Age Gradient in Women’s Crime: The Role of Welfare Reform

Abstract

We investigate how welfare reform in the U.S. in the 1990s shaped the age gradient in women’s property crime arrests. Using Federal Bureau of Investigation data, we investigated the age-patterning of effects of welfare reform on women’s arrests for property crime, the type of crime that welfare reform has been shown to affect. We found that welfare reform reduced women’s property crime arrests by about 4%, with particularly strong effects for women ages 25 to 29, slightly stronger effects in states with stricter work incentives, and much stronger effects in states with high per capita criminal justice expenditures.

Keywords

connections with welfare system economic marginality female criminality women and social policy women’s desistance age gradient

Social scientists have been studying age patterns of crime and arrest rates for decades, both to increase understanding of criminal behavior and to inform public policy. A strong and dominant finding has been that criminal behavior declines with age after increasing during adolescence, although there are some deviations from this pattern across time and place and the peak ages and rates of decline vary by type of crime (Ulmer & Steffensmeier, 2014). Because the majority of crimes are committed by men, most research on the age-crime gradient has pertained to men, despite the fact that the criminal behavior of women has been increasing both in absolute terms and relative to men (Campaniello, 2014).

Building on seminal qualitative work by Carlen (1988), a body of research on gender differences in criminal behavior points to changes in economic opportunity structures as determinants of women’s crime. Steffensmeier (1993) found that women’s arrest rates for larceny, fraud, forgery and embezzlement about doubled between 1960 and 1990, and posited that economic hardship (e.g., as a result of out-of-wedlock births) creates pressure for women to commit economic crimes. Heimer (2000) pointed to the erosion of the U.S. welfare safety net as an important factor in the economic marginalization of women and hypothesized that the desire to take care of children is an important factor in women’s crime. Steffensmeier and Haynie (2000) found that area-level structural disadvantage—poverty, joblessness, and female-headed households—was associated with higher arrest rates for index crimes including homicide, robbery, assault, burglary and larceny¹ among both men and women. A review article found that employment and fear of job loss affect women’s decisions to desist from crime (McIvor et al., 2004), and a study of female offenders found that poverty status increases the odds of re-arrest or supervision violation, with housing support reducing the odds of recidivism (Holtfreter et al., 2004). This literature suggests that a major change in economic opportunity structures, such as welfare reform in the U.S., would affect women’s economic crime and that the effects are likely to vary by age.

The 1996 U.S. Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) ended entitlement to welfare under Aid to Families with Dependent Children (AFDC) and replaced the AFDC program with Temporary Assistance for Needy Families (TANF) block grants to states. This national legislation and policies in many states earlier in the 1990s (collectively referred to as welfare reform) were designed to permanently reduce dependence on cash assistance by promoting employment of women at risk of relying on welfare. The key strategies underpinning the legislation were work requirements and time limits as conditions for receipt of cash assistance; these strategies strongly incentivized women at risk for relying on public assistance (not just welfare recipients) to secure employment, by reducing the benefits of welfare reliance compared to work and eliminating the practical option of long-term reliance on public assistance. Welfare reform was considered successful in that welfare caseloads plunged by 73% between 1994 and 2019 (Meyer & Floyd, 2020), with substantial coinciding increases in employment of low-skilled women (Fang & Keane, 2004; Ziliak, 2016).

Corman et al. (2014) exploited changes in the implementation of welfare reform across states and over time to estimate causal effects of welfare reform in the U.S. on adult women’s arrests. Using monthly state-level data on arrest rates compiled by the Federal Bureau of Investigation (FBI) in conjunction with dates of state welfare reform implementation and other relevant factors, the authors found that welfare reform led to decreases in women’s arrests for property crimes (by 4–5%), but that it did not affect arrests for other types of crimes. These effects were noteworthy considering recent evidence that women are less responsive than men to incentives in terms of committing property crime (Campaniello & Gavrilova, 2018).

The aggregate effects estimated by Corman et al. (2014) could mask heterogeneity by age, however. For example, the literature on economic marginalization discussed above suggests that welfare reform could lead to larger decreases in property crime among women in their 20s, who are more likely to have young children,² than among mothers of other ages—for example, because the former experienced larger increases in employment (U.S. Department of Labor, Women’s Bureau, 2020) than the latter and potentially benefited from childcare, transportation, and job search subsidies that were provided in many states to support employment. Of particular note, childcare subsidies in the U.S. more than tripled by 2000 after welfare reform was implemented (Adams & Rohacek, 2002). For some age groups, welfare reform could even lead to increases in crime, particularly in states that did not offer adequate employment supports. Even within age groups, there is likely to be heterogeneity in effects across women depending on their specific employment situations and family circumstances.

As far as we know, interactions between changes in economic opportunity structures and women’s stage of the life course in shaping crime have not previously been explored. Related research has focused on how life events or noteworthy changes in circumstances—sometimes referred to as “turning points”—can lead people to launch criminal careers or desist from criminal activity (Sampson & Laub, 1990). Turning points that have been studied include individual-level life events and circumstances, including marriage, employment, and health shocks in families (e.g., Corman et al., 2011, Horney et al., 1995; Laub et al., 1998; Sampson et al., 2006; Uggen, 2000).³ Nguyen and Loughran (2018) pointed to the major methodological challenge in this literature—that individual-level turning points are rarely random so the estimated effects on crime-related outcomes are likely to be biased—and advocated for focusing on events that are unrelated to individual level circumstances or decision-making (i.e., exogenous). At the same time, Sampson and Laub (2016) and Laub and Sampson (2020) pointed to a need for research on how broad societal changes (which can have large societal impacts) affect criminal participation and trajectories. The large-scale policy shift under welfare reform can reasonably be viewed as an exogenous and substantial change in the economic opportunity structures of low-income women.

In this study, we use arrest data from the FBI to investigate age-patterning of the effects of welfare reform on women’s arrests for property crime, the type of crime that women are most likely to commit (Campaniello, 2014), that welfare reform has been shown to affect (Corman et al., 2014), and that the literature on economic marginalization and women’s crime suggests would be particularly salient (e.g., Carlen, 1988; Heimer, 2000; Holtfreter et al., 2004; Steffensmeier, 1993; Steffensmeier & Haynie, 2000).⁴ That is, we explore the extent to which welfare reform had differential effects on women’s property crime arrests by age. We also explore the extent to which state-level welfare and criminal justice policies shaped the age patterning of the effects of welfare reform on women’s property crime arrests. Although we cannot directly explore welfare reform as a turning point because we do not follow women over time, the findings are suggestive of the extent to which a broad societal change in women’s economic opportunity structure shapes women’s criminal careers. As in studies such as Steffensmeier and Haynie (2000) and Corman et al. (2014), we make cautious inferences about individual behaviors from analysis of aggregated data.

Theory

Following the economic model of crime formulated by Becker (1968), extended by Ehrlich (1973), and modified by Corman et al. (2014), potential criminals act to maximize an expected utility function depicted below in equation (1):

E (U) = (1 - p) U (X_{1}, t_{c 1}) + p U (X_{2}, t_{c 2}),

(1)

wherein p is the probability of apprehension and conviction and (1-p) is the probability of not being apprehended. If the individual is not apprehended, income will equal X₁, which as shown in equation (2) is a function of the wage in the legal sector (W_l) times the amount of time spent working in the legal sector (t_l), plus the return (wage) in the illegal sector (W_i) times the time devoted to the illegal sector (t_i). W₀ represents unearned income. If the individual is apprehended, income will be X₂, as shown in equation (3) wherein income is reduced by F, which may include a fine, legal expenses, and incarceration, as well as reduced future wages due to a criminal record. The terms (t_c) represents leisure (or consumption) time depending on apprehension status.

X_{1} = W_{i} (t_{i}) + W_{1} (t_{1}) + W_{0}

(2)

X_{2} = W_{i} (t_{i}) + W_{1} (t_{1}) - F_{i} (t_{i}) + W_{0}

(3)

Welfare reform imposed strong incentives to engage in legal work, which can be a substitute for illegal work. As indicated above, an arrest results in a lower income in the current period, if caught. Additionally, an arrest may result in a lower wage in the future, which may be more consequential for women under welfare reform, since they will no longer have the option of long-term welfare receipt. An arrest and conviction for a felony offense may also make individuals ineligible for future welfare receipt and public assistance in certain states, and thus also lead to a reduction in the expected value of unearned income (W₀).

We expect relatively large negative effects of welfare reform on crime for women in their 20s because employment has been found to reduce criminal offenses for both genders (Verbruggen et al., 2012), the employment increases as welfare reform unfolded were particularly high for mothers with young children (U.S. Department of Labor, Women’s Bureau, 2020), women in their 20s are at peak ages for having young children (see endnote #2) and benefiting from childcare subsidies that were part of welfare reform, the literature on women’s economic marginalization suggests that desire to take care of children is an important factor in women’s economic crime (Heimer, 2000), and welfare reform led to decreases in women’s arrests for property (economic) crime but not arrests for other types of crimes (Corman et al., 2014).

Negative effects of welfare reform for women of relatively young ages should lead to large overall reductions in crime, since younger women commit the most crimes and their non-participation in crime or earlier desistance would generate large life course effects, given what is known about individuals’ criminal trajectories (e.g., Ulmer & Steffensmeier, 2014). However, welfare reform may not have led to reductions in crime, or may have even led to increases, among women in their teens, for two reasons. First, welfare reform led to decreases in teen childbearing (Lopoo & Raissian, 2012), so the work incentives and work supports under welfare reform would be relevant for fewer women in this age group. Second, a recent study found that welfare reform led to increases in substance use and skipping school among 10th and 12th grade girls (Dave et al., 2020), and these behaviors can be precursors to lower occupational status and unemployment (Tanner et al., 1999) and crime (Loeber et al., 2013).

In our consideration of differential effects by strength of welfare and criminal justice policies, we expect that welfare reform will have stronger negative effects on women’s property crime arrests in states with stricter work incentive policies in their welfare programs, because of the stronger “treatment.” That is, given that previous research found that welfare reform resulted in an overall decrease in women’s property crime arrests (Corman et al., 2014), we would expect the effects to be stronger in states that provided stronger incentives to work while enforcing stricter sanctions for not working. We would also expect larger decreases in women’s property crime arrests in states with stricter criminal justice policies, because the cost of engaging in criminal behavior would be much higher in those states. In terms of the age-crime gradient, we would expect the effects of welfare reform on property crime arrests to be stronger for women at peak childbearing ages in states with strict policies.

Data

The two main sources of data for this study are: (1) Uniform Crime Reporting Program arrests from the Monthly Master Files from the U.S. Department of Justice Federal Bureau of Investigation (FBI), which provide the number of arrests by age and gender for each month/offense category/reporting agency; and (2) implementation dates of welfare reform at the state level. The former is used to create measures of arrests and the latter is used to characterize welfare reform, as described below. We use arrest data from 1990 through 2002, the period during which welfare reform was implemented by all states. Given that welfare reform only affects women with dependent children, our analysis is restricted to adult women in that age range. Thus, we restrict our analysis to arrests of women ages 18 through 49 years, and then separately analyze arrests for different age groups within this broader range.

Comprehensive reviews found that Uniform Crime Reports of arrests are valid indicators of serious crimes (Gove et al., 1985; Nagin, 1998), and much of the criminology literature uses arrests as proxies for crime. For ease of discussion, we at times use the terms “crime” and “arrests” interchangeably from this point forward, acknowledging that not all crimes result in arrests, that people can be arrested for crimes they did not commit, and that it is possible that the same individual could be arrested more than once during a given month (our time unit of observation) and/or over our sample period.

Measures of Arrests

The FBI data include a record for each criminal justice agency in the U.S. for each month. For those that report, the monthly record includes the number of arrests by crime category, age category, and sex. We limit our sample to larger criminal justice agencies that cover at least 50,000 individuals.⁵ In 1996, the year that PRWORA was enacted, agencies with populations of 50,000 or more people covered approximately 55% of the total U.S. population (147 million/268 million, calculated by the authors from the FBI and U.S. Census data). From these agency-based observations, we aggregated the data to the month/year/state level. Even among large criminal justice agencies, not all agencies reported in all months. For example, in 1996, of the 147 million people in the U.S. residing under the jurisdiction of agencies of 50,000 people or more, approximately 106 million people (~72%) were covered by agencies that reported arrests to the FBI in all 12 months.⁶ We include both the total population in all agencies covering populations of at least 50,000 in the given state/month/year and the total state population on the right-hand side in our models. These population measures account for differences in the size of the underlying population that is represented or covered by the FBI reports.⁷ We consider arrests for burglary, larceny/theft (other than motor vehicle), and motor vehicle theft, which are classified as “property index crimes.”

In the original FBI data, arrests for adults are reported by individual years of age through age 24 and in 5-year increments starting at age 25. We thus consider the following age categories: 18–20, 21–24, 25–29, 30–34, 35–39, and 40–49 years, with the last category spanning 10 years because of the relatively low number of arrests in this age group (see Table 1). In some analyses, we further consolidate the age categories by decades: 21–29, 30–39, and 40–49, in order to draw out key patterns and maximize statistical power.

Table 1.

Property Crime Arrest Rates by Age Range. FBI Arrests, Women 1988–2002.

Age range (years)	1988–1991	% difference from next younger cohort	1992–1997	% difference from next younger cohort	1998–2002	% difference from next younger cohort
Age range (years)	1	2	3	4	5	6
15–17	1710.7		1984.3		1600.1
18–20	1224.3	−28.4	1303.4	−34.3	1153.2	−27.9
21–24	1024.8	−16.3	960.2	−26.3	774.2	−32.9
25–29	860.6	−16.0	795.8	−17.1	580.2	−25.1
30–34	670.2	−22.1	662.1	−16.8	511.0	−11.9
35–39	472.7	−29.5	495.9	−25.1	427.5	−16.3
40–44	302.3	−36.0	321.9	−35.1	306.3	−28.3
45–49	213.8	−29.3	202.2	−37.2	192.6	−37.1

Note. Arrest rates per 100,000 (columns 1, 3, and 5) are for “index” crimes of burglary, larceny and motor vehicle theft for all criminal justice agencies representing populations of 50,000 or more and which reported for at least half of the year. Columns 2, 4, and 6 compare the arrest rate for each age group to the next younger age group in the same period. Shaded figures indicate arrest rates for a selected age cohort over time.

Measures of Welfare Reform Implementation

The first phase of welfare reform consisted of pre-PRWORA waivers, which allowed states to implement experimental changes to their AFDC programs. Although not federally mandated, pre-PRWORA waivers were implemented in the majority of states by the time the federal PRWORA was enacted in 1996 (Schoeni & Blank, 2000). The second phase of welfare reform came with the enactment of PRWORA. States implemented their TANF programs, which were required to meet (but could be stricter than) federal guidelines between September 1996 and January 1998 (USDHHS, 1999). Specifically, waivers were introduced in 29 states over a period of 53 months, and TANF was implemented in all states over a period of 17 months. Combining both waivers and TANF, states implemented any welfare reform over a period of 64 months, spanning from October 1992 through January 1998.

Following Corman et al. (2014), and consistent with the convention in the welfare reform literature (reviewed in Blank, 2002), we exploit differences in the timing of welfare reform implementation across states. For waivers, we consider whether, in a given year and month, a given state had a statewide AFDC waiver in place that substantially altered the nature of AFDC with regard to time limits, work requirements, earnings disregards, sanctions, and/or family caps. For TANF, we consider whether, in a given year and month, the state had implemented TANF post-PRWORA. Many studies include separate measures for AFDC waivers and TANF, since they represent distinct phases of welfare reform. We estimate and report on specifications with the separate indicators in order to assess differential effects of the separate phases, but in most of our models we use a single indicator for any welfare reform (i.e., AFDC waiver or TANF) in order to maximize statistical power.

Other Key Measures

In secondary runs, we consider differential effects by state policy variations in the implementation of welfare reform post-PRWORA (a more nuanced welfare reform effect) and by state criminal justice policies (to assess the effects of welfare reform under different criminal justice contexts). Blank and Schmidt (2001) characterized state work incentives according to four significant aspects of their post-PRWORA welfare programs—benefit generosity, earnings disregards, sanctions, and time limits—as strong, medium, or weak, and combined this information to classify states as having strong, mixed, or weak work incentives. High benefit generosity, low earnings disregards, lenient sanctions, and lenient time limits are indicators of weak work incentives. Blank and Schmidt categorized states with weak incentives in at least one of the four categories and strong incentives in no other categories as having weak work incentives, and states with strong incentives in at least one category and weak incentives in no other categories as having strong work incentives. All other states were categorized as having mixed work incentives. For this study, we dichotomously classified states as having weak (versus mixed or strong) work incentives using the Blank and Schmidt classifications, resulting in 10 states (including DC) being classified as having weak work incentives and all others being classified as having mixed or strong work incentives.

Welfare reform has been successful in reducing caseloads while increasing employment among low-skilled women; over our sample period spanning 1990 to 2002, the median state experienced a decline in the number of welfare recipients of over 55%. While some of this decline reflected the strength of the economy over the 1990s, studies have consistently found that as much as one-third to one-half of the caseload decline can be explained by welfare reform (Dave et al., 2012; Grogger & Karoly, 2005; Loprest, 2012). As an alternative proxy for differential work incentives, we assess whether states that experienced larger declines in welfare caseloads also experienced larger declines in property crime arrests, and whether these differences were compounded or moderated across age. Specifically, we classify states into equal tertiles, based on the percentage decrease in welfare caseloads between 1990 and 2002 (low: decline <43%; medium: decline 43–60%; and high: decline >60%) and decompose the average effects of welfare reform across these sets of states.⁸

A substantial body of evidence indicates that increases in criminal justice resources lead to lower crime rates (Eide, 2000; Nagin, 2013). Using data from the U.S. Department of Justice Bureau of Justice Statistics, we characterize state criminal justice policy using a broad measure of per capita total justice system expenditures by state and local governments for each state/year and by per capita total full-time employees in criminal justice. There are three components of these expenditures: police, courts, and corrections, with police representing the largest share (U.S. Department of Justice, Bureau of Justice Statistics, 2017).

Methods

We exploit variation in the timing of welfare reform implementation across states within a quasi-experimental difference-in-differences (DD) research design, to estimate overall effects of welfare reform on women’s property crime, assess how the effects of this broad pro-employment policy shift differed by age, and explore how the age-specific effects differed by relevant state policies. Following Corman et al. (2014), we begin with the following baseline specification, akin to a reduced-form crime production function, relating changes in arrests for property index crimes to welfare reform:

\begin{array}{l} {Ln A}_{smt} = {Welfare}_{smt} Π + Z_{st} β + {State}_{s} Ω + ({Year}_{t} * {Month}_{m}) Ψ + \\ [({State}_{s} * Time Pre - {Welfare Reform}_{mt}) Θ] + \\ [({State}_{s} * {Time}_{mt}) Φ] + ε_{smt} \end{array}

(4)

In equation (4), the outcome represents the natural log of total arrests related to property index offenses for women ages 18 to 49 in state s, during month m and year t.⁹ Arrests are a function of welfare reform (Welfare), characterized separately by the enactment of major waivers to AFDC and implementation of TANF, and in alternative specifications by an indicator for the implementation of any welfare reform (either a major AFDC waiver or TANF). We also control for an extensive vector of time-varying state covariates (Z), including measures of the state’s economic and labor market conditions (unemployment rate, per capital personal income, poverty rate, minimum wage), relevant population base and data coverage (log of the state’s total population, log of the state’s female population for the relevant age group, log of the agency population for months with arrest reports, percent of the population covered by the reporting agencies), and criminal justice system (log of the state’s criminal justice expenditures, log of full-time equivalent number of police officers). The parameters of interest are Π, which capture the “reduced form” or total effect of welfare reform on property crime arrests, operating through all potential competing and/or reinforcing pathways. The parameter ε reflects a state-time disturbance term, which we assume to be correlated within states over time. Hence, we report and draw inferences based on state-clustered standard errors, which adjust for any arbitrary correlation within state cells over time.

All specifications further include state indicators (State), which control for all unmeasured time-invariant state heterogeneity, and year-by-month indicators (Year*Month), which control for any seasonality and national trends in criminal behavior and arrests, and are weighted by the state’s female population (ages 18–49). Unweighted models would assign each state-year-month equal weight, and the DD effect would capture an average causal effect over states rather than over individuals. Population-based weighting to state-aggregated panels would yield an average policy response over individuals. Population weighting in such models can also improve the precision of the estimates since arrest rates in a small state may be more variable from year-to-year than in a larger state; population-based weights can thus give a larger weight to more precise measurements over time (Angrist & Pischke, 2014).¹⁰

A key methodological challenge in identifying a plausibly causal effect in the context of a DD model lies in separating the effect of welfare reform from that of other state-level time-varying factors that may be related to women’s property crime arrests. To the extent that these factors uniformly affected women’s property crime across the nation, they would be captured by the time fixed effects. However, the concern relates to the possibility that these factors may have differentially affected women across states.

We address this concern regarding unobserved state-specific time-varying confounding factors in a number of ways. First, all models already include a rich set of covariates (Z) relating to the state’s economic, labor market, and criminal justice conditions. Second, we extend the specifications to include both the natural log of arrests for all offenses and that for property crimes among men. These measures capture all time-varying unmeasured factors in a state affecting overall crime, and specifically property crime, in that state.¹¹ Third, we control for lagged state-level economic indicators and welfare caseloads, in order to account for potential policy endogeneity. Fourth, we control for state-specific linear pre-policy trends, by including an interaction between each state indicator and the number of months prior to when the state implemented welfare reform (State*Time Pre-Welfare Reform). These controls parametrically account for any unmeasured systematic differential trends across states prior to policy implementation and help to address deviations from the parallel-trends assumption underlying the DD analysis.

Finally, in some specifications, we supplement the vector of controls with state-specific linear trends to further account for any state-level factors that may have coincided with welfare reform and were associated with women’s crime over the sample period (State*Time). In particular, potential state-specific shifts in the age distribution of births could bias the estimated effects of welfare reform on crime and distort comparisons across age groups. The inclusion of state-specific linear trends addresses such potentially confounding trends.

To test the hypotheses laid out earlier regarding differential effects of welfare reform by age, we estimate equation (4) separately for the following age groups (18–20, 21–24, 25–29, 30–34, 35–39, and 40–49 years). We also assess how the age-specific effects of welfare reform on women’s property crime interact with the strength of state work incentives under welfare reform as well as by features of state criminal justice systems.

Results

Table 1 shows women’s property crime arrest rates, measured as the combined rates for burglary, larceny and motor vehicle theft, by age category during three relevant time periods: Pre-welfare reform (1988–1991), during implementation of welfare reform (1992–1997), and after full implementation of welfare reform (1998–2002). As expected, there was a large age gradient in each of the three periods; columns 1, 3, and 5 show that arrest rates decreased with age and columns 2, 4, and 6 indicate the percentage decreases from one age group to the next. Arrest rates also declined in cohorts as they aged. For example, the women who were 18 to 20 years old in the first period were 21 to 24 in the second period and 25 to 29 in the third period, and the decline in arrest rates of this cohort can be seen in the shaded cells. The age gradient in property crime also changed over time. For example, in the pre-welfare reform period, the arrest rate for women ages 21 to 24 was about 16% lower than that for women ages 18 to 20, while the corresponding difference in the fully-implemented period was 33% (in the same direction).

Table 2 presents estimates from models corresponding to equation (4) that estimate effects of welfare reform on property crime arrests among women ages 18 to 49. The estimate in Model 1 is from a parsimonious specification that only includes basic controls in conjunction with state and time fixed effects, suggesting an insignificant and small decline (1.7%) in property crime arrests associated with welfare reform. As we progressively add controls for unmeasured time-varying state heterogeneity, with measures of men’s arrests in Model 2 and lagged economic conditions and welfare caseloads in Model 3, the effect becomes stronger and statistically significant. These models suggest that welfare reform led to an approximate 4% decline in arrests for property crime of women.¹² We decompose this overall effect into effects stemming from early AFDC waivers and federal welfare reform (TANF) in Model 4. Both aspects of welfare reform were associated with significant reductions in arrests; while the magnitude is somewhat larger for TANF (4.9% relative to 3.7% for AFDC waivers), the confidence intervals overlap considerably and we are not able to reject the null hypothesis of equal effect sizes. This is perhaps not surprising as most policies that characterized the early reform efforts were later incorporated into TANF. The final two specifications control for differential pre-policy trends and state linear trends, respectively. The robustness of the estimates to these controls provides validation of the research design and the results when including state linear trends in particular suggest that the estimates are not biased by unobserved state-specific trends concurrent with state implementation of welfare reform, such as potential shifts in the age distribution of births.¹³ Overall and consistent with findings by Corman et al. (2014) for 21 to 49 year olds, the estimates in Table 2 indicate a non-negligible decline in property crime among women ages 18 to 49, on the order of about 4% to 5%, as a result of welfare reform.

Table 2.

Effects of Welfare Reform on Property Crime. FBI Arrests. Women, Ages 18 to 49 years, 1990 to 2002.

Outcome	Ln arrests for property index crimes
Specification	1	2	3	4	5	6
Any welfare reform	−0.0165 (0.0246)	−0.0286 (0.0177)	−0.0388** (0.0184)	—	−0.0361** (0.0172)	−0.0451** (0.0190)
AFDC waiver	—	—	—	−0.0365* (0.0192)	—	—
TANF	—	—	—	−0.0493*** (0.0181)	—	—
State indicators	Yes	Yes	Yes	Yes	Yes	Yes
Year*Month indicators	Yes	Yes	Yes	Yes	Yes	Yes
Measures of male arrests	No	Yes	Yes	Yes	Yes	Yes
Lagged economic conditions and welfare caseloads	No	No	Yes	Yes	Yes	Yes
State-specific linear pre-policy trends	No	No	No	No	Yes	No
State-specific linear trends	No	No	No	No	No	Yes
Observations (state/month/year)	6298	6298	6298	6298	6298	6298

In Table 3, we explore heterogeneity in this average effect across the age distribution and assess how welfare reform affected the age-crime gradient. For parsimony of presentation, we focus on Specification 3 from Table 2, although estimates and patterns are insensitive to this decision. Model 1 in Table 3 replicates the estimate for women ages 18 to 49 for ease of comparison, and the other models present estimates for narrower age bands. While estimates are largely negative across all age groups, there is considerable heterogeneity in effect sizes.

Table 3.

Effects of Welfare Reform on Property Crime. FBI Arrests, Women, 1990 to 2002. Differential Effects across Age Distribution.

Outcome	Ln arrests for property index crimes
Specification	1	2	3	4	5	6	7	8	9
Age group (years)	18–49	18–20	21–24	25–29	21–29	30–34	35–39	30–39	40–49
Any welfare reform	−0.0388** (0.0184)	0.0009 (0.0026)	−0.0278 (0.0270)	−0.0439* (0.0237)	−0.0378* (0.0215)	−0.0361 (0.0258)	−0.0052 (0.0275)	−0.0247 (0.0213)	−0.0607** (0.0297)
State indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year*Month indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Measures of male arrests	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Lagged economic conditions and welfare caseloads	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Equality of WR effect across age groups (Models 1=2=3=4=6=7=9)		30.74*** [0.000]
Equality of WR effect across age groups (Models 1=2=5=8=9)		11.73*** [0.019]
Equality of WR effect across given age group and average effect (ages 18–49)	—	4.39** [0.036]	0.54 [0.464]	0.15 [0.700]	0.01 [0.914]	0.03 [0.860]	3.31* [0.069]	2.27 [0.132]	1.24 [0.266]
Observations (state/month/year)	6298	6298	6298	6298	6298	6298	6298	6298	6298

Note. Asterisks denote significance as follows: ***p ≤ .01; **.01 < p ≤ .05; *.05 < p ≤ .10. Coefficients from OLS semi-log models are presented, weighted by the state’s female population ages 18 to 49. Standard errors are adjusted for arbitrary correlation within state cells, and reported in parentheses. The regression models are based on Specification 3 from Table 1. All models control for indicators for state and year*month, in addition to the state unemployment rate, state real per capita personal income, log of total state population, log of the agency population for months with arrest reports, log state criminal justice expenditures, log of full time equivalent number of police officers, state minimum wage, state poverty rate, and percent of the population covered by reporting agencies. Measures of male arrests include the log of the male arrests for all criminal offenses (ages 18–49) and the log of male arrests for property index crimes (relevant age group). Lagged covariates include 1-year lags of the state unemployment rate, real personal income per capita, and welfare caseloads (total and child-only caseloads). Sample is limited to agencies with a reported coverage of at least 50%. For the tests of equality in the effects across age groups, Chi-squared statistic is reported with the corresponding p-value in square brackets below.

Effects for the youngest women (ages 18–20) are extremely small (.09%) and not statistically significant; those for the next youngest group, women ages 21 to 24, are small and statistically insignificant as well. Most women under age 25 have not yet had children and therefore would have been ineligible for welfare and not directly affected by welfare reform. In other words, the “treated” population among women ages 18 to 20 and 21 to 24 is likely to be small, and thus the average policy response over this age group would be muted.¹⁴ As expected, we find substantially stronger effects for women in their mid-to-late 20s, indicating a welfare reform-induced decline in property crime of about 4.4%. The larger effect for this age group was expected because these women were at peak childbearing ages (hence the treated population in this group was relatively large) and faced long time horizons potentially subject to the new welfare regime of time-limited benefits and no possibility of long-term welfare reliance. In Model 5 we aggregated all women in their 20s for comparison with later estimates. Expectedly, the effect magnitude is an average of those for women ages 21 to 24 and a 25 to 29 and indicates a statistically significant mean decline in property crime attributed to welfare reform of approximately 3.8%.

For women in their 30s (Model 8), welfare reform may have had some negative effect of welfare reform on property crime, but the estimate is not statistically significant. This potential effect appears to be driven mostly by women in their early 30s (Model 6) and suggests a potential welfare reform-associated decrease in property crime of ~3.6%. The largest effect sizes of welfare reform were for women ages 40 to 49, suggesting a decrease in property crime arrests of ~6.1%. This is a surprising result, but it is important to consider that women still committing crimes in their 40s constitute a small and anomalous group. In other words, since arrests for women in their 40s are so low (by about 25–30% compared to women in their 20s; see Table 1), the implied absolute decrease in property crime is also small. Thus, the larger relative decline (5.9% vs. 3.7%) among older women does not translate into a larger absolute decline. As such, not too much should be read into this particular finding.

We perform chi-square tests to assess differences in estimates across the age distribution and can strongly reject the null hypothesis that the estimated effects are equivalent across all age groups (see chi-square and p values reported in Table 3). When specifically testing each of the age groups against the overall mean effect across the entire age distribution (18–49 years), large standard errors and overlapping confidence intervals limit power. We are not able to reject the null hypothesis of equal effects for women in their 20s, early 30s, and 40s. Overall, the estimates in Table 3 suggest that welfare reform may have steepened the age-crime gradient for women in their 20s and 30s.¹⁵

In Table 4, we consider potential heterogeneity in the effects of welfare reform on property crime by the strength of the work incentives embedded in each state’s welfare system post-reform. Given that we found no evidence of any effects of welfare reform on property crime for the 18 to 20-year-old age group, we confine these and subsequent analyzes to women ages 21 and older. As described earlier, we follow the classification scheme developed by Blank and Schmidt (2001) and followed in prior work (Corman et al., 2014; Dave et al., 2011, 2012), which categorizes states as having strong, mixed/medium, or weak work incentives based on state variation in time limits, benefits generosity, earnings disregard, and sanctions. As before, we estimate differential effects of welfare reform across the age distribution, but now interact the welfare reform indicator with whether the state’s welfare regime contains strong or medium versus weak work incentives (Models 1–4).

Table 4.

Effects of Welfare Reform on Property Crime. FBI Arrests, Women, 1990–2002. Differential Effects across Age Distribution and Work Incentives.

Outcome	Ln arrests for property index crimes
Specification	1	2	3	4	5	6	7	8
Age group (years)	21–49	21–29	30–39	40–49	21–49	21–29	30–39	40–49
Any welfare reform	−0.0163 (0.0235)	0.0168 (0.0289)	−0.0088 (0.0376)	−0.0115 (0.0437)	_	_	_	_
Any welfare reform*Strict/Medium work incentives	−0.0280 (0.0288)	−0.0540 (0.0346)	−0.0205 (0.0417)	−0.0498 (0.0522)	_	_	_	_
Any welfare reform*Change in caseloads-low	_	_	_	_	−0.0270 (0.0263)	−0.0370 (0.0292)	−0.0087 (0.0284)	−0.0334 (0.0415)
Any welfare reform*Change in caseloads-medium	_	_	_	_	−0.0204 (0.0239)	−0.0131 (0.0291)	−0.0255 (0.0261)	−0.0587 (0.0377)
Any welfare reform*Change in caseloads-high	_	_	_	_	−0.0570** (0.0257)	−0.0479 (0.0316)	−0.0250 (0.0284)	−0.0673* (0.0372)
State indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year*Month indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Measures of male arrests	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Lagged economic conditions and welfare caseloads	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Equality of WR*Incentives effect across age groups	2.77 [0.428]				9.52** [0.023]
Equality of WR*Incentives effect across given age group and average effect (ages 21–49)	—	2.15 [0.143]	0.08 [0.774]	0.20 [0.658]	—	0.29 [0.593]	5.30** [0.021]	0.17 [0.676]
Observations (state/month/year)	6298	6298	6298	6298	6298	6298	6298	6298

Note. Asterisks denote significance as follows: ***p ≤ .01; **.01 < p ≤ .05; *.05 < p ≤ .10. Coefficients from OLS semi-log models are presented, weighted by the state’s female population ages 21 to 49. Standard errors are adjusted for arbitrary correlation within state cells, and reported in parentheses. The regression models are based on Specification 3 from Table 1. All models control for indicators for state and year*month, in addition to the state unemployment rate, state real per capita personal income, log of total state population, log of the agency population for months with arrest reports, log state criminal justice expenditures, log of full-time equivalent number of police officers, state minimum wage, state poverty rate, and percent of the population covered by reporting agencies. We also allow the year effects to differ across states with weak vs. strict/medium work incentives (Models 1–4) and across states with low/medium/high caseload change (Models 5–8). Measures of male arrests include the log of the male arrests for all criminal offenses (ages 21–49) and the log of male arrests for property index crimes (relevant age group). Lagged covariates include one-year lags of the state unemployment rate, real personal income per capita, and welfare caseloads (total and child-only caseloads). Sample is limited to agencies with a reported coverage of at least 50%. States are classified as having strict or medium (relative to weak) work incentives, based on Blank and Schmidt (2001). States are classified as low, medium, or high in terms of change in caseloads, based on terciles of the percentage decrease in caseloads between 1990 and 2002. For the tests of equality in the effects across age groups, Chi-squared statistic is reported with the corresponding p-value in square brackets below.

For the broadest age group, 21 to 49, welfare reform led to about a 1.6% decrease in property crime among states with weak work incentives; states with stronger incentives and more pro-employment features experienced an additional decline of 2.8%, for a total decrease of about 4.4%.¹⁶ Overall, this pattern is validating of a dose-response check; if it is indeed welfare reform and its pro-employment push that was driving the decline in women’s crime, then the effects should be stronger in states with regimes that incorporated stronger work incentives—and it is.

This pattern persists when we disaggregate by age: ages 21 to 29 (Model 2), ages 30 to 39 (Model 3), and ages 40 to 49 (Model 4). As before, we find stronger effects of welfare reform on property crime arrests in the youngest group (and in the oldest but anomalous group), but now see that these effects appeared to be concentrated in states with relatively stronger work incentives. These estimates again suggest a steepening of the age-crime gradient;¹⁷ however; while the magnitudes are sizable (combining the coefficients of the welfare reform indicator and the interaction term, the estimates suggest that welfare reform led to 3.7, 2.9, and 6.1% declines in property crime among women in their 20s, 30s, and 40s, respectively), they do not pass the threshold of statistical significance.

In Models 5–8, we assess these patterns in an alternative manner, by directly classifying states that experienced larger versus smaller declines in welfare caseloads over the sample period. Specifically, we interact the welfare reform indicator with indicators for whether the states were in the lowest, middle, or top tercile in terms of caseload decline. For the broadest age group, we find that reform is associated with a decrease in property crime among all three sets of states; however, the negative effect is larger among states that experienced the largest decrease in caseloads, about 5.7% relative to 2–2.7% for the other states. This pattern continues as we explore heterogeneity across the age groups. The largest declines, as before, are for women in their 20s (4.8%) and in their 40s (6.7%), particularly in states that experienced the largest decrease in caseloads. However, again it is important to note that most of the comparisons in Table 4 are imprecise due to the parsing of estimates across the age gradient and levels of work incentives and should therefore be viewed only as suggestive. That said, based on the chi-square test, we are able to strongly reject the null hypothesis that the effects are equal across all age groups when considering heterogeneity based on the change in welfare caseloads.

The effects of welfare reform may be further moderated by the state’s criminal justice system. Table 5 presents models that explore heterogeneity based on whether the state had high or low (relative to the median) per capital criminal justice spending (Models 1–4) and high or low (relative to the median) per capita full-time equivalent number of people employed in the criminal justice system (Models 5–8). In order to circumvent endogenous sample selection, we classify states as above/below the median based on levels in 1990, which predated welfare reform. Across all models and age groups, we find consistent evidence that welfare reform led to a decrease in property crime; however, these effects materialized mainly among states that had relatively high levels of criminal justice spending and staffing.¹⁸ For both sets of heterogeneity analyzes, we can reject (at the 1% significance level) the null hypothesis of equal effect sizes across all age groups. Depending on the measure of the state’s criminal justice system, we find statistically significant differences in the response for women in their 20s in states with high criminal justice spending (relative to the average effect) and for women in their 30s in states with high full-time employment in the criminal justice system (relative to the average effect).

Table 5.

Effects of Welfare Reform on Property Crime. FBI Arrests, Women, 1990–2002. Differential Effects across Age Distribution and Criminal Justice Spending.

Outcome	Ln arrests for property index crimes
Specification	1	2	3	4	5	6	7	8
Age group (years)	21–49	21–29	30–39	40–49	21–49	21–29	30–39	40–49
Any welfare reform	0.0108	0.0491**	0.0332	−0.0058	−0.0085	0.0018	0.0354	0.0001
	(0.0203)	(0.0227)	(0.0236)	(0.0356)	(0.0225)	(0.0289)	(0.0215)	(0.0376)
Any welfare reform*Criminal justice spending-high	−0.0987*** (0.0251)	−0.1583*** (0.0285)	−0.1053*** (0.0334)	−0.0980* (0.0507)	—	—	—	—
Any welfare reform*Criminal justice FTE-high	—	—	—	—	−0.0612** (0.0291)	−0.0687* (0.0403)	−0.1063*** (0.0315)	−0.1062* (0.0541)
State indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year*Month indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Measures of male arrests	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Lagged economic conditions and welfare caseloads	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Equality of WR*CJ-high effect across age groups	19.85*** [0.000]				13.99*** [0.003]
Equality of WR*CJ-high effect across given age group and average effect (Ages 21–49)	—	15.81*** [0.000]	0.14 [0.703]	0.00 [0.984]	—	0.15 [0.699]	9.94** [0.002]	1.53 [0.216]
Observations (state/month/year)	6298	6298	6298	6298	6298	6298	6298	6298

Note. Asterisks denote significance as follows: ***p ≤ .01; **.01 < p ≤ .05; *.05 < p ≤ .10. Coefficients from OLS semi-log models are presented, weighted by the state’s female population ages 21 to 49. Standard errors are adjusted for arbitrary correlation within state cells, and reported in parentheses. The regression models are based on Specification 3 from Table 1. All models control for indicators for state and year*month, in addition to the state unemployment rate, state real per capita personal income, log of total state population, log of the agency population for months with arrest reports, log state criminal justice expenditures, log of full-time equivalent number of police officers, state minimum wage, state poverty rate, and percent of the population covered by reporting agencies. Measures of male arrests include the log of the male arrests for all criminal offenses (ages 21–49) and the log of male arrests for property index crimes (relevant age group). Lagged covariates include one-year lags of the state unemployment rate, real personal income per capita, and welfare caseloads (total and child-only caseloads). Sample is limited to agencies with a reported coverage of at least 50%. States are classified as having high criminal justice spending or full-time employment (FTE) if they are above the median in per capita terms during the baseline period of the sample (1990). For the tests of equality in the effects across age groups, Chi-squared statistic is reported with the corresponding p-value in square brackets below.

Furthermore, unlike before, we find that welfare reform led to a significant decline in property crime arrests among women ages 30 to 39, but the decrease was, again, concentrated in states with larger criminal justice systems at baseline. These estimates suggest that any disincentives to engage in criminal behavior provided by welfare reform were possibly reinforced by any additional incentives to refrain from crime in states with larger criminal justice systems. This interactive effect is present for all age groups. While these patterns are suggestive, they warrant further study since states with larger criminal justice systems in 1990 also tended to have higher levels of crime in 1990, including women’s property crime. Hence, these effects also suggest that welfare reform led to larger declines in women’s property crime in states that had higher levels of crime at baseline.

In Table 6, we considered potential effects of welfare reform on men’s property crime arrests as a placebo check. Men are generally not eligible for welfare, and thus would not be directly impacted by the reforms. Any effects would be indirect (through effects on female household members) and would be second- or third-order effects. Hence, we would expect much smaller, and possibly nil, effects of welfare reform on property crime arrests among men. As expected, we found no evidence of welfare-reform associated decreases in property crime among men for any age group.

Table 6.

Effects of Welfare Reform on Property Crime. FBI Arrests, Men, 1990–2002.

Outcome	Ln arrests for property index crimes
Specification	1	2	3	4	5	6	7	8
Age group (years)	21–49	21–29	30–39	40–49	21–49	21–29	30–39	40–49
Any welfare reform	0.0206 (0.0199)	0.0170 (0.0190)	0.0166 (0.0266)	0.0456* (0.0252)	0.0114 (0.0248)	0.0070 (0.0249)	0.0119 (0.0281)	0.0205 (0.0225)
State indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year*Month indicators	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Measures of male arrests	No	No	No	No	No	No	No	No
Lagged economic conditions and welfare caseloads	No	No	No	No	Yes	Yes	Yes	Yes
State-specific linear rends	No	No	No	No	Yes	Yes	Yes	Yes
Observations (state/month/year)	6298	6298	6298	6298	6298	6298	6298	6298

Note. Asterisks denote significance as follows: ***p ≤ .01; **.01 < p ≤ .05; *.05 < p ≤ .10. Coefficients from OLS semi–log models are presented, weighted by the state’s female population ages 21 to 49. Standard errors are adjusted for arbitrary correlation within state cells, and reported in parentheses. Models 1 through 4 are based on Specification 1 from Table 1, with the exclusion of male arrests. Models 5 through 8 are based on Specification 6 from Table 1, with the exclusion of male arrests. All models control for indicators for state and year*month, in addition to the state unemployment rate, state real per capita personal income, log of total state population, log of the agency population for months with arrest reports, log state criminal justice expenditures, log of full-time equivalent number of police officers, state minimum wage, state poverty rate, and percent of the population covered by reporting agencies. Lagged covariates include one-year lags of the state unemployment rate, real personal income per capita, and welfare caseloads (total and child-only caseloads). Sample is limited to agencies with a reported coverage of at least 50%.

Additional Specification and Robustness Checks

We implemented several additional checks in order to verify that the results are robust (results not shown). First, we tested alternative functional forms and model specifications that: (1) expressed the outcome as a rate, as the natural log of the arrest rate (per 1000 women in the relevant age group); (2) changed the outcome to a logistic transformation based on the natural log of the odds of the arrest rate; (3) utilized non-logged measure of the arrest rate or total arrests as outcomes; and (4) extended the models to include state-specific quadratic time trends.

Second, we added robbery arrests to property index crime arrests in order to include all serious crimes with an economic motive. Estimates were very similar to those presented in Tables 2 through 6, likely because women’s robbery arrests are rare compared to the traditional property crime category.

Third, the analyzes were insensitive to aggregating the monthly arrest data to the annual level. While annual aggregation may smooth out some of the noise in the monthly data as well as seasonal factors, it also results in loss of meaningful variation from month-specific implementation of AFDC waivers and TANF. Nevertheless, the magnitudes of the effects, patterns, and inferences were unchanged when utilizing state-annual aggregates.

Finally, we re-estimated all models without utilizing population weights and confirmed that our estimates were not substantially different across the weighted and unweighted models. We did gain some efficiency, however, from the use of the population weights and the standard errors were lower in some cases.

Estimates in Context

Given established patterns of steep declines in criminal behavior with age, our estimates unlikely capture the full potential reductions in women’s property crime that could be plausibly be attributed to welfare reform, particularly for women in the 25 to 29-year age range. For example, if welfare reform increased desistance among women in their 20s, for whom we found evidence of a reduction in arrests of about 4% (for those in their mid-to-late 20s), we would expect the life course trajectories in crime to be dampened for that cohort. We thus project the potential overall effect of welfare reform for this group, taking into consideration potential compounding effects over the life course, as described below. This exercise is intended to illustrate potential compounding effects, given what we know about age gradients in crime and desistence. It uses aggregated data to make inferences about individual behavior, ignores potential incarceration effects, assumes that the direct effects of welfare reform persist throughout the woman’s adult life, and assumes that the effect sizes do not vary by cohort.

If the direct effect of welfare reform persists (we estimated a 2% reduction for women in their 30s), there will be a smaller criminal cohort of women in their 30s (owing to the greater share who would have desisted in their 20s), on the order of about 3%. To arrive at this figure, we assumed (based on Table 1) that about 78% of the mid-to-late 20s cohort would have persisted into their 30s. That is, the number of women engaging in crime would have been 78% of the original cohort in the absence of welfare reform, but now this number is reduced by 4% of the 74%, or by about 3%. We thus infer that not only did welfare reform lead to 4% fewer arrests of women in their mid-to-late 20s, but that is also may have led to 5% fewer arrests of these women in their 30s.

This exercise provides suggestive evidence that the overall effects of welfare reform have been substantially larger than what is suggested by our short-run estimates. However, these calculations must be viewed as speculative since our assumptions about desistence were based on age patterns in cross-sectional data.

Conclusion

This study explored how a major public policy change—implementation of welfare reform in the U.S. in the 1990s—shaped the age gradient in women’s property crime arrests, both overall and by different state-level welfare and criminal justice policy contexts. We found that women’s property crime arrest rates declined over the age span and that welfare reform led to an overall reduction in women’s property crime arrests of about 4%, with stronger effects for women ages 25 to 29. The effects were slightly stronger effects in states with stricter work incentives and were much stronger in states with high levels of per capita criminal justice expenditures and staffing, for all age cohorts.

The results from this study provide empirical support for the theoretical literature on economic marginalization and women’s crime by quantifying the effects of an exogenous, large-scale, and permanent change in economic opportunity structures faced by poor women on economic crime and how those effects varied by when during the life course (i.e., age) women were confronted with that change. The effects we found for women in their mid-to-late 20s were expected, given evidence that employment and intentions to work in the legal sector attenuate criminal behavior, welfare reform dramatically increased employment of mothers with young children, women in their mid-to-late 20s are at peak ages for having young children, and the literature on women’s economic marginalization suggests that taking care of children is an important factor in women’s economic crime.

The findings also contribute to the literature on the effects of turning points in criminal careers by focusing on the effects of a large societal change, as opposed to changes in individuals’ circumstances. Although we considered patterns of crime by age rather than changes in individuals’ criminal behavior over time (as have many previous studies of the lifecycle of crime, owing to lack of individual-level longitudinal data with sufficient sample sizes and measures of crime) and therefore cannot make strong inferences about individual’s criminal trajectories from our data, the findings from our study—in conjunction with what is known about the life cycle of crime—suggest that welfare reform affected women’s decisions to desist from crime altogether. That said, we note the lack of individual-level data on crime trajectories as a limitation of our study. Other limitations that are the findings cannot be generalized to other substantial macro-level events, such as the Great Recession in the late 2000s, or to men; our focus was on a specific (but serious, relatively prevalent, and relevant) category of crime; and we used arrests as a proxy for criminal behavior (in keeping with much of the literature), so it is possible that the findings reflect welfare-reform associated changes in the probability of arrest rather than underlying changes in behavior.

The results from this study also add to the growing knowledge about the indirect or unintended (that is, not directly intended) effects of welfare reform in the U.S. The findings strongly suggest that welfare reform has conferred societal benefits in terms of a reduction in women’s property crime that is not only likely to persist (because welfare reform is still in effect), but is also likely to compound in the future (owing to the strong effects for relatively young women). That said, this socially desirable outcome must be considered in light of the strong economy during the period studied and in conjunction with the effects of welfare reform on other relevant outcomes—both for mothers and their children—before coming to the general conclusion that society is better off as a result of this major policy shift.

Footnotes

Acknowledgements

The authors are grateful to Erik Adamcik for excellent research assistance.

Authors’ Note

Key contributions are the investigation of effects of an exogenous, large-scale, and permanent change in opportunity structures faced by poor women on economic crime and how those effects varied by age; consideration of effects of a broad societal change as a potential turning point in criminal careers; and identification of an indirect effect of welfare reform in the U.S.— a reduction in women’s property crime that is not only likely to persist (because welfare reform is still in effect), but is also likely to compound over time (owing to strong effects for relatively young women).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by a grant to Hope Corman at Rider University by the Charles Koch Foundation. It was also supported by the National Center for Advancing Translational Sciences, a component of the National Institutes of Health under award number UL1TR003017; the U.S. Department of Health and Human Services/Health Resources and Service Administration under award number U3DMD32755; and the Robert Wood Johnson Foundation through its support of the Child Health Institute of New Jersey (Grant 74260).

ORCID iD

Hope Corman

Notes

Author Biographies

Hope Corman is Professor Emerita of Economics at Rider University in and Research Associate at the National Bureau of Economic Research. Her research interests include economics of criminal behavior, and effects of public policies on crime.

Dhaval M. Dave is Stanton Research Professor in Economics at Bentley University and Research Associate at the National Bureau of Economic Research. His research interests lie in program evaluation at the intersection of health and labor economics, with a focus on crime, health behaviors and outcomes, and human capital.

Nancy E. Reichman is a Professor of Pediatrics at Rutgers University, Robert Wood Johnson Medical School and Visiting Professor of Economics at Princeton University. Her research focuses on maternal and child health, socioeconomic status and health, and effects of public policies and other contexts on child and family well-being.

References

Adams

Rohacek

(2002). Child care and welfare reform. In Weil

Finegold

(Eds.), Welfare Reform: The Next Act (pp. 21–37). The Urban Institute.

Angrist

J. D.

Pischke

J. S.

(2014). Mastering ‘metrics: The path from cause to effect. Princeton University Press.

Becker

G. S.

(1968). Crime and punishment: An economic approach. Journal of Political Economy, 76(2), 169–217.

Blank

R. M.

(2002). Evaluating welfare reform in the United States. Journal of Economic Literature, 40(4), 1105–1166.

Blank

R. M.

Schmidt

(2001). Work, wages, and welfare. In Blank

R. M.

Haskins

(Eds.), The new world of welfare (pp. 70–102). Brookings Institution Press.

Bushway

S. D.

Reuter

(2011). Labor markets and crime. In Wilson

J. Q.

Petersilis

(Eds.), Crime and public policy (pp. 183–209). Oxford University Press.

Campaniello

(2014). Women in crime. IZA world of labor 2014: 105. https://wol.iza.org/uploads/articles/105/pdfs/women-in-crime.pdf

Campaniello

Gavrilova

(2018). Uncovering the gender participation gap in crime. European Economic Review, 109(C), 289–304.

Carlen

(1988). Women, crime, and poverty. Open University Press.

10.

Colman

Dave

(2016). Econometric methods in health and human biology. In Komlos

Kelly

(Eds.), Oxford handbook of economics and human biology. Oxford University Press.

11.

Corman

Dave

Reichman

N. E.

(2014). Effects of welfare reform on women’s crime. International Review of Law and Economics, 40, 1–14.

12.

Corman

Noonan

Reichman

N. E.

Schwartz-Soicher

(2011). Life shocks and crime: A test of the “turning point” hypothesis. Demography, 48(3), 1177–1202.

13.

Dave

D. M.

Corman

Reichman

N. E.

(2012). Effects of welfare reform on education acquisition of adult women. Journal of Labor Research, 33(2), 251–282.

14.

Dave

D. M.

Corman

Kalil

Schwartz-Soicher

Reichman

N. E.

(2020). Intergenerational effects of welfare reform: Adolescent delinquent and risky behaviors. Economic Inquiry, 59(1), 199–216.

15.

Dave

D. M.

Decker

S. L.

Kaestner

Simon

K. I.

(2010). The effect of Medicaid expansions on the health insurance coverage of pregnant women: An analysis using deliveries. INQUIRY: The Journal of Health Care Organization, Provision, and Financing, 47(4), 315–330.

16.

Dave

D. M.

Reichman

N. E.

Corman

Das

(2011). Effects of welfare reform on vocational education and training. Economics of Education Review, 30(6), 1399–1415.

17.

Ehrlich

(1973). Participation in illegitimate activities: A theoretical and empirical investigation. Journal of Political Economy, 81(3), 521–565.

18.

Eide

(2000). Economics of criminal behavior. In Bouckaert

De Geest

(Eds.), Encyclopedia of law and economics (pp. 345–389). Edward Elgar.

19.

Fang

Keane

M. P.

(2004). Assessing the impact of welfare reform on single mothers. Brookings Papers on Economic Activity, 2004(1), 1–116. Retrieved April 23, 2017 from https://www.brookings.edu/bpea-articles/assessing-the-impact-of-welfare-reform-on-single-mothers/

20.

Gould

E. D.

Weinberg

B. A.

Mustard

D. B.

(2002). Crime rates and local labor market opportunities in the United States: 1979–1997. Review of Economics and Statistics, 84(1), 45–61.

21.

Gove

Hughes

Geerken

(1985). Are uniform crime reports a valid indicator of the index crimes? An affirmative answer with minor qualifications. Criminology, 23(3), 451–502.

22.

Grogger

Karoly

(2005). Welfare reform: Effects of a decade of change. Harvard University Press.

23.

Heimer

(2000). Changes in the gender gap in crime and women’s economic marginalization. Criminal Justice, 1, 427–483.

24.

Holtfreter

Reisig

M. D.

Morash

(2004). Poverty, state capital, and recidivism among women offenders. Criminology & Public Policy, 3(2), 185–208. https://doi.org/10.1111/j.1745–9133.2004.tb00035.x

25.

Horney

Osgood

Marshall

I. H.

(1995). Criminal careers in the short-term: intra-individual variability in crime and its relation to local life circumstances. American Sociological Review, 60, 655–673.

26.

Laub

J. H.

Nagin

D. S.

Sampson

R. J.

(1998). Trajectories of change in criminal offending: Good marriages and the desistance process. American Sociological Review, 63, 225–238.

27.

Laub

J. H.

Sampson

R. J.

(2020). Life-course and developmental criminology: looking back, moving forward—ASC division of developmental and life–course criminology inaugural David P. Farrington Lecture, 2017. Journal of Developmental and Life–Course Criminology, 6(2), 158–171. https://doi.org/10.1007/s40865-019-00110-x

28.

Levitt

S. D.

(1998). Why do increased arrest rates appear to reduce crime: Deterrence, incapacitation, or measurement error? Economic Inquiry, 36(3), 353–372.

29.

Loeber

Farrington

D. P.

Petechuk

(2013). From juvenile delinquency to young adult offending (Study Group on the Transitions Between Juvenile Delinquency and Adult Crime). The National Institute of Justice, Office of Justice Programs, U.S. Department of Justice.

30.

Lopoo

L. M.

Raissian

K. M.

(2012). Natalist policies in the United States. Journal of Policy Analysis and Management, 31(4), 905–946.

31.

Loprest

P. J.

(2012). How has the TANF caseload changed over time? Urban Institute. Retrieved April 23, 2017, from http://www.urban.org/UploadedPDF/412565-How-Has-the-TANF-Caseload-Changed-Over-Time.pdf

32.

Maltz

M. D.

(1999). Bridging gaps in police crime data: A discussion paper from the BJS fellows program. U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

33.

Maltz

M. D.

Targonski

(2002). A note on the use of county-level UCR data. Journal of Quantitative Criminology, 18(3), 297–318.

34.

McIvor

Murray

Jamieson

(2004). Desistance from crime: Is it different for women and girls (pp. 181–197). After Crime and Punishment: Pathways to Offender Reintegration.

35.

Meyer

Floyd

(2020). Cash Assistance Should Reach Millions More Families to Lessen Hardship: Families’ Access Limited by Policies Rooted in Racism. Retrieved January 22, 2021, from https://www.cbpp.org/research/family-income-support/cash-assistance-should-reach-millions-more-families-to-lessen

36.

Nagin

D. S.

(1998). Criminal deterrence research at the outset of the twenty-first century. Crime and Justice, 23, 1–42.

37.

Nagin

D. S.

(2013). Deterrence in the twenty-first century. Crime and Justice, 42(1), 199–263.

38.

Nguyen

Loughran

T. A.

(2018). On the measurement and identification of turning points in criminology. Annual Review of Criminology, 1, 335–358.

39.

Sampson

R. J.

Laub

J. H.

(1990). Crime and deviance over the life course: The salience of adult social bonds. American Sociological Review, 55, 609–627.

40.

Sampson

R. J.

Laub

J. H.

(2016). Turning points and the future of life-course criminology: Reflections on the 1986 criminal careers report. Journal of Research in Crime and Delinquency, 53(3), 321–335.

41.

Sampson

R. J.

Laub

J. H.

Wimer

(2006). Does marriage reduce crime? A counterfactual approach to within-individual causal effects. Criminology, 44, 465–508.

42.

Schoeni

R. F.

Blank

(2000). What has welfare reform accomplished? Impacts on welfare participation, employment, income, poverty, and family structure (Working Paper #7627). National Bureau of Economic Research.

43.

Snyder

H. N.

Cooper

A. D.

Mulako-Wangota

(2018). Bureau of justice Statistics (U.S. Arrest Estimates: Arrest Rates of Female Adults). Generated using the Arrest Data Analysis Tool. Retrieved January 28, 2018, from www.bjs.gov

44.

Steffensmeier

(1993). National trends in female arrests, 1960?1990: Assessment and recommendations for research. Journal of Quantitative Criminology, 9(4), 411–441. https://doi.org/10.1007/bf01064111

45.

Steffensmeier

Haynie

(2000). Gender, structural disadvantage, and urban crime: Do macrosocial variables also explain female offending rates? Criminology, 38(2), 403–438. https://doi.org/10.1111/j.1745-9125.2000.tb00895.x

46.

Tanner

Davies

O’Grady

(1999). Whatever happened to yesterday’s rebels? Longitudinal effects of youth delinquency on education and employment. Social Problems, 46(2), 250–274.

47.

U.S. Department of Health and Human Services. (1999). State implementation of major changes to welfare policies, 1992–1998. Office of the Assistant Secretary for Planning and Evaluation. Retrieved February 29, 2012, from http://aspe.hhs.gov/hsp/waiver-policies99/Table_A.htm

48.

U.S. Department of Justice, Bureau of Justice Statistics. (2017). Specific tables were “Justice system expenditure, by character, State, and type of government” and “ Justice system expenditure, by character, State, and type of government”. Retrieved March 21, 2017, from https://www.bjs.gov/index.cfm?ty=pbse&sid=33

49.

U.S. Department of Labor, Women’s Bureau. (2020). Labor force participation of mother by age of youngest child. Retrieved November 3, 2020, from https://www.dol.gov/agencies/wb/stats/labor-force-participation-youngest-child

50.

Uggen

(2000). Work as a turning point in the life course of criminals: A duration model of age, employment, and recidivism. American Sociological Review, 65(4), 529–546.

51.

Ulmer

J. T.

Steffensmeier

(2014). The age and crime relationship: Social variation, social explanations. In Beaver

Boutwell

Barnes

J. C.

(Eds.), The nurture versus biosocial debate in criminology (pp. 377–396). SAGE. https://www-sagepub-com-s.web.bisu.edu.cn/sites/default/files/upm-binaries/60294_Chapter_23.pdf

52.

Verbruggen

Blokland

A. A.

Van der Geest

V. R.

(2012). Effects of employment and unemployment on serious offending in a high-risk sample of men and women from ages 18 to 32 in the Netherlands. British Journal of Criminology, 52(5), 845–869.

53.

Wehby

G. L.

Dave

D. M.

Kaestner

(2020). Effects of the minimum wage on infant health. Journal of Policy Analysis and Management, 39(2), 411–443.

54.

Ziliak

J. P.

(2016). Temporary assistance for needy families. In Moffitt

R. A.

(Ed.), Economics of means-tested transfer programs in the United States (volume 1, pp. 303–393). University of Chicago Press.