Which Democratic candidate is the most electable in the electoral college?

Ian Ayres

Ian Ayres and Zachary Shelley

In a recent coauthored LA Times op-ed, we discuss the importance of polling on head-to-head matchups with Trump for determining electability of the Democratic primary candidates. Using state-level polls on these matchups with Trump, we find that Joe Biden and Bernie Sanders would be expected to defeat Trump in the general election. These two candidates would be expected to flip 106 and 84 electoral votes, respectively. Democrats only need to flip a net of 38 electoral votes from 2016 in order to win.

We note, however, that Joe Biden has the largest electoral college lead of these candidates, which provides a buffer against the possibility that polls incorrectly measure Trump’s support (as was the case in 2016).

The graphs below show how several candidates (including Elizabeth Warren and Pete Buttigieg, who have dropped out of the race) poll compared to Clinton’s 2016 election results.







Biden polls ahead of Clinton’s electoral margin in each of the 26 states for which polls are available, except for New Mexico, New York, and California, which are all blue enough that his relative performance wouldn’t cost any electoral votes. Meanwhile, Sanders polls worse than Clinton’s electoral performance in four states, including a disadvantage in Delaware large enough to put the state in play for Trump.

Particularly crucial are states that swing to Republicans (in the lower-right) and states that swing to Democrats (in the upper-left). Both of the leading Democrats left in the race are projected to win all of the states that Clinton captured in 2016.

Current polling suggests that Biden is expected to flip eight states blue (Ohio, Michigan, Pennsylvania, North Carolina, Wisconsin, Florida, Georgia, and Arizona) and Sanders would flip six (Michigan, Ohio, Pennsylvania, North Carolina, Wisconsin and Florida).

To determine the expected electoral college votes for each candidate, we use reported poll means and standard deviations to determine the probability that each candidate would win the state’s electoral votes based on that poll. We then calculated the average likelihood that each candidate would win a given state, weighting by the number of days since the start of 2019 (in order to give recent polls more importance). Having calculated the expected probability of winning each state, we determined the expected number of electoral votes a candidate would receive from each state with polling data. (Since all of the states without polling data are considered either safely Republican or safely Democrat, we assume that their electoral votes will go to the party that won the state in 2016.)

We also determine exactly how much polls would have to incorrectly measure Trump’s support in order for Trump to win the presidency again. To do this, we first calculate the average Democrat polling margin in each state – again weighting by the number of days since the beginning of 2019. We then calculate the number of electoral votes that each candidate would garner with the assumption that polls are perfectly accurate. We iterate on this process, changing the assumption to decrease the Democratic candidate’s vote margin in each state by 0.1% increments and determine the smallest change necessary for Trump to win an election against each candidate. While our method shifts the vote margin in every state, it is only necessary that the polls underestimate Trump’s support by this amount in states that swing from expected Democrat to expected Republican. It could be the case that the polls over- or under-estimate Trump’s support in other states, so long as those changes do not cause the state to flip (i.e., polls could overestimate Trump’s margin in Alabama by 10 points, but even this swing would not change the fact that Alabama would still be solidly Republican).

Using this method, we find that polls in swing states would have to underestimate Trump’s support in a competition with Joe Biden by 5.9% in order for Trump to win reelection. In head-to-head competitions with Bernie Sanders, swing state polls would only have to be off by 3.8% for Trump to win.

Using this data on matchups against Trump provides stronger evidence on electability than national polls or unsupported arguments about voter turnout.

You can find our full analysis here.




Posted 1:58 PM by Ian Ayres [link]