Rescaled Additively Non-ignorable (RAN) Model of Attrition and Substitution

We modify the Additively Non-ignorable (AN) model of Hirano et. al. (2001) so that it is suitable for data collection efforts that have a short panel component. Our modification yields a convenient semi-parametric bias correction framework for handling endogenous attrition and substitution behavior that can emerge when multiple visits to the same unit are planned. We apply our methodology to data from the Household Labor Force Survey (HLFS) in Turkey, which shares a key design feature (namely a rotating sample frame) of popular surveys such as the Current Population Survey and the European Union Labor Force Survey. The correction amounts to adjusting the observed joint distribution over the state space using reflation factors expressed as parametric functions of the states occupied in subsequent rounds. Unlike standard weighting schemes, our method produces a unique set of corrected joint probabilities that are consistent with the margins used for computing the published cross-section statistics. Inference about the nature of the bias is implemented via Bootstrap methods. Our empirical results show that attrition/substitution in HLFS is a statistically and substantially important concern.