The impact of diabetes on labour market outcomes

ABSTRACT

This study estimates the effect of diabetes on labour market outcomes (employment, unemployment and labour force participation) in South Africa using data from the South African General Household Survey (2018). We first examine the possibility that diabetes status is endogenous through the application of heteroscedasticity-based instruments. Internal instruments meet the underlying diagnostic expectations, but do not consistently accept the endogeneity hypothesis. Thus we turn to multinomial logit models, ignoring endogeneity, to estimate the effect of diabetes. Our findings indicate that diabetes has differential effects for men and women, where the magnitude of the effect tends to be larger (in absolute value) for women.

KEYWORDS:

• Diabetes
• labour market
• endogeneity

Acknowledgments

The authors would like to thank John Ataguba, Heinrich Bohlmann, Frikkie Booysen and Zoë McLaren for their comments on earlier drafts of this research. However, we remain responsible for any and all remaining errors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The five instrumental variables incorporate paternal diabetes (as well as having died from the disease), maternal diabetes (also, having died from it) and sibling diabetes.

2 In preliminary analysis, we followed Li & Racine (2004), which is implemented by the np package (Hayfield & Racine, 2008) in R (R Core Team, 2020), to estimate the residual from the first stage. We applied a Gaussian kernel for the continuous variables and the Li & Racine (2007) kernel, which is an extension of the kernel proposed by Wang & van Ryzin (1981) that works well for both ordered and unordered discrete variables. Our estimated residual was neither continuous nor offered a large support, and, therefore, we did not consider this avenue any further.

3 In additional analysis not reported, we examined the sensitivity of our endogeneity conclusion. If we limited our internal instruments to only household wealth, no endogeneity was uncovered; for all other combinations of internal instruments, endogeneity was robust.

4 The official statistics point to unemployment rates closer to 27% (http://www.statssa.gov.za/?p=11897)[http://www.statssa.gov.za/?p=11897], and these figures are based on labour force surveys, rather than the general household survey used for this analysis.