A friend of mine posted on Facebook that she’d taken a web quiz to tell her which presidential candidate best lined up with her stand on the issues. She was outraged that the web site thought she would vote the way it did. I’m not surprised (by the outrage, not her choice)…it is a case of a badly applied choice technique.
Basically the quiz worked by asking a series of questions to see where she stood on the issues. It then aligns her choices against the stand taken by the candidate (if you want to try one, here is one from the GOP Primaries this year). In essence it is a Configurator. Instead of building the perfect product for you (as you would with a Configurator) you build the perfect candidate. There are a couple of problems with this application.
First, Configurators allow you to build the ideal but generally don’t give a clear idea of what choices you might make if that ideal were not available (our proprietary Texo™ helps overcome that issue). In politics it is not unusual for voting decisions to hinge on a single issue and unlike products you can’t decide to add or subtract an important feature.
A bigger problem is that a candidate, just like a brand or certain products, might present us with stark realities that we don’t want to accept. For example, think of the middle aged father of 3 who knows a minivan is the best choice but still sees himself in a convertible. I might agree with a candidate on every issue, but still not want to see the actual candidate as a reflection of who I am.
So, it isn’t surprising that a Configurator doesn’t work, but what about discrete choice ? Easy enough to imagine things like party or various issues as “features” and to construct an experiment in which you ask respondents to choose which combination of features they would vote for. You could even include things like the candidate’s age, gender and race to see if it makes a difference. You’d end up with utilities for each feature and be able to construct the candidate who would get the most votes.
Of course, this doesn’t fully account for the “minivan vs. convertible” issue. The image of candidates can supersede their stands on issues. Unless you include them in the exercise, you can’t account for this, anymore than you can do a discrete choice without considering the inclusion of brand. I suspect this leads to an even bigger challenge, that of too many exclusions.
Exclusions are used in discrete choice to remove nonsensical choices (such as a two seat convertible minivan). In politics, “nonsensical” takes on a whole new meaning. A candidate is free to say they want to eliminate taxes, increase spending and balance the budget if they want. Voters might see this as impossible (nonsensical) or go along cause they want to believe it can be done. As such you can’t exclude choices like that.
At the same time, voters don’t like candidates who change positions drastically so while on some issues candidates will attempt to straddle the fence a bit or nuance their position, on others we need to exclude certain positions from coming up with certain candidates.
If your goal is to predict election outcomes, an even bigger obstacle would be accounting for the level of knowledge of the average voter vs. those given full information in the discrete choice survey. This is a weakness of survey research that is not unique to politics, but I believe it is a far bigger issue in presidential politics. Someone marketing a new product might well face competitors' advertising or web chatter on social media which counters their carefully crafted message, but nowhere near what politicians face. Think about the worst ad you’ve ever seen attacking a product or service and compare it to what you see in politics. Think about the sorts of comments that political posts on Facebook get as compared to those endorsing a product or service. Finally, think about how often cable news talks about products vs. politics. Accounting for all these challenges in a political conjoint would be extremely difficult.
I’m sure we could apply the same expertise we’ve used to overcome these and other challenges, but for now I’ll stick to doing research on products and services and remain an interested amateur political junkie. Of course if a future candidate were to call….
Rich brings a passion for quantitative data and the use of choice to understand consumer behavior to his blog entries. His unique perspective has allowed him to muse on subjects as far afield as Dinosaurs and advanced technology with insight into what each can teach us about doing better research.