Small Area Estimation-Based Prediction Methods to Track Poverty : Validation and Applications

Christiaensen, Luc ; Lanjouw, Peter ; Luoto, Jill ; Stifel, David

关键词: ABSOLUTE TERMS;    ABSOLUTE VALUE;    ARREARS;    ASSET CLASS;    ASSET CLASSES;   

DOI  :  10.1596/1813-9450-5683
RP-ID  :  WPS5683

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Tracking poverty is predicated on theavailability of comparable consumption data and reliableprice deflators. However, regular series of strictlycomparable data are only rarely available. Price deflatorsare also often missing or disputed. In response, povertyprediction methods that track consumption correlates asopposed to consumption itself have been developed. Thesemethods typically assume that the estimated relation betweenconsumption and its predictors is stable over time -- anassumption that cannot usually be tested directly. Thisstudy analyzes the performance of poverty prediction modelsbased on small area estimation techniques. Predicted povertyestimates are compared with directly observed levels in twocountry settings where data comparability over time is not aproblem. Prediction models that employ either non-staplefood or non-food expenditures or a full set of assets aspredictors are found to yield poverty estimates that matchobserved poverty well. This offers some support to the useof such methods to approximate the evolution of poverty. Twofurther country examples illustrate how an application ofthe method employing models based on household assets canhelp to adjudicate between alternative price deflators.