Introduction

The credibility revolution in empirical economics has produced a body of causal evidence on policy questions that was simply unavailable to researchers twenty years ago. In health, labour, and criminal justice, the accumulation of well-identified studies has increasingly shifted the debate from "is there an effect?" to "how large is the effect, for whom, and through what mechanisms?" This issue's Results section surveys five studies that illustrate the quality and range of evidence that quasi-experimental methods can produce.

Paper 1: Finkelstein et al.\ (2012) — Oregon Health Insurance Experiment

The Paper

Finkelstein et al.(2012) — "The Oregon Health Insurance Experiment: Evidence from the First Year," Quarterly Journal of Economics 127(3):1057–1106.

Identification Strategy

The state of Oregon expanded its Medicaid programme in 2008 by lottery, randomly selecting approximately 30,000 individuals from a waiting list of 90,000 to be allowed to apply. The lottery provides random variation in access to Medicaid, which serves as an instrument for actual coverage. The IV estimand is the LATE for lottery-induced Medicaid enrollees (compliers). The first stage is approximately 0.25: winning the lottery raised the probability of Medicaid coverage by about 25 percentage points.

Main Results

After the first year of coverage, Medicaid led to: a 35% increase in the probability of having a primary care visit; a 15 percentage point reduction in the probability of having an unpaid medical bill in collections; and a 9 percentage point reduction in the probability of screening positive for depression. There was no statistically significant effect on blood pressure, cholesterol, or blood glucose in the two-year follow-up study (Baicker et al.(2013)).

Contribution

The Oregon experiment is the only large-scale randomised evidence on the effects of Medicaid for adults. It established that the primary benefits of coverage over a two-year horizon are financial protection and mental health, while leaving open the question of long-run physical health effects. It triggered a large subsequent literature using DiD designs to study the ACA Medicaid expansion.

Paper 2: Miller, Johnson, and Wherry (2021) — Medicaid and Mortality

The Paper

Miller et al.(2021) — "Medicaid and Mortality: New Evidence from Linked Survey and Administrative Data," Quarterly Journal of Economics 136(3):1783–1829.

Identification Strategy

The paper exploits the ACA Medicaid expansion of 2014, which extended eligibility to adults with incomes up to 138% of the federal poverty line in states that chose to expand. Using a difference-in-differences design comparing expansion and non-expansion states, linked to longitudinal survey and mortality data, the paper estimates the effect of Medicaid eligibility on all-cause mortality.

Main Results

Medicaid eligibility under the ACA expansion reduced all-cause mortality among newly eligible adults by 0.13 percentage points per year, a 21% reduction relative to the comparison group mortality rate. The effects are concentrated among adults aged 55–64 and are driven by reductions in mortality from diseases amenable to healthcare intervention. These findings directly address the puzzle left by the Oregon experiment: over longer time horizons and in a broader population, Medicaid does reduce mortality.

Contribution

This paper is the most credible evidence to date that public health insurance reduces adult mortality in the United States. It reconciles the Oregon null results with the long-run health benefits of coverage by showing that the relevant time horizon for mortality effects exceeds two years. The DiD design, while relying on the parallel trends assumption, is well-supported by extensive pre-trend analysis and robustness checks.

Paper 3: Dube, Lester, and Reich (2010) — Minimum Wages Across State Borders

The Paper

Dube et al.(2010) — "Minimum Wage Effects Across State Borders: Estimates Using Contiguous Counties," Review of Economics and Statistics 92(4):945–964.

Identification Strategy

The paper uses all pairs of contiguous US counties straddling a state border where the two states had different minimum wages, covering 1990–2006. The identifying assumption is that adjacent county pairs share a common economic environment and would follow parallel trends in employment in the absence of minimum wage differences. Within-pair time fixed effects absorb local labour market trends, and state-level minimum wage changes provide the identifying variation.

Main Results

Minimum wage increases have no statistically significant negative effect on restaurant employment when identified from variation within county pairs. Point estimates are close to zero, with tight confidence intervals that rule out the large negative employment effects predicted by simple competitive models and found in earlier cross-state studies. The result is robust to controlling for observable county characteristics and to alternative definitions of the comparison window.

Contribution

The paper made the contiguous county pair design the standard approach for minimum wage research, addressing the critique that states choosing to raise their minimum wages differ systematically from states that do not. It substantially weakened the case for large negative employment effects and contributed to the emerging consensus (reinforced by Cengiz et al.(2019)) that moderate minimum wage increases have small employment effects.

Paper 4: Angrist and Evans (1998) — Children and Labour Supply

The Paper

Angrist and Evans(1998) — "Children and Their Parents' Labor Supply: Evidence from Exogenous Variation in Family Size," American Economic Review 88(3):450–477.

Identification Strategy

The paper uses two instrumental variables for family size: (i) the sex composition of the first two children — parents who had two children of the same sex are significantly more likely to have a third child; and (ii) twin births at the second birth. Both instruments satisfy the exclusion restriction under the assumption that child sex composition and twinning affect parental labour supply only through their effect on family size. The LATE identifies the effect of an additional child for parents induced to have a third child by these instruments.

Main Results

Having a third child causes a 6–9 percentage point reduction in the probability of working and a 12–15% reduction in annual earnings for mothers, with much smaller and statistically insignificant effects for fathers. The effects are driven by mothers who previously worked and are concentrated in the year of the additional birth and the years immediately following.

Contribution

The paper is a textbook example of IV using natural instruments. It demonstrated that the sex composition instrument — an instance of "twins and sex" instruments that have since been widely used — has a strong and plausible first stage, and that the resulting LATE estimates are robust across specifications. The paper influenced a generation of research on the labour supply consequences of fertility and helped establish IV as the preferred design for estimating causal effects of family size.

Paper 5: Stevenson and Wolfers (2006) — Divorce Laws and Domestic Violence

The Paper

Stevenson and Wolfers(2006) — "Bargaining in the Shadow of the Law: Divorce Laws and Family Distress," Quarterly Journal of Economics 121(1):267–288.

Identification Strategy

The paper exploits the staggered adoption of unilateral ("no-fault") divorce laws across US states between 1968 and 1988. These laws changed the legal default in divorce negotiations by allowing either spouse to initiate divorce without the consent of the other. The DiD design compares states before and after adoption to states that had not yet adopted unilateral divorce laws, identifying the causal effect of the legal change on family outcomes including domestic violence, suicide, and homicide.

Main Results

The adoption of unilateral divorce laws led to significant reductions in domestic violence, suicide among women, and spousal homicide. Female suicides fell by approximately 8–16%, domestic violence against women fell by 30%, and spousal homicide of women fell by about 10% after adoption of unilateral divorce. The effects persist over time, consistent with a permanent change in bargaining positions within marriages.

Contribution

The paper provided the most compelling causal evidence on the welfare effects of divorce law reform, using DiD to identify effects that observational comparisons of divorce rates and family outcomes could not credibly establish. It demonstrated that legal defaults in family law can have large effects on welfare, particularly for women in abusive relationships, and has become a canonical reference in the empirical literature on family economics.

Conclusion

The five papers reviewed in this section span three policy domains and four identification strategies (RCT/IV, DiD, contiguous county pairs, IV). What unites them is a commitment to using credible variation to answer well-defined causal questions. Each study has generated a substantial follow-on literature and influenced both academic debate and policy. The accumulation of such evidence — across settings, designs, and time periods — is the mechanism by which empirical social science makes progress.