Tuesday, October 23, 2012

Could Romney Win Popular Vote But Lose Election?

One of the peculiar things about this election is the seeming discrepancy between popular vote polling and state polling. Today's batch of polls especially was amazing to behold considering how they just do not match the state of the race and the state of the race according to their own polls. Take for instance a very reputable firm that has shown incredible accuracy in the past...Survey USA. Their most recent popular vote poll shows Romney by 3. Yet, this same firm shows Obama winning Ohio by 3 and the only recent poll showing Obama ahead in Florida by a narrow one point margin. If this was a problem with just one or two polls then we would have reason to be skeptical, but it is across the board. Rasmussen is now showing Romney by 4 in the popular vote and still losing Ohio by one point. It is practically impossible for Romney to be ahead by 3-4 points nationally and be losing Ohio, much less Florida. Could the national polls be right. There are many theories...

1.) The consensus of national polling could be right, which would mean something rather odd. The average of all the polls show approximately a one point lead for Romney and their have been occasions in which the deciding state which placed the candidate at the needed 270 EV was up to 2 points off from the popular vote margin. So in this election the deciding state that places either candidate in the winning column will likely mean Ohio. This means that their are numerous occasions in which this differential is up to two points. Recently we have been noticing this gap almost widening. In 2008 Colorado was the state that pushed Obama over the top and he won it by 8.5 but only won the popular vote by 7.3. Rarely has the discrepancy coincided with an election that was extremely close and as a result it has not mattered until 2000 that is. If you adjusted the national average by a mere two points then suddenly we have a case in which Obama is narrowly winning the popular vote and also winning Ohio by basically the same margin. In short Romney could be up by 1 point and end up losing the election due to the closeness of so many states and the narrow margin that Obama is holding onto in Ohio.

2.) That the state polls are biased but the popular vote polls are not. -OR- The Popular vote polls are biased, but the state polls are not. What is amazing is that both democrats and republicans seems to like this theory. Democrats point to the fact that many of the popular vote polls that are published have a known, proven republican bias. Republicans (Because they are the ones on the short end of the stick in the state polling) suddenly declare that the state polls must be biased. Their are a few major problems with this theory. First, how can the same poll be considered accurate in the popular vote, but biased in the state polls. Second, there are those terrible things called facts and track records. Historically, popular vote polls are not only more inaccurate then state polling, but more biased as well. Look at the chart below...

Two things strike me about this chart...
1.) The incredible accuracy of polling averages. They have been less then one point off from the end result on average over the past 10 elections.
2.) Over the past 10 elections that there has been a democrat bias 5 times and a republican bias 4 times. -AND- The bias has been less then one percent on average. Yet, with as good as these polls have been let us consider the same chart, but only with state polling averages. This depicts the third theory and one that I lean towards being the most reasonable and that is the state polling this election cycle will once again be the most accurate of the two types

No comments:

Post a Comment