25 replies to this topic

[Deisel]

Acworth - whats your thought on these guys Models? They have been right since 1980. They predicted Romney to win 52% of the vote.
A state-by-state analysis of the presidential race conducted by two University of Colorado professors predicts that Mitt Romney will be our next president. The analysis, released today, by political science professors Kenneth Bickers of CU-Boulder and Michael Berry of CU Denver, is based on economic and other factors within each of the 50 states. This same study has correctly predicted the winners of the last eight presidential elections, starting with the 1980 election won by Ronald Reagan. http://www.examiner....esidential-race

[AcworthFalcFan]

Michael Berry is one of the top names in the field and Bickers is top notch as well.  I've talked with Berry a few times over the phone, actually.

Their estimate of the popular vote seems high to me, but only by about 1% or so.  My own model has it much closer, but then I'm not using a state-by-state model (yet).  

Overall, I think this is a very credible forecasting model by two of extremely good political scientists. I'd be surprised if the popular victory for Romney was that wide, but the electoral college prediction seems fairly reasonable to me.

[Deisel]

AcworthFalcFan, on 23 August 2012 - 01:19 PM, said:

Michael Berry is one of the top names in the field and Bickers is top notch as well.  I've talked with Berry a few times over the phone, actually.

Their estimate of the popular vote seems high to me, but only by about 1% or so.  My own model has it much closer, but then I'm not using a state-by-state model (yet).  

Overall, I think this is a very credible forecasting model by two of extremely good political scientists. I'd be surprised if the popular victory for Romney was that wide, but the electoral college prediction seems fairly reasonable to me.

Thanks - as you'r in this field I felt you'd know who these guys were.

[holymoses]

AcworthFalcFan, on 23 August 2012 - 01:19 PM, said:

Michael Berry is one of the top names in the field and Bickers is top notch as well.  I've talked with Berry a few times over the phone, actually.

Their estimate of the popular vote seems high to me, but only by about 1% or so.  My own model has it much closer, but then I'm not using a state-by-state model (yet).  

Overall, I think this is a very credible forecasting model by two of extremely good political scientists. I'd be surprised if the popular victory for Romney was that wide, but the electoral college prediction seems fairly reasonable to me.

AcworthFalcFan, on 23 August 2012 - 01:19 PM, said:

Michael Berry is one of the top names in the field and Bickers is top notch as well.  I've talked with Berry a few times over the phone, actually.

Their estimate of the popular vote seems high to me, but only by about 1% or so.  My own model has it much closer, but then I'm not using a state-by-state model (yet).  

Overall, I think this is a very credible forecasting model by two of extremely good political scientists. I'd be surprised if the popular victory for Romney was that wide, but the electoral college prediction seems fairly reasonable to me.

AcworthFalcFan, on 23 August 2012 - 01:19 PM, said:

Michael Berry is one of the top names in the field and Bickers is top notch as well.  I've talked with Berry a few times over the phone, actually.

Their estimate of the popular vote seems high to me, but only by about 1% or so.  My own model has it much closer, but then I'm not using a state-by-state model (yet).  

Overall, I think this is a very credible forecasting model by two of extremely good political scientists. I'd be surprised if the popular victory for Romney was that wide, but the electoral college prediction seems fairly reasonable to me.

Blah blah blah . . . Mr. BIGSHOT!!!!  Get on the phone with NATE and I'll be impressed.

In all seriousness, and without being TOO technical, can you explain the modeling differences between these guys and Nate Silver?  I know they are all crunching numbers and weighting them based on past results . . .

Does Silver weight demographics more relative to economic indicators?

A man so nice, I quoted him thrice!

[AcworthFalcFan]

holymoses, on 23 August 2012 - 01:39 PM, said:

Blah blah blah . . . Mr. BIGSHOT!!!!  Get on the phone with NATE and I'll be impressed.

In all seriousness, and without being TOO technical, can you explain the modeling differences between these guys and Nate Silver?  I know they are all crunching numbers and weighting them based on past results . . .

Does Silver weight demographics more relative to economic indicators?

A man so nice, I quoted him thrice!

Yeah, I said that I've talked with him a few times.  That doesn't mean that he'd know me from Adam's cat, though.  

I'm about to head out for the day, but here's a short answer until I get time for a more detailed one:

I'm pretty sure that Nate Silver's forecasting model also weights according to recent polls.  He starts with some of the same structural factors (economy, approval) but incorporates things like recent poll results and perhaps even ideology of the candidates.  

Most political scientists rely on a "structural only" model that doesn't take into account the small but sometimes important campaign effects.  

Until the PS: Politics and Political Science edition with all of the forecasts is released in October, it's hard to assess what factors are being used for each model.  Some, such as Hibbs' "bread and peace" model or Jim Campbell's economy/approval model are pretty universal from one year to the next.  Others crop up each year that weight things differently.  As with polls, I think the best thing is to look at all of the forecasts together to see what they say on average.  A few of the models, especially Campbell's, are really accurate across all the years.  A few models have mispredicted.  But on the whole, across all the models they tend to be extremely accurate at projecting the winner.

[AcworthFalcFan]

I'll also poke around and see if I can find Berry and Bickers' data for replication.

[AcworthFalcFan]

Last thing for the day...

Great thread, Deisel!  This isn't as relevant to the specific forecasting models, but it's a good place to pimp a couple of free ebook chapters for an upcoming book on the 2012 presidential election.

http://press.princet...ers/s2_9920.pdf

http://press.princet...ers/s3_9921.pdf

It's written by a couple of good political scientists (those who frequent Pollster might recognize one of the names) and it's an excellent non-partisan, non-opinionated analysis of the presidential race so far.

[Deisel]

AcworthFalcFan, on 23 August 2012 - 02:26 PM, said:

Last thing for the day...

Great thread, Deisel!  This isn't as relevant to the specific forecasting models, but it's a good place to pimp a couple of free ebook chapters for an upcoming book on the 2012 presidential election.

http://press.princet...ers/s2_9920.pdf

http://press.princet...ers/s3_9921.pdf

It's written by a couple of good political scientists (those who frequent Pollster might recognize one of the names) and it's an excellent non-partisan, non-opinionated analysis of the presidential race so far.

I thought this would hit a grounder to ya. Interesting stuff.

[kicker]

I just don't see Obama losing Colorado and New Hampshire.

[Spacey C]

I gotta say they may be right, but" being right since 1980" Isn't a lot to brag on. I don't think anyone could have been wrong predicting Reagan over Carter,GH Bush over Dukakis, Clinton over Bush and GW over nobody.

The only real surprise was idiots voting GW in twice, but as I said, that was more the Dems fault for putting unelectable people up.

[Heisenberg]

I was scrolling through all the forums of this message board and when I got down to Anything But Football the preview for the first topic in this forum read University of Colorado Anal...

This thread falls monumentally short of my expectations.

[kicker]

Heisenberg, on 23 August 2012 - 06:08 PM, said:

I was scrolling through all the forums of this message board and when I got down to Anything But Football the preview for the first topic in this forum read University of Colorado Anal...

This thread falls monumentally short of my expectations.

http://www.uch.edu/c...er/anal-cancer/

[holymoses]

Heisenberg, on 23 August 2012 - 06:08 PM, said:

I was scrolling through all the forums of this message board and when I got down to Anything But Football the preview for the first topic in this forum read University of Colorado Anal...

This thread falls monumentally short of my expectations.

Particularly disappointing on a Thursday, of course.

[Spacey C]

Heisenberg, on 23 August 2012 - 06:08 PM, said:

I was scrolling through all the forums of this message board and when I got down to Anything But Football the preview for the first topic in this forum read University of Colorado Anal...

This thread falls monumentally short of my expectations.

[AcworthFalcFan]

TheThinker, on 23 August 2012 - 06:04 PM, said:

I gotta say they may be right, but" being right since 1980" Isn't a lot to brag on. I don't think anyone could have been wrong predicting Reagan over Carter,GH Bush over Dukakis, Clinton over Bush and GW over nobody.

The only real surprise was idiots voting GW in twice, but as I said, that was more the Dems fault for putting unelectable people up.

In hindsight, all of that is true.  The problem is that it's only been the last 2-3 presidential elections where these forecasting models were taken seriously.  Back in 1984, people actually believed that Reagan was in serious trouble of losing because of he looked "too old" in one of his debates.  Every political scientist knew it would be a bloodbath, but the media and the pundits thought Reagan could lose up until the second debate.

[AcworthFalcFan]

A few other thoughts on forecasting models in general this election cycle.  I wonder how much these models might be off due to a statistical "glitch" in the data concerning economic performance.  Going back to 1980, and perhaps longer, I can't remember when the overall economic picture has been ambiguous.  In 1980, it sucked.  In 1984, it was great. Same for 1992 and 1996 respectively.  Even in 2000, the economy wasn't bad and it wasn't great, but generally it was okay enough for the incumbent party to win the popular vote (e.g., Gore).  

This year, it's not so clear.  We're no longer in a recession.  We've had millions of jobs added.  Yet the rate of growth is still very stagnant and has been for a long time.  IOW, for most elections the economic cues communicated to voters was clear and concise--things are good or things suck.  This year, the news is muddled, mixed, and unclear.  

The forecasting models make their predictions based on past results, and the data from the past don't include values in the middle range of the variable (economic performance).  That has the potential to really throw off the forecasting predictions when you use a model based on unambiguous good/bad economic times to predict what happens during a period with an ambiguous and unclear economic situation.

Edit: To put all of this as simply as possible, we know what happens very clearly when things are going great (incumbent wins) and when things really suck (incumbent loses).  So people blame the incumbent for bad times and reward the incumbent for good times.  But how do people react when things are neither good nor bad?  We don't know and so our ability to predict could be seriously off.  It's possible that people give the incumbent the benefit of the doubt during mediocre recoveries.  Or it could be that people get impatient with mediocrity in the incumbent and punish him and his party for not doing a better job.  Either of these could throw the forecasts off by several percentage points.

[falcon_bob]

Heisenberg, on 23 August 2012 - 06:08 PM, said:

I was scrolling through all the forums of this message board and when I got down to Anything But Football the preview for the first topic in this forum read University of Colorado Anal...

This thread falls monumentally short of my expectations.

It's ABF in an election year. You should've known better to expect anything other than a political thread.

[AcworthFalcFan]

falcon_bob, on 23 August 2012 - 10:40 PM, said:

It's ABF in an election year. You should've known better to expect anything other than a political thread.

F*** you, liberal hippie communist seaweed sucking godless babykiller!

[AcworthFalcFan]

kicker, on 23 August 2012 - 03:04 PM, said:

I just don't see Obama losing Colorado and New Hampshire.

Colorado is interesting precisely because it is surrounded by some very Republican states.  Nobody thinks, for instance, that Obama has a chance in h*** of winning Nebraska.  Yet, OfA are organizing extremely heavily in western Nebraska to get volunteers to call and canvass neighborhoods...in Northern Colorado.  Romney, on the other hand, has totally written off the state and doesn't have strong operations there.  That means that Romney in Colorado is facing battle lines not only inside of Colorado, but also facing reinforcements from Democrats in all of those heavily-Republican states surrounding it on all sides.  

Obama is being way more strategic about allocating campaign resources than Romney in the mountain west right now.  That might change, but those are the kinds of things that can shift the results by a few percent in a state like Colorado.

[holymoses]

FOX News 8/19 - 8/21 1007 LV 3.0 44 45 Romney +1
http://www.realclear...obama-1171.html

I missed this late one yesterday, but it is very interesting . . . .

BAD for Obama because it shows a 10 point change in favor of Romney from the last Fox poll just prior to the Ryan announcement.  That's WAY beyond any other observed bounce.

But it was also just before the Akin SHTF, so it might not be SO bad.

Of course, it is one poll, and, most certainly, the early August Fox poll was an outlier.    But it reinforces my theory that there actually was a  pretty decent Ryan bounce.