Our 2017 article in this Journal stresses the pitfalls of using choice-based samples in economic history. A prominent example is the literature addressing the so-called antebellum puzzle. Heights researchers claim that Americans grew shorter in the first half of the nineteenth century, a period of robust economic growth. We argue that this result relies on choice-based samples. Without knowing the process that led to inclusion in the sample, researchers cannot properly estimate conditional mean heights. We proposed a diagnostic that can detect, but not correct for, selection bias. Komlos and A’Hearn’s interpretation of our analysis confuses diagnosis with cure. We dispute their view that selection bias has been appreciated in the heights literature.
Heterogeneous Causal Effects and Sample Selection Bias
