Is the Time-Series Evidence on Minimum Wage Effects Contaminated by Publication Bias?

Publication bias in economics may lead to selective specification searches that result in overreporting in the published literature of results consistent with economists' priors. In reassessing the published time-series studies on the employment effects of minimum wages, some recent research has reported evidence consistent with publication bias, and concluded that the most plausible explanation of this evidence is editors' and authors' tendencies to look for negative and statistically significant estimates of the employment effect of the minimum wage, (Card and Krueger, 1995a, p. 242). We present results indicating that the evidence is more consistent with a change in the estimated minimum wage effect over time than with publication bias. More generally, we demonstrate that existing approaches to testing for publication bias may generate spurious evidence of such bias when there are structural changes in some parameters. We then suggest an alternative strategy for testing for publication bias that is more immune to structural change. Although changing parameters may be uncommon in clinical trials on which most of the existing literature on publication bias is based, they are much more plausible in economics.