Around the intertubes from the time President Obama was sworn into office there was something that felt off about Rasmussen Reports. They had earned, and well earned IMO, a reputation for dead on balls accuracy in their polls nailing the 2006 and 2008 cycles. Few pollsters performed better. That all went out the windows as it seems from their first week every single poll they released was hostile to Democrats and out of step with the vast majority of other independent pollsters. For those that worship the ground Rassmussen walks on(Hi, Nova). I'd like to see what you think about this. I mean, how do you miss a poll by 40 points a week before the election in the favor of the Republican candidate unless that's what you want. Courtesy of Nate Silver at FiveThirtyEight:
One bad cycle is not enough to damn them. I would watch their performance in 2011 and 2012. If they remain error ridden then I'd drop them.
Rasmussen seems to have a systematic problem with it's polling methodology which, intentionally or not, leads them to poll more Republican voters then Democrats or independents or at least misjudge who is a "likely voter". I have read in some publications, but don't know if it is true, that Rasmussen still relies almost exclusively on land lines to select people to poll. If true then that would mean they're missing the 30% or so of people 35 and under who only use cell phones. Such an error would indeed skew their results towards older and more conservative people taking part in their polling by several percentage points.
Where's Gallup? I seem to recall they had some wild swings in their polling leading up to the election.
I presume it's a modeling error. Most poll results are driven by the assumptions made about the makeup of the electorate. I'll wait to hear something from the company themselves about what THEY think went wrong. also, I see a couple mentions about being off by 40 points but the specific incident is not cited or described, I find that astonishingly difficult to believe.
Other than two of them, the deviation was so small as to be nothing but a incidental deviance. It's is statistically unlikely that so many would tilt the same way, but statistically unlikely things occasionally happen - in any case, the degree of tilt is so small as to make it inconsequential (assuring the chart is correct for the sake of your question)
Hawaii Senate.http://www.rasmussenreports.com/pub..._elections/hawaii/election_2010_hawaii_senate Taken Oct 17, the only poll that showed it under 30 points. Rasmussen had it at 13. Inouye won by 50.
Being much more wrong that other independent pollsters and almost always skewing it to be favorable to the GOP is not small. And I have a hard time believing it was an accident for such a veteran pollster. Also, the "model" being off is a crock, Rasmussen was pulling out favorable GOP polls in the guise of Obama favorablity numbers less than a week into his presidency.
I do agree that Rasmussen has consistently been skewed to the right on... well... everything (thus their popularity among right wing posters on many sites) but I'm going to give them the benefit of the doubt right now and just guess their methodologies are bad.
you misunderstand. I was replying to an observation that all but one of the pollsters on the chart erred towards the R My reply was that of all the companies on the chart, only two deviated by more than .1% and most were under 1% and, as such, it was probably a fluke that almost all of them leaned slightly R Has anyone done a calculation that adjusted the Rasmussen divance without that one Hawaii race? IF you were going to purposely cook the book to help the GOP, it seems wildly unlikely that THAT would be the race you'd pull the stops out on - that race alone is an indicator to me that you are not dealing with an intentional manipulation of the numbers (albeit, the model might be intentionally skewed) In any case, a 37 point deviance is not something anyone tries to sneak by intentionally, it's an indication of something going wildly wrong. but i digress - what i intended to ask was this - what would the Rasmussen deviance look like over all races EXCEPT the wild outlier? if there were 105 polls and the average error was 5.8, then the sum of the errors was 609...minus 37 is 572 divided by 104 is 5.5 so if i understand the methodology here that one race alone takes you to 3.5 This moves me to wonder whether there are not another 2 or 3 wildly wrong polls which drive their average away from the pack. Finally, i note that the average deviation on that list is 4.3 - the best result there was +1, the worst result was -1.6 (1.2 without the Hawaii outlier) while I'm troubled the company I respect was the worst performer in this cycle, being less than 2 points off the average is not massive.
For those answers you will have to wait a few weeks for Nate to do his more precise pollster rankings where it takes into account what you are talking about. The list I speak of forthcoming is the one that found that Reasearch 2000 was a fraud.
From Rasmussen's charts of their own results: http://www.rasmussenreports.com/pub..._rasmussen_poll_results_2010_senate_elections Senate State - GOP actual percentage - final poll GOP percentage AL - 65 - 58 AK - 34 - 35 AZ - 59 - 52 AR - 58 - 55 CA - 46 - 43 CO - 48 - 47 CT - 46 - 44 DE - 40 - 40 FL - 49 - 50 GA - 59 - 58 HI - 22 - 40 ID - 71 - 63 IL - 48 - 46 IN - 56 - 52 IA - 65 - 55 KS - 70 - 61 KY - 56 - 53 LA - 57 - 54 MD - 36 - 38 MO - 54 - 52 NV - 45 - 49 NH - 60 - 56 NY - 33 - 31 NY - 37 - 33 NC - 55 - 52 ND - 76 - 72 OH - 57 - 57 OK - 71 - 68 OR - 39 - 42 PA - 51 - 50 SC - 62 - 58 UT - 62 - 61 VT - 31 - 32 WA - 49 - 48 WV - 43 - 46 WI - 52 - 53 36 races Races in which Republicans polled more than one point lower than the final result (i.e varied more than one point against GOP) - 20 Racesin which Republicans polled more than one point higher than the final result (varied more than one point in favor of GOP) - 5 Hmmmm.... Governors - same system AL - 58 - 55 AK - 58 - 52 AZ - 55 - 53 AR - 33 - 38 CA - 41 - 45 CO - 11 - 5 CT - 50 - 48 FL - 49 - 48 GA - 53 - 49 HI - 41 - 47 ID - 59 - 52 IL - 46 - 45 IA - 53 - 55 KS - 57 - 53 ME - 38 - 40 MD - 42 - 42 MA - 42 - 44 MI - 58 - 54 MT - 43 - 41 NE - 74 - 66 NV - 53 - 58 NH - 45 - 45 NM - 54 - 52 NY - 34 - 37 OH - 49 - 48 OK - 60 - 60 OR - 48 - 49 PA - 55 - 52 RI - 34 - 25 SC - 51 - 47 SD - 62 - 55 TN - 65 - 59 TX - 55 - 51 UT - 64 - 66 VT - 48 - 45 WI - 52 - 52 WY - 72 - 61 37 races Races in which Republicans polled more than one point lower than the final result (i.e varied more than one point against GOP) - 20 Racesin which Republicans polled more than one point higher than the final result (varied more than one point in favor of GOP) - 9 And one point is WAY under the standard margin of error. if we assign a margin of error of, say, 3 points (most polls are higher) then the Senate totals go from 20-5 to 9-1 (<that Hawaii outlier) and the Governor races go to 13-2 (<neither race ever in any real doubt) So...ya know...while I might worry about the spread in some cases, i sure as shit don't see any big tilt to the Republicans in those races. There's no chart on the House races but I'm not going to lay awake tonight worrying that they applied different practices in the House polling than they did elsewhere. My conclusion - someone is looking for a boogey man where none exists.
