Joe Miller is not yet assured of victory over Senator Lisa Murkowski of Alaska: With just 1,668 votes separating the candidates, the official result may not be known until September when at least 7,600 absentee ballots are counted. But a victory for Mr. Miller is considerably more likely than not, and even if he were to fall a few votes short, the election will have been far closer than anyone anticipated.
But a closer look at the race finds that there were signals that Mr. Miller had a tangible possibility of achieving the upset.
Start with the polling, of which there wasn’t much. The only public poll of the race, conducted in early July by Ivan Moore Research, had Ms. Murkowski leading Mr. Miller, 62 percent to 30 percent, among likely Republican primary voters. But that same poll also found that only 46 percent of the state’s registered voters knew who Mr. Miller was at that point — and only 31 percent had developed a definitive impression, positive or negative, about him. In contrast, 98 percent of registered voters knew Ms. Murkowski’s name, and 82 percent had formed an impression of her.
These name recognition numbers were among all registered voters in Alaska; it’s possible that Alaskans inclined to vote in the Republican primary were more likely to be familiar with Mr. Miller by then. But suppose that, by early July when this survey was conducted, only 50 percent of Republican primary voters had learned enough about Mr. Miller to have an impression of him. Nevertheless, 30 percent of Republican primary voters (whether they were familiar with Mr. Miller or not) had already decided that they preferred him to Ms. Murkowski. That is hardly such a bad result for him: it seems likely that, of Republican voters who were familiar with both candidates, Mr. Miller may already have been running about even with Ms. Murkowski.
These “early adopters” of Mr. Miller, therefore, may have foretold of the pattern that would take hold once more Republican voters became familiar with his candidacy. And – in contrast to many insurgent candidates — Mr. Miller had plenty of opportunity to ensure that voters would indeed get to know him by the time Tuesday’s primary rolled around. In particular, the Tea Party Express purchased more than $500,000 worth of advertising on Mr. Miller’s behalf. That would be a fairly large sum for any senate primary. But it’s a huge amount of money in Alaska, a low-population state in which only about 100,000 votes will ultimately be counted in the Republican primary (on a per-voter basis, Mr. Miller’s $500,000 was the equivalent of an $8.5 million ad buy in the Republican primary in California).
Here, then, is the formula by which we might detect in advance a win as “shocking” as Mr. Miller’s. If a candidate has low name recognition, but is running relatively well among voters who have become familiar with him or her, and has a strategy – like the ability to spend lots of money — for increasing name recognition, then we should not be so surprised if that candidate performs much better than expected on Election Day.
This might seem an obscure set of circumstances, but it is hardly so uncommon in Senate and gubernatorial primaries. In Pennsylvania’s Senate Democratic primary in May, for instance, Representative Joe Sestak had been running as many as 20 points behind Senator Arlen Specter in polls conducted just two months before the primary. But the same polls showed that Mr. Sestak was running even or ahead of Mr. Specter among voters who were familiar with both candidates. Mr. Sestak conserved his advertising dollars and surged late in the race, eventually beating Specter by 8 percentage points.
The first Quinnipiac poll conducted in the Florida Senate primary, meanwhile, found Marco Rubio trailing Gov. Charlie Crist, 54 percent to 23 percent. But as Mr. Rubio’s name recognition increased, the percentage of Republican voters expressing a preference for him over Mr. Crist increased proportionately, until he eventually took such a dominant position in the polling that Mr. Crist was forced to depart the Republican primary and run instead as an independent.
The only difference between a case like Pennsylvania and one like Alaska is that, in Pennsylvania there was an abundance of polling to examine – the race was surveyed about 15 times in the final six weeks of the campaign. Thus, we were able to see Mr. Sestak’s surge as it happened. In Alaska, on the other hand, the six-week-old Ivan Moore Research poll was the last one we had.
As I remarked the other day, a six-week-old primary poll is about as useful as a six-week-old tuna fish sandwich – particularly when the voters aren’t equally familiar with the two candidates.