The Polls Say The Democrats Will Control The Senate And They Are Probably Right

I've written about polls before and how much I dislike them. Their track record for presidential primaries isn't very good at all and I think presidential campaigns, the public and the news services get too caught up in the data. They try to see trends in time and space that aren't there. But I have to give credit where credit is due, and while the success of polls can be scattershot, the fact is that polls for US Senate races have - at least in the aggregate - done a very good job of predicting winners over the last two elections.

In 2006 and 2008 combined, there were 26 senatorial races where significant polling took place before the election. Using the averages of those polls to predict winners resulted in only one off pick, the 2008 Minnesota race. Now a good deal of the success of the polls was the result of many races never being close in the first place. Only 10 of those races had an average poll prediction with less than a seven point margin of victory. The polls were successful in picking the winner in three out of the four races with a predicted margin of victory less than four percent. That's not enough of a sample to say how good polls are at picking the close battles. My guess is that poll stated error bounds in predictions are too small.

If you look at all of the races, you find that the predicted margin of victory is often significantly off, and there is a strong tendency for polls to under-predict victory margins by a few percentage points of more. The biggest errors tend to be associated with races that were predicted to be blowouts.

What does this mean for the 2010 Senate? If poll results can be believed - and I think they can be believed - there are seven very tight races this year. For five of them at this point in time the predicted margin of victory is less than any reasonable expected error in polling, 2.6 percent. Four of them, if the averages are to be believed, lean slightly Republican. One leans Democratic. The other two races that are tight lean Democratic in a way that's close to the bounds of expected error.

In a nutshell, for the Republicans to win the US Senate, they are going to have to win all four of the Republican leaners - Colorado, Nevada, Pennsylvania, and Illinois - and the one Democratic leaner, Washington. All five races are statistical dead heats, so that's the equivalent of tossing heads five times in a row. Then the Republicans are going to have to win one out of two states, California and West Virginia, where polls so far show a margin of victory for Democrats that's in the three to five percent range. What are the chances of all this happening? Slim and none. The Democrats will hold the Senate. The House is another matter entirely.

I note that I recently gave some money to Russ Feingold, the Democratic candidate for Senate in Wisconsin. The polls heretofore show him to have a snowball's chance in hell of winning. Of particular note is a Daily Kos Democratic poll that shows that a majority of the electorate disapproves of his performance.

Have I thrown my money away? Maybe. But this is a race where I think the polls may be misjudging the electorate. The Green Bay Press Gazette - a conservative newspaper - finds Feingold's opponent Ron Johnson to be so much of an empty suit that even they can't support him. They state, "We think that with time, Johnson could become a viable candidate for national office...." Ouch.

There is one week to go. The nationwide Tea Party fervor ginned up by the press seems to be waning. Voters seem to be moving from being mad as hell to just being resigned and tired. I'm half thinking and half hoping that my money was well spent. But maybe this is just one of those elections where stupid wins and does so with ease.