Posted on November 15, 2018

Grandchildren of Low-Skill Immigrants Have Lagging Education and Earnings

Jason Richwine, Center for Immigration Studies, November 15, 2018


The intergenerational assimilation of low-skill immigrants is an important issue in the broader immigration debate. If the children and grandchildren of low-skill immigrants eventually rise to the same socioeconomic level as natives, then the poverty-related problems caused by low-skill immigration, though painful today, will dissipate over time.

Past research on the assimilation of Americans who have ancestors from Mexico (the largest source of low-skill immigration to the United States) indicates that Mexican-Americans continue to lag behind in the third generation and beyond. However, reliance on survey respondents self-identifying as Mexican-American has made this research less than definitive, as not everyone with Mexican-born grandparents has retained a Mexican-American identity.

To solve this “ethnic attrition” problem, researchers have turned to the 1997 panel of the National Longitudinal Survey of Youth (NLSY-97). Unique among government data sets, the NLSY-97 includes grandparent birth data that can objectively identify the Mexican-American third generation with no ethnic attrition. In replicating and extending a recent NBER working paper that uses the NLSY-97, this report affirms that although Mexican-Americans make progress over time, the third generation still has significantly lower education and earnings compared to fourth-plus generation white Americans.

Unfortunately, because the NLSY-97 data set is relatively small, the magnitudes of these difference are subject to considerable uncertainty. If grandparent birthplaces were incorporated into a larger data set such as the Current Population Survey, much more detailed and precise estimates could be calculated.


Although the average education level of immigrants to the United States has been rising, the post-1965 immigration wave remains markedly less skilled than the native population. Immigrants currently account for 40 percent of all working-age Americans without a high school diploma, more than double their percentage of the general population.

The impact of this low-skill immigration has been mixed. Like workers of all education levels, low-skill immigrants help reduce the price of goods and services that require their labor. However, their low earnings exacerbate the problems of poverty, welfare dependence, poor academic performance, and social isolation, and they compete in the labor market with U.S. natives who struggle with similar issues. Immigration restrictionists have long argued that these costs outweigh the benefits.

Many supporters of low-skill immigration view the costs as only temporary, however. They predict that the children and grandchildren of low-skill immigrants will climb the ladder, eventually matching natives on key measures such as education and income. As evidence, they often cite the experience of European immigrants from a century ago. “Nativists said the same thing about the Irish!” is a familiar rebuttal to doubts about the long-term assimilation of post-1965 immigrants, the least skilled of whom have come mainly from Mexico and other parts of Latin America. But the long-term assimilation of Latin Americans has always been more a matter of faith than of data. This report adds to the literature showing that education and earnings deficits persist for the grandchildren of Mexican immigrants.



All results are displayed in Table 1. Rows 1 to 4 list the different Mexican generations analyzed here. The “1.5 generation” refers to respondents born in Mexico, with the “.5” serving as a reminder that, unlike most immigrants, NLSY-97 immigrants arrived in the United States as children. As described earlier, a second-generation Mexican-American is the child of at least one Mexican immigrant, and the third-generation refers to the grandchildren of at least one Mexican immigrant. The two third-generation groups are the cross-section and the ethnic sample, which are defined in the previous section. Rows 5 and 6 are the white and black fourth-plus generation groups, respectively, which serve as points of comparison for the Mexican American generations.

(Click on the table to enlarge it.)

In rows 1 to 6, the average value of an outcome measure is displayed at the top of the cell, with the standard error below it in parentheses, and the sample size below that in the form of “n=”. For example, third-generation Mexican Americans from the ethnic sample have an average AFQT percentile score of 40.2. The standard error of that average score is 2.7, and the sample size is 128.

Rows 7 and 8 give the differences between the fourth-plus generation whites and the two third-generation Mexican American groups, along with a confidence interval. For example, fourth-plus generation whites score 17.1 AFQT percentile points higher than third-generation Mexican-Americans from the ethnic sample, and the 90 percent confidence interval around that difference extends 4.8 percentile points in either direction — i.e., from 12.3 to 21.9. Because the confidence interval does not contain zero, the difference is statistically significant at the 90 percent level.

Table 1 replicates Duncan et al.’s finding that third-generation Mexican-Americans have essentially the same high school graduation rate as fourth-plus generation white Americans, but their college graduation rate remains well behind. Moving beyond Duncan et al.’s findings, there is a substantial difference in AFQT scores between third-generation Mexican-Americans and fourth-plus generation whites.8 Similarly, although the number of work weeks are about equal, fourth-plus generation whites have greater weekly earnings than third-generation Mexican-Americans. As the confidence intervals in rows 7 and 8 indicate, the white-Mexican differences in college graduation rate, AFQT score, and weekly earnings are all statistically significant, but the exact magnitudes of the differences exhibit a high degree of uncertainty because of the small sample sizes.

None of the differences between the two third-generation Mexican American groups are statistically significant, but they are still worth noting. The cross-section’s apparent advantage over the ethnic sample in college graduation is consistent with the expectation that higher-achieving members of the third generation are less likely to identify as Hispanic. The cross-section’s advantage in AFQT is less pronounced, however, and the work and income differences appear to go in the wrong direction. More research is needed before drawing strong conclusions about the effect of ethnic attrition. The cross-sectional results do show, however, that ethnic attrition is not somehow concealing the complete assimilation of people with Mexican-born grandparents. Even with a sample that has no ethnic attrition, third-generation deficits remain significant for Mexican Americans.



Based on third-generation data constructed without regard to ethnic self-identity, the grandchildren of Mexican Americans improve upon their parents’ low average education. However, significant deficits remain. Compared to fourth-plus generation white Americans, third-generation Mexican Americans graduate from college less often, score lower on the AFQT, and earn less income. Though all of these differences are statistically significant, the exact magnitudes are subject to considerable uncertainty due to the small sample size in the NLSY-97. Including grandparent birth data in a larger data set such as the Current Population Survey would facilitate more detailed and precise comparisons. For now, the best available evidence on the grandchildren of Mexican immigrants suggests that low-skill immigrants have long-term impacts, as skill deficits persist at least into the third generation.

[Editor’s Note: The original article includes information on previous research.]