Why Election Polling was so Wrong


On the heels of an election upset, market researchers are questioning how political polling predicted the wrong outcome this election season. “The prevailing talking point in the market research industry the day after the election was, ‘How did we get it so wrong? And how can we do better moving forward?’ There’s a clear call to action among research professionals to address the lack of accuracy that characterized this season’s election polls,” noted Kerry Edelstein, President and Founder of Research Narrative,  Inc.

Methodological  Changes

When Bill Clinton ran for President, political polling was an easier task. Notes Edelstein, “In the 1990s, we were still in the era of phone-based research. There was such a thing as ‘random samples’ of the population, caller ID was uncommon, and people more often answered their phones to take surveys. Niche populations were possible to reach. That began to change by 2000, with a shift toward online panels. And what we’re finding in 2016 is that online samples often are no longer representative of the voting population. And neither is a sample of phone survey respondents who pick up their phone without screening the call. Younger, non-white, and rural populations can be especially hard to reach and often go  under-counted.”

The Role of  Celebrity

But celebrity can also play a role. Notes Edelstein, “When it comes to elections, never underestimate the element of fame. We saw this with Reagan; we saw this again with Schwarzenegger in 2003 with the California governor recall election. Trump swept the primaries in a way that few expected, and yet I don’t think researchers sufficiently considered the role that celebrity still had on the general election outcome. Familiarity is a critical first step in getting elected, and Donald Trump helmed a popular TV show for 14 seasons and has his name on buildings across the country. Americans came into the election already very familiar with his name, his brand, and his leadership  style.”

New  Solutions

As researchers discuss how to improve prediction accuracy, one thing they’re not disputing is that old mechanisms of projecting election outcomes are growing less reliable. “Voters have more access to information than ever, and that means their decision process isn’t as clear cut as it might have been in past years.” notes Edelstein. “I’m hearing veteran researchers ask, ‘Are we even asking the right questions? Do we need more sophisticated models, with new inputs?’ To me, that’s super exciting. It means that research innovation is on the  horizon.”

About Research Narrative,  Inc.

Research Narrative, Inc. is a Los-Angeles based media research and analytics consulting firm. The company offers a wide range of market research and analytics consulting services focused on data  storytelling.

All opinions expressed on USDR are those of the author and not necessarily those of US Daily Review.