Gender Differences in Attitudes toward Debt and Financial Position: The Impact of the Great Recession Julie Lyn Routzahn McDaniel College and Mary Eschelbach Hansen American University This version: October 16, 2014 Keywords: Gender, attitudes towards debt, household debt, Great Recession, feminist methods JEL Codes: B54, D12, J16 Acknowledgements: Comments from Tom Husted, Robert Feinberg, and Martha Starr are much appreciated. Tiffany Chang provided research assistance. 1 Abstract We use the 2007 and 2010 Surveys of Consumer Finances (SCF) and a difference-in-differences approach to investigate whether the Great Recession differentially affected the attitudes of men and women towards taking on debt, as well as whether it affected the relative financial position of women. In order to separate the influence of gender per se from the influence of household composition and intra-household bargaining, we compare never-married women to nevermarried men. We also compare never-married women to all households. We find that the Great Recession gave never-married women the confidence to use debt to mitigate short-term threats to living standards while making them more cautious about taking on debt for non-essentials. The financial position of women in 2010 improved relative to the position of men and relative to all households 2 Gender Differences in Attitudes toward Debt and Financial Position: The Impact of the Great Recession Introduction We investigate gender patterns in the use of consumer debt and attitudes towards it. Alternatives for consumer borrowers have been expanding since the 1980s (Ryan, Trumbull, and Tufano 2011), and the share of household income going towards required payments on debt has tripled between 1983 and 2007 (Dynan 2009). On the eve of the Great Recession, households had more debt than they had before previous downturns (Moore and Palumbo 2010), but there was no growth in total debt during the recession. Pay-down of mortgage debt was significant, while there was a small increase in average non-mortgage debt (Becker and Shabani, 2009; Chakrabarti et al. 2011). Poorer consumers and younger consumers, in particular, continued to increase their leverage (Wolff 2011). We find that patterns differ among women and men. We show that, after controlling for family composition, women did not leverage up and their debt-to-income ratios improved. We argue that a central reason why women’s financial position improved relative to men’s is that the experience of the Great Recession made women more cautious about borrowing for nonessentials. We offer a new strategy for identifying such gender differences. We abstract from the “blackbox” of household decision-making by comparing the attitudes and financial positions of never-married women to those of never-married men. This strategy minimizes the impact of partners on attitudes and financial positions. We thereby establish a baseline of behavior to which families, in all their complexity, can be compared. 3 Framework Like most work on personal finance, our investigation has at its foundation the permanent income hypothesis and life cycle models: A person maximizes lifetime utility by saving and borrowing to smooth consumption (Friedman 1957, Modigliani and Brumberg 1954). Preferences such as “impatience” lead some people to borrow more than others, although creditors are not always willing to lend as much as a borrower desires (Deaton 1992). Additionally, the stream of income and expenses is uncertain, so some cushion of savings may be desirable, even during stages of the life-cycle when borrowing dominates behavior (Carroll and Summers 1991, Carroll 1992, Deaton 1991). In a fully-fleshed-out buffer-stock model, when people with “prudent” preferences (that is, people who prefer to ensure that consumption does not fall below some floor) face a long-lasting increase in uncertainty, they increase their desired level of savings at all stages in the life cycle. The empirical work that follows is motivated by such a buffer-stock model of behavior, in which preferences such as “impatience,” “prudence,” and attitudes about the acceptability of using debt, or debt “tolerance,” are considered to be closely intertwined. Preferences and attitudes are allowed to change over time as a result of macroeconomic phenomena, personal experience, and cultural change. Preferences and attitudes are also allowed to differ by gender at a point in time and across time (Giannetti 2014). An improvement in financial position is expected to result from increased prudence or reduced tolerance for debt, all other things equal. The Tolerance of Women for Debt The empirical literature on the determinants of debt has documented life-cycle patterns of borrowing. Being in the middle stages of the life-cycle—with higher income, homeownership, and a larger number of children at home—is associated with more debt. But the level of debt 4 differs widely across people at the same stage of the life cycle (Crook 2001). Risk preferences and impatience matter (Bertaut and Starr-McCluer 2002, Brown et al. 2005, Meier and Sprenger 2010). Expectations seem to contribute to within-stage differences: Optimism about one’s financial future is associated with higher levels of debt (Brown et al. 2005, 2008). The literature on debt tolerance as a determinant of borrowing behavior is surprisingly thin, particularly in light of the importance of consumer credit to the economy (Glick and Lansing 2009, Mian and Sufi 2013) and the concerns of regulators about high levels of consumer debt (for example, Durkin et al. 2014). The few studies that directly evaluate the role of attitudes towards debt yield mixed results. In both American and British cross-sections, tolerance of debt is correlated with higher debt-to-asset and debt-to-income ratios (Livingstone and Lunt 1992, Schooley and Worden 2010). People seem to acclimate to debt—they develop positive attitudes towards debt when they successfully use it (Davies and Lea 1995). However, these findings are inconsistent with the observation that, as consumer debt grew in the 1980s and 1990s, attitudes towards debt in U.S. became more negative (Bloom and Steen 1987, Goodwin 1997). Our paper contributes by measuring the extent to which gender influences attitudes towards debt. Previous studies of gender differences in financial matters have examined relative risk aversion and its influence on investment behavior (for example, Prince 1993, Jianakoplos and Bernasek 2007) and gender differences in financing among entrepreneurs (examples include Coleman and Robb 2009: Madill, Riding, and Haines Jr. 2006). However, little is known about how gender influences attitudes toward debt. There is emerging evidence that women feel more burdened by debt after it is incurred (Keese 2012) and that women use part of newly-acquired welfare payments to pay down debt while men do not (Lyons and Fisher 2006; Shaefer, Song, and 5 Shanks 2013). These findings lead us to suspect that women’s responses to the Great Recession may be different from men’s. Strategy for Detecting Gender Differences in Tolerance for Debt We are particularly interested in separating the effects of gender per se from other influences on women’s attitudes and behavior. Among potential influences, the most difficult and important to separate from gender is the influence of family dynamics and intra-household bargaining. Within families money matters have strong gender dimensions. For example, Goode (2010) demonstrates that women are more likely to manage household finances in poor families where it is a burden rather than a source of power. Similarly, Thorne (2010) finds that financial chores of financially distressed households are almost always performed by women. Zagorsky (2003) even finds that husbands report more assets and less debt than their wives. As noted by Doss et al. (2008), to conduct the best possible gender research, it would be desirable to survey extensively all members of the household about assets, debts, and attitudes. However, no such survey of U.S. households is available. Building on Routzahn (2011), we suggest that it is possible to separate family dynamics from other aspects of gender by measuring the differences between women and men who have never been married and who are not currently cohabiting. While our strategy cannot remove the influence of past relationships or childhood experiences, it does eliminate the direct impact of spousal opinions and decisions. It establishes a baseline of behavior among never-married individuals to which future research can be compared. We apply this strategy to data from the 2007 and 2010 Surveys of Consumer Finances (SCF). 6 Data from the Surveys of Consumer Finances The Surveys of Consumer Finances (SCF) are among the most comprehensive financial surveys available (Lindamood et al. 2007; Board of Governors 2009, 2012). The surveys solicit responses from the one member of the household that the interviewer establishes as the most financially knowledgeable. From this respondent, the interviewer collects detailed information about all forms of household assets and liabilities, as well as responses to questions on the attitudes towards debt of the respondent himself or herself.1 To study gender differences and changes in attitudes toward credit, we use answers of respondents to questions about how they feel about borrowing to finance different kinds of expenditures. Respondents answered “yes” or “no” to each part of the following question: Tell me whether you feel it is all right for someone like yourself to borrow money: To finance the purchase of a car. To finance educational expenses. To cover living expenses when income is cut. To finance the purchase of a fur coat or jewelry.2 To cover the expenses of a vacation. 1 Because the SCF questions involve sensitive financial data that respondents are often reluctant to provide, many variables have missing information. If the respondent does not give an exact answer, interviewers try to get an approximation using ranges of values, but some respondents still do not provide answers to some questions. To account for missing information and to ensure the privacy of respondents, each response is replicated five times. For details, see Rubin (1987). We use the procedures for using the replicates described in Board of Governors (2009, 2012). Additionally, since the survey oversamples the wealthy, all reported results are weighted using the population weight assigned to each observation by the SCF. Use of multiple imputations is gaining popularity as researchers need a reliable way to deal with missing data and ensure privacy. However, , some concepts that are well-defined in simple regression analysis do not have a comparable in multiple imputation analysis (StataCorp 2009). Goodness-of-fit statistics are one such example. We report the average of the goodness-of-fit indicator over all five imputations (see for example, Lee and Carlin 2010). 2 Previous research using this question for example, Chien & Devaney (2001) interprets the question as asking about attitudes about borrowing to finance the purchase of a luxury, but the specific wording may be gender-biased. For example, a male respondent might feel differently about borrowing to finance jewelry than he would feel about borrowing to finance a boat or motorcycle. 7 Figure 1 summarizes the responses. The majority of all respondents in both the 2007 and 2010 SCF believed it was acceptable to borrow money for a vehicle and for education, but only singledigit percentages believed it was acceptable to borrow for a luxury. The most positive attitudes about any kind of debt were those of never-married women regarding debt-finance of education: In 2007 over 88 percent found student debt acceptable. In 2010, 86 percent found it acceptable. There was only one other statistically significant increase in debt tolerance between 2007 and 2010: never-married men became more tolerant of debt-finance of vacations. [Figure 1 about here] Of course, part of the difference between the attitudes of never-married respondents—men or women—and married respondents may be due to differences in their average ages. The average age of married respondents was just shy of 50, while the average age of never-married men and women was about 38. The average age of respondents in 2010 did not differ meaningfully from the age of respondents in 2007. Education levels, which also may be related to preferences, also did not differ appreciably between respondents of the two surveys. Never-married women and men had a fraction of a year more education on average than married respondents. However, the household income of never-married respondents was less than half the household income of married households. Descriptive statistics of demographics, socio-economic status, and other controls included in the analysis below appear in Appendix A. To study changes and differences in the financial positions of households, we use the SCF data to compute standard financial ratios including loan balances relative to total assets, loan balances relative to total income, and monthly loan payments relative to monthly income. We use the definitions that are used in the Bulletin (Board of Governors 2009, 2012) in order for this work to 8 be comparable to the greater literature that uses the SCF. Assets include both financial and nonfinancial assets. Financial assets include certificates of deposit, stocks, bonds, savings accounts, money market accounts, and so on. Non-financial assets include residential real estate, other real estate investments, and business assets. All income sources are included: wage income, interest and dividend income, self-employment and business income, pension income, realized capital gains, government program income (social security and food stamps), and alimony. The means of the major financial ratios, by respondent group, are shown in Figure 2. Households headed by never-married women do not have high debt-to-income ratios, but they are the most leveraged by far. This mainly reflects their low levels of assets. Women tend to hold fewer assets than men, resulting in a gender asset gap (Deere and Doss 2006). Assets are important to households because they can appreciate in value, generate income, and serve as a source of collateral. Additionally, assets provide a level of security in that they buffer against emergencies. They can also serve as a way to increase productivity and improve the ability of the owner to earn a living. With less income and fewer assets, women are more financially vulnerable than men. The asset gap is partly explained by gender differences in preferences: Women exhibit more risk-averse behavior than men in the choice of their asset portfolio, which results in lower returns than men (Jianakoplos and Bernasek 2007). It also appears that women experience statistical discrimination (Phelps 1972) when they receive financial and legal advice for investments, insurance, loans, and bankruptcy, since they are viewed as the “safer sex” (Lefgren and McIntyre 2009). Women are more likely to use financial planners than men found, either because of their lack of financial education or because of greater concern about the future (MacGregor and Slovic 9 1999). Men feel more competent in financial matters than women, and consequently they are willing to take risks to amass wealth (Prince 1993). The main take-away from Figure 2 is that the average household of a married couple or of a never-married man in the 2010 survey was in worse position than the corresponding households had been in 2007. In contrast, the households of never-married women were less leveraged in 2010 than their counterparts had been in 2007. The average income-to-debt ratio of nevermarried women remained about the same, but in 2010 the average never-married man had a higher monthly debt-to-income ratio than he had in 2007. Moreover, in 2010 monthly debt burden of the typical never-married man was higher than the burden carried by typical nevermarried woman that year. [Figure 2 about here] This is a dramatic reversal in financial status of never-married women relative to men. It further supports research showing that women fared better during the most recent recession. For example, they had relatively favorable employment outcomes and higher labor force participation than men (Elsby, Hobijn, and Sahin 2010; Starr 2014). These favorable employment outcomes are attributed to gender differences in industry and occupation (Hoyes et al. 2012, Albanesi and Sahin 2013). That is, women were employed in areas that were not hurt as much by the recession, whereas mean were employed in occupations that were more affected by it. 10 Gender Differences in Attitudes toward Debt We use a difference-in-difference approach to evaluate whether the Great Recession had a differential impact on the attitudes towards debt of never-married women and men. Call a positive response to a question about attitudes P. We estimate by probit 𝑃𝑟𝑜𝑏(𝑃𝑖𝑡 ) = 𝛽0 + 𝛽1 𝐹𝑖 + 𝛽2 𝑌𝑒𝑎𝑟𝑡 + 𝛽3 𝑌𝑒𝑎𝑟𝑡 ∗ 𝐹𝑖 + 𝛾𝑋𝑖 + 𝑢𝑖𝑡 , where F indicates that the respondent is a never-married female and Year indicates whether the observation is from the 2010 SCF. The estimate of particular interest is 𝛽3, the coefficient on the interaction between being a never-married female and being observed in 2010, after the Great Recession. If 𝛽3 is negative and statistically significant, it we conclude that the attitudes towards credit of never-married women grew more conservative than did the attitudes of never-married men over the course of the recession. The vector X includes respondent-specific and household-specific controls. In addition to sex and marital status of the respondent, we control for age, education, race, and number of children. We also control for expectations about the future. Expectations are represented by self-reported life expectancy, current health status, opinion on future economic conditions, and respondent’s prediction of the direction of change of interest rates. Additionally, we control for employment status (whether the respondent has been unemployed in the last 12 months), credit worthiness (whether the respondent has been turned down for credit in the last five years), and job quality (proxied by access to employer-sponsored retirement and health insurance plans). In the top panel of Table 1 we show the marginal effects when the estimation sample is restricted to never-married respondents. Again, this is our strategy for isolating the effects of gender and gender norms from the effects of the dynamics of the respondent’s adult relationships. In the 11 bottom panel we show results comparing the households of respondents who are never-married women to the households of all other respondents in the SCF. [Table 1 about here] Consider first the top panel of Table 1. At baseline, never-married women were about as accepting of loans for cars, vacations, and living expenses as never-married men, but more accepting of student loans and less accepting of loans for luxuries. Attitudes towards credit were more conservative in 2010 than they had been in 2007, which is consistent with the work of Malmendier and Nagel (2011) who find that an experience of economic hardship reduces acceptance of borrowing. The probability that a respondent approved of loans for cars and living expenses fell by 7 to 8 percentage points. Only attitudes towards educational debt were unaffected by the recession. The recession did not have a gender-specific effect on attitudes about borrowing for cars, education, or luxuries. However, compared to never-married men, never-married women became more tolerant of debt to finance living expenses when income is cut and less tolerant of debt to finance vacations. Women’s experiences during the recession gave them the confidence to use debt to mitigate short-term threats to living standards, but made them more cautious about financing non-essentials. This result mirrors the employment success women experienced during the recession as compared to men. This change in attitudes is not driven by the composition of the sample: the demographic and socio-economic attributes of never-married women in the 2007 sample was not different from the attributes of women observed in the 2010. (See Appendix A.) We know that documented gender differences exist in wage income, investment strategy, risk propensity, asset ownership, 12 availability of health insurance and retirement benefits, just to name a few. Men and women also differ in their responsibilities to care for children and older family members. Ultimately these differences result in a disparity of available resources between never-married women and nevermarried men. Therefore, it is not possible to capture every component of possible difference in the regression analysis. Turning to the results from the full sample shown in the bottom panel of Table 1, we see that never-married women were more accepting of debt in 2007 than were all respondents. This may be because never-married women must rely on the marketplace to a greater extent than married and cohabitating respondents, who may have more secure safety nets through larger extended families, the possibility of having two incomes, and greater access to better employment benefits, such as health insurance and retirement plans. The trend towards more conservative attitudes was not limited to never-married respondents, but the size of change in attitudes between samples is smaller when all respondents are considered than when only never-married respondents are considered. Again, this likely reflects the younger age of the never-married group, who were among the most hard-hit during the recession. The diff-in-diff coefficient in the bottom panel is negative in four of the five regressions, though it is statistically significant in only the regression for attitudes about borrowing for luxuries. This reinforces the results from the restricted sample: the experience of the recession caused nevermarried women to become more cautious about borrowing for non-essentials. Gender Differences in Financial Position To investigate whether the changes in attitudes that we observe resulted in changes in the financial position of never-married women, we use OLS to estimate a difference-in-difference 13 specification similar to the logit specification above, but with household financial ratios as the dependent variables and attitudes towards debt as explanatory variables. Table 2 shows results for the leverage ratio and debt-to-income ratios. Again, we follow the SCF variable definitions to make our work comparable to the greater literature. These dependent variables are calculated as ratios and are shown as percentages in the summary tables. Table 3 shows results for three types of debts: car loans, student loans, and credit card debt. In addition to the demographic and socio-economic controls used in the probit models of Table 1, the models and also control for credit-worthiness. [Table 2 about here] As we saw in Figure 2, at baseline never-married women were in worse financial health than never-married men. This is true whether we consider leverage, overall debt-to-income, or monthly obligations-to-income. The overall de-leveraging in response to the Great Recession that has been noted elsewhere (Becker and Shabani 2010; Chakrabarti et al. 2011) is not evident for never-married men, but it is evident in the full sample and for never-married women. These households have been ridding themselves of debt through bankruptcy and the short sale of homes. At the end of 2009, household debt stood at 122.5% of disposable income, down from its peak of 130.6% (Whitehouse 2010). Never-married women reduced both outstanding obligations relative to annual income and monthly required payments relative to monthly income by an economically significant extent. Because never-married women have few assets, the debt-to-income measures most accurately 14 reflect their relative financial position.3 The reduction of debt can come from two sources: Individuals can pay down the debt or they can rid themselves of the asset that resulted in the debt. We know that many households rid themselves of debt associated with housing. When we take both gender and the gender*recession contribution into account, we find that never married women reduced their debt to income ratio by 40%. For the average never married female, that amounts to a drop in debt of approximately $13,000. Income changed very little. The majority of deleveraging came from the reduction of mortgage debt. This has both positive and negative effects. While it is true that women improved their household balance sheets by decreasing debt it was a type of debt directly tied to an asset that (historically) would have appreciated in value over the long term. When we look more closely at the monthly debt to monthly income ratio, we see that nevermarried women reduced their monthly debt payments by 2 percent of income. This is not as large as one may expect considering the drop in the leverage ratio. However, it was probably the case that never married-women were at an early stage in their mortgage amortization. They were likely paying down very little debt each month, with the majority of the payment going to interest. It will be a matter for future work to discover whether the short run gain from smaller monthly payments was worth the sacrifice of the underlying housing assets for these women. To control for attitudes in the regressions for overall household financial ratios, we sum the individual attitudes responses to create an index of debt tolerance. We consider a respondent with one or zero positive responses to the attitudes questions to have conservative attitudes 3 Given that asset values, especially real estate and stock values, dropped substantially during the recession, more analysis is necessary to determine the overall financial health of the household balance sheet. Such an analysis of changes in asset portfolios is beyond the scope of the current paper. 15 towards debt. Those with two or three positive responses have moderate attitudes; those with four or more positive responses have liberal attitudes. Although it is intuitive to expect more liberal attitudes towards credit to be associated with higher debt relative to assets and income (see, for example, Rutherford and DeVaney 2009), we find no statistically significant relationships for the sub-sample of never-married respondents. The signs of the coefficients are negative, however, indicating that never-married respondents with moderate-to-liberal attitudes are most likely credit-constrained: they might like more credit, but cannot obtain it. Among all respondents, liberal credit attitudes have the expected positive effect. Lastly, Table 3 considers the types of debt most relevant to the greatest number of never-married respondents: car loans, student loans, and credit card debt. At baseline, never-married women had greater obligations than never-married men in all three categories, with the exception of car loans to total assets. Compared to all respondents, never-married women had fewer obligations in all three areas at baseline. [Table 3 about here] By 2010, however, never-married women had slightly larger car loans than never-married men or other respondents. Note that these loans did not represent more of a drain on income. Primarily, this reflects the lower income of never-married men in 2010 compared to 2007, since attitudes about the acceptability of auto loans became more conservative among all groups between 2007 and 2010. Never-married women observed in 2010 did not have more student loan debt compared to assets. Moreover, women’s student loans in 2010 placed less of a drag on their incomes than had been the case in 2007. The differential is quite large—21 to 26 percent of income. It is important to 16 note again that the average years of education among women surveyed did not fall; in fact, the distribution of years of education shifted slightly to the right. We see that education debt to assets increased for all three groups, as many individuals returned to school or continued their education during the recession. It appears that the entry of recently college graduates into the labor force and into sample was more than offset by the pay-down of student debt by other women. In Table 1, we saw that the Great Recession caused never-married women to become more accepting of using credit to smooth essential consumption, but less accepting of using it for nonessentials. The final set of regressions in Table 3, which show differences and changes in credit card debt, show that women’s increased conservatism dominated their monthly budgeting decisions. Never married women and men both decreased their dependence on credit card debt relative to their assets and total. Moreover, although never-married women were more burdened by credit card debt in 2007, they decreased their credit card obligations relative to total income and monthly income. Never-married men did not. Never married women even reduced their monthly obligations by more than all households did. In sum, we find that the relative position of the finances of never-married women improved over the course of the Great Recession. These women’s incomes changed little, but they paid down their debt obligations significantly. Among households generally, the story is less rosy. Overall, the debt-to-income ratio fell as income fell, but the debt-to-asset ratios did not change. This implies that many households used assets to pay down debt. Fewer assets means less insurance against future risk, whether those risks are idiosyncratic (such as health risks) or systemic (such as unemployment). 17 Conclusion We have shown that the Great Recession increased the tolerance of never-married women for borrowing to meet living expenses but decreased their tolerance for borrowing to obtain nonessential items. The latter finding is reinforced by comparison of the attitudes of never-married women to all households. A reduced tolerance for debt-finance of luxuries, combined with the relatively favorable employment outcomes and higher labor force participation of women during the Great Recession , led to an improvement in the relative financial position of never-married women from 2007 to 2010. These results are important on several fronts. First, they illustrate how essential gender analysis is for the understanding economic outcomes. Our work adds to a growing body of literature that reinforces the need for additional research through a gender-sensitive lens. Second, the results make an important contribution to gender-sensitive research in financial economics. Previous research has emphasized that women are more vulnerable to economic shocks because they own fewer assets, are more likely to be underemployed, take less risk, and take on more caring responsibilities. This research shows that, despite these vulnerabilities, women are able to leverage opportunities to improve their balance sheets relative to men when circumstances permit. The results have several implications for women’s future status. The improvement in financial stability that we observe for women has the potential to lead to more favorable outcomes for them in a variety of areas. For example, we expect that improved financial stability will make it easier for women to acquire assets in the future, thus reducing financial insecurity in old age. We also expect that improved financial stability will increase the bargaining power of never- 18 married women entering into marriage, thus reducing domestic abuse and divorce while improving outcomes for children. Finally, the methodology used in this paper is important for feminist economics. We demonstrate that it is possible to establish a baseline for measuring gender influences on behavior by comparing never-married women to never-married men. This method can be used to move other applied gender research forward while we await the development of gender disaggregated surveys and the dissemination of their results. Future research may focus more directly on gender differences in the debt-finance of assets. The Survey of Consumer Finances, for example, includes information on installment debt, revolving debt and other debt (which consists of loans against pensions, life insurance, and margin loans). Identifying gender differences in financing will be critical for future recommendations on credit market regulation and for financial education. 19 References Becker, Thomas A., and Reza Shabani. 2010. “Outstanding Debt and the Household Portfolio." Review of Financial Studies 23(7): 2900-2934. Bertaut, Carol C., and Martha Starr-McCluer. 2002. “Household Portfolios in the United States.” In Luigi Guiso, Michael Haliassos, Tullio Jappell (eds.) Household Portfolios, pp. 181218. Cambridge: MIT Press. Bloom, David E. and Todd P. Steen. 1987. “Living on Credit,” American Demographics 9 (101): 22-29. Brown, Sarah, Gaia Garino, Karl Taylor, and Stephen Wheatley Price. 2005. “Debt and Financial Expectations: An Individual and Household-Level Analysis.” Economic Inquiry 43(1): 100-120. Brown, Sarah, Gaia Garino, and Karl Taylor. 2008. “Mortgages and Financial Expectations: A Household-Level Analysis." Southern Economic Journal 74(3): 857-878. Carroll, Christopher. 1992. “The Buffer Stock Theory of Saving: Some Macroeconomic Evidence.” Brookings Papers on Economic Activity, 2: 61–135. Carroll, Christopher, and Lawrence Summers. 1991. “Consumption Growth Parallels Income Growth: Some New Evidence,” in B. Douglas Bernheim and John B. Shoven. (eds.) National Saving and Economic Performance, pp. 305-348. Chicago: University of Chicago Press. Chakrabarti, Rajashri; Donghoon Lee; Wilbert van der Klaauw, Basit Zafar, Basit. 2011. “Household Debt and Saving during the 2007 Recession.” Staff Report, Federal Reserve Bank of New York, No. 482, http://hdl.handle.net/10419/60813. Coleman, Susan, and Alicia Robb. 2009. “A Comparison of New Firm Financing by Gender: Evidence from the Kauffman Firm Survey Data.” Small Business Economics 33(3): 397411. Crook, Jonathan. 2001. “The Demand for Household Debt in the USA: Evidence from the 1995 Survey of Consumer Finance.” Applied Financial Economics 11(1): 83-91. Davies, Emma, and Stephen E.G. Lea. 1995. “Student Attitudes to Student Debt.” Journal of Economic Psychology 16(4): 663-679. Deaton, Angus. 1991. “Saving and Liquidity Constraints.” Econometrica, 59(5): 1221–48. Deaton, Angus. 1992. Understanding Consumption. Oxford University Press Inc. Doss, C., C. Grown, and C. D. Deere. 2008. “Gender and Asset Ownership: A Guide to Collecting Individual-Level Data.” World Bank. Available: 20 https://openknowledge.worldbank.com/handle/10986/3468 (last accessed October 16, 2014). Dynan, Karen E. 2009. “Changing Household Financial Opportunities and Economic Security.” The Journal of Economic Perspectives 23(4): 49-68. Durkin, Thomas A., Gregory Elliehausen, Michael E. Staten, and Todd J. Zywicki. 2014. Consumer credit and the American economy. New York: Oxford University Press. Elsby, Michael W. L., Bart Hobijn, and Aysegul Sahin. 2010. “The Labor Market in the Great Recession,” Brookings Papers on Economic Activity, Economic Studies Program, The Brookings Institution 41(1): 1-69. Friedman, Milton. 1957. A Theory of the Consumption Function. Princeton: Princeton University Press. Giannetti, Caterina, 2014. “Time Preference Instability, Financial and Working Status.” Quaderni - Working Paper DSE N° 924. Available at SSRN: http://ssrn.com/abstract=2394541 or http://dx.doi.org/10.2139/ssrn.2394541 Glick, Reuven, and Kevin J. Lansing.. 2009. “U.S. Household Deleveraging and Future Consumption Growth,” Federal Reserve Bank of San Francisco (FRBSF) Newsletter, No. 2009-16 (May 2009): 1 Goode, Jackie. 2010. “The Role of Gender Dynamics in Decisions on Credit and Debt in Low Income Families.” Critical Social Policy 30(1): 99-119. Godwin, Deborah D. 1997. “Dynamics of Households' Income, Debt, and Attitudes Toward Credit, 1983–1989." Journal of Consumer Affairs 31(2): 303-325. Jianakoplos, N. A., and A. Bernasek. 2007. “Are Women More Risk Averse?” Economic Inquiry 36(4): 620-630. Keese, Matthias. 2012. “Who Feels Constrained by High Debt Burdens? Subjective vs. Objective Measures of Household Debt.” Journal of Economic Psychology 33(1): 125-141. Board of Governors of the Federal Reserve System. 2009. “Codebook for the 2007 Survey of Consumer Finances.” Available: http://www.federalreserve.gov/econresdata/scf/scf_2007documentation.htm (last accessed October 16, 2014). Board of Governors of the Federal Reserve System. 2012. “Codebook for the 2007 Survey of Consumer Finances.” Available: http://www.federalreserve.gov/econresdata/scf/scf_2010documentation.htm (last accessed October 16, 2014). 21 Lee, Katherine J., and John B. Carlin. 2010. “Multiple Imputation for Missing Data: Fully Conditional Specification Versus Multivariate Normal Imputation.” American Journal of Epidemiology 171(5): 624-632. Lindamood, Suzanne, Sherman D. Hanna, and Lan Bi. 2007. “Using the Survey of Consumer Finances: Some methodological considerations and issues.” Journal of Consumer Affairs 41(2):: 195-222. Livingstone, Sonia M., and Peter K. Lunt. 1992. “Predicting Personal Debt and Debt Repayment: Psychological, Social and Economic Determinants.” Journal of Economic Psychology 13(1): 111-134. Lyons, Angela C., and Jonathan Fisher. 2006. “Gender Differences in Debt Repayment Problems after Divorce." Journal of Consumer Affairs 40(2): 324-346. Madill, Judith J., Allan L. Riding, and George H. Haines Jr. 2006. “Women Entrepreneurs: Debt Financing and Banking Relationships." Journal of Small Business & Entrepreneurship 19(2): 121-142. Malmendier, Ulrike, and Stefan Nagel. 2011. "Depression Babies: Do Macroeconomic Experiences Affect Risk Taking?*." The Quarterly Journal of Economics 126(1): 373416. Meier, Stephan, and Charles Sprenger. 2010. “Present-Biased Preferences and Credit Card Borrowing.” American Economic Journal: Applied Economics 2(1): 193-210. Mian, Atif, Kamalesh Rao, and Amir Sufi. 2013. “Household Balance Sheets, Consumption, and the Economic Slump.” The Quarterly Journal of Economics 128(4): 1687-1726. Modigliani, Franco, and Richard Brumberg. 1954. “Utility Analysis and the Consumption Function: An Interpretation of the Cross-Section Data,” in K. Kurihara (ed.) PostKeynesian Economics. New Brunswick: Rutgers University Press. Moore, Kevin B., and Michael G. Palumbo. 2010. “The Finances of American Households in the Past Three Recessions: Evidence from the Survey of Consumer Finances.” Division of Research & Statistics and Monetary Affairs, Federal Reserve Board. Available: http://www.federalreserve.gov/PUBS/FEDS/2010/201006/201006pap.pdf (last accessed August 25, 2014). Prince, Melvin. 1993. “Women, Men and Money Styles.” Journal of Economic Psychology 14(1): 175-182. Routzahn, Julie Lyn. 2011. “Gender Diferences in Attitudes towards Credit, Terms of Trade, and the Household Balance Sheet, " PhD dissertation, American University. Rubin, Donald B. 1987. Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons. 22 Ryan, Andrea, Gunnar Trumbull, and Peter Tufano. 2011. “A Brief Postwar History of U.S. Consumer Finance.” Business History Review 85(3): 461-498. Schooley, Diane K., and Debra Drecnik Worden. 2010. “Fueling the Credit Crisis: Who Uses Consumer Credit and What Drives Debt Burden.” Business Economics 45(4): 266-276. Shaefer, H. Luke, Xiaoqing Song, and Trina R. Williams Shanks. 2013. “Do Single Mothers in the United States Use the Earned Income Tax Credit to Reduce Unsecured Debt?" Review of Economics of the Household 11(4): 659-680. Starr, Martha A. 2014. “Gender, Added-Worker Effects, and the 2007–2009 Recession: Looking within the Household.” Review of Economics of the Household 12(2): 209-235. StataCorp. 2009. Stata Multipleimputation Reference Manual Release 11. College Station, TX: Stata Press. Thorne, Deborah. 2010. “Extreme Financial Strain: Emergent Chores, Gender Inequality and Emotional Distress.” Journal of Family and Economic Issues 31(2): 185-197. Wolff, Edward Nathan. 2010. “Recent Trends in Household Wealth, 1983-2009: The Irresistible Rise of Household Debt.” Review of Economics and Institutions 2(1): Article 4. Available: http://www.rei.unipg.it/rei/article/view/26/31 (last accessed August 25, 2014). Zagorsky, Jay L. 2003. “Husbands’ and Wives’ View of the Family Finances.” The Journal of Socio-Economics 32(2): 127-146. 23 Table 1. Gender Differences in Attitudes towards Debt Car Educa-tion OK to Borrow for Living Exp. if Income Cut Luxury Vacation Never-married Respondents (n=1,431) Never-married Female 2010 2010 * Never-married Female Pseudo R2 0.0008 0.0297 ** 0.0008 -0.0255 ** 0.0193 0.0149 0.0232 0.0111 -0.0668 *** -0.0023 -0.0794 *** -0.0219 ** 0.0170 0.0125 0.0204 0.0101 0.0899 *** 0.0205 0.0189 0.0466 0.0165 ** 0.0364 -0.0086 -0.0117 -0.0614 0.0229 0.0180 0.0273 0.0138 0.0221 ** 0.0438 0.0778 0.0664 0.0483 0.0360 0.0189 0.0344 ** 0.0552 *** 0.0130 * 0.0261 ** All Respondents (n=10,910) Never-married Female 2010 2010 * Never-married Female Pseudo R2 0.0132 0.0142 0.0179 0.0075 0.0123 -0.0166 *** -0.0122 *** -0.0208 *** -0.0022 -0.0031 0.0046 0.0042 0.0061 0.0028 0.0045 -0.0115 -0.0117 0.0202 -0.0284 *** -0.0124 0.0153 0.0163 0.0212 0.0094 0.0145 0.0573 0.0539 0.0412 0.0302 0.0170 Notes: Reported standard errors (in italics) are linearized. Level of significance, p <.10 - *, p<.05 - **, p <.01 - *** Adjusted R2 is averaged over five imputations. 24 Table 2. Gender Differences in Household Financial Ratios Debt/ Assets Debt/ Income Monthly Debt/ Income Never-married Respondents (n=1,431) Never-married Female 2010 2010 * Never-married Female Moderate Credit Attitudes Liberal Credit Attitudes 2 R 0.0045 *** 0.4898 *** 0.0139 0.0118 0.1610 0.0140 0.0027 -0.1143 -0.0025 0.0125 0.1007 0.0092 -0.0188 -0.8932 0.0153 0.2140 0.0179 0.0014 -0.0902 -0.0142 0.0108 0.1367 0.0135 -0.0146 -0.0937 -0.0024 0.0130 0.1801 0.0164 0.8031 0.1721 0.1942 -0.0311 -0.0571 0.0084 -0.1895 0.0148 0.0037 -0.1919 0.0024 0.0360 -0.0094 -0.8030 0.0109 0.1932 *** -0.0336 * All Respondents (n=10,910) Never-married Female 2010 2010 * Never-married Female Moderate Credit Attitudes 0.0194 0.0163 ** *** 0.0027 Liberal Credit Attitudes 2 R 0.0221 -0.0986 *** 0.3113 *** -0.0201 0.0149 *** -0.0046 0.0036 * 0.0358 0.0046 0.1179 0.0118 0.7061 0.0414 0.1909 Notes: Reported standard errors (in italics) are linearized. Level of significance, p <.10 - *, p<.05 - **, p <.01 - *** Adjusted R2 is averaged over five imputations. 25 *** 0.0037 0.0378 *** -0.0257 *** *** Table 3. Gender Differences in Financial Ratios by Type of Debt Car Loans Debt/ Assets Never-married Respondents (n=1,431) Never-married Female -0.0107 *** 0.0032 2010 -0.0037 0.0026 2010 * Never-married Female 0.0169 *** 0.0037 OK to borrow for car? 0.0056 *** 0.0021 OK to borrow for education? Debt/ Income 0.0168 ** 0.1442 0.0098 ** 0.0043 -0.0093 0.0084 -0.0090 0.0055 Student Loans Monthly Debt/ Income 0.0109 *** 0.0030 -0.0003 0.0015 -0.0062 0.0040 -0.0099 *** 0.0031 Debt/ Assets Monthly Debt/ Income Debt/ Income 0.0197 *** 0.1977 *** 0.0063 0.0692 0.0073 ** -0.0023 0.0037 0.0188 -0.0168 * -0.2626 *** 0.0090 0.0964 -0.0009 0.0011 0.0013 0.0009 -0.0021 * 0.0018 -0.0008 0.0031 -0.0001 0.0006 0.0131 0.0231 OK to borrow for living exp.? OK to borrow for luxuries? OK to borrow for vacation? R2 All Respondents (n=10,910) Never-married Female 2010 2010 * Never-married Female OK to borrow for car? 0.5182 -0.0101 *** 0.0023 0.0019 *** 0.0004 0.0127 *** 0.0034 0.0006 0.0005 0.6531 0.6590 -0.0340 *** -0.0014 0.0084 0.0035 -0.0013 -0.0007 0.0020 0.0007 -0.0026 -0.0052 0.0092 0.0038 0.0025 -0.0003 0.0022 0.0008 OK to borrow for education? 0.0009 0.1299 0.4865 0.0010 0.1466 * 0.0065 0.0768 0.0060 *** 0.0004 0.0016 0.0075 -0.0142 * -0.2139 ** 0.0078 0.0826 -0.0009 0.0008 0.0006 *** 0.0002 -0.0011 0.0009 0.0097 *** 0.0017 0.0002 * 0.0001 0.0341 *** 0.0066 OK to borrow for living exp.? OK to borrow for luxuries? OK to borrow for vacation? 2 R 0.4926 0.4412 0.4229 0.0884 0.0691 0.3561 Notes: Reported standard errors (in italics) are linearized. Level of significance, p <.10 - *, p<.05 - **, p <.01 - *** Adjusted R2 is averaged over five imputations. See text for control variables. 26 Debt/ Assets Credit Card Debt Monthly Debt/ Debt/ Income Income -0.0008 0.0046 -0.0118 ** 0.0057 0.0056 0.0048 0.0295 *** 0.0085 *** 0.0112 0.0032 -0.0114 -0.0028 0.0079 0.0023 -0.0345 ** -0.0098 ** 0.0158 0.0045 0.0069 ** 0.0027 0.0012 0.0024 0.0024 0.0028 0.0314 *** 0.0091 *** 0.0084 0.0024 -0.0101 -0.0027 0.0080 0.0024 0.0200 *** 0.0059 *** 0.0073 0.0021 0.0928 0.0143 0.0151 -0.0012 -0.0235 ** -0.0069 ** 0.0026 0.0117 0.0033 0.0019 *** 0.0233 *** -0.0060 *** 0.0006 0.0023 0.0007 -0.0021 -0.0203 -0.0059 * 0.0031 0.0124 0.0036 0.0048 *** 0.0223 *** 0.0005 0.0024 0.0031 *** 0.0088 0.0014 0.0061 0.0055 *** 0.0401 *** 0.0009 0.0042 0.0066 *** 0.0007 0.0026 0.0018 0.0116 *** 0.0012 0.0864 0.0786 0.0785 Figure 1. Attitudes towards Debt in 2007 and 2010 by Respondent Gender and Marital Status OK to Borrow for Education OK to Borrow for Car 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Never-married Women Never-married Men Married Households Never-married Women All Respondents Never-married Men Married Households All Respondents OK to Borrow for Living Expenses 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Never-married Women Never-married Men Married Households All Respondents OK to Borrow for Vacation OK to Borrow for Luxury 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Never-married Women Never-married Men Married Households All Respondents 2007 Never-married Women 2010 27 Never-married Men Married Households All Respondents Figure 2. Household Financial Ratios in 2007 and 2010 Debt/Assets 40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 Never-married Women Never-married Men Married All Households Debt/Income 180% 160% 140% 120% 100% 80% 60% 40% 20% 0% Never-married Women Never-married Men Married All Households Monthly Debt/Monthly Income 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Never-married Women Never-married Men 2007 Married 2010 28 All Households Appendix A. Descriptive Statistics of Control Variables Dichotomous Variables 2007 % 2010 C.I. n % C.I. n Race of respondent - White Never-married Women 52% 49% 55% 248 46% 44% 48% 497 Never-married Men 68% 66% 71% 238 70% 68% 72% 448 Married 83% 82% 83% 2984 78% 77% 78% 4072 All 80% 79% 80% 4418 73% 73% 74% 6492 Race of respondent - Black/African American Never-married Women 36% 33% 39% 248 35% 33% 37% 497 Never-married Men 16% 14% 19% 238 15% 13% 16% 448 Married 6% 5% 6% 2984 8% 7% 8% 4072 All 9% 9% 10% 4418 12% 12% 13% 6492 Never-married Women 8% 7% 10% 248 15% 14% 17% 497 Never-married Men 8% 7% 10% 238 10% 8% 11% 448 Married 7% 7% 8% 2984 10% 9% 10% 4072 All 7% 7% 7% 4418 10% 10% 10% 6492 Never-married Women 4% 3% 5% 248 4% 3% 5% 497 Never-married Men 7% 5% 8% 238 6% 5% 7% 448 Married 4% 4% 5% 2984 5% 5% 5% 4072 All 4% 4% 4% 4418 5% 4% 5% 6492 Never-married Women 2% 1% 2% 248 4% 3% 4% 497 Never-married Men 5% 4% 7% 238 8% 7% 9% 448 Married 6% 6% 6% 2984 5% 5% 5% 4072 All 5% 5% 5% 4418 5% 4% 5% 6492 Race of respondent - Hispanic Race of respondent - Asian/Other Take substaintial financial risks Household had any debt payments more than 60 days past due in last year Never-married Women 7% 6% 8% 248 10% 9% 11% 497 Never-married Men 5% 4% 6% 238 6% 5% 7% 448 Married 3% 3% 4% 2984 7% 6% 7% 4072 29 All 4% 4% 4% 4418 7% 7% 10% 6492 Never-married Women 6% 5% 8% 248 7% 6% 9% 497 Never-married Men 8% 7% 10% 238 5% 4% 6% 448 Married 8% 8% 9% 2984 11% 10% 11% 4072 10% 9% 10% 4418 11% 11% 12% 6492 Ever bankrupt All Been turned down for credit in last 5 yrs Never-married Women 25% 22% 27% 248 25% 24% 27% 497 Never-married Men 17% 15% 19% 238 21% 19% 23% 448 Married 15% 14% 15% 2984 20% 19% 20% 4072 All 16% 15% 16% 4418 20% 20% 20% 6492 Continuous Variables 2007 Mean 2010 C.I. n Mean C.I. n Age of head of household Never-married Women 37.51 36.66 38.37 248 37.72 37.12 38.31 497 Never-married Men 38.72 37.87 39.57 238 37.87 37.25 38.48 448 Married 48.54 48.29 48.79 2984 49.84 49.63 50.06 4072 All 50.02 49.79 50.24 4418 50.53 50.34 50.71 6492 Never-married Women 33,245 31,326 35,164 248 32,335 30,504 34,165 497 Never-married Men 47,447 36,222 58,673 238 44,310 37,677 50,944 448 Married 119,185 111,710 126,659 2984 105,683 101,089 110,277 4072 All 88,162 83,134 93,191 4418 78,332 75,258 81,407 6492 Total Income 30 Years of Education Never-married Women 13.58 13.44 13.72 248 13.66 13.57 13.76 497 Never-married Men 13.81 13.66 13.96 238 14.00 13.90 14.10 448 Married 13.38 13.33 13.42 2984 13.49 13.45 13.53 4072 All 13.26 13.22 13.30 4418 13.41 13.39 13.44 6492 Never-married Women 0.73 0.67 0.78 248 0.71 0.66 0.75 497 Never-married Men 0.11 0.08 0.14 238 0.06 0.05 0.07 448 Married 1.06 1.04 1.08 2984 1.09 1.07 1.11 4072 All 0.83 0.82 0.85 4418 0.84 0.83 0.85 6492 Total Number of Children N is number of direct observations. Binomial confidence intervals are at 95%. Confidence intervals are at 95%. Married households includes both legally married and co-habitating households. 31