Money Stuff: Robot Retail Investors

People do not like to lose money on stocks. Sometimes a stock goes down, and people who own it sell it to avoid losing more money, so it goes down more. This behavior is to some extent predictable, so there are circumstances in which you might think “if this stock goes below $50, it will probably keep going down to $45.” You might think that as a pure matter of investor psychology: “If the stock goes below $50, a lot of people who bought it at $50 will have losses and panic and hit the ‘sell’ button, driving it down more.”

You might have a slightly different theory. Because people do not like to lose money on stocks, and are not monitoring their stocks constantly, they might use stop-loss orders, where they tell their broker “if this stock goes below $50, sell it immediately.” Intuitively, people who do this are more likely to use round numbers like $50 than, like, “if this stock goes below $51.37, sell it,” so there is some clustering of stop-loss orders; there are more of them at some numbers than at others. And so if you are trying to predict market behavior, you might think “if this stock goes below $50, a lot of people will automatically sell it, and it will go down to $45.” None of this is investing or technical-analysis advice, but a certain amount of “stop hunting” probably goes on in financial markets: If you know or suspect that a lot of people have stop-loss orders to sell if a stock falls to $50, you might try to short the stock until it hits $50, expecting to trigger the stops and buy it back at $45.

In a sense, the previous two paragraphs express the same idea: They both say “people have irrational biases that make them anchor on round numbers and fear losses, so if the stock falls to $50 there will be further irrational selling driving it down to $45.” The difference is the mechanism: In the first paragraph, the stock hits $50 and then people decide to sell; in the second paragraph, people plan ahead to sell if the stock hits $50, and automate that selling. But the second paragraph feels more real, more predictable: “I know there are stop orders to sell at $50, so if the stock falls below $50 it will keep going down” has a bit more science to it than “I think people will panic at $50 and sell more.”

This point might generalize. Like:

1. Everything that happens in financial markets reflects, in some way, investor psychology. If you have a good aggregate sense of how investors think about stuff, and stuff happens, then you can predict whether asset prices will go up or down. But this is very hard and somewhat fuzzy, and if you said “I have perfect insight into how investors think so give me money to start a hedge fund,” people might not believe you. (If you have empirical insight into some narrower behavioral point — “Indian retail investors buy way more options than underlying shares,” say — then that might be a useful strategy.)
2. Some things that happen in financial markets — some small but notable subset of the things in the previous paragraph— reflect automated or rule-based investor psychology. Investors think “if X happens I should do Y,” and they set up an automated process to do Y if X happens, and if you have a good aggregate sense of what processes they have automated and how, you can predict whether asset prices will go up or down. This is not easy, necessarily, but it feels like more of an exact science than simply predicting investor psychology, and you probably can do it at a hedge fund. “When new stocks are added to the index, index funds mechanically sell the old stocks and buy the new ones”: That is a quasi-automation of investor desires, and predicting it is definitely a hedge fund strategy. “When a bond goes from a BBB- rating to BB+, investment-grade bond funds have to sell it, and its price falls more than is justified by fundamentals”: That again is a story about investor psychology (people like to own safe bonds) that has been turned into a mechanical rule (investment-grade funds have to own bonds with investment-grade ratings).

The examples in the previous paragraph — index rebalancing, fallen-angel bonds — are about institutional investors, because institutional investors are, you know, institutions. They have rules and procedures and mandates and regulation; they behave predictably. Retail investors just do what they want. They sometimes act predictably, but they tend not to automate their predictable behavior, aside from the occasional stop-loss order. [1] You might think “retail investors tend to buy the dip, so if the stock market is down more than 2% near the close today, I should buy some stock to profit from tomorrow’s retail dip-buying,” and you might be right. But the retail buyers might all change their minds or forget; their dip-buying is just a descriptive matter of psychology. It’s not like they have all set up automated rules-based systems to buy the dip.

Last week the Wall Street Journal had a story with the headline “Buying the Dip? This AI Agent Will Do It for You”:

Public, a privately held brokerage firm, is rolling out a feature allowing customers to use AI to automate their investing tactics and execute trades.  

AI agents could be deployed to buy protective puts should oil spike, hedging against potential stock losses, and automatically sweep customers’ cash into higher-yielding assets, like bonds. Account holders could also use the agents to add a 20% stop-loss order on their trades, limiting potential losses by automatically selling a security when it drops to a certain price, Public co-Chief Executive Jannick Malling said. ...

Public users start by typing in a strategy or investment thesis, then refine its parameters with a series of follow-up questions, including which accounts the agent will trade from, which assets to buy or sell and whether the trade will be one-time or recurring. 

Once the parameters are set, the tool offers a workflow for traders to edit, review and approve before the agent goes live. 

Here’s Public’s web page describing the agents, with sample prompts like “Whenever SPY drops 3% or more in a single trading day, invest $10,000 at the next market open” and “If the market volatility index (VIX) hits $25, buy a put option on the S&P 500 to hedge my portfolio.”

I don’t know. That’s cool? Probably:

• Most retail investors will not end up using artificial intelligence agents to automate their investing tactics; and
• The ones that do will automate lots of idiosyncratic investing tactics that tend to cancel each other out, so it’s not like they’ll all automatically do the same things.

Still, we have talked a few times before about the possibilities for AI to coordinate retail behavior. If retail investors tend to get investment ideas from AI chatbots, I have written, those chatbots might tend to send all the retail investors to the same ideas, so the stocks favored by the chatbots will go up. Similarly, here, if retail investing becomes increasingly agentic, the agents might tend to automate stereotypical retail behaviors. These days, the most stereotypical retail behavior probably is buying the dip, [2]  and maybe that is the first one that will be automated. And then if the S&P 500 drops 2% or more in a trading session, every hedge fund will buy at the close, knowing that there’s billions of dollars of agentic retail demand coming the next morning.

Elsewhere in retail coordination via agentic AI:

Crypto influencer and entrepreneur Anthony Pompliano[’s] ... company ProCap Financial is launching a new business focused on AI-generated research reports for individual investors. 

The plan is to have a fleet of AI agents scanning the markets, analyzing trends and drafting reports with titles like “3 Stocks That Win From Both Tariff Refunds and the Iran Oil Shock,” according to an example shared with The Wall Street Journal. ...

The agents are capable of generating hundreds of reports a day, Pompliano added—but the company will start by circulating just a few daily to avoid flooding investors’ inboxes.

Reports distributed by the service, called ProCap Insights, will cover single-name stocks, thematic trends and macro analyses and won’t offer any specific “buy” or “sell” recommendations, according to the company. A ProCap Insights annual subscription will cost investors $2,500 a year.

Pompliano, who is ProCap Financial’s chief executive, said that AI-generated research notes are quicker and cheaper to produce than those assembled by a team of analysts at, say, a major investment bank. His company built ProCap Insights over the course of two weeks, he added. He said it cost only a couple thousand dollars and there was just one employee overseeing the project.

So they spent a couple thousand dollars to build a system that automatically generates reports with no effort, and they’re charging $2,500 for those reports?

At the New York Times, John Carreyrou more or less claims to have identified Satoshi Nakamoto, the pseudonymous inventor of Bitcoin, as Adam Back, a British cryptographer and Bitcoin guy. The evidence — word choices on listservs, taste in cryptography, posting about Bitcoin, etc. — is not quite conclusive; Back has denied being Satoshi, and my Bloomberg colleague Joe Weisenthal is not convinced. 

But I appreciated another aspect of Carreyrou’s evidence. I joke a lot around here about claims that investment analysts use “tone and body language” in executive meetings to make investing decisions; the investment analysts are not CIA interrogators. But I suppose it’s possible that investigative journalists are in fact good at using that sort of evidence, and Carreyrou seems to rely on it. He was put on Back’s trail at the beginning of the story, and more-or-less confirmed in his suspicions at the end, by Back’s body language:

[In a documentary Carreyrou watched,] Adam Back, a British cryptographer and leading figure in the Bitcoin movement, sat on a park bench in Riga, Latvia, his shirt untucked under a brown coat. The filmmaker casually rattled off the names of several Satoshi suspects. At the mention of his own name, Mr. Back tensed up, strenuously denied he was Satoshi and asked that the conversation be kept off the record.

Having encountered my share of liars and developed something of an expertise in their tells, Mr. Back’s demeanor — his shifty eyes, his awkward chuckle, the jerky movement of his left hand — struck me as fishy. 

And at the end:

Over the next two hours, I presented my evidence piece by piece. In his soft British lilt, Mr. Back insisted he wasn’t Satoshi and chalked it all up to a series of coincidences. But at times, his body language told a different story. His face reddened and he shifted uncomfortably in his seat when confronted with things that were harder to explain away.

It’s the $5 wrench attack of Satoshi Nakamoto identification: The best way to prove that someone is Satoshi is to go up to him and say “are you Satoshi” and see if he starts looking shifty.

Anyway, in his day job, Back is now the chief executive officer of Bitcoin Standard Treasury Company, a digital asset treasury company planning to go public by merger with a special-purpose acquisition company, of course. I once wrote about the rise of digital asset treasury companies around the world:

What would Satoshi Nakamoto think? What a strange vision of crypto this is. In the future, in every country, you will be able to go to your locally regulated stockbroker and pay a premium of 100% or more to buy shares of stock of a trusted local company, denominated in the local currency, that will hold Bitcoin for you. If you want to transfer your Bitcoin across national borders you can … I don’t know, sell the stock on the exchange through your broker, do a foreign exchange transaction to convert rupees into dirham, find a stockbroker in the target country, open an account, pass know-your-customer checks, fund the account with local currency and then buy stock in that country’s local Bitcoin company (at a 100% or more premium). Seems like it might be easier to buy Bitcoin? But what do I know.

Maybe now we know what Satoshi Nakamoto would think!

One more point. Carreyrou writes:

As chief executive of the merged company, Mr. Back was required under U.S. securities law to disclose any information that was material to its investors. A secret stash of 1.1 million coins that could crash the Bitcoin market if it were suddenly sold, for example, would probably be considered material.

Everything, I often say around here, is securities fraud. It would be satisfying if being Satoshi Nakamoto is also securities fraud.

Elsewhere in hedge fund trades, Bloomberg’s Justina Lee reports:

The dispersion trade, which buys options on individual US stocks while selling those on the broader index, suffered a 4.9% loss last month, the steepest since 2011 in backtested data, an index from JPMorgan Chase & Co. shows. Bank swaps tied to the approach fell 2.6%, according to Premialab, which aggregates industry data.

The losses lay bare the strategy’s vulnerability to such a major geopolitical event. As the conflict in Iran escalated, investors rushed to hedge at the index level, driving implied volatility on the S&P 500 sharply higher. At the same time, individual stocks began moving in lockstep — collapsing precisely the divergence between single-name and benchmark volatility that the dispersion trade is designed to harvest.

“The recent events had macro consequences on oil, commodities, inflation, rates, rather than micro impacts at the single-stock level,” Luca Valitutti, managing director for exotics and hybrids trading at Citigroup Inc., wrote in an email. “Correlation between single stocks has spiked.”

We talked about this at the beginning of last month. The dispersion trade is, classically, a bet on stereotypical behavior by investors: They like to buy index options to hedge broad market risk, and they like to sell single-stock options for income. Therefore index options trade at a premium and single-stock options trade at a discount, and you can manufacture index options out of single-stock options to capture that spread. [3]

But it is also a fundamental bet on correlation. If individual stocks are volatile but uncorrelated, then their volatility will cancel each other out: Some stocks will go up 2%, some will go down 1.5%, and the overall stock index will go up 0.5%. Index volatility will be lower than single-stock volatility, and a strategy of buying single-stock options and selling index options will make a lot of money. [4] If individual stocks all move together, then they’ll all go up 2% today and down 1.5% tomorrow, the overall index will go up 2% today and down 1.5% tomorrow, index volatility will be high and you’ll lose money on the dispersion trade. [5]

As we discussed last month, a reasonable fundamental story in February was something like “the rise of artificial intelligence is going to create huge winners and huge losers, so there will be a lot of dispersion among stocks.” And February was a great month for the dispersion trade. And then a reasonable fundamental story in March was something like “oh man are you watching this Iran war?” The war has similar effects on all of the stocks, so correlation has gone up.

There are various sophisticated forms of accounting fraud, where you can exploit nuances of accounting rules and blind spots of auditors to inflate your company’s assets or revenues. But there is also the simplest form of accounting fraud, where you have 1,000 widgets in inventory and just write down that you have 10,000. That is not especially difficult for your auditors to spot: They can just go to your widget warehouse and count the widgets. But that’s easy for me to say, here, typing on my computer. If you are the auditor sent out to the widget warehouse, “just count 10,000 widgets” might not sound that easy.. Here is a Wall Street Journal story about auditors praying for AI to replace them:

Accounting firms and their clients are increasingly using artificial intelligence and drones to do work long handled by humans. But so far, there’s been no technological solution to what is often the dirtiest part of an audit: counting inventory. That means the messy, bizarre field trips remain a rite of passage for young professionals in an otherwise deskbound field. ...

Auditors are often tasked with traveling to the middle of nowhere and tallying up a large amount of unusual things, from chickens and pigs to quarry rocks, corn, traffic lights and telephone poles. They complain about ending up covered in manure or dust, or shivering in a freezer. 

The large accounting firms typically rely on junior auditors to do much of the dirty work. The trips provide an opportunity to learn more about clients. But in the moment, they stink. Gen Z auditors share horror stories on social media. “I just counted thousands and thousands of nuts and bolts. What did you do today??” one posted on TikTok. A Reddit user said they were sent to count rocks in below-freezing weather, including a pile in a snake-infested quarry.

We used to talk all the time about the idea that “the blockchain” would prevent fraud, guarantee the provenance of goods, etc., and I was always skeptical: Really good computer technology can prevent all sorts of computer-based fraud and manipulation, but there is always going to be some nexus between the computers and the physical world, and you can’t count on the computer to monitor that nexus. Eventually I suppose AI will get good enough to police that nexus — humanoid robots with sophisticated eyes will walk into the warehouse and instantly say “I count 162,385 nuts and 237,423 bolts, what’s next?” — but for now it is, uh, maybe the last job for humans.

Also, I know I will get emails about it, so I will obligatorily mention the Salad Oil Scandal of 1963, which I suppose could have been prevented if a junior auditor had gone for a swim in the tanks of soybean oil. [6]

Tesla/SpaceX merger sure why not.  Polymarket’s Iran Bets Draw Fresh Disputes and Insider Scrutiny. Insurers’ $1 Trillion Buildup in Private Credit Is Leaving Regulators in the Dust. Blue Owl Fund Outlook Cut to Negative by Moody’s on Outflows.  Barings Caps Redemptions as Private Credit Holders Seek 11%. Citi sets aggressive targets for bankers in wealth management unit. Commerzbank Doesn’t See Basis For Deal With UniCredit Following Talks. Iran demands crypto fees for ships passing Hormuz during ceasefire. A Fire Sale Has U.S. Office Buildings Going for 90% Off. Chile Uncovers $917 Million Copper-Theft Ring Shipping to China. ‘Freak Out’ Indicator Soars to Record With War Sparking Trader Anxiety. Humans Are Losing the Fight Against Flying Fish. “Donald Trump … has threatened to destroy a civilization. How does an investor process that? Is it a bigger upside risk or downside risk?”

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!