Don’t Shoot The Messenger: CU Study Predicts Romney Win

( – promoted by Colorado Pols)

A new University of Colorado study produced a model that would have correctly predicted every Presidential election since 1980, and it predicts that Willard “Mitt” Romney will be the 45th President of the United States. The study, done by Political Science Professors Michael Berry from CU Denver and Ken Bickers from CU Boulder, uses different indicators than most other models. Instead of looking at one measurement of economic health for each state, it looks at two — per capita income, as well as that state’s unemployment number. According to the model, every battleground state will go to Romney, including Florida, Virginia, North Carolina, Ohio, Pennsylvania, Wisconsin, Minnesota, New Hampshire and Colorado, making the grand total Romney 320 to Obama 218.

The model also correctly predicted what happened in 2000, when Al Gore won the popular vote, and George W. Bush won the electoral college.

It is unclear from their press release whether the model included the many recent Republican “War On Women” gaffes, or considers the huge influx of money into the GOP campaigns post-Citizens United. It also does not state if it takes into account Koch brothers’ PACS buying up most of the air time on television.


My own anecdotal evidence from knocking on doors tells me there seems to be a sizable number of independents and unaffiliateds who seem less enthusiastic about the President than they were four years ago. They claim to be “disappointed” but when asked about specific reasons why, have very little to say (it has crossed my mind more than once that Obama’s PR team may have been more policy wonks than great marketers these past four years).

There also seems to be greater complacency among political activists who supported Obama four years ago. With the number of Republican gaffes and Tea Party members of Congress horror stories, are many Obama supporters so sure of a win for the Dems they have yet to put in any real effort to make it happen?

About nancycronk

Nancy Cronk is a longtime community activist and women's leader living in Arapahoe County. Six months before the historic "red sweep" election of 2014, she was recruited to run as a "placeholder" in HD37, and managed to bring in 40K from 500 small donors, and 42% of the vote -- just one point lower than the previous candidate who ran in a presidential year.

37 Community Comments, Facebook Comments

  1. thiokuutoo says:

    Say what? A model that predicted the Supreme Court decision to install the village idiot as president? Anyone who comes up with that and insists it is real has to be partying nude in the Sea of Galilee.

  2. dwyer says:

    It is called Colorado Pols.  There was a vigorous discussion of this announcement from CU, yesterday.

    And a pollster – sxp151- analysized the so-called study.  You might find it interesting

    OK, they just posted a new link to the paper

    I don’t know the standards of the field, so I can’t compare how this model holds up against any other model, but essentially what they did is to take eight (8) data points and a statistical analysis of a bunch of variables. The model was only recently devised, so as I suspected it has never successfully predicted an election outcome (only been fit to retroactively predict what happened).

    From the official CU press release:

    “For the last eight presidential elections, this model has correctly predicted the winner,” said Berry. “The economy has seen some improvement since President Obama took office. What remains to be seen is whether voters will consider the economy in relative or absolute terms. If it’s the former, the president may receive credit for the economy’s trajectory and win a second term. In the latter case, Romney should pick up a number of states Obama won in 2008.”

    Their model correctly predicted all elections since 1980, including two years when independent candidates ran strongly, 1980 and 1992. It also correctly predicted the outcome in 2000, when Al Gore received the most popular vote but George W. Bush won the election.

    This is absolutely misleading, and at first I thought perhaps some uninformed publicity writer was responsible for it. But no, it’s one of the authors of the study implying that his model has been successfully used for 32 years to predict Presidential outcomes. Yes, that’s how you get publicity for your paper, but it obviously leads to confusion among people who are not sure how linear regressions work.

    It’s like if I suddenly decided tonight to count Libertad’s posts per day for the past eight days, then confidently predicted he will post 14.27 times on Thursday based on a statistical analysis of the temperature outside, Sean Hannity ratings, and word count of the day’s Wall Street Journal editorials. And then I claimed that my model successfully predicted the last eight days already, so we’re all in trouble. Well no it didn’t.

    In fairness, they do use election data from all 50 states, so it’s not as bad as using only eight data points, and they do note at the end that the model has never been successfully tested, but I’m pretty confident that using a different set of variables with the same analysis could produce a different result. In the end their success rate with state-by-state predictions is about 90%, measuring only which candidate wins or loses, not by how much. Neglecting everything except swing states reduces that success rate a lot. Yet somehow at the end of the paper they make predictions to four digits of each state’s vote share. FOUR DIGITS (e.g., Obama is projected to get 48.19% of the vote in Colorado).

    In short I would trust an election model that’s already made a single successful prediction over this.

    “Mitt Romney and Paul Ryan” = “My ultimate Ayn Rand porn”

    by: sxp151 @ Wed Aug 22, 2012 at 16:57:55 PM MDT

  3. parsingreality says:

    Besides lopsideness of R PAC money (Whare ARE those infamous union bosses when you need them?) I would count:

    1.  Romney losing women’s votes by even more than he never had, thank you Akins.

    2.  The lack of income and tax transparency of Rmoney; lots of average Joe’s will remain pissed off about that.

    No model could have incorporated those and other factors.  

  4. LemonLyman says:

    The renowned statistician has a considerably different take:

  5. Dark Cloud says:

    In the Daily Camera, it was stated this method HAD predicted correctly every election since 1980.  Here, it’s “would have predicted….” Which?

    And they pad their prediction by not mentioning all data was corralled five months previous to the election (that would be….April)and they’ll update it closer in.

    There is not enough liquor in Colorado to allow a three digit IQ to value this. The Coors Curia for Conservative Catholic Chickenhawks in Golden – and those face down in their lap – will give it a go, though.  

  6. ArapaGOP says:

    I don’t think it’s gospel either, but it should be taken into consideration.

  7. dwyer says:

    And, it is being promoted on some newscast as a model that has successfully predicted election outcomes over the last thirty years….that is atrocious…..and it reflects badly on CU, IMHO.

    I really appreciate sxp151 analysis…..I wish to hell he had an agent….

    • sxp151 says:

      that Nate Silver had many of the same objections and has a much larger audience. So it’s not like nobody will ever hear that there are serious problems with it. Hard to do much public critiquing as an anonymous jerk on a blog though.

        • MADCO says:

          I’m not outing anyone to do it though.

          Think about this though-

          Colorado is approximately 1/3 D, 1/3 R, 1/3 U.

          And has been for some time.

          An outsider who didn’t understand CO would think we split most of the time.

          But we don’t.

          Our U’s are slightly less likely to vote than our D’s and R’s. (maybe they don’t get their general ballot because they didn’t vote in the primary – IDK)

          And when they do vote, they tend to vote R.  Not always. But generally.

          That’s why Bennet (an others) was smart to use a Colorado polling company in 2010 – they understand CO.

          But if you’re 538 /Silver you gotta get your insider dope somewhere.

Leave a Reply

Comment from your Facebook account

You may comment with your Colorado Pols account above (click here to register), or via Facebook below.