E. Gakidou’s scientific contributions

Household surveys in developing countries often lack modules on income and ex- penditure. Even in cases where they may be included, the resulting estimates show substantial measurement error and are subject to systematic reporting biases. In order to overcome these problems, some analysts have constructed indices based on factor or principal components analysis of indicator variables such as asset ownership. These in- dices do not provide information on the level of income at which different durable goods or services are likely to be acquired, nor do they provide any prospective guidance on identifying the best indicators for obtaining more refined estimates of permanent in- come in future surveys. In this paper, we show that these limitations can be overcome through an approach based on av ariant of the hierarchical ordered probit (HOPIT) model. The model produces a series of indicator-specific cut-points on a latent scale (permanent income or wealth). These cut-points are values on the latent scale above which respondents are more likely to respond affirmatively than not. When combined with an individual household's responses to the questions, the cut-points can be used to estimate the permanent income of the household. This analysis compares estimates of permanent income using the above approach with estimates resulting from factor or principal components analysis using household survey data from Greece, Peru and Pakistan. Although estimates of permanent income using the probit method are comparable with those of the comparison method (in terms of rank correlation with reported income or expenditure), we show that one of the key advantages of the former method is it's compatibility with item reduction methods. Thus, the approach is particularly useful by allowing development of sensitive combi- nations of indicator questions that will yield the most refined estimates of household permanent income in different survey settings.

This paper addresses the issue of cross-population comparability of economic status indices estimated from household-level indicator variables on consumer durables using latent variable statistical models. The problem is similar to the one faced in estimation of PPPs. The same basket of consumer durables need not imply the same level of economic status across countries. This is because, even for the same level of the true economic status, the likelihood of owership of a given consumer durables is not the same across countries. This may be due to different prices structures, preferences, or other enviornmental factors. The problem we face is similar to that faced by psychometricians in intelligence testing using question banks: even for the same level of intelligence some questions exhibit differential item functioning (DIF) in that the likelihood of responding correctly for these questions varies significantly across socio-demographic groups. We propose one way of making the estimates of economic status comparable across countries which relies on assuming that the entire set of consumer durables not exhibit DIF. This enables us to estimate economic status on ac ommon scale across countries. The results seem promising meriting further research for testing and validating using better-quality data on consumer durables.