Extant interurban research in life expectancy greatly suffers from an underestimation of the role of place. Place is often conceptualized as a level of geography; this view ignores categorical differences between types of places. In addition, despite advances in theory and research that support their use, many important place-based factors remain under-utilized as control variables. We use multivariate analyses of life expectancy for the top and bottom quartiles of household income by sex in 148 US counties to compare the strengths of seventeen diverse variables. We find that cities' built, natural, and social environments play strong roles in life expectancy disparity among cities; many place-based variables consistently compare in strength to well-known control variables such as race, education, and behaviors. Furthermore, we find that place impacts men and women differently, even within the same income quartile. Indeed, some factors are associated with higher life expectancy in some demographic groups, and lower life expectancy in others. Researchers can protect against omitted variable bias when investigating public health outcomes by using a wider range of control variables. Researchers should also use better measures of place, and consider selecting specific cases to study.
Keywords: Interurban research; Life expectancy; Model building; Quantitative methods; Variable selection.
Copyright © 2019 Elsevier Ltd. All rights reserved.