When A Good Poll Goes Horribly Bad


When A Good Poll Goes Horribly Bad

CREATED Sep. 20, 2012

There are good polls, bad polls and some polls that are not worth the paper their results are printed on.

For the last year and a half, the Marquette University Law School poll has been the Gold standard for political polling in Wisconsin.  For example, while some polls had Scott Walker in a dead heat with Tom Barrett during the recall race, the Marquette poll absolutely nailed Walker's ultimate margin of victory.

That's why it's so disappointing when a pollster who you respect goes horribly off the rails.

Last month, the Marquette poll had Tommy Thompson leading Democrat Tammy Baldwin by 9 points (50 to 41).  Yesterday, the September poll had Baldwin leading Thompson by the same margin (50 to 41).  In other words, the poll claims that there has been an 18 point swing in one month.

The poll also shows Obama leading Romney by 14 points (up from a 3 point spread in August).

There's no nice way to say this.  I believe the poll is embarrassingly inaccurate and it makes me wonder what's going on at Marquette.

I have no doubt that the Thompson/Baldwin race has narrowed.  Baldwin has been running television ads for the better part of a month essentially claiming that Tommy Thompson hates Wisconsin and eats baby kittens for lunch.  The Thompson campaign has, by and large, not responded.  If you get pounded every day and don't fight back, you'll take some hits - but not a swing of 18 points.

Another poll released yesterday showed the Thompson/Baldwin race tied and Obama up by 6.  Candidly, I think these numbers are a lot closer to reality.

The real question to me is why Marquette put out these obviously bogus numbers.  In their sample, 51% of those polled leaned Democrat, 40% leaned Republican.  That doesn't come close to being representative of the electorate.  In other words, from a statistical purpose - garbage in, garbage out.

It's like going to a Hershey plant and asking workers whether they prefer chocolate or vanilla ice cream or going to a Harley factory and asking workers whether they prefer Japanese or American motorcycles?  Think the results might be skewed?

When Marquette got these bizarre results, you have to wonder why pollsters didn't go back and try again - just to see if their second survey results matched the first?  The fact that they didn't makes you wonder whether accuracy matters - or whether the goal is simply to get a poll out and get some attention?

I continue to believe that Thompson will win handily in November.  Still, these numbers (no matter how absurd) should serve as a wake-up call to the Thompson campaign that it can't take anything for granted.

As for the Marquette poll though, I'm afraid it's time to go off the Gold standard.  That's unfortunate.  I hate it when a good poll goes horribly bad.