The Ecological Fallacy of the Abortion-Crime Thesis

Bryan L. Sykes, University of California, Berkeley
Dominik Hangartner, University of Bern
Earl Hathaway, University of Wisconsin at Madison

Recently some scholars have asserted that abortion legalization lowered crime through time lagged effects. While economists have both substantiated and challenged these findings, sociologists and demographers have been mute on the topic. We show that the link between abortion and crime is the result of omitted variables bias and researchers' inability to distinguish between age-period-cohort effects. We correct these problems and use propensity score matching and quasi-experimental methods to retest the causal argument in the case of homicide. Using data from the NSFG, Census, CDC, SHR, Donahue and Levitt, and Fryer et al., we find that when period-cohorts are compared, abortion legalization did not have any measurable effect on crime once race, sex, the sex ratio at birth, divorce rates, contraceptive use, and the crack-cocaine epidemic are controlled. Our findings suggest that any drop in crime is the result of a mixture of unmeasured period and cohort effects and not abortion.