The importance of measurement error for the measured versus true dynamics of indices of poverty, income and unemployment has long been recognized in projects using data from developed countries. Often such studies have found the measurement error process to be less persistent than the underlying true process, making the measured poverty or unemployment dynamics (for example), appear less persistent than they truly are. Owing to the lack of appropriate data for developing countries, however, these methodologies have not been widely imported into studies on poverty and income dynamics in developing countries. This is unfortunate as the conclusions drawn from developed countries are likely not generalizable to developing economies where income processes are generally less persistent owing to their dependence on weather shocks, illness, and other `high frequency' shocks. In addition, the statistical methodologies which are commonly employed for data on households in developed economies may also not be appropriate for the developing country context. Households in developing economies are often rather diverse enterprises in which household members are likely to be comparatively less informed of the overall household activities as compared to the smaller, `nuclear' household in the developed country context. As such, the use of survey responses by multiple household members on a given household activity are unlikely to di®er by only a pure classical `measurement' or response error. We instead construct a statistical framework that allows for a behavioral component of mis-reporting that allows the multiple reports to di®er by more than just a classical measurement error term, but also due to asymmetric information regarding household activities, for example.
Poverty Measuresurement and Dynamics presented at Staying Poor: Chronic Poverty and Development Policy, Institute for Development Policy and Management, University of Manchester, 7-9 April 2003. Chronic Poverty Research Centre (CPRC), Manchester, UK, 42 pp.