A New York Times article suggests that New Hampshire voters are capricious types who really make their choice in the last few days. Any evidence to support this in a recent general election?

Let's take a stroll through the archives of the University of New Hampshire Survey Center, which displays polls for the last four elections:

2000

My colleague Andy Smith did not do tracking polls that year, but his last poll, which left the field two days before the election, had George W. Bush up six points over Al Gore, 45 percent to 39 percent, with Ralph Nader at 5 percent and undecideds at 6 percent. Bush won the election 48-47-4.

Jeanne Shaheen, John Sununu and Charlie Bass were all ahead in this final poll, and they all won their races.

2002

The closest race that year, of course, was the first matchup between Jeanne Shaheen and John Sununu for U. S. Senate. The UNH final poll that year left the field on November 2, three days before the election. It showed Sununu up just one percentage point, 46-45. On Election Day, Sununu won 51-47.

The final UNH poll also had Craig Benson, Jeb Bradley and Charlie Bass ahead in their races, and they all won easily.

2004

This graph of the UNH tracking poll pretty much says it all.

John Kerry kept a narrow lead over George W. Bush throughout the tracking poll, and he won narrowly, 50-49.

Governor Benson was up 1 point, 46-45, in the last tracking poll report. Smith called this race too close to call. John Lynch won this race 51-49.

Races for Congress were all lopsided that year.

2006

This election provides the only example in recent history of "late deciders" reversing a poll that was not "too close to call." That was the race in the First Congressional District. The final UNH tracking poll, which left the field two days before the election, showed Congressman Jeb Bradley ahead 49-40. Shea-Porter won the election 51-49. (In retrospect, a look at the graph indicates that the poll may have picked up a narrowing race in the final day in the field.)

In the other major competitive race that year, UNH had Paul Hodes up 48-39 after the final day of polling. Hodes won 53-46 on Election Day.

To conclude:

Yes, we all know that Obama blew a big lead in this state not so long ago. (Believe me, this dumb pundit won't forget anytime soon.) But we also all know that primary elections are much more volatile than general-election contests because political party identification, of course, does not offer any cues to voters in primaries. But people tend to vote their party in general elections, and this one is no exception.

There is no evidence that large numbers of late-deciding Granite Staters changed the outcome of either of the last two presidential races, both of which were closer than this one is. Indeed, Obama is over the 50 percent mark, according to Pollster.com, so even if every single undecided voter out there moves to John McCain, that would not be enough. There is evidence in the current UNH tracking poll that Republicans are "coming home" to John McCain. That won't be enough, however, with independents favoring Obama by a 2-to-1 margin.

And when one looks to past races downticket for the last four elections, there is exactly one example of late-deciders creating a surprise ending: Carol Shea-Porter in 2006.

2006, of course, was a huge Democratic year, and Shea-Porter had national gale-force winds at her back. Does anyone detect such winds at McCain's back as he heads to Peterborough tomorrow?