Misallocation or mismeasurement?
What is this research about and why did you do it?
A determinant of aggregate productivity differences, both across countries and within countries over time, is how well resources such as capital and labour are allocated across firms. The importance of such resource misallocation for aggregate productivity is often inferred from the dispersion in average revenue products (revenues over inputs) in firm-level data. Measurement error, for example due to omitted or double-counted revenue or inputs, implies that measured gaps in average products do not reflect true gaps in marginal products. This could thereby amplify dispersion in measured average products and lead researchers to overstate the magnitude of misallocation.
How did you answer this question?
We propose a methodology to correct estimates of misallocation for measurement error in revenue and inputs. Our approach exploits how revenue growth is less sensitive to input growth when a plant’s average products are overstated by measurement error. The key assumption is that the measurement errors are orthogonal to (i.e., uncorrelated with) the true marginal products. We apply our methodology to data on Indian manufacturing plants from 1985 to 2013 and U.S. Census data on manufacturing plants from 1978 to 2013.
Data sources are the Indian ASI and U.S. LRD. The India sub-figure shows uncorrected and corrected allocative efficiency for years 1985 to 2013. Average uncorrected allocative efficiency is 47.7% while average corrected allocative efficiency is 53.4%. The U.S. sub-figure show uncorrected and corrected allocative efficiency for years 1978 to 2013. Average uncorrected allocative efficiency is 47.6% while average corrected allocative efficiency is 67.4%.
What did you find?
We find that our correction lowers the potential gains from reallocation by 20% in India, but that measurement error is even more severe in the U.S. On average, our correction lowers the potential gains from reallocation in the U.S. by 60% between 1978 and 2013. Strikingly, measured revenue productivity dispersion in the U.S. exhibits a sharp upward trend, seemingly implying that misallocation increased dramatically from 1978 to 2013. However, we find that rising measurement error in U.S. plant-level data accounts for most of this trend, and that corrected allocative efficiency declined by considerably less.
What implications does this have for the research on wealth concentration or economic inequality?
Many policy choices involve trade-offs between social goals, such as reducing inequality and economic efficiency. For example, flexible labor laws may make it easier for firms to hire and fire workers but may not be desirable for workers seeking job security. Our methodology shows that the estimated efficiency gains from reallocation can be overstated due to measurement error and provides researchers a set of tools to estimate these more accurately.
What are the next steps in your agenda?
Future research should further decompose dispersion in average products into distortions which result in misallocation and can be addressed by policy, and other factors which don’t reflect true misallocation, such as measurement error or unavoidable adjustment costs or transportation costs.
This paper can be cited as follows: Bils, M., Klenow, P. J., and Ruane, C. (2021) "Misallocation or mismeasurement?" Journal of Monetary Economics, 124 (Supplement), pp. S39-S56.