Computing Capital Stocks in the German Social Security Records and Quantifying Their Role for Wage Inequality

Abstract We develop a method to impute capital stocks from investments for a sub-sample of firms in the German social security records and implement a machine-learning algorithm to predict capital stocks for the universe of firms. These capital stocks explain 40% of the variation in capital stocks of the Bureau van Dijk data. We make our data available for other researchers. We find that these capital stocks explain a sizeable fraction of wage inequality by extending the variance decomposition of Card et al. (2013), suggesting that rising firm heterogeneity in capital intensity may further amplify wage inequality. (JEL codes: C81, D24, and J31)