A growing thread of research uses Biracials—those who exist at the intersection of our major social cleavages (racial groups)—to reveal the current nature and future trajectory of our racial hierarchy. Specifically, researchers explore whether Minority-White Biracials (those with one White parent and one Minority parent) tend to be more similar to either Whites or to their Minor- ity counterparts. The former circumstance would suggest a trajectory of assimilation for racial minority groups and waning intergroup prejudice, while the latter augurs enduring racial group boundaries and continued minority subjugation. Existing studies provide tremendous contribu- tions to this genre, but are constrained in their data and methodology. In this study, I offer new data which measures Biracials by parentage (an important circumvention of endogeneity) and a machine learning approach which can use hundreds of variables at a time in order to measure how Biracials compare to their single-race counterparts. In terms of political attitudes, Black- White Biracials are more simliar to Blacks, while Asian-Whites exhibit political thinking approximating that of single-race Whites. Latino-Whites remain ”in-between” their coutererpart groups.

The purpose of this dissertation is to introduce and test a new theoretical framework for explaining and predicting the political behavior of mixed-race individuals I call relative racial salience (RRS). Relative racial salience is defined as the relative salience of one of the racial groups within a mixed-race individual’s heritage, as compared to the other, within a given social context. Through the RRS framework I argue that racial group boundaries are relatively permeable for mixed-race individuals, meaning they can avail themselves of the different racial identities in their heritage as they vary contextually in fit and accessibility. In the first section, I draw from an unprecedented wealth of new and independently crafted datasets to demonstrate that racial context is indeed a basal component in the construction and expression of mixed-race individuals’ racial identity choices, partisanship, political attitudes, and candidate evaluations. In the second section, I use voter files to introduce a unique model of causal assessment that substantiates the likely causal effect of racial context on Biracials’ partisanship. In the third and final empirical section, I leverage ensemble machine learning categorization which takes into account every single variable available in my two largest datasets (800+ variables) to provide the most comprehensive empirical portrait of the general political characteristics of mixed-race individuals to date. In doing so, I demonstrate that Latino-Whites and Asian-Whites are actually more similar to Whites than they are to their single-race minority counterparts (Latinos and Asians, respectively), while Blacks-Whites remain resolute in their resemblance of Blacks. I comport these findings as evidence that the continuation of current demographic trends is likely to lead to fading boundaries between Latinos, Asians, and Whites, and the reification of the racial group boundary separating Blacks from all others as the most salient color line in the United States.