The journals are full of great studies but can we believe the statistics? Revisiting the mass privatisation - mortality debate

Abstract

Cross-national statistical analyses based on country-level panel data are increasingly popular in social epidemiology. To provide reliable results on the societal determinants of health, analysts must give very careful consideration to conceptual and methodological issues: aggregate (historical) data are typically compatible with multiple alternative stories of the data-generating process. Studies in this field which fail to relate their empirical approach to the true underlying data-generating process are likely to produce misleading results if, for example, they misspecify their models by failing to explore the statistical properties of the longitudinal aspect of their data or by ignoring endogeneity issues. We illustrate the importance of this extra need for care with reference to a recent debate on whether discussing the role of rapid mass privatisation can explain post-communist mortality fluctuations. We demonstrate that the finding that rapid mass privatisation was a "crucial determinant" of male mortality fluctuations in the post-communist world is rejected once better consideration is given to the way in which the data are generated.