Consequences of Partner Incarceration for Women's Employment

Abstract

Research has documented the limited opportunities men have to earn income while in prison and the barriers to securing employment and decent wages upon release. However, little research has considered the relationship between men's incarceration and the employment of the women in their lives. Economic theory suggests that family members of incarcerated individuals may attempt to smooth income fluctuation resulting from incarceration by increasing their labor supply. This study used data from the Fragile Families and Child Wellbeing Study (N = 3,780) to investigate how men's incarceration is associated with the number of hours their female partners work as well as variation in this association. Results showed that, on average, women's hours of work were not significantly impacted by the incarceration of their partners. However, there was a positive relationship between partner incarceration and employment among more advantaged groups of women (e.g., married women, white women).

The rapid expansion of incarceration in the United States (U.S.) has continued, mostly unabated, since the mid-1970s. The growing number of men who spend time in penal institutions as well as the uneven distribution of this experience across the population have lead researchers to investigate the impact of imprisonment on the life chances of former inmates and the implications for social and economic inequality (e.g., Manza & Uggen, 2006; Massoglia, 2008; Western, 2006; Wildeman & Muller, 2012). A particularly important area of this research has focused on the influence of imprisonment on labor market outcomes. Social scientists have amassed considerable evidence that incarceration not only pushes inmates out of jobs but also reduces employment prospects upon release (Freeman, 1992; Pager, 2003; Raphael, 2007; Western, 2006). Moreover, research suggests that black men pay an even higher penalty than similar white men for having a criminal record. Employer discrimination against former inmates is more pronounced for black men, and the effect of incarceration on wages and earnings is stronger (Pager, 2003, 2007; Western, 2006).

Although we know that incarceration has detrimental consequences for the employment of currently and formerly incarcerated individuals, we know little about how incarceration impacts the employment of their family members. Many incarcerated men are members of families and households that are subjected to the ramifications of their incarceration and ex-offender status (Bureau of Justice Statistics, 1997; B. E. Carlson & Cervera, 1991; Grinstead, Zack, Faigeles, Grossman, & Blea, 1999; Jorgensen, Hernandez, & Warren, 1986; Western, 2006). Indeed, several studies have documented the reduced income and substantial costs associated with incarceration (Comfort, 2008; deVuono-powell, Schweidler, Walters, & Zohrabi, 2015; Geller, Garfinkel, & Western, 2011; Grinstead, Faigeles, Bancroft, & Zack, 2001; Johnson, 2008). In response to instability, households and families often engage in activities that attempt to smooth income fluctuation and stabilize economic well-being. Economic theory has suggested that a common strategy to address income loss from one family member's reduced labor supply is for another family member to increase their labor supply (Lundberg, 1985; Mattingly & Smith, 2010; Western, Bloome, Sosnaud, & Tach, 2012). This leads to the question: does the erosion of men's labor force participation due to mass incarceration have its counterweight in their female partners' increased employment?

Using data from the Fragile Families and Wellbeing Study, this paper is among the first to empirically evaluate the relationship between the incarceration of women's romantic partners and their employment. It pursued two research questions. First, how is men's incarceration associated with the number of hours their female partners, or women with whom they share children, work? Drawing on theory and recent research (Chiricos, Barrick, Bales, & Bontrager, 2007; Massoglia, Firebaugh, & Warner, 2013; Swisher & Waller, 2008; Turney & Wildeman, 2015; Wildeman, Turney, & Yi, 2016), this paper also examined whether the association between partner incarceration and work hours varied by race/ethnicity, family composition, and conditions of the incarceration.

This study contributes to social science research in a number of ways. First, the exclusive focus on the employment and earnings of formerly incarcerated individuals means that we may have underestimated the impact of incarceration on labor market outcomes. Additionally, we may have underestimated its impact on social inequality. Although scholars have explored the role of incarceration in fostering racial and economic inequality (Manza & Uggen, 2006; Massoglia, 2008; Western, 2006; Wildeman & Muller, 2012), little research has considered its role in reproducing gender inequality (see Davis, 2003 for an exception). An examination of women's employment does exactly that. Incarceration is a gendered institution – it “reflects and further entrenches the gendered structure of the larger society” (Davis, 2003, p. 61). Incarceration intensifies the marginalization of disadvantaged families and communities, and the burden of these hardships is shouldered primarily by women (Christian & Thomas, 2009; Roberts, 2004). By removing men from marriages and relationships – which discourages shared responsibility for children, the home, and the household economy – prisons, jails and justice system processes reproduce gender inequality. Women have long shouldered responsibility for balancing work and family, but incarceration may exacerbate this responsibility or, at the very least, maintain it. If we ignore the ways in which incarceration constrains not only disadvantaged men but also disadvantaged women we fail to consider a potentially significant, and for some women, dominant institution in their lives (Comfort, 2008).

Background

The primary reason to expect that men's incarceration will increase their female partners' employment lies in what is known about the adverse consequences of incarceration for family income. Most men are employed before incarceration and cite wages and salary as their main source of income (Mumola, 2000). Although these men's incomes may be modest, they do report providing primary financial support for their families (Braman, 2004; Glaze & Maruschak, 2010). Thus, when a man is removed from household wage earning via incarceration, his family experiences an immediate reduction in economic resources (Arditti, Lamber-Shute, & Joest, 2003; Geller et al., 2011; Johnson, 2008). The effects on employment are not limited to periods of incarceration; they continue after release (Holzer, 2009; Kling, 2006; Pager, 2003; Pager, Bonikowski, & Western, 2009; Pager, Western, & Sugie, 2009; Western, 2006). About two-thirds of former prison inmates, mostly young men without a college education, remain out of work a year after release (Langan & Levin, 2002; Pager, Western, et al., 2009; Petersilia, 2003; Travis, 2005). Although usually much shorter in duration, jail stays also have consequences for job attainment and maintenance. By preventing individuals from attending work, the incapacitating effect of even a night or a few weeks in jail can jeopardize employment (Fernandes, 2015; Grogger, 1995; Sullivan, 1989).

In addition to reducing resources, ethnographic research has shown that incarceration is associated with a number of financial costs for families derived from replacing in-kind contributions to household labor, maintaining contact with prisoners, and supporting them financially (Arditti et al., 2003; Braman, 2004; Comfort, 2008; Grinstead et al., 2001). The amount of money family members spend on travel for prison visits, phone calls, quarterly packages, and commissary has received the most attention (Comfort, 2008; deVuono-powell et al., 2015; Grinstead et al., 2001), but families often take responsibility for attorney fees, fines, and legal debt associated with involvement in the criminal justice system (Comfort, 2008; deVuono-powell et al., 2015; Harris, 2016; Harris, Evans, & Beckett, 2010). Monetary sanctions are imposed by judges across felony and misdemeanor offenses in amounts much higher than the expected earnings of people who come into contact with the criminal justice system. Thus, for disadvantaged families, even minimal criminal justice involvement (i.e., short jail stays) can create long-term expenses (Harris, 2016; Harris et al., 2010).

The incarceration of women's partners alters the economic resources available to them and may shift responsibility for wage earning. Incarceration, then, is not unlike other family transitions that reduce household resources. Both divorced women and women's whose husbands have become unemployed face substantial reductions in household income, and these events, under certain circumstances, have been shown to increase women's labor supply (Bradbury & Katz, 2002; Lundberg, 1985; Mattingly & Smith, 2010; Moehling, 2001; van Damme, Kalmijn, & Uunk, 2009). In the case of husband unemployment, wives labor market response has been termed the “added worker effect.” The basic premise is that wives enter the labor market or increase the hours they spend working in order the smooth out fluctuation in family income (Lundberg, 1985; Mattingly & Smith, 2010; Moehling, 2001). Although incarceration has in common with divorce and husband unemployment a potential to reduce economic resources, incarceration is distinct for a few reasons. First, the families of incarcerated men are extremely disadvantaged – even more so than families likely to experience divorce and unemployment. The poverty and racial inequality these families already face may limit the applicability of an economic model that balances one wage-earner's income loss with gains of another. Even when the added worker response was directly tested, Lundberg (1985) found only a small effect for white women from low-income households and none at all for low-income black women.

Second, incarceration is incapacitating; it not only removes individuals from household wage earning, it also places them in institutions that circumscribe their ability to participate in family life (Western, 2006). Men often share childrearing responsibilities with the mothers of their children prior to incarceration (Nurse, 2002), and the loss of this support, whether it be for a jail stay of a few weeks or a prison sentence of several years, puts constraints on the time women can devote to work activities. Furthermore, maintaining connections between children and incarcerated fathers is time consuming; prisons in particular are often a distance from families' homes, and the responsibility for child-father visits rests on women's shoulders. The incapacitation of prisoners and the additional caregiving responsibilities faced by their partners, may limit women's ability to respond to income loss by increasing their labor supply. Previous findings, although limited, regarding the relationship between women's employment and family member incarceration suggest this is the case. Interviews with prison visitors providing primary care for inmates' children indicated a reduction in employment, despite substantial financial strain resulting from family member absence. The authors suggested that the unavailability of the incarcerated individual for childcare may increase work-family conflict and lead to labor market exits (Arditti et al., 2003).

It is also possible that partner incarceration does not increase women's employment simply because employment is not the only way to cushion income loss. Families can respond to financial strain by cutting back on expenditures, and ethnographic research has shown that partners of incarcerated men sometimes forfeit their own consumption (e.g., tightening their personal food budgets) in order to cover the costs of maintaining contact with an inmate and providing him with basic comforts (Comfort, 2008; Fishman, 1988). Families of incarcerated men also generate near-cash income through participation in public assistance programs such as food stamps and Medicaid (Sugie, 2012). They may rely on work in the informal labor market – such as cleaning houses, babysitting, and doing hair – which can offer flexibility to meet the care needs of their families (Edin & Lein, 1997; Harris, 1993). Additionally, re-partnering and living with other family members can be sources of income as well. Women may move to their parents' home to reduce costs, and cohabiting with a new, employed partner may limit women's need to increase employment (van Damme et al., 2009).

Variation in the Association between Partner Incarceration and Women's Employment

It is possible that incarceration will increase employment for some women but not others. For one, the association between the incarceration of women's partners and their employment may depend on race/ethnicity. Previous research has suggested that the consequences of incarceration may be relatively minor for racial/ethnic minority women given pre-existing disparities in the economic outcomes and exposure to incarceration (Chiricos et al., 2007; Massoglia et al., 2013; Swisher & Waller, 2008; Turney & Wildeman, 2015). Massoglia and colleagues (2013), who found a significant negative effect of incarceration on neighborhood attainment for white men but not black men, argued that because black inmates come from such disadvantaged neighborhoods they have much less to lose than white inmates. Blacks experience what has been referred to as a “floor effect.” Turney and Wildeman (2015) used a similar explanation for heterogeneity in the effect of maternal incarceration on child wellbeing. The authors reasoned that incarceration may be least consequential for children most likely to experience incarceration because these children, who already face significant disadvantages, cannot accumulate additional adverse consequences; they have reached a point of saturation.

The concept of a “floor effect” or “reaching saturation” can be applied to women with incarcerated partners in two ways. First, racial/ethnic minority women must contend with economic barriers due to racial discrimination even if their partners never go to prison. Thus, incarceration may do little to change their economic situation. They may have reached saturation, in terms of accumulation of economic disadvantage. Second, racial/ethnic minority women face disadvantages not only in general economic terms but specifically in the labor market. Black and Hispanic women are often relegated to low-wage service sector and caregiving jobs that provide few flexibilities when family circumstances change and women need more hours (Burton & Tucker, 2009; Golden, 2005; Lambert, Haley-Lock, & Henly, 2012; Reid, 2002; Reskin, 1999; Swanberg, Catsouphes, & Drescher-Burke, 2005). It may be that racial/ethnic minority women are “at capacity” – they are working as much as they can at their current jobs, given the structure of low-wage work. Additionally, the compensatory effect of women's employment may be smaller for black women whose labor force attachment has been high relative to white women's for decades (Amott & Matthaei, 1991; Casper & Bianchi, 2002). The household economies of black families have long depended on women's employment (Collins, 2011; Furstenburg, 2007; Glenn, 1994), and perhaps loss of income due to incarceration cannot easily be balanced with additional labor supply of an already employed woman. Finally, disadvantaged families make decisions about employment in the context of public assistance receipt. Potential reductions in these benefits when earnings increase (i.e., marginal tax rates) may keep highly disadvantaged women from increasing their hours (Harris, 1993).

Variation in the association between partner incarceration and employment may also derive from conditions of the incarceration itself, such as the length of the incarceration spell. Returning to the concept of the added worker effect, one of the most important attributes of this model is that changes in wives' employment are timed to smooth out fluctuation in household income – they are a response to temporary loss of income, rather than long-term loss. Permanent decreases in income are expected to cause an adjustment in consumption which takes the place of an employment response (Lundberg, 1985). In this study, I measured permanence versus transience of incarceration in two ways. The first and most straightforward approach used information on the duration of the incarceration spell. Following the assumptions of the added worker effect, short spells of incarceration will be associated with higher employment, but longer spells will not. We can think about permanence and transience in a second way. Suppose the first time incarceration of a woman's partner is more likely to represent a transitory income reduction, while a second or third incarceration is a signal of a reoccurring or long-term problem associated with low permanent income. Then, women will respond to first time incarcerations with an employment increase but will adjust in other ways to a repeat occurrence.

Finally, the impact of partner incarceration on women's employment may depend on family composition. Economic models regarding family labor supply decisions tend to focus on households headed by married couples, but families have become increasing complex, and divergence from the married bio-parent family model is particularly common among some disadvantaged groups (Bumpass & Lu, 2000; Cancian & Haskins, 2014; M. J. Carlson & Furstenberg, 2006; Guzzo & Furstenberg, 2007; Raley, Sweeney, & Wondra, 2015; Sykes & Pettit, 2014). Family composition may affect the degree of the economic instability families of incarcerated men experience, the resources available in times of need, and, in turn, how women use their employment as a resource to manage financial strain. In this study, I captured family composition in two ways: the status of women's relationship with her child's father and multiple-partner fertility. When a man's economic resources are concentrated on a single household – perhaps when he is married and/or has fathered children with only one woman – the loss of those resources during a period of incarceration may be particularly challenging. The economic consequences of men's imprisonment likely reverberate in more complex families – families characterized by multiple partner fertility, cohabitation or non-residential relationships – because they too receive formal and informal financial support (Geller et al., 2011). However, family complexity may mean that men's economic resources are diffused across households (Cancian & Meyer, 2011; M. J. Carlson & Meyer, 2014; Sinkewicz & Garfinkel, 2009), and incarceration is less consequential for each family's economic well-being. In addition, women whose families are complex may have structures in place, such as new partners and female kin, to support fluctuations in the contributions of their children's fathers (Edin & Lein, 1997; Kalil & Ryan, 2010; Patillo-McCoy, 1999; Stack, 1974). For example, women who have children with multiple fathers may call on other fathers to step in when one father is unable to provide support.

In sum, the incarceration of women's partners creates a need for income from employment by removing men from household wage earning, creating barriers to men's employment post-release, and exacerbating economic instability for their families. However, women's ability to increase employment may be constrained if women's time is consumed by childcare responsibilities and efforts to keep their families intact despite separation. In addition, they may find other ways to make ends meet. Theory and recent research also suggests that the association between partner incarceration and women's employment may depend on how factors such as race/ethnicity, conditions of incarceration and family composition structure the degree of economic loss and resources available to women in times of need.

Additional Characteristics Associated with Partner Incarceration and Women's Employment

The multivariate models included a number of control variables that are related to incarceration and women's employment. Immigrants are less likely than individuals born in the U.S. to be incarcerated (Butcher & Piehl, 1998), and immigrant women are less likely to be employed (MacPherson & Stewart, 1989; Schoeni, 1998). Age is associated with incarceration (Pettit & Western, 2004). Childhood family structure is also associated with incarceration (Harper & McLanahan, 2004), and it is an important mechanism for the reproduction of economic inequality which may impact adult resources and employment (McLanahan & Percheski, 2008). Human capital is an key determinant of women's employment, and education is considered one of the most important indicators of human capital (Becker, 1981; Percheski, 2008; Reid, 2002). Education is also associated with incarceration (Carson, 2015). Other dimensions of human capital include hourly wages, cognitive skills, and impulsivity (Percheski, 2008; Reid, 2002). Intimate partner violence impedes a woman's ability to hold a steady job (Riger, Staggs, & Schewe, 2004; Tolman & Wang, 2005), as might having a drug or alcohol problem, or a history of incarceration (Blitz, 2006; Rose, Michalsen, Wiest, & Fabian, 2008; Tonkin, Dickie, Alemagno, & Grove, 2004). These characteristics are also correlated with incarceration (Mumola & Karberg, 2006; Western, 2006).

Data, Measures, and Analytic Strategy

Data

I used data from the Fragile Families and Child Wellbeing Study, a longitudinal survey that follows a cohort of new and mostly unmarried parents living in 20 large U.S. cities. The study began with interviews of nearly 5,000 parents, conducted between February 1998 and September 2000 (Reichman, Teitler, Garfinkel, & McLanahan, 2001). Mothers were interviewed in the hospital following the birth of their child, and fathers were interviewed in the hospital or shortly thereafter. Subsequent interviews were conducted by phone one, three, five, nine, and fifteen years later. Fragile Families oversampled unmarried parents, which means the sample was economically disadvantaged and included a substantial number of men who came into contact with the penal system prior to or during the course of the study. These data were particularly suitable for studying the impact of partners' incarceration on women's employment because they included substantial information about respondents' family life and economic activity as well as factors that may affect the likelihood of both partner incarceration and employment.

The final analytic sample for this paper consisted of 3,780 observations. I relied on data primarily from the three- and five-year surveys because men's incarceration is most accurately measured between these two surveys. Thus, of the 4,898 observations in the baseline sample, I dropped 1,051 (21%) women who did not participate in both the three- and five-year surveys. I dropped an additional 30 (<1%) observations in which the child's father was no longer living at the three-year survey. I excluded an additional 37 (1%) observations missing data on outcome variables. I used multiple imputation by chained equations to preserve observations missing values for other variables (White, Royston, & Wood, 2009). Few observations were missing data on the independent variables. About 9 percent of observations were missing values for men's incarceration. All other control variables were missing for one percent or fewer of the observations, with the exception of women's most recent hourly wage and men's pre-incarceration financial contribution. Models using complete cases show nearly identical results (results available upon request).

Measures

Women's employment

Women participating in the Fragile Families study were asked whether they did any regular work for pay in the week prior to the survey as well as the number of hours per week they usually worked at their most recent job. Because a most recent job is not necessarily a current job, I used information from both survey items regarding employment to construct an outcome variable, hours of work, that represented the number of hours a woman worked per week in her current, formal sector job. Some analyses included a lagged measure of the dependent variable.

Explanatory variable

My primary explanatory variable was recent partner incarceration. A woman experienced the recent incarceration of her partner if the father of her child was incarcerated between the three- and five-year surveys, including incarceration at the five-year survey. Incarceration was captured in the Fragile Families survey in three ways: through women's direct reports; through administration of men's survey in prison or jail; and from indirect reports that a woman's partner is or was incarcerated (e.g., reports incarceration as a reason the man was unable to find a job) (Turney & Wildeman, 2013). Consistent with other research in the area, I used both men's and women's reports of incarceration at and between surveys to construct the measure of recent partner incarceration.

The inclusion of multiple measures of incarceration in the survey means that the data provided an excellent opportunity to investigate the consequences of men's incarceration for their families; however, the measure of recent incarceration is subject to limitations. The data did not include information about whether a woman's partner was incarcerated in jail or prison, and these two types of incarceration may differentially impact women's employment. As with many longitudinal studies, the precise timing of events that occurred between surveys was a problem across variables. Thus, information regarding the exact timing of incarceration between the three- and five-year surveys was not available. In addition, information about the duration of incarceration was available for only 81 percent of observations in which women experienced the recent incarceration of their partner, and this information was not missing at random. Because the duration of an incarceration spell may shape the economic impact on a family and thus, women's employment response, I used this information in supplemental analyses which are described below. Further data collection efforts should attend to duration more thoroughly so that stronger conclusions can be drawn.

Control variables

The analyses adjusted for several individual-level characteristics associated with women's employment and attachment to incarcerated men. These variables were measured at or before the three-year survey, prior to the measurement of recent incarceration. Demographic controls included women's race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic other race) and two dichotomous variables indicating a woman was born outside the United States and the woman lived with both parents at age 15. All were measured at the baseline interview. Age was a continuous variable measured at the three-year survey. The analyses also controlled for several measures of human capital. Educational attainment was a categorical variable with four mutually exclusive categories (less than high school diploma, high school diploma or GED, some college, and college degree) measured at the three-year survey. Hourly wage was a continuous variable indicating a woman's hourly wage at her current or most recent job. Women's cognitive scores were derived from a set of eight questions taken from the Similarities Subtest of the Weschler Adult Intelligence Scale – Revised (WAIS-R). Impulsivity was based on an average of each woman's responses (1 = strongly disagree to 4 = strongly agree) to an abbreviated form of (Dickman, 1990) impulsivity scale. The six questions included in the scale were: “I will often say whatever comes into my head without thinking first;” “Often, I don't spend enough time thinking over a situation before I act;” “I often say and do things without considering the consequences;” “I often get into trouble because I don't think before I act;” “Many times, the plans I make don't work out because I haven't gone over them carefully enough in advance;” and “I often make up my mind without taking the time to consider the situation from all angles” (α = .84) (Wildeman et al., 2012). A dichotomous variable indicated whether the respondent reported she was in fair or poor health. Depression and drug or alcohol problem were both assessed using women's responses to the Composite International Diagnostic Interview - Short Form (CIDI-SF) (Kessler, Andrews, Mroczek, Ustun, & Wittchen, 1998). Own incarceration history was based on the woman's report that she has spent time in a correctional institution as of the three-year survey.

An array of variables measuring family economic resources were also included. Earnings from informal work measured women's total previous year earnings from “off the books” work, work in the illicit labor market, and work in her own business that was not reported as regular work. Household income was a measure representing household income in relation to the poverty line, which is established by the Census Bureau and based on household size and composition. The measure corresponded to the year prior to the three-year survey and was included in models as a series of five dummy variables: less than 50 percent of the poverty line, 50-99 percent of the poverty line, 100-199 percent of the poverty line, 200-299 percent of the poverty line, and 300 percent of the poverty line or higher. Use of public assistance was a dichotomous variable that indicated whether a woman received TANF, Food Stamps, Medicaid, or Social Security Income (SSI) in the year prior to the three-year interview. Partner financial support was a dichotomous variable that measured whether or not the partner provided financial support to the woman in the 12 months leading up to the three-year survey. A partner was considered to be providing financial support if he was living with the woman and employed in either the formal or informal labor market, or he was not living with the woman and the woman reported he provided formal or informal child support (Geller et al., 2011). Although researchers have calculated the amount of partners' financial contribution at the five-year survey, it was not possible to construct the same continuous measure at the three-year survey due to an inconsistency in survey questions. Partner prior incarceration indicated whether or not a woman's partner had been incarcerated at some point before the three-year survey. This measure was distinct from recent incarceration; each measure referred to a separate time period in which incarceration occurred. Some women had partners who experienced both recent and prior incarceration.

Other household and family characteristics included number of children under the age of five in the household at year three. Relationship status was a categorical variable indicating women's relationship with her child's father. Four mutually exclusive categories (separated, married, cohabiting, non-residential romantic relationship) made up the measure. Involved with a new partner was a binary measure indicating whether a woman was in a romantic relationship with a partner other than the father of the focal child at the three-year survey. Own multiple partner fertility indicated whether a woman had children with men other than the partner/focal child's father, and partner multiple partner fertility indicated whether the child's father had children with other women. Finally, intimate partner violence was based on the woman's report that her partner at the time of the three-year interview (either her child's father if she was romantically involved with him or a new partner) tried to keep her from seeing or talking to friends and family, tried to prevent her from going to work or school, withheld money, slapped, kicked or hit her, or tried to make her have sex or do sexual things she did not want to do (McLanahan et al., 2014).

Moderators

In some analyses, I considered five moderators of the association between recent partner incarceration and women's hours of work: race/ethnicity, partner prior incarceration, relationship status, women's own multiple partner fertility, and partner's multiple partner fertility. All are measured as stated above.

Analytic Strategy

In the first analytic stage, I estimated women's hours of work as a function of recent partner incarceration. Hours of work was a continuous dependent variable, so I used ordinary least squares (OLS) regression. Hours of work was not normally distributed, (a large number of women worked zero hours last week); thus, I also performed a Tobit analysis, which yielded substantively similar results (available upon request). Results from OLS models are presented because they are more easily interpretable. Model 1 included recent partner incarceration, the key explanatory variable, and partner's prior incarceration as a control. Model 2 added a wide range of control variables measuring women's individual and family circumstances as described above. In Model 3, I added a lagged measure of the dependent variable. Thus, any persistent association between partner incarceration and women's hours of work at the five-year survey was net of work hours at the three-year survey, or before the start of the most recent incarceration. Although the inclusion of a lagged dependent variable cannot fully rule out the possibility that an association between partner incarceration and hours of work is spurious, considered in conjunction with a rich set of controls, a significant association between incarceration and hours of work in this model would provide more convincing evidence of a causal effect (Finkel, 1995).

In the second analytic stage, I considered five potential moderators of the association between recent partner incarceration and women's hours of work: race/ethnicity, partner prior incarceration, relationship status, women's multiple partner fertility, and their partners' multiple partner fertility. In the third and final analytic stage, I considered the relationship between duration of incarceration and women's hours of work. To do so, I utilized an alternative specification of the explanatory variable, recent partner incarceration, that includes information about length of the incarceration spell (see Table 4). This variable consisted of four categories: recent incarceration less than three months (n = 228), recent incarceration three months or greater (n = 450), duration missing (n = 177), and no recent incarceration (n = 2,925).

Table 4. OLS Regression Models Estimating Women's Hours of Work with Alternative Specification of Recent Partner Incarceration: Duration.

	Model 1	Model 2	Model 3
	Unadjusted	+ Controls	+ Lagged DV
Recent partner incarceration (ref. no incarceration)
Less than 3 months	4.491**	5.105**	3.716**
	(1.555)	(1.500)	1.420
3 months or greater	-0.733	0.306	-0.152
	(1.186)	(1.174)	1.090
Missing duration	-1.362	0.016	0.608
	(1.779)	(1.746)	(1.640)
Race/Ethnicity (ref. Non-Hispanic white)
Non-Hispanic Black		5.066***	2.581**
		(0.920)	(0.865)
Hispanic		3.507**	1.857†
		(1.050)	(0.983)
Non-Hispanic other race/ethnicity		2.902	1.533
		(1.888)	(1.763)
Age		-0.076	-0.091
		(0.063)	(0.059)
Education (ref. Less than High School)
HS Diploma or GED		4.158***	2.440**
		(0.866)	(0.813)
Some college		4.696***	2.236*
		(0.930)	(0.875)
Bachelors degree or higher		3.710*	1.344
		(1.459)	(1.370)
Lived with both parents at age 15		-0.061	-0.098
		(0.697)	(0.651)
Immigrant		-0.813	-0.625
		(1.091)	(1.020)
Hourly wage at most recent job		0.144**	0.157**
		(0.055)	(0.049)
Earnings from informal work		0.00006	0.00007
		(0.00006)	(0.00005)
Cognitive score		0.125	0.173
		(0.133)	(0.125)
Impulsivity		-1.055†	-0.683
		(0.542)	(0.505)
Fair or poor health		-2.166*	-1.378
		(0.961)	(0.898)
Depression		-1.600†	-0.767
		(0.944)	(0.882)
Drug or alcohol problem		-8.045*	-5.482†
		(3.368)	(3.152)
Own Incarceration history		-4.473*	-3.251†
		(1.919)	(1.788)
Intimate partner violence		0.814	0.888
		(1.016)	(0.948)
No. of children <age 5 in household		-0.709†	-0.283
		(0.410)	(0.383)
Relationship status (ref. Separated)
Married		-6.118***	-2.671**
		(1.064)	(1.004)
Cohabiting		-1.943†	-1.009
		(0.994)	(0.929)
Non-residential romantic		1.285	1.214
		(1.849)	(1.728)
Involved with new partner		-0.944	-0.672
		(0.989)	(0.923)
Own multiple partner fertility		-0.568	-0.706
		(0.741)	(0.692)
Partner multiple partner fertility		1.519*	1.212†
		(0.730)	(0.679)
Household income (ref. <50% of poverty line)
50-99% of poverty line		5.169***	2.458**
		(0.983)	(0.925)
100-199% of poverty line		7.284***	2.864**
		(0.979)	(0.933)
200-299% of poverty line		11.888***	5.944***
		(1.246)	(1.190)
300+% of poverty line		11.165***	5.038***
		(1.355)	(1.289)
Public assistance		-1.141	0.635
		(0.842)	(0.790)
Partner provides financial support		1.599†	1.178
		(0.834)	(0.773)
Partner prior incarceration	-1.261†	-0.263	-0.455
	(0.757)	(0.781)	(0.729)
Hours of work (y3)			0.378***
			(0.016)
Constant	22.367***	13.861	9.259**
Adjusted R-squared	0.003	0.098	0.214
N	3,780	3,780	3,780

^†p<.10

^*p<.05

^**p<.01

^***p<.001

Sample Description

Table 1 shows descriptive statistics for all variables. Including those who did not work, women in the sample worked, on average, 22 hours per week at the five-year survey. The sample was relatively disadvantaged across a range of demographic and socioeconomic characteristics. Recent incarceration of women's partners was common; 23 percent of women were attached to incarcerated men. About 78 percent of women were racial/ethnic minorities, and nearly 15 percent were born outside the U.S. More than half (57 percent) of women had no more than a high school diploma or GED. About 13 percent of women reported fair or poor health, 14 percent reported depression, and 12 percent had experienced intimate partner violence. More than half were in a romantic relationship with their partner/child's father, and 18 percent were involved with a new partner. About 42 percent of women's household incomes were below the poverty line, and nearly two-thirds received some form of public assistance.

Table 1. Descriptive Statistics of All Variables Included in Analysis.

	Full Sample
	Mean or Percent	SD
Dependent variable
Hours of work (y5)	21.94	20.12
Key independent variable
Partner recent incarceration (y5)	22.7
Control Variables
Race/Ethnicity (b)
non-Hispanic White	21.7
non-Hispanic Black	48.7
Hispanic	26.1
Non-Hispanic other race/ethnicity	3.6
Age (y3)	28.17	6.05
Education (y3)
Less than High School	27.9
HS Diploma or GED	29.5
Some college	30.3
Bachelors degree or higher	12.3
Lived with both parents at age 15 (b)	43.0
Immigrant	14.6
Hourly wage at most recent job	11.19	6.91
Earnings from informal work	595.95	5,504.38
Cognitive score	6.77	2.65
Impulsivity	2.03	.61
Fair or poor health	13.4
Depression	14.1
Drug or alcohol problem	.9
Own incarceration history	2.9
Intimate partner violence	12.3
No. of children <age 5 in household	1.61	.81
Relationship status (with partner)
Separated	42.2
Married	32.3
Cohabiting	22.2
Non-residential romantic	3.3
Involved with new partner	18.4
Own multiple partner fertility	39.3
Partner multiple partner fertility	40.6
Household income
<50% of poverty line	22.2
50-99% of poverty line	19.5
100-199% of poverty line	25.2
200-299% of poverty line	13.8
300+% of poverty line	19.3
Public assistance	62.5
Partner provides financial support	73.9
Partner prior incarceration	43.4
N	3,780

Results

Estimating Women's Hours of Work

Table 2 shows multivariate results from OLS models estimating women's hours of work as a function of recent partner incarceration. Model 1, which adjusted only for partner's prior incarceration, showed no statistically significant relationship between recent partner incarceration and women's hours of work. Model 2 adjusted for the full set of control variables, and recent partner incarceration was associated with about two additional hours of work per week (p < .10). The increase in magnitude of the relationship between recent incarceration and hours of work suggests that some control variables acted as suppressors. For example, recent incarceration was positively associated with hours of work (see Model 1), but recent incarceration was negatively associated with women's education, and education was positively associated with hours of work. When I adjusted for the lagged dependent variable in Model 3, the size of the coefficient for recent incarceration decreased and was not statistically significant, even at the .10 level.

Table 2. OLS Regression Models Estimating Women's Hours of Work.

	Model 1	Model 2	Model 3
	Unadjusted	+ Controls	+ Lagged DV
Recent partner incarceration	0.659	1.735†	1.191
	(0.931)	(0.945)	(0.879)
Race/Ethnicity (ref. Non-Hispanic white)
Non-Hispanic Black		4.956***	2.491**
		(0.920)	(0.864)
Hispanic		3.474**	1.836†
		(1.051)	(0.983)
Non-Hispanic other race/ethnicity		2.863	1.502
		(1.891)	(1.765)
Age		-0.075	-0.088
		(0.063)	(0.059)
Education (ref. Less than High School)
HS Diploma or GED		4.220***	2.471**
		(0.867)	(0.813)
Some college		4.841***	2.324**
		(0.930)	(0.875)
Bachelors degree or higher		3.775*	1.365
		(1.459)	(1.370)
Lived with both parents at age 15		-0.071	-0.110
		(0.697)	(0.652)
Immigrant		-0.754	-0.582
		(1.091)	(1.020)
Hourly wage at most recent job		0.142*	0.156**
		(0.055)	(0.049)
Earnings from informal work		0.00006	0.00007
		(0.00006)	(0.00005)
Cognitive score		0.127	0.174
		(0.133)	(0.125)
Impulsivity		-1.074*	-0.710
		(0.542)	(0.505)
Fair or poor health		-2.121*	-1.345
		(0.962)	(0.899)
Depression		-1.537	-0.705
		(0.944)	(0.882)
Drug or alcohol problem		-7.905*	-5.315†
		(3.364)	(3.150)
Own Incarceration history		-4.536*	-3.303†
		(1.920)	(1.789)
Intimate partner violence		1.008	1.027
		(1.014)	(0.946)
No. of children <age 5 in household		-0.671	-0.254
		(0.410)	(0.383)
Relationship status (ref. Separated)
Married		-6.028***	-2.600*
		(1.065)	(1.004)
Cohabiting		-1.721†	-0.846
		(0.993)	(0.927)
Non-residential romantic		1.397	1.278
		(1.851)	(1.728)
Involved with new partner		-0.828	-0.576
		(0.989)	(0.923)
Own multiple partner fertility		-0.620	-0.740
		(0.741)	(0.692)
Partner multiple partner fertility		1.548*	1.239†
		(0.730)	(0.679)
Household income (ref. <50% of poverty line)
50-99% of poverty line		5.322***	2.560**
		(0.982)	(0.924)
100-199% of poverty line		7.400***	2.928**
		(0.979)	(0.933)
200-299% of poverty line		12.023***	6.023***
		(1.247)	(1.190)
300+% of poverty line		11.167***	5.023***
		(1.356)	(1.289)
Public assistance		-1.151	0.624
		(0.842)	(0.790)
Partner provides financial support		1.820*	1.325†
		(0.826)	(0.767)
Partner prior incarceration	-1.496	-0.429	-0.595
	(0.755)	(0.780)	(0.728)
Hours of work (y3)			0.380***
			(0.016)
Constant	22.441	13.440	8.964**
Adjusted R-squared	0.0006	0.096	0.213
N	3,780	3,780	3,780

^†p<.10

^*p<.05

^**p<.01

^***p<.001

Variation in Association between Partner Incarceration and Women's Hours of Work

In the second analytic stage I considered how the association between recent partner incarceration and women's hours of work varied by individual and partner characteristics. Figures 1 shows predicted hours of work based on five models that included separately: 1) recent partner incarceration X race/ethnicity, 2) recent partner incarceration X partner prior incarceration, 3) recent partner incarceration X relationship status, 4) recent partner incarceration X multiple partner fertility, and 5) recent partner incarceration X partner's multiple partner fertility. Statistically significant differences (at p < .05), indicated by open markers, were determined by estimating first differences (or the effect of changing an average observation from no recent incarceration to recent incarceration) rather than by confidence intervals. Table 3 presents coefficients for the variables of interest and their interaction terms.

Figure 1. Predicted Hours of Work as a Function of Recent Partner Incarceration and Interaction Terms.

Note: Open markers indicate significant differences (p<0.05)

Table 3. OLS Regression Models Estimating Women's Hours of Work with Interactions.

	Model 3		Model 7		Model 8		Model 9		Model 10		Model 11
	All controls and LDV		+ Race/ethnicity interaction		+ Incarceration history interaction		+ Relationship status interaction		+ Own multiple partner fertility interaction		+ Partner multiple partner fertility inter.
	β	SE	β	SE	β	SE	β	SE	β	SE	β	SE
Recent partner incarceration	1.191	(.879)	5.077	(1.978)	5.073	(1.787)	1.106	(1.110)	2.425	(1.146)	2.309	(1.235)
Race/ethnicity (ref. Non-Hispanic white)
Non-Hispanic Black			3.325	(0.978)
Hispanic			2.894	(1.092)
Non-Hispanic other race/ethnicity			2.126	(1.923)
Race/ethnicity x recent partner incarceration
Non-Hispanic Black x recent partner inc.			-4.258	(2.187)
Hispanic x recent partner incarceration			-5.557	(2.504)
Non-Hispanic other race/ethnicity			-3.345	(5.035)
Incarceration history					0.200	(0.787)
Incarceration history x recent partner inc.					-5.007	(2.010)
Residential Status (ref. is separated)
Married							-3.156	(1.055)
Cohabiting							-0.429	(1.014)
Non-residential romantic							1.313	(2.136)
Residential Status x recent incarceration
Married							5.966	(2.524)
Cohabiting							-2.740	(2.004)
Non-residential romantic							-0.046	(3.534)
Own multiple partner fertility									-0.122	(0.795)
Own multiple partner fertility x recent partner inc.									-2.413	(1.536)
Partner multiple partner fertility											1.761	(0.785)
Partner multiple partner fertility x recent inc.											-2.027	(1.616)
Constant	8.964		8.075		8.490		9.048		8.647		8.731
Adjusted R-squared	0.213		0.213		0.214		0.214		0.213		0.213
N	3,780		3,780		3,780		3,780		3,780		3,780

Turning first to variation by race/ethnicity. Panel A in Figure 1 shows that white women who had recently incarcerated partners worked significantly more hours than white women who did not have recently incarcerated partners (p < .05). White women with recently incarcerated partners worked about 24.3 hours, and white women whose partners had not been recently incarcerated worked about 19.2 hours. Panel B shows predicted hours of work from a model that included the interaction between recent and prior incarceration. Among women whose partners had no history of incarceration, women whose partners were incarcerated for the first time worked about 26.7 hours, and women whose partners were not incarcerated worked about 21.7 hours. There was no significant relationship between recent incarceration and hours of work for women whose partner had a history of incarceration.

Panels C through E in Figure 1 show variation in the relationship between partner incarceration and hours of work by three measures of family complexity. Regarding relationship status, a significant difference in predicted hours of work was found only among women married to the incarcerated man at the three-year survey (p < .01). Married women with recently incarcerated partners worked about 26.7 hours, and married women whose partners had not been recently incarcerated work about 19.6 hours. In addition, women who had children only with a recently incarcerated partner worked 24.2 hours, while other women who had children with one father worked 21.7 hours. There was no statistically significant variation by partner's multiple partner fertility.

To further explore variation by conditions of the incarceration, I considered an alternative specification of the recent partner incarceration variable that included information on duration (see Table 4). Incarceration spells lasting less than three months were more strongly associated with increases in women's employment than spells last three months or more. In Model 6, which included all control variables and a lagged measure of the dependent variable, the coefficient for incarceration lasting less than three months was 3.7. Thus, women whose partners were incarcerated for less than three months worked nearly four hours more than women whose partners were not incarcerated (p < .01). The coefficient for incarceration lasting three months or longer was -.15 and statistically different from the coefficient for incarceration lasting less than 3 months (p < .05). Because a large number of observations were missing information on duration, and the nature of missingness was non-random, these results should be considered preliminary.

Discussion and Conclusion

Research has documented the diminished employment opportunities and earnings associated with a having a criminal record (Holzer, 2009; Pager, 2003; Pager, Bonikowski, et al., 2009; Pager, Western, et al., 2009) as well as the economic consequences of men's incarceration for their families (deVuono-powell et al., 2015; Geller et al., 2011; Schwartz-Soicher, Geller, & Garfinkel, 2011). Yet, little research has considered how men's incarceration is associated with the employment of the women they leave behind. This is an unfortunate oversight; given the loss of household income and costs associated with imprisonment (deVuono-powell et al., 2015; Grinstead et al., 2001; Johnson, 2008), incarceration has the potential to alter the employment of not only incarcerated individuals but also the women to whom they are connected. In addition, this omission leaves the gendered consequences of incarceration for employment unexplored.

I used data from the Fragile Families and Child Wellbeing Study to investigate the relationship between men's imprisonment and their partners' employment as well as variation in the association. I found no statistically significant relationship between partner incarceration and women's hours of work, but, consistent with previous research (Chiricos et al., 2007; Massoglia et al., 2013; Swisher & Waller, 2008; Turney & Wildeman, 2015), attention solely to the “average effects” of incarceration masked heterogeneity. A significant, positive association was evident among white women, married women, women who had children only with the incarcerated partner, women experiencing their partners' first incarcerations and women whose partners were incarcerated for less than three months. Among these groups, women whose partners had been recently incarcerated work more hours than their counterparts, even when accounting for the number of hours they worked prior to incarceration.

It is notable that women most impacted by the incarceration of their partners tended to belong to more socially and economically advantaged groups – and the groups least likely to experience the incarceration of their partners. There are several possible explanations for these results. First, women with economic advantages, relative to other women who experience partner incarceration, may have more to lose when their partners are imprisoned (Chiricos et al., 2007; Massoglia et al., 2013). Men's pre-incarceration financial contributions among the recently incarcerated group suggests this may be the case. Receipt of financial support from their partners was more common among groups of women where the effects of incarceration were strongest (i.e., whites, married women, women who share children with only one father, and women whose partners have no history of incarceration) (see Table A1). Women in more advantaged groups may be more likely to lose not only the contributions of their incarcerated partners but also a set of economic resources concentrated on them. For example, a married man or a father who shares children with only one mother likely focuses his economic resources on a single household. Because his resources are not dispersed, the actual amount of his financial contribution and the resultant loss due to incarceration may be greater. A second possibility is that women from more advantaged groups may have been working fewer hours than possible at their current employers. Because wives have the advantage of sharing resources with their partners, they may have worked fewer hours than their jobs allowed. If their employment was not “at capacity,” there may have been “room to grow” in the wake of their partners' incarceration. Similarly, white women's labor market positioning relative to other women who have incarcerated partners may mean that they were working better jobs where their hours could be expanded.

That partner incarceration and employment were not linked for women most likely to experience the incarceration of their partners suggests that their employment was driven by other factors. Incarceration is a racialized phenomenon. For black women, involvement with the penal system via the fathers of their children creates a marginalized status that intersects with their other marginalized statuses (Christian & Thomas, 2009). Thus, black women with incarcerated partners comprise a highly disadvantaged group. Fewer black women in the analytic sample were employed in either the formal or informal labor market prior to their partner's most recent incarceration. In addition, more black women lived in poverty and receive public assistance (see Table A1). Thus, the economic barriers black women face anyway, including chronic unemployment of the men in their lives, their own positioning in the labor market, and their involvement in means-tested public assistance programs, may attenuate the impact of incarceration on their families' finances and constitute more significant drivers of the amount of time they spend working.

Although similar arguments could be applied to women whose partners were incarcerated for the first time and for relatively short amounts of time, the concept of the added worker effect provides a more straightforward explanation. First-time and shorter incarcerations may represent a temporary reduction in household income that women respond to by increasing their employment. Higher levels of employment, particularly in the absence of a partner, may be more difficult to sustain in the long-term, so women adjust household consumption when incarceration appears more permanent. Women experiencing first-time and short-term incarceration of their partners are likely to have partners incarcerated in jail. Although I was unable to distinguish jail from prison stays in the Fragile Families data, these results suggest that future research should investigate how various forms and stages of criminal justice contact differentially impact family economic instability and women's employment response. The temporary versus permanent expectation of the added worker model also lends itself to an explanation regarding positive findings among whites and null findings for blacks and Hispanics. The incarceration of white men may be more likely to represent a short-term reduction in family income, but the incarceration of black men may be a signal of low permanent income (Lundberg, 1985). This explanation is supported by what we know about underemployment among racial/ethnic minority men as well as racial/ethnic disparities in sentencing and employment prospects upon release from prison (Malveaux, 1981, 1990, Pager, 2003, 2007; Pager, Bonikowski, et al., 2009; Western, 2006).

Study Limitations

The Fragile Families data set is frequently used to assess the impact of incarceration on children and families; however, it is subject to some important limitations. First, the study was not originally intended for use in examining the effects of incarceration on families; thus it is limited in how much information it contains about the conditions of incarceration. For example, information about the type of facility in which the partner is incarcerated is not available in the public release data, although it was gathered at the one-, three-, and five-year surveys. Researchers have gained access to this information, but a great deal of the data is missing and has been used cautiously (Wildeman et al., 2016). Information about the duration of incarceration is more readily available but still incomplete. This is particularly important for the current study, which incorporates duration of incarceration into the analysis. The results from these models should be considered preliminary, as additional data is needed for confirmation. Other information such as whether and how often women visited their partners as well as the distance of correctional facilities from families is simply unavailable. These factors likely have important consequences for the economic impact of incarceration families and, in turn, how women use employment as a resource. Without additional information about the conditions surrounding the incarceration, and particularly costs incurred by families, the mechanisms linking partner incarceration to employment for certain groups remain unclear. To be sure, our understanding of the collateral consequences of incarceration for families would be enhanced by the collection of more reliable data on the conditions of incarceration.

Additionally, the Fragile Families sample is relatively small when compared to other longitudinal surveys (e.g., the National Longitudinal Survey of Youth, the National Longitudinal Study of Adolescent to Adult Health), but it is the only dataset suitable for the analyses presented here, as it includes information about both incarcerated men and their female partners. The relatively small sample size means that subdividing the sample to examine moderators of the association between partner incarceration and women's employment results in a small number of cases in the relevant subgroups (see Table A1). Therefore, the findings of this study should be interpreted with caution, but they do provide a consistent signal that the relationship between partner incarceration and women's work hours is evident only among more advantaged groups of women.

Finally, longitudinal data are ideal for identifying causal relationships, but in these analyses threats to causal inference have not been eliminated entirely. I took several steps to reduce concern about such threats, such as appropriately time-ordering variables, controlling for a wide range of potential confounders, and including a lagged measure of the dependent variable. Although lagged dependent variable models bring us one step closer to identifying causal associations, such models do not capture the ongoing process through which women make decisions about their employment in response to their economic resources and family lives. The analyses suggested a link, among some groups, between the incarceration of women's partners and the number of hours they work, but it would be unwise to make causal conclusions about this relationship.

Conclusion

The results of this paper indicated a null average association between partner incarceration and the number of hours women worked. This suggests that a large proportion of the disadvantaged women whose partners are removed from their family systems via incarceration do not use employment as an modifiable resource during a time that both qualitative and quantitative research have characterized as financially destabilizing. Thus, for the most part, it appears that the reduction in men's employment due to incarceration does not have its counterweight in their female partners' increased employment. Nonetheless, consideration of heterogeneity in the impact of incarceration showed that women in more advantaged positions – women for whom incarceration may be less common and a more notable economic and social shock – did work additional hours when their partners were imprisoned. Thus, the findings of this study contribute conceptually to scholarship on the collateral consequences of incarceration by providing additional evidence that the impact of incarceration on family outcomes is shaped, in part, by the conditions of confinement and by the array of advantages and disadvantages conferred by racial/ethnic group membership and family composition – two predictors of a wide variety of family and individual outcomes. This has important implications for the reproduction of inequality. That women who have some economic advantages (e.g., white women, married women) may respond to partner incarceration with more hours at work, and women who have fewer economic advantages (e.g., black women, unmarried women) may not, suggests at least a maintenance, if not a deepening, of economic inequality. Because women most at risk of partner incarceration are an extremely disadvantaged group, and the disadvantages they face regardless of their partners' incarceration may be the more significant drivers of their labor force participation, efforts to support such women may require greater attention poverty and the conditions of low-wage employment.

It may be that most women cannot increase their work hours in response to changes in income and expenses associated with their partners' incarceration simply because they already have too much to do. They already shoulder the burden of balancing their families' economic and caregiving needs. Partner incarceration may increase the need for additional income from employment, but it strips families of the human resources that can make meeting both the financial needs and the caregiving needs of their families easier. Future research should examine more directly the consequences of family member incarceration for the work-family interface. This study contributes to a growing body of literature that has documented the consequences of mass incarceration for families. It highlights racial inequalities, gender inequalities and inequalities between families. Given the predominant focus on former prisoners' employment in previous research and the importance of employment for making ends meet in the context of diminishing public safety nets, understanding the consequences, or lack thereof, of men's incarceration for women's employment, adds a new dimension to our understanding of the collateral consequences of incarceration.

Appendix

Table A1. Variation in Employment and Economic Circumstances among Women with Recently Incarcerated Partners.

	Employed in formal labor market	Hourly wages	Employed in informal labor market	Household income below poverty line	Public assistance receipt	Hours of work in formal labor market	Partner provides financial support
	Mean or Percent							n
Race/ethnicity
Non-Hispanic White	62.83	9.78	25.26	43.28	72.02	23.25	58.02	119
Non-Hispanic Black	52.72	9.39	19.33	63.13	86.11	19.58	48.68	548
Hispanic	54.58	9.10	14.35	60.35	76.64	20.11	51.67	170
Non-Hispanic Other	54.44	14.60	5.31	43.92	69.80	19.61	61.79	19
Partner prior incarceration
No prior incarceration	57.15	9.89	11.89	51.7	72.58	21.94	66.79	128
Prior incarceration	54.08	9.43	20.08	60.73	83.55	19.9	48.06	728
Relationship Status
Separated	54.64	9.56	17.82	61.56	82.65	20.26	38.47	600
Married	57.80	10.05	20.20	44.23	66.75	23.06	85.03	72
Cohabiting	54.94	9.20	21.59	52.71	81.49	19.38	81.78	138
Romantic	46.82	8.68	22.25	74.97	97.45	17.40	66.62	45
Own multiple partner fertility
One father	62.20	9.50	19.04	47.89	76.26	22.86	55.16	399
Multiple fathers	47.84	9.50	18.70	69.45	86.86	17.87	47.09	456
Partner's multiple partner fertility
One mother	55.66	9.47	19.10	59.29	81.64	20.62	59.20	357
Multiple mothers	53.73	9.52	18.68	59.45	82.10	19.90	44.87	498
Incarceration duration
Less than 3 months	59.96	9.56	19.70	50.58	74.80	23.12	68.00	228
3 months or greater	54.67	9.48	18.52	61.78	85.67	19.87	47.95	450
Missing duration	47.20	9.46	18.62	64.64	81.51	17.27	36.21	177

Note: all employment and economic circumstances are measured at the three-year survey. The number of observations varies across imputed datasets; the values in the final column are an average across datasets.