Modeling and Forecasting Income Inequality in the United States

Recently, an idea has emerged that “the rich are getting richer and the poor are getting poorer”. Using tax data from Piketty, Saez, and Zucman (2017) (updated in the World Wealth& Income Database) and internal microdata from the Current Population Survey (1975-2015),this paper models inequality and performs pseudo-out-of-sample (2012-2015) and true out-of-sample (2016-2018) forecasts for 5 income inequality measures. The lowest forecast errors from the best models are found for distributional metrics, as compared to top income shares. While macroeconomic indicators, human capital, and labor force metrics often enhance models, measures of skill biased technological change are not found to be robust predictors of inequality trends. Näıve approaches often outperform more complex models and forecasts differ between models by <4% for all variables.