The Trumpoll Problem

The history of political polling is rife with examples of polls being wrong. To be sure, they are often right, or right enough, but there are many examples of polls being dramatically wrong and so we understand that polls are not votes. The emergence of online environments and all that they add or subtract from opinion making must, one can imagine, have an impact on the nature of polling. I’m speculating.

I should acknowledge that I have long been more interested and attracted to the failures inherent in social research, including polling, than I am its success as a tool for understanding humans. In fact, I have been guilty of spending more time explaining why a piece of research I have conducted might not be helpful than I spend talking about what we can learn from it.

(“So, are we paying you to tell us why you can’t really tell us anything?”)

So and then it will come as no surprise to you, because you have been thinking the same thing, that I wonder about the polling numbers for Donald Trump. At this point we should pause and take a moment to remove all the energy we now experience around Donald Trump, or set it aside, tie it to a helium balloon, whatever works for you. Whether your energy is negative or positive or simply confused, that is not the thing I am interested in right now.

People lie to pollsters, or tell half formed truths, maybe. It happens. For example, it is well established, though not without caveats, that conservatives are sometimes reticent in offering their true opinions (this is called the “Shy Tory” effect). Another problem is that people might not answer truthfully when one of the candidates is a minority (the “Bradley Effect”). These are relatively exotic problems with polling. Others are more common and mundane, the most common being the sample error relative to size, which polls attempt to correct for by offering a margin of error. The margin of error is a statistical recognition of the fact that you cannot ask everyone the question. If you cannot ask every single adult in America, you are obliged to acknowledge the potential for inaccuracy, which is greater the smaller the sample size.

There are other errors and many of them have to do with how well the sample that is standing in for the general population serves as a microcosm of that population. These can be really interesting because randomness requires discipline, ironically maybe. If you are stopping people on the street to ask them questions, how does your personal preference not become part of the sample? People who look like they’ll kill you if you ask them a question are part of the population, but may not be accurately represented in a sample group.

The biggest problem to my mind, and the one I think is most relevant when it comes to looking at Trump’s poll numbers, is the assumption, both necessary and flawed, that the people who refused to answer your silly questions would have answered them in the same way as those who agreed to answer them.

We know that people who agree to participate in polling are not necessarily accurate representations of everyone. Generally speaking, the reluctance to speak with a pollster is not considered a statistically relevant variable or one that is corrected in the margin of error related to sample size.

Simply put, it is my opinion, based on nothing whatsoever beyond my own personal observations, that Trump supporters, much like their candidate, are very eager to share their opinion; whereas, your average Republican, more thoughtful, has not yet formed an opinion because they are not yet sure what has become of their party or where to assert their support. Trump is simply a more extreme example of how the Tea Party minority has taken a seat at the table just by being consistently vocal.

I live in the most conservative county in the state of Georgia, which is saying something, and yet I can count on one hand the number of Trump bumper stickers I have seen. By this time in the last election I was seeing Romney bumper stickers by the dozens. They were everywhere, at every stop light.

My opinion, and I will not be surprised if I am wrong, is that we are seeing polling anomalies that pace Trump’s incongruence as a candidate. I mean, how many Trump supporters have you met personally and how does that number compete with how many Republicans you know?

I have not met one Trump supporter. I keep thinking about what advertising innovator David Ogilvy, a real life madman, said about focus groups 30 years ago. He said talking to seven of his friends was as good as a sample group. I’ve always believed that was true. If Donald Trump was really going to be relevant in any meaningful way, I think he would have come up in one real world conversation with a supporter by now, or more than two bumper stickers in the most conservative county in Georgia.

It is my opinion that Trump supporters are eager, excited, assertive and want everyone to know what they think and are disproportionately represented in polls at a time when most Republicans don’t know what to think.

And this is not taking into account numerous other issues such as how questions are written or how pollsters contact people. Landlines (inexpensive to reach) will skew Republican. Cell phones (expensive to reach) will skew democratic. Fewer and fewer people are willing to sit through a 30 minute set of polling questions (more questions, more accuracy) so how do people who are willing to sit through 30 minutes skew the results. Are they more passionate? More decided?

Look, I’m not a statistician or a professional researcher or a pollster. But I’ll tell you what, Donald Trump’s seeming popularity does not square with my daily experience. I have this feeling that we’re looking at smoke and mirrors. To what end I cannot even guess. If Trump turned out, as Jeb Bush has suggested, to be a spoiler on behalf of Hilary Clinton, I honestly would not be surprised. Everything is too out of whack not to expect the unexpected.

Finally, any one candidate’s actual chances are difficult to define in a crowded race. There is a large peloton behind Trump. I use this bike racing term as a metaphor because if Trump is truly in the lead then he is winning as Lance Armstrong.




Leave a Reply

Your email address will not be published. Required fields are marked *