Nowcasting the Income Distribution

Abstract

This paper proposes a framework for nowcasting the income distribution in real time. Decile means of the annual U.S. household income distribution are combined with a panel of about 80 quarterly macroeconomic indicators in a mixed-frequency dynamic factor model (DFM) and evaluated pseudo-real-time from 1990 to 2024. Relative to an autoregressive benchmark, the DFM reduces decile-level nowcast errors by 16% on average and by up to 25% in the middle of the distribution; improvements are smallest at the two tails, with the volatile bottom decile the hardest to nowcast. The DFM also substantially improves directional accuracy, raising the share of detected turning points from 7% under the benchmark to 45%. The signal is available about eighteen months before the official release, and a news decomposition shows that labor-market releases carry more weight at the bottom of the distribution and price releases at the top. For inequality measures, the DFM does not improve on the benchmark. The framework complements microsimulation and functional approaches by supplying timely, evaluable decile-level signals from standard state-space tools for distributional surveillance and as real-time priors when micro data are delayed.