Parsing the Indiana and North Carolina Primaries

Yesterday, Democrats voted in Indiana and North Carolina. My latest W.S.J. column parses the results. A few highlights:

With Barack Obama and Hillary Clinton trading victories in North Carolina and Indiana, it’s tempting to call Tuesday’s primary vote a split decision. Instead, political prediction markets have declared Senator Obama a clear winner.

Senator Obama began Election Day rated a 76 percent chance to win the Democratic nomination. By the time the poll results were finalized, the markets had reassessed his chances at 89 percent. This 13 percentage-point rise makes Tuesday’s primaries the best day of the campaign for Senator Obama since the surge of momentum he enjoyed after the Iowa caucuses.

On the flipside, Senator Clinton’s chances of securing the nomination were cut in half, falling from 22 percent to 10 percent.

The political junkies amongst us will be easily recognized today: we are the bleary-eyed folks hovering over the coffee pot.

The Indiana race turned out to be surprisingly close, and while C.B.S. confidently called the race minutes after 8 p.m., the other networks waited until about 1:15 a.m. While C.B.S. was ultimately correct, you can bet there were some white knuckles as the vote count came in closer than expected:

Interestingly, when C.B.S. made its early call, prediction markets still rated Senator Obama a 1-in-14 chance to win. His stock dipped briefly in response to that call, but within 15 minutes, it became clear that C.B.S. didn’t have any special information, and Senator Obama’s stock quickly recovered to its previous level.

Thus, political prediction markets clearly suggest that C.B.S. analysts took a pretty big gamble.

The propensity of committees of expert analysts to be overconfident is a widely documented phenomenon, and if the C.B.S. bigwigs are interested, I’m willing to bet that greater reliance on the wisdom of crowds can serve as a useful counterweight.

It seems pretty unlikely that the network executives intended their station to risk a 14-to-1 bet.

A full wrap-up is available here.

Jirka Lahvicka

A quick note on Intrade prices; although Intrade claims that "91.1 means the market predicts there is a 91.1% chance that this event happens.", it is not correct once you take into account transaction fees and the interest rate. Current Barack Obama nomination price of 91.1 indicates more than 93 percent chance of his becoming a nominee (if traders are fully rational).

Jirka Lahvicka

Re [6] The InTrade prediction market did a very good job (or the traders did :-). Obama win in Indiana traded at one point (hours after CBS called it) for about 30 (indicating 30 percent chance of winning) - we were all waiting for Lake County results and it was clear that Obama could erase the whole Clinton margin. Once it became clear that he would not, the price crashed close to zero. Do not underestimate traders - the make a very good job of aggregating all information (not just TV news) and various analyses floating around.

If you want to see an example of a good Indiana analysis (which is probably just one source that many traders are aggregating), go here:

(By the way, it is fascinating that the best election analysis is found on blogs such as this and not in mainstream media or even here).

S. Heaton

It's not clear to me the point Professor Wolfers is trying to make. Are prediction markets doing a good job or not? How do they work well when they led to surprise in Indiana?

I don't see why he's getting on CBS - so they called early at the same time tradesports was priced at 90.


I suspect political analysts were expecting Clinton to take them both if her theories about momentum shifting in her direction were correct.

Since it was so close in Indiana and Obama won handily in SC, it basically showed that not much has changed and Obama is still comfortably in the lead despite the Wright controversy and Clinton's implication that the bloom had left Obama's rose.


it's tempting to call Tuesday's primary vote a split decision. Instead, political prediction markets have declared Senator Obama a clear winner.

It's not really tempting to call it a split decision, because Clinton needed to win both states by big margins to have even the slightest chance remaining. The fact that Obama had a net gain of delegates seals the deal. It's only a split in the sense that Clinton won the popular vote in Indiana, and that was a squeaker.

S. Heaton

Prof. Wolfers' post doesn't make much sense to me. I can't understand the point he's trying to make - did the prediction market mispredict in Indiana and NC? (YES). They don't seem to be predicting anything other than the ability to count votes on


Someone needs to tell Hillary it's over.
I still can't believe how long her party has allowed her to sabotage their chances in November.

This should have been a slam dunk election year for the Dems, and now it looks like it will be a nail biter.

The coming months will definitely be interesting.

Franchise Whale

Really enjoyed it, I wanted to click out and
you kept pulling me back in! Many thanks
and keep up the great work!

G. Owen Schaefer

I'm not so sure CBS necessarily erred simply because prediction markets hovered at 1-14 odds for Obama in Indiana when they called it for Clinton. That's because prediction markets (at least during voting nights) are only accurate in virtue of news organizations like CBS itself.

Consider: *very* few of the bettors are doing up-to-date statistical analysis of the results, or considering carefully the remaining voting population's propensities. I'm pretty sure most of them are, in fact, relying on an amalgamation of news sources to count such wagers. The fact that no one else other than CBS called the election is probably what led the dip to be only temporary.

While prediction markets are very good at measuring how various media as a whole are indicating winners, I don't think they can adequately reveal how one particular network was deficient in its analysis. Better to compare the actual methods CBS used with other networks, and see which is a better method.


Steve Schran

It is strange indeed that the markets move so much on primary results, as the information they provide is essentially unrelated to the likelihood of either candidate gaining the nomination. The probability of Hillary Clinton gaining the nomination through the current preferences of the electorate (both pledged and super delegates) has been approximately zero since Obama's run of victories after Super Tuesday (and arguably since the day after). That future primaries can't change this picture (all Rev. Wright issues being equal) is not a controversial notion. But it should also be obvious that the superdelegates will not come to Hillary's rescue.

Unless one is willing to make incredibly strong assumptions that the undeclared supers are closet Clinton fanatics (in which case, why haven't they declared for her already?) the probability that super-majorities will flock to her approximates zero. Suppose the preferences of undeclared supers are 50-50. Flip a fair coin 300+ times (once for each undeclared super). What is the chance of winning 2/3 of these? Approximately zero. And then there is the more likely fact (see Prof. Schaffner's work as reported on that Barack is likely to get 60% of the undeclareds. Now instead of a fair coin, roll a die 300+ times with Hillary winning on rolls of five or six. Approximately zero just got a lot smaller.

The bottom line is that the probability of a Clinton victory is equal to the probability of scandal toppling Obama. No new information, it seems to me, comes from election results that have no impact on the margin.


S. Heaton

Jirka, thanks for the great response to my comment. I think you're right about it being a strong predictor at 12 am. But at the time that the polls closed, wasn't intrade calling pricing at something like 90 for HRC? If that's the case, based on Wolfers' estimates, I guess she should have won by a large margin around 10%. Yet the market didn't really start moving in this direction until the Lake County results were coming in on TV. And at that point, I would argue, who cares whether the market is getting this right? The more important thing is what this market was doing compared to other methods like polling. In this case markets didn't seem to be doing any better than polling.


So there seems to be this contradiction in that the results were not THAT different from expected, yet Obama moved so drastically from 76% to 89%. The reason is that included in the 24% chance that the bettors were giving Clinton was the unlikelt possibility of a big upset, like NC going for Clinton. If Clinton won NC and IN, it's possible that Clinton's chances spike to 40% or higher.

Therefore, when the expected result happens instead of the improbable, that still shifts the overall results in favor of the expected winner, Obama.

A couple other notes. According to IEM, McCain is on 93% to win the Republican Nomination. Of course, the real likelihood is over 99%, but there are issues with liquidity and time value of money on futures that keep McCain bets below 99%. That said, Obama selling at 88% or 89% means that he's only about 5% less likely than McCain to be nominated.

Second, I would argue that even if Obama won by 6% in NC and lost by 8% in IN, he still would have improved in the markets.


John Phillips

The primaries in North Carolina and Indiana gave Hillary the chance to say something that would disqualify her from being hired as a CEO for any company--probably from being hired in any position in today's workplace.