Decomposing the Rise in Markups

A recent literature argues that the rise in aggregate markups is due to economic activity reallocating toward high-markup firms rather than an increase in firm-level markups. I show that this result is biased by three types of measurement error. First, standard decomposition methods estimate a positive reallocation effect even when the allocation does not change. Second, firm-level markups are mismeasured when the production function is misspecified. If this is the case, the decomposition can give opposite results depending on whether actual or measured markups are used. Third, the decomposition is sensitive to measurement error of the variables sales and costs. Correcting for each type of measurement error suggests that the reallocation effect is smaller than previously thought and perhaps even zero or negative. Moreover, correcting for measurement error of sales lowers the estimated sales-weighted average markup, and I find that the estimated increase in the sales-weighted markup over the last 60 years in Compustat is in its entirety due to a rise in measurement error. The cost-weighted average markup is not affected by measurement error and is therefore a more robust estimate of market power.