On the issue of causes of mortality crises. The use of matching and inverse probability weighting to analyze mortality in 19th century southern Sweden

Tommy Bengtsson, Lund University
Göran RA Broström, Umeå University

In a time series approach mortality crises are “outliers” and the question is whether their causes differ from “normal” variation. Causality is addressed by decomposing the explanatory variable(s) into trend and cyclical components. This approach is very useful since it is easy to establish causality. Still it is possible that another factor influence both fluctuations in the explanatory variable and in mortality. Temperature, for example, may affect both food prices and mortality, which may create biased estimates of the effect of food prices on mortality. In this paper we use matching and inverse probability weighting in order to estimate causal effects. The purpose of this is to remove confounding without introducing selection bias. Software has been developed to treat weights that vary over risksets. The data comes from the Scanian Demographic Database and covers five rural parishes for the period 1813 to 1895.

  See paper

Presented in Session 201: A historical demography of epidemics