Live

Watch CBSN Live

The Problems With Polls

This column was written by Mark Hemingway.

The Internet is ablaze with allegations of vote fraud in New Hampshire. As I write this, the most popular link on Reddit — a popular online social bookmarking service where millions of visitors a month rank interesting stories on the web — is a story claiming that the election was stolen from Barack Obama.

Intrepid amateurs, sleuthing from their couches, are claiming that Hillary Clinton enjoyed a five-point advantage in New Hampshire towns where votes were counted by Diebold voting machines, compared to those where the votes were counted by hand. As further proof of vote tampering, they also note that there was a wide and inexplicably large difference between pre-election polling and the primary outcome.

Fewer than 24 hours after the election, the story was spreading like wildfire across the Internet — see here, here, here, here — and there's a lot more conspiratorial nonsense where that came from. Ron Paul supporters are also claiming they have "CONFIRMED" evidence of voter fraud against their candidate, but I'll spare you the less-than-worthwhile links.

These claims of fraud are baseless paranoia; it has long been alleged that the dastardly Diebold voting machines are puppeteered by the company's management, who are said to be loyal Republican foot soldiers. How, and why, the evil GOP electioneers might suddenly do favors for the Clintons is an issue that none of these conspiracy mongers seems to want to face.

Trending News

As for their disparity argument, if you were following the polls I can't say that these conspiracy theories are sprung from an entirely delusional mindset. The major discrepancy between the polls and the election result could very well be evidence of voter fraud. After all, didn't the media assure us that Obama would coast to an easy victory in New Hampshire? All of the polls confirmed this; even Clinton's campaign seemed convinced of her loss ahead of time. The day of the primary, reports were coming from all directions that the Clinton campaign was likely to have a major staff shakeup in response to the impending loss.

Respected polling firm Zogby also had Obama up 13 points. Tim Russert reported on the night of the primary that the Obama campaign's internal polling had him ahead of Hillary Clinton by 14 points, and that Clinton's own poll had her losing by 11 points. In the world of polling, it's hard to get any more wrong than that.

But to use this discrepancy as evidence of voter fraud, you have to assume the initial polls were reliable. And you know what they say happens when you assume.

In fact, the exit polls matched up nicely with the end result. In particular, they showed that Hillary Clinton enjoyed a 13-point advantage among women, who made up more than half of the voters in the primary. Based on the exit poll data, her victory was not at all mystifying.

The candidates, electorate, and the media alike may all have to get used to the idea that polling — never a bet-the-farm science to begin with — is increasingly unreliable.

The reasons for this are myriad. Increasing dependence on cell phones is one commonly cited explanation. Cell numbers frequently are not listed in open directories, and this cuts out a huge swath of the population from polling samples.

At the same time there is pollster fatigue. Call centers and information technology have reduced the cost of conducting a poll considerably. Thousands of companies and candidates that were once unable to afford polling, can now fund polling. One result of that shift is that voters may avoid answering the calls of pollsters. The day of the Iowa caucus, for instance, I spoke to a grandmother from Des Moines who noted that she and her friends had been so pestered by phone calls this election, they'd stopped answering their phones in the evening over a month ago. Sure enough, most polls in Iowa greatly underestimated Huckabee's margin of victory and a few even had Romney winning.

Moreover, opining to a stranger is quite different than casting a vote in private. The phenomenon, called the spiral of silence, refers to poll respondents feeling pressured to give a more popular answer or to stay silent about their true support. On the other hand, perhaps New Hampshire voters were eager to publicly declare themselves in support of Obama's fresh-faced, inspirational campaign but then in the solemn solitude of the voting booth felt more comfortable supporting the politically-more-substantive war horse, Hillary Clinton.

So while Zogby, Gallup, Rasmussen, et al., scramble to figure what's gone wrong with their polls this primary season, I'm hoping that their polls stay broken. Media coverage of polls dominates, and not enough people are asking whether the obsession with polling is a good thing. In 1976 — the year that CBS and The New York Times announced their historic polling partnership — there were just a few national opinion polls conducted by the media each year. By 1996, there were over 300 public opinion polls conducted in just the final two months of the presidential election. I shudder to think how many polls are conducted now.

Now I'm not saying that polls aren't valuable; as a political journalist I must admit that they are useful. But the polls shouldn't be the story in and of themselves.

Matt Robinson examines the problems that stem from the media's over-reliance on polls in his 2002 book, Mobocracy: How the Media's Obsession with Polling Twists the News Alters Elections, and Undermines Democracy. Despite some significant problems with question wording, sampling error, and response bias, news organizations treat survey results as the Gospel. That, in turn, means pundits, the media, and voters create political narratives to justify the results. As Robinson observes, they commit the logical fallacy of reasoning from effect to cause.

I can think of no better illustration of the problems with this approach than the crackpot theories about New Hampshire cited above. Those disappointed with Obama's loss reasoned that the polls were correct and tried to reconcile that with the election result, rather than the other way around. This lead to such misguided mental gymnastics that it produced conspiracy theories about voter fraud, rather than acceptance that the polling numbers were egregiously wrong.

Even people who traffic in polling data for a living admit that the media's reliance on polls is troublesome. Kellyanne Conway, President of The Polling Company has said, "Polls have become a substitute for thought, for reporting, and for principles." Pollster John Zogby has expressed particular concern with the rapid, overnight cycle of political polling, saying, "A combination of public demand and media demand requires that we get results immediately the next day." Overnight polls are problematic, he says, because getting results so quickly means it's very hard to get a truly random sample.

Further, poll results themselves influence public opinion. For instance, it's possible that polls showing Obama cruising to an overwhelming victory created a blowback effect, which was further bolstered by the sympathy engendered by Clinton's emotional outburst the day before the primary.

For all the talk about how to "empower voters" perhaps the easiest way to do exactly that, would be to stop bombarding the electorate with news about who the likely winner will be months, weeks, days, and hours in advance of actually voting. What could be more electrifying than stepping into a voting booth with no other guide to casting a ballot than your intellect and conscience?

De-emphasizing the importance of polls in our elections is, of course, no guarantee that voters will then demonstrate that they have either an intellect or a conscience. But fewer — and better — polls might force the media to spend more time covering the substance of a candidate's positions rather than snapshot reactions to the candidate's focus-grouped soundbites. This might produce slightly more informed voters who in turn elect slightly better selected bozos to public office. It may not amount to much, but given the uncertainty surrounding the most wide-open election in 80 years, this might be exactly the right time to reassess how much polling we want and need.
By Mark Hemingway
Reprinted with permission from National Review Online

View CBS News In