Democratic reform and health : interpreting causal estimates

In The Lancet Global Health, Hannah Pieters and colleagues (September, 2016)1 analyse the effect of democratic reforms on child mortality across the world. We wish to highlight, however, that even with sophisticated causal inference techniques, such results cannot necessarily be interpreted as causal effects. First, the results are compatible with a number of different theories including that democratic reforms have no effect on health ceteris paribus (ie, holding everything else fixed). Consider the cases of South Africa, Zambia, Mozambique, and Zimbabwe, all notably missing from the analyses but experiencing substantial democratic changes, analysed here using a similar synthetic control analysis (figure).1, 2 No change is observed in South Africa after the end of apartheid in 1994. In Zambia, after reform in 1991, a reduction is observed but not until the price of copper tripled and GDP per capita doubled. In Mozambique, the large fall is likely attributable to the cessation of the civil war in 1993. And in Zimbabwe, democratic restrictions in 1987 did not precipitate an increase in child mortality