Diabetes and labour market outcomes in Mexico: a panel data analysis

Abstract: Diabetes is increasingly recognized as a major health threat, also in low and middle income countries, with potentially strong adverse economic effects. However, studies of these effects typically need to confront various statistical issues to arrive at credible causal inference. One such issue is omitted variable bias. For the study of the effects of diabetes on labour market outcomes the method of choice - despite its limitations - has been the use of the family history of diabetes as an instrumental variable (see e.g. Seuring/Goryakin/Suhrcke 2015). Our study proposes a different route, taking advantage of Mexican panel data to account for time-invariant omitted variables. Measurement error poses an additional statistical issue, when using self-reported diabetes information. We therefore use biomarker data in a cross-sectional analysis to objectively measure diabetes and to assess the extent of undiagnosed diabetes as well as the potential for measurement error to cause biased estimates. The results of the panel-data analysis indicate a strong negative relationship of self-reported diabetes and employment chances which are reduced by about 5 percentage points. We find no adverse relationship with wages and working hours. Further, we find that when diabetes is measured objectively, the adverse association with employment remains (albeit slightly reduced in size) and is mainly driven by those with diagnosed diabetes. Our findings suggest a strong adverse effect of self-reported diabetes on employment chances, however, once people have selected into employment, self-reported diabetes does not appear to impact their productivity. Further, future studies of the labour market consequences of diabetes ought to take into account the possibility that using self-reported diabetes might lead to an overstatement of the employment effect of diabetes.