Working Paper: NBER ID: w19955
Authors: Howard Bodenhorn; Timothy Guinnane; Thomas Mroz
Abstract: Much of the research on height in historical populations relies on convenience samples. A crucial question with convenience samples is whether the sample accurately reflects the characteristics of the population; if not, then estimated parameters will be affected by sample selection bias. This paper applies a simple test for selection biased developed in Bodenhorn, Guinnane, and Mroz (2013) to several historical samples of prisoners, freed slaves, and college students. We reject the hypothesis of no selection bias in all cases. Using Roy's (1951) model of occupational choice, we interpret these findings as reflecting the economic forces that lead individuals to take the actions the led to inclusion in the sample. Our findings suggest that much of the evidence on the "industrialization puzzle" during the nineteenth century could reflect changing selection into the samples rather than changes in population heights.
Keywords: No keywords provided
JEL Codes: N01; N31
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| height (Y60) | occupational choice (J29) |
| sample selection bias (C83) | interpretation of historical height data (Y10) |
| height-based selection (C52) | conclusions about population health and wellbeing during industrialization (I14) |
| military enlistment practices (H56) | height-based selection (C52) |
| economic conditions (E66) | likelihood of individuals of different heights to enlist (J45) |
| height selection (Y10) | observed heights in military samples (J45) |
| macroeconomic conditions (E66) | height selection over time (C41) |
| selection process (C52) | spurious trends in height data (J11) |