Time Aggregation and State Dependence in Welfare Receipt

This article tests the validity of the conditional Markov assumption in a dynamic discrete‐choice model of welfare benefit receipt dynamics, using Norwegian administrative data. Exploiting its implied time aggregation properties, we find that the model is seriously mis‐specified. Estimated state dependence varies strongly with the time unit of analysis, with the average treatment effect of past benefit receipt increasing with the level of aggregation. When permitting richer types of benefit dynamics, we find evidence for duration and occurrence dependence in benefit receipt and disparities in the effects of observed and persistent unobserved characteristics on the benefit entry and persistence processes.