The 2016 Presidential election results are still a political hot potato. So let’s leave political questions out of this and ask the other question that begs to be answered:
How did the polls miss the mark so badly?
We have written before about some of the challenges in modern polling, and you can read about them here. The fact is that technology is changing virtually everything on the planet, and polling is no exception. Clearly, the 2016 Presidential election results show that overall, the polling industry has not been keeping up with the political technological times. Consider:
- Hillary Clinton was consistently given a comfortable lead over Donald Trump in the weeks leading up to the election in polls conducted by Bloomberg Politics; CBS News; Fox News; Reuters/Ispos; USA Today/Suffolk; Quinnipiac; Monmouth; Economist/YouGov; and NBC News/SM
- Of 61 national polls conducted since October 2016 that tracked both candidates, 55 of those polls predicted a win by Hillary Clinton.
- That means only six polls conducted a month or so before the election predicted Donald Trump’s win.
And all six of those polls came from one place: The Los Angeles Times/University of Southern California-Dornsife “Daybreak” tracking poll.
The Daybreak poll was established as part of the ongoing Understanding America Study (UAS) at the University of Southern California’s (USC) Dornsife Center for Economic and Social Research, in partnership with the Jesse M. Unruh Institute of Politics and the Los Angeles Times. The poll was panel-based: each day one-seventh of the members of the panel were asked to answer three questions:
- What is the percent chance that you will vote in the presidential election?
- What is the percent chance that you will vote for Clinton, Trump, or someone else?
- What is the percent chance that Clinton, Trump or someone else will win?
As the answers came in, the charts were updated daily with an average of all of the prior week’s responses.
Political pundits and the assorted media didn’t seem to give the poll much credence before the election. Responses ran from complete disregard (Huffington Post did not include the Daybreak poll in their general election coverage at all) to criticism. The New York Times singled out the survey’s weighting methodology – the way in which responses were manipulated to try and more accurately represent the diversity of the population, so that (hopefully) a more accurate prediction can be extrapolated – as problematic.
The L.A. Times/USC poll does indeed use a weighting plan that is more complicated than that of most other surveys. Specifically cited as cause for concern was the weighting of the responses of a Trump supporter who happened to be a black male. Said the paper, the survey was “distorting national polling averages,” as the most “over-weighted” person in the survey was “unrepresentative of his demographic group.”
But the Daybreak pollsters stuck to their guns. The L.A. Times published an op-ed piece about the poll on November 7th – the day before the election – urging readers not to “dismiss its data.” And on election day, the proof was in the pudding.
But was weighting the source of Daybreak’s accuracy in this election where everyone else failed? Or was it something else? Or something more?
Polls typically use one of three methodologies to collect voter opinions: traditional live-caller telephone polls; automated polls (robo-call polls); and internet polls. Daybreak used the internet. And the anonymity of the online answering may have made all, or at least a significant chunk of, the difference.
Arie Kapteyn, Director of the University of Southern California’s (USC) Dornsife Center for Economic and Social Research believes some voters may have felt uncomfortable reporting that they would vote for Trump to a live person on the phone. She said,
“There’s some suggestion that Clinton supporters are more likely to say they’re a Clinton supporter than Trump supporters are to say they’re a Trump supporter.”
Whether it was social pressure, fear of being regarded in a potentially negative light, or just mistrust of the polling process; something led to a big difference in what people reported they were going to do depending on the type of poll being conducted.
Traditional telephone polls involving human-to-human contact yielded mainly Clinton-leaning responses. The Washington Post cited this trend when they published the headline, “When it Comes to the Polls, Cell phones are not Donald Trump’s Friend.”
But give responders the anonymity of the Daybreak survey’s internet methodology, and their responses leaned more definitely towards Trump.
Here’s a well-known fact: what people say they will do and what they actually do are often different. In psychology, this is called the attitudinal versus behavioral dimension.
Behavior is external to a person – what a person actually does yields results that are visible to others, and therefore that can easily be measured. Attitude, on the other hand, is a person’s mindset. It’s an internal belief or feeling, which may or may not be expressed out loud (or in a Tweet, or on Facebook, or in a poll) and can be extremely difficult to gauge accurately. Data gathered to indicate attitude is tougher to rely upon, and it can absolutely be affected by the ways and means in which it is gathered.
Polling is a scientific field with many variables to measure. Pollsters are in the business of predictions. Accurate predictions. For the unique set of circumstances that led up to the 2016 Presidential election, the Daybreak poll’s internet data gathering and response weighting yielded the most accurate results. Certainly, a takeaway for observers is that technology, voter behavior, and attitudes are changing. Using the former to accurately gather and interpret data on the latter two is becoming more and more important to politics in the US.
At 5i Solutions Inc. our business is making sure you have access to your data and the ability to manipulate it in the ways you need to make evaluations and predictions to move your business forward.
We use technology to connect people, computers, and businesses to the information they need, no matter the location or device. We have the tools to make sure you can access it in the way that is most effective for you: the behavior of the system matches the attitudes and needs of the user. And the internet and cloud data solutions link it all together, allowing your business access to the information needed to make better business decisions today.
5i Solutions Inc. offers custom-tailored document management systems that can find any document instantly; tools that turn handwriting into digital smart data; even software solutions to allow legal teams to work smarter and analyze tremendous amounts of different types of documents and communications. Using today’s technology to access today’s data can keep your business’s projections on point, and help you anticipate the future.
Data is nothing without security, and 5i Solutions’ Cloud Vault offers site-to-site VPN encryption, network security, and key management. Custom-tailored levels of identity verification and access control let those who need to know in, and keeps those who don’t out. Threat management, penetration testing, monitoring, logging, and on-demand reports ensure the continued integrity of your data.
5i Solutions Inc. has the tools and technology to help you grow your business, and the know-how to customize those tools and technology to perfectly fit what you do.
5i Solutions. One single, secure point of intake, access, and storage. One singular solution.
5i Solutions, Inc.