Opinion polls did well
One thing that hasnt been said yet - but bears mentioning considering the scorn and ridicule that opinion polls often meet here - is that the polls did pretty well measuring the mood of the electorate in the various states.
Of course, it always remains important to look at the trend over several polls from different pollsters, rather than going on the one single poll. But those who did exactly that, could have gotten a pretty correct idea of how the races would work out.
At pollster.com, they did that - updating graphs for every Senate race, which tracked the average of the last five polls out. Yesterday, going just on the preliminary results of course, they did a quickie overview of how that average compared to the actual outcome. Here it is:
(Note that, "For brevity's sake, the table above displays the results for Joe Lieberman in the Republican column")
The average of the last five polls out had gotten the winner right in every of these 12 competitive races. What is more, it had gotten the margin of victory within 3% in all but two races. Only in Maryland (Cardin vs Steele) and in Pennsylvania (Casey vs Santorum) did the margin of victory turn out to be much larger: 10 instead of 3 points and 18 instead of 10 points, respectively.
Those two examples also highlight something else. I had been following the Pollster graphs, but my prediction of the outcome was nevertheless off by no less than 3 seats. This was because I tend to take small Dem leads in opinion polls with a grain of salt. Too often (Bush vs Kerry being a prime example), the actual result turns out to be 2-3% worse, in the difference between the two candidates, for the Dem in the race. Hence me brushing off 1-3% leads for the Dem in poll averages for Virginia, Missouri and Montana.
But as we now know, this time that didnt happen. In fact, the Dems overall did slightly
better on election night than they had done in the final polls. Their average margin in the 12 states altogether (see table) was 5 rather than 3 points. And while Cardin and Casey in particular did much better than the polls had suggested, two candidates who did 3 points better than polled were also both Democratic: Klobuchar in Minnesota and Ford in Tennessee.
All of this may have something to do with what I saw dubbed somewhere as a "late countersurge". Somewhere - but Ive tried in vain for 10 minutes to find back where - I read an analysis of exit poll numbers in Missouri, which showed that voters who made up their mind 2-5 days before the election had broken towards Republican incumbent Talent, but that those who made up their mind only on election day itself had overwhelmingly broken to Democrat McGaskill. If that happened everywhere, it explains how the Democrats did even better than in the last polls out.
(It bears mentioning that in 1994, too, the defending party - the Democrats - appeared to have something of a surge the last few days before the elections - but it didnt help them none. Same with what had looked like a last-few-days surge for the Republicans now.)