Money Stuff: The Tariff Refund Trade

What is the probability that the US Supreme Court will strike down President Donald Trump’s broad tariffs? There are a number of ways of thinking about that question. One is what you might call legal analysis: You can look at the text of the US Constitution and the relevant statutes, as well as relevant legal precedents, and try to figure out whether the tariffs are legal or not. If they seem very illegal, then the probability that the Supreme Court will strike them down is high; if they seem very legal, low; if the law is genuinely ambiguous then maybe it’s 50/50. In previous columns I have sort of gestured in the direction of this analysis, but honestly this might be the worst way of thinking about the question. 

A more promising approach is what you might call legal realism: You notice who sits on the Supreme Court and what their preferences are about partisan politics and executive power, and you try to figure out what that predicts for their ruling. I have gestured in this direction as well, and it seems a bit more, um, realistic than relying on the law.

A third approach is what you might call market-implied probabilities: You look at the market price of some traded instrument and try to back out a probability from that, assuming that markets are generally efficient and that the aggregated opinions of people with skin in the game are probably better than your own legal or political analyses. In general I am a fan of this sort of approach. But: what instrument? You can pick some public companies that are particularly exposed to tariffs, regress their prices against tariff news, and try to figure out what changes in their prices say about the likely legality of the tariffs. But that is very noisy: Those stocks will move for lots of reasons unrelated to tariffs, and even their tariff-related moves will be compounded of (1) the probability that the tariffs stay in place and (2) the impact of the tariffs on their business. Maybe Treasury bond prices reflect tariffs in some way — if the tariffs are struck down that will probably increase the budget deficit — but again the effects are noisy and ambiguous.

This is, classically, what prediction markets are for. You can just go on Kalshi or Polymarket and see the prices of contracts that pay out $1 if the Supreme Court rules in favor of the tariffs or $0 if it rules against them. The price of that contract is a reasonably pure market estimate of probability. [1]  These contracts trade in the neighborhood of 40 cents on the dollar, which translates fairly directly into an implied probability of about 40% that the tariffs will be upheld. [2]  Not quite a coin toss, but a biased coin toss.

Why would you trade that market? Well, you might have a view: You might have used legal analysis or legal realism or an insider source on the Supreme Court or whatever to come to a view that the court will uphold the tariffs, or that it won’t, or that it’s a true 50/50 tossup so the current 40% price is too low. Different people will use different (or the same) methods to come up with different answers; the ones who think the court will uphold the tariffs will buy the “Yes” contracts and the ones who think it won’t will buy the “No” contract and the market will trade to a price that balances both sides.

But there is another reason you might trade that market. You might have economic exposure to tariffs. You might be a company that imports stuff into the US and has to pay tariffs; if the Supreme Court upholds the tariffs your costs will remain high, but if it strikes them down your costs will go down. You have no idea what will happen, but you need to plan. You could spend $4 million to buy 10 million “Yes” contracts. If the Supreme Court rules in favor of the tariffs, those contracts will pay out $10 million, offsetting your increased costs. If the Supreme Court rules against the tariffs, you will lose your $4 million, but your costs will be lower. (In fact, if the Supreme Court rules against the tariffs, and you have already paid a lot of tariffs, you might get a refund, which might cover this loss.) You have hedged some of your tariff risk.

This is a good use of prediction markets: There is some uncertain future event with significant economic consequences to real businesses. Those businesses could trade prediction market contracts to hedge those consequences. Doing this hedging does not necessarily require them to have a strong view on the merits: You might say “I am a simple widget importer, I have no idea what will happen in the Supreme Court, but paying $4 million today to take some tariff risk off the table will help me sleep at night so I’ll do it.” And because these hedging flows are relatively price-insensitive, there are lucrative opportunities for informed investors to trade against them, so there is an incentive to become informed.

I am just kidding about all of this. I mean, I’m not, this is how prediction markets should work, but there is a problem, which is that the volumes are tiny. As of this morning, Kalshi shows less than $250,000 of total volume on its tariff contract, Polymarket less than $400,000. The order books show resting orders in the single-digit thousands of contracts. The biggest position on Polymarket is about $70,000 notional amount on “No.” If you wanted to hedge $10,000 of tariff risk, you could, at the cost of moving the price several percentage points. If you wanted to hedge $10 million of tariff risk, no chance.

Anyway here’s a Bloomberg story about actual tariff risk transfer:

Wall Street banks are arranging bets on President Donald Trump’s tariffs being struck down by the Supreme Court — long-shot trades that could pay off handsomely for hedge funds betting against the legality of the administration’s flagship policy.

Jefferies Financial Group Inc. and Oppenheimer & Co. are among firms brokering the deals, matching investors with companies that have paid tariffs to import goods into the US, according to people with knowledge of the matter and correspondence seen by Bloomberg News. …

In the trades, the importing companies essentially sell to investors any future rights to claim refunds on their tariff bills, which could come if the nation’s top court sides with an ongoing legal challenge to Trump’s tariffs. The companies sell at a discount to their expected refunds, meaning investors would reap the upside in a ruling favorable to them. The banks arranging the deals take a cut. …

For example, a hedge fund might pay somewhere between 20 to 40 cents for each dollar of claims they could get back in refunds, giving them an upside of several times their bet, according to the correspondence and some of the people, who asked not to be identified discussing potential terms. Most of the trades range in size from $2 million to $20 million, with few over $100 million, one of the people said.

The smallest of these trades is bigger than the total tariff volume on prediction markets. [3]  I suppose we are still in early days, but so far, the prediction markets arguably are not “expanding how individuals and institutions use probabilities to understand and price the future” as far as tariffs go. Whereas the traditional financial sector is: It is arranging large trades between economically motivated speculators and real-economy hedgers, trades that both help companies hedge risks and give some market-implied indication of probabilities.

Kalshi had $12.5 million of volume on yesterday’s Jets game.

But can the traditional financial sector do this?

[Kalshi is] hiring tarot card readers, palm readers, psychics, fortune tellers, oracles, wizards, and individuals born with divine intuition.

If you are familiar with magic and the dark arts I'm hiring for something big. 

(YOU HAVE TO BE UNEMPLOYED - No job havers please). 

That’s from an X post by Salman Sohani at Kalshi (via Emily Sundberg). Is he kidding? I have no idea. In general there is not a ton of evidence that tarot card reading is useful in predicting financial asset prices; there is an xkcd. But prediction markets are a fairly new sort of financial asset, and maybe the tarot has more power there. Fun to find out.

By the way, we talked last week about a bunch of people who were charged with federal crimes for allegedly using inside information about professional basketball games to place bets at sportsbooks. My basic point was that “insider betting” is an odd sort of crime: They were charged with wire fraud, that is, using a “scheme or artifice to defraud” the sportsbooks. How did they defraud the sportsbooks? Well, there are some intuitive answers like “come on betting on inside information is fraud,” but the history of insider trading does not entirely support that theory. There is a simpler answer, though, which is that the sportsbooks’ terms of service forbid insider betting. The prosecution’s theory is approximately that the bettors were lying to the sportsbooks — by acknowledging the terms of service, which include saying “I won’t use inside information,” and then using inside information anyway — and that this lie is what makes it fraud.

But another way to put it is that the sportsbooks’ terms of service can make things into crime. If the sportsbook says “you can’t use inside information,” and you do, that’s wire fraud. If the sportsbook says “you can’t use sophisticated statistical models,” and you do, would that be wire fraud? I don’t know. The terms of service don’t say that. I don’t think? DraftKings’ terms of service do say that it’s a violation to use “automated means including, without limitation, bots, scripts, artificial intelligence, programs, parser, spiders, screen scrapers tools, or applications, to interact with the Services or Website, including, without limitation, collecting any information or content from the Services or Website, automatically placing bets or wagers, automatically entering any poker or live dealer Game, or automation of game play decisions.” Does that make it a federal crime to use a computer to automate your betting decisions? I suppose we’ll find out eventually.

Anyway my point here is that I do not think that any sportsbook currently bans occult powers or divine intuition in their terms of service. But if they did, and you used your occult powers to make bets, would that be a crime?

Man, I don’t know, if you see a television ad that is like “The World’s Most Sophisticated Investors Want to Invest Your Money, Act Now, Supplies Are Limited,” what conclusions should you draw? An ad like “The World’s Best Smartphone Company Wants to Sell You a Smartphone” is fine, you know, the more volume the better. Investing isn’t quite like that. Every strategy has some capacity, and historically one way that, for instance, private equity firms made buckets of money for themselves and their investors is that they picked deals. If there are thousands of companies, and your job is to buy 20 of them, and you are very sophisticated, it is not hard to understand how you could achieve a market-beating return. You could evaluate the companies to find the best ones, you could build good relationships so they will sell to you, you could negotiate fiercely to buy them cheaply, you could lean on your lenders to get attractive credit terms, you could focus on improving their operations to increase their value, and when you have done all that you could sell them back to the public markets at a premium that reflects their quality and scarcity. 

But if there are thousands of companies and your job is to buy, you know, 2,000 of them, all of that is attenuated. You might still be really good and smart at picking companies and negotiating deals and improving operations. But your returns are just naturally going to end up being closer to the broad market return. And if you are still hiring lots of smart people to work very hard to pick the best companies and negotiate the best deals, your costs will be high. People can get market returns with index funds; why should they pay you a lot to get similar returns?

And the answer is perhaps “because they are retail investors and you can advertise your history of good returns”? I wrote last month:

I suppose the natural endpoint of this is a complete leveling, where “being owned by private equity” and “being owned by public shareholders” are essentially equivalent: In each category you’d get the same sorts of companies with the same sorts of capital structures and the same sorts of operations.

Also, why not, the same ultimate owners. We have talked a lot about the push to get ordinary investors’ retirement savings into private assets. If private equity is just a normal way to own companies, you would expect “being owned by private equity” to mean, ultimately, “being owned by ordinary investors in their 401(k)s.” This is in part because individual retirement savings are a big source of cash, and as private equity gets big enough it will require all the biggest sources of cash.

But it is also in part because individual retirement savings are, like, the most normie source of cash. People’s 401(k)s are the last place you would expect to find a ton of differentiated alpha. In the beginning, when private equity was a novel risky arbitrage to correct systematic mispricing of companies, it was funded by sophisticated investors who could understand and appreciate that bet. If private equity is just “owning companies,” it will be funded by everyone’s retirement savings.

But to make that transition happen, you need two things. One is regulatory reform to make it easier for private equity firms to sell to retail investors and 401(k)s. That is very much happening!

The other thing you need is a television advertising campaign. Bloomberg’s Loukia Gyftopoulou and Preeti Singh report on a Blackstone Group TV ad:

Scene: The Gold Rush, 1849. Prospectors gaze up at a snow-capped peak.

“Now let’s go get that gold!” a newcomer exclaims, trudging up the trail to the twang of spaghetti-western music.

“I got a better idea,” says another. His brainstorm: Rather than hunt for gold, they’ll get rich selling picks and shovels.

Welcome to “Blackstone Town,” as a crew member later calls it – the set of an old-timey TV western that’s been repurposed for, of all things, an advertisement for the world’s largest private equity firm.

Okay! The article goes on to point out that nobody is exactly asking for this:

For Main Street, the timing looks iffy. Pension funds and endowments that have long powered private equity have begun to pull back. Returns don’t look as good as they used to, particularly given the record boom in stocks. Private credit has hit a few bumps lately, too, and faces concerns about rising defaults in a slowing economy.

If all that weren’t enough, average Janes and Joes aren’t knocking down the door. According to a recent Harris Poll for the Wall Street Journal, only 10% of 401(k) savers are dissatisfied with their investment choices and want less traditional options.

But if nobody is buying, that makes it all the more important to sell:

Interviews with more than two dozen wealth advisers and retirement plan advisers paint a picture of an industry in full sales mode. Wealth advisers — gatekeepers to the high-net-worth crowd — say they’ve been inundated with one come-on after another: spam emails, LinkedIn messages, cold calls and more. Some say they’ve even gotten happy birthday messages. Retirement plan advisers, who help companies with their 401(k) offerings, say the same.

I want to make a parody ad. Scene: The private-markets gold rush, 2025. Alternative investment managers in suits huddle around a campfire, eating takeout sushi. “Now let’s go earn our incentive fees by making market-beating investments,” a newcomer explains, firing up a spreadsheet. “I got a better idea,” says another. His brainstorm: Rather than beating the market, they’ll get rich selling private assets to retail investors.

There are two theories of artificial intelligence in the workplace. One theory is that AI will take over a lot of the intellectual drudgework of many white-collar jobs, freeing up humans to think higher-level thoughts and pursue deeper understanding. The other theory is that AI will take all of the good jobs and do all of the fun interesting important higher-level thinking, leaving humans to do the less intellectual work of fixing AI’s mistakes and going to the in-person sales meetings. [4]

We talked last week about OpenAI’s efforts to hire former investment bankers to make its chatbot really good at financial modeling. “AI can build a financial model in seconds that would previously have taken an investment banking analyst days” is an ambiguous claim. Building that model would have felt, to the analyst, a lot like drudgery, but also analyzing the drivers of a company’s financial value is the intellectual content of investment banking, and if you hand that over to the AI, what is left? Shaking hands with clients and babysitting the AI? 

On the other hand:

JPMorgan Chase has given employees the option to use its in-house artificial intelligence system to help write year-end performance reviews, underscoring how AI-generated text is proliferating in corporate America. 

The tool allows employees to use the US bank’s large language model to generate a review based on prompts they give it, according to people familiar with the matter. 

It is a shortcut to the often painstaking process of writing multiple reviews that are typically required by large companies.

Ahahaha. Obviously, yes, writing year-end performance reviews is perhaps the best-known example of intellectual drudgery in white-collar employment. Letting an AI do it is a strict improvement for the people writing the reviews. Then one assumes that the reviews are also read by AIs, and increasingly people’s salaries and job security will be set by AIs talking to AIs rather than by human beings. Still that’s a tradeoff a lot of people would take if it means not writing performance reviews.

Similarly, here are two theories of how AI will affect various knowledge industries:

1. The incumbents in the industry — consulting firms, investment banks, etc. — will get really good at deploying AI, improving their work product and profit margins and making themselves indispensable to clients; or
2. The clients will get good enough at deploying AI to dispense with the incumbents entirely. Instead of asking a consultant for advice, you just ask ChatGPT and get a good enough answer.

We discussed these two theories in connection with consulting a few weeks ago. Consulting strikes me as particularly vulnerable to AI because the work product of a consultant (advice, reports) is broadly similar to that of an AI chatbot.

Investment banking seems a bit safer: The product of an investment bank is a deal, and AI cannot yet do deals. Still. If you are the chief executive officer of a big company and you want to have a bit of fun, you could fire up a chatbot and type in questions like “what strategic buyers or private equity firms would be good buyers of my company” and “how much should they be willing to pay for my company” and “what is the best way for me to persuade them to pay the most for my company” and “please write an information memo for potential buyers incorporating a robust financial model and pretty charts” and “please write a merger agreement for the potential buyer to sign.” I do not think that, in 2025, you will get outputs that are good enough to put into production. But, again, OpenAI is hiring people to improve that. Give it a few years. You never know.

OpenAI itself, for instance, is post-investment banking:

OpenAI chief Sam Altman and an inner circle of executives masterminded deals worth as much as $1.5tn with little input from external advisers, despite unusual structures that tie the start-up’s fortunes to some of the world’s largest companies.

Altman largely shunned OpenAI’s bankers and lawyers to negotiate huge multiyear deals with Nvidia, Oracle, AMD and Broadcom to supply chips and other computing infrastructure.

Instead, he has leaned on a few lieutenants including the start-up’s president, Greg Brockman, chief financial officer Sarah Friar and Peter Hoeschele, recently promoted to a new role leading infrastructure financing, said people close to the company.

The unconventional dealmaking process has resulted in agreements that analysts have criticised for their lack of detailed financial terms — as well as circular structures that tie together suppliers, investors and customers.

Well, I am on record saying that “OpenAI Is Good at Deals”; it’s not obvious to me that “detailed financial terms” are what OpenAI wants.

I am joking a little here, in that there is no evidence that OpenAI asked ChatGPT to structure its deals. The deals were done with a combination of in-house expertise, personal relationships, and the general hubris of a company that is changing the world; it’s not like ChatGPT wrote up a pitchbook. Still, we live in a world in which many kinds of expertise are increasingly (1) devalued and (2) getting acquired by AI bots. OpenAI is at the cutting edge of both trends.

One generic form of career advice for young people is along the lines of “skate where the puck is going.” Look around, try to predict what skills and sectors will be hot in five or 10 years, and then get good at that stuff. This is hard, because predicting the future is hard for anyone and particularly hard for college freshmen, but if you get it right the payoff is enormous. This is the sort of advice that leads people to start billion-dollar companies in their dorm rooms, or at least get deep learning PhDs in 2019 so that now they can get $100 million pay packages.

Another generic form of career advice is along the lines of “be countercyclical.” Look around, try to see what skills and sectors are hot right now, and try to see what other skills and sectors are ice-cold. Then get good at the ice-cold ones, figuring that they will make a comeback eventually, you will be ready and the competition will be sparse. This is risky, too, since some things remain ice-cold forever; don’t spend your 20s getting really good at hand-crafting buggy whips. But this is the sort of advice that leads people to become, like, Cobol programmers, or to get into distressed-debt trading in a boom where there is no distressed debt to trade. But just you wait.

Anyway I know nothing about the pipeline for gold traders but I have to assume that … you know, over the last 10 years, if you had somewhat crankish monetary inclinations and wanted to be a trader, you got into crypto trading, and if you had less crankish inclinations you got into stocks or bonds or natural gas or electricity or cocoa or, you know, kind of anything else. “The barbarous relic,” people call gold, often proudly and crankily. But gold is super hot now, and if you did get into gold trading a few years ago, now you can name your price: 

Trading houses, hedge funds and banks are on a hiring spree for specialist gold traders as interest in the metal soars, creating a battle for talent that is driving up pay packages in what has historically been a niche market.

Major commodity traders Trafigura Group and Gunvor Group have brought in teams of precious-metals traders this year, while rivals IXM and Mercuria Energy Group Ltd. have also been looking to hire in the sector, according to people familiar with the matter. Numerous hedge funds, banks and industrial companies like refiners are also either trying to break into precious metals or expand their teams, headhunters and industry executives said.

While gold is a high-profile market, with the equivalent of hundreds of billions of dollars changing hands every week in the main hub of London, it has traditionally been dominated by a handful of banks such as JPMorgan Chase & Co., HSBC Holdings Plc and UBS Group AG, operating with lean teams of traders. As prices soared this year, new market participants seeking to break into the sector had only a small pool of experienced traders to draw on. …

“Precious metals has been, for many years, seen as more on the fringe,” LBMA Chief Executive Officer Ruth Crowell told the conference on Monday. “But I think given the growth in not just trading but client interest and participation, there certainly is a mainstream feeling and potentially a new chapter.”

Right what you want is to get into an industry when it’s on the fringe, and then cash in when there is a mainstream feeling.

Private Jets and Car Washes Are the Latest Tax Shields for the Ultrarich. Inside Steve Cohen’s Cubist shake-up: How a quiet search and chance timing led to the surprising change at Point72. The AI Startup Fueling ChatGPT’s Expertise Is Now Valued at $10 Billion. Keurig Dr Pepper Turns to Private Equity to Back $18 Billion Deal. Argentine Assets Surge After Decisive Milei Election Win. Huntington Strikes $7.4 Billion Deal to Expand in the South. It’s decorative gourd season.

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Or you can subscribe to Money Stuff and other great Bloomberg newsletters here. Thanks!