A closer look at the gap. A comment on Cooper and Willis' 'mind the gap' paper

Recently, there has been a lively debate between Cooper and Willis (2001,2002,2003a, 2003b) and Caballero and Engel (2004) about the apropriateness of the so-called 'gap approach' to labor adjustment. Cooper and Willis claim that the gap approach is unable to identify non- convex adjustment costs because of a measurement error under the alternative hypothesis of convex costs. This comment assesses the validity of Cooper and Willis' claim by providing evidence from a number of Monte-Carlo experiments. In contrast to Cooper and Willis findings from single simulations, the experiments reveal no tendency to falsely reject the convex-cost hypothesis if one uses the correct one-sided test for non-convexities. In fact, the parameter estimates are typically biased against the hypothesis of non-convex costs. Consequently, there is no tendency to falsely reject although the estimates show substantial excess dispersion as a result of a spurious regression problem.