The main source of information for this volume is an original data set of federal politicians’ statements that relate to racial and ethnic minority concerns. The data consist of presidential statements publicly available through the Public Papers of the Presidents of the United States, provided by the U.S. Government Printing Office. Congress members’ statements are drawn from digitized documents in the Congressional Record that are publicly available through the U.S. Government Printing Office's Federal Digital Systems at the time of writing this book. The Government Printing Office provides the most comprehensive source of federal politicians’ statements and is widely used among academic scholars (see, e.g., Wood 2007; Coe and Schmidt 2012; Canon 1999).

Using these two sources, I first restricted the data to only oral remarks. This is particularly important for the Congressional Record, given that the Government Printing Office includes the written language of bills that representatives submit as well as congressional procedures. Using the paragraph structure provided by the clerks and staff at the Government Printing Office, I later divided oral remarks into paragraphs to indicate separate statements.

Once the data were divided into separate statements, I used a combination of human coding and computer programs to classify politicians’ remarks by whether they specifically addressed racial and ethnic minority issues. In the first stage of this process, two research assistants classified a random sample of presidential remarks (300 paragraphs for each presidential administration) and congressional statements (1,000 for each congressional session). The human coders classified only those statements that specifically referenced a racial group (e.g., African Americans or blacks), a race-specific policy (e.g., Affirmative Action), a celebrated minority figure (e.g., Dr. Martin Luther King or Rosa Parks), or a race-specific issue (e.g., racial profiling). Broader political statements on crime, education, and health that disproportionately impact racial and ethnic minorities were coded as dealing with race only when they specifically referenced a racial group, a racial policy, or a racial issue. Given that only explicit statements were used, the intercoder reliability was high at 97 percent. I was the final arbitrator of the conflicting statements and classified these remarks.

In the second stage of the process, I programmed computer-based classifiers to identify race-related statements based on the classification process the research assistants had created. The computer classifiers coded the entire data set of statements.