Immigration Undermines Affirmative Action in American States

A new argument against liberalized immigration has recently emerged: more immigration will increase affirmative action. There are two arguments for why more immigration could increase the scale and scope of affirmative action. The first is that most immigrants are not white. As a result, they would benefit from affirmative action and, thus, demand it. The second is that immigrants mostly vote for Democrats who support expanding affirmative action. To answer whether more immigration is correlated with affirmative action, we look at whether a larger immigrant population on the state level is correlated with affirmative action policies. Dominique J. Baker, assistant professor of education policy at Southern Methodist University, wrote a paper examining why some states ban affirmative action in state universities, government agencies, and public contracts. She uses a statistical method called discrete‐​time survival analysis to analyze a data set of 47 states from 1995 to 2012 to investigate which variables predict a state’s likelihood of adopting a statewide affirmative action ban. She concludes that a paucity of spots in higher education for white students is a good predictor of affirmative action bans. In other words, she finds that states are less likely to ban affirmative action when whites are a higher proportion of students in state flagship universities. This blog post borrows Baker’s methods to analyze how immigration affects whether states ban affirmative action. We identified the nine states with affirmative action bans from Baker’s paper. They are California, Texas, Washington, Florida, Michigan, Nebraska, Arizona, New Hampshire, and Oklahoma. Baker studied 47 states, excluding Texas, Hawaii, and Alaska. We included those three states in a separate analysis but they did not change the results, so we stuck with the 47 states that she used for this blog post to make the findings more comparable. Baker excluded Texas because after it banned affirmative action, the ban was then reversed by a court decision resulting in two Texas public universities restarting affirmative action admissions policies. We also extended the time horizon studied by Baker to the end of 2019. We then included data on the percentage of non‐​citizens, the share of the population that was white, and diversity as measured by a fractionalization index on the state level based on data from the IPUMS American Community Survey (ACS). Data are not available for intercensal years, so we linearly interpolated missing years to impute the annual share of non‐​citizens by state. We ran a multi‐​record parametric model with adopting a ban as the failure condition. Once a state adopts a ban on affirmative action, it does not “survive” and it exits the model. We code all 47 states as entering the dataset in 1995 and the observation period ends in 2019. For all three of the potential determinants (diversity, share non‐​citizens, and share whites), we run the model with no lag and with lags of 1 year, 2 years, 3 years, and 5 years. We use no controls. Our results differ from Baker’s results. The share of the white population is not statistically significant in any specification and the diversity measure is only significant at the 10 percent level in the 3‐​and‐​5‐​year lags (Table 1). As the number of lags increases, the coefficient on the share of the population that is foreign‐​born climbs in significance and generally in magnitude. The share of a state’s population who are non‐​citizens is significant at the 1 percent level in every lag. The magnitude of the coefficient on the share of non‐​citizens stays within a range of 0.27–0.34, meaning that a 1 percent increase in the share of non‐​citizens is associated with a 27–34 percent increase in the probability of the state banning affirmative action.