New Majority Index

Methodology

The New Majority Index is the first elections rating system to incorporate racial demographics and disaggregated racial voter participation data into its electoral analysis. I am deeply indebted to the data scientist Dr. Julie Martinez Ortega for lending her statistical expertise, number-crunching wizardry, and culturally-informed political acumen to the project. This page will describe the methodology behind the NMI ratings.

Our rating system, like that used to calculate the Democratic Performance Index (“DPI”) and the Partisan Voting Index (“PVI”) compares partisan election results in individual districts to the national average, and presents a score indicating how close or far that district is from the national average. The traditional DPI is based entirely on the past performance of each major party within each specific state or district. The average of the partisan split found in the most recent two presidential cycles at the national level, using the official result tallies provided by the Federal Election Commission, form the basis for all comparisons when calculating the DPI. 

The NMI recognizes the utility of the DPI’s overall framework by beginning its calculations with this well-established approach to determining a district’s partisan lean. Since the Cook Political Report has a long and respected track record calculating its Partisan Voting Index, and since they rely on publicly available data through Dave’s Redistricting App for their calculations, we accept their determination of the district-level voting percentages for the respective presidential candidates in 2016 and 2020. We then use a different formula from Cook in calculating the composite partisan voting average for each district in that we believe that the 50/50 average remains relevant and revealing. And then we go two steps further by incorporating data on racial demographics. 

First, we add in the racial composition of each congressional district by looking at that district’s Citizen Voting Age Population (CVAP), disaggregated by racial group. Next, we weight the relative size of the individual groups of color within that district’s overall population of People of Color. We then analyze the voter turnout rates for each racial group within that district and arrive at an overall voter turnout rate for whites and People of Color. Those numbers were used to determine the racial voter gap in each district, and then we combine the DPI, the CVAP, and the racial voter gap to arrive at the NMI. 

In developing this index, we used multiple data sources in our analyses.

The biennial Current Population Survey’s November Supplement on Voting and Registration P20 tables provided information about the self-reported registration status and turnout rates (whether they voted) in the most recent election by racial group. The Citizen Voting Age Population (CVAP) Special Tabulation from the 2016-2020 5-Year American Community Survey provided information on the underlying racial composition of a district’s eligible voters. 

Each state produces and publishes data on its federal election results as well as on its registered voters, usually via its Secretary of State office. The quality and quantity of information varies greatly between the states, but information is available publicly for all 50 states.

Each state produces at least minimal information about the process used in its redistricting efforts. Some states, such as Arizona, that use a non-partisan election commission for redistricting, provide a great deal of data on the newly drawn districts, including the racial and partisan preferences information on those new districts.  In other states, some state legislatures produce and share extensive data via their redistricting committees. We relied upon all such relevant data that we were able to secure to inform our analysis.   

The recent changes made to congressional district lines post-redistricting based on the 2020 Census presented a challenge unique to the specific time during which we were working. Many states had not yet released sufficient data for all factors we were interested in analyzing on the newly drawn districts. In particular, 2016 and 2020 partisan preferences, CVAP by race, and the turnout rates by race were data points that we needed by congressional district, but that did not yet exist for all 435 congressional districts in a consistent format. We commissioned TargetSmart to calculate this information for each of the 435 CDs, all 50 states, and the state legislative districts in all 50 states. 

In addition to the information on each congressional district, we used the Federal Election Commission’s Official Presidential General Election results for the past two presidential election cycles to confirm the national partisan shares won by each major party. This is standard practice when doing traditional calculations of DPI.

The underlying data for all 435 congressional districts can be found in this spreadsheet.