关键词: 代理工具测试模型, ROC曲线, 低保贫困瞄准

Abstract: Based on the data of China’s Family Panel Surveys (CFPS), this paper evaluates the targeting of minimum subsistence allowance by using proxy means testing (PMT) model and receiver operating characteristics (ROC) curve simultaneously. The results show that toilet types and computer possession are good individual indicators of poverty in urban areas, while age of household head and refrigerator possession are effective individual indicators of poverty in rural areas. Changing the cut-off of poverty probability will affect the targeting result of public budget transfers. If a low value for the cutoff is chosen, the corresponding rate of coverage of the poor population by public budget transferred payment and inclusion of the non-poor are both relatively high. When policy-makers attach equal weights to the coverage of the poor and the inclusion of nonpoor, the cut-off of poverty probability is around 0.5 for urban and rural areas. This paper also finds that PMT-based targeting results in large errors when poverty rate is low and the chosen targeting rate is the same as the former. With a given poverty rate, the accuracy of poverty targeting is increasing as the beneficiary share (inclusion rate) increases. When the percentage of the universal budget drops, the coverage and leakage rates also show a declining trend, and the poverty rate is gradually increasing; the percentage of the budget for coverage increases, and the proportion of the budgeted leakage decreases. Changes in the poverty line will affect coverage and leakage rates.

Key words: Proxy Means Testing (PMT), Receiver Operating Characteristics (ROC) Curve, Poverty Targeting for Minimum Subsistence Allowance