Inequality of Opportunity in Health among the Middle-aged and Elderly People in China:New Evidence from Machine Learning Methods
With the continuous acceleration of China's ageing process,the issue of health inequality among the elderly has become increasingly prominent and important.Drawing on data from 2011-2018 China Health and Retirement Longitudinal Survey,this study employs three estimation methods,namely,linear regression,conditional inference tree and conditional forest,to measure the inequality of opportunity in health adaptive load and self-rated helth among middle-aged and elderly adults aged 45 and above in China.It also measures the relative contribution of each circumstance variable to the inequality of opportunity in health.The results show that the relative values of inequality of opportunity for allostatic load(self-reported health)ranges from 3.21%(5.15%)to 7.76%(10.44%)respectively.The decomposition results further indicate that demographic characteristics(age and gender)and childhood socioeconomic status are the key contributors for inequality of opportunity in both allostatic load and self-reported health.Unlike linear regression results where region/province of birth is the most important factor,the two machine learning estimates show that demographic characteristics(age and gender)and childhood socioeconomic conditions are the two main factors accounting for the opportunity inequality of allostatic load.This paper proves that the health opportunity inequality measurement based on conditional forest is better than the traditional linear regression.This result remains robust for individual objective health indicators and the adjustment for observable environmental variables.Health opportunity inequality is the underlying reasons behind health inequality,and a comprehensive evaluation of health inequality for middle-aged and elderly people in China is of great practical significance for the introduction of effective public policies to reduce health inequalities for the elderly.
inequality of opportunity in healthShapley-value decompositionconditional inference treeconditional forest