Very few demographic surveys in developing countries gather information on household income or consumption expenditure, despite the theoretical importance of these measures. Consequently, researchers have been forced to rely on ad hoc collections of proxy measures for living standards, and the properties of these proxies have not been systematically analyzed. In this research, we ask what hypotheses can be tested using proxy variables, and evaluate the performance of proxy measures in relation to consumption expenditures per adult, our preferred measure of living standards. We find that the proxy variables commonly employed in demographic research are very weak predictors of consumption per adult, having partial R^2 values that are extremely low. Nevertheless, when other factors are taken into account, we show that tests based on proxy variables are likely to be sufficiently powerful to merit consideration. In an examination of fertility, child mortality, and children’s schooling, we compare coefficient estimates based on consumption per adult to alternative estimates based on proxies, and find that the proxy-based estimates provide generally reliable guidance to the sign and magnitude of the preferred estimates.
Montgomery, Mark R., Michele Gragnolati, Kathleen A. Burke, and Edmundo Paredes. 1999. "Measuring living standards with proxy variables," Policy Research Division Working Paper no. 129. New York: Population Council. Version of record: https://doi.org/10.2307/2648118