Association for Public Policy Analysis & Management

Estimates, projections, and demographic profiles of the Hispanic population are central to both academic and public policy debates about immigration, immigrant assimilation, and minority mobility.  Participants in these debates use population counts and forward-looking projections to portray both pessimistic (Coulter 2015; Krikorian 2008) and optimistic accounts (Frey 2014; Myers 2007) of what the growing Hispanic population portends for the future of the United States. 

The conclusions that analysts draw may be profoundly shaped by choices that respondents make about how to identify themselves (Waters 2002). Ample qualitative research and several quantitative studies show that these choices are, especially for Hispanics, fluid, situational, and shift over the life course (Lee and Bean 2010; Roth 2012; Smith 2014; Waters 1990; Saperstein and Penner 2012; Liebler et al. 2014). This research points to a need to examine national-level patterns and factors related to individual-level change in racial and ethnic identification in survey and census data. Researchers are limited in doing so, however, either because of small sample sizes inherent in longitudinal surveys or data restrictions on large government surveys.

We have access to restricted data from the 2000 and 2010 censuses that enable us to overcome the limitations of previous research. These data are unique in that individual records from the 2000 and 2010 censuses are linked and allow use to compare responses to the Hispanic origin question. These data reveal whether one stays, joins, or leaves the broad Hispanic category as well as a specific categories of Hispanic origin.

We investigate three main questions: To what extent does individual ethnic identification change over time and what patterns of change predominate? What proportion of such changes are associated with data collection and survey measurement issues such as non-response? And who is most likely to change between categories of ethnic identification?

We find that 7.2 million individuals, or 31 percent of linked records that identified as Hispanic in either census, had different responses to the Hispanic origin question. While identification changes resulted in net growth of the Hispanic population only by about 640,000, or 4.2 percent of total Hispanic population growth, the implications of identification changes for specific categories of Hispanic origin are more substantial. For example, changes account for net growth of 1.3 million in the Mexican origin population, 12 percent of total growth, and a net increase of 530,000 in the population with Central American origins, which accounts for 23 percent of the growth in this population.

We also model identification changes to understand what characteristics are most strongly associated with individual change. Our results show that those who change ethnic identification also often report inconsistent ancestry and ethnic origins, are minor children, are native born, and have experienced significant life changes such as divorce or state-to-state migration. Given the relative youth of the Hispanic population and its geographic dispersion in recent decades, our findings suggest that individual ethnic identification changes over time carry important implications for Hispanic population estimates and projections and thus both academic research and public policy debates.