Search This Blog

Saturday, September 15, 2012

Strange Polls, Bizarre Polls and Weird Polls

As any reader of this blog knows, I study the election polls that are constantly being released. This week we have seen a wave of new polling data that will build in the next few weeks into a tsunami of polls that will continue until election day. The numbers in the recent polls are all over the lot. There are, however, a few things that are clear from the recent data:

1) The bounce that Romney got from the GOP convention is long gone. He briefly moved up about 5% in the tracking polls from where he began, but by the time the Democrats moved into high gear at their convention, the Romney bounce disappeared.

2) The bounce that Obama got from the Democrat convention is also pretty much gone. It seems to have disappeared by September 12th, the Wednesday after the end of the convention. Thus, the Rasmussen tracking poll saw Romney move back into the lead already. The Gallup tracking poll which includes the average of seven days of polling is also coming back to parity. Today, Obama is up by 4%, yesterday it was 5%, the day before 6% and the day before 7%. The latest results (4%) include four days while the bounce was still in Obama's numbers and three after it was totally gone. Most likely, we will see the Gallup figures come back to a lead no greater than 1% for one candidate or the other. Many of the other national polls that have been released show larger margins for Obama, but these are polls that were taken before the bounce dissipated.

3) A high percentage of polls seem to be rigged. That may seem like an outrageous charge, but there is no other possible conclusion. We have seen poll after poll that somehow sample too many Democrats. Let me put it this way: Rasmussen has polled about 15,000 people per month so far this year. Each one is asked to self idendify by party. The results of close to 130,000 interviews have been that the numbers of Republicans and Democrats are very nearly the same. In the last month, Republicans are ahead of Democrats by just over 1%. So, a random poll of Americans ought to produce a sample that has about the same number of Democrats and Republicans. Nevertheless, in the last two months, we have seen national polls that have had 14% more Democrats than Republicans and many others that came close to that type of lead for the Democrats in party identification. In all that time, I have yet to see any national poll that over samples Republicans. As a result, Obama has done better in most polls that he ought to do.

Now, if the higher number of Democrats in some polls were just a statistical coincidence, there ought to be an equal number of polls with too many Republicans. The idea that there are literally dozens of polls that over sample Democrats and none that over sample Republicans means one of three things: a)the polls are rigged; or b)the Rasmussen poll of 130,000 people chosen at random is somehow way off; or c)the Republicans refuse to answer questions from the pollsters. Given those choices, it is not hard to settle on choice "a", namely that some of the polls are rigged. I have written many times about the phony polls from Public Policy Polling which seem to show a lead for the Democrats in state races where every other pollster shows a lead for the Republican. But even beyond PPP, there are polls that can be explained on if one accepts that the results are rigged.

4)We are getting close to the point where the polls will have to become accurate. Each of the pollsters wants very much to come as close as possible to the results on election day. It is the only thing that preserves their ability to sell their polls in the future. that means that the rigging of polls has to be reduced as we move into October. It will be interesting to see what happens as actual reality is injected in the polling results.

By the way, I want to make clear that I do not believe that all polls are being rigged. Some pollsters like Gallup and Rasmussen seem to be trying for accuracy rather than to influence the outcome of the elections. Others, however, cannot be so described.


No comments: