Hours before reporters rang in the New Year in Des Moines, the topic of conversation among many was a new poll released by the Des Moines Register suggesting not only that Barack Obama had a big lead there but that turnout would exceed anyone's expectations. No one could believe it was true, as every other public poll showed a tighter race.

The pollster who conducted the Register's poll, Ann Selzer, is considered by many to be the best at surveying the tricky terrain of the Iowa caucuses. In the following days, she withstood blistering shots from top strategists to Hillary Clinton and John Edwards, who attacked the poll as fatally flawed and a clear outlier.

But come caucus night, Selzer was proven right: Her margin was closest to the outcome, and her turnout predictions proved true: Most campaigns expected anywhere from 150,000 to 200,000 on the Democratic side. Nearly 240,000 people actually turned out. What the media called a close race with record turnout became an Obama blowout with nearly double the turnout of the previous Democratic record.

In New Hampshire a week later, the media predicted another blowout and an Obama win by double digits. That didn't happen either, as Clinton scored a two-point upset that some have called one of the greatest comebacks in American political history. So how did pollsters and the media get it so wrong both times?

The truth is, everyone gets polls wrong sometimes. And the results and reporting out of Iowa and New Hampshire were no different from any other election.

There are many reasons for Obama's under-performance in New Hampshire, and few would be obvious before an election occurred. The youth vote, which went heavily for Obama in Iowa, was significantly reduced in New Hampshire -- 18-29 year olds made up 22% of the electorate in the Hawkeye State and only 18% up north, dropping Obama's vote total.

Bolstering Clinton, married women not only turned out in greater numbers, they also gave Clinton a bigger plurality in New Hampshire, with 45% going for her as opposed to just 32% in Iowa. In fact, while women comprised 57% of the electorate in both Iowa and New Hampshire, they gave Clinton a whopping 46% of the vote in New Hampshire, compared with just 30% in Iowa.

Anyone can call a thousand people, ask them who they're voting for and call it a poll. As Selzer showed, it takes a lot more talent to correctly predict the turnout and the demographics of those who will show up to vote. Polling New Hampshire is easier than polling Iowa, which would explain why, media hype to the contrary, several pollsters actually did get it right.

Polls come equipped with a margin of error for a reason. Not everyone can be accurately polled, but academic statistics are such that, in nineteen out of twenty times, a sample's result will show the actual state of public opinion within that margin of error. By the end of the race, two pollsters, Research 2000 and Mason-Dixon, actually called the race correctly. Polling for the Concord Monitor, Research 2000 said Obama led by one; polling for MSNBC and McClatchy, Mason-Dixon had Obama up two. Both Obama leads were within the 4% margins of error. Not everyone, in short, got it wrong.

Sixteen polls were released either on January 6 or 7, showing results as divergent as Obama leading by one and Obama leading by 13. It is ironic that the poll-mania came just weeks -- sometimes days -- after the media reported that so many voters were undecided going into both the caucuses and the primaries. This reporter witnessed several people signing supporter cards for one candidate while openly expressing their lack of a firm decision. Voters in Iowa and New Hampshire reserve the right to change their minds, and they take advantage of that right.

In fact, it was those stories that were correct: In New Hampshire, 17% of voters decided for whom they would cast a ballot on Election Day. No polls were conducted on election day, so 17% of the electorate should have been undecided in pre-Election Day surveys. But a poll that shows a large portion of the electorate undecided is not a sexy poll -- just ask anyone who conducts polls in New Jersey, where voters are notoriously reluctant to identify their choices. Media clients want a clear picture of the race, so pollsters push for "leaners."

If a voter is undecided, towards whom are they leaning? That skews results, and at times badly. Final polls probably did not take into account last-minute developments in the race -- Clinton's tears, Obama telling Clinton she was "likable enough" in the final debate -- that either solidified support or cost someone else votes.

Those who decided on their candidate on Election Day were, in fact, most reflective of the electorate's mood. Clinton won 39% to 37% on Election night. Those who made a final decision that day broke for Clinton, 39% to 36%. That puts Mason-Dixon and Research 2000 in an even better light.

To be fair, the Clinton campaign's pollster -- the same one who criticized Selzer's poll -- must have gotten New Hampshire equally wrong. Rumors abound today that the Clinton team thought it was heading for a big defeat, and that they planned to undergo a shakeup that same night, in order to fold the news into one campaign cycle and get over the bad news.

The real culprit, many felt, was the Beltway media, which initially expected a big win for Obama. It's understandable that they did: To many eyes, including my own, Obama's crowds were bigger, lines to attend his events longer and excitement higher. But looks can be deceiving: Mitt Romney held a series of house parties in New Hampshire that drew a hundred or so people, while rallies held by Mike Huckabee drew three or four times more. No one was under the illusion that Huckabee was going to beat Romney.

The basic difference between polls and prognosticators: Predicting is different from reporting. The coverage from New Hampshire was spot-on; something happened that made women break for Clinton. Whether it was her choking up, Obama's likability comment at the debate or a couple of hecklers screaming "Iron my shirt," women clearly rallied to Clinton and away from Obama.

In a way, the incorrect predictions are encouraging. Candidates who fall behind often say that voters in Iowa and New Hampshire do not like being told what they will do. That sounds like it smacks of desperation: My numbers are better than you Washington insiders are saying they are. But New Hampshire voters did not do what the pundits said they would, offering hope that many -- perhaps more than the media thinks -- actually pay attention to the race and make an informed, educated decision.

As Dennis Kucinich and some in the blogosphere flail around for an excuse as to why something was crooked or corrupt, they should not blame pollsters, two of whom got it right while others screwed up not their mathematics but their turnout projections (Obama, as he took over the lead in many polls, wisely said that he didn't pay attention to polls when they were behind, and he wouldn't pay attention when they are ahead. If that's true, one outlier or bad sample can ruin a campaign).

The Kucinich backers should not blame reporters, who reported what was happening on the ground and, largely, why. Women had to break for some reason. Younger voters did not, in fact, show up in the record numbers they did in Iowa. And the great weather encouraged older voters to turn out, which boosted Clinton even further.

And they should not blame the pundits. Yes, the pundits got it wrong, but what's the point of actually going to vote if pundits always get it right? It is happily reassuring to know that people's votes actually do matter, and that elections should not be decided through sponsored polls, but rather at the polls. Occasionally, it takes a blunder on the part of a few in order to remind the many that their voices are the ones that matter.