Second, the fixed effect 5 * control for common factors that affect the rate of wage growth of immigrants who arrived at the same time and at the same age. The inclusion of these fixed effects effectively implies that the regression coefficient 6 would be numerically identical if the dependent variable had been defined in terms of the rate of wage growth of immigrants relative to that of comparably aged natives, or Avjjk{t,t’). The reason is that the native rate of wage growth is constant within a particular age group. The regression results reported below, therefore, can be interpreted as analyzing the determinants of the rate of wage convergence between immigrants and natives.

Finally, to ensure that the convergence regressions use the log “entry” wage as the independent variable, the analysis is restricted to the rate of wage growth observed during an immigrant cohort’s first ten years in the United States. As a result, the cohorts that arrived in the 1960s contribute only one observation to the sample, giving the wage growth between 1970 and 1980; and the cohorts that arrived in the 1970s also contribute one observation to the sample, giving the wage growth observed between 1980 and 1990. Of course, the wage at time t (the beginning of the decade) is a much better approximation of the entry wage for the immigrants who arrived in the last half of the preceding decade. Consider, for example, the cohort that arrived between 1965 and 1970. Equation (15) then relates the rate of wage growth over the period 1970-80 to the 1970 log wage. The 1970 wage, however, is not as good an approximation of the entry wage for the immigrants who arrived in the first half of the 1960s. I will show below that this rough approximation does not impart a serious bias on the analysis.