384. Abortion and Crime, Revisited

Episode Summary

Executive Summary: This episode examines Steve Levitt and John Donahue’s controversial theory that legalized abortion helped drive the 1990s U.S. crime decline. It traces the original evidence, the backlash and methodological critique, a later update showing the pattern persisted, and a parallel explanation centered on lead removal. The episode also broadens into how social science should communicate causality, uncertainty, and multicausal explanations.

Main Topics: Abortion legalization and crime decline (Priority: 5/5): The core thesis argues that legalized abortion reduced the number of unwanted births, and that unwanted children were statistically more likely to commit crime later in life; thus Roe-era abortion access may have lowered crime a generation later. The original crime-drop puzzle (Priority: 5/5): Levitt’s broader attempt to explain why violent crime fell sharply after 1990 ruled out several common explanations and highlighted a set of factors that only partially explained the decline. Evidence, natural experiments, and methodological limits (Priority: 5/5): The episode details how Levitt and Donahue used state-level abortion variation, cohort analysis, and imperfect observational data to infer causality in the absence of randomized experiments. Backlash, criticism, and the coding error (Priority: 4/5): The paper drew intense criticism from pro-life and pro-choice audiences, and a later critique found a table/coding description error that weakened but did not overturn the results. Lead exposure as a parallel crime explanation (Priority: 4/5): Jessica Reyes’s research on leaded gasoline suggests environmental lead reductions also contributed materially to falling crime, supporting a multicausal view rather than a single-cause story. How science should be communicated (Priority: 4/5): Levitt and others argue that public debate should start from facts and data, not just rival hypotheses, because media and politics often flatten complex causal evidence into binary narratives. Policy implications and unwantedness (Priority: 4/5): The discussion closes by emphasizing child welfare, family support, and reducing unwanted pregnancies as more constructive policy directions than using the research as a direct abortion-policy argument.

Key Arguments: Levitt and Donahue argued that legalized abortion reduced unwanted births, and unwanted children were more likely to experience adverse life outcomes including criminal behavior. Their 2001 analysis suggested legalized abortion could explain as much as 50% of the recent U.S. crime drop, based on state-level and age-cohort patterns. A later update using data through 2014 found the same directional relationship, with high-abortion states seeing much larger crime declines than low-abortion states. The evidence is observational and imperfect, but multiple independent data patterns pointed in the same direction, strengthening the inference. A coding/description error in the original paper weakened the estimated effect but did not eliminate the core finding. Jessica Reyes’s lead research indicates that lead exposure reduction also helped drive crime down, and her results do not negate the abortion hypothesis. The episode argues that multiple causes can be true at once; crime trends are likely explained by several interacting factors, not a single dominant one. The research should not be treated as a simple pro-life or pro-choice policy argument because the moral premises differ sharply from the empirical question. If unwantedness is a major life-shaping force, policy should focus more on wanted children, parenting support, and prevention of unwanted pregnancies than on punitive crime policy alone. Public discourse often mishandles causal inference by privileging ideological identity and headline-friendly binaries over a careful reading of the data.

Data Points: U.S. violent crime decline: more than 30% from 1991 to 2001 - Levitt’s benchmark for the major national crime drop he was trying to explain. New York City homicides: more than 2,200 in 1990; fewer than 300 in recent years - Used as a concrete example of the long-run crime decline. Peak annual U.S. abortions: 1.5 million per year - Shown as the scale of legalized abortion that could plausibly affect later social outcomes. Live births at peak abortion period: roughly 4 million live births per year - Compared with 1.5 million abortions to show abortion’s large share of pregnancies. Abortions per 1,000 live births: 345 - Back-of-the-envelope measure used by Levitt and Donahue to estimate potential crime effects. Early-legalization states: 5 states - New York, California, Washington, Alaska, and Hawaii legalized abortion before Roe v. Wade. Crime drop across high vs. low abortion states: about 30% difference by 1997 - Found in the original state-group comparison. Abortion hypothesis share of crime drop: as much as 50% - Claim in the 2001 Quarterly Journal of Economics paper. Abortion effect in updated analysis: high-abortion states saw crime fall about 60% more than low-abortion states - Result from the revisit using data from 1997 to 2014. Updated cumulative claim: about 80–90% of the entire decline in crime - Levitt’s description of how large the implied effect becomes in the longer-run analysis. Lead phase-out period: 1975 to 1985 - EPA leaded-gasoline phaseout used by Jessica Reyes as another natural experiment. Abortion and lead studies relationship: each effect remained similar magnitude when controlling for the other - Reyes said her lead results reaffirm the abortion findings rather than refute them. Teen pregnancy decline: more than 60% over the past quarter century - Raised as evidence that unwanted pregnancies have already fallen substantially in the U.S. Current annual abortions: about 640,000 - Compared with the peak of 1.5 million to show the present-day decline in abortion volume. Estimated homicide reduction from abortion: 5,000 to 10,000 fewer homicides - Levitt’s rough estimate of the public-safety effect implied by the research.

Pivotal Quotes: "I think maybe legalized abortion might have reduced crime." — Steve Levitt: The central hypothesis that emerged from his and John Donahue’s puzzle-solving about the 1990s crime decline. "what our data suggests is that women are pretty good at choosing when they can bring kids into the world who they can provide good environments for." — Steve Levitt: Levitt’s rebuttal to eugenics-related criticism and explanation of the mechanism. "I think we should flip that discussion on its head." — Steve Levitt: His argument that social-science reporting should start with facts and data rather than competing narratives.

Implications: The episode suggests crime is multicausal and that abortion policy debates should not be reduced to one empirical study. More broadly, it argues for policies that reduce unwanted pregnancies and support wanted children, while treating social-science claims with nuance.