Why do prediction markets seem more bullish on Biden than most of the news sources I read? I'm sure there must be a good reason, but the polls currently don't favor Biden, and I doubt perceptions of the economy are going to change much in the next 6 months. Is the bet that the trials will hurt Trump? That Biden still has room to consolidate his base?

(Serious question, I'm not just arguing that Biden is doomed)

@WilliamDewey Manifold is left-biased. Other prediction markets are not as bullish on Biden.

Metaculus:

Polymarket:

https://polymarket.com/event/presidential-election-winner-2024?tid=1715948026766

@WilliamDewey I think it's both of those things, plus the fact that it's still too far from the election to take polls at face value, and that RFK Jr. will likely lose support as we get closer to the election, as third-party candidates tend to, and will bleed support to Biden. It's worth noting that most polls don't have either candidate above 50% because there are too many undecided voters. I think if the polls still show Trump winning by the same margins on Labor Day, his predicted probability of winning will be much higher than it is now.

There's also polling errors you have to take into account. Even if the election were held today, Biden needs only a small polling error to beat Trump in Pennsylvania, Michigan, and Wisconsin (much smaller than the polling errors in previous elections). Of course, if you believe that there's some factor that makes the polls systematically underestimate Trump, as they did in 2016 and 2020, and believe that it's still true today, then maybe you don't care about that because you don't think there will be that kind of polling error. But historically, polling errors are random, and two elections is nowhere near a good enough sample size to prove that this trend is over. And if we factor in priors, it does seem likely that the polls are overestimating Trump right now, at least in some states (e.g., Nevada).

@Bair And yes, of course it's possible that it's all just wishcasting and all of these are retroactive justifications for a misguided probability estimate. But I do think that any news sources that look at the polls right now and conclude that Biden is definitely doomed are way overconfident.

@WilliamDewey (Imo) other sources are overweighting polls vs fundamentals a bit atm given distance to the election

@WilliamDewey polymarket is giving Biden 42%, with a betting volume of 128 million US dollars.

So efficient market hypothesis and all, I think you should take it more seriously than Manifold

@colorednoise When it comes to political campaigns with deep pockets that spend on perception because perception influences voter turnout, $128M total market capitalization ain't shit. In 2020 US presidential campaign spending was $3.16B for Democrats and less than $1B for Republicans. That doesn't include dark money pools... Expect a higher generalized spend this year.

@becauseyoudo is there a reason to expect that prediction markets cannot work on events that have a lot of money spent on them? Kalshi seems to do pretty well predicting interest rates which are a consequence of trillions in the macroeconomy.

@SemioticRivalry The value of accurately predicting future interest rates is a whole lot higher than the value of predicting the next president (trillions vs low billions). The value of publicly releasing those accurate predictions is also much higher for interest rates.

Lenders want borrowers to be able to plan for and pay back their debt. Political campaigns don't want the other side to know their plans, tactics or even their projected outcomes if possible. Political campaigns are also in the business of spending large sums of money to convince voters to cast a ballot on election day (perception drives voter turnout).

Since political campaigns have this higher preference for information asymmetry between each other and the voters they are trying to influence, it make sense that they would attempt to extend that asymmetry to polls, prediction markets and other sources of public information.

I'm not sure if its even possible to prevent the manipulation of public sources of information but increasing the cost of the manipulation (> market capitalization of the prediction market) seems like a reasonable place to start.

@becauseyoudo So your position is that political campaigns are spending millions of dollars to manipulate prediction markets? This is just clearly not happening. I think prediction markets are accurate for reasons totally orthogonal to the 'value' of the forecast, they're accurate because people have a strong incentive to be right.

@SemioticRivalry Why not? It seems like an easy spend to change the perception that your candidate is ahead, which is their function. The incentive to be right, or the incentive to make money from being right competes directly with the incentive to control the perception of future outcomes.

@becauseyoudo Very few people actually care about prediction market prices. There are many easier ways to spend millions of dollars to actually try to help your cause. And it's not necessarily obvious that you would want to show your candidate as being ahead. In 2016, many Democrats felt like they had already won the election, so they failed to actually vote.

@SemioticRivalry Good points but prediction markets are gaining in popularity.

https://www.forbes.com/sites/zacharyfolk/2024/05/14/peter-thiel-invests-in-polymarket-political-betting-platform-but-the-future-of-gambling-on-elections-remains-unclear/?sh=3f09787b5b80

Fun fact, you can get better prices for both Trump and Biden by betting here, as long as you don't think either of them will tweet the word "rationalussy"

brooo please joe

NYT sinema doesn't even move the needle by a bit?

What's the argument? Pools have been favoring republicans in the 22 midterms and 2023 special elections, so we discount them completely?

@MP I think it's less that they underestimated Republicans recently and more that it's still very far out from the election, and the responses still have tons of undecideds and third party supporters who will likely switch to one of the major candidates before Election Day.

Wtf is going on with the rust belt markets lol

okay now Kennedy is overvalued on this market. really wish he had gone up to 2% before the pivot was announced and I donated all my shares and mana.

Why are Kennedy and Obama so much lower here than on Polymarket?

@Bair Gambling addicts

@Shai ok, reasonable answer. Wouldn't you expect the same behaviour here after the pivot? Wouldn't it make sense to buy below 1% early and sell when gamblers drive it up to 5% or so?

@Bair Entry costs + Opportunity Cost against just investing in ETFs means that betting "NO" isn't profitable at a ~4% potential gain, so there isn't any pressure to push it down

@Bair at that point you are timing the market, i.e. "why don't I just get into this cryptocurrency early, and then sell it when other speculators drive it up"

@Bair It's just a silly guess. @StopPunting's take sounds more reasonable

@Bair Manifold usually has more accurate probabilities for low probability events because loans mean you can bet NO on them without tying up all your mana until resolution. So it can actually be profitable to bet on much more extreme probabilities than it is in real life. This won't work the same way after the pivot, but it sounds like Manifold is trying to come up with a replacement for loans.

@PlasmaBallin Yes. Not only is this effect amplified for long term markets, but it’s actually a pretty big issue with prediction markets in general. Like, with a 4% ROI (which is easy to make) you would make about 17% in 4 years, which means there is no incentive to bet below about 15% (or above 85%) on a 4 year market. And it is greatly diminished as the market gets near that value. Even for 1 year markets extreme probabilities are pointless.

@PlasmaBallin Yes, these are all valid reasons, but I still doubt they can explain 10x difference in levels, especially given that there will be no loans anymore.

@Bair The Manifold market had less liquidity so big users betting the top options was bringing the others down quickly. I think they might use a different formula to adjust the probabilities as well

@StopPunting there are still people buying NO for Newsom at 0.4 cents on Polymarket, even though it's much worse than investing in ETFs. There are no such people for Kennedy.

@Bair Then that’s just dumb idk

@ShadowyZephyr it is dumb of course. People make dumb decisions on markets all the time. My point is that there are such people willing to buy NO at really low chances for most candidates on Polymarket, but levels for RFK and Obama are significantly higher, and this is a signal.

@Bair Signal that RFK and Obama actually have a chance ? Do you really think so ?

@Odoacre I think that their chances are low, but certainly higher than 1%.

@Bair There is a higher chance that I get struck by lightning in the next five minutes than that either of them wins. And I'm inside right now.

@PlasmaBallin I think RFK has a chance of about 0.7% - say Biden has health issues and needs to drop, if he really wants to get Trump out he will endorse RFK since he is the only one who even has a chance of being Trump 1 on 1. Or Trump has to drop for some reason, but doesn't pick his VP.
Obama on the other hand, no chance. If Biden dies, Kamala Harris takes over the nomination, not her. I don't even think she wants to run.

@ShadowyZephyr If Biden has health issues and needs to drop, I would expect him to endorse Harris. Or for the Democratic Party to just pick a new nominee, who wouldn't be Kennedy.

@PlasmaBallin Harris doesn't win. There is almost no chance that anyone else could win. RFK already has the support to beat Trump 1 on 1, although it is unlikely (like maybe 25%).

@PlasmaBallin I agree that the Dems probably won't endorse RFK, but... what would you estimate as the probabilities of:

1. P(Biden has severe health problems and has to drop out shortly before the election)

2. P(Dems rally around RFK | Biden drops out)

3. P(RFK wins | Dems rally around RFK)

I'd estimate those at maybe 5%, 20%, and 70% respectively - so already >0.7% chance RFK wins. Unless you think one of those probabilities is several orders of magnitude below what I said, we're still way above lightning-strike-level chances. And that's not the only possible way he could win.

No way is 2. 20%.

@ZaneMiller if you really think 2. is that high, go make some mana here. 0.2% currently

@ZaneMiller you are delusional.
1. 3% at absolute most, I would say lower
2. 0%, similar chance of Michelle Obama or Oprah taking over in the next few months
3. 0%, smaller chance than Michelle Obama or Oprah

@ZaneMiller
1. 3%
2. 20%
3. 30%

Also possibilities for him if Trump drops out, one or both die, but lower. If both Trump and Biden die (~0.08%) he is very likely to win. My probability is steadily dropping though considering all the recent news articles about RFK hasn't converted into anything.

@Tumbles 3. is way >0%, he beats Trump (pop. vote) by 2 percentage points in national polls in a 1v1

@Bair So, I'm on Polymarket. I think Michelle Obama has <0.1% chance of winning the presidential election this year. I could bet NO on there, limit order at about 3.9%.
I still have to leave that money in there till at least November 5th, likely later if the election results are contested. So that's about half a year. That's 8% return on interest.
I can also put that money in an index fund and not worry about the risk of either USDC depegging or Polymarket still existing. There's no financial incentives towards fixing prices below about 7% per year.

@Fedor I understand all that. Yet there are people doing exactly what you don't want to do, betting against other candidates at 0.5% and lower.

@ShadowyZephyr There is easily a <1% chance RFK will 1v1 Trump. Multiply that by his chance of winning and you pretty much have 0%. Suggesting 30% means you are completely disconnected from reality.

@robm Realized I was making a bit of a mistake with my numbers there - conflating Biden dropping out between now and the election, with Biden dropping out just before the election.

When I said 20%, I was thinking in terms of what would happen if Biden suddenly died near the end of October. But it's not actually valid for me to multiply that 20% with the chance Biden drops out between right now and the election (because if Biden drops out right now, the DNC will almost certainly choose someone else, such as Harris.)

@Tumbles Are you familiar with the notation P(A|B) for probability that A happens, conditional on B happening? Are you saying, for number 3, that there is a ~0% chance that RFK wins given that Biden drops out and the Democrats support RFK instead? Because that was the number for which I said 70% and shadowyZephyr said 30% - not RFK's overall chance of winning.

Since hypothetical 1v1 polls between RFK and Trump show RFK a little ahead, I'm not sure why you would think he had a ~0% chance in that scenario.

@ZaneMiller Ah yes you're right, I was misreading the notation. In that case

1. <3%
2. <5%
3. <20%

Multiplying these together gives approximately 0%. Polls about how RFK would beat Trump are meaningless, people will say they would vote for ham sandwich over Trump when it's just a crazy hypothetical and not a real choice to consider. The democratic establishment would never rally around someone as wacky as stupid as RFK. If Biden drops dead they will go with Newsom or Harris or some other person in the vague shape of a real candidate.

@Tumbles Dem establishment will never, but Biden himself could endorse RFK if he thinks that is the best chance of beating Trump. And yes the polls are hypothetical, that's why it's 25-30% instead of 50%. That being said my probability is lowering

@ZaneMiller Lightning-strike level chances is perhaps an exaggeration but you have underestimated all three of those probabilities.

1. The probability that Biden drops out due to health reasons is probably something like 2% (based on his age, actuarial tables would give him something like a 3% chance of dying before the election, but his real chance is less than that. A health issue that prevents him from running but doesn't kill him is probably around the same probability).

2. The probability that he endorses Kennedy, given that he drops out, is extremely low, probably 1% at most. Even if he chose not to endorse Harris because he thinks she would lose, he would just endorse some other Democrat to be the party's nominee.

3. The probability that Kennedy wins even given all that is still a tossup at best. Maybe you could argue that it's a little higher than 50%, but I see no warrant for 70%. He's only ahead of everyone in head-to-head polls because he's the "Other" option. If he was actually the Democratic nominee, or the de-facto Democratic nominee, he's no longer the "Other" option and people would start to scrutinize his actual positions rather than just thinking, "I like him because he's not Trump or Biden." Plus, even if Biden drops out and endorses him, that doesn't mean Democrats will rally around him. I could easily imagine the Democratic Party just ignoring Biden's endorsement and choosing a new nominee, or a situation where many Democrats are already so turned of from Kennedy that they vote for Jill Stein or something.

If I combine all these probabilities, I get 0.01%, quite a bit lower than the market's current estimate.

@ZaneMiller Just realized I used slightly different things than you as probabilities 1, 2, and 3, but the overall probability is still the same.

@ShadowyZephyr I don't think RFK is likely to win if Biden and Trump both die. If it happens after Trump chooses a VP, then it probably becomes an election between Harris and Trump's running mate. Otherwise, the Republican Party would probably choose either DeSantis or Haley at the convention.

@PlasmaBallin I think 2 is above 1% because Kennedy already offered him the spoiler deal and I don't believe any other Dem who wants to run can win here. I mostly agree with your analysis though, it's surely below 1% total chance.

@ShadowyZephyr why can no other dem win? both trump and biden have historically terrible approval ratings

@SemioticRivalry all the people who would be running also have low approval ratings and wouldn't beat trump, especially harris. I just can't see them garnering the support they need. (well, it's possible but unlikely. I think Biden, if he drops, might just ask everyone to vote RFK)

@ShadowyZephyr What kind of person would vote for RFK over Trump, but wouldn't vote for Harris or Newsom over Trump?

@Tumbles Maybe an antivaxxer?

@Tumbles
1. Conspiracy theorists
2. Republicans who hate Trump ("Never Trumpers")
3. Other center-right Republicans annoyed by Trump's rhetoric
4. Center-right independents
5. People who vote for Independents as a protest vote because they hate the main parties
6. Libertarians