Recession, Man-cession, or Mom-cession? Gender Inequality in Reemployment Outcomes

Transcription

1 1 Recession, Man-cession, or Mom-cession? Gender Inequality in Reemployment Outcomes Disaggregated by Marital and Parental Status ABSTRACT During the Great Recession the public generally focused on the situation of displaced male workers, which is evident in the often-used terms of he-cession and mancession. We address the relative silence that has surrounded the economic fates of displaced female workers during the recession by examining two reemployment outcomes: weeks of unemployment and earnings upon reemployment. Instead of testing for only direct gender effects, we use four waves of the CPS Displaced Workers Supplement to advance a disaggregated approach that distinguishes female and male workers by marital and parental statuses. Though conventional evidence suggests that women experienced longer unemployment durations up until the most recent recession, our results demonstrate that the aggregated approach masks important within-gender inequalities. We find that mothers and wives, but not always single women, are consistently disadvantaged in reemployment outcomes even during the recent recession. Fathers and husbands, however, experience far better outcomes. The recent Great Recession of is widely considered to be the most trying period for American workers since the Great Depression of the 1930s. In the course of two years, millions of workers lost their jobs, leaving one out of ten unemployed during the recession s peak. Unemployment was hardly a mild setback for workers displaced by plant closures and layoffs; 58 percent of displaced workers reported that they were still unable to find reemployment after three years a ten-year record high (Authors 2011). It is still unclear how

2 2 this dramatic shift in workers economic wellbeing will play out in the future, but all signs suggest that the recession s aftermath and the effects of chronic unemployment will be felt for many years to come (Grusky, Western, and Wimer 2011). As the recession ran its course, it became clear that the effects of the economic downturn were not evenly distributed across groups of workers, and this was particularly true for women and men. Alarmingly, men s employment rate dropped by 7.2 percentage points compared to a troubling, but less dramatic, 3.2-percentage-point decrease for women. For men, unemployment outflows halted, causing the male unemployment rate to increase to a high of 11 percent, while women s unemployment remained at 8.3 percent (Sahin, Song, and Hobijn 2010; Sum, Trubskyy, and McLaughlin 2011). This gender gap in unemployment was the largest reported in the postwar era (Sahin et al. 2010), leading one commentator to speculate that the image of the male breadwinner had become all but obsolete (Rosin 2010). The recession s impact on men s employment prospects was in no small part a result of their overrepresentation in the financial and goods producing sectors the industries most adversely affected by the recession (Sum et al. 2011). The media, quick to reduce the troubling trend into appealing sound bites, began to refer to the economic downturn as the he-cession or man-cession, portraying men as victims fathers and husbands who were unable to fulfill the traditional provider role. Meanwhile, the public paid less attention to the recession s effects on working women (Williams and Tait 2011). Why the silence concerning women s economic outcomes? Some argued that the public s fixation on men s economic losses resulted from a stubbornly persistent gender ideology that still views masculinity to be synonymous with the provider role (Hertz 2010; Williams and Tait 2011), despite the economic reality that wives contribute approximately 47 percent of household earnings (Mattingly and Smith 2011; Smith 2009). Additionally, there is little evidence that

3 3 unemployed women, as a whole, took longer than men in finding reemployment (Farber 2011). In fact, the Great Recession ended a several decade long trend of gender inequalities in postdisplacement unemployment durations that historically favored men (Farber 2005). In this paper, we address the relative silence that has surrounded the economic fates of displaced female workers during the recent recession by providing a more nuanced examination of women and men s reemployment outcomes. When researchers look solely at aggregate gender differences in post-displacement employment outcomes, they miss important ways that gender interacts with other life-course characteristics to generate unobserved inequalities. Earnings and employment research indicates that family statuses can influence people s reemployment prospects, but the direction of the effect depends on the gender of the respondent (Budig and England 2001; Hodges and Budig 2010). Instead of simply testing for direct gender effects, we advance a disaggregated approach that allows us to further distinguish female and male workers by marital and parental statuses. To accomplish this, we use four waves of the Displaced Workers Supplement (DWS) a nationally representative sample of displaced workers that is a recurring module of the U.S. Current Population Survey (CPS). The four waves cover major economic periods of the last decade, allowing us to examine how the gendered effects of marital and parental status are shaped by the strength of the greater economy in which they are embedded. We begin this paper by reviewing supply and demand side explanations for gender disparities in the labor market. We then integrate these perspectives with queuing theory and outline our central hypotheses. After describing our data and methods, we present findings that illustrate continuing gaps in reemployment outcomes by gender, marital, and parental statuses. Though prior evidence suggests that women have experienced longer unemployment spells up

4 4 until the most recent recession, our results demonstrate that the aggregated approach masks important within-gender inequalities. Mothers and wives, but not single childless women, have consistently been disadvantaged in reemployment outcomes over the last decade, but fathers and husbands experienced the shortest post-displacement unemployment durations. REEMPLOYMENT OUTCOMES FOR DISPLACED WORKERS Our research centers on a specific subset of the working-age population: unemployed displaced workers. Whereas other forms of job loss can vary due to individual circumstances and employer preferences, displacement due to factory closings, mass layoffs, or insufficient work systematically affects large groups of workers. The analysis of post-displacement unemployment, then, provides a useful test of labor market efficiencies and discrimination, since unobservable individual characteristics associated with other types of job separation are largely accounted for (Madden 1987). We therefore focus on this group of workers in our analysis, though persistent gender inequalities exist through displaced women s higher likelihood of dropping out of the labor force (Farber 2011). Researchers have repeatedly demonstrated that reemployment outcomes, such as unemployment duration and later earnings, vary by displaced workers demographic characteristics. Generally, more disadvantaged workers, in particular female, minority, lesseducated, and older workers, spend more time unemployed and face additional earnings losses upon reemployment when compared to members of more advantaged groups (Couch, Jolly, and Placzek 2009; Crossley, Jones, and Kuhn 1994; Authors 2011). Surprisingly, the media and many social scientists have focused on men s unemployment and earnings losses in the Great Recession, often overlooking the plight of female displaced workers. In our research we

5 5 investigate whether the recent recession truly presents a switch in employment gaps by gender for displaced workers. We address the following research questions: Do male or female displaced workers fare worse in terms of unemployment duration and earnings losses? How do these gender differences vary in relation to displaced workers parental and martial statuses? And, how have these relationships changed over the past decade? Gender Disparities in Reemployment Outcomes Gender inequalities in the labor market are well documented. Despite much progress over the past century, a gender gap in earnings remains and occupational gender segregation persists (BLS 2011). Although occupational segregation often creates earnings gaps between women and men, gender inequality in the labor market also stems from worker characteristics and preferences; intentional, subtle, and statistical discrimination by employers; organizational processes; policy choices; and ideology surrounding gender differences (Blau and Kahn 2006; Davis and Greenstein 2009; Kaufman 2010; Leicht 2008; Reskin 2001). Despite this large literature on gender discrimination in the labor market, few researchers have investigated variation in employment outcomes by gender and familial status until recently. This absence is particularly conspicuous in the research on reemployment outcomes for displaced workers a population of workers united by the experience of involuntary job loss and, presumably, the preference to seek reemployment. In this section, we first address supply-side explanations for post-displacement inequalities by gender and parental status that emphasize how workers employment and family preferences affect the length of unemployment and reemployment earnings. We then turn to demand-side explanations that focus on the employer s role in generating and perpetuating post-

6 6 displacement inequalities through hiring discrimination. To explain how the ranking and sorting of displaced workers plays out at the aggregate, we discuss how labor queues result from employee and employer preferences that structure labor market inequalities at the national level. Supply-Side Explanations of Post-Displacement Inequalities: Workers Gendered Preferences Among economists, human capital theory dominates approaches used to explain labor market inequalities, including those experienced by displaced workers. Drawing from Becker s (1964) seminal work on the topic, proponents of human capital theory assume that workers defer present earnings in order to develop (or invest in) their education, skills, and training. This human capital provides workers with a competitive advantage in the labor market because high levels of human capital enhance workers productivity and therefore their attractiveness to future employers. Economists typically differentiate between two forms of human capital: general skills, such as investments in higher education, that transfer across employers, and firm-specific skills learned on the job that do not transfer to other employment contexts. Upon involuntary job loss, firm-specific capital, usually measured as previous job tenure, is negatively associated with labor market outcomes, while general human capital investments in education improve postdisplacement chances (Addison and Portugal 1989; Arulapalam, Gregg, and Gregory 2001; Crossley, Jones and Kuhn 1994; Houle and Van Audenrode 1995). Human capital theory also predicts gender differences in reemployment outcomes. According to Becker s (1985, 1991) new home economics, women and men are rational actors who specialize in different skills with the shared goal of maximizing household productivity. Men, comparatively advantaged in the labor market, increase their earnings capacity by accruing human capital across the life course through investments in education, and importantly,

7 7 continuous employment. Women, on the other hand, must anticipate sizable job interruptions due to childbirth and childrearing. Ostensibly, women allocate a considerable sum of energy to childrelated activities leading them to seek employment that offers fewer monetary rewards in exchange for flexibility, fewer hours, and less stress. Becker (1985, 1991) explained this tradeoff using the language of compensating differentials. Women prefer to avoid high commitment jobs that typically reward employees with more hours and higher pay in favor of lower-paying or part-time jobs that better prepare them to balance work and family demands. Taken together, some argue that these non-discriminatory factors (i.e., preferences) explain a large portion of the gender gap in wages over time (O Neill 2003). According to this theory, mothers earnings capacity and labor market competitiveness should be especially low. First, compared to childless women and men, mothers bring the least amount of expendable energy to paid work, as the presence of children considerably increases time spent on household labor (Bianchi et al. 2000). Second, while women in general may anticipate child-related job interruptions, mothers live through these human capital losses with each additional child. Indeed, about one third of the motherhood penalty in wages can be attributed to job experience differences between mothers and non-mothers (Budig and England 2001). The motherhood penalty is especially steep for highly educated mothers who tend to occupy high-skilled jobs where skill depreciation is especially costly (Anderson, Binder, and Krause 2003). When the population of interest is restricted to displaced workers, however, the logic of human capital theory suggests an alternative pattern of gender inequality: differential skill investments may lead to less severe economic penalties for displaced women compared to men. Because women are assumed to anticipate more job changes over the life course, Maxwell and

8 8 D Amico (1986) speculated that women were more likely to invest in general capital (education) compared to firm-specific capital. Therefore, one might expect women s unemployment durations to be shorter and earnings losses fewer because, unlike firm-specific capital, general skills are more valuable to future employers. Despite this prediction, economists have yet to find consistent evidence to substantiate this claim. Most tests of this hypothesis have indicated that displaced women fare considerably worse in the labor market compared to their displaced male counterparts (Madden 1987; Maxwell 1989; Maxwell and D Amico 1986). Women s poorer reemployment outcomes could reflect their choice to, at least temporarily, re-focus their time on motherhood and other family-related obligations. Differential engagement in paid work and housework stems from more than pure rational calculations it acts as a powerful way in which women and men do gender to reinforce gender boundaries and identities (Brines 1994; West and Zimmerman 1987). Mothers, in particular, may feel pressured to conform to gendered expectations about femininity and domestic work, as women s conformity to intensive mothering remains a stubbornly persistent trend in the U.S. context, regardless of women s employment preferences (Hays 1996). According to these perspectives, which we refer to as the gender role socialization argument, displaced women with spousal and child obligations may conform to pressures and embrace the family-sphere, delaying their search for reemployment or dropping out of the labor force all together. This strategy may be less costly for unemployed women with working husbands, a group who make up almost a quarter of all married households with an unemployed spouse (BLS 2012). The potential for displaced women to gravitate toward family responsibilities may also partially explain why more women than men chose to leave the labor force during the recent economic downturn (Farber 2011).

9 9 Recent research, however, casts doubt on the argument that family obligations and childcare commitments weaken women s labor force attachment. To begin with, the length of time that women spend out of the labor force to engage in parenting has decreased dramatically since the 1970s (Boushey 2008), especially for women in managerial and professional occupations (Percheski 2008). Approximately 70 percent of mothers with children under the age of 18 actively participate in the labor force, and the percentage of dual earner households actually increases when children are present, likely due to child-rearing expenses (BLS 2012). Although little research has examined how involuntary unemployment affects the gendered work-family preferences of displaced workers, at least one study of married and cohabiting couples suggests that women and men respond rationally when a spouse experiences involuntary job loss. Using couple-level longitudinal data from the Panel Study of Income Dynamics (PSID), Gough and Killewald (2011) found that, regardless of gender, the unemployed spouse increased his or her relative share of household labor. Gender was not entirely absent from couples decisions to reallocate housework; unemployed wives weekly housework contribution still doubled that of their husbands. Nonetheless, although small gender differences persisted, both husbands and wives with unemployed spouses did not appear to permanently reevaluate workfamily roles; instead, spouses responded rationally and reallocated their labor in a (mostly) equitable manner (Gough and Killewald 2011). In sum, supply-side explanations suggest that workers gendered and family-related preferences will drive any inequalities in displaced women and men s subsequent employment outcomes. Human capital theory suggests that, although women and mothers generally prefer to invest in the family-sphere, the net effect of job displacement will be less costly for women and mothers because women, anticipating job interruptions, invest in transferable general skills.

10 10 However, human capital theory ignores the fact that displaced women and men may base workfamily decisions on gender expectations, rather than on the logic of the market. Displaced married women s and mothers decision to conform to gender expectations may result in reemployment inequalities by gender and family status, as these women will extend (or avoid) their job search and possibly choose less demanding but low-paying jobs. Demand-side Perspectives: Employers Preferences and Discrimination In virtually all studies of gender and family inequality in hiring and pay, gaps remain even when accounting for human capital measures and job characteristics (e.g., Budig and England 2001). Further, low-paying, predominantly female jobs rarely provide the flexibility or stress free environment necessary for women to better balance competing work and family obligations (Glass and Camarigg 1992). More recently, Kmec (2011) found that working mothers did not differ from non-parents on multiple measures of pro-work orientations and job demands, and mothers were actually more likely than non-parents to report higher levels of job engagement and work intensity. Taken together, these studies cast doubt on the core assumptions of human capital theory s explanations for gender and family status labor market inequalities, particularly the notion that women and mothers compensate for low-paying employment by taking flexible and low intensity jobs. Thus, researchers generally acknowledge that institutional factors most commonly discrimination due to employers preferences play a significant part in patterning women and men s labor market outcomes. These demand-side explanations generally incorporate employers explicit and implicit preferences for workers (Kaufman 2010). In hiring or assigning people to jobs, employers often rely on characteristics that are not direct measures of ability to predict the performance of future

11 11 workers (Holzer 1996; Moss and Tilly 2001). This can lead to gender discrimination in hiring when preferences stem from certain gender ideologies. For example, employers might assume that women are less attached to the labor force than men, making them less productive and less desirable workers. Employers unconscious or implicit biases can also lead to in-group favoritism, which is not necessarily negative, but can still foster discrimination (Greenwald and Krieger 2006; Tajfel 1982). In the case of gender, the net effect is discriminatory for the women because they are essentially left out and overlooked. A demand-side perspective also suggests ways that marital or parental statuses might result in labor market advantages and disadvantages. Researchers have consistently demonstrated a motherhood penalty in wages that cannot be accounted for by worker and job characteristics typically associated with pay (Budig and England 2001; Budig and Hodges 2010). In contrast, parental status positively affects men s opportunities for rehire through a daddy bonus in the form of higher earnings and preferential hiring (Correll, Benard, and Paik 2007; Glauber 2008; Hodges and Budig 2010). Recent experimental studies have illustrated the potential for hiring discrimination based on gender and parental status. When undergraduates evaluated and ranked resumes of same-gendered job applicants that differed by only parental status, they were less likely to recommend mothers for hiring and job promotions, deemed mothers less competent and committed, and offered them lower starting salaries than equally qualified women without children (Correll et al. 2007; Fuegan et al. 2004). Confirming their findings in an audit study of 600 employers, married mothers were half as likely to receive a call back compared to nonmothers, whereas parental status did not affect men s chances (Correll et al. 2007). Correll and colleagues (2007) concluded that these results provided compelling evidence that motherhood acts as a salient status characteristic that interacts with cultural beliefs about mothers ability and

12 12 willingness to be productive workers, which results in discrimination. They also speculated that because current gender ideology identifies men as breadwinners, men were seen as more deserving and perhaps more responsible than their childless counterparts (Correll et al. 2007). In addition, employers might judge mothers more harshly or hold them to a higher standard insofar as they believe that motherhood represents a personal choice to invest in the family instead of paid work. To examine this hypothesis, Kricheli-Katz (2011) tested for motherhood penalties in hiring and wages in contexts that varied in the degree that motherhood was viewed of as a choice. In this laboratory experiment, undergraduates were more likely to engage in hiring and pay discrimination against mothers compared to childless women when researchers primed them with literature that emphasized the importance of choice in determining individuals economic and social outcomes. Interestingly, when researchers primed undergraduates with literature emphasizing disadvantage as the result of uncontrollable constraints, they gave preferential treatment to mothers instead of non-mothers. In a second analysis, Kricheli-Katz (2011) analyzed CPS data from 1988 through 2004 and found larger motherhood penalties in wages in states where motherhood was most viewed as a choice; that is, in states with high proportions of childless women and progressive pro-choice abortion policies. In addition to parental status, marital status can widen inequalities between women and men. Employers may assume that married women will be less productive, committed, and competent, as the transition into marriage may signal the gendered decision to embrace the family sphere, and perhaps an intention to have children. Married men, like fathers, may experience a wage boost and preferential treatment in the hiring process because, according to the gendered expectations associated with male breadwinner status, their marital status may signal responsibility and deservingness. Researchers have documented that men receive a wage

13 13 premium that cannot entirely be explained by human capital and other relevant intra-household behaviors (e.g., Rodgers and Stratton 2010), though the effect seems stronger when young dependents are present (Hersch and Stratton 2000). Perhaps more telling, laboratory experiments similar to the ones described above provide evidence that evaluators view married women as less deserving and suitable for employment, while married men enjoy preferential treatment (Jordan and Zitek 2012). Taken together, these studies suggest that displaced mothers and married women will spend more time unemployed and experience larger earnings losses upon reemployment than married women with and without children. But despite consistent evidence of labor market discrimination against women with family obligations, the few studies of the effects of gender and family status on post-displacement employment outcomes were mostly unable to find support for motherhood and marriage penalties. For example, Mazerolle and Singh (2004) followed displaced former employees after an Ontario factory closing and found that, although women did experience longer unemployment durations, having a dependent actually increased the likelihood of reemployment. Presumably, the increased costs associated with child raising pressured parents back into the labor market (a finding that is at odds with the argument that displaced mothers might choose to temporarily embrace intensive mothering). In Koeber and Wright s (2006) analysis of the 2000 DWS women were also less likely than men to be reemployed at the time of the survey. Parenthood did not affect women and men s reemployment prospects, but marriage improved the likelihood of reemployment for men at the time of the survey. Among women, unmarried women displaced from full-time or high-skilled jobs were significantly more likely to be rehired than their married counterparts. Apparently, employers

14 14 preferred to hire career-oriented women workers when they were less attached to the familysphere through marriage commitments. These studies suggest that motherhood does not affect reemployment outcomes, but the findings are limited because they are based on either single-year data or convenience samples. Despite not focusing on displaced workers, the laboratory experiments described above provide the most compelling evidence of disparate labor market outcomes for parents, whether due to employers stereotypes about productivity or the view that motherhood disadvantages are the result of women s individual choices. And we are aware of no evidence to suggest that displaced married women and mothers differ substantially in the job search process from women who left their job due to other circumstances. Motherhood and marriage appear to be salient status characteristics that systematically disadvantage women in the labor market. Given the reality that employers maintain a preference hierarchy by gender and family status and the nature of our data, we require a theory that can explain this unequal ranking among displaced workers in the aggregate. Queuing theory a perspective that views the sorting of workers as a structural feature of labor markets provides a useful way of understanding how supply and demand-side explanations of inequality by gender and parental status play out at the national level. Toward a Structural Explanation: Job Queues, Gender Queues, and Family Queues Queuing theories that incorporate all types of employer preferences, along with the preferences of potential employees, help to explain differential reemployment outcomes for women and men, as well as other groups. In short, queuing theory posits that employers, lacking perfect information on whom to hire, rank and sort potential employees by observable characteristics, including race, gender, and indicators of human capital. Workers whose visible

15 15 traits signal unproductivity, either resulting from human capital deficits or membership to a group that is believed to be less productive, occupy positions at the tail end of the hiring queue. Workers whose traits signal the best fit in many cases, highly educated white men enjoy a privileged position in the hierarchy of applicants. Thus, hiring queues are a structural feature of the labor market and perpetuate inequalities (or labor market inefficiencies) based on employer s preferences for productive workers preferences that are susceptible to stereotypes and misinformation (Thurow 1975). Workers also generate job queues by sorting potential jobs into a hierarchy based on their preferences (Reskin and Roos 1990). For example, if women as a group prefer jobs with schedule flexibility and shorter hours, jobs with these characteristics will enjoy a privileged position in women s job queues. Thus, job queues interact with employer-generated hiring queues to shape workers labor market destinies, though employers preferences play a larger role than workers preferences in generating inequalities (Kaufman 2010; Reskin and Roos 1990). Researchers have used queuing theory to explain the persistence of gender segregation across occupations (McTague, Stainback, and Tomaskovic-Devey 2009; Reskin and Roos 1990), the feminization of the U.S. banking industry (Rich 1995), the growth of sales occupations (Bellas and Coventry 2001), the earnings gap among male and female law school graduates (Noonan, Corcoran, and Courant 2005), and inequalities in displaced workers reemployment outcomes (Koeber and Wright 2006; Spalter-Roth and Deitch 1999). In their analysis of the 1996 DWS, Spalter-Roth and Deitch (1999) found that, net of human capital and job characteristic measures, displaced white men were viewed as most desirable, as indicated by their shortened unemployment durations and future job prospects, followed by white women, women of color,

16 16 and finally black men (see also Authors 2011). Similarly, Koeber and Wright (2006) used the 2000 DWS and argued that displaced women and men were sorted into gender queues; displaced women were less likely than men to find reemployment, holding constant other key worker characteristics. In both studies, employers preferences provided whites and men with a competitive advantage over similarly situated women and nonwhites. As we have noted, however, gender disparities in post-displacement outcomes persisted through the 21 st century and then diminished during the Great Recession, suggesting that displaced women and men have shared a similar fate in recent years (Farber 2005, 2011). This may be because, unlike men who were overrepresented in the hard hit manufacturing, construction, and financial industries, women better weathered the storm, given their overrepresentation in the relatively stable service sector. We therefore incorporate workers previous occupations into our models to account for this difference. Because gendered employer and employee preferences can still affect reemployment outcomes regardless of occupation, we expect that unemployment durations will be longer and earnings will be lower for displaced women relative to men in every period examined, except during the recent recession (Hypothesis 1). However, we believe that this hypothesis will only hold true in the aggregate. Just as discrimination based on employers preferences results in gender queuing, employers can generate hiring queues based on family statuses. As previously discussed, stereotypes and beliefs about working mothers operate to disadvantage mothers in multiple important job-related outcomes, such hiring and wage setting. In the language of queuing theory, displaced mothers occupy a position in the hiring queue behind childless women who enjoy preferential treatment by employers. Not only will this affect the length of unemployment, but mothers may have to settle for jobs with less competitive hiring queues jobs that will invariably

17 17 offer lower starting wages. Explicitly, we expect that motherhood will increase displaced women s unemployment duration and decrease later earnings relative to women without children (Hypothesis 2). Fathers, however, occupy a favorable position in the labor market relative to their childless counterparts. The gendered ideology that equates fatherhood with the breadwinner or provider role may lead employers to view fathers as more hardworking and deserving compared to childless men (Correll et al. 2007; Glauber 2008; Hodges and Budig 2010). This preferential treatment, combined with pressures on men to provide for their families, places fathers at the front of the hiring queue, allowing them to secure the most attractive jobs with relative ease. We therefore expect that fatherhood will decrease displaced men s unemployment duration and increase later earnings relative to men without children (Hypothesis 3). As noted above, an emerging literature suggests that the motherhood penalty and fatherhood bonus patterns may extend to marriage more generally. Marriage may exacerbate gendered preferences for employment, and employers may judge married men and women differently. Employers may view married women, like mothers, as less committed and competent, as marriage signals increased family demands and potential child-related obligations. Employers may prefer married men, however, because they hold the gendered belief that married men are more hardworking, reliable, and deserving of job opportunities, as they are expected to act as the primary source of household income. Therefore, we predict that marriage will increase displaced women s unemployment durations and decrease later earnings relative to unmarried women (Hypothesis 4), but marriage will decrease displaced men s unemployment durations and increase later earnings relative to unmarried men (Hypothesis 5).

18 18 DATA AND MEASURES We test these five hypotheses using data from the January 2012, 2010, 2008, and 2006 Displaced Worker Supplement (DWS) to the Current Population Survey (CPS), a monthly labor force survey conducted by the U.S. Census Bureau. The DWS consists of a sample of workers age 20 years or older who involuntary lost or left their jobs due to a plant or company closing, insufficient work, or a shift ending in the three calendar years preceding the survey. The four waves cover displaced workers in the three years prior to 2006 and to 2008, two periods of relative labor market stability; the three years prior to 2010, a recessionary period characterized by labor market turmoil; and the three years prior to 2012, a recovery period following the recent recession. We analyzed each sample separately to compare the outcomes for unemployed men and women searching for work across varying economic contexts. 1 After removing missing data, we obtained samples of 3,575 displaced workers in 2012, 4,397 displaced workers in 2010, 2,378 displaced workers in 2008, and 2,587 workers in The smaller samples for 2008 and 2006 are representative of the better economic conditions and fewer unemployed persons in these time periods. We present descriptive statistics for our data and measures in Table 1. [Table 1] Outcome Variables We include two outcome variables in this analysis that capture the adverse effects of job displacement: the respondent s time to reemployment and logged weekly earnings upon reemployment. We measure the respondent s time to reemployment in weeks and report most

19 19 findings as hazards or risks of reemployment. We log our earnings outcome variable in order to better fit with regression model assumptions. Despite similar demographic characteristics of unemployed persons in the samples, the likelihood of reemployment was much lower and the earnings losses were larger during the recent recession (Table 1). Additionally, more men than women found reemployment in the stable and recovery period samples, but the gender gaps were smaller during the recession. Predictor Variables Our primary individual-level predictor variable describes the respondent s gender, which we indicate as a binary variable of male or female. In our analyses we also include measures for the respondent s marital status and presence of children, which we interact with gender to disaggregate some of the gender effects that we find. In order to simplify our comparison of men and women, our marital status variable indicates only whether the respondent was currently married. We also indicate the presence of children in the respondent s household as a binary variable. Across sample years, a slightly larger percentage of displaced men than women were married, but a similar percentage of both groups had children (Table 1). We include controls for the respondent s race, citizenship status, age, and educational attainment. We indicate race as two binary variables of black or non-black and Hispanic or non- Hispanic. We measure citizenship status with a categorical variable that indicates whether the respondent was a native U.S. citizen, a foreign-born naturalized U.S. citizen, or not a U.S. citizen. We measure age in years and include a squared term to account for its non-linear relationship with employment and earnings. We measure educational attainment with a categorical variable that indicates whether the respondent completed high school, attended some

20 20 college, completed college with a bachelor s degree, or went on to obtain additional education through professional or graduate school. We also include covariates related to the respondent s previous job to control for prior human capital and experience, and we incorporate measures of the conditions of displacement to account for the job loss situation that could affect a respondent s reemployment opportunities. As indicators of a person s previous job we use his or her weekly earnings, full-time/part-time status, job class, and job tenure. We log weekly earnings and measure job tenure in years. Fulltime/part-time status is a binary variable. Job class is a categorical variable that indicates whether the respondent worked at a for-profit, government, or non-profit firm. Across all samples, male displaced workers had higher earnings than female displaced workers and were more likely to be displaced from full-time jobs and for-profit firms, but the gender differences in job tenure were negligible (Table 1). We use four indicators of the conditions of the respondent s displacement: year of displacement, advance notice of displacement, reason for displacement, and receipt of unemployment insurance. We measure receipt of unemployment insurance and advanced notice of displacement as dichotomous variables. As shown in Table 1, about one-third of workers received advanced notice across sample years, and more workers received unemployment insurance during the recession. The respondent s reason for displacement is a categorical variable with three levels: insufficient work, plant or company closing, and position or shift abolished. Generally, men were more likely than women to lose employment due to insufficient work and were less likely to receive any advance notice of job loss. Occupation, State, and Temporal Fixed Effects

21 21 Occupational gender segregation is a fundamental cause of labor market inequality by gender (Charles and Grusky 2004; Padavic and Reskin 2002). Male and female displaced workers different representations in certain industries and occupations can also affect their potential for being rehired, particularly during recessionary periods when some parts of the economy lose more jobs than others. Therefore, we include fixed effects for the broad occupational classification of the respondent s last job. 3 We incorporate indicator variables for the 23 broad categories that the DWS classified occupations into our models. 4 States also vary in important ways that can affect men and women s employment prospects. Their different demographics, economic circumstances, labor market strength, public sector size, and unemployment policies can all shape a displaced worker s reemployment outcomes (Beggs 1995; Esping-Anderson 1990). We therefore include state-level fixed effects by incorporating state indicator variables into each model to account for any unobserved statelevel factors that could influence our outcomes (Allison 2009). 5 In addition, by analyzing each sample separately we account for demographic, economic, and labor market aspects that can the affect the structure of labor queues over these four time periods. METHODS We use event history methods, particularly the Cox proportional hazards model, to determine a respondent s risk of reemployment. The Cox model is a semi-parametric method for determining the relationship of a survival distribution to covariates (Box-Steffensmeier and Jones 2004). This model offers an alternative to a parametric model that does not specify an underlying distribution for the data. Therefore, the Cox model does not calculate a baseline hazard, but individual coefficients are interpreted as increasing or decreasing the risk of

22 22 reemployment. Within the Cox model, the hazard rate for the ith individual is described by Equation 1: ( ) ( ) ( ) (1) where ( ) is the baseline hazard function and represent the regression coefficients and covariates. 6 In addition to the Cox model, we use quantile (or median) regression models (QRM) to estimate a respondent s weekly logged earnings after reemployment coupled with techniques to adjust for selection into reemployment (Hao and Naiman 2007; Koenker 2005). Instead of minimizing the sum of squared residuals to estimate a sample mean, as in OLS models, quantile regression models minimize the sum of absolute squared residuals in order to estimate a sample median conditional on a set of covariates (Koenker 2005). 7 Equation 2 describes the quantile regression model: ( )( ) ( ) (2) where ( ) represent the set of covariates and regression coefficients and τ = 0.5 because were are estimating the conditional median of logged weekly earnings, y. Because they estimate a sample median, quantile regression models are more robust to outliers than OLS models. Unlike OLS models, the monotone equivariance property also holds for QRMs, which allows us to easily discuss covariate effects expressed in log-scale or absolute terms (Hao and Naiman 2007). We also account for respondents without work by imputing wages for non-employed respondents. In cases where the outcome variable of interest - weekly earnings - is observed only for a subset of the sample - reemployed displaced workers - estimates can be biased due to selection into a certain outcome. In these situations, researchers often rely solely on data for

23 23 employed persons that omit workers and can bias results, or use techniques, such as the twostage Heckman correction to correct for this selection bias (Heckman 1979). 8 Instead of omitting non-employed respondents from this part of our analysis, we follow Neal (2004) and Johnson, Kitamura, and Neal (2000) and impute wages for these respondents. We use the respondent s gender, education level, year of job loss, and previous earnings to impute wages for these nonemployed respondents based on the median wages for their employed counterparts. 9 We then weight each case by the respondent s probability of reemployment, as calculated through probit models predicting employment using the same covariates as those in our Cox PH models. By incorporating weights into the models, we allow cases where the outcome is known to influence our estimates more than those that we imputed. Our models also incorporate robust standard errors using a kernel sandwich estimator as proposed by Powell (1991). RESULTS As expected, unemployed men and women in the current recession experienced longer periods of unemployment and earned less once they found employment when compared with displaced workers who lost their jobs before and after the recession, even when controlling for individual and state characteristics. In terms of gender differences, there were smaller gaps in time to reemployment and earnings upon reemployment between men and women in 2010 than there were in other survey years. This pattern continued once we included control variables. However, after disaggregating gender by marital status and presence of children, we found significant variation in reemployment outcomes for married and single men and women, and for men and women with and without children.

24 24 Time to Reemployment by Gender Men s and women s overall probabilities of reemployment were smaller and their time to reemployment was longer in the recession than in previous years, as illustrated by the survival curves in Figure 1. [Figure 1] These survival curves plot a person s probability of continued unemployment over time measured in weeks. The solid lines represent the curves for women and the dashed lines represent the curves for men. In general, the 2012, 2008, and 2006 curves are steeper than the 2010 curve and show larger gaps between men and women. The earlier curves indicate that about 72 percent of men and 65 percent of women will become reemployed within 30 weeks of losing their jobs, and 80 percent of men and 72 percent of women will find work within two years. However, the curves for men and women converged during the recession, estimating that only 50 percent of both groups will become reemployed within 30 weeks and 58 percent within two years. Reemployment outcomes diverge again during the recovery, though, with growing gaps between men and women. The hazard coefficients presented in Table 2, where the figures are modeled from, also describe the reemployment gap for men and women. In the 2010 sample this coefficient was not statistically significant and much smaller than the coefficients in other years. [Table 2]

25 25 The addition of covariates and fixed effects to the models in Table 3 shows a similar pattern for men and women, but the decreased magnitude of the coefficients indicates that the added covariates explain at least some of the gap in reemployment time between men and women. Net of demographic and employment controls, as well as occupation and state fixed effects, women had a 13 to 17 percent lower risk of reemployment than men in the earlier samples, but the risk difference was not statistically significant in the recession and recovery samples, which partially supports our first hypothesis. 10 Additionally, the risk of reemployment for unmarried persons was lower than that of currently married persons in the 2010 and 2006 samples, but the difference was not significant in the 2012 and 2008 periods. There were no differences in outcomes by parental status, either. [Table 3] Other individual-level variables showed larger gaps between groups in their time to reemployment under conditions of a severe recession. For example, blacks had a lower risk of reemployment across most samples; however, the risk was 7 to 12 percentage points lower during the recent recession than the risk in other years. Naturalized U.S. citizens also had a lower risk of reemployment compared to native citizens in the recessionary period, but not in the other time periods. The effects of control variables related to the person s previous job varied over the four time periods as well. For instance, the risk of reemployment decreased as a person s previous income increased in all samples except for Previous job tenure also decreased the risk of reemployment in non-recession years, indicating that workers who had been with the same

26 26 employer for longer periods of time had a longer time to reemployment, a finding consistent with human capital theory. Interestingly, in the 2010 sample, part-time workers had a 20-percent higher risk of reemployment than full-time workers, but not in other sample years. Finally, covariates related to the conditions of workers displacement showed similar relationships with reemployment in most time periods in terms of the significance, direction, and magnitude. Time to Reemployment by Gender, Marital Status, and Presence of Children In order to disaggregate the gender differences in time to reemployment, we estimated a model that included interaction terms between gender and marital status and between gender and the presence of children, along with a second model that included a three-way interaction term between these variables. Table 4 presents the hazard coefficients from these regressions, which include all control variables and fixed effects from the models in Table 3. [Table 4] Although the significance levels of the covariates fluctuated across samples, this model now presents a more consistent pattern in gender differences in duration of unemployment across sample years. In the simple interaction model, the significant interaction terms demonstrate the need to consider both marital and parental status when investigating gender differences in reemployment outcomes. The three-way interaction model, also shown in this table, included fewer statistically significant terms, but did indicate a need to consider marital and parental status in modeling gender differences in time to reemployment.

27 27 In order to facilitate the interpretation of these interaction terms, Figure 2 presents the results of the Table 4 interaction models. Each estimate represents the percent change in the hazard of reemployment, net of control variables, when comparing the first group to the second. The first four estimates compare men and women of similar marital and parental statuses, the next two estimates compare men and women by their parental status, and the last two compare males and females by their marital status. [Figure 2] The first section of the figure, comparing women and men, shows that the gender gap in unemployment duration partially reflects the larger gaps between married mothers and fathers. Across all samples, the largest gaps occurred between married women and men with children. In the 2010 sample, married mothers had a 31 percent lower risk of reemployment than married fathers. In other years, this difference ranged from 29 to 45 percent. In addition, the aggregate gender findings for 2010 appear to have been driven by apparent labor market benefits for single childless women. Unmarried women without children had a 29 percent higher risk of reemployment than unmarried men without children in the 2010 sample and a 22 percent higher risk in the 2008 sample, but this relationship did not hold later on. Although the coefficients were negative, there were also no significant differences between married men and women without children and between unmarried men and women with children. The results from these interaction models support many of our other hypotheses as well. In terms of differences in unemployment duration based on parental status, we expected that motherhood would increase the duration when comparing mothers to similar non-mothers (H2),