auto title loans and refund anticipation loans than Army members. Relative to the soldiers in the Army (Col. 2), Marines (Col. 11) are, on average, younger, less often female, less likely to be married, less likely to be divorced, less likely to have children, less likely to be a minority, comparably educated and to have comparable incomes. They also have comparable financial literacy scores and financial confidence index scores relative to their Army counterparts. The financial outcomes (Panel B) suggest that Marine Corps members, relative to their Army counterparts, are more likely to report spending greater than income, equally likely to report difficulty covering expenses and less likely to report having declared bankruptcy. They have fewer credit cards, comparable credit behaviors, are less likely to have a 3 month emergency fund and are less likely to have non-retirement investments. Finally, with respect to AFS use, Marines are less likely to use payday loans, auto title loans and refund anticipation loans than Army members. The unconditional outcomes provided above are interesting, but they do not account for the many differences in observable characteristics between the services suggested by the means in Table 2A Panel A. To more precisely estimate the average differences between members of the different services I complete a multivariate regression analysis described in Equation 2. = (2) is the financial outcome of interest for individual., and are indicators that take on the value of 1 if an individual is in that service and equals 0 otherwise. Members of the Army are the comparison group (omitted category) in this regression. is the 15

16 same vector of individual characteristics as above, represents state fixed effects and represents a 2012 fixed effect. represents fixed effects for each military service component (Reserve and National Guard; Active Duty is omitted). Here, and are the parameters of interest and they reflect the average difference in each outcome between members of the Army and each of the other military services. Since the military service an individual joins is not randomly assigned and the coefficients should not be interpreted causally. Instead, they represent a combination of all factors that make members of each military service differ for each financial outcome. These differences could include individual differences among enlistees in cognitive abilities (e.g., average Armed Forces Qualification Test [AFQT] scores are higher for Air Force and Navy entrants than for the Army), in time preferences (e.g., individuals who join the Navy may be more career oriented than those who join the Army or Marines) and any institutional factors that differ by military service (e.g., financial education or counseling services). In Table 2B, I present ordinary least squares regression results for Equation 2 and I only report the coefficient estimates for the parameters of interest ( - ). [Insert Table 2B about here] Focusing on the potential differences between each Service and the Army, the regression results for the financial outcomes in Panel A suggest that, on average, members of the different services have few differences in their recent financial conditions. Members of the Navy, Air Force and Marines do not differ from their Army counterparts in the probability that they spend more than their income each month (Col. 1) or in their probability of having difficulty paying their bills (Col. 2); all of these estimates are statistically insignificant. With respect to the probability of declaring bankruptcy (Col. 3), Navy and Air Force members do not differ from their Army counterparts but Marine Corps members are 7.93 percentage points (139%) less 16

17 likely to have declared bankruptcy in the past 2 years than Army members and the difference is statistically significant (p<0.01). This difference might be explained by selection into the Marine Corps or by financial training and support services available to Marines. The regression results for Panel B also suggest very few differences in routine financial activities by service. Members of the Navy, Air Force and Marine Corps do not differ in statistically significant ways in their number of credit cards (Col. 1), their credit card behaviors (Col. 2), their probability of having a 3 month emergency fund (Col. 3) or the probability of having non-retirement investments (Col. 4). Finally, the results in Panel C suggest very few differences in AFS use by Service. Members of the Navy, Air Force and Marine Corps do not differ from Army members in the probability of having used payday loans (Col. 1). Air Force members, on average, have a 8.08 percentage point (53%) lower probability of having used an auto title loan (Col. 2) than their Army counterparts but the difference is marginally statistically significant (p=0.066). Navy and Marine Corps members do not differ significantly on this dimension. Navy members, on average, have a 6.37 percentage point (57%) lower probability of having used a refund anticipation loan (Col. 3) but this difference is also marginally statistically significant (p=0.060). Air Force and Marine Corps members do not have statistically significant differences in their reported use of refund anticipation loans (Col. 3) relative to Army members. The regression results from Table 2B suggest that the financial conditions for military members appear very similar across a wide variety of outcomes. To be clear, there are a few cases in which Air Force (i.e., lower use of auto title loans), Marine Corps (i.e., lower probability of having declared bankruptcy), and Navy (lower use of refund anticipation loans) personnel appear better off than their Army counterparts. Conversely, for none of the ten outcomes 17

18 evaluated here do Army personnel have statistically significant beneficial differences relative to members of the other services. This analysis cannot determine if this is due to differences in the unobserved characteristics of Army members (e.g., time preferences or self-control problems) or if the financial assistance available to Army members differs from those in the other military services (e.g., financial education, counseling services, or military relief society assistance). 3.3 How do financial outcomes and behaviors differ between military members by Component? The military service components differ substantially in their requirements and experiences for servicemembers. Individuals in the Active Duty component serve full time in the military, often live on a military installation, typically relocate to a new duty station every few years and deploy as required by their unit. Reserve and Guard component members normally serve in a more limited capacity, completing their service requirements during one weekend per month and a several week training period each year. However, these individuals can also be required to deploy overseas or to serve on longer mobilizations in the U.S. and many have done so over the past decade. They often hold civilian jobs or pursue education while serving and some reservists are full time military employees. Reservists and Guardsmen typically live near their unit, though these locations may be smaller than active duty military installations. Reserve forces units are directed and resourced by the federal authority, while National Guard units are directed primarily by their state. The military pay and benefits also differ by component commensurate with the service requirements. Given these different requirements and conditions, we would expect servicemembers to endogenously sort into the different components and to differ from one another in their observable and unobservable characteristics. To briefly highlight the observable differences, in 18

19 Table 3A, I present summary statistics on the observable characteristics by component. I will compare the differences in the individual characteristics and financial outcomes between members of the Reserve and National Guard components with members in the Active Duty. [Insert Table 3A about here] The results in Panel A suggest that on average, members of the Reserve component are younger, more often female, less likely to be married, have children or be a minority, comparably likely to be divorced, slightly more educated and to have lower income. They have lower financial literacy index and financial confidence index scores relative to their Active Duty counterparts. With respect to the economic outcomes evaluated in Table 3A (Panel B), Reserve members, relative to their Active Duty counterparts, are comparably likely to report spending greater than income, more likely to report difficulty covering expenses and less likely to report having declared bankruptcy. They have fewer credit cards, fewer poor credit behaviors, are less likely to have a 3 month emergency fund and are less likely to have non-retirement investments. Finally, Reservists are slightly more likely to use payday loans, but about equally likely to use auto title loans and refund anticipation loans. Relative to their Active Duty counterparts, members of the National Guard are older, more often female, less likely to be married, less likely to be divorced, less likely to have children, less likely to be a minority, less educated and they have lower incomes.. Like the National Guard members, Reservists have lower financial literacy index and financial confidence index scores relative to their Active Duty counterparts. In their economic outcomes (Table 3A Panel B), National Guard members, are more likely to report spending greater than income, more likely to report difficulty covering expenses and more likely to report having declared bankruptcy relative to their Active Duty counterparts. They have more credit cards, fewer poor credit behaviors, are 19

20 equally likely to have a 3 month emergency fund and are equally likely to have non-retirement investments. Finally, with respect to AFS use, National Guard members are more likely to use payday loans, auto title loans and refund anticipation loans As with the previous analyses, these unconditional differences in financial outcomes fail to account for the many differences in the members by component. To more precisely estimate these average differences I complete regression analysis as described by Equation 3. = (3) represents the financial outcome of interest for individual. and are indicators that take on the value of 1 if an individual is in that service component and equals 0 otherwise. Members of the Active Duty component are the omitted category in this regression. is the vector of individual characteristics, represents state fixed effects and represents a 2012 fixed effect. represents fixed effects for each military service (Navy, Air Force and Marines; the Army is omitted). Here and are the parameters of interest and they reflect the average difference in each outcome between members of the Active Duty component and the other military service components. Military service components are endogenously determined and so the coefficients do not have causal interpretations. They represent a combination of all factors that make members of each military service component differ for each financial outcome. These differences could include differences in cognitive abilities (e.g., Reserve and National Guard members may have more financial experience since they are typically older and often hold civilian jobs), risk preferences (e.g., Active Duty members typically deploy more than their Guard and Reserve counterparts) and any institutional factors that differ by military service (e.g., Active Duty members may have access to more financial services since they typically work at lager installations with more resources). In 20

21 Table 3B, I present ordinary least squares regression results for Equation 3, restricting my attention to the parameters of interest (, ). [Insert Table 3B about here] I focus my service component analysis on the potential differences between individuals serving on Active Duty and those serving in the Reserves or the National Guard, as opposed to all of the potential inter-component differences. 16 The results in Panel A suggest that on average, members of the different service components have few differences in their recent financial conditions. Members of the Reserves and National Guard do not differ, on average, from Active Duty members with respect to spending more than their monthly income (Col. 1) or having difficulty paying their bills (Col. 2). Reserve component members, on average, have a 7.46 percentage point (131%) lower probability of having declared bankruptcy in past two years and the difference is statistically significant (p<0.05), but National Guard members do not have statistically different probabilities of declaring bankruptcy than Active Duty members. The results for the differences in Panel B suggest some important differences between members from the difference components with respect to more routine financial behaviors. On average, Reserve component members have 0.75 (23%) fewer credit cards (Col. 1) than members of the Active Duty and the difference is statistically significant (p<0.01). On average, Reserve component members have 0.43 (21%) fewer bad credit behaviors (Col. 2) and are 9.17 percentage points (20%) less likely to have non-retirement investments (Col. 4) than members of the Active Duty, but the differences are marginally statistically significant (p=0.059 and p=0.068 respectively). Reservists have comparable probabilities of having a 3 month emergency fund 16 I classified military servicemembers as Reservists or Guardsmen based on their self-identified component (variable AM7), regardless of if they were activated and serving on Active Duty at the time of survey response. This coding could make the estimated differences in this paper lower bound estimates of the actual differences in service members outcomes and decisions by component. 21

22 (Col. 3). National Guard members do not differ from their Active Duty counterparts in the number of credit cards, the probability of having a 3 month emergency fund or the probability of having non-retirement investments. They do have 0.60 (28%) fewer bad credit behaviors (Col. 2) than members of the Active Duty, but the difference is marginally statistically significant (p=0.087). Finally, the results from Panel C suggest that there are no differences in AFS use by members of the different components. The point estimates for Reservists suggest overall lower use of payday loans (Col. 1), auto title loans (Col. 2) and refund anticipation loans (Col. 3) relative to Active Duty members, but none of the differences are statistically significant. Conversely, the point estimates for National Guard members suggest overall higher use of payday loans, auto title loans and refund anticipation loans relative to Active Duty members, but none of these differences are statistically significant either. 4. Discussion & Summary Summary Using data from the 2009 and 2012 National Financial Capability Study Military and State-by-State Surveys and multivariate regression analysis, this research investigates three important questions related to household financial decision-making and finds a nuanced set of answers to each question. First, how do financial outcomes and behaviors differ between military and civilian households? I find that military members, relative to their civilian counterparts, are less likely to have difficulty paying their bills, more likely to have a 3 month emergency fund and more likely to have non-retirement investments. However, military members also have more credit cards and poorer credit card behaviors. The groups are comparable on a number of 22

23 dimensions including the probability of spending more than their income, the probability of declaring bankruptcy and their use of alternative financial services. Second, how do financial outcomes and behaviors differ between military members by Service? I find that overall, there are few differences in behavior by members of the different Services. Members of the Navy, compared to members of the Army, have fewer credit cards and are less likely to have used a refund anticipation loan; their behavior is comparable for the other eight outcomes analyzed. Members of the Air Force, compared to members of the Army, are less likely to have used an auto title loan; their behavior is comparable for the other nine outcomes analyzed. And members of the Marine Corps, compared to members of the Army, are less likely to have declared bankruptcy; their behavior is comparable for the other nine outcomes analyzed. Finally, how do financial outcomes and behaviors differ between military members by Component? In this case I find more similarities than differences by component. Members of the Reserve component, compared to Active Duty members, are less likely to have declared bankruptcy, they have fewer credit cards and they have fewer poor credit behaviors. But they are less likely to have non-retirement savings. Their behavior is comparable for the other six outcomes I analyze. Members of the National Guard have fewer poor credit card behaviors than Active Duty members, but their behavior is comparable for the other nine outcomes I analyze. Research Challenges It is important to highlight several shortcomings in this analysis. First, as mentioned above, survey respondents may not be representative of the groups to which they belong (e.g., representative of their individual Services or the DOD population as a whole). In the case of the military and civilian comparisons, the fact that the survey was completed online may exclude some of the lowest income individuals who lack access to computers and the internet. This 23