I have a long-standing dispute with Tyler Cowen about the epistemic value of betting. To my mind, my position is modest: Bets advance our knowledge by clarifying contentions and raising the price of error. While isolated bets don’t “prove” anything, they tip the argumentative scales in favor of the winner. And a long track record of successful betting is a strong sign of reliable judgment. Furthermore, the refusal to make bets at all shows that deep down, you know that you don’t know what you’re talking about.
In contrast, Tyler’s position is extreme: Bets have zero epistemic value.
Why would he think this? While Tyler has multiple arguments against the epistemic value of betting, his “nuclear option” is just “If you’re so foresighted, why aren’t you the proud owner of a successful hedge fund?” Instead of betting your friends that the unemployment rate will sharply fall in a few years, why haven’t you mortgaged your home to buy the S&P? (Better yet, buy option contracts that only pay if the S&P sharply rises). I’ve already heavily criticized this argument, but now I’d like to share yet another reason why Tyler is deeply mistaken.
Suppose I say, “Unemployment will fall to 4% by January 1, 2022.” You say I’m wrong – and use Tyler’s nuclear option against me. “If that’s right, you should leverage every penny you can on the S&P.” Sounds good, right?
Think again. It’s totally possible that the market already agrees with me! The current level of the S&P could be exactly the level you’d expect if the unemployment rate were destined to equal 4% on January 1, 2022. If so, my claims about unemployment and my disinterest in buying stock options are fully consistent.
In sharp contrast, it makes no sense to simultaneously hold my view on the path of unemployment and refuse to bet on unemployment. Why? Because as long as I’m right and you disagree, I will profit.
I don’t need to worry about “what the market has already figured out.”
I don’t need to understand a thousand other factors that influence the S&P.
I don’t need a Theory of Everything.
As long as I genuinely know one disputed fact about the future, and my opponents are intellectually honest, they will bet me and I will take their money.
End of story.
P.S. In 14 days, I expect my betting track record will hit 22 for 22.
READER COMMENTS
Oscar Cunningham
Jan 6 2021 at 10:46am
I agree that you win this one, but you should have imagined more possibilities when wording the bet, because Trump could still be removed from office in a future term.
Warren Platts
Jan 6 2021 at 9:31pm
lol! After today, it is somewhat not unlikely that he may be removed before Inauguration day!
Philo
Jan 7 2021 at 1:59pm
Caplan’s bet still looks good, but not nearly as good as it did a couple of days ago.
Garrett
Jan 6 2021 at 12:32pm
And even if you knew something the markets didn’t, it’s still hard to make money off of it. Knowing about every press release before the market does is still “only” worth a 77% hit rate.
J Mann
Jan 6 2021 at 1:03pm
True story: when I was a young teen, my brother and I placed 3 or 4 bets, each in cases where I was 100% sure I was right. Often, they were historical – who wrote such and such book, etc.
I lost all of the bets. For under $10, I learned (a) not to overestimate my own knowledge, and (b) that when my brother was willing to bet, he was probably right. A cheap lesson, even at 20th century prices.
Dylan
Jan 6 2021 at 1:12pm
I have a similar story betting with my step father, and our bets tended to be a little bigger, generally $100. I had a perfect record (which to be fair, was less than a dozen bets) before my step father learned a similar lesson.
But, even at a young age, I knew to only make bets about things that were already objective facts. Otherwise I took Yogi’s advice about predictions about the future, and have mostly refrained from making them.
Mark Z
Jan 6 2021 at 3:11pm
In these kinds of scenarios, the epistemic value of betting is due to the hidden knowledge of the party willing to bet. However, in the context of a debate, where all facts are readily verifiable, I think the epistemic value of betting wanes.
E.g., if two people without internet access disagree over what the capital of Turkey is, and one is willing to bet and the other isn’t, that counts a lot in favor of the former. But if two people are debating in a public forum over the effect of some policy and one is willing to bet and the other isn’t, it seems less likely that differences is because of some concealed knowledge that each side knows but isn’t sharing that would affect the outcome. It’s possible the side unwilling to bet is just arguing his position from social desirability bias, or that the issue is so complicated that there is inevitably a lot information left in the minds of experts that’s inaccessible to the public, but I could see betting willingness in policy debates being largely driven by how risk averse the participants are and not just how right they are, particularly when almost all bets are for nominal amounts.
Dylan
Jan 6 2021 at 1:05pm
This is what I take from your betting record, you essentially bet on the status quo being maintained and are able to find people that will generally give you even odds on this. That’s a good bet, and impressive in its own right, but it doesn’t mirror the way the world works very well. In the real world, those kinds of bets get you a small but steady return on your money, but the one time you’re wrong can wipe out all the gains you made on previous trades (I know this from experience at a hedge fund, fwiw)
What would impress me far more is if you took your betting prowess to a betting market where odds were determined by the wisdom of crowds. If you were able to continue your streak there in total winnings, then I think you would have something worth bragging about.
J Mann
Jan 6 2021 at 1:10pm
That’s a good reframing of the Caplan-Cowen disagreement, IMHO.
Caplan thinks its informative that he is generally (so far always) right when he bets against a set of individuals.
Cowen thinks it would only be informative if Caplan could successfully and consistently bet against the market.
If you’re an EMH proponent, you could say “I already use the market as my touchpoint; assuming that I’m correct to do so, then it’s not interesting that Caplan can beat individuals in betting, at least as long as Caplan is betting in the market’s direction and his counterparties are betting against it.”
Aaron G
Jan 6 2021 at 3:28pm
I admittedly haven’t followed any back and forth between Caplan and Cowen on the betting issue, but I doubt yours is the correct framing of their disagreement. It seems that Caplan would think it a very positive advance in discourse if the pundits would actively (and non-anonymously) participate in betting markets on the issues they speak so authoritatively about. If that were the case, you would certainly learn how strongly the pundits believe what they are saying, and it would establish a track record for their performance on concrete predictions.
DeservingPorcupine
Jan 6 2021 at 6:33pm
Well, Caplan’s betting still reveals something about the world, if nothing else the degree of overconfidence displayed on an individual level.
Daniel Carroll
Jan 6 2021 at 1:49pm
I believe Bryan Caplan uses betting markets to inform his bets, so betting “against” the market would be out of character.
The problem with betting is it’s binary nature. That leads to simplistic, mostly linear bets. Most states of the world are not binary, and nonlinearities are where the real money is made (and where the real information about the world resides). Most states of the world involve distributions of possibilities, so a binary bet is not usually informative. The presence of odds somewhat mitigates that problem, but only in the aggregate – somewhat, because we can never *know* if the odds were accurate, since odds by definition account for the possibility of loss.
The problem with betting markets is the lack of liquidity and the vulnerability to manipulation. The possibility of a non-linear / discontinuous event combined with human psychology will push betting markets “overprice” those events (“the lottery effect”, hubris), making for easy wins “most” of the time.
MarkW
Jan 6 2021 at 2:05pm
The absurd thing about the S&P retort is that Tyler is asking you to prove your mettle by betting against a betting market. You could have perfect knowledge about any number of things discussed by pundits and still not know what being right about those things would mean for the S&P. Suppose you knew two weeks ago that the Democrats were certain to win the Georgia runoff elections? What obvious couldn’t-lose bet would you have made about the direction of the S&P?
Incidentally, Tyler doesn’t bet because he thinks he might lose and it would harm his status (and status is obviously a key obsession for him). Most pundits gain status by making seemingly bold predictions while carefully inserting hedges to make sure they can never be pinned down and have to own their failures. But when you make bets, your opponents will refuse to accept your hedges, and that puts the ‘bold prediction’ pundit at a disadvantage relative to an opponent making the boring ‘nothing dramatic will happen’ prediction. So making bets is a no-win prospect for bold, ‘big thinker’ pundits like TC.
Michael Stack
Jan 7 2021 at 2:02pm
Tyler doesn’t seem to like betting, and while I don’t agree with his criticism of Bryan, it seems unfair to conclude that Tyler’s aversion to betting is simply fear of loss of status.
Phil H
Jan 7 2021 at 1:15am
“It’s totally possible that the market already agrees with me!”
If true, then presumably Cowen’s argument would be: then you don’t know anything. If your knowledge is identical to the knowledge of the market, then your claim is that you have zero insight.
Now, this is no bad thing. Most of us have zero insight on most topics. And being able to realise that you have zero insight is a very important bit of epistemological progress. And following the markets in that situation is clearly the correct thing to do, so you’re acting in the best way.
But it’s not anything to crow about, either. If that’s the case, then your winning of a bet amounts to nothing more and nothing less than finding a sucker. And again, I’d agree with the Cowen critique: If you’re good at finding suckers, find more of them and make enough money to retire already.
MarkW
Jan 7 2021 at 7:13pm
<i>If true, then presumably Cowen’s argument would be: then you don’t know anything.</i>
Are you actually arguing that no knowledge or prediction is useful unless it can forecast the future value of the S&P 500?
James
Jan 7 2021 at 10:54am
I think the best criticism of Tyler’s view on betting is that he has never given a single reason to believe it apart from repeated assertion.
If you have a view on whether Biden gets reelected you could express that through a bet on whether Biden gets reelected, or you could take a position in some stock index. Tyler should explain why the latter is a better way to express your conviction.
Even if your reasons to value betting are less than compelling, that still beats no reasons.
Alabamian
Jan 8 2021 at 9:25am
There can be epistemic value in betting, provided that you are betting with prices determined with a large liquid market. But that’s not what you do. You bet inconsequential sums of money in a small illiquid market with odds that you make up. That’s fine as an amusing diversion, but Hayekian observations about markets as information aggregators don’t apply.
To reframe it slightly: there isn’t a market price for the bets you are making, we don’t know what the market price would be, and there isn’t any reason to think that you have any knowledge or understanding than is superior to that hypothetical market. You seem to acknowledge as much, since you concede that a large liquid market like the S&P has probably already priced in your insights.
James
Jan 8 2021 at 4:11pm
Great point about the lack of liquidity but that’s a point against Cowen’s view. If the problem with betting is the lack of liquidity, then this is a complaint that bettors (who are providing some liquidity) should raise against non-bettors (who provide no liquidity).
Alabamian
Jan 8 2021 at 11:18pm
I don’t make any claim about whether there *should* be a liquid market or think that the existence or non-existence of one is a point for any side. I am just making the observation that we should attach a different epistemic value to someone beating a liquid market for real money versus someone making $50 personal bets for conjured odds.
As to the existence of a liquid market, there are plenty of reasons why you might not have one. It doesn’t take much consideration of insurance or financial markets to know that not every contract that *could* be made in theory *is* made in fact, presumably because the transaction costs eat up the value and make it an unattractive endeavor for one or both. The market is too small to justify the underwriting and administration, the stakes too low to justify the investment, etc.
On some level it’s an insoluble problem. There are a lot more people willing to plunk down $50 on the Steelers because they think they know something about football and it’s fun to watch the game than there are people wanting to bet Bryan on the direction of the EU so that they can find out if they win in three years…I make no moral judgment on either; but in terms of information discovery, small bets against individuals rather than markets odds don’t tell you much.
MarkW
Jan 9 2021 at 7:04am
But that’s not what you do. You bet inconsequential sums of money in a small illiquid market with odds that you make up. That’s fine as an amusing diversion, but Hayekian observations about markets as information aggregators don’t apply.
But that’s not the point. If there were liquid predictions markets, Caplan wouldn’t have to bet because those markets would do his work for him — by seeing which bets won and lost (and perhaps by winning and losing themselves), people would get a more accurate, less extreme, less panicky view of the world. Caplan’s ‘one simple trick’ is betting against against extreme events, predictions of doom, and against the world changing dramatically. And it’s interesting that that is almost always a winning bet.
Alabamian
Jan 10 2021 at 4:54pm
I disagree. Bryan’s claim is that “a long track record of successful betting is a strong sign of reliable judgment“, and that his track record of winning bets implies that his opinions are more likely to be correct. But the relevant question is “as compared to what?” And that’s the critical thing.
The sort of betting in which Bryan engages doesn’t tell us anything about that. What’s the domain of expertise? Over what group should I value Bryan’s opinion more highly? When should I listen to Bryan over an expert in the field? Should it affect my assessment of Bryan’s judgment that he apparently bets so as to optimize his chance of winning rather than optimizing for expected value? Why doesn’t Bryan take the odds more often?
There are easy ways for Bryan to validate his apparent claim to good judgment if he is serious about it. First note that, given the sums involved, Bryan’s skin in the game is reputational and not monetary. So why not do something like Scott Alexander? Bryan can make 50 or 100 predictions for 2021; assign them probabilities; post an assessment next year to see how he did.
AMT
Jan 10 2021 at 4:12am
This is by far the most important thing. Tyler’s argument is just garbage. He doesn’t bet because he just wants to seem smart and not commit to anything, because he’s (at least) smart enough to know he’s not that smart. As you said, “you know that you don’t know what you’re talking about.”
Comments are closed.