Democratic ticket has an 84% chance of winning the election

Electoral vote predictions:
305 Biden-Harris
233 Trump-Pence

Reload for current forecast from The Virtual Tout® (updated hourly)


Forecasts in time audio:

Saturday, October 31, 2020, 02:20:04 AM Pacific


Better and faster forecasts without polls

Political scientists often complain about opinion polls, as do politicians and analysts. Polls are costly and inefficient. Polls are out-of-date as soon as they are published. And, given the great variability from one poll to the next, we must wonder why election forecasters rely on opinion polls as their primary data source.

Prediction markets provide a meaningful alternative to political opinion polls. We say prediction markets are better than polls because prediction markets can respond to all factors relevant to an election. While opinion polls try to anticipate the behavior of likely voters on election day, prediction markets consider voting practices and contingencies as well as voter intentions. Prediction markets can also reflect voter demographics and economic conditions.

We say prediction markets are better than polls because prediction markets are fast to respond to current events. Prediction markets provide relevant, up-to-date information, responding to all events in the public sphere. If a candidate gives an especially good speech or convention events play well in the media, prediction market prices respond. If a candidate misspeaks or says something that reflects poorly on his or her character, that, too, can move prediction market prices. Election day weather forecasts can affect prediction markets, as can election regulations, election meddling, manipulation of voter registration rolls, vote-by-mail restrictions, and reductions in voting locations or voting hours.

Prediction markets reflect collective information across all people willing to place bets on the outcomes of political contests. Anything known to the public can affect prediction market prices. Prediction markets reflect the "wisdom of the crowd."

Forecasts in time show the effects of campaign events and news. The time series leading up to and following the first presidential debate and the time series immediately after Trump's COVID-19 tweet suggest that the probability of a Democratic victory in the Electoral College increases dramatically. Download figures showing hourly election forecasts in a recent two-week period: (Probability forecasts in time) (Electoral vote forecasts in time)

Methods of election forecasting audio:

This figure provides a summary of what we know about opinion polls versus prediction markets.


Understanding election simulations audio:

Methods employed by The Virtual Tout®

The Virtual Tout®, an event forecasting service of Research Publishers LLC, draws on betting or prediction market prices to pick winning teams. For the US presidential election, prices obtained from Predictlt correspond to 56 Electoral College markets: 48 of the 50 states, three regions in Maine, four regions in Nebraska, and the District of Columbia. These prices change in response to betting on election outcomes in each of the Electoral College markets. We convert prediction market prices for the Democratic and Republican tickets into estimated probabilities of winning.

We use statistical simulation to generate one million hypothetical elections. We note the total electoral votes for the Democratic and Republican tickets across these elections. Our simulation methods are consistent with methods employed by other election forecasters. What is different about The Virtual Tout® is its reliance on prediction markets rather than opinion polls.

The figure below summarizes study results from one million hypothetical presidential elections based on prediction market data. Because 270 electoral votes are needed to win the election, the probability of the Democratic ticket getting fewer than 270 electoral votes represents the probability of a Republican victory, which is small for the upcoming election on November 3, 2020. We can also look at the median number of Democratic electoral votes. Half of the hypothetical elections in our study fall above the median and half below. We use the median as our predicted number of Democratic electoral votes. The predicted number of Republican electoral votes is 538 minus the predicted number of Democratic electoral votes. We see from the figure that the median number of Democratic electoral votes is well above 270.

Understanding the figure audio:

Forecasting uncertainty audio:

Results updated on Saturday, October 31, 2020, 02:20:04 AM Pacific

Democratic ticket has a/an 84% chance of winning the election

Republican ticket has a/an 16% chance of winning the election

Electoral vote point estimates (medians):

305 Biden-Harris (Democratic ticket)

233 Trump-Pence (Republican ticket)

Electoral vote empirical confidence intervals:

90 percent sure of seeing 245 to 364 Democratic electoral votes

95 percent sure of seeing 232 to 374 Democratic electoral votes

About Data Science Quarterly

Data Science Quarterly from Research Publishers LLC promotes data science as a discipline, showing its relevance to social and political discourse. Data Science Quarterly combines the best attributes of academic journals and popular news magazines. Its articles draw on high-quality empirical research written in clear, concise language that anyone can understand

To trust the news, we want to see the data behind it. Show us the research. Show us the science. To trust an argument, we want to see evidentiary support and correct logic. These values are reflected in the journal's tagline: Following the data, leading with science.

The print version of Data Science Quarterly will be available beginning January 2021, with printed paperbound copies available for distribution in the United States and with electronic file downloads available worldwide. The print version will be published each January, April, July, and October. Articles in the print version of Data Science Quarterly are organized under four major sections:

  • • The Opinion section comprises op-eds;
  • • The Research section presents research studies much like we see in peer-reviewed academic journals;
  • • The Cases section describes practical problems, setting the stage for subsequent research; and
  • • The Reviews section includes commentary on technical books and data sources.

Authors, see the call for papers.

The online version of Data Science Quarterly became active August 20, 2020 with the initial publication of The Virtual Tout ® election forecast. The online version is free for all Internet users and will be expanded with additional news and articles after the November 3, 2020 election.

ISSN 2693-8871 (print)
ISSN 2693-8863 (online)

Introducing Data Science Quarterly audio:

Sample articles from Volume 1, Number 1 (January 2021)

Message from the Editor-in-Chief. Introducing Data Science Quarterly
Read More
Opinion. Seeking Responsible Social Media: The Facebook Case
Read More
Opinion. Protect Public Health Data Now
Read More
Case. Model of an Epidemic
Read More
Case. Willingness to Fly
Read More
Research. Predicting the 2020 Presidential Election
Read More

Learn more at the RP's website

Founded in 2002, Research Publishers LLC publishes books, periodicals, and research reports. RP also offers research, measurement, forecasting, and natural language processing services and solutions.