Decomposing Differences in Productivity Distributions

Abstract

I analyse the post-crisis slowdown in UK productivity growth using a novel decomposition framework, applied to firm-level data. The framework tracks flexibly defined distributions over time, and links changes in the shape of these distributions to aggregate movements. It encompasses many existing methods, which typically track firms over time, and also provides opportunities for various new types of analysis, particularly where firms are not repeatedly observed in survey data. In my application, I show that the slowdown in productivity growth is driven entirely by post-crisis reallocations of workers to firms with less-productive characteristics, rather than changes in the productivity associated with these characteristics (which have actually supported growth since the crisis). I further show that the puzzle is located in the top tail of the distribution, as is the negative contribution from these allocation effects.