I think the third objection is the strongest - in addition to the ability to make goals about a good prediction market, you also need a constituency which cares about making aggressive prediction market goals and then reaching them. Need to create (and somehow enforce?) a culture.
My first thought is that this would work better if there was a bipartisan consensus on the need to reduce emissions. If Biden fails to meet the goals, what recourse do you have as someone who cares about climate change. "Blaming Biden" in this case seems likely to lead to less action to cut emissions, not more.
The interesting thing about prediction markets with a long term payout is that it's both a prediction about the thing in question but also an implicit prediction about inflation rates. If inflation is 10% a year for ten years, a bet that pays off at $2.59 at that point is breakeven with a dollar wagered now.
>If yes, that's the mechanism which corrects the market. Mitch Hedberg used to say that “escalators can’t break, they can only become stairs”; in the same way, prediction markets can’t break, they can only give you free money.
The lure of "free" money is strong, but it's not infinite - there are real costs both in time and capital investment for the participants. That gives you an upper bound on accuracy, depending on the effectiveness of competing investments and how many unsophisticated investors are in the market. (This was my rationale for exiting, yours too right?)
More broadly, if giving away that money is a cost someone's willing to pay, manipulating policy becomes straightforward. Maybe that price tag is going to be prohibitively high for things like presidential elections where there's massive engagement, but I can't believe there won't be narrow areas where manipulating projections becomes positive expected value. And then it's more or less just legalized bribery of the predictors.
The easiest win for prediction markets is using them to guide business decisions. Rather than Biden pledging to move the prediction markets, any ordinary CEO could do it (or better yet, companies could adopt Hanson's [fire the CEO markets](https://www.overcomingbias.com/2008/04/if-i-had-a-mill.html)). There are thousands in the United States that could do this, and probably hundreds of companies with quirky CEOs who would be willing to try. Yet as far as I can tell, no one has yet tried, since hardly anyone cares enough about non-legible expertise. Political prediction markets are a good idea, but unfortunately, we aren't even close yet.
> I trust the combination of this pressure plus a system of accountability to produce something kind of sane
From where I'm standing, it seems like the 50% goal is nowhere near the neighborhood of "something kind of sane". So I expect one of three things to happen under your proposal of using a betting market:
1. Environmental groups put up pressure, Biden doesn't make the announcement they want.
2. Environmental groups put up pressure, Biden makes the announcement they want, proceeds to ignore it.
3. Environmental groups put up pressure, Biden makes the announcement they want and achieves it, economy craters.
1. This skews the responsiveness of politicians towards the white/male/young/educated/affluent voters who are more likely to have the time, money, and energy to participate in predication markets, and we already have a way (campaign contributions) in which they exert (too much?) influence.
2. Since putting money at risk is the usual way we ensure that people who participate in prediction markets take their "vote" seriously, this heightens the influence of money on politics, and what we need is *less* of that, not more. Perhaps political campaigns would neutralize each other in the effort to buy the prediction -- but the amount of money they would spend doing so is unfortunate, and would have a regrettable influence on politics in general, e.g. the influence of campaign donors on political parties would get higher, which is the wrong direction.
3. Don't we already have a mechanism for this, in the form of ordinary political polling? Why add another mechanism which is more complicated, inherently suffers from self-selection bias, and which injects more money into the process?
Not saying I personally favor any of these objections, but I can see them being raised, and being effective.
Cynically, one could say that if the voters genuinely cared that Biden reduce US emissions, then they already have ways to hold him to that, and that if he in fact makes vacuous promises that he has no intention of keeping, it's because that strategy has the largest expected payoff for him -- which says something about what voters really want.
I think #2 is a stronger concern that you are giving it credit. Although it seems difficult to you at this moment, and you would likely make a lot of money in such a world, the system can definitely still be gamed.
One, it can be gamed in a kind of mundane way even in the exact mechanism you mention. Lets say that the "natural" environment of predictions says that Biden is likely to meet his goal by 2030, 53%. By putting a lot of money on a counterbet, his adversaries will tilt that number substantially. Even with lots of people buying up the "free money" options when it's 1%, that number will eventually go back up, and maybe sits at 45% instead of 53%. Now it looks like Biden is less likely to meet his goal. It's not free money from 45% to 53%, as even before the outside shocks Biden was almost equally likely to fail based on the prediction market. Betting enough to get it back to 53% would be expensive and quite possibly a losing proposition. Therefore the gaming of the system succeeded.
Two, you've made a measure of a goal the goal itself. There are lots of ways to affect that goal that have little to do with the underlying premises of his promise. For instance, you could hack the database. Or, more commonly, "make friends" with the owner of the prediction market and have them adjust the questions being asked to skew the results.
To make this work you would likely need billions of dollars on each question that are *not* ideologically spent, to set a baseline. You would also need very strong controls on who gets to ask the questions, who measures the response, etc. It would almost have to be a real part of the federal government.
I think this misses the point. At least "cut carbon emissions 50% by 2030" is a SMART goal, but it's still just a goal. The issue here is setting a goal is only the first step in the long, hard process of moving from a vision to a strategy and breaking that down into operational details with their own SMART sub-goals. Problem decomposition. It involves too many unaligned actors all working at odds with each other, and that will remain the case no matter how you change electoral promise culture.
Proposing that a goal, instead of being SMART, be about prediction markets instead, is trying to incentivize politicians to actually do the further hard work of problem decomposition, since that is what would move a prediction market.
But this isn't just about incentives. It's about feasibility. The complaint here is it won't be possible to hold Biden to task if the 2030 promise isn't held, because he'll definitely be out of office and will probably be dead by then. But if a shorter timebound promise about prediction markets isn't achieved, I still don't see what voters can do. In the general election, you can vote for the GOP candidate, who isn't going to promise to move prediction markets in the direction of lower emissions, but also isn't going to promise to lower emissions, which is presumably what voters really care about. Trivially, a politician can just promise nothing and then never fail to deliver. But voters want them to at least try to move policy in the direction voters wish. Trying and failing is no worse in terms of outcomes than not even trying.
There's the primary process, but even there, what exactly is the stark contrast? No matter who you vote for, they're likely picking the same staffers and they're stuck with the same Congress and the same agency permanent staffs, and I don't think a ton of novel ideas are going to change because one candidate wins a nomination instead of another. This stuff is so heavily determined by the parties themselves at this point that even someone as prima facie radical as Trump was barely able to move outside of standard GOP policy actions, at least in terms of actual legislation, not Twitter outbursts.
So if the goal of your proposal is more churn in who holds office, maybe you'll achieve that. But if you're trying to create a political climate in which politicians will actually be better able to achieve long term goals, I doubt it. There's a lot more than bad electoral incentives preventing that.
Fun idea. Moreover, if candidate pledges to move the target to 51%, and then you press them on why it hasn't happened, they will probably try to say, "well the market is just wrong / being manipulated / irrationally weighting unlikely risks / etc."
Which allows you to reply, "So how much of your own money have you put on the other side, so we can actually see your personal confidence in your own pledge?"
To which their natural rhetorical gambit will be, "Oh, but it's uncouth to gamble money on such things, and I'm offended you asked!" echoing the weird backlash that hit Nate Silver after he proposed a wager (for charity) with Joe Scarborough to make their confidence levels explicit.
At that point, though, if they're publicly rejecting the utility of prediction markets, then we can simply exclude the candidate from serious consideration (absent other overriding considerations).
Unlikely to have much effect on outcomes, given the microscopic size (and tbh lack of ideological cohesion) for the rationalist block. But might be self satisfying.
My objection isn't to the idea itself. If we really wanted to hold politicians accountable for their promises, then this is as good an idea as any I've heard. It'll never happen though. Obviously politicians have zero interest in being held accountable, but I don't see any evidence that voters have any interest in it either. Sure, there are activists who want to see results, but most voters these days would rather blame their opponents for obstructing progress. Actually making progress might get in the way of that, so they ignore all the broken promises.
People generally vote for candidates that have their own interests at heart. For example, "strength of the economy" often is the most important issue for voters according to polls. Policies that have long term benefits don't seem to be important to voters. Perhaps, a more effective approach (in terms of winning elections and achieving long term benefits) would be something like "achieving 51% likelihood of achieving long term climate goal, conditional on 100% short term likelihood of improving everyone's life"
> One way around this is to use a conditional prediction market - "what’s the probability that emissions will be below X, given that Biden's plan is enacted and stays in place until 2030?"
One problem is, if the probability the plan stays in place is close to 0, people have little incentive to place money-making bets, so the market gets dominated by people trying to move the numbers for political reasons.
The time value of money will severely reduce liquidity in longer-term markets if they're denominated in USD. Supposing you can earn 6% from an index fund, you need at least a 79% advantage on a 10 year bet for it to be worth betting. To get around this problem, the bets should be denominated in shares of SPY, or bitcoin.
Prediction markets will fail to correct proportional to the loss currently inherent in prediction markets: fees, inflation, loss of other potential returns, taxes on winning, regulations, and so on and so forth.
Yes, in the abstract, this idea would be fantastic. In viable real-world prediction markets, I suspect the friction involved would make objection #2 inevitable.
> Second objection: once the prediction becomes a target, aren't a lot of people going to try to distort it in order to make their candidate look good or bad? Probably. This is why we need good prediction markets.
Prediction markets will just become a new form of PR, just like only the analysts, pundits and think tanks that agree with a certain politician are used in their campgains.
So Biden won't campaign touting a prediction that makes him look bad, but will tout ones that are phrased in just the right way to make him look good compared to other candidates.
Then prediction markets will also probably get flooded with cash from PACs, distoring the prediction accuracy. Political campaigns have a lot cash to burn, and if prediction markets become meaningful, they'll burn their money there.
While I understand the intuitive appeal, there's a very, very, very obvious flaw in this plan:
> Second objection: once the prediction becomes a target, aren't a lot of people going to try to distort it in order to make their candidate look good or bad? Probably. This is why we need good prediction markets. If a prediction market is big enough and liquid enough, it should be robust against these kinds of distortion attempts.
Our _actual_ market is not resistant to these distortion attempts. More concerningly, a large amount of these distortions are _mandated by law_ and therefore inescapable within the United States.
If you started holding elected officials accountable to the results of a prediction market, I _promise_ you that arbitrary-seeming regulations on those markets will start getting passed that benefit incumbent politicians at the expense of newcomers.
You cannot engineer your way out of a public choice question. It can't be done. If someone claims to be accomplishing that, they are tricking you.
I think this is a wonderful idea. I'm skeptical that it would work in practice, because the idea strikes me as too subtle to capture the attention of a large share of voters. And if voters do not care, then why would politicians. Nor do I think it is in the interest of politicians to persuade voters of this idea. Politicians don't like accountability. But, assuming sufficiently rational voters, this is a very clever idea!
I just want to make a small remark about the first objection: I think that is really not a valid objection at all. You're essentially saying "what if Biden cannot do what he pledges, even if he really tries?". The natural response would be "don't make promises you cannot keep". So this is a feature, not a bug. (I understand that politicians may want to signal their intentions. But intentions are cheap - and hence bad signals.)
The status of prediction markets needs to be raised first. I'd guess most American voters don't know what prediction markets are, and, that among those who do, less than half view them as meaningful. (Most people are not good at thinking in terms of percentages.)
Does Joe Biden even know what prediction markets are?
This seems like you're primarily just fantasizing about a big public figure endorsing a thing that you like.
If ordinary people already took prediction markets seriously, then you could use them to hold Biden accountable whether he mentions them or not--in fact, in that case I suggest it's better if he doesn't, so that he avoids painting a bull's-eye on any single market that could be manipulated. The difference between that world and this world is that you're hoping that if Biden mentions Metaculus then more people will start taking Metaculus seriously.
(Also, it's really weird that you used Metaculus as your example when it is especially vulnerable to your objection #2.)
This feels like another instance of trying to find a technical solution to a social problem. If we geeks just develop the right technology/app/whatever, we can magically solve this big issue. (And I say this as a geek who tends to like technological solutions.)
Even if we successfully explain betting markets to the general public, even if we convince presidential candidates to make quantifiable pledges on said betting markets... none of this will make people care. They're still going to pick their presidential vote by whatever (irrational!) process they've been using all along.
This would only work if voters would be willing to vote based on who delivered on promised prediction markets outcome, and I think it is pretty obvious they are not willing to do that.
Australia, Brazil and others failed in their climate targets presumably because while promising action on climate is popular with (part of) the electorate, steps that would actually achieve promised emission reductions would be horrendously unpopular, given current culture and technology.
> "In order to get Metaculus (or some alternative prediction market) to show a 51% chance of meeting emissions targets, Biden would have to pass a credible package of legislation that puts us on the path to achieving that goal, and makes everyone think it’s more likely than not."
This is making a very optimistic assumption; the actual easiest way to get Metaculus (or some other prediction market) to show a 51% chance of meeting emissions targets would be to either nationalize the target market, or to in some other way subvert the owners of the company to altering their algorithm to show a different number.
The point of making promised using measurements about the things you care about rather than proxies is that it limits your realm of options; in a similar manner, if your goal is to improve the quality of life for the majority of Americans by improving their economic status, you could set a target that actually measures that, or you could set a target of "increase the value of the stock market" assuming that that is a sufficient measure of economic quality for peoples' lives. However, stock prices generally speaking *don't* directly affect most peoples' lives, and thus setting a target to improve that number opens you to a range of activities that are good for improving that target number, regardless of whether that improves the measure you care about or actually makes it worse.
Here's one way to bootstrap this. One year after a candidate makes a promise, the press could ask them how come the prediction markets aren't showing any confidence in said promises?"
I'm worried there may be a bad interaction between problems 2 and 1.
If the Republicans put a few hundred million at the market to drive the market prediction down and that gets the Biden unelected, they could dismantle his policies making their manipulated policy come true.
Overall I'm still cautiously optimistic about prediction markets.
Anybody interested in learning about how prediction markets work, and what it feels like to participate in one (you suddenly get VERY VERY GOOD at putting a number on your gut feelings of probabilities)...
...you might be interested to play Murder She Bet! https://murdershebet.com/ it's a game where you get together with friends and watch a murder mystery TV show, and buy and sell bets (using fake money) on who you think did it 🙂
As the game goes on you get new information, you "update", and adjust your bets accordingly, at any given time the market price is a representation of the info the audience has, etc. Originally designed by Hanson, implemented for the web by me, over the last few months!
"If Metaculus gives a lower number than this, we can consider Biden to have failed in his pledge, and we can hold it against him when he tries to get re-elected."
And then we get the "Oh Look, A Squirrel!" strategy from the Democratic Party. I don't think Biden is going to run again in 2024 but whoever the candidate might be (Kamala Harris? U.N. Known?) the party will simply go "Okay, so we didn't hit the targets. But that wasn't our fault! And besides Metaculus is biased against us! And who are you going to vote for instead, a Republican? Yeah, right! Only Trump-loving Trump-supporting Trump-voting types would reject a Democrat President! If you don't vote for our candidate, then Trump has won!"
All these Accords Held In World Capitals are simply political games. The people are saying that Something Must Be Done, so this is Something. By the time the bill comes due, we'll be out of power. Or maybe the horse will have learned to sing.
Biden is the first politician I've seen in a while who is taking the standard advice to under-promise and over-deliver, with all his vaccination claims - 1 million a day, and 100 million by the first 100 days, upped eventually to 150 million. Only the very last ask, of 70% of adults vaccinated by July 4, looks unlikely to be made, and that one was phrased as an ask, rather than a promise.
"Australia's 2009 Copenhagen summit pledge to decrease emissions 5% by 2020 (in fact, they increased 17%)."
Huh? I think you need to recheck your figures. Emissions are down, not up, on the Copenhagen context. The only sticking point is measuring Kyoto carry credits. Your results source covers only energy and cement production.
The relationship between whatever Australia was doing in 2009 and what it achieved in 2020 is also irrelevant to your thesis. Australia does not have term limits (Australia's longest serving PM spent 18 and a half years in office). The party which held power from the 2008 election produced a hung parliament in the 2010 federal election after infighting and lost convincingly in the 2013 election, in no small part due to the carbon tax it imposed in 2011.
Your examples must come from places with term limits to have any relevance to your thesis.
I think a cousin of the third objection is: even if you *can* hold a president accountable to a campaign promise, it doesn't do much. I don't think presidents really bother trying to avoid setting goals they can't be graded on, so much as they aren't judged about hitting the goal in the first place.
I like the idea, but I don't think the problem is that presidents' promises can't be evaluated. It's that most voters don't pay attention or care about them, and even if they did, it almost certainly won't be enough to change their opinion to vote for somebody else.
I ran a vaguely-similar idea by Robin Hanson earlier this year; aside from his work with Eliezer, he's also known as perhaps the most knowledgeable person around on the topic of Prediction Markets. The TL;DR of his feedback was, more or less: It could work, but only if someone has a strong incentive to pay for it. I got the sense that that was a very general objection and I get the sense that it's probably the main stumbling point here too.
This seems poorly suited to the US political system in particular. In a country where the party that wins the election can do what it wants until the next election, your proposal addresses a democratic accountability problem with pledges the party makes about future outcomes. But in the US, the major democratic accountability problem is that so many different actors have a say in policy that voters don't know who to credit/blame for outcomes they like/dislike. If you create the expectation that the president is always responsible for meeting specified goals and that failure has real political consequence, you increase incentives for obstruction and sabotage. These perverse incentives already exist but making them an even bigger element of our system seems bad.
"Stock markets can stay longer irrational than you can stay solvent" is a common saying among (value) stock investors.
So political prediction markets can surely stay longer irrational than politicians can stay in power.
Yes, people who recognize that irrationality might make a killing, just like Micheal Burry made a killing with shorting the housing market in 2008. That does not make the market sane.
This proposal is based on the not entirely explicitly stated premise that Biden's pledge is a promise made by him personally to the American people. And having noted the problems with that interpretation, you conclude that people are just Not Doing It Right. However, another conclusion one could draw upon seeing how problematic interpreting such pledges as personal promises is is to conclude that such an interpretation is not correct. To wit, Biden is the democratically elected representative of The American People, and Biden, as an agent of The American People, is pledging on their behalf to cut emissions. In making the pledge, Biden is asserting that he will plan his policies around this being a priority, and he is committing The American People to either cut emissions, or have their country be on bloggers' lists of Countries That Make Emission Pledge That They Totally Didn't Fulfill. If the electorate feels that Biden should be accountable for how far the country has come by the time end of his first term, they are free to make such an evaluation and vote accordingly regardless of whether there's a prediction market.
If you really want to have more responsive politicians -- and I'm not sure we do, there's a lot to be said for gridlock, unless you feel At One with the madding crowd 100% of the time -- then we should just do what business does: make executive pay largely bonuses. Every Presidential election year minus two we have a giant plebiscite in which The People vote on the bonuses they'll pay for certain objectives attained. Then everyone runs for office two years later and their pay is $0 (or whatever the practical de minimis is) + bonuses they achieve.
And if you *really* want to make the motivation effective on both sides, meaning both politicians and voters take the setting and achieving of goals seriously, then the way you vote for bonuses is you kick in your own money. So everyone who feels passionately about cutting CO2 emissions can throw $1000 of his own money into a pot to be divided among the Congressman, Senators, and/or President who achieve the specified goal within the specified time span.
"If Metaculus gives a lower number than this, we can consider Biden to have failed in his pledge, and we can hold it against him when he tries to get re-elected."
An interesting proposition, but of limited utility since most people don't seem to base their assessments of politicians on hard facts, sound projections, etc.
This is pretty close to Hanson's futurarchy proposals but importantly different since the market is the output not an input to a decision (so you avoid the worry about manipulatability and predictability by other states).
I don't think any of the objections raised are that convincing but actually getting this to work seems impossible. We can't even get the SEC to approve a prediction market with enough liquidity with real money to even have an underlying market.
Also no way you'll get progressives not to distrust anything this market oriented.
1). This seems like a feature not a bug. Biden can't control what a successor might do but he can adjust the amount he pledges to move the price to account for this (if he can only impact chance of hitting goal by 5% pick that goal).
2). This isn't a problem as it just creates arbitrage opportunities provided the market doesn't cap participation or positions.
3). Pols already make promises about higher stock market values. What's the difference?
15% is so preposterous that I was driven to create a Metaculus account to downvote it. So you hae successfully applied Cunningham's law to boost their userbase.
The second objection seems the strongest here. Given the importance of political victory to millions of people, and the staggering sums of money involved, free money might not be enough to put things back to true estimates.
The first and third can be dealt with fairly easily - a promise that's too weak is a joke, not appealing.
But I think you missed the fourth argument. People can't even understand the stock market as a prediction market, and think of it as being some kind of random money-stealing machine. Even some very smart people. How on earth will you get a dorky niche site like Metaculus to become big enough to be relevant to big-league politics? I doubt you could even get the Andrew Yang campaign to try that - they'd think it'd make them look like dweebs, never mind someone like Trump or Biden.
I think your point is valid, but at least one of the facts you used is incorrect because Australia did meet its 2020 pledge target. The outcomes data you link to does not include emissions from land use change while Australia's target was for total emissions including from land use change.
1. This is needlessly complicated. Right now, if a politician makes an old-fashioned campaign promise, the sort of nerds and wonks who trust prediction markets can use prediction markets to evaluate whether those promises are being kept and the politicians in question should be reelected. And everybody else can use whatever it is they already use for the same purpose. So the current system can extract whatever utility prediction markets can provide, whether prediction markets are a fringe thing only a few nerds and wonks care about or something as revered as the NYSE, or anywhere in between. And it can *also* extract whatever value politicians and their constituents feel they are getting from traditional campaign promises, which by their actions they seem to feel is substantial. Your version provides value *only* to the extent that prediction markets are respected, which would seem to make it strictly inferior.
2. In addition to the prospect of the Evil Billionaires(tm) on the Other Side using their billions to drive the prediction market in a way that makes the incumbents look bad, there's also going to be the Gamestop Brigade doing the same thing, on both side. "Our Guy(tm) needs us to go invest in the Our Guy's Telling The Truth portfolio, and if enough of us do this and stay the course, Our Guy wins *and* we all make lots of money from the Evil Billionaires who thought they could manipulate the market! Let's do this!"
3. The Gamestop Brigade will do this, and the Evil Billionaires will be believed to have done this whether they did it or not. And in the end, some of them may lose a lot of money for having done it. But *until* the end, each side has a perfectly good excuse to say, "the market would totally have proven us right except for all those cynical market manipulators rigging the game", and nobody is actually going to e.g. vote Biden out of office because the prediction markets were putting him at only 30% to reduce carbon emissions.
Goodhart's Law and Campbell's Law are real. If you think prediction markets are useful, or likely to be come useful in the future, then don't do this because it will break them. Particularly don't do it now, or any time before prediction markets are as large and robust as the NYSE.
I am intrigued by Scott Alexanders long-standing advocacy of prediction markets. He is an impressively clever guy, so I am totally open to the possibility that there is something vital about such markets I am too dim - or uninformed - to get.
...and with such an intro there is bound to be a but, of course.
Here is my but:
...but is not the only way a prediction market can yield important/useful information about what is likely to happen in the future, a prediction market limited to bettors that have SERIOUS skin in the game?
To elaborate: If betting is restricted to those willing to put 10.000 USD behind their bet (that's just a number - they point is that it must be a serious amount of money for most people), people are incentivized to put in information-gathering costs at least equal to USD 10.000, to have an "enlightened" opinion on what they are betting to happen.
If, by contrast, people can bet sums upwards from USD 1 (or even nothing), the whole betting schtick will be swamped by non-informed bettors who bet just for the fun of it. It will be impossible to distinguish the signal (the bets of seriously informed bettors) from the noise (the uninformed bets by everyone who participate just for the fun of it).
If I am right here, and we - to avoid that - limit betting to those wo have put serious money where their mouth is, we unfortunately run into another problem:
....Most policy-related decisions/events are driven by a ton of factors (think Brexit or the fall of the USSR), which make the random component of what is going to happen, much larger than in most business-related bets (will the share value of Amazon go up or down next year and suchlike). Implying that no sane, ordinary person will bet as much as USD 10.000 on any policy-related future decision/issue taking place.
...Which in its turn implies that you will get a selection effect, where the only bettors willing to back their bets by USD 10.000 or more, are either non-sane people (whose predictions you for a variety of reasons should not have much confidence in); or those rich enough to regard USD 10.000 as a trivial sum to spend just for the fun of it (whose predictions you should not have much confidence in either, since they are so rich they do not have an incentive to get themselves an informed opinion before they place their bet).
That's my problem(s) with the assumed usefulness of prediction markets. But again, Scott Alexander is an extraordinary clever person. What do I miss? Where do I go wrong in the above line of reasoning?
And for the record: I am not at all trying to be smug or anything! Really. I am genuinely interested in whether, and if so why, the above objections miss the mark; or whether there are some benefits to prediction markets I do not see, that trumps these objections.
I think first you would have to make the average voter aware that there is such as thing as prediction markets, what they do and how the current ones are weak. This should take at least the rest of Biden's first term (since I don't think the average voter knows already, this seems safe).
Also "figuring out how to prevent your successors from dismantling it". Judging by what currently is bulletproof, one needs to craft legislation that gives every legislator something. For instance, the current programs funding NASA seem to be mostly interested in funding something in every state that could block it. Or look at whose ox would be gored if any given welfare rule is changed - the ones that are not changed are the ones where any change hurts someone who can block the change.
Trying to hold people accountable for their work results runs into all kinds of trouble, including campbell's law/goodhart's law. The more complex, and difficult to evaluate the work results are, the worse the troubles will be.
If we hold politicians accountable for a number of easily quantifiable goals, they will pursue these goals at the cost of everything else. And cook the books. And play all kinds of nasty games with blame-shifting and sabotaging their opponents.
I think we would get better results if we were to hold politicians LESS accountable.
I would love for this to work, but I think that looking at actual IRL stock markets suggests that it wouldn't.
Politicians do in fact pledge to grow the economy and stock market, but don't consistently do so. Instead, they claim success whenever it goes up and blame their opponents whenever it goes down. Since the economy follows semi-predictable but difficult to time business cycles, it's hard to say whether they're even having an effect or not.
In individual stocks, we see all sorts of BS. It's more profitable to try and cash in on a bubble than to short it. Stock markets seem to largely function as a Keynesian beauty contest with only occasional reality checks. "Crazy" fluctuations based on memes, passing news stories, people misreading the name of the stock, etc are *entirely rational for investors to buy into* as long as they think some fraction of people are idiots they can fleece, amplifying rather than suppressing the dumbest voices.
End result? At the end of Biden's presidency, some news story comes out that's bad for his climate targets. Rational investors judge that it's nonsense, but some fraction of people will believe it, so they buy tons of "Biden won't hit his target" contacts, sending the price soaring (and vice versa for the opposite contract). They're right, and morons start buying it, sending the price even higher. Trump supporters start a meme to #HODL "Biden plan won't hit climate targets", diamond hands will win Trump 2024! The price climbs to absurd levels suggesting Biden will personally destroy the planet; this becomes it's own news story, driving the price still higher.
Biden's term ends and he's narrowly reelected, but he blames his failure to hit the prediction market targets on Trump supporters gaming the market and other factors outside his control. Prediction markets are declared racist.
A year later, the smart money quietly sells at the peak of the market. The dumb money panics and sells, mostly at a loss, and the bubble collapses.
In your discussion of Objection 2, you bring up endogeneity ("The Republicans will spend money to artificially deflate the prediction to make Biden look bad and lose re-election") but then seem to ignore it later on in the same discussion ("If the prediction market number is low, everyone will buy because it'll be free money!").
Won't it only be "free money" if the market ends up high enough for Biden to win and continue enacting his policies? Doesn't that in alter the prediction market into a dollar-based popularity contest, or possibly something with similar dynamics to a "dollar auction"?
The first objection is a feature not a bug. Presumably, a market would put greater weight on policies with broad support and less weight on policies enacted by slim temporary majorities. Exactly the sort of compromise policy making that the wings of both (US) major parties have discouraged over the past decade or so.
Would this not produce a huge incentive for obstruction? It's almost like for this to work, we need to replace checks and balances with prediction markets?
Nitpick, but let me take issue with the claim that Biden will probably be dead by 2030. He is currently 78. According to this table (https://www.ssa.gov/oact/STATS/table4c6.html) he would be expected to live an other 9.37 years, which puts us well into 2030. He is probably healthier than the average 78-year-old, so I'd say he will probably still be alive in 2030.
When I explained to a friend how absurdly expensive the Green New Deal would be, she replied that it wasn't actually a real goal, it was like haggling--you always open with an offer you don't expect to be accepted. It also has to be too extreme in order to "motivate the base". Also, achieving politically divisive goals would reduce the fanaticism of your own side, and increase that of the other side.
Holding politicians accountable via prediction markets would be great, but I'm generally more worried about politicians achieving their stupid grand goals than about not achieving them. If by monitoring prediction markets as they monitor opinion polls, politicians learn how to achieve their goals, that could be a net loss.
"If a prediction market is big enough and liquid enough, it should be robust against these kinds of distortion attempts."
Prediction markets are robust against people trying to change them by *trading inside* the market, for exactly the reason stated (outcome A still pays off an expected ~$1, outcome B ~$0).
However, they are not robust against a credible promise to distribute a subsidy proportionately among holders of outcome B, because doing so legitimately changes the expected value of holding that position. To what degree it changes the price depends on how much capital is in (or can get into) the market, how credible the promise is, how many people hear about it in the first place, etc.
This can be combined with taking a position in whatever the prediction market controls, if it's directly hooked up to something of interest, to pump more than enough money out of the system to cover the subsidy plus profit. E.g. there used to be calls to replace the Fed with a prediction market -- in that case, one can take a position in the forex market and then subsidize the predictiFed to pull the price in the appropriate direction.
"Imagine Biden pledges that some prediction market will have a 51% chance of reaching his 2030 emissions target by the time he leaves office. He passes a carbon tax, and the market shoots up from 15% to 30%. Now he knows he’s on the right track, but still has to do more. So he bans a bunch of coal power plants, and it goes up to 45%. He's still not quite there, so he gives big subsidies to solar panels a few days before the campaign season kicks off, the prediction market reaches 51%, and he's able to say he fulfilled his pledge."
This feedback loop is almost entirely blunted if predictors expect Biden to actually care about the prediction market. Once Biden makes that pledge, the prediction market should take it mostly for granted that he's going to take reasonable measures to hit the goal (if possible). If your example's numbers are reasonable absent a pledge, then the market should take for granted that he'll pass a carbon tax, etc. Just making that pledge will cause the market to trade way over 15%, over 30% if he's perceived as serious. If he passes the carbon tax and the market just sits still at 45%, it doesn't provide him much information. If he passes a useless bill, maybe the market ticks down from 45% to 44% or something, but it should price in that he's going to come around to a better bill eventually.
"accept that part of the work of making a global warming plan is figuring out how to prevent your successors from dismantling it"
There is the issue that this would create an even stronger incentive to find ways to lock in policies in ways that violate the spirit of the Constitution or law.
I think the third objection is the strongest - in addition to the ability to make goals about a good prediction market, you also need a constituency which cares about making aggressive prediction market goals and then reaching them. Need to create (and somehow enforce?) a culture.
Is this a Scott Sumner guest post? Seriously though, using deep/efficient prediction markets to hold government policy to account is a great idea.
My first thought is that this would work better if there was a bipartisan consensus on the need to reduce emissions. If Biden fails to meet the goals, what recourse do you have as someone who cares about climate change. "Blaming Biden" in this case seems likely to lead to less action to cut emissions, not more.
The interesting thing about prediction markets with a long term payout is that it's both a prediction about the thing in question but also an implicit prediction about inflation rates. If inflation is 10% a year for ten years, a bet that pays off at $2.59 at that point is breakeven with a dollar wagered now.
>If yes, that's the mechanism which corrects the market. Mitch Hedberg used to say that “escalators can’t break, they can only become stairs”; in the same way, prediction markets can’t break, they can only give you free money.
The lure of "free" money is strong, but it's not infinite - there are real costs both in time and capital investment for the participants. That gives you an upper bound on accuracy, depending on the effectiveness of competing investments and how many unsophisticated investors are in the market. (This was my rationale for exiting, yours too right?)
More broadly, if giving away that money is a cost someone's willing to pay, manipulating policy becomes straightforward. Maybe that price tag is going to be prohibitively high for things like presidential elections where there's massive engagement, but I can't believe there won't be narrow areas where manipulating projections becomes positive expected value. And then it's more or less just legalized bribery of the predictors.
The easiest win for prediction markets is using them to guide business decisions. Rather than Biden pledging to move the prediction markets, any ordinary CEO could do it (or better yet, companies could adopt Hanson's [fire the CEO markets](https://www.overcomingbias.com/2008/04/if-i-had-a-mill.html)). There are thousands in the United States that could do this, and probably hundreds of companies with quirky CEOs who would be willing to try. Yet as far as I can tell, no one has yet tried, since hardly anyone cares enough about non-legible expertise. Political prediction markets are a good idea, but unfortunately, we aren't even close yet.
> I trust the combination of this pressure plus a system of accountability to produce something kind of sane
From where I'm standing, it seems like the 50% goal is nowhere near the neighborhood of "something kind of sane". So I expect one of three things to happen under your proposal of using a betting market:
1. Environmental groups put up pressure, Biden doesn't make the announcement they want.
2. Environmental groups put up pressure, Biden makes the announcement they want, proceeds to ignore it.
3. Environmental groups put up pressure, Biden makes the announcement they want and achieves it, economy craters.
The main objections I can see people making are:
1. This skews the responsiveness of politicians towards the white/male/young/educated/affluent voters who are more likely to have the time, money, and energy to participate in predication markets, and we already have a way (campaign contributions) in which they exert (too much?) influence.
2. Since putting money at risk is the usual way we ensure that people who participate in prediction markets take their "vote" seriously, this heightens the influence of money on politics, and what we need is *less* of that, not more. Perhaps political campaigns would neutralize each other in the effort to buy the prediction -- but the amount of money they would spend doing so is unfortunate, and would have a regrettable influence on politics in general, e.g. the influence of campaign donors on political parties would get higher, which is the wrong direction.
3. Don't we already have a mechanism for this, in the form of ordinary political polling? Why add another mechanism which is more complicated, inherently suffers from self-selection bias, and which injects more money into the process?
Not saying I personally favor any of these objections, but I can see them being raised, and being effective.
Cynically, one could say that if the voters genuinely cared that Biden reduce US emissions, then they already have ways to hold him to that, and that if he in fact makes vacuous promises that he has no intention of keeping, it's because that strategy has the largest expected payoff for him -- which says something about what voters really want.
I think #2 is a stronger concern that you are giving it credit. Although it seems difficult to you at this moment, and you would likely make a lot of money in such a world, the system can definitely still be gamed.
One, it can be gamed in a kind of mundane way even in the exact mechanism you mention. Lets say that the "natural" environment of predictions says that Biden is likely to meet his goal by 2030, 53%. By putting a lot of money on a counterbet, his adversaries will tilt that number substantially. Even with lots of people buying up the "free money" options when it's 1%, that number will eventually go back up, and maybe sits at 45% instead of 53%. Now it looks like Biden is less likely to meet his goal. It's not free money from 45% to 53%, as even before the outside shocks Biden was almost equally likely to fail based on the prediction market. Betting enough to get it back to 53% would be expensive and quite possibly a losing proposition. Therefore the gaming of the system succeeded.
Two, you've made a measure of a goal the goal itself. There are lots of ways to affect that goal that have little to do with the underlying premises of his promise. For instance, you could hack the database. Or, more commonly, "make friends" with the owner of the prediction market and have them adjust the questions being asked to skew the results.
To make this work you would likely need billions of dollars on each question that are *not* ideologically spent, to set a baseline. You would also need very strong controls on who gets to ask the questions, who measures the response, etc. It would almost have to be a real part of the federal government.
I think this misses the point. At least "cut carbon emissions 50% by 2030" is a SMART goal, but it's still just a goal. The issue here is setting a goal is only the first step in the long, hard process of moving from a vision to a strategy and breaking that down into operational details with their own SMART sub-goals. Problem decomposition. It involves too many unaligned actors all working at odds with each other, and that will remain the case no matter how you change electoral promise culture.
Proposing that a goal, instead of being SMART, be about prediction markets instead, is trying to incentivize politicians to actually do the further hard work of problem decomposition, since that is what would move a prediction market.
But this isn't just about incentives. It's about feasibility. The complaint here is it won't be possible to hold Biden to task if the 2030 promise isn't held, because he'll definitely be out of office and will probably be dead by then. But if a shorter timebound promise about prediction markets isn't achieved, I still don't see what voters can do. In the general election, you can vote for the GOP candidate, who isn't going to promise to move prediction markets in the direction of lower emissions, but also isn't going to promise to lower emissions, which is presumably what voters really care about. Trivially, a politician can just promise nothing and then never fail to deliver. But voters want them to at least try to move policy in the direction voters wish. Trying and failing is no worse in terms of outcomes than not even trying.
There's the primary process, but even there, what exactly is the stark contrast? No matter who you vote for, they're likely picking the same staffers and they're stuck with the same Congress and the same agency permanent staffs, and I don't think a ton of novel ideas are going to change because one candidate wins a nomination instead of another. This stuff is so heavily determined by the parties themselves at this point that even someone as prima facie radical as Trump was barely able to move outside of standard GOP policy actions, at least in terms of actual legislation, not Twitter outbursts.
So if the goal of your proposal is more churn in who holds office, maybe you'll achieve that. But if you're trying to create a political climate in which politicians will actually be better able to achieve long term goals, I doubt it. There's a lot more than bad electoral incentives preventing that.
Fun idea. Moreover, if candidate pledges to move the target to 51%, and then you press them on why it hasn't happened, they will probably try to say, "well the market is just wrong / being manipulated / irrationally weighting unlikely risks / etc."
Which allows you to reply, "So how much of your own money have you put on the other side, so we can actually see your personal confidence in your own pledge?"
To which their natural rhetorical gambit will be, "Oh, but it's uncouth to gamble money on such things, and I'm offended you asked!" echoing the weird backlash that hit Nate Silver after he proposed a wager (for charity) with Joe Scarborough to make their confidence levels explicit.
At that point, though, if they're publicly rejecting the utility of prediction markets, then we can simply exclude the candidate from serious consideration (absent other overriding considerations).
Unlikely to have much effect on outcomes, given the microscopic size (and tbh lack of ideological cohesion) for the rationalist block. But might be self satisfying.
This feels related to (though not identical to) http://mason.gmu.edu/~rhanson/futarchy.html / http://mason.gmu.edu/~rhanson/futarchy2013.pdf, which Scott has also talked about before.
You walk up the escalator stairs so you can get into the Sandwich Club.
My objection isn't to the idea itself. If we really wanted to hold politicians accountable for their promises, then this is as good an idea as any I've heard. It'll never happen though. Obviously politicians have zero interest in being held accountable, but I don't see any evidence that voters have any interest in it either. Sure, there are activists who want to see results, but most voters these days would rather blame their opponents for obstructing progress. Actually making progress might get in the way of that, so they ignore all the broken promises.
People generally vote for candidates that have their own interests at heart. For example, "strength of the economy" often is the most important issue for voters according to polls. Policies that have long term benefits don't seem to be important to voters. Perhaps, a more effective approach (in terms of winning elections and achieving long term benefits) would be something like "achieving 51% likelihood of achieving long term climate goal, conditional on 100% short term likelihood of improving everyone's life"
> One way around this is to use a conditional prediction market - "what’s the probability that emissions will be below X, given that Biden's plan is enacted and stays in place until 2030?"
One problem is, if the probability the plan stays in place is close to 0, people have little incentive to place money-making bets, so the market gets dominated by people trying to move the numbers for political reasons.
The time value of money will severely reduce liquidity in longer-term markets if they're denominated in USD. Supposing you can earn 6% from an index fund, you need at least a 79% advantage on a 10 year bet for it to be worth betting. To get around this problem, the bets should be denominated in shares of SPY, or bitcoin.
Prediction markets will fail to correct proportional to the loss currently inherent in prediction markets: fees, inflation, loss of other potential returns, taxes on winning, regulations, and so on and so forth.
Yes, in the abstract, this idea would be fantastic. In viable real-world prediction markets, I suspect the friction involved would make objection #2 inevitable.
Look at the difficulties encountered by somebody who is at the literal pinnacle of understanding how to avoid prediction market friction: https://vitalik.ca/general/2021/02/18/election.html
> Second objection: once the prediction becomes a target, aren't a lot of people going to try to distort it in order to make their candidate look good or bad? Probably. This is why we need good prediction markets.
Prediction markets will just become a new form of PR, just like only the analysts, pundits and think tanks that agree with a certain politician are used in their campgains.
So Biden won't campaign touting a prediction that makes him look bad, but will tout ones that are phrased in just the right way to make him look good compared to other candidates.
Then prediction markets will also probably get flooded with cash from PACs, distoring the prediction accuracy. Political campaigns have a lot cash to burn, and if prediction markets become meaningful, they'll burn their money there.
While I understand the intuitive appeal, there's a very, very, very obvious flaw in this plan:
> Second objection: once the prediction becomes a target, aren't a lot of people going to try to distort it in order to make their candidate look good or bad? Probably. This is why we need good prediction markets. If a prediction market is big enough and liquid enough, it should be robust against these kinds of distortion attempts.
Our _actual_ market is not resistant to these distortion attempts. More concerningly, a large amount of these distortions are _mandated by law_ and therefore inescapable within the United States.
If you started holding elected officials accountable to the results of a prediction market, I _promise_ you that arbitrary-seeming regulations on those markets will start getting passed that benefit incumbent politicians at the expense of newcomers.
You cannot engineer your way out of a public choice question. It can't be done. If someone claims to be accomplishing that, they are tricking you.
I think this is a wonderful idea. I'm skeptical that it would work in practice, because the idea strikes me as too subtle to capture the attention of a large share of voters. And if voters do not care, then why would politicians. Nor do I think it is in the interest of politicians to persuade voters of this idea. Politicians don't like accountability. But, assuming sufficiently rational voters, this is a very clever idea!
I just want to make a small remark about the first objection: I think that is really not a valid objection at all. You're essentially saying "what if Biden cannot do what he pledges, even if he really tries?". The natural response would be "don't make promises you cannot keep". So this is a feature, not a bug. (I understand that politicians may want to signal their intentions. But intentions are cheap - and hence bad signals.)
The status of prediction markets needs to be raised first. I'd guess most American voters don't know what prediction markets are, and, that among those who do, less than half view them as meaningful. (Most people are not good at thinking in terms of percentages.)
Does Joe Biden even know what prediction markets are?
This seems like you're primarily just fantasizing about a big public figure endorsing a thing that you like.
If ordinary people already took prediction markets seriously, then you could use them to hold Biden accountable whether he mentions them or not--in fact, in that case I suggest it's better if he doesn't, so that he avoids painting a bull's-eye on any single market that could be manipulated. The difference between that world and this world is that you're hoping that if Biden mentions Metaculus then more people will start taking Metaculus seriously.
(Also, it's really weird that you used Metaculus as your example when it is especially vulnerable to your objection #2.)
This feels like another instance of trying to find a technical solution to a social problem. If we geeks just develop the right technology/app/whatever, we can magically solve this big issue. (And I say this as a geek who tends to like technological solutions.)
Even if we successfully explain betting markets to the general public, even if we convince presidential candidates to make quantifiable pledges on said betting markets... none of this will make people care. They're still going to pick their presidential vote by whatever (irrational!) process they've been using all along.
Or maybe I'm just too cynical?
This would only work if voters would be willing to vote based on who delivered on promised prediction markets outcome, and I think it is pretty obvious they are not willing to do that.
Australia, Brazil and others failed in their climate targets presumably because while promising action on climate is popular with (part of) the electorate, steps that would actually achieve promised emission reductions would be horrendously unpopular, given current culture and technology.
> "In order to get Metaculus (or some alternative prediction market) to show a 51% chance of meeting emissions targets, Biden would have to pass a credible package of legislation that puts us on the path to achieving that goal, and makes everyone think it’s more likely than not."
This is making a very optimistic assumption; the actual easiest way to get Metaculus (or some other prediction market) to show a 51% chance of meeting emissions targets would be to either nationalize the target market, or to in some other way subvert the owners of the company to altering their algorithm to show a different number.
The point of making promised using measurements about the things you care about rather than proxies is that it limits your realm of options; in a similar manner, if your goal is to improve the quality of life for the majority of Americans by improving their economic status, you could set a target that actually measures that, or you could set a target of "increase the value of the stock market" assuming that that is a sufficient measure of economic quality for peoples' lives. However, stock prices generally speaking *don't* directly affect most peoples' lives, and thus setting a target to improve that number opens you to a range of activities that are good for improving that target number, regardless of whether that improves the measure you care about or actually makes it worse.
Here's one way to bootstrap this. One year after a candidate makes a promise, the press could ask them how come the prediction markets aren't showing any confidence in said promises?"
I'm worried there may be a bad interaction between problems 2 and 1.
If the Republicans put a few hundred million at the market to drive the market prediction down and that gets the Biden unelected, they could dismantle his policies making their manipulated policy come true.
Overall I'm still cautiously optimistic about prediction markets.
Anybody interested in learning about how prediction markets work, and what it feels like to participate in one (you suddenly get VERY VERY GOOD at putting a number on your gut feelings of probabilities)...
...you might be interested to play Murder She Bet! https://murdershebet.com/ it's a game where you get together with friends and watch a murder mystery TV show, and buy and sell bets (using fake money) on who you think did it 🙂
As the game goes on you get new information, you "update", and adjust your bets accordingly, at any given time the market price is a representation of the info the audience has, etc. Originally designed by Hanson, implemented for the web by me, over the last few months!
"If Metaculus gives a lower number than this, we can consider Biden to have failed in his pledge, and we can hold it against him when he tries to get re-elected."
And then we get the "Oh Look, A Squirrel!" strategy from the Democratic Party. I don't think Biden is going to run again in 2024 but whoever the candidate might be (Kamala Harris? U.N. Known?) the party will simply go "Okay, so we didn't hit the targets. But that wasn't our fault! And besides Metaculus is biased against us! And who are you going to vote for instead, a Republican? Yeah, right! Only Trump-loving Trump-supporting Trump-voting types would reject a Democrat President! If you don't vote for our candidate, then Trump has won!"
All these Accords Held In World Capitals are simply political games. The people are saying that Something Must Be Done, so this is Something. By the time the bill comes due, we'll be out of power. Or maybe the horse will have learned to sing.
Biden is the first politician I've seen in a while who is taking the standard advice to under-promise and over-deliver, with all his vaccination claims - 1 million a day, and 100 million by the first 100 days, upped eventually to 150 million. Only the very last ask, of 70% of adults vaccinated by July 4, looks unlikely to be made, and that one was phrased as an ask, rather than a promise.
"Australia's 2009 Copenhagen summit pledge to decrease emissions 5% by 2020 (in fact, they increased 17%)."
Huh? I think you need to recheck your figures. Emissions are down, not up, on the Copenhagen context. The only sticking point is measuring Kyoto carry credits. Your results source covers only energy and cement production.
The relationship between whatever Australia was doing in 2009 and what it achieved in 2020 is also irrelevant to your thesis. Australia does not have term limits (Australia's longest serving PM spent 18 and a half years in office). The party which held power from the 2008 election produced a hung parliament in the 2010 federal election after infighting and lost convincingly in the 2013 election, in no small part due to the carbon tax it imposed in 2011.
Your examples must come from places with term limits to have any relevance to your thesis.
I agree with your main point, about promises being hollow. However, your first example is incorrect. Australia's promise to reduce emissions included changes in land use (https://web.archive.org/web/20091027003859/http://climatechange.gov.au/~/media/Files/minister/wong/2009/media-releases/May/mr20090504c.ashx), whereas the data from Our World in Data excludes land use.
Australia including land use in its emissions counting is a consistent source of controversy.
I think a cousin of the third objection is: even if you *can* hold a president accountable to a campaign promise, it doesn't do much. I don't think presidents really bother trying to avoid setting goals they can't be graded on, so much as they aren't judged about hitting the goal in the first place.
I like the idea, but I don't think the problem is that presidents' promises can't be evaluated. It's that most voters don't pay attention or care about them, and even if they did, it almost certainly won't be enough to change their opinion to vote for somebody else.
I ran a vaguely-similar idea by Robin Hanson earlier this year; aside from his work with Eliezer, he's also known as perhaps the most knowledgeable person around on the topic of Prediction Markets. The TL;DR of his feedback was, more or less: It could work, but only if someone has a strong incentive to pay for it. I got the sense that that was a very general objection and I get the sense that it's probably the main stumbling point here too.
This seems poorly suited to the US political system in particular. In a country where the party that wins the election can do what it wants until the next election, your proposal addresses a democratic accountability problem with pledges the party makes about future outcomes. But in the US, the major democratic accountability problem is that so many different actors have a say in policy that voters don't know who to credit/blame for outcomes they like/dislike. If you create the expectation that the president is always responsible for meeting specified goals and that failure has real political consequence, you increase incentives for obstruction and sabotage. These perverse incentives already exist but making them an even bigger element of our system seems bad.
"Stock markets can stay longer irrational than you can stay solvent" is a common saying among (value) stock investors.
So political prediction markets can surely stay longer irrational than politicians can stay in power.
Yes, people who recognize that irrationality might make a killing, just like Micheal Burry made a killing with shorting the housing market in 2008. That does not make the market sane.
I feel like you're underestimating the power a bunch of people have to move markets if they really want to. Look at Gamestop and AMC.
This proposal is based on the not entirely explicitly stated premise that Biden's pledge is a promise made by him personally to the American people. And having noted the problems with that interpretation, you conclude that people are just Not Doing It Right. However, another conclusion one could draw upon seeing how problematic interpreting such pledges as personal promises is is to conclude that such an interpretation is not correct. To wit, Biden is the democratically elected representative of The American People, and Biden, as an agent of The American People, is pledging on their behalf to cut emissions. In making the pledge, Biden is asserting that he will plan his policies around this being a priority, and he is committing The American People to either cut emissions, or have their country be on bloggers' lists of Countries That Make Emission Pledge That They Totally Didn't Fulfill. If the electorate feels that Biden should be accountable for how far the country has come by the time end of his first term, they are free to make such an evaluation and vote accordingly regardless of whether there's a prediction market.
How about "at least prediction market predicts that a majority of the population believes that Biden tried his very best to reduce emissions by 30%"?
If you really want to have more responsive politicians -- and I'm not sure we do, there's a lot to be said for gridlock, unless you feel At One with the madding crowd 100% of the time -- then we should just do what business does: make executive pay largely bonuses. Every Presidential election year minus two we have a giant plebiscite in which The People vote on the bonuses they'll pay for certain objectives attained. Then everyone runs for office two years later and their pay is $0 (or whatever the practical de minimis is) + bonuses they achieve.
And if you *really* want to make the motivation effective on both sides, meaning both politicians and voters take the setting and achieving of goals seriously, then the way you vote for bonuses is you kick in your own money. So everyone who feels passionately about cutting CO2 emissions can throw $1000 of his own money into a pot to be divided among the Congressman, Senators, and/or President who achieve the specified goal within the specified time span.
“escalators can’t break, they can only become stairs”
Always assume that things can break more dramatically and more completely than you think they can.
https://www.youtube.com/watch?v=zM8q5tya6w4
https://www.youtube.com/watch?v=TFI5768nt-E
You're creating an incentive for people to bet billions that CO2 emissions will increase such that they lose money if that isn't the case.
Mix that with political hyper polarisation and that's a recepie for incentivise people to throw molotov cocktails into coal mines.
"If Metaculus gives a lower number than this, we can consider Biden to have failed in his pledge, and we can hold it against him when he tries to get re-elected."
An interesting proposition, but of limited utility since most people don't seem to base their assessments of politicians on hard facts, sound projections, etc.
This is pretty close to Hanson's futurarchy proposals but importantly different since the market is the output not an input to a decision (so you avoid the worry about manipulatability and predictability by other states).
I don't think any of the objections raised are that convincing but actually getting this to work seems impossible. We can't even get the SEC to approve a prediction market with enough liquidity with real money to even have an underlying market.
Also no way you'll get progressives not to distrust anything this market oriented.
1). This seems like a feature not a bug. Biden can't control what a successor might do but he can adjust the amount he pledges to move the price to account for this (if he can only impact chance of hitting goal by 5% pick that goal).
2). This isn't a problem as it just creates arbitrage opportunities provided the market doesn't cap participation or positions.
3). Pols already make promises about higher stock market values. What's the difference?
15% is so preposterous that I was driven to create a Metaculus account to downvote it. So you hae successfully applied Cunningham's law to boost their userbase.
The second objection seems the strongest here. Given the importance of political victory to millions of people, and the staggering sums of money involved, free money might not be enough to put things back to true estimates.
The first and third can be dealt with fairly easily - a promise that's too weak is a joke, not appealing.
But I think you missed the fourth argument. People can't even understand the stock market as a prediction market, and think of it as being some kind of random money-stealing machine. Even some very smart people. How on earth will you get a dorky niche site like Metaculus to become big enough to be relevant to big-league politics? I doubt you could even get the Andrew Yang campaign to try that - they'd think it'd make them look like dweebs, never mind someone like Trump or Biden.
I think your point is valid, but at least one of the facts you used is incorrect because Australia did meet its 2020 pledge target. The outcomes data you link to does not include emissions from land use change while Australia's target was for total emissions including from land use change.
1. This is needlessly complicated. Right now, if a politician makes an old-fashioned campaign promise, the sort of nerds and wonks who trust prediction markets can use prediction markets to evaluate whether those promises are being kept and the politicians in question should be reelected. And everybody else can use whatever it is they already use for the same purpose. So the current system can extract whatever utility prediction markets can provide, whether prediction markets are a fringe thing only a few nerds and wonks care about or something as revered as the NYSE, or anywhere in between. And it can *also* extract whatever value politicians and their constituents feel they are getting from traditional campaign promises, which by their actions they seem to feel is substantial. Your version provides value *only* to the extent that prediction markets are respected, which would seem to make it strictly inferior.
2. In addition to the prospect of the Evil Billionaires(tm) on the Other Side using their billions to drive the prediction market in a way that makes the incumbents look bad, there's also going to be the Gamestop Brigade doing the same thing, on both side. "Our Guy(tm) needs us to go invest in the Our Guy's Telling The Truth portfolio, and if enough of us do this and stay the course, Our Guy wins *and* we all make lots of money from the Evil Billionaires who thought they could manipulate the market! Let's do this!"
3. The Gamestop Brigade will do this, and the Evil Billionaires will be believed to have done this whether they did it or not. And in the end, some of them may lose a lot of money for having done it. But *until* the end, each side has a perfectly good excuse to say, "the market would totally have proven us right except for all those cynical market manipulators rigging the game", and nobody is actually going to e.g. vote Biden out of office because the prediction markets were putting him at only 30% to reduce carbon emissions.
Goodhart's Law and Campbell's Law are real. If you think prediction markets are useful, or likely to be come useful in the future, then don't do this because it will break them. Particularly don't do it now, or any time before prediction markets are as large and robust as the NYSE.
I am intrigued by Scott Alexanders long-standing advocacy of prediction markets. He is an impressively clever guy, so I am totally open to the possibility that there is something vital about such markets I am too dim - or uninformed - to get.
...and with such an intro there is bound to be a but, of course.
Here is my but:
...but is not the only way a prediction market can yield important/useful information about what is likely to happen in the future, a prediction market limited to bettors that have SERIOUS skin in the game?
To elaborate: If betting is restricted to those willing to put 10.000 USD behind their bet (that's just a number - they point is that it must be a serious amount of money for most people), people are incentivized to put in information-gathering costs at least equal to USD 10.000, to have an "enlightened" opinion on what they are betting to happen.
If, by contrast, people can bet sums upwards from USD 1 (or even nothing), the whole betting schtick will be swamped by non-informed bettors who bet just for the fun of it. It will be impossible to distinguish the signal (the bets of seriously informed bettors) from the noise (the uninformed bets by everyone who participate just for the fun of it).
If I am right here, and we - to avoid that - limit betting to those wo have put serious money where their mouth is, we unfortunately run into another problem:
....Most policy-related decisions/events are driven by a ton of factors (think Brexit or the fall of the USSR), which make the random component of what is going to happen, much larger than in most business-related bets (will the share value of Amazon go up or down next year and suchlike). Implying that no sane, ordinary person will bet as much as USD 10.000 on any policy-related future decision/issue taking place.
...Which in its turn implies that you will get a selection effect, where the only bettors willing to back their bets by USD 10.000 or more, are either non-sane people (whose predictions you for a variety of reasons should not have much confidence in); or those rich enough to regard USD 10.000 as a trivial sum to spend just for the fun of it (whose predictions you should not have much confidence in either, since they are so rich they do not have an incentive to get themselves an informed opinion before they place their bet).
That's my problem(s) with the assumed usefulness of prediction markets. But again, Scott Alexander is an extraordinary clever person. What do I miss? Where do I go wrong in the above line of reasoning?
And for the record: I am not at all trying to be smug or anything! Really. I am genuinely interested in whether, and if so why, the above objections miss the mark; or whether there are some benefits to prediction markets I do not see, that trumps these objections.
I think first you would have to make the average voter aware that there is such as thing as prediction markets, what they do and how the current ones are weak. This should take at least the rest of Biden's first term (since I don't think the average voter knows already, this seems safe).
Also "figuring out how to prevent your successors from dismantling it". Judging by what currently is bulletproof, one needs to craft legislation that gives every legislator something. For instance, the current programs funding NASA seem to be mostly interested in funding something in every state that could block it. Or look at whose ox would be gored if any given welfare rule is changed - the ones that are not changed are the ones where any change hurts someone who can block the change.
Trying to hold people accountable for their work results runs into all kinds of trouble, including campbell's law/goodhart's law. The more complex, and difficult to evaluate the work results are, the worse the troubles will be.
If we hold politicians accountable for a number of easily quantifiable goals, they will pursue these goals at the cost of everything else. And cook the books. And play all kinds of nasty games with blame-shifting and sabotaging their opponents.
I think we would get better results if we were to hold politicians LESS accountable.
I would love for this to work, but I think that looking at actual IRL stock markets suggests that it wouldn't.
Politicians do in fact pledge to grow the economy and stock market, but don't consistently do so. Instead, they claim success whenever it goes up and blame their opponents whenever it goes down. Since the economy follows semi-predictable but difficult to time business cycles, it's hard to say whether they're even having an effect or not.
In individual stocks, we see all sorts of BS. It's more profitable to try and cash in on a bubble than to short it. Stock markets seem to largely function as a Keynesian beauty contest with only occasional reality checks. "Crazy" fluctuations based on memes, passing news stories, people misreading the name of the stock, etc are *entirely rational for investors to buy into* as long as they think some fraction of people are idiots they can fleece, amplifying rather than suppressing the dumbest voices.
End result? At the end of Biden's presidency, some news story comes out that's bad for his climate targets. Rational investors judge that it's nonsense, but some fraction of people will believe it, so they buy tons of "Biden won't hit his target" contacts, sending the price soaring (and vice versa for the opposite contract). They're right, and morons start buying it, sending the price even higher. Trump supporters start a meme to #HODL "Biden plan won't hit climate targets", diamond hands will win Trump 2024! The price climbs to absurd levels suggesting Biden will personally destroy the planet; this becomes it's own news story, driving the price still higher.
Biden's term ends and he's narrowly reelected, but he blames his failure to hit the prediction market targets on Trump supporters gaming the market and other factors outside his control. Prediction markets are declared racist.
A year later, the smart money quietly sells at the peak of the market. The dumb money panics and sells, mostly at a loss, and the bubble collapses.
Prediction markets have enormous potential for good, especially if people could get interested in them to the same degree as meme-stocks.
In your discussion of Objection 2, you bring up endogeneity ("The Republicans will spend money to artificially deflate the prediction to make Biden look bad and lose re-election") but then seem to ignore it later on in the same discussion ("If the prediction market number is low, everyone will buy because it'll be free money!").
Won't it only be "free money" if the market ends up high enough for Biden to win and continue enacting his policies? Doesn't that in alter the prediction market into a dollar-based popularity contest, or possibly something with similar dynamics to a "dollar auction"?
This exercise assumes that Biden has honest intentions, that he cares about results more than optics. He does not.
No, that is not an endorsement of either Trump or Team R.
The first objection is a feature not a bug. Presumably, a market would put greater weight on policies with broad support and less weight on policies enacted by slim temporary majorities. Exactly the sort of compromise policy making that the wings of both (US) major parties have discouraged over the past decade or so.
Would this not produce a huge incentive for obstruction? It's almost like for this to work, we need to replace checks and balances with prediction markets?
Nitpick, but let me take issue with the claim that Biden will probably be dead by 2030. He is currently 78. According to this table (https://www.ssa.gov/oact/STATS/table4c6.html) he would be expected to live an other 9.37 years, which puts us well into 2030. He is probably healthier than the average 78-year-old, so I'd say he will probably still be alive in 2030.
When I explained to a friend how absurdly expensive the Green New Deal would be, she replied that it wasn't actually a real goal, it was like haggling--you always open with an offer you don't expect to be accepted. It also has to be too extreme in order to "motivate the base". Also, achieving politically divisive goals would reduce the fanaticism of your own side, and increase that of the other side.
Holding politicians accountable via prediction markets would be great, but I'm generally more worried about politicians achieving their stupid grand goals than about not achieving them. If by monitoring prediction markets as they monitor opinion polls, politicians learn how to achieve their goals, that could be a net loss.
"If a prediction market is big enough and liquid enough, it should be robust against these kinds of distortion attempts."
Prediction markets are robust against people trying to change them by *trading inside* the market, for exactly the reason stated (outcome A still pays off an expected ~$1, outcome B ~$0).
However, they are not robust against a credible promise to distribute a subsidy proportionately among holders of outcome B, because doing so legitimately changes the expected value of holding that position. To what degree it changes the price depends on how much capital is in (or can get into) the market, how credible the promise is, how many people hear about it in the first place, etc.
This can be combined with taking a position in whatever the prediction market controls, if it's directly hooked up to something of interest, to pump more than enough money out of the system to cover the subsidy plus profit. E.g. there used to be calls to replace the Fed with a prediction market -- in that case, one can take a position in the forex market and then subsidize the predictiFed to pull the price in the appropriate direction.
"Imagine Biden pledges that some prediction market will have a 51% chance of reaching his 2030 emissions target by the time he leaves office. He passes a carbon tax, and the market shoots up from 15% to 30%. Now he knows he’s on the right track, but still has to do more. So he bans a bunch of coal power plants, and it goes up to 45%. He's still not quite there, so he gives big subsidies to solar panels a few days before the campaign season kicks off, the prediction market reaches 51%, and he's able to say he fulfilled his pledge."
This feedback loop is almost entirely blunted if predictors expect Biden to actually care about the prediction market. Once Biden makes that pledge, the prediction market should take it mostly for granted that he's going to take reasonable measures to hit the goal (if possible). If your example's numbers are reasonable absent a pledge, then the market should take for granted that he'll pass a carbon tax, etc. Just making that pledge will cause the market to trade way over 15%, over 30% if he's perceived as serious. If he passes the carbon tax and the market just sits still at 45%, it doesn't provide him much information. If he passes a useless bill, maybe the market ticks down from 45% to 44% or something, but it should price in that he's going to come around to a better bill eventually.
"accept that part of the work of making a global warming plan is figuring out how to prevent your successors from dismantling it"
There is the issue that this would create an even stronger incentive to find ways to lock in policies in ways that violate the spirit of the Constitution or law.