Lori Shelby, Ph.D., Founder & CEO
The last blog post in our series on voting methodology for #MethodsMondays takes a look at the challenges of polling accurately for U.S. presidential elections. Although the topic is of general interest to survey research geeks like myself, before I start on the details I wanted to clarify that no polling method at this historic time should be considered infallible. It is human nature to choose not to take the time to vote when on the news they are already calling who is expected to win in your state. I would encourage you this year to vote when you can and ignore the poll results, which may or may not be correct. Let's Vote!
Don't get me wrong, there is a great deal of science behind polling and as someone who has built a career on survey research, I believe in the research. However, we all remember Hillary Clinton was expected to win by approximately 85% in the last election. It is reasonable to infer that this suppressed voters who just assumed their vote wasn't important, thus helping President Trump to win. The American Association of Public Opinion Research released a well researched and considered report: An Evaluation of 2016 Election Polls in the U.S. Reasons for the discrepancy included late decisions by voters on who they were going to vote for (about 14% of voters in Michigan, Wisconsin and Pennsylvania in the final week). Also, there was an overrepresentation of college graduates in the polls, which were not properly adjusted during the analysis. Most pollsters have now fixed this mistake.
However, the need for survey researchers to learn from their mistakes is concerning in an election where there is significant and unprecedented events occuring. For example, the uncertainty of Covid#19, increased voting using alternative methods, a growing fear of violence at polling places on election day, and the potential for an unprecedented voter turnout. Here is a great article explaining how the polling modeling has improved over the last 4 years and electoral college math that explains why Biden may be up in the polls, but how President Trump could still win. The Pew Research Center released on article on what we can trust the 2020 election polls to tell us a few days ago.
In my opinion, surveys are a valuable tool. But, perhaps in this case they are of more value for understanding why people are voting as they did, and why election results happened as they did. We will hopefully see final election results very soon and that is when we can start discussing and learning from the accuracy of the 2020 presidential election polls.