President (To Win Colorado) See Full Big Line

(D) Kamala Harris

(R) Donald Trump

80%↑

20%

CO-01 (Denver) See Full Big Line

(D) Diana DeGette*

(R) V. Archuleta

98%

2%

CO-02 (Boulder-ish) See Full Big Line

(D) Joe Neguse*

(R) Marshall Dawson

95%

5%

CO-03 (West & Southern CO) See Full Big Line

(R) Jeff Hurd

(D) Adam Frisch

50%

50%

CO-04 (Northeast-ish Colorado) See Full Big Line

(R) Lauren Boebert

(D) Trisha Calvarese

90%

10%

CO-05 (Colorado Springs) See Full Big Line

(R) Jeff Crank

(D) River Gassen

80%

20%

CO-06 (Aurora) See Full Big Line

(D) Jason Crow*

(R) John Fabbricatore

90%

10%

CO-07 (Jefferson County) See Full Big Line

(D) B. Pettersen

(R) Sergei Matveyuk

90%

10%

CO-08 (Northern Colo.) See Full Big Line

(D) Yadira Caraveo

(R) Gabe Evans

52%↑

48%↓

State Senate Majority See Full Big Line

DEMOCRATS

REPUBLICANS

80%

20%

State House Majority See Full Big Line

DEMOCRATS

REPUBLICANS

95%

5%

Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors
September 29, 2010 09:59 PM UTC

A Razzy For Your Wednesday

  • 71 Comments
  • by: Colorado Pols

It’s an outlier based on every other poll showing this race well inside the margin of error–and we apparently have to pay a fine to somebody for even mentioning–but we’re obligated to call your attention to Rasmussen Reports’ latest poll on the Colorado Senate race, out today:

The latest Rasmussen Reports telephone survey of Likely Voters in Colorado shows Buck with 51% support, while Bennet earns 43% of the vote with leaners included. Two percent (2%) prefer some other candidate, and four percent (4%) remain undecided.

These findings move the race from a Toss-Up to Leans GOP in the Rasmussen Reports Election 2010 Senate Balance of Power rankings. Two weeks ago, Buck held a 49% to 45% lead when leaners were part of the totals…

The survey of 750 Likely Voters in Colorado was conducted on September 27, 2010 by Rasmussen Reports. The margin of sampling error is +/-4 percentage points with a 95% level of confidence.

UPDATE: Some germane food for thought on the expansive question of “likely voters,” and how presumptions and demographic circumstance weight polls–from Politico, H/T to Michael Booth of the Denver newspaper:

This Election Day is poised to send a message about how valid the methodologies behind the surveys are – and prompt questions about how reliant the 24-hour news cycle ought to be on polls for headlines that end up influencing everything from strategy to campaign donations to the perception of viability.

Republican pollster Bill McInturff of Public Opinion Strategies put it bluntly: “Lots of them are simply lousy polls; they don’t accurately reflect younger voters, African-Americans and Latinos. This all contributes to underrepresenting Democrat support. Having said that, there’s a real debate about what will be the appropriate composition of the electorate.”

He said that too many surveys are omitting minor-party candidates on the ballot when they test and that the press ought to make a point of demanding that all demographic data and questions be released with the top-line numbers…

Tweet
Share

Comments

71 thoughts on “A Razzy For Your Wednesday”

      1. and I heard Frazier will get $5 million in 527 spending to bolster his 15-point lead in a poll I heard about.

        — Oops, there I go scooping H-man and Libertad again.

          1. He’s winning. He has it won. It’s ovah.    Right?

            All the polls say so. You say so. Rassy, Libby, wilson – you all say so.  That NRSC is just throwing money around recklessly.  Why don’t they go win Florida or Delaware?

  3. As Madco noted above, Ras reports numerous polls up until the election and then suddenly goes silent.  So, there’s no way to verify the accuracy of their polls, except noting they always have a GOP tilt.

    This is so dishonest that I don’t even believe they do actual polling (have they ever been audited?).  Instead, I would not be the slightest surprised to learn that they are merely exercising a political dirty trick to promote GOP talking points.  

    So, yeah Pols, you owe me $25.

  4. I’m pretty skeptical of this poll and its numbers.  I suppose it is possible that the race has hit a tipping point, but to believe it you have to believe that Bennet has cratered in Denver, Pueblo, Larimer and Boulder counties – and I just don’t think that is the case.

    I think on a pure gut level that Buck probably has a solid 47% and Bennet has a solid 46%.

  6. It was 100 percent robocall, which virtually eliminates the cellphone-only demographic, a relatively young and liberal group.  That goes a long way to explain razzy’s pro republican bias.  Please not that plus or minus 4 percent at the 95 percent confidence level refers only to the statistical odds of sampling.  It makes no allowance for sampling error itself.  

      The infamous Literary Digest Poll that had Alf Landon beating FDR relied on telephones — at a time when mostly rich people had telephones.  Ignoring cell phones is an error of comparable magnitude.

          1. Did you leave a cell number in connection with a contribution or other communication?  You might be considered to have an existing business relationship.

            Also, was the push poll a robopoll or a real human?  That makes a difference.

            1. about two weeks ago from Magellan. Questions were on HD34 candidates and candidates for Gov. The GOP is searching for another race to sink late money into in hopes of getting the majority. I doubt they liked the results of the HD34 poll so they probably moved on to another HD race.

        2. from a phony group claiming to report on credit card info…they say “there is nothing wrong with your account” and hop[e you can’t hear that before they pitch for a new credit card.  

          Obviously, they break that law, as they do when they hit my home phone, which is on the no-call list.

            Politicians, charities and polls, however, exempted themselves from the no-call list and thus may account for some of the legal robocalls.  But to the extent that ralphie is right, it obviously makes the cell-phone problem even worse for pollsters.

      2. 1-The best source of most likely to vote lists are people who actually voted in past elections.  If they list phone numbers, it is usually a land line, rarely a cell phone.  In a case like me, with a cell and a land line, only my land line is listed.

          2-If true random digit dialing is being used, there is still a problem because that gives me two chances of being contacted — my land line or my cell phone.  In contract, a cellphone only user only has one chance, in theory.  If old guys like me have twice as much chance of being contacted as young folks, who are most of the cellphone onlies, the demographics suck.

        3-sort of a subset of 1 and 2 is that I almost never take calls on my cellphone and neither does my daughter, a cellphone only user or her husband, another cell only.  Instead, they let people leave a message and call back later (if the number is onl their contact list and it says Your Wife, they may answer directly.  Robo calls, obviously, don’t leave messages and thus don’t get call backs.

          As cellphones and cell-only households grow, polling gets dicier and dicier.

    2. In practice, none of the studies that have been done to date have discerned any systemic inaccuracies introduced into telephone polling as a result of cell phone effects, presumably because cell phone users and non-cell phone users have been similar enough poitically.  But, there is always a first time, and the more common cell phones become, and the more people who are atypical of the public at large (e.g. younger age demographics) cease to have land lines, the more it will matter.

      FWIW, I think that caller ID (which telecom company bundles has made much more common) and the Do Not Call List, which have together fostered the attitude that people don’t have a duty to pick up the phone and talk to everyone who calls them, have probably done more to skew telephone polling than cell phones have, and it is my sense that the proportion of people who screen their calls has gone way up in the last few years.

      Non-response rates for telephone polls have always been very high, and continue each year to get higher.  Through 2006, at least, there is very little evidence that non-responses were skewing the samples of telephone polls one way or the other politically, but we may have reached the point where non-responses are starting to affect the accuracy of telephone polling.

      The other thing that is screwing up likely voter models is the dilema pollers face between looking at historical turnout numbers and acknowledging the increasing turnout trends created by mail in balloting and early voting, and shifts in the partisan make up of registered voters.  In Colorado, just about the only assumptions that are save, in my opinion, are that turnout will be lower in 2010 than it was in 2008, since this isn’t a Presidential election, and that it will be higher in 2010 than it was in 2006, because of a trend towards voting methods that produce higher turnout.  But, that still leaves a lot of wiggle room.

      In the medium term, I think that the days of true random polling as a useful tool are fading.  Structured samples, that break the electorate into subgroups of registered Republicans, registered unaffiliated voters, and registered Democrats, based upon likelihood of voting in the past and demographic and geographic variables, and then make a concerted effort to get random samples within each subset by multiple means (phone, mail, e-mail, knocking on doors), and then weight each subsample, seems to be the only way to really get accuracy as sampling issues skew random sampling methods that try to capture the entire electorate.

      Indeed, in partisan races, it may be possible to be pretty efficient by devoting all of your resources to figuring out turnout likelihoods in surveys of voters registered with a political party, devoting only token resources to identifying exceptional races where members of one or the other political party are particularly likely to vote for someone other than they’re party’s candidate, in order to leave more resources free to get a more accurate count of the preferences of unaffiliated likely voters.

      Usually, most of the meaningful information that a true random poll has to offer can be summed up by the partisan makeup of likely voters that it reveals and the candidate preferences of unaffiliated voters.  Most of the rest of the information is noise, and if the partisan makeup of likely voters (something that survey questions aren’t particularly good at discerning) is wrong, the bottom line numbers in the poll are worthless.

      The share of the population that always votes and never votes is actually pretty high and fairly easy to identify for public records, and it also doesn’t take that much analysis to get a half decent model to predict likelihood of voting within the population of people who are registered to vote “sometimes voters.”

      Another fruitful avenue for predicting the way that unaffiliated voters will vote is to look at how the unaffiliated vote in localized geographic areas (ideally precincts) varies with the partisan make up of voters who vote or voter registration there.  Neighbor effects can be a pretty powerful predictors at a statistical level of how unaffiliated voters will vote.

      The grunt work for this kind of analysis can be done many months in advance, with only modest work to keep the baseline weighting model updated as the election gets closer.  And, if you can devote more of your limited sample size to polling carefully identified unaffiliated likely voters, you ought to be able to get greater accuracy.

      This involves more skill in the art of polling, rather than merely the science of analyzing the results of a properly drawn sample, but we seem to have already arrived at the point where polling has proven itself to be more art than science.

      1. And they do.

        But it’s not right to make unsolicited calls that cost the recipient money.  That’s what the law says, and I agree with the law.

        In the meantime, polling companies will have to apply mathematical models to correct for the bias that results from not talking to cellphone owners–a big hole in the sample.  And those models will be subject to discussion, if not derision.

        It got so complicated that I gave up doing telephone polls.  To the small customers I used to work for, it wasn’t worth the money.  And to me, it wasn’t worth the trouble.

  7. Aren’t the midterms always about likely voters?  Maybe it is too simplistic, but if young voters, minority voters (traditionally democratic voters) turn out, Bennet wins.  If the older, more conservative voters (traditionally republican voters) turn out, Buck wins.  It isn’t rocket science. Things seemed to be pointing to democratic malaise, and republican enthusiasm.  How many first time voters in ’08 are going to be second time voters in ’10?  You can certainly count on the voters that voted for the 25th time in ’08 voting for the 26th time in ’10.

    1. Turnout is one of the two main factors, but the other is the preferences of unaffiliated voters who make up about a third of registered voters and a somewhat smaller share of people who actually vote.

      Candidates like Ken Salazar and Bill Ritter won landslide elections by winning a very large share of the unaffiliated vote.

      Also, at least according to conventional wisdom, turnout is also generally not equally important for each of the two major parties.  

      Historically, Republicans have tended to have higher turnout on average and equally important, more stable turnout, year in and year out.  Historically, Democrats have tended to have lower turnout on average, and greater variation from year to year.  Factors that influence Democratic party turnout matter more than factors that influence Republican party turnout.

      Unaffiliated voters, on average, are less likely to vote than affiliated voters, but it is harder to pin point circumstances (other than Presidential v. midterm elections) that influence their turnout percentages.  The more divided unaffiliated voters are, the less their turnout percentage matters.  If unaffiliated voters are equally split between the candidates, it doesn’t matter how many actually vote because they cancel each other out.  But, if unaffliated voters strongly favor one candidate or the other, their turnout has a strong impact.

  8. I’m just not seeing the kind of effective positive messaging combined with effective push back from Bennet that we saw in the primary race.  

    Of course as late as two weeks before the deadline Romanoff was polling ahead of Bennet and intense late spending on tough counter punch ad campaign did the trick.  Bennet finished very well then but this is a completely different set of circumstances and Dems must be increasingly concerned that he may not able to repeat for the general. Have been waiting for the campaign to come out with guns blazing but so far…crickets.

    No buyers remorse as I think AR would have done worse. Just disappointment because I think this was Bennet’s to lose. Hope to be pleasantly surprised by late stage blitz.

    1. Bennet is running the attack ads and Buck is running the “Unfair ads against poor me” rebuttal.  So the roles have been reversed.  

      I think Bennet now needs to switch strategy by stating his vision for America.  Something Romanoff never did.  Instead, he went down swinging.

            3. this proves anything.  But then neither does a Rassmussen. So settle down ‘Tad. Besides, I’m pretty sure it was not a DSCC poll I heard about on TV so maybe there are two.  Once again, not proof of anything but still relevant, just like your faves.

    2. In the primary, Bennet was outspending Romanoff 10-1, now it is an even fight.

      I thought Romanoff would have put up a better fight because 1) he would be better able to run against DC and 2)he has some personality.

      2. You just made up the 10 to one figure.  Second, once again I never said this poll proved anything. I leave that to you and Lib who clearly believe any poll you like is great and any poll you don’t is completely invalid.   Here’s where you move on to another subject when confronted with facts.  I’m sure you will not be replying with a link to any valid source for your ten to one figure.  We all know where such a link would lead; the place where  most of your facts come, a dark, dirty and disgusting place.

  9. http://www.politico.com/news/s…

    Do they really think that they can beat Ed Perlmutter?

    As to Rasmussen 538.com calls them reliable and that is good enough for me.  If you want to criticize their record just talk to Massachusetts Senator Coaxley and North Caroline Senator Dole.  Rasmussen was the first to predict their defeats.  

  10. For twenty years we had a land line for our residential phone.

    For 20 years that phone had an answering machine on it. Robots and humans hung up when the machine kicked in.

    So, for 20 years we were surveyed only if I picked up the phone quickly. A rare event.

    And, in theory, we white boomers are represented correctly in polls. Wrong!

    Last week, we dumped that land line. So, the chance of this white boomer being polled just when to zero.

    And, I have only missed one [municipal] election in 40 years so I guess I am a likely voter.

  11. It should be pointed out that Scott Rasmussen was a paid consultant for the George W. Bush campaign and his company has a political bias to push questions that elicit polling responses that favor the Republican candidates.  It is irresponsible for Colorado Pols to not state this fact in this story.  So the poll offered here is probably hogwash.

    2. that various polls are more or less biased for the right or left. Rassmussen is so notorious I usually spot the Dem 5 points. No need to especially point that out in every mention of every poll.  A good place to get the whole sweep of available polls from many pollsters along with averaging of those polls is

      http://www.realclearpolitics.com/

      At best, Bennet isn’t looking like a sure thing to say the least.  Wish it were so.

Leave a Comment

Recent Comments


Posts about

Donald Trump
SEE MORE

Posts about

Rep. Lauren Boebert
SEE MORE

Posts about

Rep. Yadira Caraveo
SEE MORE

Posts about

Colorado House
SEE MORE

Posts about

Colorado Senate
SEE MORE

117 readers online now

Newsletter

Subscribe to our monthly newsletter to stay in the loop with regular updates!