Working Papers

Trade liberalization and the opioid epidemic struck the same American communities over the same years. Using restricted county-level CDC mortality microdata for 1990 to 2024 and county exposure to NAFTA’s tariff cuts, I show that NAFTA’s mortality signal falls on firearm suicide, a cause with no connection to drug supply, and not on drug overdose. A one-standard-deviation increase in exposure raises the firearm-suicide rate by close to four percent, including where prescription opioids were scarce. The margin for policy is the local labor market, not the trade barrier.

JEL: F14, F16, I12, I18, J61.

For the American poor elderly, a Supplemental Security Income claim was once the route to help with Medicare’s premiums and cost-sharing, until the 1989 to 1991 Medicare buy-in delivered that help without the claim. Aged SSI receipt fell by some seventy percent between 1975 and 2010, and the standard account explains none of its geography. Using a 1977 to 2010 county panel built partly by digitizing fourteen years of printed SSA volumes, I show that counties with more poor elderly shed aged SSI markedly faster once the buy-in took effect, a gap that reached a fifth of the receipt rate by 2010 and reflects claims never filed, a lesson for today’s auto-enrollment reforms.

JEL: H51, H55, I38, J14, F16, C23.

U.S. adolescent childbearing has fallen by roughly seventy percent since 2007, and no established economic, policy, or social factor explains much of the decline. I ask whether broadband is part of the answer, comparing counties funded by the 2009 federal broadband stimulus with counties whose applications were rejected, using restricted-use NCHS natality records. After early implementation, teen births in funded counties fell a further 3.7 to 8.6 percent, a tenth to a fifth of their overall decline, with the response concentrated among 15-to-17-year-olds. Pre-award trends are flat back to 2003.

JEL: I18, J13, I12, J18, H54, L96.

Between 1976 and 2016, most U.S. states adopted laws authorizing or requiring police to make warrantless arrests at domestic violence calls. These laws were meant to protect victims but may also deter them from calling police. Using CPS microdata on more than 6 million adults, corrected law classifications, and modern staggered difference-in-differences methods, we find no detectable effect of mandatory, preferred, or discretionary arrest provisions on marriage or divorce, and we can rule out effects as large as those of unilateral divorce reform.

JEL: J12, K36, K42, I18. Keywords: domestic violence, warrantless arrest, mandatory arrest, divorce, marriage, difference-in-differences.

We examine how political majoritarianism affects gender-based violence in India, using the 2013 rise of the Bharatiya Janata Party (BJP) to national prominence as an institutional shock. Using administrative data from 36 states and union territories for 2006 to 2019, we find that BJP-majority states experienced a 47 percent increase in gender-based violence relative to other states after 2013, with no change in non-gendered crimes. Rising domestic-violence-related female suicides, falling indictment rates, and growing acceptance of domestic violence point to weaker institutional deterrence and shifting gender norms.

JEL: D63, K42, J11, O53.

Federal broadband programs are promoted partly on rural-health grounds, and the $42 billion BEAD program makes the question pressing. I study the 2009 federal broadband stimulus with a rejected-applicant design that compares 2,448 funded counties with 263 matched counties whose applications were rejected. The award raised fiber availability by 2.8 percentage points, yet across nineteen County Health Rankings outcomes, from healthcare access to mortality and mental health, I find no robust effect, with most estimates precise nulls. The results bound the short-run health returns to broadband, an input to the cost-benefit case for BEAD.

JEL: I18, I14, H54, L96, R11.

What are rural households willing to pay for broadband? I answer through the housing market, studying the $7.2 billion 2009 broadband stimulus, the largest place-based broadband program in U.S. history. Comparing 2,448 funded counties with 263 counties that applied for the same funding but were rejected, I find that broadband funding raised rural gross rents by about 1.7 percent, with flat pre-trends and a premium that persists through 2021. The rent response provides a revealed-preference benefit measure for rural broadband investment, relevant to the ongoing BEAD program.

JEL: R31, R11, L96, H54.

Does refugee dispersal raise sexually transmitted infection in receiving communities? Most countries cannot tell, because screening of arrivals adds diagnosed cases even when no one is newly infected. England publishes tests alongside diagnoses for each local authority, which separates detection from transmission. Studying the April 2022 Full Dispersal mandate, which required every local authority to house asylum seekers, I find no change in diagnoses, testing, positivity, or symptomatic infection, and equivalence tests rule out increases above five percent. Earlier apparent effects are consistent with selection in which authorities volunteered, not transmission.

JEL: I18, J15, F22, I12.

A trade shock’s demographic footprint depends on the gender of the jobs it destroys. NAFTA’s deepest tariff cuts fell on textiles and apparel, an industry 62 percent female in 1990, so the shock fell on women’s own jobs. Using the restricted, county-identified universe of U.S. births for 1990 to 2024, extended back to 1979, I find that NAFTA’s effect on births runs through this female-intensive textile shock rather than the male-intensive manufacturing the literature has emphasized. Textile exposure raised preterm births, concentrated after 1998, while overall fertility held steady.

JEL: J13, J16, J12, I12, F16.

National trade policy is set in Washington; its social costs are borne locally. We estimate one such cost, property crime, in U.S. counties exposed to NAFTA’s tariff phase-in and to Chinese import competition. A one-standard-deviation increase in NAFTA exposure raises the county burglary rate by 5 percent of the mean, and the China shock has a statistically indistinguishable effect. The response tracks each industry’s tariff phase-in year by year and is consistent with lost local employment. These locally borne costs are absent from standard evaluations of trade policy and from federal trade-adjustment compensation.

JEL: K42, H23, H75, F14, F16, R23.

Incomplete take-up of safety-net programs is often blamed on the cost of applying, which online enrollment is meant to lower. If broadband reduces that cost, broadband expansion should raise participation. Using the 2009 federal broadband stimulus and a rejected-applicant design comparing 2,448 funded counties with 263 matched rejected applicants, I estimate precise nulls across SNAP, Medicaid-eligible insurance, the EITC, and cash assistance, ruling out SNAP effects above about 6.5 percent of baseline. The findings caution against expecting broadband programs such as ReConnect and BEAD to raise rural take-up.

JEL: H53, H75, I38, L96, R11.

Place-based program evaluations increasingly compare funded units with rejected applicants. This note shows that in such designs the first stage often erodes, because rejected applicants later obtain the treatment elsewhere. For the 2009 federal broadband stimulus, the funded-versus-rejected gap in broadband availability peaks at about 4.6 percentage points in 2016 and fades by 2021. A simple framework shows that estimates for flow outcomes inherit the fading gap, while stock outcomes build and persist, a pattern that holds across seven outcomes of the stimulus. The lesson guides outcome and window choice in rejected-applicant designs.

JEL: C21, C23, H54, R58, L96.

Federal broadband programs are increasingly promoted as public-health investments, yet whether connectivity affects deaths of despair is unknown. Comparing counties funded by the 2009 broadband stimulus with matched rejected applicants, and measuring working-age deaths of despair from NCHS restricted-use vital records for 2010 to 2020, I find a precise null: equivalence tests rule out effects larger than about six percent of baseline, with no robust change in drug overdose, alcohol-induced mortality, or suicide. Broadband infrastructure on its own did not measurably change deaths of despair in funded counties.

JEL: I12, I18, H54, R58, O18.

Between 3 and 5 million LGBTQ persons in the U.S. live with disabilities. This study is the first to examine whether living as a sexual minority affects participation in Social Security disability programs. Using the American Community Survey (2008 to 2024) and the March Current Population Survey (2008 to 2025), we find that same-sex coupled adults with disabilities are 16 to 25 percent less likely to receive Supplemental Security Income than otherwise similar adults, a gap that grows after controlling for income and household characteristics. Differences in SSDI are small, pointing to program-specific frictions.

JEL: I38, J14, J71.

Title IX’s expansion of girls’ high school athletics is widely credited with raising women’s human capital, but its welfare consequences have not been examined. Using Stevenson’s (2010) identification strategy, I show that the policy’s effects were sharply heterogeneous. For Black women and women in states with smaller boys’ programs, Title IX reduced teen fertility and raised schooling and employment. For predominantly white women in states with large boys’ programs, it raised teen births and later welfare receipt, an effect concentrated in states with restrictive pre-Roe reproductive policy.

JEL: I18, I38, J13, J16, J18. Keywords: Title IX, athletic participation, teen fertility, welfare participation, instrumental variables.

The dollars families pay out of pocket are invisible in the claims data that coverage evaluations rely on. This paper provides the first household-budget evaluation of the federal Medicaid postpartum extension, which let states lengthen pregnancy-related coverage from 60 days to 12 months. Using the Consumer Expenditure Survey Interview (2013 to 2024) and a triple difference comparing families with and without an infant, I find that the extension reduced infant households’ out-of-pocket medical spending by about $39 per quarter, roughly 17 percent. The policy reached the household budget.

JEL: I13, I18, I38, H51, D12.

Do family cash, credit, and reproductive-policy shocks change what households buy for their infants? Using Consumer Expenditure Survey Diary microdata for 2013 to 2024 and a triple-difference design, I study refundable state EITCs, state Child Tax Credits, and post-Dobbs abortion bans across baby food, infant clothing, out-of-pocket medical spending, and food away from home. A minimum-detectable-effect discipline separates precise nulls from underpowered ones: nine of twelve policy-outcome cells are precise nulls, a portfolio of credibly estimated zeros for the infant-specific basket.

JEL: I18, I38, H24, H75, J13, D12.

How large must the pass-through of a diaper or menstrual-product sales-tax exemption be before public survey data could detect it? Using state-by-quarter Consumer Expenditure Survey data and staggered difference-in-differences estimators, I compute each design’s minimum detectable effect and benchmark it against the largest change full pass-through could produce. The detectable effect is three to nineteen times that benchmark, so these data cannot measure the policy’s household benefit. The note provides that bound and a scanner-data design that can.

JEL: H22, H71, H25, D12, I18.

Food away from home is on average more calorie-dense and of lower diet quality than food prepared at home, so a shift toward it changes an infant household’s food environment, not only its budget. Using the Consumer Expenditure Survey Diary (2013 to 2024) and a triple-difference design, I find that state paid family leave is associated with an increase in infant households’ weekly spending on food away from home of about $51, three-quarters of baseline, and a decline in prepared baby food: the signature of provisioning shifting toward the market.

JEL: I12, I18, D12, D13, J13, J22.

All 50 U.S. states and D.C. have adopted anti-bullying laws, which reduce suicidal behavior among teenage girls. We document an unintended consequence: fewer suicide-based teen organ donations. Using the universe of U.S. deceased organ donations from the UNOS STAR files (1993 to 2024) and staggered adoption of the laws, we find that anti-bullying laws reduced suicide-based organ donation among girls aged 14 to 18 by 88 percent relative to the mean, with no effect for boys, mirroring the laws’ effect on suicide. The cost is small against the lives saved, but it reveals an overlooked link between suicide prevention and the transplant system.

JEL: I12, I18, J13.

Does the electoral success of a majoritarian party increase violence against historically marginalized groups? Using administrative crime data from India for 2006 to 2019 and a difference-in-differences design, we estimate the impact of the BJP’s 2014 electoral victory on crimes against Scheduled Castes. States where the BJP won majority vote shares saw an approximately 40 percent increase in atrocities against Scheduled Castes, alongside declining chargesheeting and arrest rates, pointing to weakened institutional deterrence.

JEL: D63, K42, J11, O53.

Age-based COVID-19 vaccine eligibility rules created sharp differences in access across states, ages, and dates, suggesting a natural instrument for vaccination. Combining state eligibility dates with CDC vaccination data and restricted-use NCHS mortality records by state, single year of age, and month, I show that eligibility raised first-dose coverage by 4.91 percentage points within two weeks, a strong first stage. I then assess whether the design can credibly identify mortality effects, showing why instrument relevance, exclusion, and outcome precision must be evaluated separately before reporting a causal IV estimate.

JEL: C26, I12, I18. Keywords: COVID-19 vaccination, mortality, age eligibility, instrumental variables, design feasibility.