Bryan Caplan  

What the Primaries Mean

More from Lester Thurow... Japan's catastrophic yen-quake...
Donald Trump has wildly exceeded electoral expectations at every stage of the election.  But betting markets still say he has but a 20% chance of winning the presidency.  Isn't Trump's record of success after success strong evidence that he's far more popular than normal measures suggest?  Most obviously, couldn't there be preference falsification at work on a mass scale?

Maybe, but there's a simpler story.  Trump has an enviable track record in the Republican primaries.  But hardly anyone votes in those primaries!  His victories notwithstanding, most of Trump's victories were backed by less than 10% of registered voters.  These results from late March omit the latest primaries, but still show the big picture.


Of course, the same goes for every primary candidate.  Hillary too has received few actual votes so far.  The key distinction: For well-established candidates, we have prior measures of their national appeal.  Not so for Trump.  Each of his victories has confirmed his unexpectedly strong appeal to politically active Republicans; it's no fluke.  But he's yet to demonstrate appeal to a broader audience, so betting markets remain the smart bet.

Disagree?  I'm happy to bet against Trump at even odds with my usual interface.

Comments and Sharing

COMMENTS (29 to date)
Sam Armstrong writes:

He's going to total more votes than any other republican presidential candidate in the history of primaries. Does that change your betting odds?

Andrew_FL writes:

Guess you're gonna have to cover your loss in the bet with Nathaniel Bechhofer somehow, since Trump losing of course means Hillary wins.

Of course, it's still possible that the GOP somehow pulls it together to deny Trump the nomination, but it's quite unlikely at this point. But I can't bet against you on this one. He's clearly going down harder than any candidate in recent memory. You'll be tickled pink, open borders is the big winner of this election. Well, open borders and socialism, of course. They go hand in hand.

E. Harding writes:

"But I can't bet against you on this one. He's clearly going down harder than any candidate in recent memory."

-No chance. 2008 was the perfect storm of electoral disaster for the GOP. And third time is, with one lone exception two hundred years ago, never the charm for the party controlling the Presidency. There is no conceivable way any Republican candidate could perform worse than McCain in 2008, even if he's Lucifer himself.

And Romney was a terrible candidate, having the stench of elitism and un-Americanism all over him. I couldn't see how anyone could vote for that guy. Turned out he won independents percentage-wise, but won fewer votes than Bush in 2004. That's because he thought up Obamacare, and expected us to believe that he had any intent of repealing it. Turned out he was too right-wing for Obama voters and too left-wing for true Republicans, thus leading to him having low enthusiasm among voters come November.

Trump is notoriously non-ideological, so he can bridge that partisan divide. He was a massive Bush critic in 2007, and is eager to dredge up any issue, no matter how low, to appeal to the lowest common denominator. Romney couldn't do that, because he was too elitist.

I do not expect Clinton to win, but it's close to even. There's no doubt Trump is currently unpopular among the general public, but Sanders has been notoriously weak at attacking Clinton. Trump has proven himself to be the candidate most able to stand up against his opponents. The theory is he'll attack Clinton from forty-five different directions every hour, thus crippling her favorability rating and driving her chances in battleground states into the sewer.

Trump's biggest hurdle is having to win a state that went less than 47% for Romney in 2012 to win the Presidency. Pennsylvania offers a really good opportunity, though, as it's been trending Republican. Trump also performs better in the primaries in the Eastern battleground states than any other candidate (except Kasich in Ohio).

Trump has been performing better among independents than self-identified Republicans, which also helps his chances in the general election.

A Trump-Kasich ticket would work wonders. Ohio would be a shoo-in, and only Florida and Pennsylvania, both trending Republican, and North Carolina, trending Democrat, would be serious battlegrounds. Kasich could even help the establishment vote in Virginia.

E. Harding writes:

Also, the number for Alabama is clearly wrong.

Jo VB writes:

At even odds? But then you turn around to take the opposite side with the bookmakers, rational as you are. Sounds more like arbitrage than betting.

Andrew_FL writes:

@E. Harding- You'll be taking the bet, then?

Josh Wexler writes:

Laying even odds isn't supporting your position very strongly. iirc, a Democrat winning the presidency was 3:2 well before Trump became the clear front runner. Given that the political context favored Dems from the outset, you ought to lay Trump at 40% at the shortest in order to prove your point.

You also should make the bet contingent on Trump winning the GOP nomination, as he's still at 10-15% to lose the nomination based on the possibility of party maneuvering at the convention (or at least bake this possibility into your wager and lay him at yet longer odds).

Josh Wexler writes:

...And even at 3:2 (or 3:1 for that matter), how fair is it to call Trump believers chicken when they turn down a price that's so far above what's being offered on the market?

Now, I'd certainly be dismissive if yoir gainsayers wouldn't take the betfair price. But why isn't it reasonable for them to reply to you, "I legitimately believe he has a 51% chance to win, but I'm not going to take your exorbitant price when I can get a much larger margin on even heavily juiced betting markets."?

Mike D writes:

[Comment removed pending confirmation of email address. Email the to request restoring this comment. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Phil writes:

you say his odds at 20% are right on, so you'll give odds of 50/50 if I disagree?

are you serious?

even if I did disagree, why would I give you a 30 point premium over the rest of the market?


I think the betting odds have fairly consistently been slightly too negative on Trump, I would put his odds at about 30%

if you want to bet on Trump at 30% odds, I'll do that up to $100

I'll give you a 10% premium over the market based on the public nature of your position,

if I really wanted return on my money, that's not competitive with what I could get from public liquid markets, but the bragging rights would give me a nice non-monetary reward, so I'd do that

a 30% premium is ludicrous

David R. Henderson writes:

Hillary too has received few actual votes so far. The key distinction: For well-established candidates, we have prior measures of their national appeal.
But Hillary is in the same boat. We have no prior measures of her national appeal. She, like Donald, has simply won primaries, this year and in 2008.
Also, it does look like an arbitrage play on your part. But give me substantially better odds and I’m interested.

Effem writes:

Closer to 40% Bryan (per predictit)

Brad writes:

I suggest adding a line to your bettors creed.

"I shall not offer odds well away from observable market and pretend it proves a point unless someone specifically states a probability different than market."

John Brennan writes:

If you give me 5:1 odds (based on the stated 20 percent Trump probability and your implication that the primaries mean nothing), I will bet you $100.

Andrew_FL writes:

People are citing PredictIt as more favorable to Trump neglecting to add its total odds for all candidates add up to over 100%

Richard writes:

So what if few people vote in the primaries? In theory, the markets and pundits should've known that and taken it into account.

Phil writes:



Jon Murphy writes:

Trump has an enviable track record in the Republican primaries. But hardly anyone votes in those primaries!

I'd add that those primaries are not the same question.

In the primaries, you're asking "Who is the better option: Trump, Cruz, Kasich?" In the general election, you're asking "Who is the better option: Trump or Clinton?"

Trump may be preferable to Cruz and/or Kasich, but he may not be preferable to Clinton (and vice versa: Clinton may be preferable to Sanders, but may not be preferable to Trump).

LD Bottorff writes:

Enviable record? By this point, he should have won several states with more than 50% of the Republican vote. I'm not convinced he's a Republican, but he may be slightly preferable to the other Democrat.

Glenn writes:
Each of his victories has confirmed his unexpectedly strong appeal to politically active Republicans; it's no fluke.

This is demonstrable false.

Trump's unexpected electoral success has come precisely because of his inability to attractive politically inactive independents, who are politically alienated but philosophically conservative. I don't know who the politically alienated philosophically liberal voters are doing this cycle (abstaining, maybe? Bernie? even Trump is plausible; I haven't cared enough to explore the question).

In that sense, the rise of Donald mirrors very well the unexpected rise of Obama eight years ago against Hillary (the favorite of politically active Democrats at the time).

Your making this claim presumably without even a cursory look at the data is a textbook example of confirmation bias at work. You have established beliefs about Republican primary voters, and distorted primary results to satisfy that belief.

If you look, though, you'll quickly find that the data do not support this. The easiest case does not require any detailed analysis of electoral turnout at all: Trump loses caucus states, which favor party stalwarts and dedicated voters, and wins primary states - particularly open primary states.

Andrew_FL writes:

@Jon Murphy-I would suggest the fact that the population of people who vote in Republican primaries being different from those who vote in the general election, is far more important than the fact that the primaries ask "a different question." Most of the people voting in the Republican primaries, presumably prefer any Republican to any Democrat. If you asked those voters the question of the general election, they'd probably favor the Republican overwhelmingly. The reason this says nothing about the general election is because people who do not vote in the Republican primaries cannot be expected to have their opinions being anywhere close to accurately predicted by assuming they track the sample of Republican primary voters.

Trump has done a lot to turn non GOP primary voters against the GOP this year. In fact he's done a lot to turn GOP primary voters off the GOP this year. I would not be surprised, on the off chance he won, that he would have negative coattails. Most optimistic scenarios for Trump almost certainly involve getting people to vote for him who will vote Democrat down ballot, and by bashing the GOP so much during the primary process, he's practically told them to.

With any luck the Coalition House and Democrat Senate would impeach and convict him on, you'll forgive the pun, Trumped up charges.

E. Harding writes:

"By this point, he should have won several states with more than 50% of the Republican vote."

-Pennsylvania, New York, Rhode Island, Connecticut, Delaware, Maryland, etc., aren't states?

If Hillary is inaugurated, I'll buy one of Caplan's books.

John S writes:

Once Trump locks up the GOP nomination with a win in Indiana (probability: 95%), his general election odds will improve in the markets as the GOP establishment coalesces around his campaign.

I know, prediction markets should already have taken this into account, etc etc. But Trump's campaign has repeatedly shown that betting market punters still have a lingering doubt about him until the results actually come in, which in turn gives him a spike in the markets.

This has happened several times after Trump won a primary he was widely expected to win. Doesn't make sense, but it has something to do the psychology of people's reactions to something that was wildly unexpected (last year, hardly anyone gave him a chance) actually happening. It takes some time for it to sink it that THIS IS FOR REAL.

But that feeling will largely disappear by November, when Trump's candidacy will be seen as "normal." His chances will rise to at least 40% on Predictwise by October (but I think 50%+ is unlikely).

blink writes:

Even odds, Bryan? Come on! If the betting markets are giving Trump 20%, your odds demands make you look like a stalwart supporter.

If not with your wallet, at least stick your neck out with a prediction: At betting market odds, which side do you take?

Jon Murphy writes:

@Andrew_FL, good point!

Andrew_FL writes:

@E. Harding-"Pennsylvania, New York, Rhode Island, Connecticut, Delaware, Maryland, etc., aren't states?"

By this point a front runner who's all but certain to be the nominee should be getting 60% of the vote at a minimum in the late contests. Romney and McCain were weak candidates, not well liked by many Republicans. But they managed that much. Trump's consolidating, but his consolidation is historically weak.

People seem very confident Trump can and will unite the party behind him. But they're all basing that on "business as usual" assumptions, the same kind of assumptions about this year that said Trump would never get anywhere. You can't say the conventional wisdom has been bad this year when it was against Trump, and then embrace the conventional wisdom when it supports him.

Christopher Chang writes:

At the time Caplan posted this, had Trump in the 20-22%, not 45-55%, range. Thus, the only reasonable interpretation of Caplan's offer to "bet against Trump at even odds" is that it's open until election day: he's either confident that Trump's betting market odds will never rise above 50%, or is comfortable being one side of an arbitrage opportunity in the event that they do.

If this isn't actually what Caplan means, and instead he's offering a completely vacuous proposition, he should clarify that soon.

ChrisA writes:

Has Bryan already bet for Trump in the online better market? If he bets 1 dollar for Trump at 20% odds in the online market, and against Trump at 1 dollar at 50% odds here, doesn't that mean if Trump wins he gets $5 less $1 stake on the second bet = $4 net? OTOH if Trump loses he gets $2 back from his second bet so his stake is covered. How can he lose? Have I got this wrong?

Rocinante writes:

He's only totaled more votes than other GOP primary candidates because meaningful contests went on longer. As a percent of the vote, Trump has the absolute lowest since primaries counted. He has 40% or so and others have 50 or 60%.

Comments for this entry have been closed
Return to top