Content uploaded by Thomas M. Holbrook
Author content
All content in this area was uploaded by Thomas M. Holbrook on Apr 08, 2019
Content may be subject to copyright.
Forecasting Us Presidential Elections
Abstract
This article offers an assessment of the current state of US presidential election forecasting models. It pays attention to presidential forecasting models from the last three election cycles. It starts by exploring 'under the hood' and describes the specifics of the most widely known models from the 2004 election. In addition, the predictions made by these models are addressed and the determinants of forecasting accuracy from 1996 to 2004 are identified. Moreover, the article explores the lessons learned from the 2000 campaign and the alternatives to the dominant aggregate-national forecasting models: electronic markets, citizen forecasts, and state-level forecasting models. From a forecasting perspective, the 2008 election outcome was business as usual. Some models were more accurate than others, as is always the case, but the average error was somewhat lower than in the past two elections cycles.
... As noted above, past accuracy is often a poor indicator for a forecasting model's future accuracy; and forecasting U.S. presidential elections is no exception. Holbrook (2010) reports that the accuracy of nine established econometric models varied considerably across the three elections from 1996 to 2004. Models that were among the most accurate in one election were often among the least accurate in another. ...
... In 6 of the 9 elections, the correlation was negative; the median correlation across the nine elections was -.20. These results conform to Holbrook (2010): models that were among the most accurate ones in the previous election tended to be among the least accurate in the succeeding election. For a more conservative estimate and to protect against outliers, we also calculated the correlation of the models' absolute error and their mean absolute errors across the previous five elections. ...
The present study shows that the predictive performance of Ensemble Bayesian Model Averaging (EBMA) strongly depends on the conditions of the forecasting problem. EBMA is of limited value when uncertainty is high, a situation that is common for social science problems. In such situations, one should avoid methods that bear the risk of overfitting. Instead, one should acknowledge the uncertainty in the environment and use conservative methods that are robust when predicting new data. For combining forecasts, consider calculating simple (unweighted) averages of the component forecasts. A vast prior literature finds that simple averages yield forecasts that are often at least as accurate as those from more complex combining methods. A reanalysis and extension of a prior study on US presidential election forecasting, which had the purpose to demonstrate the usefulness of EBMA, shows that the simple average reduced the error of the combined EBMA forecasts by 25%. Simple averages produce accurate forecasts, are easy to describe, easy to understand, and easy to use. Researchers who develop new methods for combining forecasts need to compare the accuracy of their method to this widely established benchmark method. Forecasting practitioners should favor simple averages over more complex methods unless there is strong evidence in support of differential weights.
... While many studies have focused on perceptions of economic conditions, we should note that scholars have also examined whether people accurately assess other conditions related to political life, including perceived levels of political competition and projections of election outcomes. For instance, some scholars (Guinjoan et al. 2014;McDonald and Tolbert 2012) have found that perceived levels of political competition correspond to actual levels of competition in congressional elections, and others have found that, in aggregate, survey respondents do a pretty good job predicting actual election outcomes (Holbrook 2010;Lewis-Beck and Skalaban 1989). ...
Using two unique surveys, one that includes over 6,000 respondents interviewed across 39 cities and another that includes over 47,000 respondents interviewed across 26 U.S. cities, we investigate the extent to which perceptions of local conditions—the state of the local economy, the quality of local schools, and local crime—reflect actual local conditions. We examine individual-level differences in the accuracy of perceptions of local conditions using two different frameworks, one that emphasizes factors that limit information acquisition and may exacerbate political inequalities, and another that emphasizes motivations for information processing. Objective conditions influence perceptions of conditions, but the relationship between objective and perceived local conditions is strongest among individuals with high levels of education and preexisting knowledge. In addition, we find that partisanship plays a role in shaping perceptions of local conditions. While the partisan match between a respondent and the mayor of their city has little effect on local perceptions, the match between a respondent’s partisanship and the president’s party has a strong effect on perceptions of the local economy.
... "The Primary Model" [23] is a method that relies on presidential primaries as a predictor of the vote in the general election; it also makes use of a swing of the electoral pendulum that is useful for forecasting. As in the case of the "The Keys to the White House", "The Primary Model" also predicted the victory of Donald Trump [24]. However, "The Primary Model" predicted a Trump win over the popular vote, which turned out to not be accurate. ...
This study examined the accuracy of the various forecasting methods of the 2016 US Presidential Elections. The findings revealed a high accuracy in predicting the popular vote. However, this is most suitable in an electoral system which is not divided into constituencies. Instead, due to the Electoral College method used in the US elections, forecasting should focus on predicting the winner in every state separately. Nevertheless, miss-predicted results in only a few states led to false forecasting of the elected president in 2016. The current methods proved less accurate in predicting the vote in states that are less urbanized and with less diverse society regarding race, ethnicity, and religion. The most challenging was predicting the vote of people who are White, Protestant Christians, and highly religious. In order to improve pre-election polls, this study suggests a few changes to the current methods, mainly to adopt the “Cleavage Sampling” method that can better predict the expected turnout of specific social groups, thus leading to higher accuracy of pre-election polling.
... One study compared responses to the ANES vote expectation question to forecasts from the IEM vote-share markets during the past two weeks prior to each of the five U.S. presidential elections from 1988 to 2004. The relationship between the respective forecasts and the final vote was slightly higher for the vote expectations compared to the IEM forecasts (Holbrook 2010). ...
Partisan preferences usually stand out as the major driving force behind voters' expectations about election outcomes. Apart from partisan preferences however, purely individual‐level factors appear to be only weakly associated with forecasting ability. Some studies argue that we need to move from the strictly personal sphere to the interpersonal one to better understand the underpinnings of individuals' forecasting ability. This paper leverages data from 77 elections at the district, municipal, regional, and/or national levels in ten different countries to assess the impact of social networks and social interactions on the accuracy of citizens' electoral expectations. The results cast doubt on the capacity of social interactions to influence citizens' forecasting skills. This article is protected by copyright. All rights reserved
Las encuestas de intención de voto son posiblemente el método más difundido para generar estimaciones del resultado de las elecciones. La precisión de un método está fuertemente influenciada por las idiosincrasias de una elección en particular, varía entre elecciones e incluso dentro de una misma campaña. En este artículo comparamos los principales métodos de estimación para las últimas cuatro elecciones generales en España (2015, 2016, abril y noviembre de 2019). Sobre esta evidencia empírica, discutimos la relación entre las características de una campaña electoral y la pertinencia de los indicadores más utilizados (intención de voto, simpatía, probabilidad de ir a votar) y su ponderación por el recuerdo de voto.
These politics articles were commissioned by an editorial board as part of our former online-only review article series. We are offering them here as a freely available collection.
We add to the literature on citizen forecasting by examining the 2013 German federal election and, for the first time, trying to predict vote shares. A random sample of voters was asked to predict the vote shares for each party in telephone interviews. We examine the accuracy of individuals' expectations and analyze the influence of wishful thinking and published vote intention polls on voters' expectations. Individual forecasts are aggregated, assuming a wise crowd will make a precise forecast. Expectations do not yield a forecast of satisfactory accuracy and they are inferior to vote intention polls. High-ability subgroups do not yield better forecasts than the whole sample. Issues of sampling and measurement are addressed. We conclude that asking voters to predict eight interdependent vote shares is probably too difficult a task.
This article investigates the impact of interest group membership on presidential approval. Data come from the 2006 Cooperative Congressional Election Study, which asked respondents about membership in 11 interest groups. Distinguishing between easy and hard issues, I argue that interest groups will tighten the relationship between issue position and approval for hard issues because of the information that groups provide it members. Analysis looks at interest group effects on five issues: abortion and the Iraq War, both deemed easy; the environment and Social Security Reform, which are harder issues; and stem cell research, which is harder for some. As hypothesized, membership in abortion or veterans groups has no impact on members' approval either directly or through the relevant issues. But for members of environmental and senior groups, the relevant issues are found to have statistically significant impact on approval. And membership in abortion rights groups affects stem cell research. The conclusions put the findings into perspective and discuss directions for future research.
This research updates, revises, and extends a forecasting equation of the presidential vote in the states. The original equation was composed of sixteen predictors available well before the election and estimated with data from 531 state elections from 1948 to 1988. The equation was empirically strong, based on objective predictors, and more parsimonious than previous equations. Reexamining the equation with 200 additional state elections from 1992, 1996, 2000, and 2004 indicates that the equation remains well supported, but suggests several opportunities for improvement. A revised equation has a mean absolute error of 3.2 percentage points and correctly predicts 87 percent of all electoral votes. The extension of the analysis adapts the forecast equation to predict electoral vote winners, conducting a logit analysis that takes into account both the size of the state and the closeness of its previous election. This produces more accurate forecasts of both electoral vote winners in the states and the division of the aggregate national electoral vote.
The forecasting model presented here is a revised version of a model developed prior to the 1996 election (Holbrook 1996) and is essentially a referendum model. The original model regressed the incumbent party percent of the two-party vote on presidential popularity, an aggregate measure of satisfaction with personal finances, and a dummy variable coded “1” for years in which the incumbent party had held the White House for at least two consecutive terms and “0” for all other years. The first two variables are intended to capture the political and economic performance of the incumbent administration, while the latter variable (borrowed from Abramowitz 1988) is based on the idea that it may be easier to convince voters that it is “time for a change” if the incumbent party has held the White House for at least two consecutive terms. This model provided a fairly accurate forecast of the 1996 election and also had close out-of-sample post-dictions of elections from 1952–1992. However, the 2000 election represented a significant bump in the road, and the model over-predicted Gore's percent of the vote by approximately 10 percentage points.
As the experience of 2000 shows, forecasting presidential elections is an inexact science. Everybody knows that “the economy” matters, but simple projection from economic conditions at the time of the forecast is not enough. And the most important economic shocks to the economy are the late shocks, which may arrive too late to be measured by the forecaster. Other events also impact, such as (in 2004) the Iraq war. Incorporating presidential approval into the model helps to control for “other” events that economic indicators ignore, but obviously only those that are observable by the time of the latest approval reading. They also don't reveal much about voters' comparative judgments of the two candidates. Of course one can forecast the presidential race from trial-heat polls available at the moment rather than trying to capture the fundamentals that will matter on Election Day. But the whole purpose of forecasting is to present information about the voters' future behavior that is not yet evident in the trialheat polls. Besides, early polls only tell us little about the final election outcome.
Forecasting presidential elections has become a growth industry in political science. Econometric models used to predict the presidential election that were once viewed as part of “recreational political science” are now being taken seriously. The widespread belief that these models have been highly successful at predicting election outcomes has enticed a score of political scientists to propose new models hoping to share in the triumph of political science over the pundits and polls. Unfortunately, there has been very little critical examination of the models used to forecast presidential elections. A close review reveals that existing quantitative models are not useful predictors of presidential races. In addition, most of the new proposed models have adopted an approach that is unlikely to lead to better forecasts.
There is probably no subject which has been studied more thoroughly by political scientists than American presidential elections. A vast literature has developed examining the effects of attitudes toward the parties, candidates, and issues on voter decision-making in these quadrennial contests (for a comprehensive review of this literature see Asher, 1988). Despite the proliferation of literature on this topic, however, relatively little research has addressed what is perhaps the most basic question about presidential elections: who wins and who loses?
A few scholars have developed models for predicting the national outcomes of presidential elections. Brody and Sigelman (1983) proposed a model based on the incumbent president's approval rating in the final Gallup Poll before the election. This extremely simple model yielded an unadjusted R ² of .71. Hibbs (1982) proposed a different bivariate model based entirely on the trend in real per capita disposable income since the last presidential election. This model yielded an unadjusted R ² of .63. Thus, neither of these bivariate models proved to be highly accurate.
Lewis-Beck and Rice (1984) have developed a forecasting model which combines economic conditions and presidential popularity, Their model, which uses the president's approval rating in May and the change in real per capita GNP during the second quarter of the election year to predict the popular vote for president, yields an unadjusted R ² of .82.
This research examines an equation developed to forecast the two-party presidential vote in the states. The equation is composed of 16 independent variables measured at the national, regional, and state levels. It includes national trial-heat polls from early in the campaign, the growth in the national economy, presidential incumbency, the state's voting record in the previous two presidential elections, home state and regional advantages for the candidates, the partisan division of the state legislature following the last midterm election, an index of state ideology based on the roll call voting of its House delegation, and five regional trend variables to take into account partisan shifts over time. The equation is estimated over 531 state elections from 1948 to 1988. The overall fit of the equation is quite good. It accounts for nearly 85% of the variance, leaves a standard error of less than four percentage points, and has an average error of just three percentage points of the vote. This level of accuracy is comparable to that of Steven Rosenstone's equation in his Forecasting Presidential Elections (1983). Moreover, the equation is quite parsimonious and yields forecasts by early September of the election year.
Yogi Berra might have said it: the best predictor of an election is, well, an election. Not a trial-heat conducted by opinion polls, but a real election of voters going to the polls. In the U.S., at least, what is known as a “general” election is preceded by a “primary” election, and that has been the case for presidential contests since 1912. So is the voting in presidential primaries a leading indicator of the vote in November? Remarkably so, as it turns out. How well presidential candidates do in primary elections foretells their prospects in the November election with great accuracy. What is more, the use of primaries as a vote predictor makes it possible to include in the forecast model both the candidate of the incumbent party and the candidate of the party out of the White House. The forecast for 2004 uses candidate vote shares in primaries, not just a win-lose dichotomy as done in the model used to predict the vote in 2000 (Norpoth 2001).
This symposium presents 10 articles forecasting the 2008 U.S. national elections. The core of this collection is the seven presidential-vote forecasting models that were presented in this space before the 2004 election. Added to that group are one additional presidential forecasting model, one state-level elections forecasting model, and one model forecasting the relationship between congressional votes and seats won by the parties. Some of the articles that are focused on the presidential race have also taken the opportunity to forecast the congressional elections as well.