Measuring income inequality using survey data: the case of China

The purpose of this paper is to raise awareness of missing data when we evaluate income inequality using survey data. If the income data are not missing completely at random, the calculated income inequalities are more likely to be biased, which may lead to inappropriate conclusions and policy recommendations. To handle the missing data on income, a multiple imputation approach is utilized. In particular, we propose an extended approach to correct the possible sample selection bias in the imputation process. A case study using China’s household survey suggests that extended imputation corrects for biases effectively in the calculation of Gini coefficients and results in gains in efficiency as well.