Time to jettison horse race polls

Time to jettison horse race polls
© Getty Images

The pollsters badly missed in 2016 when they all but promised that Hillary Clinton would beat Donald Trump and become the first woman president of the United States. The Princeton Election Consortium placed her odds of victory at just under 100 percent. “It is totally over. If Trump wins more than 240 electoral votes, I will eat a bug,” declared Sam Wang. But to his credit, he did just that. Nate Silver, the prominent founder for Five Thirty Eight, was less sanguine over the prospects for Clinton. However, he still believed she had over a 70 percent chance to beat Trump.

The pollsters insisted that in 2020 they had remedied such historic errors as they claimed to have increased their response rates, assembled better samples, and ferreted out the hidden vote of Trump and Republicans. But an article by Five Thirty Eight found that Doug Kaplan was the sole one of many pollsters who said he was worried about the hidden vote. However, the pollsters were even worse off this time. They correctly indicated a win for Joe Biden, but missed on everything else, from the Senate and House results to the national popular vote to battleground states.

The errors were not random but consistently underestimated support for Trump and Republicans as the pollster did in 2016. The Princeton Election Consortium and Silver both forecasted that Democrats would take control of the Senate with over 50 seats. Yet Democrats won 48 seats and have a slim chance of winning two additional seats in the runoffs in Georgia. The pollsters badly estimated the losses of Democrats. The Princeton Election Consortium indicated that Democrats would raise their House majority by five seats, and Silver estimated an increase of seven seats.


Silver also set the lead for Biden in the national popular vote at 8 points, which was almost double the actual lead for Biden at less than 4 points. Silver has claimed that his forecasts for president were highly accurate because they correctly called the winner in 48 states. This is a deceitful metric as the election has not been seriously contested for most states. There were less than a dozen true battleground states that were won by either Trump or Clinton by less than 5 points back in 2016.

Further, the Princeton Election Consortium and Silver both miscalled the battleground states of Florida and North Carolina, where they forecasted wins for Biden. The pollsters also had us think that Ohio, Iowa, and Texas, which Trump won by 8 points to 9 points in 2016, were also battleground states since their forecasts showed a thin lead for Trump. But he still won these states with margins of 6 points to 8 points this year.

Given their false assurances for 2020, we have no reasons to believe that the pollsters will somehow find the holy grail of accuracy for 2024. Horse race numbers, moreover, turn our attention to the daily ups and downs of the campaign instead of the dynamics in an election for president, which is a verdict for how well the incumbent has led the nation.

By watching the larger picture of incumbent strength with my keys to the White House, I was able to predict a victory for Biden in early August, well before the pollsters were of any use. Once liberated from the detriment of horse race numbers, the candidates could run a different campaign which is focused not on sound bites or attacks but on building a policy mandate to govern. In reaction to the criticism, Silver said Five Thirty Eight did fine. He needs to find another career or stick to sports betting.

Allan Lichtman is an election forecaster and a distinguished professor of history at American University. He is the author of “The Embattled Vote in America: From the Founding to the Present.” He tweets @AllanLichtman.