15We test the sensitivity of the findings to perturbations in the estimated optimal productivity by decreasing or increasing by 1%, 5%, and 10% the optimal productivity estimates to determine therobustness of the conclusions to different levels of error. The results show the changes in 1 are only in the third decimal place, and the t-values are essentially unchanged, indicating that the test for H1 is robust to error in estimating optimal productivity.16These values are interpretable as the average P* in Equation 22.17To test the robustness of the time trend in optimal productivity, we also calculated the mean estimated optimal productivity for 2005 (.272) and 2009 (1.136). Together with the results from 2002 and 2007, the four years have significantly different means (F = 504.67, p < .001), the time ordering of the productivity levels is as expected, and the trend is in the hypothesized direction.