Production Function Estimation in Stata using Inputs to Control for Unobservables

A key issue in the estimation of production functions is the correlation between unobservable productivity shocks and input levels. Profit-maximizing firms respond to positive productivity shocks by expanding output, which requires additional inputs. Negative shocks lead firms to pare back output, decreasing their input usage. Olley and Pakes (1996) develop an estimator that uses investment as a proxy for these unobservable shocks. More recently, Levinsohn and Petrin (2003a) introduce an estimator that uses intermediate inputs as proxies, arguing that intermediates may respond more smoothly to productivity shocks. This paper reviews Levinsohn and Petrin's approach and introduces a Stata command that implements it.