Articles and work in progress

Population policies are often evaluated by their demographic effects, but coercive implementation may leave social consequences of its own. We study India’s Emergency-era sterilization campaign, a 10-month population-control program that predominantly targeted men. We combine newly digitized sterilization records with administrative crime data and exploit geographic variation in campaign exposure, measured by the growth in the share of married couples sterilized between 1975--76 and 1976--77, the year in which coercive measures were implemented. We find that high-exposure districts experienced a 14.5% increase in violent crime after the campaign, driven by increases of 24% in murder and 103% in reported rape. We find no similarly robust effects on property crime or other non-violent offenses. The age profile of perpetrators suggests that the increase in violence came not only from directly exposed cohorts but also from younger, indirectly exposed cohorts. The campaign may have increased violence by normalizing its use, altering gender norms, and harming male well-being. Consistent with this, we find that greater campaign exposure is associated with greater acceptance of intimate partner violence, lower female bargaining power and mobility, and poorer male well-being. We also examine fertility as a potential mechanism. More exposed districts experienced a modest fertility increase after the campaign, contrary to the campaign's population-control objective, but the magnitude is too small to explain the rise in violence. Together, the findings suggest that the program generated persistent social and psychological effects that may explain the long-run increase in violence.

Presented at ASREC Australasia 2023, LAGV 2023, ICDE 2023, World Cliometrics 2023, EDGE Jamboree 2023, 2023 European Winter Meeting of the Econometric Society, 2nd Diversity and Human Capital Workshop, CEPR Economic History Annual Symposium 2024

This paper studies barriers to female entrepreneurship in a developing country context. We develop a theoretical framework that rationalizes investment decisions within households and links intra-household bargaining dynamics with gender norms. We show that spouses’ incentives to invest may diverge, and that relative income shifts must be offset by power redistribution to sustain agreement. Embedding traditional gender norms into this framework, we analyze norms that increase the utility cost of female investment, potentially deterring efficient outcomes. We test the model using two empirical strategies. First, in a field experiment in India, we show that expanded women’s access to microcredit increases investment in income generating activities  and personal consumption for women from castes more supportive of women’s work, but those women also report declines in autonomy. Second, we exploit India’s 2005 GATT accession and find that the removal of textile quotas—benefiting female-intensive sectors—increased female employment in garment but worsened women’s health outcomes in caste and district contexts more supportive of women’s work. Our findings highlight how gender norms and intra-household dynamics can limit both investment and female well-being.

Presented at LAGV 2022, JMA 2022, Workshop in Gender and Family Economics 2022, LAGV 2024, NOVAFRICA 2024, AFSE 2024, ICDE 2024

Governments in low- and middle-income countries have invested heavily to expand access to health services; however, access alone may not be the binding constraint to care seeking and utilization. Effective implementation and service delivery by trained personnel can play a critical role, particularly when health outcomes depend heavily on the interaction between providers and clients. We assess whether implementation quality is an important policy margin to improve the effectiveness of health interventions in the case of family planning (FP) counseling. By means of a field experiment in urban Malawi, we exploit the random assignment of 901 women to one of six FP counselors. Using an examiner design, we construct an instrument for the quality of counseling that a woman received based on reported satisfaction and the range of topics her counselor covered on average. Higher-quality counseling reduces the number of sessions that women receive while increasing the amount of information delivered within each session. FP adoption increased sharply by 13 percentage points (p.p.) in the pooled sample and by 19 p.p. among postpartum women, largely concentrated in the adoption of injectables after childbirth. Other reproductive outcomes diverge by baseline pregnancy status: postpartum women exposed to higher-quality counseling had more pregnancies and births within 24 months, while women who were pregnant at baseline had fewer subsequent pregnancies. This divergence turns on a single counselor, and we read it cautiously, whereas the take-up effects hold when we drop any one counselor in turn. Our findings suggest that improving the quality with which FP programs are delivered may represent an important policy margin, particularly in resource-constrained settings where expanding or redesigning the program itself is costly. 

Presented at ASHEcon 2026

Family planning (FP) programs have been historically centered on averting negative reproductive outcomes, such as unintended pregnancies and unwanted births, even as these same programs call for the equal promotion of positive reproductive outcomes, in particular sexual health and well-being. In this study, we examine the causal effect of FP on sexual well-being using evidence from a randomized controlled trial in urban Malawi. Married women aged 18 to 35 who were either pregnant or had given birth within the previous six months were recruited and randomly assigned to an intervention arm or a pure control arm. Women who were assigned to the intervention arm received a multicomponent FP intervention over a two year period. On the extensive margin, intervention women initiate sex sooner but not necessarily more often: the FP intervention raises the likelihood of recent sexual activity but not coital frequency among already sexually active women. Among women who had sex in the past week, the intervention significantly increases reported enjoyment of sex. When overall sexual satisfaction is measured with a single global item asked of all women, the effect is positive but marginal, suggesting that coarser, less frequent measures may underestimate sexual well-being, and that intensive-margin measures should complement global ones to more comprehensively measure welfare, more broadly.

Family planning (FP) programs are widely promoted on the premise that they reduce unintended pregnancy, yet direct experimental evidence on this link is scarce, and the field has not converged on how unintended pregnancy should be measured. We use longitudinal data from a randomized FP trial in urban Malawi that collected both a prospective (ex-ante) and a retrospective (ex-post) measure of pregnancy intention in the same sample, along with detailed birth timing preferences. We show that a null treatment effect on a binary ex-ante measure of \emph{unwanted} pregnancy - which flags only women who wanted no more children - reflects that narrower construct rather than an absence of intervention impact. Extending the ex-ante measure to also capture mistimed pregnancies over a 24-month horizon yields a treatment effect of 3.8 percentage points (p.p.), or a 44% relative reduction, that closely reproduces an ex-post retrospective estimate of unintended pregnancy (3.7 p.p.), and concordance between the two measures rises accordingly. An otherwise identical timing-inclusive measure that is built on a 12-month horizon reproduces the null result, which we attribute to low fecundability at 12 months among our immediate postpartum and pregnant-at-baseline sample rather than to a flaw in the timing-inclusive approach itself. These results offer causal estimates of FP's effect on unintended pregnancy specifically and indicate that reconciling ex-ante and ex-post approaches to measuring it depends both on incorporating a timing dimension and on allowing a follow-up horizon long enough to observe pregnancy risk in the population of interest

We propose an indicator of contraceptive concordance that identifies the alignment between stated preferences for contraception and concurrent contraceptive behavior. Our indicator departs from traditional approaches to measurement in family planning that infer concordance to be the alignment between women's contraceptive (non-)use and their fertility preferences; indicators that are based on these approaches (e.g. unmet need) have been extensively critiqued for not reflecting women’s actual demand for contraception. We estimate our indicator of contraceptive concordance using data from a cross-sectional survey that was conducted with 1,958 married women in rural India. More than half of all women in our sample (51.2 percent) report that they are currently using a contraceptive method, with almost twice as many users reporting that they are using traditional methods relative to users who report using a modern method. More than 3 in 5 women (60.8 percent) were classified as wanting to use a contraceptive method at the time of the survey. We find that 60 percent of women in our sample are classified to be concordant (either wanted users or wanted non-users), while almost 1 in 4 women (24.8 percent) have a preference for using contraception but are not users (unwanted non-users), and 15.2 percent of women in our sample are estimated to have a preference for not using contraception but are users (unwanted users). We discuss the comparative advantages and limitations of our approach relative to traditional measures and other recently developed indicators. 

Published Paper, Replication Files, World Bank Working Paper, IZA Working Paper