Mind the (Approximation) Gap: A Robustness Analysis

This note continues the discussion of the results reported by Ricardo Caballero and Eduardo Engel (1993), hereafter CE, and Ricardo Caballero, Eduardo Engel, and John Haltiwanger (1997), hereafter CEH, by responding to the results reported in Christian Bayer (2008). Russell Cooper and Jonathan Willis (2004), hereafter CW, find that the aggregate nonlinearities reported in CE and CEH may be the consequence of mismeasurement of the employment gap rather than nonlinearities in plant-level adjustment. Bayer reassesses this finding in the context of the CE model in the case where static employment gaps are observed and concludes that the CW result is not robust to alternative shock processes. We concur with Bayer's assessment that the nonlinearity finding is sensitive to the aggregate profitability shock process. We argue, however, that Bayer's finding does not imply that the mismeasurement problem goes away. Instead, the nonlinearity created by mismeasurement is directly related to the level of the aggregate shock. Once the empirical specification properly incorporates the aggregate shock, the nonlinearity test is robust to alternative shock processes and confirms the results in CW. More importantly, we demonstrate that the CW findings are robust to alternative shock processes for the natural case of unobserved gaps as examined by CE and CEH.