By Madison Hall, Matthew Chen, and Samuel Chiu October 11, 2016
As the most unusual election in modern American history approaches election day, citizens around the country begin to decide whom to vote for. At Stanford Predicts, we use a probability model based entirely on recent state-by-state polling data to project the winning probabilities for the two candidates. For our October 11 model, we input the most recent and reliable polls based on our judgement, including those taken after the first presidential debate, to provide our first probability estimate to predict this election's victor.
Currently, our model predicts that Democratic nominee Hillary Clinton has more than a 99% chance of winning the election, if it were held today.
As you may know, the winner of the election is determined by the electoral college - and there are over one quadrillion combinations of state-by-state election results which would each result in a different makeup of the electoral college. Any combination of electoral votes totalling at least 270 for a given candidate would lead to that candidate becoming the next President of the United States. Our model predicts that the most likely outcomes of the election would result in Clinton winning 328, 337, 343, or 352 electoral votes. (We've done similar predictions of electoral vote counts in previous years - check out the About Us section if you're interested in seeing how we did in past years, and the Methodology section if you’re curious about how we find the possible combinations of electoral votes.) By summing the probabilities of the outcomes where Clinton wins at least 270 electoral votes, we see that the probability of Clinton winning at least 270 electoral votes is greater than 99%.
In addition to running our model based on state-by-state polling data, we also ran the model using the state-by-state win probabilities for each candidate produced by FiveThirtyEight as the only data source, bypassing the first two steps of our model as shown below. (Specifically, we used the polls-only forecast that FiveThirtyEight publishes.) Interestingly, we again arrive at a 99%-plus win probability for Clinton, while FiveThirtyEight publishes a win probability of 83.5% using the exact same data. Since their methodology is not public, it's not clear how they arrive at this conclusion from their own state-by-state win probabilities.
Over the upcoming weeks, we will continue to publish new histograms as we update our model with new polling data. In the meantime, get excited for the final presidential debate; as with any election, unforeseen breaking news may change the polling and impact our prediction, so stay tuned!