Work in progress


formerly titled "The Cost Share Approach to Production Functions"

paper   slides   Stata code

Abstract: I introduce a new method to estimate heterogeneous output elasticities, markups, and revenue productivity using only standard firm-level revenue and expenditure data, which avoids assumptions on demand, the evolution of productivity, and output measurement. My approach circumvents common biases that affect existing methods in non-competitive settings by exploiting firms' optimizing behavior. To correct for bias from measurement error in input costs (e.g., dynamic inputs like capital), I develop a two-stage procedure that purges this error non-parametrically, and then uses corrected cost shares to estimate firm-time-specific elasticities, markups, and revenue productivity. Monte Carlo simulations confirm the estimator's accuracy, even with substantial measurement error from multiple sources. Applying my method to Compustat, I find markup dispersion has increased almost four-fold since the 1960s and drives around 73% of revenue productivity dispersion. Among Compustat firms, superstar outcomes in markups and productivity are associated not with the largest firms, but with smaller, more sunk-cost-intensive companies.

Winner of the FJPB Young Economist Prize (JEI 2025)


paper (old)   slides (new)

Abstract: Under the assumptions of constant returns to scale production technologies and competitive input markets, the sum of revenue elasticities across all inputs is the reciprocal of the price-cost markup for a profit-maximising firm. Profit maximisation also implies non-parametric bounds for arithmetic means of revenue elasticities in a large sample of firms. These bounds are available for both flexible and predetermined inputs, and require data only on revenue shares (i.e. input costs as a share of revenue). Summing over inputs, we thus obtain empirical bounds for harmonic means of price-cost markups. Using company accounts data from Compustat, we calculate these bounds for publicly traded North American firms over the period 1955-2022. Our results are consistent with a modest rise in the sales-weighted harmonic mean of markups since the early 1980s, and a considerable increase in the cross-section dispersion of markups.