Is predicted data a viable alternative to real data ?

It is costly to collect the household- and individual-level data that underlies official estimates of poverty and health. For this reason, developing countries often do not have the budget to update their estimates of poverty and health regularly, even though these estimates are most needed there. One way to reduce the financial burden is to substitute some of the real data with predicted data. An approach referred to as double sampling collects the expensive outcome variable for a sub-sample only while collecting the covariates used for prediction for the full sample. The objective of this study is to determine if this would indeed allow for realizing meaningful reductions in financial costs while preserving statistical precision. The study does this using analytical calculations that allow for considering a wide range of parameter values that are plausible to real applications. The benefits of using double sampling are found to be modest. There are circumstances for which the gains can be more substantial, but the study conjectures that these denote the exceptions rather than the rule. The recommendation is to rely on real data whenever there is a need for new data, and use the prediction estimator to leverage existing data.