Times Have Changed by Such a Degree that Polling Can’t Account for it

By Michael Todd

Something went wrong with polling in 2016, and they haven’t been able to account for the discrepancy. Before anything can be done to correct the inaccuracies in polling, the problem needs to be recognized in order to account for the issue. If you are selecting the wrong set of data, then selection bias will occur. In instances such as polling, it is vital to discern between the total number of participants and the number of participants who directly relate to the data set needed to yield a prediction. An example of this would be likely voters over registered voters.

In 2016 pollsters said every poll would have to be wrong for Trump to win, which came to fruition. Why cant it happen again? Something changed in the electorate that pollsters cannot account for in their models. One such variable is the increased reliance on social media. This has altered how people receive their news, making it harder to account for changing perceptions during an election. Pollsters don’t just need to account for change, but now they need to account for the rate at which change occurs.  With all its downsides, social media has helped voters obtain differing positions on issues that significantly affect their lives.    

Enthusiasm for Biden is at a historic low — so low that Democrats run the risk of replaying 2016. Only 43 percent of Democrats strongly support Joe Biden. Given the same question, 75 percent of Republicans said they strongly support President Trump. When voters are enthusiastic about their candidate, they tend to come out in more significant numbers instead of their prime motivation being the opponent’s defeat. It’s also worth noting that 89% of voters chose the economy as their most important election issue, which bodes well for Trump because people do not blame him for the economic downturn; they blame COVID. The handling of the coronavirus came in third, double digits behind the economy. Many lifelong Democrats feel their party has gone off the deep end and no longer represents them.  This is not indicative of the polling numbers we see that show Biden ahead by 9 points.

Early voting data shows an entirely different story from the national polls. For instance, the Democrats were projected to have a sizable early voting advantage, yet Republicans are outperforming predictions in many key battleground states. This is one of many variables affecting the outcome of the predictions being made.  Many Trump voters haven’t voted yet. Based on CBS’s polling in crucial battleground states, more people who support Trump over Biden 59-40, haven’t voted yet. Polling USA presidential poll shows that among those who haven’t voted in Florida, Georgia, and North Carolina, Trump leads by 19, 10, and 17 points. What this shows is that Trump is most likely going to get a significant boost on election day.  This coupled with early voting numbers showing Republicans outperforming expectations, should worry Democrats. High-propensity Trump voters will increasingly start showing up in force. Democrats are exhausting far more of their high-propensity voters, and the margins are expected to begin tightening considerably, as they have in Florida.

Newly registered voters and Democrat defectors are not being adequately accounted for. In many of the key battleground states, Republicans have seen a 2-1 new voter registration advantage over Democrats. Polling firms are using generalized percentages to calculate party affiliation, thus oversampling Democrats by wide margins. Biden’s chance of winning decreases dramatically when factoring in the current landscape and removing the Democrat oversampling. Proportions matter, and slight variations can make a big difference in polling.

In the Senate tracking polls, Republicans have doubled their early voting numbers and are expected to have a very high turnout on election day. This is significant because Republicans generally vote in higher numbers on election day than early voting.  It seems to show that Republicans have a great shot at retaining the Senate. In presidential battleground states,  48 percent of the votes cast in 2016 have been received. The data only indicates party registration, not which candidate voters support. Nate Silver at 538 also said that many of the mail-in votes that come in late would be neutral or favor Trump. With most Republican voters expected to vote on election day, we could see record turnout for Trump and the GOP. 

Accurately forecasting the election can be quite challenging when factoring in all the variables needed to provide a precise prediction. People are also very aware of what can happen when they speak up for their beliefs, so they watch from the sidelines, not feeling comfortable to say what they think. In a highly divided country, censorship from social media has forced people to become more reclusive with their views, making it harder to account for in polling. One thing to consider is that the poles have been wrong many times before. In 1988, Michael Dukakis enjoyed a lead in the polls over George H.W. Bush that was nearly double Biden’s lead over Trump in the RCP average. Bush would go on the win 426 electoral votes to Dukakis’s 111, securing the presidency.   

Leave a Reply