’Offensive Medicine’ or Aggregation Bias: Reversal Paradox in the Case of Cesarean Sections

Economists have considered the Cesarean section (c-section) a medical procedure whose use by physicians is amenable to legal and economic incentives. Increasing c-section rates in the United States have been attributed, in part, to physicians seeking to avoid malpractice tort claims by practicing defensive medicine.  Economic studies have found mixed evidence of the impact of tort laws on physicians’ use of c-sections and the presence of defensive medicine. Two studies have found that decreasing physician expected liability through tort reform increases cesarean sections, contrary to what one would expect if physicians were practicing defensive medicine. With a couple of exceptions, these studies have generally not accounted for observed heterogeneity (in terms of maternal and fetal conditions) among pregnant women or have accounted for such heterogeneity by including dummy variable regressors for such conditions. The studies have generally not estimated separate regressions for sub-populations of pregnant women (such as women who have had a prior c-section or women with plural births) who differ according to such conditions. This paper focuses on the natural experiments created when, in 1996, the Oregon Court of Appeals found that a cap on economic damages enacted by the state legislature violated the state constitution and, in 1999, the Oregon Supreme Court affirmed the Court of Appeals ruling. We obtain separate difference-in-difference (DD) estimates for the period between the Court of Appeals and the Supreme Court rulings, and for the period after the Supreme Court ruling. We explicitly account for the impact of the foregoing observed heterogeneity on the average partial effects of DD estimates of the Oregon court decisions by neestimating separate regressions for the different sub-populations of pregnant women. We also gain insight into the presence of unobserved heterogeneity by explicitly modeling unobserved heterogeneity with a discrete factor model. We find that estimating separate regressions for the sub-populations of women with the different medical (or fetal) conditions may be preferable to accounting for those conditions by including dummy variable regressors for those conditions. The results for a regression which combines all pregnant women suggest that removal of the cap decreased c-section rates while separate regressions for the different sub-populations of women indicate that c-section rates either increased or did not decrease across the sub-populations. The contradictory results appear to be a result of a change over time in composition of maternal conditions, with a greater percentage of pregnant women presenting with conditions which have a lower probability of c-section. This finding may explain prior empirical findings that a decrease in expected medical malpractice liability increased c-section rates.  We seek to explain these findings in the context of a theoretical model presented in Currie and MacLeod (2008), emphasizing the differing effects of medical and legal risks on physician behavior. Our regression results also suggest that unobserved heterogeneity does not affect average partial effects of the DD estimates.