diabetes and cardiovascular disease. In addition since diabetes is a risk factor for cardiovascular disease there is an additional indirect effect from smoking through the pathway of diabetes risk to cardiovascular disease prevalence and then to labour force participation. We estimate the marginal effect of all of the exogenous and endogenous variables on labour force participation and their confidence intervals separating out the direct and indirect treatment effects through these pathways of influence. The treatment effects were calculated from the difference in the predicted probabilities conditional on marginal changes for continuous regressors and zero and one for discrete variables in each equation. Standard errors were calculated using the Delta method taking account of the correlation in errors across equations. (Greene 2003) and tested at the 10% significance level.
The treatment effect of being diagnosed with cardiovascular disease is calculated as the difference in the predicted conditional probability of being in the labour force without cardiovascular disease less the predicted conditional probability of being in the labour force with cardiovascular disease after controlling for exogenous variables. The calculation of the treatment effect on labour force participation of diabetes and the exogenous variables is more complex involving as they do the indirect effect on labour supply through the risk of cardiovascular disease.
The data come from the AusDiab survey, a population-based cross-sectional survey of national diabetes mellitus prevalence and associated risk factors in people aged 25 years, conducted between May 1999 and December 2000 in the six states and the Northern Territory of Australia (Dunstan, Zimmet et al. 2002). The study involved an initial household interview, followed by a biomedical examination that included an oral glucose tolerance test, standard anthropometric tests, blood pressure measurements and the administration of questionnaires. Of the 20 347 eligible people (aged 25 years and resident at the address for more than 6 months) who completed a household interview, 11 247 (55.3%) attended for the biomedical examination. To account for the clustering and stratification of the survey design, and to adjust for non-response,
the data have been weighted to match the age and gender distribution of the 1998 estimated
residential population of Australia aged
The key advantages of the AusDiab survey from the point of view of this study is that it offers a large representative sample of the Australian population over the age of 25 with detailed information on individual self reported and health and clinically measured chronic disease status along with employment status. Importantly it includes clinically measured risk and treatment status for two major chronic diseases, diabetes and cardiovascular disease (blood pressure and cholesterol), self reported family risk information on diabetes, age, as well as lifestyle information on weight and exercise, smoking, and socio-economic information on education status, children and marital status. The richness of the data on chronic disease risk factors allows us to specify detailed structural equations that explain the prevalence of diabetes and cardiovascular disease independently of the participation decision. This allows us to adjust for the potential endogeneity of disease status associated with unobserved common factors that might explain the prevalence of cardiovascular disease, diabetes, and labour force participation.
The definition of those who are diabetic in the AusDiab survey is a combination of people on insulin or tablets who have been told they have diabetes plus those who say they have never been told they have diabetes but have a diabetic glucose value. A diabetic glucose value is
Chronic disease and labour force participation in Australia: an endogenous multivariate probit analysis of clinical prevalence data