We evaluate the empirical evidence for costs that penalize changes in investment using U.S. industry data. In aggregate models, such investment adjustment costs have been introduced to help account for a variety of business cycle and asset market phenomena. So far no attempt has been made to estimate these costs directly at a disaggregated level. We consider an industry model with investment adjustment costs and estimate its parameters using generalized methods of moments. The findings indicate small costs associated with changing the flow of investment at the industry level. The weighted average of the industry elasticities with respect to the shadow price of capital, which depends inversely on the adjustment cost parameter, is eight times larger than the largest estimate reported in Levin et al. (2006). We examine a variety of factors that may account for this discrepancy, but a substantial part of it remains unexplained. Our results therefore suggest that more caution is needed when giving policy advice that hinges on a structural interpretation of large investment adjustment costs.