as time or risk preference that also impact on labour market decisions. Social security eligibility conditions may encourage those surveyed individuals who do not participate in the labour force to overstate their ill health as a justification for their non participation, although the empirical literature has not found strong evidence of this effect so far (Stern 1989; Dwyer and Mitchell 1999; Cai and Kalb 2006). Poor health may also be a consequence of labour market outcomes raising the possibility of simultaneity bias.
An obvious alternative to measures of self reported health is to use more objective measures either from generic health status measures or clinical measures. Other researchers have examined the effects on labour market performance of specific chronic health conditions such as diabetes (Kahn 1998; Bastida and Pagan 2002; Brown and Pagán Elena Bastida 2005) or mental health problems (Waghorn and Chant 2005). Although specific health conditions may also be self- reported, relative to self-assessed global health status, they are much less likely to be subject to reporting/measurement errors. There are limitations in looking only at specific conditions in isolation. Many chronic diseases are interrelated through common risk factors and physiology. For example diabetes is associated with long-term conditions such as diseases of the circulatory system and eyes. In particular, people with diabetes are at an increased risk of developing coronary heart disease, stroke and peripheral vascular disease (Australian Institute of Health and Welfare 2006). In the 2004 national health survey 20% of persons reporting diabetes also reported heart, stroke or vascular disease; 18% of those aged 45-64 years, and 27% of those aged over 65 years, had one or more of these circulatory conditions (Australian Bureau of Statistics 2006). Among the studies on specific chronic health conditions and labour market performance, most have focused on one particular chronic illness and treated the incidences of other chronic illnesses as exogenous. However, incidences of chronic conditions such as diabetes and cardiovascular disease are likely to be in part the results of lifestyle behaviour and other unobservable individual characteristics that also influence labour market outcomes. Any degree of measurement errors due to the subjectivity of the self-reported medical conditions would also mean that the incidences of such reported chronic diseases are potentially endogenous. There are some exceptions in this literature such as (Brown and Pagán Elena Bastida 2005) who studied the impact of diabetes on employment, (MacDonald and Shields 2004) who looked at the impact of problem drinking on employment and (Morris 2007) who investigated the relationship between obesity and employment studied using a recursive bivariate probit model to account for endogeneity. In all cases however, only a single self-reported disease was used.
In contrast this paper uses clinically measured diagnoses of chronic diseases and risk factors to provide evidence on relationship between labour market participation and chronic disease and adopts a multivariate approach to estimation that account for multiple sources of endogeneity across two chronic diseases: diabetes and cardiovascular disease. The multivariate approach is similar to that in (Zhao and Harris 2004) for drugs, but with the addition of the simultaneous determination of two of the three dependent variables.
The underlying theoretical model is one of a conventional household labour supply. The individual is assumed to make a trade off between labour and leisure subject to a constraint of full income. The decision to enter the labour force is the first part in a two part decision on hours supplied. The final outcome of hours worked will depend on both labour supply and the demand for that individuals labour. In this paper we do not model the second part of the decision but restrict the analysis to the participation part of the labour supply decision. This is done to simplify the model
Chronic disease and labour force participation in Australia: an endogenous multivariate probit analysis of clinical prevalence data