For the life of me I cannot understand why, when a political analyst produces a range for their prediction of election results, Ed Fitzgerald would arbitarily pick the lowest number of the range. Logic would suggest that if a fair-minded person were going to reduce a range to a single number, they would pick the mid-point in that range.
My predections were, 4-6 seat loss for Republicans in the Senate (the midpoint would be 5, the final result was 6), a GOP loss of 6-8 governorships (midpoint 7, final 6) and 20-35 in the House (midpoint 27.5, as of 11/19, it is 29).
Is someone wants to quibble about us being off by one seat in the Senate and governorships and 1.5 seats off in the House, fine, go ahead, but to pick the lowest number in a range is methodologically nonsense and grossly unfair.
Thank you for your note.
I did not approach this election as a non-partisan analyst, I approached it as a very concerned Democrat (as I clear stated in my explanatory notes) with a vital interest in understanding what was likely to happen in the election. As such, and since I included prediction of both partisan and non-partisan analyists, I did not want to be lulled by perhaps overly optimistic predictions of a Democratic win, I wanted instead as conservative an estimate of what would occur as possible.
This is why I decided to use the lower number whenever an analyst provided a range of numbers except when the analyst specifically flagged a number in the middle of the range as the most likely result, in which case I used that number.
(In the penultimate report of my survey, I also ran the numbers using the highest number in the ranges and found that the resulting averages where not significantly different from those that resulted using the lower estimates. This was because only a few analysts -- 9 to be exact -- resorted to ranges instead of specific projections. [Correction: Although 9 analysts had ranges, only 4 of them, including Cook, did not indicate a preferred number for the House pick-ups.])
Larry Sabato, for instance, in his prediction for the House had a range of 25-33, but said that 29 was the most probable, so I used 29 for his line. If you thought that the most likely outcome within your range of 20-35 was something close to the midpoint of that range ( 27 or 28 ), perhaps you might have made that clear in your report.
I will concede this, however, that in my aftermath report, I did not make the conceptual leap to change from the conservative prediction model I had been using before the election to one designed to best represent the projection of each analyst, and so I simply re-used the same numbers I had used before. (The chart was, basically, simply a re-sort of the previous one.) I regert that error, and I will make an effort to correct in when I re-do the chart again -- which I plan to do once all the House races are decided.
As for the contention that the midpoint best represents your prediction, I'm not sure I totally agree with you. 27.5, after all, is the midpoint not only of your prediction (20-35), but also of an infinite number of other predictions: 19-36, 18-37, 17-38, 16-39 etc. If Joe Blogger had predicted a Democratic gain of 14-41 seats, would you believe that he should be credited with the same accuracy as you, simply because your projections both shared a midpoint of 27.5?
But, really, it seems to me that the best way to insure that your predictions aren't mispresented is to make them a little more specific and less broadly general. A 15 seat range, after all, is hardly the model of pinpoint accuracy.
One correction: Mr. Cook's prediction of a 20-35 seat pick-up by the Democrats encompasses 16 different possible outcomes, and not 15.
Since Mr. Cook didn't state a preference for any particular result within his fairly broad range as being more likely than any other (and such a choice would not necessarily have to be the mid-point, since the possible outcomes could, conceivably, be weighted, with some more likely than others), than all the results are fair fodder for analysis. If the pick-up had been 20 seats, surely Mr. Cook would be claimiing some measure of success for having included that within his predicted range, just as he would have if the result had been 35 seats. With that in mind, I don't see that using 20 seats for his prediction in an effort to generate a conservative projection of the election's likely results can be considered a gross misrepresentation.
Update: I was a little perplexed about what brought Mr. Cook's attention to my survey after all this time, and provoked his ire in the process, but perhaps it was this.
Update: I've had a reply via e-mail from Mr. Cook. I don't as yet have permission to quote directly from it, but it basically said that I shouldn't use his numbers in the future if I was going to misrepresent them. He also pointed out that he became aware of my survey through journalism.org, presumably the post I referred to above, and suggested, somewhat sarcastically, that this was an example of "excellence in journalism". (If Mr. Cook responds with permission, I'll post his e-mail here in full. [See below. - Ed])
My reply was this:
I'm sorry, your reply assumes that I have misrepresented your figures, and I don't believe that is the case, for reasons that I think I've adequately explained. In any event, your figures are publicly available, and as long as they remain so, I will continue to use them in any way that I deem appropriate.
Since you have been kind enough to offer me advice, please allow me to reciprocate. My suggestion to you, is that you might consider working a bit harder to provide predictions that are more specific and less broad and general, since it opens you up to the suspicion that you are deliberately playing it safe and hedging your bets. That's understandable, I suppose, but you are a professional analyst (the dean, according to some) who supposedly understands the political process better than most, with inside information and insight drawn from experience, and yet your predictions were so broad as to be practically useless, and you didn't end up doing all that much better than many amateur analysts.
Finally, if you believe that the failings, whatever they may be, of journalism.com are the sine qua non of the criminal lack of journalistic excellence in our political media, I suggest you take a giant step backwards and reconsider the contemporary media landscape, which includes the preference for conventional wisdom over skeptical inquiry, the dissemination of partisan received ideas in the guise of neutral reporting, collusion with official leaking for obvious political purposes, the profusion of anonymous sourcing even by outlets that in theory disallow it, print journalists doubling as pundits on TV, and steganography masquerading as reporting. It's a cesspool that in no way serves the vital function that the fourth estate was intended to provide, of keeping the public informed and acting as a brake on the use and misuse of government power. That a minor website repeated the, perhaps, errant work of an even more insignificant blogger, just doesn't stack up against that, however annoyed you personally might be about it.
Update: As I said above, I wasn't planning on doing another chart until the House races had been decided, but given Mr. Cook's objections to my representation of his election prediction (objections which have some validity in regards to my aftermath chart, if not in respect to the pre-election survey reports), I decided to re-run the chart, using the midpoint to represent Mr. Cook's prediction, as well as the prediction of the 3 other sites (Kos, Tradesports and Rothenberg) which also gave ranges without specifying a likely result.)
Comparison of this chart with the original one shows that of the 4 sites that had their numbers changed, Kos improved the most, moving from the middle of the pack to just above the Nov 9th estimate of what the final result may be, Rothenberg moved from close to that estimate to farther away from it (above), and Tradesports and Cook moved from the bottom third to the middle third, still well away from the jackpot. (Cook, specifically, moved from the top of the bottom third to the middle of the middle third, moving up 7 slots.)
So if Mr. Cook's ire stems from how I "misrepresented" his House prediction in a way that understated its accuracy, it doesn't seem that using his preferred representation, utilizing the mid-point of his range, really does an awful lot to salvage it. He still comes in significantly far away from whatever the final result ends up being, bettered by other professionals as well as inexperienced and unheralded amateurs.
Update: Mr. Cook responded to curtly deny permission to post his e-mail. Given the policy stated in the sidebar -- "All e-mail received is subject to being published on unfutz without identifying names or addresses." -- I don't actually need his permission to publish it, but, what the heck, I don't want to be petty about it. In any case, my 51-word description above of his 58-word e-mail more than adequately conveys its content. If questions should come up about its actual contents in the future, I'll reconsider my decision and publish it should it seem advisable.
Update: On Election Day night, after the exit polls were available to the networks, but before actual results came out, Charlie Cook, interviewed on MSNBC by Chris Matthews, said that if someone told him that his estimate of 20-35 was wrong, he'd bet that it was low, that 35+ was more likely than under 20. Asked about his Senate prediction, he said that if it turned out to be 35+ it would be likely that the Democrats would take the Senate, but if it was around 22-25, the Republicans would likely hold on to the Senate.
This illustrates the difficulty of crediting Cook with any degree of accuracy, given the broadness and generality of his predictions. It turns out that his overall range, 20-35 will turn out to be correct, since in the unlikely event that all the disputed races so their way the Democrats would pick up exactly 35 seats, but since he now seems to feel that his prediction is best captured by the midpoint of the prediction, 27.5, that will turn out to be low -- how low depends on on how the disputed seats fall out.
Such non-specific punditry is a pretty good example of what I mean by Cook hedging his bets.
One thing is certain, though -- regardless of how accurate or non-specific Mr. Cook's predictions should turn out to be, we'll still be seeing him on TV interviewed by the likes of Matthews and Russert, while the amateurs who did as well or somewhat better than him will likely be nowhere to be found.
hostile to science
lacking in empathy
lacking in public spirit
out of control
Thanks to: Breeze, Chuck, Ivan Raikov, Kaiju, Kathy, Roger, Shirley, S.M. Dixon
i've got a little list...
Steven Abrams (Kansas BofE)
Howard Fieldstead Ahmanson
Roger Ailes (FNC)
Alan Bonsell (Dover BofE)
Bill Buckingham (Dover BofE)
George W. Bush
Bruce Chapman (DI)
The Coors Family
William A. Dembski
Leonard Downie (WaPo)
John Gibson (FNC)
Fred Hiatt (WaPo)
James F. Inhofe
Philip E. Johnson
by Joel Pelletier
(click on image for more info)
Stephen C. Meyer (DI)
Judith Miller (ex-NYT)
Sun Myung Moon
Elspeth Reeve (TNR)
Martin Peretz (TNR)
Richard Mellon Scaife
Susan Schmidt (WaPo)
John Solomon (WaPo)
Richard Thompson (TMLC)
Bob Woodward (WaPo)
All the fine sites I've
Be sure to visit them all!!
Arthur C. Clarke
Daniel C. Dennett
Philip K. Dick
Stephen Jay Gould
"The Harder They Come"
Ursula K. LeGuin
The Marx Brothers
Michael C. Penta
Michael Powell & Emeric Pressburger
"The Red Shoes"
"Singin' in the Rain"
Talking Heads/David Byrne
Hunter S. Thompson
"2001: A Space Odyssey"
If you read unfutz at least once a week, without fail, your teeth will be whiter and your love life more satisfying.
If you read it daily, I will come to your house, kiss you on the forehead, bathe your feet, and cook pancakes for you, with yummy syrup and everything.
(You might want to keep a watch on me, though, just to avoid the syrup ending up on your feet and the pancakes on your forehead.)
Finally, on a more mundane level, since I don't believe that anyone actually reads this stuff, I make this offer: I'll give five bucks to the first person who contacts me and asks for it -- and, believe me, right now five bucks might as well be five hundred, so this is no trivial offer.