Estimating income poverty in the presence of measurement error and missing data problems
The aim of this paper is estimating income poverty in the presence of missing data and measurement error problems. We use the conventional definition of poverty as having an income below a poverty line, 60% of median household income, where income is scaled to take account of household composition and size.
Income measures are usually among the variables with the highest nonresponse rates. Both survey methodologists and applied econometricians suggest that ignoring the nonresponse problem can cause a serious selection bias. Moreover, since responses to surveys are not perfectly reliable, income poverty measures are generally plagued by measurement errors too.
Point estimation approaches taking account of missing data and/or measurement error problems impose usually restrictive and non-testable assumptions. On the contrary, in this paper we do not impose any restrictive assumption but we provide bound estimates instead of point estimates of the poverty rate. In other words, we provide an upper and a lower bound which defines the range of logically possible values for the poverty rate.
By using the European Community Household Panel (ECHP) we compute bound estimates for the poverty rates in Belgium, Denmark, France, Germany, Greece, Ireland, Italy, the Netherlands, Portugal, Spain, and the UK. Our bound estimates seem to be in line with official statistics published by Eurostat where Greece, Italy and Portugal are the countries with the highest poverty rates, around 0.23, and where Belgium, Denmark, Germany and the Netherlands have the lowest poverty rates, usually lower than 0.15. In fact, the estimated lower bounds of the poverty rate are higher than 0.15 in Greece, Italy and Portugal; while the estimated upper bounds are lower than 0.23 in Belgium, Denmark, Germany and the Netherlands. Moreover, we find that the missing data problem is more relevant than the measurement error problem in all countries.
Journal of Business and Economic Statistics
Volume and page numbers
29 , 61 -73
Albert Sloman Library Periodicals *restricted to Univ. Essex registered users*