By Mark Rhoads -
Twelve years ago back in 2000, I was one of those people who was very enthusiastic when options trader John McIntyre and advertising exccutive Tom Bevan first started the Real Clear Politics web site in Chicago. They filled an important need and they were willing to give the conservative point of view a fair shake which was a novelty for that time among political blogs. In the years since the RCP site has become a huge success and expanded to markets and other topics so that even MSM reporters consult it frequently.
Even today, after Steve Forbes bought a majority share, but kept McIntyre and Bevans as managers, while some RCP writers have a more liberal perspecive than others do, on the whole the site still gives a fair break to conservatives. But unfortunately the RCP Average of polls has aquired an imaginary gravitas with media pundits that it simply does not deserve. An average of junk is still just junk.
The RCP average tracks 8 national polling organizations with wildly different reputations, different weighted sampling methods by partisan leanings, and taken on different dates with different sample sizes with different screens for likely voters. Two of the national polling groups, Rasmussen and Gallup, have fairly good track records over many years. But the rest that include cover brand names from news groups that contract the actual poll out to research groups of various reputations, are far less reliable than the first two and they all release results at different times for snap shots or sometimes rolling averages over 3 to 7 days. An average of such wildly different polling reports is not of very much practical value but nevertheless readers are misled into thinking that the top number is an honest average of many polls with similar samples and methods and the readers therefore think the average is a better number than individual polls offer. It really is not and RCP does not offer the caveats it should offer.
The fact that Mitt Romney has gained a lot of ground in many polls since the first debate on Oct. 3 does not necessarily mean the sample methods are any less flawed than they were when Obama was leading on the national level. If you read carefully enough regarding the sample partisan weighting methods, you still often see a heavy tilt to Democrats based on 2008 voting patterns and that year was not a typical year. If that is correct, it would mean that Romney is doing even better than some polls indicate that still use a 2008 turnout model. When Gallup switched from reporting only registered voters to a likely voter screen, Romney's numbers went up in that poll but it was hard to isolate that factor since his general approval numbers were also going up after the first debate. While a likely voter screen is a very good screen to have to make a poll more accurate, it is hard to further refine the likely voters as to the degree of their intensity which also differs greatly among groups and individuals. In general, the Tea Party and conservative voters had the most intensity to turn out in the 2010 congressional elections and that helped to drive the Tea Party victories in many House races. It took a very long time this year for Mitt Romney to finally close the deal with conservative voters who wanted to be for him and not only just anti-Obama but I think the Oct. 3 debate did seal that deal for the most part. The Democratic National Covention and the early part of September also reminded conservatives us of just how hostile Obama is to the concepts of personal freedom and responsibility vs. socialistic paternalism by government. The anti-Democratic intensity factor is also up among conservative GOP voters--maybe not as high as the 2010 levels--but getting there fast. In general, the national and state by state polling trend lines are all in the direction of Romney since Oct. 3. That is all good news to date. But beware of attaching too much importance to the RCP average of polls number because it does not deserve the disproportionate attention it gets. Gallup and Rasmussen are better national indicators and Rasmussen state polls also do have a good reputation in general.
Maybe another way to make the point I want to make is just to simply say that an average of flawed polls is no less flawed just because it is an average.
