Who’s On Track For The GOP Nomination?

We’ve estimated how many delegates each candidate needs to win in each contest.

By G. Elliott Morris, Katie Marriner and Aaron Bycoffe

Read moreDownload data

*Guam, Montana, New Mexico and South Dakota are not included because they will not allocate their delegates, which will proceed to the national convention unbound. Michigan will have two separate elections for allocating delegates: The state's 16 at-large delegates will be allocated based on the results of Michigan's Feb. 27 primary, and its 39 congressional-district delegates will be allocated based on the results of district conventions at the party's March 2 state convention. Texas will also have two separate elections: 150 of the state’s 161 delegates will be allocated based on the results of Texas’s March 5 primary, while the remaining 11 (all at-large) will be allocated at the party’s May 23-25 state convention. Pennsylvania will allocate its 16 at-large delegates to the statewide winner of its April 23 primary, but its 51 congressional-district delegates will go to the national convention unbound. Oregon will formally allocate delegates at the party's May 25 state convention, but the convention will vote to award all 31 of its delegates to the statewide winner of the May 21 primary. Missouri and Wyoming will formally allocate delegates at their state conventions, but their caucuses and county conventions will indicate the likely delegate hauls.

Methodology: These benchmarks are based on estimates of how many delegates each candidate would need to win in each state and territory to secure a majority of the 2,429 delegates at the Republican National Convention in July. We calculate these benchmarks by taking the (real or imputed) polling average in each jurisdiction and adjusting each candidate’s level of support until they reach a delegate majority.

For jurisdictions that have enough polls for us to run our full poll-aggregation model, we start with the latest data from our polling averages page. For other jurisdictions, we use their demographic and political characteristics to predict (or “impute”) what the polling average would be. To do this, we used 2018-22 American Community Survey data and 2020 U.S. decennial census data (for American Samoa, Guam, the Northern Mariana Islands and the U.S. Virgin Islands) to estimate the percentages of people in each state and territory who are white, Black, Hispanic and Asian American and Pacific Islander as well as the percentage of people age 25 or older who are white and have attended any college. We also included as a variable Trump’s vote share in each 2016 primary contest as well as the percentage of each jurisdiction that is white evangelical. We use state-level religion data from the 2022 American Values Atlas published by the Public Religion Research Institute, except in Washington, D.C., where the latest available data is from their 2020 survey, and Puerto Rico, where we approximated the share based on a Pew Research analysis. We imputed any missing state- or territory-level religion, polling averages and Republican primary results using a procedure called multiple imputation with chained equations, which predicts those variables based on values in geographies with similar demographics.

We then input these variables into a K-nearest-neighbors algorithm to find the five jurisdictions that are the closest match to each of the 56 jurisdictions holding Republican primaries, caucuses or conventions in 2024. The predicted polling average in a jurisdiction is equal to the average of the polling averages (or, in jurisdictions without polling averages, individual polls, weighted by sample size and recency) in its five closest matches. In this average, polling averages are given three times as much weight as individual polls. If a jurisdiction has already held its nominating contest, we use the actual results, which receive a weight of five polls.

We then convert our estimates of each candidate’s support into estimated delegate totals according to each jurisdiction’s delegate rules. From there, we run a simulation model that estimates how many delegates each candidate would win assuming they gained or lost a given amount of support in each state. This model takes votes away from other candidates proportionally. From this, the model can calculate the implied national vote swing that results in each candidate winning at least a majority of delegates as well as how many delegates they would win from each state on their way to doing so.

We generated our first set of benchmarks before any states voted. As election results come in, we will update our benchmarks for jurisdictions that have not yet voted using the real popular-vote results from jurisdictions that already voted. The benchmarks shown for jurisdictions that have already voted, however, will be frozen in time with their last pre-contest delegate benchmarks.*

*A previous version of our model used the post-election delegate estimates from ABC News to update our benchmarks as the primary season progressed. However, this caused bugs in the way our 538 delegate estimates added up for certain candidates, so this feature was removed on Jan. 23, 2024.