After moving to the U.S., research has shown that some ethnic groups fare better than others. Specifically, research has suggested that Asian immigrants tend to do better once arriving in the U.S.—quickly earning college degrees and climbing the economic ladder—than Hispanic immigrants. But it’s also true that as some immigrant families become more assimilated, children and grandchildren may cease to identify with their country and ethnicity of origin. And that can make the data used to determine progress a bit more difficult to accurately interpret, according to a new paper from National Bureau of Economic Research.

The study’s authors, Brian Duncan and Stephen J. Trejo, identify this phenomenon—when future generations no longer identify with the race or ethnicity of their parents or grandparents—as “ethnic attrition.”

Most methods for evaluating how different ethnic groups are faring require only self identification and the country of an individual’s birth, with little data about the immigration status of previous generations. Thus when the children and grandchildren of immigrants don’t identify themselves as a part of a particular ethnic group, data about economic progress can be skewed. The lack of ethnic identification among descendants is pretty important because the most significant progress in socioeconomic mobility happens across generations—from parents, to children, to grandchildren—not within one individual’s lifespan.

In order to get a better picture of immigrant groups, the authors parse detailed data from the Current Population Survey, which asks individuals not only about their country of birth, but their parents as well (information not provided to other surveys such as the Census or the American Community survey). By putting together more than a decades’ worth of CPS data the authors are able to piece together a more robust three-generation portrait of families that includes the more concrete measure of country of origin, in addition to subjective identification. (The data is nevertheless limited for capturing the third generation because researchers included only biological children of married parents.)

They find that attrition is fairly common among the children and grandchildren of Asian and Hispanic immigrants. Ninety-nine percent of first-generation  Hispanic immigrants identify with their ethnicity, compared to 82 percent of the third generation.The decline in identifying with the country of one’s family is even more precipitous for Asian Americans. About 96 percent of first-generation Asian immigrants participated in ethnic identification, but by the third generation that number was only 58 percent. Attrition also differs based on country of origin, for instance Cubans and Indians are less likely to ethnically identify, while Mexican and Chinese immigrants are more.

These patterns are not random but follow along socioeconomic divides. The researchers note that educational achievement is often the best way to predict economic attainment and class. They find that more educated Hispanics who are second and third generation immigrants are less likely to ethnically identify. But the reverse was true for Asians, a group where the less educated cohort was more likely to cast off their ethnic identifiers. The authors suggest that findings such as this may cause an underestimation of the performance of Hispanic Americans in the U.S. and an overestimation of the success of Asian Americans, since the most educated, economically successful are either largely left out of the data representing a specific group, or make up a disproportionately large share of it.

The authors hypothesize a few causes for ethnic attrition. For starters, the differences in how the children and grandchildren of immigrants identify may have origins based on the ethnic makeup of the other members of their family. Ethnic identification is more persistent among the second- and third-generation immigrants when more of their familial line hails from a particular country. For instance, having two sets of grandparents who immigrated from China, or if both parents parents hail from Mexico, an individual is more likely to also identify with their parent’s ethnicity than if their parents intermarried with other ethnic groups. This fact along with a disparity in the propensity to marry outside of one’s culture may also help explain the differences in what groups of Hispanic and Asian Americans choose to identify or not. Hispanics who intermarry tend to have higher educational attainment than those who marry  within their same culture, but for Asians the opposite is true. When added to the fact that intermarriage generally decreases ethnic identification, the result is that second- and third-generation Hispanics who have been more successful in upward socioeconomic mobility may not be included in counts of intergenerational mobility.

These rates of identification are important when considering the statistical and demographic characterizations of ethnic and immigrant groups. For instance, according to the study, the high-school dropout rate among third-generation immigrants is 25 percent higher for those who identify as Mexican. But nearly one-third of third-generation immigrants whose families hail from the country don’t even identify as Mexican, creating a potentially negatively skewed picture of the third-generation Mexican immigrant drop-out rate. That’s why the authors suggest involving more care and nuance in both obtaining and extrapolating data about immigrant populations and their ability to thrive in the U.S.