After the 2016 election I felt the need to both defend the polls and point out how they differ from what we do at TRC…market research. To paraphrase Ronald Reagan…”Here we go again!”
With the surprise victory by Donald Trump in 2016 there was much soul searching in the polling world. For this cycle we were assured that they learned from the last cycle. What they determined was that they undercounted working class whites so they fixed that. In most cases though they were still off and often outside the margin of error. I think generals call this “fighting the last war”.
One issue is that politics can get ugly fast. People may prefer Coke or Pepsi but that preference is unlikely to cause a family rift at Thanksgiving. While it is impossible to prove, it stands to reason that some people might be afraid to tell an interviewer (or even a web survey) the truth about their voting intention. How much this impacts things is hard to say.
While people are less likely to fear giving their opinion in a market research survey, we still need to be on guard. For example, we know that price laddering studies create bias that doesn’t exist in a monadic design (and is much reduced by the use of things like conjoint analysis). That’s why careful design in research is critical.
In the end though, my argument last time still holds…polls can only take us so far. The country is very close to evenly divided. In an election with the biggest turnout in history and highest percent of eligible voters in over 100 years (meaning going back to a time when a much smaller percent of the population was eligible as women were not allowed to vote) it shouldn’t surprise us that polls are often “wrong”. Many are within the margin of error mind you (therefore not wrong) and other differences come down to issues related to turnout (such as the surprising turnout of white working class men last time).
When we do market research we face our own “turnout” issues. Even established markets require us to determine things like “likely car buyers in the next six months”. We know that intent and reality don’t always line up, but they get close enough for our needs. After all, in market research it is not all or nothing like it is in voting. Only one candidate will be the next President whereas win or lose Coke or Pepsi will still sell a lot of soda.
Our “turnout” issues become really challenging in new or rapidly changing markets. For example, if you needed to understand the electronic payments market you have to figure out who the customers are (new ones start up every day) and who the competition is (this isn’t limited to direct competitors but also includes traditional players like credit cards or banks AND new players who might not have handled payments in the past. We’ve helped clients figure out (and size) markets like these…in essence, made sure that they don’t ignore the equivalent of working class voters in their market (if you’d like to understand how we’ve successfully done this, let’s have a chat… [email protected]).
Defining markets like this requires a carefully planned customized approach…and I am speaking from experience as we’ve done this in multiple markets (happy to have a call to discuss how it might help in your marketplace).
Another difference between market research and polling is we don’t have to face weeks of counting, recounts, court battles and so on. For now, I’m going to focus on helping my clients navigate the weeks ahead. I suggest you all do the same…the politics will take care of itself.
I suppose, given the nature of this blog, that you might want to know where I stand. Honestly I am really split. I like the slightly sweeter taste of Pepsi, but there is something special about drinking Coke from one of those small bottles right from the fridge. Read More