Journal Issue: Opportunity in America Volume 16 Number 2 Fall 2006
The Persistence of Ethnic Wage Differences
As noted, ethnic groups vary widely in socioeconomic status both in the first generation and in the second. Some national origin groups, typically those from advanced economies, do quite well in the U.S. labor market, while others, typically those from poorer countries, fare much worse. Table 2 shows some of this variation for selected groups.11 In 1970, for instance, immigrants from Canada earned 18.5 percent more than the typical worker in the baseline third generation, while immigrants from Mexico earned 31.6 percent less. By 2000, second-generation workers from Canada earned 16.8 percent more than the typical third-generation worker, while second-generation workers from Mexico earned 14.7 percent less.
To determine the extent to which ethnic wage differences among immigrants persist into the second generation, some studies estimate statistical models that relate the relative wage of a second-generation national origin group to that of its first-generation counterpart.12 These analyses account for the fact that first and second-generation workers in a single cross-section of data have little biological connection by linking the relative earnings of second- generation workers at a particular time (for example, the 2000 cross-section) to the earnings of first-generation workers a few decades past (for example, the 1970 census).
Before summarizing the evidence, I will illustrate the nature of the exercise. Figure 1 shows the intergenerational link for male workers from many national origin groups between 1970 and 2000.13 The horizontal axis gives the age-adjusted relative wage of men in the immigrant generation, using data from the 1970 census. The vertical axis gives the age-adjusted relative wage of men in the second generation, using data from the 2000 cross-section. The correlation between the average wages of workers in the two generations is obviously strong and positive; the groups that fared well economically in the first generation also did well in the second.
The upward-sloping line in figure 1 summarizes the statistical relationship that links the relative wages of particular national origin groups across the two generations. The slope of this line, often called the intergenerational correlation in relative wages, measures the extent of regression toward the mean across generations. A relatively flat line would show little connection between the average skills of the ethnic groups in the second generation and those of the immigrant groups. Put differently, all second-generation groups would have relatively similar wages, regardless of the economic performance of their parents. The intergenerational correlation would be near zero, and the regression toward the mean would be complete. A relatively steep line, by contrast, would show a close link between relative wages in the first and second generations. If the intergenerational correlation were equal to one, for example, the relative status of the ethnic groups in the two generations would be identical and the line would have a 45 degree slope. In an extreme case, if the typical worker in a particular immigrant group earns 30 percent more than a third-generation worker, the typical second-generation descendant of that group would also earn 30 percent more than the third generation. There is no regression toward the mean because the ethnic differences remain the same from generation to generation.
Table 3 reports the estimated intergenerational correlations over both 1940–70 and 1970–2000. Among working men, the correlation (row 2) is 0.511 for 1940–70 and 0.560 for 1970–2000. Among working women (row 3), the correlations are smaller: 0.242 for 1940–70 and 0.280 for 1970–2000. As noted, however, the intergenerational changes in the relative wage of women may reflect the dramatic increase in the rate of female labor force participation over these six decades.
The bottom panel of table 3 reports the estimated intergenerational correlations after earnings are also adjusted for differences in schooling. This adjustment weakens the correlations: the correlation in the sample of working men, for example, drops by nearly half to 0.287 during 1940–70 and to 0.245 during 1970–2000. In short, a primary reason why ethnic wage differences persist across generations is the persistence of ethnic differences in schooling.
The estimated correlations for working men reported in the top panel of the table suggest two important conclusions. First, the inter generational correlation between the skills of the first and second generations is about halfway between zero and one. A correlation of 0.5 implies that half of the wage difference between any two national origin groups in the first generation persists into the second generation. If the average wage of two ethnic groups is 30 percentage points apart in the first generation, it is expected to be about 15 percentage points apart in the second. There is some social mobility, therefore, but ethnicity continues to have a large effect on labor market outcomes in the second generation.
Second, the estimates of the intergenerational correlation between 1940 and 2000 are remarkably stable. The process linking the economic performance of first- and second-generation ethnic groups was quite similar over those six decades, despite major changes in economic and social conditions, as well as in immigration policy.14
It is also important to determine whether the ethnic wage differences that remain in the second generation are transmitted to the immigrants' grandchildren. To establish how long a person's ethnic background matters for wage outcomes, one can track the economic performance of the grandchildren of the immigrants who entered the United States during the “first” Great Migration, at the beginning of the twentieth century.15 One would use data from the 1910 census to get information on the skill level of the national origin groups that made up the first Great Migration and use the General Social Surveys to get information on the sample of American-born workers (around 1985) with at least one grandparent born outside the United States.
Figure 2 summarizes some of the data and shows clear differences in relative wages among third-generation ethnic groups—although the differences are far smaller than those in the first generation. Nevertheless, even after three-quarters of a century, a positive correlation remains between the relative wage of the original immigrant groups and that of the corresponding third-generation ethnic groups. The slope of the line linking the relative wage of the first and the third generations implies an intergenerational correlation of 0.22, meaning that 22 percent of the wage gap between any two groups in the immigrant generation persisted into the third. Recall that roughly half of the wage gap between any two immigrant groups disappears between the first and second generations. It seems that about half of what remains in the second generation disappears between the second and the third.16
The historical record suggests one broad generalization. The “half-life” of ethnic skill differences is roughly one generation: about half of the ethnic differences in relative wages disappears in each generation. Put differently, a 20 percentage point wage gap among ethnic groups in the immigrant generation implies a 10 point gap among second-generation groups and a 5 point gap among third-generation groups.