Abstract

While the labor market woes of low-skilled male workers in the United States over the past several decades have been well documented, the academic literature identifying causal factors leading to declines in labor force participation (LFP) by young, low-skilled males remains scant. To address this gap, I use the timing and characteristics of welfare-reform policies implemented during the 1990s and fixed-effects, instrumental variable regression modeling to show that policies seeking to increase LFP rates for low-skilled single mothers inadvertently led to labor force exit by young, low-skilled single males. Using data from the Current Population Survey and a bundle of work inducements enacted by states throughout the 1990s as exogenous variation in a quasi-experimental design, I find that the roughly 10 percentage point increase in LFP for low-skilled single mothers facilitated by welfare reform resulted in a statistically significant 2.8 percentage point decline in LFP for young, low-skilled single males. After conducting a series of robustness checks, I conclude that this result is driven entirely by white males, who responded to welfare-reform policies with a 3.7 percentage point decline in labor supply. Young black males, as well as other groups of potentially affected workers, appear to be uninfluenced by the labor supply response of less-educated single mothers to welfare reform. Impacts on young, single white males are large and economically significant, suggesting that nearly 150,000 males departed the formal labor market in response to directed welfare-reform policies.

Introduction

Recent media accounts have highlighted the plight of a group of low-skilled young men in the United States who are becoming increasingly disconnected and left behind by society. The aptly titled Economist article, “Men Adrift: Badly Educated Men in Rich Countries Have Not Adapted Well to Trade, Technology or Feminism” (2015), chronicles how changing economic and social conditions over the past 40 or more years have disproportionately impacted young men with low levels of education. The article notes that although high-skilled males still retain significant labor market advantages, their low-skilled male counterparts have struggled to find stable employment with wages comparable to those earned by working-class males during the post–World War II manufacturing boom in the United States. Moreover, males continue to lag behind their average female counterparts in terms of educational performance and attainment (see “The Weaker Sex” 2015), making it unlikely that this downward trajectory will reverse anytime soon without some significant public intervention. The long-term consequences of these decreased work opportunities, as well as the general underperformance by male students, have had profound impacts on topics central to the studies of demographers within low-income communities, including the declining economic stability of the working class, the decrease in the number of traditional nuclear families,1 and the increasing allure of the black market for a set of low-skilled males.

Although the primary source of these “adrift” men are individuals failing to adapt to an economy no longer producing a large number of well-paid, low-skilled manufacturing positions, this article examines how the targeting of public policies may have aggravated the situation. More specifically, I investigate how directed welfare-reform policies contributed to the decline of labor supply by young, low-skilled male workers over the 1989 to 2002 period.2 While several scholars have examined the decline in labor force participation (LFP) rates for young, less-educated men in the United States since the early 1980s (Blank 2009; Holzer and Offner 2006; Holzer et al. 2005), few studies have presented compelling causal evidence identifying distinct factors contributing to these trends.

I seek to fill this gap in the literature by exploiting the timing and characteristics of welfare reform in the United States in a quasi-experimental research design. Reforms took place throughout the 1990s as states converted their Aid to Families with Dependent Children (AFDC) programs to the new, federally mandated Temporary Aid to Needy Families (TANF) cash assistance program. After President Clinton’s vow to “end welfare as we know it,” states encouraged low-skilled single women with children to enter the labor force via work requirements, time limits, and work incentives. The results were quite striking: U.S. caseloads fell by 56.5 % from 1994 to 2000, and LFP rates for single mothers with children under the age of 18 increased from approximately 68 % in 1994 to almost 78 % in 2000 (Blank 2002).

Although social policy scholars have devoted an enormous amount of attention to the legislative changes of the 1990s, the question of whether welfare-reform policies accelerated labor supply declines by low-skilled male labor has been understudied (see also Bartik 2002; Blank and Gelbach 2006). In this article, I investigate whether a significant number of males left the legitimate labor market because of cash assistance reform policies and, moreover, whether two occurrences—the increase in labor force participation by single mothers and the decrease in labor supply for young, low-skilled single males—can be linked causally. Utilizing a fixed-effects, instrumental variable (FE-IV) research design, I show that targeted policies enacted under AFDC state waivers, the implementation of TANF, and the Earned Income Tax Credit (EITC) expansions led young, low-skilled single males to exit the labor force. To the best of my knowledge, this study is the first to show that the subsidies and work inducements embedded in reform policies targeting low-skilled single mothers unintentionally led to a decline in the labor supply of young, low-skilled single males.

Estimates of labor supply declines among young, low-skilled single males aged 16 to 29 prompted by welfare reform are large and robust across a number of specifications. When estimated for all U.S. males, two-stage least squares (2SLS) modeling reveals a roughly 2.8 percentage point decrease in LFP rates for single males for each 10 percentage point increase in LFP for single mothers. Modeling by racial group makes evident that young white males drive these statistically significant findings: estimates of the drop in LFP are larger and more precise for younger, single white males, roughly on the order of a 3.7 percentage point decline in labor supply for each 10 percentage point increase in LFP by low-skilled single mothers. Meanwhile, young black males appear to be relatively unaffected by the labor supply response of less-educated single mothers to welfare reform. This last conclusion is consistent with the work of Holzer et al. (2005), who attributed much of the decline in labor supply for this group since 1980 to increased incarceration rates and stronger child support enforcement laws. A null finding could also be reflective of sampling issues associated with modeling a hard-to-reach subpopulation.

Robustness tests indicate that the indirect effects of welfare-reform-induced declines in LFP were concentrated among young, single male workers. There is no evidence of labor supply declines for low-skilled single males aged 30 to 49, who typically have higher levels of work experience yet could have been impacted by the large influx of low-skilled single mothers. Moreover, there is no apparent response by young, low-skilled single women without children.

Although the methodology presented in this analysis cannot be used to identify the exact mechanism of these labor supply declines—such as intrahousehold labor reallocation, increases in the number of discouraged workers, or reservation wages exceeding the market price for low-skilled labor—this article presents the first robust estimates of declines in young, male labor supply stemming from welfare-reform policies. Although I build on the modeling approaches developed by Bartik (2002) and Blank and Gelbach (2006), the primary contribution of this study is in the finding that young, low-skilled single white males were particularly responsive to the market conditions created by public policies enacted during welfare reform. When the 3.7 percentage point decline in labor supply is applied to the roughly 4 million young, low-skilled single white males in the March 2002 Current Population Survey (CPS), modeling suggests a departure of approximately 150,000 men from the formal labor market. Equivalently, this loss of young male labor supply is roughly the population of a midsized U.S. city (e.g., Fort Collins, CO; Savannah, GA; or Syracuse, NY). Consequently, findings from this analysis are both statistically and economically significant.

Background and Previous Research

The 1990s were a remarkable period for the transition of individuals—mostly single mothers—from welfare to work. This decade witnessed the end of welfare programs as an entitlement, which was perceived by some as engendering a long-term, and often intergenerational, transmission of poverty for some individuals (DeParle 2004; Moffitt 1992, 2002). Instead, cash welfare assistance in the United States became a program that was time-limited; full of sanctions and requirements; and, at its core, sought to eliminate many of the disincentives of the previous program by “making work pay” (Danziger et al. 2002; DeParle 2004; Ellwood 1988). Although many aspects of welfare reform have been carefully investigated (for thorough reviews of this literature, see Blank 2002; Grogger and Karoly 2005), the possibility that government welfare-reform policies inadvertently led to the labor force exit of another vulnerable population—namely, young, low-skilled single males—has largely been ignored.

Through their efforts to encourage work and to reform a public welfare system that was widely perceived as suboptimal, federal and state government policymakers altered the nature of social assistance offered through the ADFC program. In conjunction with welfare-reform waivers from AFDC (AFDC waivers) and new stipulations under the TANF program (TANF implementation), legislators also greatly increased access to the EITC, public health insurance (i.e., Medicaid), and childcare credits throughout the 1990s. In this article, I consider all of these components to be part of the bundle that constituted reform of the public welfare system.

While increasing the scope of benefits provided by these means-tested social welfare programs, which primarily targeted low-income single-parent households, legislators also increased the level of expectations placed on recipients. Whereas AFDC was viewed as an entitlement, TANF contained several conditions for receipt of public assistance. State-level AFDC waivers and TANF programs often implemented work requirements (e.g., at a minimum, recipients must be either looking for a job or enrolled in job training), time limits on the duration of benefits received (usually a maximum of five years), and personal responsibility policies (stipulating actions such as the frequency of physician visits for children). Coupled with the greatly increased EITC, work requirements and time limits had great potential to affect the labor supply.

Both before and after the reform of the cash assistance welfare system, single mothers with educational levels at or below a high school diploma were those most likely to apply for benefits. Correspondingly, I examine how changes in their labor supply stemming from welfare reform impacted the LFP rates for single males with equivalent levels of education.3 Typically these men were categorically excluded from the package of welfare-reform programs. Conceptualizing this bundle of work incentives as a direct subsidy to single mothers, welfare-reform policies greatly fostered entrance into the labor market for low-skilled single mothers, which prompted a sharp increase in their LFP. Meanwhile, labor force entrance also tightened the labor market for low-skilled workers and reduced the relative work incentives for young, low-skilled single male workers, many of whom competed directly with these women for low-wage, entry-level positions.

The labor supply responses of young, low-skilled single males are of demographic and policymaking concern because of their several-decade decline in LFP4 and the strong statistical relationship between less-educated males and outcomes that are generally detrimental to a society. Young males who exit the legitimate labor market are at an increased risk for a number of socially undesirable outcomes, including higher probabilities of incarceration and delinquency (Blanchflower and Freeman 2000; Bloom and Haskins 2010; Harlow 2003; Pew Charitable Trusts 2010; Smeeding et al. 2011) and decreased suitability for marriage (Cherlin 2010; Edin and Kefalas 2005; Edin and Lein 1997; McLaughlin and Lichter 1997). The latter phenomenon has been linked to the decline in the traditional nuclear family within low-income communities (Carlson et al. 2004, 2013; Cherlin 2010; Edin and Kefalas 2005; Wilson 1987). This decline in the nuclear family, in turn, has been tied to a number of other social issues—for example, the decreasing role of nonresidential, less-educated fathers in their children’s child rearing and financial support (Meyer et al. 2005; Rangarajan and Gleason 1998) and the rise in the number of complex families, which often involve multipartner fertility (Cherlin 2009; Meyer et al. 2005). All these topics have important implications for those involved, especially for the development of children.

In sum, young, low-skilled males who are no longer enrolled in school and not looking for work create negative externalities not only for their families and local neighborhoods but also for the broader society.5 In addition, this general withdrawal from educational programs and labor market activities also leads to internalities: the lack of human-capital development during a critical early period of adulthood can lead to prolonged, higher levels of social isolation and lost potential for an at-risk group of young males.

Although limited, two notable studies have investigated the possibility of adverse effects from welfare reform on the labor force participation rates of low-skilled males. First, Blank and Gelbach (2006) used data from the CPS and a variety of empirical tests to examine the substitutability of less-educated males and females within the low-skilled labor market. After a series of empirical tests, they did not find consistent evidence of male labor supply crowd-out from welfare reform. This article builds on their work in two ways. Rather than using a single variable for welfare-reform policy to examine the impact of welfare reform, I examine a set of instruments that capture the substantial heterogeneity in AFDC waiver and TANF programs across states. I use these multiple instruments to uncover the exogenous increase in female labor supply stemming from welfare reform. Through an FE-IV research design with this more expansive set of instruments, I improve on the identification limitations inherent in Blank and Gelbach’s original analysis, given that this FE-IV approach better exploits the variation in, and timing of, the work incentives influencing low-skilled workers across the different state plans. This methodology increases the precision of the estimates and allows me to test the strength and validity of my instruments.

Second, Bartik (2002) used instrumental variables to control for welfare caseloads across states and estimates the labor market spillover effects of welfare reform. Using simulation, Bartik estimated the displacement and wage elasticities for less-educated workers given the increase in supply of less-educated workers, under the assumption that welfare reform led to an influx of 1.4 million individuals into the low-skilled labor market. With his 2SLS modeling, Bartik concluded that welfare reforms, which led directly to decreased caseloads, may have led to employment losses for less-educated males and reduced the wages for single mothers and male high school dropouts. Whereas Bartik’s modeling concentrated on employment rates and wages across all ages of workers, I focus on estimating the impact of welfare reform on the labor supply decision of young, less-educated males. That is, this study extends Bartik’s analysis by concentrating on the specific subset of low-skilled male workers who were most likely to be affected by welfare reform and by focusing on the effects of reform on labor supply.

Theoretical Framework: Welfare Reform and Labor Supply

Two of the explicit political goals of welfare reform were to end welfare as an entitlement and to make work pay for low-income individuals (Blank 2002; DeParle 2004). In this section, I describe why labor supply by young, single males should not be directly affected by welfare policy changes, which is critical in establishing a valid IV approach. In addition, I outline three indirect mechanisms stemming from welfare-reform policies that could have prompted labor force exit by young, low-skilled males: (1) intrahousehold labor reallocation, (2) a drop in the market price for low-skilled labor below the reservation wage for many young men, and (3) an increase in the number of discouraged workers as a result of a tightened labor market and an “envy story” rooted in behavioral economics.

Young, low-skilled single males are rarely the overt target of social welfare policies in the United States. Welfare reform was no exception. To provide a baseline perspective of program targeting before the transition from AFDC to TANF in the 1990s, only about 10 % of welfare recipients in 1990 were male and received benefits as either a head of family or a single parent. During the period of welfare-reform transitions, roughly 9 % of all less-educated female workers reported receiving benefits in 1996, while only 1 % of less-educated men did (Blank and Gelbach 2006). Thus, males—especially young single men—should have been categorically ineligible for government benefits both before and after welfare reform. Furthermore, while a subset of unmarried young, low-income males may be part of an informal household choosing to reallocate labor in response to market incentives, the change in female labor supply within these loosely structured family units is a mediating variable—that is, males formulate their labor supply decisions based on the choices of their cohabiting partner and not the government’s policies.

Welfare reform is a mediating variable beyond intrahousehold labor market reallocation. Two other probable mechanisms are also founded on the assumption that welfare reform did not affect males directly. Instead, labor supply decisions by males were first influenced by the behavioral responses of low-skilled single women with children who would have qualified for government benefits under the old AFDC entitlement system. The direction of the causal relationship for labor supply responses to welfare reform for the subset of individuals explored in this analysis is shown in Fig. 1.

Note that for this IV approach to be valid, I am explicitly claiming that there was no direct impact of welfare reform on the labor supply by young, low-skilled single males. Rather, impacts were channeled through, or meditated by, the direct response by women who potentially qualified for welfare benefits. The remainder of this section will describe these three potential mechanisms in greater detail.

The first mechanism, intrahousehold labor market reallocation, can be illustrated by considering the set of factors influencing labor supply decisions by less-educated workers. State-level public policies enacted during welfare reform sought to increase the labor supply of single mothers. Some methods, such as time limits, family caps, and sanctions, arguably more indirectly influenced supply. Others, such as the EITC, childcare subsidies, and access to public health care, could be seen as direct subsidies to working individuals with children under age 18. As reforms progressed, work requirements became an explicit part of the cash-assistance welfare program. Thus, benefit receipt became conditional on LFP, and individuals were expected to seek work to retain their monthly cash transfers from the government, as well as the in-kind transfers. Considering this package of incentives seeking to increase the labor supply in its totality, post-welfare LFP decisions by low-skilled single mothers were partly determined by (1) potential income from work, (2) welfare cash benefits, (3) available EITC, (4) childcare credits, and (5) public health insurance.

Far fewer policy incentives affected labor supply decisions by young, low-skilled single males. The primary motivating factor for LFP is income generated from the low-skilled wage, although a small EITC would be available in some years. Given that other considerations are likely to affect both groups of workers equally, the targeting of these public policies resulted in a substantial number of single mothers receiving benefits from LFP in excess of what young males competing for the same types of low-skilled, entry-level jobs would receive. For an informal household choosing to maximize income and benefits based on their option set, a male could exit the formal labor market and concentrate on domestic duties while the corresponding female participates in the labor market, thereby producing an intrahousehold reallocation of market labor.

The second and third mechanisms are, respectively, reservation wage and discouraged worker stories. These two cases would most likely apply to a broader set of young, low-skilled single males who were not cohabitating with a welfare-eligible single mother. If real wages declined for low-skilled workers because of the influx of labor supply induced by welfare reform, as Bartik (2002) noted, then government transfers and subsidies would become more important factors in labor supply (i.e., the work vs. no work decision). Young, low-skilled males would be disproportionately affected by these targeted supports. Under this scenario, the exit of low-skilled male labor facilitated by welfare reform could stem from either reservation wages or an increase in the number of discouraged workers. Under the former, the influx of low-skilled female labor would drive down the effective market wage for overlapping industries6 and encourage lower-skilled males with higher reservation wages to exit the market, especially if these individuals perceived that they had better-paying options in the informal or black market.

The second scenario occurs when the entrance of low-skilled single mothers, who have more incentives for work and thus may be better-motivated employees, leads to a disproportionate hiring of and retaining of these workers. In turn, these changes in hiring behavior and work environment could lead some unemployed young males to become discouraged and exit the formal labor force. Similarly, one could easily envision an “envy story” routed in behavioral economics whereby young males in overlapping industries learn about these target supports. Equipped with this information, they would realize that their effective wages are much lower than those of their female counterparts because of the bundle of goods offered by the government. Again, discouraged young males might be more inclined to participate in informal or black markets in order to receive better effective wages.

Unfortunately, data and methodology used in this analysis cannot distinguish between these purported explanations. Future work seeks to disentangle these mechanisms. However, understanding the potential scenarios prompting behavioral responses by young, low-skilled single males is essential for establishing the validity of the FE-IV approach outlined in the next section.

Empirical Strategy

To isolate the causal impact of increased labor supply by single mothers stemming from welfare reform on young, low-skilled male labor, I use a fixed-effect, instrumental variable (FE-IV) approach. Under this two-staged design, the dependent variable in the outcome equation is the LFP rate for a given group of males (e.g., low-skilled single males aged 16–29) at a given point in time (e.g., 1996, first quarter). A welfare-reform-induced reduction in the labor supply of young, single males would be observed when an exogenous increase in the LFP rates for females leads to a decrease in the labor supply for males.7 A strength of this FE-IV research design is that it controls for time-invariant factors affecting the male and female labor supply across both states and time; it also isolates the indirect effect of an exogenous change in female labor supply increase resulting from welfare reform as channeled through a single mediating variable.

The general estimation strategy is as follows:
LFPmalessyq=α+βLFPfemalessyq+γXsyq+δs+ζy+ψq+ωyq+εsyq,
1
where an observation is defined by a state (s), year (y), and quarter (q); X is a vector of control variables containing measures of economic growth, state minimum wage rates, child support enforcement strictness, wages for low-skilled males, male-to-female sex ratios, and male incarceration rates; δ, ζ, ψ, and ω are fixed effects for state, year, quarter, and year-quarter, respectively; and ε is the robust standard error.

In this specification, the state, year, quarter, and year-quarter fixed effects are used to detrend the labor supply variables across both states and time.8 All regressions are weighted by the corresponding number of males residing in the state (s) in a particular year (y) and quarter (q).

As specified above, LFPfemales is most likely endogenous because LFP rates for men and women are influenced by many of the same economic factors. Failure to account for simultaneously determined LFP rates in an ordinary least squares (OLS) specification produces biased coefficients on the primary variable of interest—that is, ∂LFPmales / ∂LFPfemales. This is represented by the coefficient β in Eq. (1). To derive unbiased coefficients, I require a set of instrumental variables for LFPfemales, which will allow me to use the exogenous portion of the increase in female LFP stemming from welfare reform to estimate impacts on the labor supply of young, low-skilled single males.

Under the classic definition of an instrumental variable, I need variables that are uncorrelated with LFPmales conditional on other controls within the model (i.e., the exclusion restriction) but that explain changes in LFPfemales (i.e., the relevance criterion). The welfare-reform policies of the 1990s conveniently serve this purpose: legislation and new programs provided a series of incentives for welfare recipients—most of whom were single mothers—to enter the labor force. At the same time, these policies were designed to have little or no direct impact on young, single males simply because they were categorically ineligible for these programs. Consequently, the attributes, timing, and variation of welfare-reform policies can be used as identifying instruments for changes in female LFP rates. In addition, this approach can be used to test the relevance and exogeneity of the instrumental variables. Evidence of exogeneity will mitigate concerns over simultaneity in the second-stage equation.

The first-stage equation, which models LFP rates for low-skilled single mothers aged 16–44, is as follows:
LFPfemalessyq=ω+χXsyq+ρΖsyq+φWelfareReformsyq+τMaximumCashBenefitsMaximumEITCsyq+ηWelfareReform×CashBenefitssyq+κs+υy+ςq+ιyq+μsyq,
2
where X is the same set of controls from Eq. (1); Z is a vector of attributes of state-level AFDC waivers which are in effect in year y and in quarter q; Welfare Reform is an indicator that is turned on after the first waiver implementation or the effective TANF program date; Maximum Cash Benefits are the maximum state AFDC or TANF cash benefits for a family of three at a particular point in time; Maximum EITC is the maximum state and federal earned income tax credit that could be earned in a state and year; Welfare Reform × Cash Benefits allows for a change in LFP incentives when cash benefits become directly linked to labor supply; κ, ν, ς, and ι are vectors of fixed effects for states, years, quarters, and year-quarters, respectively; and μ is the first-stage error term.

In Eq. (2), the Z vector contains the state-level characteristics of AFDC waiver programs, such as the presence of work requirements, time limits, work incentives, and personal responsibility policies.9 Because welfare reform is defined as a single indicator variable, it measures the average impact of waiver and/or TANF implementation, after the specific elements of these programs are accounted for.

The EITC is now the largest cash transfer program in the United States (Moffitt 2007) and has become a vital part of the labor supply decision for low-income workers. Correspondingly, the Maximum Cash Benefits / Maximum EITC variable captures the implicit tradeoff between cash transfers from no work (maximum AFDC / TANF cash benefits) and work (maximum state and federal EITC), a tension that would not be controlled for by the other welfare-reform variables.10 Online Resource 1 shows this ratio, which has decreased dramatically over time in all states. Note that a large ratio of cash benefit to EITC reveals that the gains to work are low if one is solely seeking to maximize the amount of transfers received from the government and bases the work versus no work decision on the amount of cash transfers received from the government. A higher ratio should foster lower LFP rates for qualifying individuals, and a ratio below 1.00 would indicate that a family of three can earn more money in transfers via the EITC than through cash benefits and no work. Thus, I expect the coefficient on this variable (τ) to be negative.

Another important explanatory variable is the interaction between cash benefits and the timing of welfare reform. This coefficient captures the very different behavioral responses elicited by welfare reform. In a pre-reform period of welfare entitlement, large cash benefits would be considered a work disincentive that would decrease LFP rates for low-skilled single mothers. Post-reform, however, work requirements are linked to aid receipt, making work essential to retaining benefits, offering single mothers great incentives to enter the labor force. Inclusion of this interaction captures the complex, changing relationship between cash benefits and LFP. As specified, I expect η to be positive.

Finally, under the fixed-effects framework used in this article, federal policies such as the Child Tax Credit cannot be uniquely identified because they lack cross-sectional variation at a given point in time. This is also true for the individual impacts of TANF programs after all states have adopted their new systems. Identification cannot occur because the year and year-quarter fixed effects subsume all variation for variables that are the same for all states in a particular period.

Data Sources and Descriptive Statistics

Measures of labor supply used in this analysis are constructed from monthly CPS data.11 Within its nationally representative, rotating sample design, the CPS contains detailed demographic and work history information for individuals residing in more than 50,000 housing units across the United States. Despite this seemingly broad scope, the number of observations in some subgroups (e.g., black men aged 16–29 living in Montana) can be extremely limited. To address this sampling issue, I aggregate data to the quarterly level to increase the precision of the estimates.

Two other important issues regarding the LFP rates used in his analysis are of note. First, as is common in the literature, I omit all individuals younger than age 25 who are enrolled in school or university full-time (see Holzer et al. 2005; Holzer and Offner 2006). Second, because the CPS excludes institutionalized individuals from its sample, these people cannot be part of LFP estimates. Thus, to the extent that incarceration rates are increasing over time within a given state and subpopulation, the LFP rates derived are an upper bound because those most prone to criminality are presumably less likely than the average low-skilled male to engage in the formal labor market. Use of this upper-bound LFP measure means that measured impacts should actually serve as a lower-bound estimate of the true effects: the observed decline in LFP would not be as pronounced as it would be when otherwise including the incarcerated. Given that incarceration could play a large underlying role in this analysis, I use two measures to account for this increase in institutionalization, as I discuss shortly.

Data range from 1989 to 2002, yielding approximately five years of data before the large increase in the maximum EITC and the initial mass of states implementing waivers in 1994, as well as roughly five years of data for the period after the last state implemented its TANF program.12 Analogous to the CPS sampling issues, quarterly data more precisely control for the timing of the AFDC waiver or the state’s TANF program implementation. Finally, it is important to reiterate that my analysis is based on individuals with a completed educational level of high school diploma or less.

Figure 2 displays the LFP rates for single males aged 16–29 and 30–49, and single mothers aged 16–44. These categories are used in the empirical modeling and were selected for the following reasons: 16–29 captures the range for younger workers who have relatively fewer years of work experience and who could be competing with single mothers for low-skilled or entry-level positions. Moreover, the age range of 16–29 facilitates more precise estimation for subgroups in smaller states, while the age range of 30–49 isolates workers with longer work histories and potentially different marginal propensities to seek informal job opportunities. For females, I examine mothers aged 16–44 with educational levels at or below a high school diploma. Although welfare qualification and participation cannot be identified in the monthly CPS data, women of childbearing age who have low levels of education and children younger than age 18 are those most likely to apply for welfare benefits and receive them.13

Figure 2 shows the aggregated trends for the United States.14 Broadly speaking, the United States witnessed relatively flat LFP rates for young, less-educated males in the earlier period, followed by a slight decline in LFP during welfare reform and then another stabilization. Although these trends foreshadow, at an aggregate level, the potential for only a slight decline in labor supply for young males stemming from welfare-reform policies, I exploit state-level trends to derive more precise estimates of the unintended consequences of the targeted policies enacted during welfare reform. Furthermore, for older, less-educated single males, rates were fairly consistent over time (hovering around 82 %), whereas the LFP rates for single mothers across the United States increased dramatically from 1994 to 1999, as reported by Blank (2002).

Figure 3 shows trends in LFP rates by race that are similar to the aggregated trends presented in Fig. 2. Again, the potential for labor supply declines appears to be relatively small when examined at the U.S. level. However, this aggregation masks significant potential heterogeneity in response to the state-level policies, which is critical for identification in this study. Stated another way, a negative labor supply response can be identified when states experiencing the largest increases in LFP by low-skilled single mothers also witness the greatest decline in labor supply by young, low-skilled single males. The forthcoming empirical models will explicitly test this proposition.

Before proceeding to estimation, it is necessary to outline the instrumental and control variables used in this analysis, which are outlined in Table 1.15 First, the implementation dates of the AFDC waivers and TANF programs were obtained from the U.S. Department of Health and Human Services (1999); the implementation dates of these waivers and the new state-level cash assistance programs showed a wide degree of heterogeneity. Using information provided by Ziliak et al. (2000), I control for AFDC waiver characteristics, such as work requirements, time limits, work incentives, and responsibility clauses.16 As with implementation, these reform efforts showed a wide range of within- and across-state variation. Finally, the maximum cash benefits for a family of three under the state-level AFDC and TANF programs and the maximum EITC—which includes both state and federal benefits—were taken from a comprehensive database complied by the University of Kentucky Center for Poverty Research (2015). Like all other monetary variables in this analysis, these data are standardized to 2002 dollars.

Other state-level control variables used in the FE-IV models include (1) the percentage growth in gross state product, (2) the state minimum wage, (3) a child support enforcement (CSE) index, (4) a lagged measure of the average weekly earnings for low-skilled males aged 16–29 who were working full-time, (5) the male-to-female sex ratio, and (6) incarceration rates by race/ethnicity. The first two covariates account for factors impacting labor market conditions for low-skilled workers: the percentage growth in gross state product controls for general economic conditions at the state level, which were generally stronger during the latter half of the 1990s; and a panel data analysis using state-level minimum wages will account for legislated increases in generosity for workers earning at the lowest levels of the wage distribution.

The third variable—the CSE index—is included to capture the differential labor supply responses by low-income parents not residing within the same household. It is also required because CSE policies are correlated with welfare reform. Although stricter CSE may increase household income and make it less likely for single mothers to seek employment, it may also drive many low-income males away from the formal labor market and toward informal opportunities because of the relatively high marginal tax rates (Holzer et al. 2005). To capture this complex relationship, I extended the CSE index developed by Huang et al. (2002). This variable ranges from 0 to 8, with a higher number indicating the presence of more state programs to enforce child support payments.

Average weekly earnings for low-skilled males aged 16–29 who were working full-time addresses potentially omitted factors contributing to the well-documented decline in wages for this particular subset of workers. I estimate low-skilled wages by state using the March CPS earnings data. I use two techniques to account for simultaneity and sampling issues in the construction of this variable. First, because of the concern that LFP rates and wages are jointly determined, earnings data are lagged by two years. Second, to avoid identifying estimates based on CPS sampling variation, I smooth estimates using a four-year moving average. These adjustments create a wage measure that can capture important factors affecting male LFP rates, which may also be correlated with the timing of welfare-reform policies.

The last two variables address the increase in mass incarceration rates in the United States. More specifically, the male-to-female sex ratio responds to growing concerns over the number of “missing males” within select communities in the United States (see Wolfers et al. 2015).17 The sex ratio is constructed as the total number of less-educated males in a particular age category (e.g., 16–29 or 30–49) divided by a biologically equivalent number of low-skilled women within the same group analyzed (e.g., all, black, or white).18 A ratio significantly above or below 1 indicates that the biological sex ratio with a state/group has become unbalanced. An increasing sex ratio imbalance may systematically affect labor supply decisions by low-skilled men and women; this covariate accounts for this variation.

Finally, similar to Holzer et al. (2005), I construct incarceration rates by race/ethnicity using National Prisoner Statistics data from the Bureau of Justice Statistics and population data from the Surveillance, Epidemiology, and End Results Program. Estimates represent the fraction of the adult population incarcerated at a particular point in time within a given state. This value is lagged by three years to reflect the fact that the average sentence during this period was roughly three years (Holzer et al. 2005). With this lagged structure, this variable captures the reentrance of former prisoners into a local economy.

Table 2 presents the unweighted summary statistics for all variables used in the empirical models. The average LFP rate is 0.852 for all young, single males (aged 16–29) and approximately 0.823 for older single males (30–49) and both are estimated with a relatively wide range of roughly 0.5 to 1. The “Min.” and “Max.” columns show some LFP rates of 0 or 1. Again, this reflects measurement error implicit within the CPS sampling design, which does not always reach select subpopulations. To further address this sampling limitation, I weight regressions by the number of corresponding males in each category in each quarter. Bias from mismeasurement is attenuated under the assumption of classical measurement error.

Note that the unweighted average LFP rate for single mothers is 0.695. This labor supply measure is much lower than the rates for single males; however, graphical analysis shows that it increases markedly over time. Additionally, a little more than half (54 %) of the observations fall in the implementation period for either the post–AFDC waiver or TANF for a given state; this is indicated by the average of the Welfare Reform (waiver or TANF) dummy variable. Summary statistics for the AFDC waiver attributes can be interpreted similarly: the mean indicates the total fraction of the unweighted sample that is affected by that variable.

Empirical Findings

To explicate the findings from my empirical modeling, I start by outlining the results from one first-stage model in Table 3. The first-stage models are the key to any causal links among welfare reform, LFP among low-skilled single mothers, and labor supply of young, low-skilled males. For the sake of clarity, I offer brief commentary on the reported findings for a solitary case: all single males. In this discussion, it is important to recall the focus of this study: the coefficient on the impact of the plausibly exogenous increase in LFP rates for low-skilled single mothers on the labor supply decisions of single, less-educated males. Given this focus, and the fact that the marginal effects are often derived from several variables, I do not interpret individual coefficients and instead comment on groups of covariates.

In Table 3, the middle column for each set of regression models contains the coefficients for the first-stage models, which estimate LFP rates for low-skilled single mothers aged 16–44. As indicated by the first-stage coefficients for the all single male models, a large portion of the within-state variation in labor supply by low-skilled single mothers can be explained by the first-stage models. This suggests that the included explanatory variables and instruments account for much of the increase in LFP rates by low-skilled single mothers depicted in Fig. 2. More specifically, policies implemented under the state-level waiver programs appear to explain a significant share of the within-state variation in female LFP during their period of enactment, with all else being equal; these are the identifying instruments in the 2SLS analysis. In particular, the trio of Welfare Reform (waiver or TANF), Maximum Cash Benefits / Maximum EITC, and Cash Benefits × Welfare Reform are all highly statistically significant and add substantial explanatory power to the model.

In terms of the strength and exogeneity of the instruments proposed in this analysis, the F statistic for the identifying instruments is 13.23, which is above the empiricist minimum of 10 required to pass the weak instruments test (Angrist and Pischke 2009). Furthermore, I can formally test the exogeneity of the instruments because my model is overidentified. As shown, the Hansen J statistic indicates that the null hypothesis that the identifying instruments are exogenous cannot be rejected at the 5 % level of statistical significance.19 Thus, the two crucial components required to conduct an instrumental variable analysis—the relevance criterion and the exclusion restrictions—are met in this analysis.

Turning to the other models in Table 3, the mediated impact of welfare reform on the LFP rates for less-educated males aged 16–29 presented for three groups of men: (1) all single males (regardless of race/ethnicity), (2) black single males, and (3) white single males. I begin with the single-equation OLS coefficient, which indicates a positive and significant relationship between LFP rates for single mothers (aged 16–44) and all single males (aged 16–29). As noted, this estimate is biased. Two-stage modeling shown in the third column reveals a negative and statistically significant relationship. The coefficient of –0.281 can be interpreted as follows: an exogenous 10 percentage point increase in LFP of low-skilled single mothers prompted by welfare-reform policies led to an approximately 2.8 percentage point decline in labor supply by young, low-skilled single male laborers. Note that the sign of the estimated relationship has changed from positive to negative when going from the OLS to the FE-IV modeling. This occurs because the IV strategy can isolate the exogenous portion of the labor supply increase that is directly attributable to welfare-reform policies. By removing a significant portion of the bias inherent within OLS estimation, the FE-IV approach can mitigate the endogeneity concerns highlighted earlier.

Examining the LFP rates for young, low-skilled black males reveals a different picture. Although the models also pass the weak instruments and exogeneity tests, the estimated effect of increases in labor supply by single mothers is statistically indistinguishable from zero. Although somewhat surprising, these findings are in line with those of Holzer et al. (2005), who attributed the decline in LFP for low-skilled black males to factors other than welfare reform and increases in low-skilled female labor supply. More specifically, Holzer and colleagues found that CSE and incarceration rates drive a large portion of the drop in the labor supply of black men aged 16–34. The research design and data panel length used in this analysis were not aimed at reproducing these findings,20 but it is important to note that black males within the period of welfare reform appear to have been unresponsive to the influx of labor supplied by low-skilled single mothers.

The last group—young, single white males—is the group that is driving the findings of labor force exit for low-skilled men. As displayed, the estimated relationship between labor supply by single mothers and young, single white males is negative and statistically significant in the 2SLS modeling. Moreover, the point estimate indicates that a 10 percentage point increase in LFP rates by low-skilled single mothers led to approximately a 3.7 percentage point decline in LFP rates by young, low-skilled single males, holding all else equal. These values are both highly statistically and economically significant: this decline over a base LFP rate for white males of 87.6 percentage points can also be interpreted as a 4.2 % decline in labor supply. In terms of the roughly 4 million young, low-skilled, single white males aged 16–29 in the March 2002 CPS, this represents a decline in supply of approximately 150,000 young men—a number that is roughly the same size as the entire population of a midsized U.S. city.

Robustness Checks

In this section, I present three robustness checks to test the sensitivity of my findings. The first set of models introduces state-level time trends to the core models previously reported in Table 3. These specifications seek to account for other time-variant omitted factors within a state that could either increase or decrease the propensity for LFP by low-skilled female and male workers. Table 4 contains the findings from this exercise. As displayed, the introduction of time trends diminishes the statistical significance of estimated decline in LFP rates for all single males aged 16–29. Although the point estimates are still negative, the coefficients are not statistically significant at conventional levels. In addition, the F statistic on excluded instruments is now well below the empiricist minimum of 10 required to mitigate the weak instruments critique. Conclusions for the two subgroups of young males are similar to those reported in Table 3. I find no evidence of labor supply changes for blacks but do find statistically and economically significant labor force exit for young white males. Under this robustness check, the magnitude of the findings is very similar to that of the previous estimates: I find an approximately 3.4 percentage point decline in labor supply for each 10 percentage point increase in LFP for single mothers.

Although not strongly supportive at the aggregated level, the results from the models with time trends provide additional evidence for causal claims for the group that appears to be driving the finding of labor force exit: young, single white males. However, modeling with state-specific time trends in this analysis, and the resulting weak first stage, can be criticized on three grounds. First, the majority of the data used to calculate the time trend resides in the post-reform period, which distorts the inferences and value of establishing a pre-reform trend because it is based primarily on post-reform data. Second, because the margin on which labor supply decisions are changing is relatively small, the use of a time trend does not leave much variation to be explained by other mechanisms, especially when considering that the model already contains state, year, quarter, and year-quarter fixed effects. Finally, data at the aggregate level (recall Fig. 2) indicate a rather quick transition of LFP rates for young males during the welfare-reform period, followed by stabilization. Time trends are an overly blunt instrument, absorbing much of the variation in LFP that should actually be attributable to welfare-reform policies. Again, this is particularly true given the other fixed effects used in the modeling. In light of these criticisms, coefficients derived from core models are more plausible and provide a more precise measurement of labor force exit stemming from an exogenous shift in low-skilled female labor supply.

The second set of robustness checks examines whether a similar subset of men could have been affected by large increases in labor supply of low-skilled single mothers. As presented in Table 5, low-skilled single males aged 30–49 do not appear to be negatively impacted by the labor supply increases of single mothers aged 16–44. I will eschew a detailed analysis of these findings and instead propose potential explanations for the difference in patterns between the younger and older low-skilled male workers. Younger males may be more responsive to market conditions because they have a shorter work history, have fewer labor force attachments, and are more likely to engage in criminal activity (Freeman 2000; Levitt 2001). Thus, they would be more influenced by changes in female labor supply. Many older males are ostensibly in the legitimate labor force, a tendency that changed very little during the welfare-reform years. In addition, these older men may be competing for positions that are unaffected by the influx of low-skilled female labor. Given these factors, the labor supply decisions of older males appear to operate at a different margin—one that is not negatively impacted by the increase in labor supply of low-skilled women stemming from welfare reform.

Finally, Table 6 contains a third set of robustness checks. Here, I apply the FE-IV strategy used in this study to two overlapping childless groups that should not have been directly impacted by welfare reform under the theoretical framework established earlier: young, low-skilled single females aged 16–29 who are without children and low-skilled single women aged 16–44 who are also without children. This examination of the responsiveness of single, childless women of each age group, using the changes in labor supply of single mothers aged 16–44 as the first-stage instrumented variable (i.e., equivalent to the second-stage model reported in Table 3), reveals no evidence of labor force exit stemming from welfare reform.

Results from these three sets of robustness checks present strong evidence that the indirect effects of welfare reform were targeted specifically on low-skilled male workers. This important differentiation reveals that young, low-skilled males may be more sensitive to monetary incentives offered by the marketplace, especially if they have more lucrative employment opportunities in the informal or black market.

Discussion and Conclusions

This research seeks to identify a more recent casual factor contributing to the decline in labor supply by young, low-skilled males, an issue that has confounded some scholars (Blank 2009; Holzer et al. 2005). In this article, I present evidence that the roughly 10 percentage point increase in labor supply of low-skilled single mothers prompted by the welfare-reform policies of the 1990s unintentionally led to a 2.8 percentage point decline in the LFP rates for all young, low-skilled single male workers. These negative impacts appear to be concentrated solely on white males, who experienced a 3.7 percentage point decline in labor supply. When this latter estimate is applied to the roughly 4 million young, low-skilled males identified in the March 2002 CPS, the scope of this decline corresponds to an exit of approximately 150,000 young males from the legitimate labor force.

Both policymakers and demographers should be concerned about the decline in young males engaging in the conventional labor market. Less-educated males who are not enrolled in school or working/seeking work are essentially withdrawing from traditional society, making them much more likely to engage in criminal activities and less likely to form stable, lasting relationships with their partners. These byproducts have contributed greatly to the long-term increases in incarceration rates in the United States, as well as the decline of stable, nuclear working-class families. Although other long-run factors, such as incarceration rates and child support, appear to contribute more significantly to the decline in labor supply by young black males, this research clearly demonstrates that young, low-skilled white males are responsive to work incentives that are embedded in targeted public policies. Young, single white males, however, responded negatively to the welfare-reform supports from which they were excluded.

Although findings presented in this article are robust across a number of alternative specifications, three limitations to this work temper the claims for a truly causal relationship between the welfare-induced increases in labor supply of low-skilled single mothers and the results presented. First, the potential for type II errors, or reporting of false negatives, could originate from sampling issues associated with modeling LFP rates for young, low-skilled black males. For example, the impact on LFP rates for these males may be too small to be statistically distinguished from zero using the methodology presented in this article. A null finding may stem from the increased institutionalization rates and the CPS sampling issues discussed earlier: noisy estimates of LFP rates would most likely lead to larger standard errors and attenuation bias in estimation. Similarly, although arguably the best tool available, the sex ratios used to address the issue of “missing” black men are limited. Most notably, the measure is a state-level aggregate. This approach could mask interesting regional or census track-level variation explaining changes in LFP rates of the low-skilled men and women central to this analysis. Unfortunately, given data limitations and the unit of analysis in the empirical modeling, there is no obvious way to further address these concerns.

Second, the FE-IV strategy cannot directly control for compositional changes in population over time. The range of data used in this analysis is fairly long: 1989–2002. Bias could enter the modeling if the subpopulation of interest at the end of the period is drastically different than the initial one and this variation is correlated with the timing of welfare-reform policies. For example, suppose that the average white male in State A with education at or below a high school diploma is markedly less employable in 2002 than in 1989. If this decrease in “employability” is consistent across other states (B, C, and so on), then the year and year-quarter fixed effects will control for this time-variant change in employment suitability. On the other hand, if the decrease in employability is both time- and region-variant and is correlated with reform, then this cannot be captured by a FE-IV model. Presumably, this scenario would lead to an overestimation of the decline in young, single male labor supply that is attributable to the labor supply changes prompted by welfare reform; in this case, estimates presented in this article would serve as an upper bound of the true impact.

Finally, other potentially omitted time-varying factors, such as a large increase in young, low-skilled individuals receiving disability insurance (DI) or supplemental security income (SSI) benefits are problematic if changes in the generosity of these programs and program participation are correlated with the timing of welfare-reform policies. The inclusion of estimates of the number of DI or SSI recipients by state and year initially appears to be a reasonable addition to the models. However, it is an outcome variable, thus meeting the definition of a “bad control” (Angrist and Pischke 2009) and likely being endogenous.21 To properly address this shortcoming, one needs to control for the time-variant behavioral factors that lead to DI or SSI claims and that are correlated with the instruments. In other words, one should not simply control for the outcome that can be jointly determined with LFP. To the best of my knowledge, these data are not available.

Those shortcomings aside, this article presents evidence that young, low-skilled single males are responsive to government policies and that there was another serious unintended consequence of the large set of welfare-reform policies enacted during the 1990s: the exacerbation of labor force exit by young, low-skilled single males. Relevant to policymaking, this article examines how resources concentrated on one particular set of disadvantaged individuals (e.g., low-skilled single mothers) adversely affected the behavior of another group (e.g., young, low-skilled males). This latter group has struggled to incorporate itself into a changing American economy and is currently linked to a host of social ills, such as black-market LFP, increased incarceration rates, and declines in low-income nuclear families (and all the other associated issues). Thus, to the extent that policymakers believe that the government has a role to play in steering young, disconnected males toward more socially desirable outcomes and choose to include these young men among the “deserving poor” (Moffitt 2015), this research supports the calls by many scholars to increase work incentives to other segments of low-wage workers. Policy levers such as extensions of the EITC to childless adults and noncustodial parents, as discussed by Blank (2009), Smeeding et al. (2011), and Pew Charitable Trusts (2010), should be given more consideration in an attempt to reverse many of the negative trends associated with this set of workers. If males respond negatively to exclusion from targeted public policies seeking to increase LFP rates, they may react positively if included in targeted supports for low-skilled workers.

Acknowledgments

PhD program funding from Syracuse University greatly supported this research. The author gratefully acknowledges Leonard M. Lopoo for his patience, helpful comments and suggestions, and encouragement throughout the various iterations of this work. Invaluable input was received from a number of other faculty at the Maxwell School, including Sarah Hamersma, Jeffery D. Kubik, Douglas Wolf, Robert Bifulco, David Popp, and Leonard Burman. Finally, the author would like to thank the editorial staff at Demography, as well as a number of anonymous referees, for their valuable feedback during the review process.

Notes

1

In particular, the decline of married, two-biological-parent households has been tied to a host of other issues, including the reduced role of low-skilled men in the family life of their biological children, the increase in the number of complex families, and the various challenges and impacts associated with child support payments.

2

In this analysis, low-skilled workers are defined as those individuals with an educational level of high school diploma or less. Furthermore, the terms “low-skilled” and “less-educated” will be used interchangeably and “high school diplomas” include both traditional and general equivalency diplomas.

3

As is common in the literature, in this article, I concentrate on LFP rather than employment. LFP is arguably a more accurate depiction of labor supply because it captures the intent to provide labor. Employment, on the other hand, can be based on a number of factors outside the individual’s control, especially the demand for labor.

4

The general decline in LFP rates for low-skilled males since the early 1980s has been documented very thoroughly by Holzer and various colleagues. In Holzer and Offner (2006), they report declines in labor supply for young, less-educated white males from approximately 92 % in 1979 to roughly 87 % in 2000. Correspondingly, rates for black males have dropped from roughly 82 % to 70 % over the same time period. In Holzer et al. (2005), the authors report that at least half of the decline in employment among less-educated black males can be attributed to increases in incarceration rates and stronger child support enforcement laws.

5

Recent estimates of the scale of the illicit drug trade in the United States indicate that it is a highly lucrative industry which is estimated to produce up to $150 billion in revenue each year (Bagley 2012; United Nations Office on Drugs and Crime 2012). Given this scope, it does not seem unreasonable to contend that low-skilled males are more likely than other groups to enter these illegal professions given that they have fewer employment options.

6

Scholars have noted their skepticism regarding whether low-skilled men and women compete in the same labor markets (Blank 2002; Blank and Gelbach 2006). However, there is seemingly enough overlap in some low-skilled sectors, such as fast-food services, custodial services, and security and retail jobs, for this supposition. For example, Card and Krueger (1994) claimed that fast-food franchises are a leading employer of low-wage workers, and low-skilled workers of either gender seem equally qualified for these entry-level positions.

7

A positive relationship could indicate peer effects, whereby the welfare-reform work inducements create positive spillovers in the form of increased LFP for applicable males residing within the household or the community.

8

The state dummy variables account for time-invariant unobserved factors that influence historical LFP rates in a particular state. The year, quarter, and year-quarter fixed effects control for omitted factors that impact labor supply rates in all states during a particular period. In addition, the quarterly variables account for seasonality in the LFP rates for young males. The CPS considers university-bound males on summer break (i.e., those between their last year of high school and first year of college) as potential labor force participants. During the summer months, this influx of short-term labor drives down the LFP rate for single males aged 16–29.

9

Other instruments were considered in this analysis but were excluded because of their weak predictive power.

10

In some states this ratio changed well before the implementation of either an AFDC waiver or their state-level TANF program. For example, New York implemented its TANF program in November 1997, but the LFP rates for single mothers in that state increased markedly before this point—presumably because of the large increases in the federal EITC beginning in 1994.

11

The vast majority of CPS data used in this analysis come from the IPUMS-CPS database (Flood et al. 2015).

12

These periods are identified with vertical lines in the forthcoming figures.

13

Upcoming modeling is not sensitive to the choice of using single mothers aged 16–44. Models using single mothers aged 16–30 produce very similar estimates.

14

As previously mentioned, it is important to control for the seasonality of LFP for young, single males given the influx of university-bound males during the summer months. Thus, the graph is seasonally adjusted.

15

I also analyzed a number of other characteristics of welfare reform, but did not include them in the final first-stage regression models because of their weak predicative power. Including them violates the relevance criterion of an instrumental variable. Examined in this analysis but not included in the final modeling were TANF attributes regarding the strictness of sanctions and time limits (Pavetti and Bloom 2001), state diversion policies under TANF (Urban Institute, Welfare Rules Database; anfdata.urban.org/wrd/wrdwelcome.cfm), and states with childcare fee waivers available through the Child Care Development Fund (Blau 2003).

16

The first three are rather self-explanatory. The personal responsibility clauses include restrictions on benefits for increasing the family size (i.e., family caps), as well as the children’s regular school attendance and health check-ups.

17

Wolfers et al. (2015) reported the number of 25- to 54-year-old black males who are “missing” based on deviations from the biological sex ratio at birth. Using Census Bureau data and a ratio of 1 male to every female, they found almost 20 % (or 1.5 million) fewer black men living in the U.S. general population. As the authors noted, the primary factors contributing to this gap are differential incarceration and mortality rates—issues that disproportionately impact low-skilled males. The authors also found that this gap exists among whites, although the impact is not as pronounced.

18

This ratio is derived from the same underlying CPS data. To obtain a more precise estimate, I construct this ratio at the annual level. In addition, I estimate it for all less-educated males and females within the group examined. This latter decision further increases the precision of the estimate and reduces bias from contemporaneous family structure decisions based on sex ratios.

19

This test reveals that the instruments in this analysis do not directly influence male labor supply but that the impact is moderated through the instrument’s influence on female LFP. In other words, they meet the exclusion restriction of a valid IV. This finding is critical to establish a valid IV research design and, as will be shown, is not typically found in the modeling for other groups.

20

Holzer et al. (2005) examined the 1979–2000 period as well as a different set of age categories.

21

Recall that DI and SSI beneficiaries, by definition, are not part of the labor market. This is also the rationale behind not including variables such as the unemployment rate in the models. Other modeling (not shown but available upon request) includes state and federal SSI generosity in both the first and second stages, but results remain substantively unaffected.

References

References
Angrist, J. D., & Pischke, J-S (
2009
).
Mostly harmless econometrics: An empiricist’s companion
.
Princeton, NJ
:
Princeton University Press
.
Bagley, B. (
2012
).
Drug trafficking and organized crime in the Americas: Major trends in the twenty-first century
(Woodrow Wilson Center Update on the Americas).
Washington, DC
:
Woodrow Wilson International Center for Scholars
. Retrieved from http://www.wilsoncenter.org/sites/default/files/BB%20Final.pdf
Bartik, T. J. (
2002
).
Instrumental variable estimates of the labor market spillover effects of welfare reform
(Upjohn Institute Working Paper No. 02-78).
Kalamazoo, MI
:
W.E. Upjohn Institute for Employment Research
. Retrieved from http://research.upjohn.org/up_workingpapers/78
Blanchflower, D. G., & Freeman, R. B. (
2000
).
Youth employment and joblessness in advanced countries
.
Chicago, IL
:
University of Chicago Press
.
Blank, R. M. (
2002
).
Evaluating welfare reform in the United States
.
Journal of Economic Literature
,
40
,
1105
1166
. 10.1257/.40.4.1105.
Blank, R. M. (
2009
).
Economic change and the structure of opportunity for less-skilled workers
.
Focus
,
26
(
2
),
14
20
. Retrieved from http://www.irp.wisc.edu/publications/focus/pdfs/foc262c.pdf
Blank, R. M., & Gelbach, J. (
2006
).
Are less-educated women crowding less-educated men out of the labor market?
. In Mincy, R. B. (Ed.),
Black males left behind
(pp.
87
119
).
Washington, DC
:
The Urban Institute Press
.
Blau, D. M. (
2003
).
Child care subsidy programs
. In Moffitt, R. (Ed.),
Means-tested transfer programs in the United States
(pp.
443
516
).
Chicago, IL
:
University of Chicago Press
.
Bloom, D., & Haskins, R. (
2010
).
Helping high school dropouts improve their prospects
(Future of Children Policy Brief, Social Genome Project Research Series No. 4).
Princeton, NJ
:
Princeton-Brookings
. Retrieved from http://www.brookings.edu/research/papers/2010/04/27-helping-dropouts-haskins.
Card, D., & Krueger, A. B. (
1994
).
Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania
.
American Economic Review
,
84
,
772
793
.
Carlson, M., McLanahan, S., & England, P. (
2004
).
Union formation in fragile families
.
Demography
,
41
,
237
261
. 10.1353/dem.2004.0012.
Carlson, M. J., VanOrman, A. G., & Pilkauskas, N. V. (
2013
).
Examining the antecedents of U.S. nonmarital fatherhood
.
Demography
,
50
,
1421
1447
. 10.1007/s13524-013-0201-9.
Cherlin, A. J. (
2009
).
The marriage-go-round: The state of marriage and the family in America today
.
New York, NY
:
Knopf Doubleday Publishing Group
.
Cherlin, A. J. (
2010
).
Demographic trends in the United States: A review of research in the 2000s
.
Journal of Marriage and Family
,
72
,
403
419
. 10.1111/j.1741-3737.2010.00710.x.
Danziger, S., Heflin, C. M., Corcoran, M. E., Oltmans, E., & Wang, H-C (
2002
).
Does it pay to move from welfare to work?
.
Journal of Policy Analysis and Management
,
21
,
671
692
. 10.1002/pam.10080.
DeParle, J. (
2004
).
American dream: Three women, ten kids, and a nation’s drive to end welfare
.
New York, NY
:
Viking Adult
.
Edin, K., & Kefalas, M. J. (
2005
).
Promises I can keep: Why poor women put motherhood before marriage
.
Berkeley
:
University of California Press
.
Edin, K., & Lein, L. (
1997
).
Work, welfare, and single mothers’ economic survival strategies
.
American Sociological Review
,
62
,
253
266
. 10.2307/2657303.
Ellwood, D. T. (
1988
).
Poor support: Poverty in the American family
.
New York, NY
:
Basic Books
.
Flood, S., King, M., Ruggles, S., & Warren, J. R. (
2015
).
Integrated Public Use Microdata Series, Current Population Survey: Version 4.0
[Machine-readable database].
Minneapolis
:
University of Minnesota
.
Freeman, R. B. (
2000
).
Disadvantaged young men and crime
. In Blanchflower, D. G. & Freeman, R. B. (Eds.),
Youth employment and joblessness in advanced countries
(pp.
215
246
).
Chicago, IL
:
University of Chicago Press
. Retrieved from http://www.nber.org/chapters/c6806
Grogger, J., & Karoly, L. (
2005
).
Welfare reform: Effects of a decade of change
.
Cambridge, MA
:
Harvard University Press
.
Harlow, C. W. (
2003
).
Education and correctional populations
(Bureau of Justice Statistics special report).
Washington, DC
:
Bureau of Justice Statistics, U.S. Department of Justice
. Retrieved from http://www.eric.ed.gov/ERICWebPortal/detail?accno=ED477377
Holzer, H. J., & Offner, P. (
2006
).
Trends in the employment outcomes of young black men, 1979–2000
. In Mincy, R. B. (Ed.),
Black males left behind
(pp.
11
37
).
Washington, DC
:
The Urban Institute Press
.
Holzer, H. J., Offner, P., & Sorensen, E. (
2005
).
Declining employment among young black less-educated men: The role of incarceration and child support
.
Journal of Policy Analysis and Management
,
24
,
329
350
. 10.1002/pam.20092.
Huang, C-C, Kunz, J., & Garfinkel, I. (
2002
).
The effect of child support on welfare exits and re-entries
.
Journal of Policy Analysis and Management
,
21
,
557
576
. 10.1002/pam.10073.
Levitt, S. D. (
2001
).
Alternative strategies for identifying the link between unemployment and crime
.
Journal of Quantitative Criminology
,
17
,
377
390
. 10.1023/A:1012541821386.
McLaughlin, D. K., & Lichter, D. T. (
1997
).
Poverty and the marital behavior of young women
.
Journal of Marriage and Family
,
59
,
582
594
. 10.2307/353947.
Men adrift: Badly educated men in rich countries have not adapted well to trade, technology, or feminism
. (
2015
,
May
30
).
The Economist
. Retrieved from http://www.economist.com/news/essays/21649050-badly-educated-men-rich-countries-have-not-adapted-well-trade-technology-or-feminism
Meyer, D. R., Cancian, M., & Cook, S. T. (
2005
).
Multiple‐partner fertility: Incidence and implications for child support policy
.
Social Service Review
,
79
,
577
601
. 10.1086/454386.
Moffitt, R. (
1992
).
Incentive effects of the U.S. welfare system: A review
.
Journal of Economic Literature
,
30
,
1
61
.
Moffitt, R. A. (
2002
).
Welfare programs and labor supply
. In Auerbach, A. J., & Feldstein, M. (Eds.),
Handbook of public economics
(pp.
2393
2430
).
Amsterdam, The Netherlands
:
Elsevier
.
Moffitt, R. A. (
2007
).
Four decades of antipoverty policy: Past developments and future directions
.
Focus
,
25
(
1
),
39
44
.
Moffitt, R. A. (
2015
).
The deserving poor, the family, and the U.S. welfare system
.
Demography
,
52
,
729
749
. 10.1007/s13524-015-0395-0.
Pavetti, L., & Bloom, D. (
2001
).
State sanctions and time limits
. In Blank, R., & Haskins, R. (Eds.),
The new world of welfare
(pp.
245
269
).
Washington, DC
:
The Brookings Institution
.
The Pew Charitable Trusts
. (
2010
).
Collateral costs: Incarceration’s effect on economic mobility
(Joint report of the Economic Mobility Project and the Public Safety Performance Project).
Washington, DC
:
The Pew Charitable Trusts
. Retrieved from http://www.pewstates.org/research/reports/collateral-costs-85899373309
Rangarajan, A., & Gleason, P. (
1998
).
Young unwed fathers of AFDC children: Do they provide support?
.
Demography
,
35
,
175
186
. 10.2307/3004050.
Smeeding, T. M., Garfinkel, I., & Mincy, R. B. (
2011
).
Young disadvantaged men: Fathers, families, poverty, and policy
.
ANNALS of the American Academy of Political and Social Science
,
635
,
6
21
. 10.1177/0002716210394774.
United Nations Office on Drugs and Crime (UNODC)
. (
2012
).
World drug report: 2012
.
Vienna, Austria
:
UNODC
. Retrieved from https://www.unodc.org/documents/data-and-analysis/WDR2012/WDR_2012_web_small.pdf
University of Kentucky Center for Poverty Research
. (
2015
).
UKCPR National Welfare Data, 1980–2014
[Data set].
Lexington, KY
:
Gatton College of Business & Economics, University of Kentucky
. Retrieved from http://www.ukcpr.org/data
U.S. Department of Health and Human Services
. (
1999
).
Table A: State implementation of major changes in welfare policies, 1992–1998
. Retrieved from http://aspe.hhs.gov/hsp/waiver-policies99/Table_A.PDF
The weaker sex: Boys are being outclassed by girls at both school and university, and the gap is widening
. (
2015
,
March
7
).
The Economist
. Retrieved from http://www.economist.com/news/leaders/21652323-blue-collar-men-rich-countries-are-trouble-they-must-learn-adapt-weaker-sex
Wilson, W. J. (
1987
).
The truly disadvantaged: The inner city, the underclass, and public policy
.
Chicago, IL
:
University of Chicago Press
.
Wolfers, J., Leonhardt, D., & Quealy, K. (
2015
,
April
20
).
1.5 million missing black men
.
New York Times
. Retrieved from http://www.nytimes.com/interactive/2015/04/20/upshot/missing-black-men.html
Ziliak, J. P., Figlio, D. N., Davis, E. E., & Connolly, L. S. (
2000
).
Accounting for the decline in AFDC caseloads: Welfare reform or the economy?
.
Journal of Human Resources
,
35
,
570
586
. 10.2307/146393.

Supplementary data