Model specification, data selection, and the employment effect of minimum wage

Abstract

This study aims to show that the effect of minimum wage on employment highly depends on the model specification. To verify this hypothesis, we use a publicly available dataset on employment and minimum wage from the article of Dube et al. (2010) for US counties in 1990–2006. We replicate the minimum wage employment equation using different model specifications and data subsets to deduce the empirical distribution of the minimum wage parameter. In addition, we verify the relationship between the sample size and the significance of the estimated minimum wage parameter. Our research confirms that the specification of the econometric model determines the results of the minimum wage employment relationship. Different model specifications lead to different results and are incomparable. The results show that allowing for differences in the time trends in different survey areas nullifies the relationship between minimum wage and employment. This was observed only in a few counties, so it was difficult to find a statistically significant relationship at the country level. In addition, our study shows that data selection has no significant impact on the results.