What is Election Analytics?
- Election Analytics tracks and analyzes polling data to forecast who will win the United States Presidency and which party will secure control of the United States Senate.
- Our student run laboratory uses Bayesian statistics and operations research to make sense of the polling data reported in the national media.
- Without any political commentary or partisan opinion, we provide a snapshot of the election—forecasting the outcome if the election was held today.
- Available since 2008, Election Analytics provides a full history of our performance.
- We have been mentioned in the Wall Street Journal and other publications.
- Reach us at election.analytics.cs@gmail.com
How does our forecast work?
- The Model
- Our mathematical model employs Bayesian estimators that use available state poll results to determine the probability that each presidential candidate will win each of the states (or the probability that each political party will win the Senate race in each state).
- These state-by-state probabilities are then used in a dynamic programming algorithm to determine a probability distribution for the number of Electoral College votes each candidate will receive (or Senate seats that each party will secure).
- Weighting Polls
- Polling data for each state is weighted based on how recently the poll was conducted.
- If three or more polls are available within the past two weeks, then polls within the past week have a weighting factor of 1, polls between one and two weeks old have a weighting factor of 0.5, and polls older than two weeks have a weighting factor of 0.
- If two or fewer polls are available within the most recent two weeks, then the three most recent polls are used, with polls within the past week have a weighting factor of 1 and polls older than one week have a weighting factor of 0.5.
- If no polls comparing the two candidates are available for a state, the results of the last election are used to estimate the outcome in the upcoming election.
- Other Considerations
- Polls often show figures that are for likely voters, as opposed to registered voters. If a greater number of registered voters show up to vote on election day, then the poll numbers may not be representative of the actual voters.
- The results presented are a direct function of the quality of the state polling data being used. Any biases in this data can lead to invalid conclusions.
- We do not forecast House races. Because of the sparsity of polling data for House races, any forecast for control of the House would be heavily dependent on the initial assumptions that were made.
How was our model developed?
- The ideas and methods used in this site all stem from this paper written by Steven E. Rigdon, Sheldon H. Jacobson, Wendy K. Tam Cho, Edward C. Sewell, and Christopher J. Rigdon.
- This paper provides an analysis of the prediction results from the 2008 Presidential Election.
- The results from the 2000 and 2004 United States Presidential Election suggested that it can be difficult to forecast the winner of the presidential election based on popular vote.
- To address this, Rigdon et. al created a new forecast model based on the electoral college vote to determine the winner of the next presidential election.
- This model was used to track and analyze the 2008 and 2012 presidential elections.
- In 2012, the model was extended to handle Senate races.
Questions about our forecasts
- How are Maine and Nebraska handled?
- Maine and Nebraska split their electoral college votes (4 and 5, respectively) based on their congressional districts. Prior to 2016, these states were treated like every other state (i.e., all or nothing). Starting in 2016, these states are subdivided in cases where congressional district polling data is available.
- Is a tie possible?
- Yes. There are numerous combinations of states that can lead to a tie (269 Electoral College votes for both candidates).
- How are independent senators handled?
- For the purposes of determining the number of seats held by the two major parties, each independent senator or candidate can be included in the party with which he or she is most likely to caucus. These values are reported in columns labeled Dems & Inds and Reps & Inds.
- What does the undecided swing customization do?
- It controls our model's assumption about how undecided voters will behave on election day. Undecided voters can have a significant role on the outcome of elections, and they are likely to be the ultimate deciders of who will win this presidential election.
- For instance, if the undecided swing is set to +10% Republican, 10% of undecided voters are assumed to vote for the Republican candidate, with the remaining 90% of the undecided voters split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
- What does the weight of polls customization do?
- It controls the influence of polls on our model. The default setting of 100% gives polls a strong influence compared to the priors, while 10% and 1% scale the poll sizes down by 0.1 and 0.01, respectively.
- How are third-party candidates incorporated in the presidential race?
- By default, the presidential forecasts on this site rely on polling data that focus on a two-candidate race, where survey questions are often of the form "If the election were held today, would you vote for the Democrat, the Republican, other, or undecided?"
- Some polling companies also provide polling data that focus on a multi-candidate race. For example, in the 2016 presidential election, a survey question might be "If the election were held today, would you vote for Hillary Clinton (D), Donald Trump (R), Gary Johnson (L), Jill Stein (G), other, or undecided?"
- The candidate combination option for custom forecasts for the 2016 presidential election allows users to determine which set of polling data should be used. If the options "with Johnson" or "with Johnson and Stein" are selected, then polling data which include these candidates will be used to construct forecasts when such data is available.
- Why do some probabilities not add up to 1.000?
- Sometimes there is a slight round-off error when decimals are truncated for display purposes.
- Why do the 2008 and 2012 forecasts on this site differ from the previous years' sites?
- Between 2012 and 2014, the code for computing each candidate's probability of winning a particular race was rewritten and the mechanism for collecting and weighting polling data was modified. These two changes have led to some minor differences between forecasts presented on this site and the official 2008 and 2012 forecasts.
Donate
If you would like to support the ongoing student development activities for this project, tax-deductible contributions can be made to the Department of Computer Science at the University of Illinois. Tax-deductible contributions can be made to the Department of Computer Science at the University of Illinois. Under "Computer Science Annual Fund", enter the gift amount. On the next page, in the "Additional Instructions" box, please type "Election Analytics Project". Thank you for your support
Our Team
- The following students and professors have made this project possible. All students are from the University of Illinois unless otherwise noted.
- Senior Project Advisor
- Industrial Engineering Graduate Student
- Senior Project Advisor
- Industrial Engineering Graduate Student
- Web Developer
- Computer Science Major
- Web Developer
- Computer Science Major
- Web Developer
- Computer Science Major
- Senior Project Advisor
- Industrial Engineering Graduate Student
- Web Developer
- Statistics & Computer Science and Political Science Major
- Web Developer
- Computer Engineering Major
- Senior Project Advisor
- Industrial Engineering Graduate Student
- Web Developer
- Computer Science Major
- Web Developer
- Computer Science Major
- Web Developer
- Computer Science Major
- Mobile App Developer
- Computer Science Major
- Senior Project Advisor
- Computer Science Graduate Student
- Undergraduate Team Lead
- Computer Science Major
- Web Developer
- Mathematics & Computer Science Major
- Senior Project Advisor
- Computer Science Graduate Student
- Public Relations
- Industrial Engineering Major
- Web Master
- Computer Science Major
- Web Master
- Computer Science Major
- Technical Analyst
- Computer Science Major
- Technical Analyst
- Mathematics & Computer Science Major
- Programmer
- Computer Science Major, Swarthmore College
- Web Master
- Computer Science Major
- Research Assistant
- Political Science Major
- Technical Analyst
- Computer Science Major
- Media Research
- Computer Science Major
Professor Sheldon H. Jacobson, Ph.D.
Department of Computer Science
University of Illinois
Urbana, Illinois
(217) 244-7275
shj@illinois.edu
@shjAnalytics
Professor Steven E. Rigdon, Ph.D.
Department of Biostatistics
St. Louis University
St. Louis, Missouri
(314) 977-8127
srigdon@slu.edu
Assistant Professor Jason Sauppe, Ph.D.
Department of Computer Science
University of Wisconsin–La Crosse
La Crosse, Wisconsin
(608) 785-6807
jsauppe@uwlax.edu