While it is tempting to infer the welfare effects of minimum wage changes from empirical observations on pre- and post-change wage distributions, in this exercise we have attempted to point out the hazards of doing so. We have focused on wage distributions in this paper, but this statement applies with equal force tot the case in which the lack of change in employment levels following a minimum wage increase is taken to imply welfare increases. The welfare criterion utilized in this paper, which is motivated by a simple equilibrium matching and bargaining model, reflects both employment probability and wage distribution effects of minimum wage changes and hence is preferable to any measure which takes into account only employment or wage information. While the value of the welfare measure we haven chosen is open to question, we would argue that whatever measure is finally chosen, a formal model of the labor market is required in order to meaningfully interpret minimum wage impacts on labor market outcomes. The small empirical applications we have presented usefully summarizes the general points we wish to make. First, the fact that the wage offer distribution in 1998 first order stochastically dominates the 1997 wage offer distribution does not necessarily imply an increase in welfare. Second, while we found no evidence of spillover resulting from the minimum wage increase of September 1997, within the context of our model this could only be taken to imply that there were no welfare effects of the minimum wage increase if the matching distribution satisfied condition 1 of Proposition 2. To determine whether or not this is the case requires that specific tests be conducted using information from the wage distribution above the minimum or that the equilibrium model be directly estimated and tested. Third, we have demonstrated that employment rate declines are perfectly consistent with increases in welfare. Finally, we have shown that the existence of spillover effects does not imply that a given minimum wage change was beneficial. Spillover can be good or bad, and can only be judged as beneficial within a particular model of the labor market.