Since 2016 (and even before then), there has been an ongoing debate about whether to trust polling and if so, to what extent. Some argue that polling is untrustworthy, while others claim it is incredibly accurate and that we should put all our trust in it. Both of these approaches are rather extreme and irrational.
Polling accuracy has varied (sometimes vastly) each election cycle, so it is best to analyze polling and its accuracy on a case-by-case basis to then further analyze how much polling can be trusted in a generalized way.
The debate about polling largely stems from the 2016 presidential election, where polling wrongly showed Former Secretary of State Hillary Clinton beating Donald Trump. It is without question true that polling in 2016 was wrong, but it is also true that most polling for the popular vote was not far off. The Real Clear Politics polling average showed Clinton up by three points nationally, one point off from the final result of Clinton winning the National Popular Vote by two points.
So yes, in 2016, polling did not correctly forecast who would win in swing states around the Country, but polling was generally relatively accurate, contrary to common perception.
The next election we can look at is the 2018 election. To analyze this, I looked specifically at the polling for the Generic Ballot (which party voters think they will vote for). While there are a variety of polls I could look at, this gives the best generalized look at polling.
In 2018, Generic Ballot polling had Democrats up by about eight–nine points. Compared to the 2018 United States House elections, this is on point. In the Popular Vote, Democrats beat Republicans by 8.6 percentage points, with polling showing spectacular accuracy.
While polling in 2016 and 2018 seemed accurate, 2020 may be the election where polling was most questionable. The Real Clear Politics polling average from 2020 showed (then) Vice President Biden beating President Trump by 7.2 points. The National Popular Vote ended up favoring Biden by 4.5 points. The 2020 election had the most inaccurate polling, even though it got the final result (a Biden win) correct.
The last election we have to analyze is the 2022 Midterm elections. In 2022, the Real Clear Politics polling average had Republicans winning the Generic ballot by 2.5 points. In the election, Republicans won the Popular vote by 2.8 points, a display of incredible accuracy.
The lesson I have learned from this is that despite the partisan chatter about polls being biased or wrong, polls remain the most accurate and trustworthy indicator of who will win an election.
With this said, polls should not be taken at face value without context. As we all know, we “exist in the context,” and it can be easy to misunderstand polls if we do not consider what has been happening at the time of the poll.
Polling only tends to become truly accurate towards the end of an election. Polling early in an election cycle often indicates one candidate is farther ahead than they are.
So my advice to everyone looking at polls this election cycle is that they are likely giving us a relatively accurate indicator, but that we should stay caught up in what polls do (and don't) say.
Photo Credit: David Paul Morris/Getty Images/KQED
Comments