Research

How do parental behaviors at different stages of childhood causally shape a child’s teenage outcomes? This paper estimates the causal impacts of observed parental behaviors (marriage/cohabitation, employment, welfare receipt, substance use, and criminal activity) on cognitive ability and on teenage educational attainment, substance use, and criminal behavior among fragile families. I jointly estimate a dynamic model of parental behaviors, child cognitive skill production, and teenage outcome production using data from the Fragile Families and Child Wellbeing Study. The model accounts for theoretically relevant and empirically tested exclusion restrictions, and permanent and time-varying family unobserved heterogeneity. A key finding is that parental behaviors shape teenage outcomes primarily through direct behavioral pathways rather than through cognitive skill formation, suggesting that policies aimed solely at boosting cognitive test scores are unlikely to undo the long-run damage of adverse parental trajectories: relative to more stable trajectories, sustained marital instability and economic disadvantage each increase adverse outcomes (school suspension, police contact, and early sexual initiation) by at least 8 percentage points. Three policy experiments, involving a sin tax, father-marriage promotion, and peer-environment upgrades, illustrate that sustained, multi-dimensional interventions in early childhood and adolescence offer the greatest potential for improving long-run outcomes in fragile families. Consistent with Heckman’s critical-window hypothesis, sustaining paternal co-residence through age five reduces grade failure and school suspension by at least 3.4 percentage points, an effect that vanishes once the same intervention begins only after this window has closed.

This paper studies the effect of state medical marijuana legalization on a broad set of family well-being outcomes using data from the Fragile Families and Child Wellbeing Study (FFCWS), a longitudinal cohort of largely unmarried, economically disadvantaged parents. Exploiting staggered state-level adoption of medical marijuana laws between 2001 and 2018, I estimate both a two-way fixed effects (TWFE) event-study specification and the state-clustered Callaway and Sant’Anna group-time average treatment effect estimator across eight outcome domains: mental health, substance use, criminal justice contact, labor market activity, relationship and family structure, intimate partner violence, parenting behavior, and housing and food insecurity. The significant father-side effects include a reduced likelihood of repartnering after separation, increased criminal justice contact, worsened mental health, and greater receipt of welfare. A mechanism check shows the mental health effect does not attenuate when criminal justice contact is added as a control, suggesting a channel operating independently of legal exposure and not through decriminalization. A heterogeneity check shows these effects are concentrated among fathers who had not used marijuana prior to legalization, consistent with the effects operating through legalization-induced changes in behavior, not simply pre-existing use. This pattern is evidence that medical marijuana legalization affects family well-being through a channel specific to fathers’ own exposure, not a generic household-level or state-level confound.

This paper estimates the causal effects of multidimensional school and household poverty on student achievement in Mathematics and English. Using data from the 2016/17 Young Lives school survey covering seventh- and eighth-grade students across 30 public schools in Ethiopia, we construct poverty indices at both levels using the Alkire-Foster framework and estimate value-added models combining instrumental variables with latent class analysis (IV+LCA), exploiting cross-subject scores as an instrument and recovering five discrete student ability types. Achievement is highly persistent across grades, but much of this persistence reflects stable differences in latent ability rather than genuine state dependence. The preferred IV+LCA estimates reveal an asymmetry between the two poverty dimensions: household poverty carries a robust negative effect on English achievement, while its Mathematics effect is not statistically significant, and school poverty loses significance once student composition is properly accounted for. This points to household-level constraints (limited assets, poor health, and disrupted schooling) as the primary barrier to learning, even after conditioning on school quality and prior achievement. Heterogeneous effects show that school poverty in Mathematics matters most for average-ability students and is significantly larger for girls than for boys, pointing to a gendered dimension of disadvantage that aggregate estimates obscure. Policy simulations indicate that eliminating school infrastructure deprivation yields a predicted Mathematics gain of around 0.46 points, equivalent to 6 percent of a standard deviation, while household interventions targeting asset poverty and health generate English gains of up to 0.13 points, equivalent to 1.8 percent of a standard deviation. The findings make a case for redirecting education policy attention toward household deprivations and demonstrate that multidimensional poverty measurement reveals levers that income-based measures would miss.

Cash transfer programs aim to reduce poverty, but their impact on labor supply in developing economies is debated, showing heterogeneous, neutral, or even positive effects. Scholars are moving past the classic labor-leisure theory to examine behavioral factors like liquidity constraints and gender dynamics, but the role of household indebtedness remains underexplored. In low-income settings, debt is a significant constraint that can alter how households respond to income transfers and distort both consumption and labor decisions, challenging standard economic models. This study examines how household indebtedness influences the labor supply effects of Ghana’s LEAP 1000 cash transfer program, given that beneficiaries had high baseline debt burdens than nationally representative rural populations. Using a quasi-experimental design, we find that indebted beneficiaries increase wage work and non-farm enterprise (NFE) formation but decrease part-time and agricultural employment. The effect further varies by debt type, investment and consumption debt increased NFE formation. Informal debt led to an increase in wage work due to repayment pressures. In contrast, formal debt was associated with an increase in entrepreneurship, especially among women. These findings highlight the need for debt-sensitive cash transfer policies, structured credit access, and gender-focused financial interventions to boost labor market inclusion and economic resilience.

Prior research suggests that college drinking adversely affects education. While several studies have examined the effect of alcohol on the quantity of education, this paper investigates the effect of alcohol use on the quality of education among college students. I examine the acute and chronic effects of alcohol use on college student performance (using GPA as a measure). Using data on 1,025 college students from NLSY97 enrolled between 2004 and 2009, I employ the diff-in-diff with individual fixed effects and the simple fixed effect model. The results show a significant effect of acute drinking on student performance among females but not males. Chronic alcohol use of drinking 10 days per month significantly decreases GPA by 0.07. The study reveals that a significant proportion of college students consume alcohol, and this is detrimental not just to their health but also to their academic performance, most especially among females.