Predicting Bad Data

Politics

SONY DSC
I rarely write a blog post in direct response to another article. But I just couldn’t help myself with a post from a The Week story about Neil Newhouse’s predictions for the 2014 congressional elections.

When the 2012 election approached, I became quite obsessed with the polling data trying to get a clue about who would win. And I mean obsessed. I would hover around 538, Huffington Post, Real Clear Politics and any aggregate pollster site I could find. I had a Gallup app on my phone which I checked hourly. I read stories about polling and how some were more accurate than others. I also paid for a pricey copy of The Signal and the Noise by Nate Silver.

Not surprising, most of the pollsters called it right in the days leading up to the election. The one state that gave people fits was Florida, which was fitting given that Obama won it by a hair and the total vote wasn’t even complete until days after the election.

Contrary to what people think about polling, they are very accurate when it comes to elections. I think this is largely because pollsters are dealing with definitive data. Voters have to choose one or the other or none. Election polling isn’t the same as judging presidential approval, consumer confidence or even television ratings, which are more subjective and fall on a scale.

But pollsters hired by politicians try quite hard to develop some inside data, information that common, news polls won’t have. They look for this nebulous information hoping it will give them some inside track on the outcome.

Newhouse worked for Romney during the campaign and had Mittens believing he would win the election up until the polls closed. When the numbers rolled in and it was obvious Romney had lost, his campaign was dumbfounded. I honestly don’t know why. Had they been reading the aggregate polls, which are available to everyone, they would have known that in the end, Romney never had a chance.

There were all sorts of finger pointing post-election. They blamed higher than expected minority turnout. They blamed the impact of Sandy. They blamed Romney’s campaign strategy.

While those are valid faults in the campaign loss, it still doesn’t explain how they were so caught off guard through their polling. Again, they were wildly off the mark compared to everyone else, save only Gallup. Surely someone who’d spent a little time on political websites had to be standing next to Romney telling him that Newhouse was off his rocker. But apparently there wasn’t.

One of the things Newhouse implemented to gain his inside track on the polling was voter enthusiasm. He developed some formula in his polling to gauge how motivated Romney’s voters were about voting for him versus Obama’s people. He rated voters’ motivation on a 1 to 10 scale. More often than not, he found Mitt’s people to be more dedicated.

Newhouse and other Romney backers kept pointing to this enthusiasm advantage as some sort of proof for momentum and a bellwether for turnout. Some were skeptical, including Nate Silver, who repeatedly said there was no momentum for Romney leading up to Election Day. (Nate, by the way, called all 50 states correctly based on his aggregate poll).

Newhouse is now using a version of his same formula to predict the 2014 elections.

While I love data, I also admit fully that I don’t completely understand the complex mining thereof. I do however, grasp basic logic. While enthusiasm might provide insight about voter motivation, at the end of the day it comes down to whether a voter shows up or not. Despite the disparity in motivation, a voter who rates their motivation at a 7 or 8 may be just as likely to show up as one who rates their motivation at a 9 or a 10. And guess what? Their votes count the same.

That’s really the nature of election polling. It’s this or that. No amount of data massaging will get you a magically accurate, contrary opinion that no one else has. It’s not like processing information on stock pricing.

Now, Newhouse isn’t really going out on a limb here by saying that 2014 doesn’t look good for Democrats. Sixth-year elections for a president’s party have never gone well. Ever. His predictions about the November elections likely will pan out barring some unforeseen disaster like a massive Aikenism or Hurricane Sandy or credit default.

We shouldn’t be surprised if he’s right. But we shouldn’t buy the motivation scale either. It just doesn’t add up.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s