Frequently Asked Questions
Where can I learn more about U.S. Elections?
Wikipedia has plenty of information to get you started. Check out the following articles:
- About the Government
- U.S. Government (Overview)
- About the Elections
- U.S. Elections (Overview)
- Presidential Election
- Senate Elections
- House Elections
- Historical Elections
What is a forecast?
A forecast is an assignment of probabilities to potential future outcomes based on currently available information. This information can come from fundamentals, such as historical voting trends and economic factors, or from polling data. Forecasts can also incorporate anticipated future events or changes when assigning probabilities. As an example, a forecast could use trends in current polling data to extrapolate voter turnout on election day.
Forecasts on this site use the results from the previous election and current polling data (without extrapolations) to construct models for voter turnout in each of the races on election day. These models are used to compute the candidates' probabilities of winning the individual races. The numbers from all races are then combined to assign probabilities to potential outcomes at the national level (e.g., electoral votes received, Senate seats secured).
What is a snapshot?
A snapshot is a summary of information available at a specific moment in time.
The forecasts on this site can be viewed as snapshots of the election because each forecast aggregates the currently available polling data without extrapolating it when constructing the election day voter turnout models.
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, ties are possible. There are numerous combinations of states that can lead to a tie (269 Electoral College votes for both candidates), though only a small fraction of such combinations are likely to occur.
How are Independents handled?
Some senators and candidates for the Senate identify as independent instead of joining a particular political party. However, these individuals will typically participate in either the Senate Democratic Caucus or the Senate Republican Conference. 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.
For the 2014 elections, the two current independent senators who are not up for re-election (Bernie Sanders in Vermont and Angus King in Maine) both caucus with the Democrats and are expected to continue to do so in the next session of Congress. Only one independent candidate is currently considered competitive, Greg Orman in Kansas. Orman is running against the incumbent, Republican Pat Roberts (the Democratic candidate, Chad Taylor, withdrew from the race). While Orman has not declared the party with whom he will caucus, the forecasts on this site include him with the Democrats by default.
For the 2012 elections, there were two independent candidates running, Bernie Sanders in Vermont and Angus King in Maine. Sanders had previously caucused with the Democrats, and King was expected to do so. Because of this, the independent candidates could be included with the Democrats when computing the probability of one party achieving a majority.
How are upcoming primaries handled?
In states where one or both primaries have not finished, each possible general election match-up between candidates is considered. The probability of a particular match-up occurring is equal to the probability that each candidate in the match-up wins their respective primary race. A weighted sum of the candidates' probabilities of winning the general election is then computed across all match-ups. This is similar to how a basketball bracket may forecast the outcome of March Madness.
What are runoff races and how are they handled?
Some states use a two-round system for voting. If no candidate in a race receives more than 50% of the vote, then the top two candidates advance to a runoff race while all other candidates are removed from the ballot.
In the forecasting system used here, the forecasts from a runoff race are incorporated in the following manner. First, the probability of a runoff occurring is computed as 1.0 minus the probability that any candidate wins 50% or more of the vote in the primary race. If the runoff probability is non-zero and the state has runoff races, then each candidate's probability of advancing to the general election is computed as the probability of the candidate winning 50% or more of the vote in the primary plus the probability of the candidate winning the runoff weighted by the probability of the runoff occurring.
This same approach is applied to open primaries.
Why are no House forecasts available?
Given 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. Because of this, House forecasts have been omitted.
What are custom forecasts?
For 2016, a new feature has been added to the site that allows you to customize the official forecasts in various ways. These customizations allow you to control the influence of undecided voters through turnout scenarios, exclude specific polling organizations from consideration, adjust the influence of polls on outcomes, and see the impact of third-party candidates on the presidential election.
What is a turnout scenario?
A turnout scenario is an 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. Several different turnout scenarios are currently available:
- Neutral Scenario: The prior distribution of the Bayesian estimators for undecided voters are split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
- Very Strong Democrat Turnout Scenario: 20% of undecided voters are assumed to vote for the Democratic candidate, with the remaining 80% of the undecided voters split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
- Strong Democrat Turnout Scenario: 10% of undecided voters are assumed to vote for the Democratic 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.
- Mild Democrat Turnout Scenario: 5% of undecided voters are assumed to vote for the Democratic candidate, with the remaining 95% of the undecided voters split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
- Mild Republican Turnout Scenario: 5% of undecided voters are assumed to vote for the Republican candidate, with the remaining 95% of the undecided voters split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
- Strong Republican Turnout Scenario: 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.
- Very Strong Republican Turnout Scenario: 20% of undecided voters are assumed to vote for the Republican candidate, with the remaining 80% of the undecided voters split 48%-48%-4% between the Democratic candidate, the Republican candidate, and the independent candidates, in all states.
Each of these scenarios is considered individually. The Neutral scenario provides an unbiased handling of undecided voters. The Very Strong turnout scenarios provide two extreme envelopes around which the results obtained can be judged and evaluated. The Strong and Mild turnout scenarios provide realistic possibilities if late breaking information surfaces that shifts voter preferences.
What does the Influence of Polls option do?
This option determines how much the polls contribute when determining candidates' chances of winning each state or senate seat. The default setting of High gives polls a strong influence compared to the priors, while Medium and Low scale the poll sizes down by 0.1 and 0.01, respectively.
Why do some polling companies show up in the filter options but not others?
The polling companies that can currently be excluded in custom forecasts are those that had released at least 10 state polls by the end of July 2016. Rasmussen is included for custom forecasts of past elections (an example of which can be found here).
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?" However, 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 (two-candidate or multi-candidate) 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 does it take so long to load custom forecasts?
When generating forecasts that use a non-neutral turnout scenario or that exclude some polling organizations, it is necessary to recompute the candidates' win probabilities for each state or Senate seat. These computations are done on our web server, and can take up to several seconds to perform depending on how close the individual races are.
Why do some probabilities not add up to 1.000?
Not all probabilities will add up to 1.000 because of slight round-off error in the computations and the need to truncate the numbers for display purposes.
How can a presidential candidate have a non-zero probability of winning 270 or more electoral votes without having any histogram bars past 269?
This apparent discrepancy is due to the need to round probabilities for display purposes. For example, a candidate might have a probability of 0.0004 for securing exactly 270 electoral votes and a probability of 0.0003 for securing exactly 271 electoral votes. Individually, these two probabilities are below the display cutoff for the histogram; however, when added together, the candidate has a probability of (at least) 0.0007 to secure 270 or more electoral votes, which would get rounded to 0.001 for display purposes.
Why are there forecasts that date back to the end of the previous election cycle?
We use the results from the previous election as an "initial" forecast for the current election. For example, the 2016 presidential election has a forecast from January 1st, 2013 that is based solely on the results of the 2012 presidential election.
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.
How can the forecasts on different web sites be different?
Each web site uses their own set of algorithms and methodologies. In a close election, two different approaches may lead to different forecast outcomes. Since our methodology only uses state polling data and ignores national popular vote polling data, it may seem that a presidential candidate is gaining in overall popularity while the candidate's probability of winning the election does not change significantly or moves in a direction contrary to what one would expect. For our methodology, forecasts are only affected when there are shifts in state polling data.
Where can I find forecasts for previous elections?
Other Forecast Sites