Follow us on Instagram
Try our daily mini crossword
Subscribe to the newsletter
Download the app

What's in a poll?

In this year's presidential race, polls reported in the news can be very confusing. Every day we hear about national and state polls showing Bush ahead, Kerry ahead or one candidate surging and overtaking the other. What is one to make of this river of information?

My field is far from politics: In my research I collect and examine laboratory data. But this chaos of reporting made me very interested in how to read polls. I face a similar problem in the laboratory — how to reduce complex data to a single snapshot. I started wondering if I could come up with a more accurate answer to the simple but vital question, If the election were today, who would win the Electoral College and the presidency?

ADVERTISEMENT

First, a basic fact: National polls are not good at giving us the answer to this question. Why not? The election is determined not by the popular vote, but by 538 electoral votes, which are decided by 51 separate races. We learned this in 2000, when Al Gore won the popular vote but lost narrowly in the Electoral College and consequently lost the presidency.

Second, it's hard to rely on any single poll because of statistical variation, and also because different pollsters may use slightly different methods. So what's a political junkie to do?

To get a snapshot of where things stand today, I decided to do a "meta-analysis" of many polls at once, using basic tools of statistics to provide greater objectivity and precision. Data are plentiful thanks to the work of many polling organizations: Usually several polls are taken every week in each battleground state. After feeding hundreds of state polls into my analysis, I have reached two conclusions.

Conclusion one: The race is nearly tied. Although national polls have reported a wide range of results, the real race — the Electoral College — has been tight ever since late June. Now, in mid-October, the race is essentially a tie. True, major events have caused swings: the release of "Fahrenheit 9/11," the addition of John Edwards to the Democratic ticket, the party conventions, the attacks on Kerry's military record and the first debate all had effects of about 30 to 40 electoral votes. But the underlying effective swings in public opinion have been tiny. This is consistent with the idea that nearly all voters have made up their minds. However, a few undecided voters remain. Undecideds usually end up voting against the incumbent, but until they vote they are yet another source of suspense.

Conclusion 2: This election will be decided in Pennsylvania, Ohio and Florida. About fourteen "battleground" states are still up for grabs. The biggest three are Florida, Ohio and Pennsylvania. A very strong predictor of victory is who will win at least two of these three states. As of today, all three states are close to tied. Now that's a cliffhanger!

The big lesson learned from conducting this analysis is that looking at the whole body of polls can give more insight than looking at one poll, or even the work of one pollster. News organizations often don't do this kind of synthesis and analysis. Political horserace coverage might be clearer if they did.

ADVERTISEMENT

If you care about the outcome, what should you do? In this election a critical factor is voter turnout. Because Princeton is so close to Pennsylvania and Ohio, you are in a perfect position to make an impact. Join up with the Kerry-Edwards campaign, the Bush-Cheney campaign, America Coming Together or a church or community organization. In the next two weeks, knock on doors to canvass for voters. Finally, vote absentee or early, so you will be free on Election Day to get out the vote.

Thinking about data statistically may not be in the regular journalistic toolbox. But that doesn't mean it can't be in yours. To that end, one final piece of advice. If you want to know how the race is either nationally or in your home state, read three polls, then decide. Sam Wang is an assistant professor of molecular biology and neuroscience. He can be reached at sswang@princeton.edu. Details of his election analysis, which is updated daily, can be found at http://election.princeton.edu.

Subscribe
Get the best of the ‘Prince’ delivered straight to your inbox. Subscribe now »