In lieu of an abstract, here is a brief excerpt of the content:

  • Comments
  • Roberto Rigobon and Kevin Lang

Roberto Rigobon: This paper studies the effect of minimum wages on the distribution of wages and on overall unemployment. This is an extremely important question, and one that has barely been analyzed in emerging markets. Brazil's minimum wages provide a unique opportunity to study this question because changes in minimum wages have been large and frequent.

The paper mainly consists of two parts: a study of the impact of minimum wages on the wage distribution and an analysis of the implications of minimum wages for unemployment.

To study the influence of minimum wages on the wage distribution, Lemos examines different measures of the tightness of minimum wages. The specification that she uses is the following:

All her results are well summarized in table 1.

To study the impact on employment, Lemos uses the same measures and evaluates the impact of its changes on total employment, total hours, and jobs. This is done to appraise the different margins in which minimum wages can work. For instance, an increase in the minimum wage could produce a decline in the hours worked while keeping the same number of jobs, or it could reduce the number of jobs while keeping the hours per worker intact. Her specification is

The results for this specification for the different measures of minimum wages and employment are presented in table 3. [End Page 253]

Her main findings are threefold. First, the wage distribution experiences significant compression regardless of the measures of minimum wage tightness used in the specification. Second, employment measures are almost unaffected—and if they are at all, the effects are very small. Third, the preferred measures for evaluating the impact of the minimum wage are the real fraction of workers affected and the spike.

I organize my comments along two lines. First, although I have no problems in principle with the spike variable, I have severe doubts that the real fraction of workers affected should be used at all. Second, I examine what type of robustness tests should be performed to ensure that the results are not driven by unobservable variables.

The objective of the fraction of workers affected variable is to measure the proportion of workers for whom the minimum wage wasn't binding at time t - 1, but is binding at time t. This is measured as the mass of workers who have a wage that satisfies equation 1:

where MW is the nominal minimum wage and w is the wage. The idea of the real fraction of workers affected is to use real wages instead of nominal ones:

Equation 1 measures the proportion of workers who have wages close to the minimum wage and who are likely to find it binding. Therefore, this is a clear measure of how tight the minimum wage is. This interpretation does not apply to real minimum wages, however. For example, movements in the inflation rate will imply changes in the real minimum wage that are not necessarily associated with changes in the tightening of the minimum wage. Moreover, the presence of inflation changes the interpretation of the results. If inflation has different effects on the nominal wages of individuals along the distribution, then changes in real wages could be the result of these differences and not the outcome of the tightening of the minimum wage. In particular, if inflation increases the nominal wages of individuals in the bottom of the distribution more than those in the top of the distribution, then the compression of the wage distribution could be explained by the omitted variable and not by the tightening of the [End Page 254] constraint—especially if the relation is nonlinear. If employment depends negatively in real wages, then increases in inflation together with increases in minimum wages will make the average effect on employment small, and possibly with the wrong sign. Lemos understands this inconvenience, and she introduces inflation on the right-hand side of the specification. Her specification is insufficient, however, if the relation between wages and inflation is not linear.

One way to test for this is to compare the level and first-difference estimates, for instance by estimating the following equations:

and

If both equations...

pdf

Share