This is Part II of my new series, Grading the Pollsters. Part I (Mason-Dixon) is here:
http://www.dailykos.com/...
Rasmussen's national tracking poll finished at 52-46, which is very close to the current national popular vote margin (53-46). The Rasmussen national tracking poll was generally consistent and except for a few blips was steadier than most.
However, when it came to the state polls, Rasmussen had a less than stellar night, which leads to a head scratching moment: how could Rasmussen be so accurate in the national poll and have state polls not match or accurately reflect that trend? Let us take a look below the flip:
1. Florida: Actual result: Obama 51 McCain 49. Final Ras poll: McCain 50 Obama 49.
Rasmussen blew it in Florida. Throughout the election cycle, their numbers in Florida were far more stingy than other pollsters. They did not pick up the swings in the I-4 to Obama, or in the Sarasota region, nor did they pick up the increased voter participation and margins in the Dem areas. McCain never had 50% of the vote in this state in most polling by other organizations. It seems that the Rasmussen formula understated Obama's support among suburban voters and Democrats, and overstated McCain's overall strength.
2. North Carolina: Actual result: Obama 50 McCain 49. Final Ras poll: McCain 50 Obama 49.
Rasmussen got this one wrong too. Admittedly, NC was the second closest race in the country and this really could have gone either way. PPP, however, got this one right and seemed to have a better read on NC than anyone.
3. Ohio: Actual result: Obama 51+ McCain 47. Final Ras poll: Obama 49 McCain 49
Ras blew it in Ohio too. They were often very stingy to Obama throughout the election cycle. Several other polls had Obama consistently hitting the 50% mark, but Ras never did. They clearly failed to notice the Obama trends in Cincinatti, Dayton, Columbus, NE Ohio and the several red counties where Obama improved on Kerry's 2004 performance.
4. Virginia: Actual result: Obama 53 McCain 47. Final Ras poll: Obama 51 McCain 47.
Rasmussen's result is very close to the final tally, though it understated Obama's margin by 2 points, but they did seem to accurately capture the mood and trends of the state. McCain did a few points better than other polls had suggested, primarily through outperforming expectations in rural VA. Rasmussen, however, did capture this support while acknowledging Obama's strength in NOVA, Southeast VA and the exurb counties. One of the great things of Obama's victory is that he did it without carrying much of rural america in the battlegrounds, especially southern battlegrounds. He got there by expanding the Democratic electorate in the cities and suburbs and making incremental improvements in some red counties. VA is the perfect example of that strategy. Plouffe is now officially a political genius.
5. Indiana: Actual result: Obama 50 McCain 49. Final Ras poll: McCain 49 Obama 46.
Ras blew another one here. This was the third closest race in the country, but several other polls, including PPP and Selzer, pointed out that Obama was doing well enough in rural and southern Indiana to pull off a win with a big turnout among African American voters in the Gary region. That is exactly what happened. Rasmussen did not provide this kind of analysis, and underestimated Obama's true level of support.
6. Missouri: Actual Result: McCain 50 Obama 49. Final Ras poll: McCain 49 Obama 49.
Ras pushed on Missouri, and it was the closest race in the country. They were accurate in the sense that the race was very close and the top line numbers were pretty close to the actual result. They didn't pick a horse in this race, however.
Two points to make about Missouri: First, I don't want to hear anyone ever say again that it is a bellweather state. They haven't been a bellweather in the last 3 elections. They have gone against the popular vote twice, and in 2004, Bush's margin of victory exceeded his national popular vote margin by nearly double. Second, I hate Ralph Nader. He took 18,000 votes in the state and probably cost Obama a chance to win.
7. Nevada: Actual Result: Obama 55 McCain 43. Final Ras poll: Obama 50 McCain 46.
The result in Nevada was a very pleasant surprise and proved that Obama's push for early voting was no gimmick. We dominated the early voting and it allowed us more room to bring more total voters to the polls on election day.
As for Ras, he picked the correct winner, but the numbers suggested a much closer race than it actually was. Ras missed Obama's number by 5 points and overstated McCain's by 2 points. There is a very big difference between a 4 point result and a 12 point result, so one can't say that Rasmussen's polling was particularly accurate.
8. Colorado: Actual Result: Obama 53 McCain 45 (and still counting Boulder, so it could be higher). Final Ras poll: Obama 51 McCain 47.
Ras missed the margin of victory by a significant 4 points, understating Obama's margin by 2 points and overstating McCain's by 3 points. Votes are still being counted in Boulder County (going for Obama 3:1) so it is quite possible that the margin will creep closer to 9 points. There were several polls that showed CO strongly in Obama's column. Ras was not one of those pollsters. This was one of the easiest calls of the night for the media. It was not a close race and Ras' poll showed a GOP house lean that wasn't supported by facts on the ground or in other credible polling.
9. Pennsylvania: Actual result: Obama 55 McCain 44. Final Ras poll: Obama 52 McCain 46.
Ras played a lot of games with its PA polling in the last few weeks, issuing a poll 5 days before the election stating that the margin had shrunk to 4 before issuing this final poll with a 6 point margin. In either case, Ras' result was off, understating Obama's percentage by 3 points and overstating McCain by 2 points. PA was one of the easiest calls of the night for the Media, and was called as soon as polls closed by MSNBC and other organizations. Clearly something was off in Rasmussen's polling. In short, Rasmussen did not provide a truly accurate picture of the PA electorate.
10. New Mexico: Actual Result: Obama 57 McCain 42. Final Ras poll: Obama 54 McCain 44.
Ras' polling did indicate a double digit margin, but again demonstrated a GOP house lean as he overstated McCain's number by 2 points and understated Obama's percentage by 3 points. Rasmussen had a clear GOP house lean in its polling that seemed consistent in state after state.
11. Wisconsin: Actual result: Obama 56 McCain 43. Final Ras poll: Obama 51 McCain 44.
Another example of the GOP House lean. Wisconsin was one of the largest margins Obama had in a 2004 battleground state.
12. Iowa: Actual result: Obama 54 McCain 45. Final Ras Poll: IA: Obama 52 McCain 44.
Rasmussen outperformed other polling in Iowa, which had suggested a double digit victory for Obama. Ras missed the margin by 1 point and had the usual house lean in Obama's numbers but was much closer than even the Selzer poll. Obama was clearly willing to sacrifice a few points in Iowa to go after Indiana and that decision paid off big time.
13. New Hampshire: Actual result: Obama 54 McCain 45. Final Ras poll: Obama 51 McCain 44.
Ras' result was much closer to the actual result, but again demonstrated a GOP house lean by understating Obama's total by 3 points. Obama's margin of victory in NH is the highest for a Democrat in many years. He won every county in the state. Not bad for a black guy campaigning in an all white state.
14. Montana: Actual Result: McCain 50 Obama 46. Final Ras poll: McCain 50 Obama 46.
Ras pretty much nailed Montana, and also seemed to get it right in the Dakotas.
15. Georgia: Actual "Result": McCain 52 Obama 47. Final Ras poll: McCain 52 Obama 47..
I don't trust the GA secretary of state, and this margin could narrow if all the votes get counted, but for the moment it looks like Ras got this one right.
Overall, Rasmussen had a pretty bad night in terms of the key battlegrounds, missing the result in Ohio, Florida, Indiana and North Carolina. Rasmussen also understated the Obama margin of victory in several states, including CO NM, NV, PA. Rasmussen also exhibited a pronounced GOP house lean, consistently understating Obama's strength of support in several states while overstating McCain's support in several. They seemed to better in red states and were less reliable in blue and purple states.
The Ras state polls did not match up with its national poll, and that is puzzling as I don't believe there is any example of a candidate winning the popular vote by over 5% and not winning over 300 EVs. Nate Silver sketched out these mathematical probabilities on 538, and Rasmussen's numbers did not match up very well to the likely outcomes. It seems to me that Ras might have assumed that undecideds would disproportionately go towards McCain (Bradley Effect factoring perhaps?), but that should have also resulted in a popular vote adjustment to around 3 points. That didn't happen, and it raises some questions about Rasmussen's methodology.
In short, Rasmussen was not a reliable indicator of the underlying trends in many battleground states. Averaging the grade for their national poll (A -) and their state polling (D), I think at best Rasmussen gets a grade of C - for this election cycle.
Update: Thanks for all the comments. Some think the grade of C- is too harsh. I could raise it to a C, but to miss 4 results among the battlegrounds and show a consistent house lean that didn't match up well to the national poll suggests less than stellar and perhaps less than average performance.