A forecasting framework that measures how voters evaluate candidates’ future leadership abilities and policy expertise accurately anticipated the closely contested outcome of the 2024 United States presidential election. The method offers an alternative to conventional models that rely on past economic performance, providing campaign strategists with specific guidance on shaping public perception. The research was published in Research and Politics.
Election forecasting has a long history within political science, frequently relying on a concept known as retrospective voting. This theory assumes that citizens act as auditors of the incumbent party, punishing or rewarding candidates based on recent economic indicators and job approval ratings. These models calculate a candidate’s chances by looking backward at the outgoing administration’s track record.
However, retrospective forecasting becomes complicated during open-seat elections where the sitting president is not on the ballot. This was the exact scenario in the 2024 election following President Joe Biden’s withdrawal from the race. Without a direct performance record for the new candidate, voters tend to shift their focus away from judgments about the past.
Instead, many voters rely on prospective voting, a process of evaluating candidates based on anticipated future performance. Researchers have found that these forward-looking assessments become a primary driver of voter behavior in open-seat contests. Voters look ahead at what policies and leadership styles the new candidates might bring to the office.
Andreas Graefe, a researcher at the Macromedia University of Applied Sciences in Munich, Germany, developed a forecasting tool known as the Issues and Leaders model to capture these forward-looking dynamics. Graefe designed the framework to address common limitations in mainstream election forecasting. Popular contemporary models, such as standard poll aggregators, often update daily to provide real-time snapshots of a race.
While these aggregators help campaigns identify where to allocate financial resources, they offer little strategic advice regarding campaign messaging. They tell observers who is currently winning a standard preference poll, but they fail to explain which policy areas or personality traits are driving those voter preferences. Graefe sought to create a tool that tracks the specific considerations underlying voter choices.
The Issues and Leaders model focuses entirely on two variables: issue-handling competence and leadership perception. To calculate issue-handling scores, the model requires three conditions to be met. Voters must be aware of an issue, they must perceive it as important, and they must trust one candidate more than the other to manage it.
To determine the relative importance of various issues, the model utilizes responses from Gallup’s regular surveys asking Americans to name the country’s most pressing problem. These problems are categorized into economic, foreign policy, and other domestic concerns. The model then weighs how much importance the electorate assigns to each category.
Once the relative weight of the issues is established, the model analyzes survey data regarding which candidate voters trust to handle them. For the 2024 analysis, Graefe utilized polling data compiled by the election analysis website FiveThirtyEight. The dataset included 586 questions regarding issue competence drawn from 87 unique surveys conducted between October 2023 and November 2024.
The model calculates a daily average of voter trust for the incumbent party’s candidate across all these issues. It uses a mathematical technique called exponential smoothing, which gives more weight to recent polling while retaining some influence from older data. This prevents the model from overreacting to minor, short-term fluctuations in public opinion.
The second component of the model measures leadership perception. This metric relies on polls asking respondents a simple, direct question regarding which candidate they believe is a stronger leader. The 2024 tracking incorporated 22 unique surveys conducted between February and October that focused explicitly on this single dimension of leadership strength.
To turn these daily scores into an actual election forecast, Graefe relied on historical data spanning thirteen U.S. presidential elections from 1972 to 2020. By analyzing past voting patterns, a statistical model determines exactly how much weight to assign to partisanship, issue-handling, and leadership scores at different stages of a campaign. The model evaluates how these variables interact with the incumbent party’s final share of the two-party popular vote.
The historical analysis reveals a notable shift in voter behavior as Election Day approaches. One hundred days out, a large percentage of intention is tied simply to underlying partisan loyalty. By the final hours of a campaign, that baseline partisan influence drops by more than half. Instead, candidate-specific evaluations concerning policy expertise and leadership gain substantial predictive weight.
When applied to the 2024 race, the model tracked evolving voter perceptions in real time. Beginning exactly 100 days before the election, Vice President Kamala Harris maintained a very slight edge over former President Donald Trump regarding overall issue competence. By contrast, she started at a massive disadvantage regarding leadership perception, trailing Trump by 20 points in late July.
Over the subsequent months, Harris steadily narrowed the leadership perception gap. By Election Eve, she had reduced Trump’s advantage in this category to less than five points. Despite this late momentum, Trump maintained just enough of an edge in perceived leadership to offset Harris’s slight advantage on policy issues.
The final forecast generated by the model on Election Eve predicted a near tie, with Trump receiving 50.2 percent of the two-party popular vote and Harris receiving 49.8 percent. This cautious projection stood in contrast to many conventional polling averages, which generally showed Harris retaining a slight lead. Ultimately, Trump won the national popular vote by approximately 1.5 percentage points.
The model’s final forecast underestimated Trump’s eventual vote share by just half a percentage point. Across the entire 100-day tracking period, the model’s average error was only 0.65 percentage points. This level of accuracy is consistent with its performance in the 2012, 2016, and 2020 election cycles, demonstrating its reliability as an out-of-sample predictive tool.
Beyond providing an accurate forecast, Graefe notes that the model offers campaigns specific strategic directions. Recognizing that voter perceptions decide elections, candidates can actively attempt to improve their reputations regarding vital policies. They can also try to steer the media narrative toward topics where they already enjoy a reputational advantage.
Because perceptions of leadership are heavily influenced by relatively fixed traits like professional background and personal demeanor, parties could use these metrics to make more informed choices during primary elections. Identifying candidates who naturally exude the leadership qualities expected by the broader electorate might give a party an early, structural advantage.
A primary limitation of the current model is its restriction to the national popular vote. Due to a lack of detailed, state-level polling on specific issue-handling and leadership questions, the model cannot generate an Electoral College forecast. In the United States system, the popular vote does not dictate the winner of the presidency, making state-by-state predictions highly desirable.
Expanding this methodology to individual states is a promising direction for future research. A state-level approach could capture regional differences in what voters care about most. For example, immigration policy might hold more weight for voters in border states, while economic manufacturing issues might dominate in industrial regions.
Access to localized data would allow researchers to refine the model’s sensitivity to these geographic differences. Such an expansion would provide campaigns with localized intelligence for tailoring advertisements and stump speeches. Until such polling becomes widely available, the national model remains an effective tool for understanding the underlying expectations that shape voting behavior.
The study, “Prospective voting and the issues and leaders model: Forecasting the 2024 U.S. presidential election,” was authored by Andreas Graefe.