Forecasting the 2024 Republican Presidential Nomination: Can the Former Heavyweight Champ Win Another Title Shot?

Abstract

Donald Trump’s bid for the 2024 Republican presidential nomination is unique in that no former president since Theodore Roosevelt in 1912 has sought the nomination of their political party, nor has a candidate sought the nomination while facing multiple criminal indictments. With data from previous nomination cycles, we use presidential nominations from 1980 to 2020 to create a forecast for the 2024 Republican primaries. The variables in the equations consist of data from the pre-primary period (e.g., money raised, cash reserves, elite endorsements, and polling results) and a second model with results of the Iowa caucuses and the New Hampshire primary to forecast the remaining primary vote. The models accurately predict Trump’s victory despite the unique nature of his candidacy.

Recent academic studies have modeled open presidential nomination outcomes with considerable success (Adkins and Dowdle 2000, 2001; Dowdle et al. 2016, 2021; Steger 2000, 2007), especially when there is a strong frontrunner during the invisible primary (Steger 2013; Steger, Dowdle, and Adkins 2012). In no small part, this occurs because few senators and governors—being strategic and opportunistic politicians—run when a nationally known candidate with strong support in national polls is expected to enter the race (Adkins et al. 2015; Steger 2016).

The 2024 Republican presidential nomination represents an interesting case for open presidential nomination forecasts. Donald Trump is the first former president since Theodore Roosevelt in 1912 to seek their party’s nomination. As a former elected president, Trump had most of the advantages of an incumbent president, including near-universal name recognition, high media coverage, a proven donor base, and having previously won the party’s nomination.

Donald Trump is the first former president since Theodore Roosevelt in 1912 to seek their party’s nomination.

Yet, Trump differs from most incumbents. He won the presidency in 2016 despite losing the popular vote by 3 million votes; in 2020, he lost by 7 million votes. Congressional candidates associating with Trump’s Make America Great Again (MAGA) faction of the Republican Party received an average of almost 10% fewer votes in the 2022 midterms than Republican congressional candidates who did not self-identify as a MAGA candidate. Trump’s presence as a candidate in the race did not deter a large field of governors and senators from declaring their candidacy. When candidates were deciding whether to run in the summer and fall of 2022, Trump had double-digit-net unfavorable ratings nationally, and he averaged less than 50% support in polls of potential candidates for the 2024 Republican nomination. Trump’s support decreased in the months after he preemptively declared his candidacy early to deter potential rivals from entering the race, averaging only 43% in November and December of 2022.

Also at that time, Trump faced the potential for multiple criminal indictments for alleged crimes from before, during, and after his presidency. That a former vice president, three governors and two former governors, a senator, and several other candidates entered the race suggests perceived vulnerability. Despite that, Trump’s support among Republicans increased throughout 2023, and most of his rivals exited the race by year’s end. As of the date of the Iowa caucuses, there remained only four other serious contenders for the Republican nomination. ¹

RESEARCH DESIGN

We used two ordinary least squares (OLS) models to generate candidate forecasts for the open presidential nomination contests from 1980 to 2020 and then applied the out-of-sample model results to the 2024 Republican race. The first model used pre-primary measures of support whereas the second model added the results from the Iowa and New Hampshire contests as indicators of early momentum. These factors provided an important advantage throughout the post–McGovern–Fraser era. Because presidential incumbents have not lost a renomination in the modern era, we excluded these nominations from the analysis. ²

The dependent variable was the percentage of the aggregate presidential primary vote that a candidate receives in each nomination cycle. In developing different models, we used several indicators to measure early support, including a candidate’s percentage share of support in pre-primary national Gallup polls, endorsements by party elites, fundraising success, and early campaign momentum (e.g., finishes in the Iowa caucuses and New Hampshire primary).

Poll Results

Poll results in the pre-primary season are commonly used to forecast presidential nominations (Adkins and Dowdle 2000; Mayer 1996; Steger 2007).

H1: The greater the percentage that an aspirant receives in pre-primary preference polls, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

The Poll Results variable was created from each candidate’s average of support among self-identified partisans and leaners in national Gallup polls conducted in the fourth quarter of the pre-primary season from 1980 to 2024. For example, we used the average rating that a candidate received in pre-primary polls taken during October, November, and December of 2019 to generate a score for each candidate for the 2020 Democratic nomination. Because Gallup stopped polling pre-primary “horse race” preferences in 2015, we used the average percentage of support that each candidate received in the CNN poll for the 2016 and 2020 contests and the Harris Poll for 2024. ³

Campaign Expenditures

Since 1980, the candidate who won the “money primary” often claimed the party’s nomination. The nominees were the aspirants who tended to raise the most money during the pre-primary period (Adkins and Dowdle 2002). To control the impact of money spent before versus after the primaries begin, we created two separate measures of fundraising success: campaign expenditures by December 31 and the remaining cash reserves as of that date.

H2: The greater the amount of money that a candidate spends in the pre-primary period relative to the rest of the field, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

The Campaign Expenditures variable measures the amount of money spent by each candidate during the pre-primary period as a percentage of the money spent by the entire candidate pool in that particular race. This allowed us to account for both inflation and the different levels of spending between various election cycles. ⁴

Cash Reserves

The second measure of fundraising success, Cash Reserves, represents an aspirant’s ability to raise funds without needing to spend it to be competitive. Adkins and Dowdle (2001) demonstrated that this measure better predicts campaign success than money raised throughout the pre-primary season or money spent before the Iowa caucuses.

H3: The greater the amount of unspent money at the end of the pre-primary period that a contender has relative to the rest of the field, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

The Cash Reserves variable is each candidate’s unspent funds at the end of the pre-primary season as a percentage of total cash reserves held by all candidates at the end of the year prior to the election (e.g., 2019 for the 2020 Democratic nomination). ⁵

Endorsements

Public endorsements by party-elite elected officials, taken in the aggregate, signal the breadth of support among party insiders who are sensitive to the preferences of party activists and groups (Whitby 2014). As Steger (2007) demonstrated, elite endorsements in the pre-primary period help to predict nomination outcomes.

H4: The greater the percentage of elite party endorsements that an individual receives relative to their opponents by the end of the pre-primary season, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

The Endorsements variable represents the unweighted percentage of governors, senators, and members of the House of Representatives who endorse a particular candidate relative to all elected officials in those positions within a party as of December 31 of the year prior to the nomination contest. ⁶

Iowa Caucuses

Previous research (Bartels 1988) revealed that the Iowa caucuses and New Hampshire primary results can impact a campaign’s overall success. Despite producing a relatively small share of the total convention delegates selected throughout the process, these early contests can generate momentum to boost campaigns. Several scholars cited performance in Iowa as a factor in overall success or failure in subsequent contests (Hull 2008; Redlawsk, Tolbert, and Donovan 2011). We used two variables to capture the effects of Iowa.

H5: The winner of the Iowa caucuses will receive a higher percentage of the vote in the presidential primaries.

The first indicator is a dichotomous variable, Iowa Win, that measures whether a candidate won the caucuses. The winner is scored “1” and other contenders are scored “0.” ⁷

H6: The greater a candidate’s percentage of the vote in the Iowa caucuses, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

The second variable, Iowa Percent, represents a candidate’s share of the vote of Iowa caucus-goers.

New Hampshire Primary

Previous studies (Adkins and Dowdle 2001; Mayer 1996; Steger 2000) concluded that the results of the New Hampshire primary impacted the nomination outcome, even when Iowa was accounted for in a prediction model. As with Iowa, we used two variables to capture the effects of the New Hampshire primary.

H7: The winner of the New Hampshire primary will receive a higher percentage of the vote in the presidential primaries.

The first is a dichotomous variable, New Hampshire Win, that measures whether a candidate won the primary. The winner is scored “1” and other contenders are scored “0.” The second variable, New Hampshire Percent, represents a candidate’s share of the New Hampshire primary vote.

H8: The greater the percentage of the vote that an individual wins in the New Hampshire primary, the higher the percentage of the vote that the candidate will receive in the presidential primaries.

Although other measures in recent contests—such as social-media following (Chen et al. 2012) and independent spending by political action committees (Gulati 2012)—may affect nomination outcomes, we excluded them for three reasons. First, incorporating them would sharply reduce the number of cases that we could include in the model because these factors have become important only in the past three or four election cycles. Second, there is some difficulty in aggregating these data in a timely enough manner to generate a forecast. Third, one of the advantages of incorporating our existing measures is that they mostly have been gathered by organizations with a long history of consistently recording and sharing data (e.g., the Federal Election Commission and Gallup). This is essential because a long-term forecasting model must be consistent over the years that it covers.

DATA ANALYSIS

The first of two OLS regression models (table 1) includes a series of measures from the pre-primary period that ends on December 31 of the year prior to the election to predict the eventual total aggregate primary vote. The variables included are Poll Results, Campaign Expenditures, Cash Reserves, and Endorsements.

Table 1 OLS Forecasting Models of Aggregate Primary Vote, 1980–2020

Note: The dependent variable is the percentage of the total aggregate primary vote—not including the results of the New Hampshire primary—that a candidate receives in each nomination cycle. Coefficients are unstandardized OLS coefficients; t scores are in parenthesesstandardized beta coefficients are in brackets [], and SEE=standard error estimate. Significant at *p<0.05, **p<0.01.

The pre-primary model accounts for variations in the dependent variable with an adjusted R² of 0.61. Two independent variables, Poll Results and Endorsements, were statistically significant at the 0.01 level and positively correlated with receiving higher shares of the total aggregate primary vote. These findings mirror our hypotheses as well as previous research. Campaign Expenditures were not significant, which is consistent with most prior studies. Although the negative relationship between spending in the pre-primary period and vote share seems counterintuitive, it is important to remember that most candidates spent money throughout the pre-primary period but failed to move up in the polls. Burning through cash early in a campaign signals difficulty in gaining traction. At the same time, Cash Reserves represent candidates who can raise money and be competitive in public opinion polling without having to spend those funds. A sizeable “war chest” also signals of viability to the press and party elites and gives a candidate flexibility during the primaries. The finding that Cash Reserves were not significant either is surprising because it is inconsistent with the findings in Dowdle et al. (2021). The bilityy of candidates to raise money quickly using technology may have ended the impact of having cash reserves.

The second OLS regression model (see table 1) includes the four independent variables in the first OLS model plus the two Iowa and New Hampshire variables: Iowa Win, Iowa Percent, New Hampshire Win, and New Hampshire Percent. The second model is better at accounting for variations in the dependent variable with an adjusted R² of 0.82. Both Poll Results and Endorsements were statistically significant at the 0.01 level and positively correlated with the independent variable. Pre-primary spending was negatively correlated with the dependent variable in the post–New Hampshire model, consistent with previous research (Dowdle et al. 2021). Cash Reserves were not significant either, although they were significant in previous studies (Dowdle et al. 2016, 2021). Three independent variables representing the early primary season (e.g., Iowa Win, New Hampshire Win, and New Hampshire Percent) were correlated with a boost in a candidate’s overall percentage of the total primary vote. By contrast, the percentage of the vote share that a candidate wins in Iowa (i.e., Iowa Percent) was not statistically significant. ⁸

It is instructive to review the success rate of both models in predicting the ordinal-level finishes for each race. The pre-primary model correctly predicted 13 of 16 (81%) of the winners of the aggregate primary vote, incorrectly identifying Jesse Jackson in 1988, Howard Dean in 2004, and Rudy Giuliani in 2008 as winners. The post–New Hampshire model correctly predicted 14 of 16 (88%) of the winners (table 2).

Table 2 Combined Model Predicted and Actual Finish, 1980–2024

Note: Underlined names indicate a correct ordinal forecast in terms of the percentage of the primary vote. In the 2008 Democratic nomination process, Hillary Clinton finished with the highest number of total aggregate primary votes, if the results of the Florida Democratic primary are included.

The second model correctly identified John Kerry in 2004 and John McCain in 2008, but it incorrectly chose both Paul Tsongas as the victor in 1992 and Bernie Sanders in 2020. It is reasonable to point to races with few entrants (e.g., the 2000 and 2016 Democratic contests) as “inflating” the accuracy rate, but the success rates of both models were well above 70%—even when eliminating those races with only a few contenders. Also noteworthy is that predictions for Democratic contests were still less accurate than for Republican contests (Dowdle et al. 2021), which may argue for splitting the two parties in future studies in which there are contested primaries in both parties.

Neither model did as well at predicting second and third place finishes in these contests. The pre-primary model correctly identified only six of 16 runners-up and four of 15 third-place finishers. ⁹ The post–New Hampshire model was somewhat more accurate and predicted 11 of 16 runners-up but only four of 15 third-place finishers.

DISCUSSION

The post–New Hampshire model in the 2020 Democratic and 2024 Republican nomination races was less accurate than the initial pre-primary model. In 2020, the pre-primary model predicted Biden would win but the post–New Hampshire model placed him as finishing third. Increasingly, these races appear to be “bumps in the road” to the nomination for the candidate leading at the end of the invisible primary (Adkins and Dowdle 2004). In 2024, both models generated predictions that Trump would win by a large margin. The pre-primary model predicted that Trump would garner 73% of the vote, the post–New Hampshire model predicted he would garner 89% of the primary vote, and his actual percentage of the primary vote was 77%.

An important question is whether presidential nominations could be forecast earlier because information for the pre-primary model is available around the end of January. Additional research is needed to replicate and extend Adkins and Dowdle’s (2005) study, which demonstrated that many nomination winners could be determined by the end of the third quarter of the year prior to the election. Their research emphasized the greater importance of early poll standings in forecasting winners. However, in 2024, Trump’s win seemed preordained to some pundits even though Ron DeSantis was within 10 points of Trump in the polls in the first quarter of 2023. Therefore, although early polling results are important indicators, they are not necessarily definitive in every race, especially within the Democratic Party (Steger 2013).

[A]lthough early polling results are important indicators, they are not necessarily definitive in every race, especially within the Democratic Party.

CONCLUSION

Donald Trump’s candidacy was much like a former heavyweight boxing champion trying to regain his title in a rematch. Despite multiple criminal indictments, his support among Republicans increased throughout 2023, causing most of his rivals to exit before the Iowa caucuses. As a former president, Trump enjoyed many advantages of incumbency, including strong name recognition, extraordinary media interest, an established donor base, and having previously won the Republican nomination—all of which contributed to his winning the title of Republican presidential nominee.

As a former president, Trump enjoyed many advantages of incumbency, including strong name recognition, extraordinary media interest, an established donor base, and having previously won the Republican nomination—all of which contributed to his winning the title of Republican presidential nominee.