Meteorologists who examine their prognosticative skill often notice a curious phenomena.
Forecasts for snow on December Twenty-fourths and Twenty-fifths have higher probabilities than forecasts for Twenty-sixths or Twenty-sevenths. This disparity is inverted on the Fourth of July, where probabilities for rain are lower than for the Fifth. Surely nature doesn’t have a precipitation date bias, so what is going on?
Wishcasting, that’s what.
Weathermen want white Christmases and dry Fourths just like everybody else, and their desire for these outcomes influence the forecasts they make. These professionals routinely forecast probabilities too high for Christmas and too low for the Fourth.
Wishcasting shows up everywhere: physician prognostications, particularly end-of-life judgments; stock broker picks, mostly in stocks owned by the brokers; sports betting, with a home-team bias; and presidential election guesses, where a form of the home-team bias also appears.
It’s is no small thing, either. The difference between an objective, evidence-based prediction and one made under the influence of desire can be enormous. Its presence is why Las Vegas stays in business and perhaps why there are many more John Does than Warren Buffets.
This year’s presidential election steamed a lot of collars, so I thought it would be fun to test for the presence of wishcasting in the race.
Just after the close of both Democrat and Republican conventions, I conducted a survey asking people who they thought would win, Romney or Obama, who they wanted to win, and how strongly they felt about those guesses. I also collected basic demographics.
Now, this being the internet, and people being people, the chance of fibbing, spoofing, and other forms of dishonesty when answering surveys is not trivial; but I decided, like all sociologists do, to ignore it. That’s what makes it science. I got 506 registered voters, some of whom self-labeled as conservative-leaning, others as progressive-leaning.
No special statistical training is needed to understand the results, except one thing: all probabilities (here and everywhere) are conditional on stated evidence. The chance of 6 when rolling a die is 1/6, because evidence tells us that we have a six-sided object, just one side of which is a 6. If we were geeks and instead had a twenty-sided object (he rolls to attack), the probability changes to 1/20 because the evidence changed.
All the magic happens in this picture:
After the conventions, there was publicly available evidence, like in a stock market, which if considered would have given us the probability Romney would win.
For example, suppose the evidence implied Romney had a 50% chance. I have no idea what the true probability was, but I don’t have to. The horizontal axis in the graph ranges over all possible values this number could have been, from 0% to 100%.
It is true that Obama won the race, but that does not imply that, given the information available at the close of the conventions, the probability he would win was 100%. This would be like saying a person who guessed a six would show on the roll of a die was claiming there was 100% chance the die would show a six.
Examine only the thick red and blue lines and suppose the probability was 50% chance of a Romney win. 80% of folks who picked Romney to win wanted him to win, and 90% of people who picked Obama to win wanted him to win. But if the true Romney win probability was 50%, then Romney supporters overshot by 30%, and Obamaites by 40%. These amounts are the wishcasting bias.
The graph tracks how much wishcasting bias supporters of both men garnered (vertical axis), indexed by the true Romney win probability (horizontal axis).
For Obama fans to have no wishcasting, the true Romney win probability would have to have been 10% (where the blue line intersects 0). For Romneyons, this was 80%. Conclusion is that Obama supporters wishcast less, but still some, if we consider the true probability was likely less than 50%.
The thin red and blue lines break the data down by sureness of picks. People who were strongly sure of their picks had higher wishcasting bias than folks who were weakly sure (this is so if the true probability was greater than 10%).
Women wishcast more than men, but age didn’t seem to matter, nor did political self-labels. There’s more to this analysis than space allows, but I think we have the gist.
William M. Briggs, the Statistician to the Stars!, is a sometime Professor of Statistics at Cornell and a most-times consultant.
DONATE
Donations tax deductible
to the full extent allowed by law.
Comments
Thanks for the post. Had never heard of ‘wish-casting’, but it certainly describes me pretty well in this past election. I certainly did wish for a Romney win and thought, almost up to the day of the election, that he had it.
In other words, you see what you want to see and pick the easiest heuristic solution available?
“The difference between an objective, evidence-based prediction and one made under the influence of desire can be enormous.”
What I really want to know is how many Republicans voters, the ones who were planning to show up at the polls, were lulled into thinking that maybe their individual vote wouldn’t matter since all the Conservative Pundits and GOP Consultants were publicly wishcasting that Romney was going to coast into the White House.
For our side : If wishes was fishes, we’d all have full stringers.
Unsure if this is related, but did anyone notice this from the New York Times over this weekend:
Facing the possibility that President Obama might not win a second term, his administration accelerated work in the weeks before the election to develop explicit rules for the targeted killing of terrorists by unmanned drones, so that a new president would inherit clear standards and procedures, according to two administration officials.
Apparently pretty late in the campaign even the administration was unsure it would be re-elected.
“It is true that Obama won the race.”
It still hurts to read that sentence. My wish-casting didn’t do a lick of good in the election so I’ve turned my wish-casting desire to the lottery tomorrow night. I have this sinking feeling wish-casting isn’t going to help me there either!! Are you SURE Obama won?
Interesting analysis. Another approach would be to assume that Democrat and Republican partisans are roughly equal in their wishcasting bias and then work backwards to find the probability that would satisfy both conditions.
Based on the above chart that would have put Romney at a 45% likelihood to win after the convention which actually seems pretty reasonable in retrospect.
This election wasn’t as close as I thought it was going to be. I think it’s a very interesting analysis you put together. Thanks for sharing.
No matter how many reports I read or statistics I saw to the contrary I had a nagging feeling Romney wasn’t going to win.
This is only antidotal but no one I knew took for granted Romney would win and every one of them was afraid not to vote. My personal opinion, the people who stayed home did so because they didn’t want to vote for Romney.
I felt he was ready to govern but never felt he knew how to win. As a typical moderate liberal he missed all the opportunities to tell people who Obama was, what he was doing (lying)and what he intended to do. Obamacare was an overwhelming minus for Obama that Romney couldn’t use because he was still enamored with his own version.
Well, you have the WISHcasting… but what about FEARcasting? By that I mean how much of the prediction pool was wrong not because they WANTED their guy to win, but because as pessimists, they FEARED that their guy would lose?
If what you’re showing is that the Right was still optimistic before the election, that’s cool. Well, better than finding out we were all Allahpundit type Eeyores ‘nStuff.
The problem with this analysis – perfectly valid in its own right – is the way it is inevitably received by the wishcasters on both sides.
The winning side naturally flees to the succor of their victory and takes full validation from it – although even a dead even 50-50 race would result in one side or the other winning and claiming validation: “See? We were right all along!”
A less clear rationalization takes place on the losing side, but one with more logical-sounding arguments. For example, Republicans are blaming Romney as a candidate, his bio, his refusal to go on the attack over Benghazi, the failure of the ORCA GOTV plan, libertarian defectors, and discouraged white conservatives – all of which may well have played some role in the defeat, but none of which change the fact that Republicans were engaged in active wishcasting from the time of the first debate, at least.
[…] InsurrectionToday’s post is over at Legal Insurrection, a classy site run by Cornell Law guy Bill […]
Who did the I Ching predict? 🙂
Are you one of those people who like to suck the life out of a party? Do you annoy people around you by pointing out the inconsistencies in a movie? You know there is such a thing as a statistical fluke.
The Redskins typically play strong in the first half and then usually flub the game in the second. (their defensive line sucks) Yet there are many die hard Redskin fans who cheer for them every game. And yet every once in a while they are rewarded with a win.
You don’t win if you don’t play but you only have to win once to demonstrate you are still in the league. The New Orleans Saints demonstrated that point.
This election was not about expectations but about technical points. Obama for all his incompetence in administering the country does one thing well and that is winning election contests. It’s the only thing he does well. How many times did Romney go for the GOP nomination and was repeatedly rejected by Conservatives? Even this election cycle Romney only won the nomination because he was the last man standing. McCain did the same thing in 2008, he was the last man standing and by default got the nomination. Romney ran on expectations, Obama ran on technical points. Romney is good at running companies, Obama is good at running sales mills like time shares and Shamwow commercials.
After all what is an election but a TV commercial with a shuck and jive announcer selling Shamwows.
http://www.dailymotion.com/video/x7zrly_shamwow-commercial_shortfilms
Mr. Briggs this is you testing Shamwow to verify Vince’s claims
http://www.dailymotion.com/video/xocpni_shamwow-unboxing-test-linus-tech-tips_tech
My point here is that just because you win an election doesn’t mean you have the competence to do the job you sold yourself for, they are two different skill sets. Remember when Democrats defended Obama in 2008 when it was pointed out Obama had zero executive experience in running anything? Their response was he successfully ran his election campaigns… Obama upon winning the presidential election doesn’t represent the country, he represents a well oiled time share sales operation.
A note on the last man standing paradigm Romney based his campaign upon. He like McCain reasoned that competence was the issue for the electorate to make their decision. They figured given the zero qualifications Obama presented that they would win the election, after all they survived the GOP nomination process by sheer will power of remaining standing.
The GOP candidates misread the electorate, they weren’t interested in hearing about competence, they were wishcasting for the best message that promised them the most. Obama understands that the majority of the electorate wants a magical solution to their problems and that the solution offered requires the least amount of effort on their personal part. One look at the youth today in their work ethic and expectations tells the story.
Romney lost because all things weren’t equal, Obama won because he knew he couldn’t win by Romney’s rules. Why play by the rules if you are going to lose? So change the rules.
Good post. Cognitive bias.
I see only one real problem with this and other evaluations of the election. They al ignore a rather serious BIAS in the final results.
Let’s start with the well known bias. Barry created 4 MILLION new citizens out of young illegals of voting age illegally. If only half of those people voted we have Barry’s win. We then look at the less well established, but, still somewhat persuasive reports since the election. Barry didn’t win any states with Voter ID laws. Reports of illegals and felons voting illegally after Acorn and other registration drives. Voting Machines defaulting to Barry. Precincts reporting from 90-150% of registered voters voting, and, of course, the usual dead turnout.
Nope, kinda hard to call a close election with that kind of BIAS!!