Zhang Xiaohui, Zhao Xueyan, Harris Anthony
Department of Econometrics and Business Statistics, Monash University, Australia.
J Health Econ. 2009 Jan;28(1):91-108. doi: 10.1016/j.jhealeco.2008.08.001. Epub 2008 Aug 19.
We examine the impact of several chronic diseases on the probability of labour force participation using data from the Australian National Health Surveys. An endogenous multivariate probit model is used to account for the potential endogeneity of the incidence of chronic conditions such as diabetes, cardiovascular diseases and mental illnesses. The cross-equation correlations are significant, rejecting the exogeneity of the chronic illnesses. Marginal effects of exogenous socio-demographic and lifestyle variables are estimated through their direct effects on labour market participation and indirect effects via the chronic diseases. The treatment effects of chronic diseases on labour force participation are estimated via conditional probabilities using five-dimensional normal distributions. The estimated effects differ by gender and age groups. Although computationally more demanding, these treatment effects are compared with results from a univariate model treating the chronic conditions exogenous and the structural effects from the multivariate probit model; both significantly overestimate the effects.
我们利用澳大利亚国民健康调查的数据，研究了几种慢性病对劳动力参与概率的影响。使用内生多元概率模型来解释糖尿病、心血管疾病和精神疾病等慢性病发病率的潜在内生性。交叉方程相关性显著，拒绝了慢性病的外生性。通过外生社会人口和生活方式变量对劳动力市场参与的直接影响以及通过慢性病的间接影响来估计其边际效应。通过使用五维正态分布的条件概率来估计慢性病对劳动力参与的治疗效果。估计的效果因性别和年龄组而异。尽管计算要求更高，但将这些治疗效果与将慢性病视为外生变量的单变量模型结果以及多元概率模型的结构效应进行了比较；两者都显著高估了效果。
