Monday, October 29, 2012

Rasmussen Now Showing Romney Winning Election

Rasmussen Reports has been a well known and well liked polling company for republicans since it started in 2003. Democrats claims they are always biased in favor of republicans while those in the GOP will tell you that they are relatively accurate and did a great job in 2008. Here's the reality...both sides are right. They do have a very good track record for accuracy in past presidential elections, but a very suspect record in other races. The problem is that they also have a long track record of republican bias which is held in greater scrutiny because Scott Rasmussen himself is a republican. The problem is that both sides have a strong tendency to over react when criticizing and praising the poll.

One of the biggest knocks against this poll in 2012 is their insistence on what is considered old methodology. They are not adopting to any of the new methods such as: calling cell phones, using an enthusiasm gap to determine voter turn out, sprinkling their state polls over a few days, etc. Could their instance on old methodology catch up with them in 2012? Perhaps, but making such a statement is rather risky until more evidence arrives. Nate Silver a professional prognosticator who was very accurate in 2008 is so sure of this assumption that he refuses to weigh their polls as he would other reputable firms. He claims that their methods have to catch up with them soon and assumes it will be in this election. I find this decision to be a major gamble that could create some obvious flaws in his numbers.

However, what everyone must remember is that they are only one poll. Whenever anyone puts too much stock in one poll it leads to terrible misconceptions. While they were one of the most accurate presidential polls in 2008 according to the ARG they also had their share of problems with republican bias. Most of these problems showed up in the important swing states and often with outlier polls. Here is a sampling of their final polls in some of the swing states four years ago and how it compared to the final results.

State Final Rasmussen Poll Final Result Bias
Colorado Obama by 4 Obama by 8.5 Republican- 4+
Florida McCain by 1 Obama by 2.5 Republican- 3+
Nevada Obama by 4 Obama by 12.4 Republican- 8+
New Hampshire Obama by 7 Obama by 9.5 Republican- 2+
Ohio TIED Obama by 4 Republican- 4
Virginia Obama by 4 Obama by 6.3 Republican- 2+
Wisconsin Obama by 7 Obama by 13.9 Republican- 6+

Of course in all fairness 2008 was a very bad year in many of these same states for many pollsters, due to underestimating democrats turn out and overestimating republican turn out. Yet, what is different about Rasmussen is that they are always showing the same bias and gives numerous outlier polls in the process. Rasmussen was far more accurate in many of the other non-swing states.

If the election were held today Rasmussen has Romney pulling out Ohio, New Hampshire, Colorado, Virginia, Florida, and tied in Wisconsin. This would be plenty enough to make Romney the victor. So why am I 75% confident that Obama would win the election if it were held today?
1.) This is one poll and I consider all reputable polls.
2.) They are the only poll showing Romney in the lead in Ohio and tied in Wisconsin.
3.) Other polls with a known republican bias still has Obama leading in key states like Ohio, New Hampshire, and Colorado.
4.) I adjust their polls one point for being republican biased unless they are an outlier in which I more heavily adjust the number.

It is not that these latest Rasmussen polls are not important, because they are...in proper context. If Romney is to win the election he will need this firm showing good numbers in states like Ohio, Colorado, and Virginia.  These polls help compensate for some of the other left-leaning polls showing Obama in front by several points. The key is found in some of the other reputable firms showing similar numbers and at this point they are not.

4 comments:

  1. Thanks for your insight. I have a question, though. What was the difference from the actual results and Realclear's average in 2008? Thanks!

    ReplyDelete
    Replies
    1. Thank you for your comments and question. Here is a link that shows state by state the average of all RCP polls compared to the final results.
      http://www.realclearpolitics.com/epolls/2008/president/battleground.html

      Delete
    2. It appears almost every poll underestimated the Democrat turn out in the last election. Could we see a repeat of that or do you think that they will underestimate the Republican turn out this time over-correcting for the last election?

      Delete
    3. I am relatively convinced as are at least half of the polling firms that we are looking at a dramatically different turnout this time around. One of the reasons Gallup and Rasmussen has had different numbers then most of the other firms is this very assumption. However, what is rather odd is that even companies assuming very high republican turnout are showing pretty similar numbers to those that are not. I will have a post up tomorrow on the voter turnout statistics and how this could impact the election. However, historically in extremely close state races with a lot of varying polls, a vastly different voter turnout then anticipated will move the numbers around two points maximum. That is why their is a huge difference between Obama being up 3 points and Obama only being up 1-2 points.

      Delete