Can We Trust What Political Polls Tell Us Or Are They Simply Too Problematic?

Why you shouldn't believe everything the pollsters tell you.

Jun 15, 2024 at 12:27 am
Go vote!
Go vote! adobe

“There are lies, damn lies and statistics”

          Benjamin Disraeli, British Prime Minister in the 19th Century 

There are political polls, polls, polls these days. I’m obsessed with these polls. Read them every morning, noon, and night. Get angry when these polls are bad for my favorite candidate(s). Get excited when they show more favorable results.  I should know much better to take them seriously this early in the presidential campaign.

For you see, political polls have been following me around since I was a graduate student in educational psychology and research. I have been analyzing them ever since. Analyzed them for my political blog. Analyzed them for political candidates. Have discussed them with national authorities on polling. 

I have learned what political polls are not. They are rarely accurate predictors of actual elections. Pollsters cannot predict with certainty who is or is not going to vote in an election. Polling corporations have developed sophisticated models of identifying likely voters for a particular election. These models can be problematic.  

Nearly all of the currently reported polls are based upon a pool of registered voters.  A percentage of these voters will not vote in the presidential election. The non-voters are driving Trump’s lead in the polls.  Don’t believe me, then believe Nate Cohn, eminent political guru for the NY Times. He recently noted in a NY Times column: (Trump’s lead in the polls is) “built on gains among voters who aren’t paying close attention to politics, who don’t follow traditional news and who don’t regularly vote.”   We thus don’t know if Trump is leading with likely voters.

Furthermore, Professor Larry Sabato of the University of Virginia has noted: “For our general election outlook, we are not taking current polling much into account right now. Biden’s national polling right now is probably worse than what our ratings reflect.” Listen to the good and wise Professor. 

Pundits tell us that political polls are a picture in time of potential voter behavior.  NOT TRUE. They are a fuzzy picture (full of possible errors). A measurement of error (MOE) is inherent in any polling data. This sampling error is usually between 1.5 and 3 points for all sampled candidates. A poll might, for example, show Trump with 47% of the vote and Biden with 45%, and has an MOE of 2.  This statistic must be interpreted (which rarely is done by pundits) as Trump having between 49 and 45 percent of the vote and Biden having between 47 and 43 percent. Biden may then be in the lead or badly trailing Trump. 

Polls may also contain internal errors that are beyond the MOE. Simply put, a poll should be a stratified sample of the voting population. Some polls may have oversampled Democratic voters and others Republican voters. You don’t know.

Don’t fret if your favorite candidate is not winning now. Don’t listen to the chit-chat on your favorite news outlet. Don’t become like me, obsessed with questionable poll data.