Hybrid survey to improve the reliability of poverty statistics in a cost-effective manner

This paper studies the benefits, in terms of reliability and frequency of poverty statistics, of conducting a hybrid survey that collects non-consumption data from all surveyed households and consumption data from only a small subsample. Collecting detailed consumption or income data for the purpose of estimating poverty is costly and many low-income countries cannot afford to carry out such surveys on a regular basis. One option is to collect only non-consumption data and use consumption models developed from a previous round of household survey data to project poverty data. Although this approach is cost-effective because collection of non-consumption data is much cheaper than collection of consumption data, it is vulnerable to a structural change between the current and previous household surveys and might produce poverty estimates that are not comparable with the previous ones. Instead, the hybrid approach creates consumption models from a subsample of the current survey and applies them to the entire survey to project consumption data for all households in the sample. This paper examines the hybrid approach with data from the Bangladesh Household Income Expenditure Surveys of 2000 and 2005. Improvements in accuracy are achieved even with subsamples of just 320 or 640 households. Budget simulations confirm that the additional cost of collecting consumption data for such small subsamples is minimal.