Presidential election forecasting: The Bush-Gore draw

Our Proxy Model of presidential election forecasting declared, from data issued six months before the November contest, that Obama would garner 52.7% of the two-party popular vote. [See the model release in ourMonkey Cageblog-post (Lewis-Beck and Tien 2012a) on September 18, 2012, and in the October issue ofPS(Lewis-Beck and Tien 2012b).] Thus, that forecast called the correct winner, with a point estimate error of only 0.9 percentage points.

Download Full-text

Election Forecasting: The Future of the Presidency and the House

Political Science and Politics ◽

10.1017/s1049096508081225 ◽

2008 ◽

Vol 41 (04) ◽

pp. 713-716 ◽

Cited By ~ 13

Author(s):

Brad Lockerbie

Keyword(s):

Political Science ◽

Annual Meeting ◽

Presidential Election ◽

Popular Vote ◽

Election Forecasting ◽

Vote Outcome ◽

The Past ◽

Actual Outcome ◽

Variable Equation ◽

Science Association

This article is about a simple two-variable equation forecasting presidential election outcomes and a three-variable equation forecasting seat change in House elections. Over the past two decades a cottage industry of political forecasting has developed (Lewis-Beck and Rice 1992; Campbell and Garand 2000). At the 1994 meeting of the Southern Political Science Association, several participants offered their forecasts of the upcoming midterm House elections. Unfortunately, not one of the forecasters was within 20 seats of the actual outcome. If, however, these forecasts had been pooled, as Gaddie (1997) points out, then they would have come remarkably close to the actual seat change that occurred. Moving forward, at the 1996 APSA Annual Meeting the collection of forecasters did a much better job with that year's presidential election. The forecasters also got the overall popular vote outcome correct at the 2000 APSA Annual Meeting for that year's presidential election. We all forecasted a victory for Al Gore, with James Campbell coming the closest to the actual total (50.2%) at 52.8%. At the panel at the 2004 APSA Annual Meeting almost every forecaster predicted the actual outcome correctly. Forecasting elections holds us accountable—we cannot go back and change our forecast for an election after it has occurred. Moreover, if we stick with one forecast, it easy to judge the overall accuracy of our equations.

Download Full-text

What, If Anything, Have We Learned from Presidential Election Forecasting?

Political Science and Politics ◽

10.1017/s1049096514000134 ◽

2014 ◽

Vol 47 (02) ◽

pp. 329-331 ◽

Cited By ~ 1

Author(s):

William G. Mayer

Keyword(s):

Presidential Election ◽

Election Forecasting

Download Full-text

Forecasting the 2012 French Presidential Election

Political Science and Politics ◽

10.1017/s1049096512000066 ◽

2012 ◽

Vol 45 (02) ◽

pp. 218-222 ◽

Cited By ~ 5

Author(s):

Martial Foucault ◽

Richard Nadeau

Keyword(s):

Political Economy ◽

Presidential Election ◽

Right Wing ◽

National Data ◽

Election Forecasting ◽

Vote Function ◽

The Right ◽

Nicolas Sarkozy

AbstractWho will win the next French presidential election? Forecasting electoral results from political-economy models is a recent tradition in France. In this article, we pursue this effort by estimating a vote function based on both local and national data for the elections held between 1981 and 2007. This approach allows us to circumvent the smallNproblem and to produce more robust and reliable results. Based on a model including economic (unemployment) and political (approval and previous results) variables, we predict the defeat, although by a relatively small margin, of the right-wing incumbent Nicolas Sarkozy in the second round of the French presidential election to be held in May 2012.

Download Full-text

The Job of President and the Jobs Model Forecast: Obama for '08?

Political Science and Politics ◽

10.1017/s1049096508081262 ◽

2008 ◽

Vol 41 (04) ◽

pp. 687-690 ◽

Cited By ~ 25

Author(s):

Michael S. Lewis-Beck ◽

Charles Tien

Keyword(s):

Social Sciences ◽

Data Mining ◽

Presidential Election ◽

Model Forecast ◽

Special Issue ◽

Election Forecasting ◽

The Social ◽

Simple Product ◽

Forecasting Errors

The statistical modelers are back. The presidential election forecasting errors of 2000 did not repeat themselves in 2004. On the contrary, the forecasts, from at least seven different teams, were generally quite accurate (Campbell 2004; Lewis-Beck 2005). Encouragingly, their prowess is receiving attention from forecasters outside the social sciences, in fields such as engineering and commerce. Noteworthy here is the recent special issue on U.S. presidential election forecasting published in theInternational Journal of Forecasting, containing some 10 different papers (Campbell and Lewis-Beck 2008). Our contribution in that special issue explored the question of whether our Jobs Model, off by only 1 percentage point in its 2004 forecast, was a simple product of data-mining (Lewis-Beck and Tien 2008).

Download Full-text

A Machine Learning Approach for Forecasting US Presidential Election

Advances in Social Sciences Research Journal ◽

10.14738/assrj.710.9210 ◽

2020 ◽

Vol 7 (10) ◽

pp. 186-198

Author(s):

Noah Loewy ◽

Ashok Singh ◽

Tina Marie Gallagher

Keyword(s):

Machine Learning ◽

Presidential Election ◽

Strong Predictor ◽

Electoral College ◽

Gradient Boosting ◽

Attractive Alternative ◽

Election Forecasting ◽

Extreme Gradient Boosting ◽

Machine Learning Approach ◽

Continuous Predictors

In this paper, we develop and compare two models for forecasting the 2020 U.S. presidential election using multiple linear regressions (MLR) and the Machine Learning method of Extreme Gradient Boosting (xgboost). We predict each state’s Republican vote share using seven continuous predictors from 1976-2016, as well as dummy columns for each state. After computing 95% confidence intervals for each prediction, we determine the candidates’ electoral college probabilities. The xgboost appears to be a very strong predictor, accounting for 98.6% of the variance with a 3.34% root mean square error (RMSE), whereas the MLR only accounts for 71.8% of the variance and leaves an RMSE of 6.35%. We observe that 1) both models predict a Democratic electoral college landslide in the 2020 elections, 2) Georgia, Iowa, Florida, North Carolina, and Ohio are crucial for the Republicans to win, and 3) Extreme Gradient Boosting is an attractive alternative to MLR in election forecasting.

Download Full-text