Predicting Benchmarked Us State Employment Data in Realtime

Abstract

US payroll employment data come from a survey of nonfarm business establishments and are therefore subject to revisions. While the revisions are generally small at the national level, they can be large enough at the state level to substantially alter assessments of current economic conditions. Researchers and policymakers must therefore exercise caution in interpreting state employment data until they are "benchmarked" against administrative data on the universe of workers some 5 to 16 months after the reference period. This paper develops and tests a state space model that predicts benchmarked US state employment data in realtime. The model has two distinct features: 1) an explicit model of the data revision process and 2) a dynamic factor model that incorporates realtime information from other state-level labor market indicators. We find that across the 50 US states, the model reduces the average size of benchmark revisions by about 9 percent. When we optimally average the model’s predictions with those of existing models, we find that we can reduce the average size of the revisions by about 15 percent.

Keywords: dynamic factor model, Employment, Data revisions, nowcasting

JEL Classification: C53, E24, R11