Three Essays on Poverty Mapping and Targeting in Rural China.
详细信息
文摘
This dissertation is a collection of three essays in poverty mapping and targeting in China. The first essay uses a recently developed small-area estimation technique to derive geographically detailed estimates of consumption-based poverty and inequality in rural Shaanxi, China. The first chapter also investigates whether including environmental variables in the equation used to predict consumption and poverty improves upon typical approaches that only use household survey and census data. I found that ignoring environmental variables in statistical analyses that predict small-area poverty rates appears likely to produce targeting errors. Using information on locations that are, in fact, receiving poverty assistance, I found evidence that official poverty policy in Shaanxi targets particular areas which may not be the poorest. The second essay uses a spatial econometric approach in estimating specifications that incorporate spatial dependence in the first stage of consumption model of the poverty mapping exercises. The results presented in this essay offer a rough test of the ELL methodology and point to some tentative conclusions that may inform future applications of the ELL poverty mapping methodology. Using geo-referenced survey data from rural Shaanxi I found the evidence of spatial autocorrelation in the data, as a consequent, the conventional methodology could significantly over-state the precision of local-level estimates of poverty in the second stage of the analysis. The empirical results also seems to suggest that spatial error framework is more effective in capturing location effect in comparison to the standard random effect model even after the inclusion of location means of household-level variables from census and environmental data. The conventional small area estimation method used for poverty analysis involves using household unit level data from a census. Researchers, however, do not always have access to the household-level census data because they are regarded as confidential. One alternative is to census data that has been aggregated to a higher level such as township or county). It is not clear to policy analysts how much reliability being traded off for easier data access. In the third essay, I generate poverty estimates using Chinese census data that have been aggregated to different levels and compare the results to those obtained from household level census data and assess the question of how much precision is lost in generating poverty maps from aggregate census data using the Chinese data set. I found that if household level census data is not available to researcher, it is still possible to get a reasonably accurate estimate of the incidence of poverty using aggregated census data. The errors due to aggregation are more likely to be small if the level of aggregation of census data is relatively low.