Of Bandwagons and Underdogs

There's a new batch of polls out today and if you cherry pick with care, your guy is ahead.

And then there is the complaining.  About polls.  About their influence on journalists and pundits and -- most important -- public opinion.

So what the hell is public opinion?  Quite simply, public opinion is what public opinion polls measure.  That's as circular a definition as you'll ever find, and a topic for another day.

Here I want to touch ever so briefly on whether polls influence people.  That is, does seeing or reading or hearing about polls make people move to the majority in a poll?

Yes.

It's called the bandwagon effect.  It exists.  It's real.

But there's also an underdog effect.  It exists.  It's real.  And it often offsets any movement to the majority by a movement to the minority in a poll.

There have been careful studies about both the bandwagon and underdog effects, even a brilliant dissertation written on the topic.  I'm not going to spend my lovely Friday afternoon digging them up and breaking them down for you.  Maybe next week.  Just trust me that polls do influence a few people -- very few -- but the influence cuts both ways, at least when it comes to opinions about a candidate and the likelihood to vote for him or her in an upcoming election.

There are, obviously, more subtle effects of poll results.  Campaign contributions, for example, tend to go to the likely winner, the person ahead in a race.  Why bet on a losing horse?  More subtle but equally important, though, is how a set of positive poll results influences the news narrative.  In other words, today's polls has at least one good result for Romney (Gallup tracking) but a bunch of fairly good ones for Obama.  Expect certain news organizations and pundits to cherry pick their favorite, or you can read someone like Nate Silver, the scary smart guy at the 538 blog, who will break it down for you in a reasonable manner.

But to return to my theme here -- yes, polls can influence people, but they tend to nudge people in both directions, basically the way they were already leaning, and the results largely cancel one another out.