(Published Oct. 22, 2008)
In a recent interview on the Today show, Sen. Barack Obama predicted his lead in the polls would shrink, because that's what happens toward the end of presidential campaigns.
So let's start with some context from Obama's exchange with Matt Lauer:
"Do you see a path to victory for John McCain?" Lauer asked.
"There's no doubt," Obama replied. "I mean, I – we think that the race will tighten, just because that's what happens at the end of campaigns. They always have. They even – when there are substantial leads. And in each of these battleground states, you've got a lot of close races. One of the messages that I've had to my team is that we don't let up. We do not let up."
The interview was taped Oct. 19, 2008, which was 16 days before the election. We looked at every election since 1944 to see if poll averages in early November were tighter than they had been at any time in the prior few weeks, a loose time frame we decided was an appropriate interpretation of Obama's somewhat vague claim that races get tighter at the "end of campaigns."
Fortunately, a pair of political scientists, Christopher Wlezien from Temple University and Robert Erikson from Columbia University, had done the hard part. They compiled averages of the incumbent-party candidate's standing – in a hypothetical match-up of the two major-party candidates – in all major polls in the run-up to the past 16 elections, broken down by day. They've used the data in several published papers, and several other political scientists told us it was the best available on the subject. Because these numbers reflect an average of several polls, even small changes in the average reflect more significant changes in the underlying polls.
Their spreadsheet enabled us to see, for instance, that in 1972, Richard Nixon was polling at 67.8 percent 17 days before the election. Just prior to Election Day, however, he was polling at 63.6 percent. So yes, that election did tighten at the end. (Nixon won with 60.7 percent of the vote, by the way.)
Some tightened far less. In 1984, Ronald Reagan was leading Walter Mondale with more than 61 percent in the average of all polls taken 11, 12 and 13 days before the election. Just prior to Election Day, his lead had shrunk to 60.2 percent. (Reagan won with 58.8 percent.) So that race tightened less – but still, it tightened. (Slight changes in polling averages are less significant the further back in time you go, because there were fewer polls, but we gave Obama the benefit of the doubt if the polling averages reflected even a minor tightening.)
Not every election got closer, though. In 1956, Dwight D. Eisenhower never topped a 59 percent average in polls until about the last week before the election. He peaked in early November at 59.4 percent.
The 1960 election didn't get any tighter at the end either. But every other one did.
That is, out of the past 16 elections, 14 tightened at least somewhat in the last few weeks.
A commonly cited explanation for this, Wlezien said, is that voters who are undecided toward the end of a campaign are not likely to feel strongly about either candidate, and thus are likely to split 50-50 between them. That pushes the race in the direction of a tie.
While we were at it, we looked in more detail at the elections since 1980, to see how the final results compared to the last few polls. Six of those seven elections have been tighter than the polls in the final few days indicated they would be.
That said, the candidate who was ahead in most polls two to three weeks before the election ended up winning every time. (Gore briefly polled better than Bush in 2000, but that was for just one out of the last 21 days. Similarly, Carter polled better than Reagan in 1980, but for only three out of the last 21 days.)
But we're not in the prediction business, we're in the fact business. And the fact is, 14 out of the last 16 elections have tightened in the final few weeks of the campaign. So Obama's claim that elections tighten at the end is Mostly True.