Money Stuff: AI Refi

A central fact of consumer finance is that big sophisticated financial companies can offer ordinary consumers irrationally good deals, because the consumers will probably mess them up for normal human reasons. The best-known example might be credit card rewards. Patrick McKenzie has a classic explanation, but the schematic story is something like:

1. If you buy stuff on a credit card, your bank collects a fee from the merchant of, call it, 1.5 cents for every $1 you spend.
2. The bank gives you a “reward” — cash back or points you can use to buy stuff — of X cents for every $Y you spend.
3. On average, the bank might pay its customers about 1 cent for every $1 they spend, keeping the other 0.5 cents for itself.
4. But there are some categories of spending, and some sorts of rewards, where the ratio is higher. “Get 5% cash back when you put gas in your car, and 1% cash back on everything else,” maybe. Or “get 1 point for every $1 you spend, the points can be converted into cash at 1 cent per point, or you can buy a first-class round-trip ticket to Paris for 100 points when the moon is waxing gibbous and Libra is in the seventh house.” Those points are worth 1 cent per $1 you spend, unless you use them the exact right way, in which case they’re worth more. 

If you have a lot of free time and a certain sort of brain, you might optimize this. You might take out a bunch of credit cards and use each one only for its most rewarding purpose. You might get the 5%-cash-back-on-gas card and use it only to buy gas, earning 5% aggregate cash back on that card. You might get the round-trip-to-Paris card, use it once to buy $100 of groceries, book the ticket at exactly the right time, and earn like 2,000% rewards on that card. People do stuff like this; there are websites and Reddit forums devoted to it. [1]

If you use each card optimally, the bank will not make much money on your account; it might even lose money. The bank might be paying you more in rewards than it makes in transaction fees. The bank is giving you a deal that is too good, a deal that might bankrupt it if everyone took the deal. But almost nobody does this. Most people just have one or two or three credit cards, and they use them for all of their stuff, and they mostly get the average reward payment and it’s fine. Some of them do even worse; they forget about the rewards and they expire, say. Some people are on the Reddit forums trying to optimize their rewards, but that’s fine. [2] Most people are busy. 

Every so often a bank will mess this up and lose money. (We have discussed an example, Bilt, a credit card for paying your rent.) But for the most part it works out well. Competitive pressure forces banks to offer deals that would be good if consumers used them optimally, and to rely on the fact that consumers mostly don’t.

That’s the story of credit cards, but it is a useful paradigm for lots of consumer finance. Let me briefly mention two of my favorite examples. One is that Capital One offered a high-yield savings account with an advertised very high interest rate that attracted a lot of customers. This was a bad deal for Capital One: It was paying more than the going rate for deposits. But then, over time, Capital One lowered the rate on the account, until it was paying much less than the going rate for those deposits. Customers could take their money out at any time without penalty, and one assumes some of them did, to put it in other banks paying higher rates. But many of them didn’t, because they were busy. They were not checking every day to make sure that the rate was still top-of-market. “We will pay you an above-market interest rate for now, and you can take your money out any time if we are no longer above-market” is on its face a bad deal for the bank, but in fact it was fine. [3]

My other favorite example is life insurance. Life insurance is often underpriced: If you buy 20-year term life insurance, your total monthly premiums over 20 years will probably add up to less than (1) your death benefit times (2) the actuarial probability that you will die during those 20 years. If 10,000 people signed up for term life insurance and paid all of their premiums, the insurance company would probably lose money.

The insurance companies are not idiots. They can offer underpriced insurance because many people do not pay all their premiums. They buy term life insurance, they pay premiums for a while, and then they stop: They need the money for other purposes, their circumstances change, they stop paying and the policies lapse. They do not “optimally exercise” their life insurance policies; they let them lapse when they shouldn’t.

We have talked about this a couple of times because, for a while, financial investors like Apollo Global Management were able to buy up a lot of these policies from insurance customers. The financial investors, unlike the ordinary customers, did optimally exercise; they kept paying the premiums whenever doing so had positive expected value. This was bad for the insurers.

It wasn’t as bad for the insurers as it could have been. You could imagine an extremely efficient world in which everyone took out millions of dollars of underpriced life insurance and immediately turned around and sold it at a profit to Apollo, which then made a lot of money at the expense of insurance companies, which then had to significantly raise the premiums they charged for life insurance. This mostly did not happen in the real world, in part for legal-risk reasons, but in large part because people are busy. If I ran up to you saying “hey you can make a quick buck by taking out some life insurance and selling it to Apollo,” you would almost certainly think I was nuts. That sounds gross and weird. What’s the catch? (Will Apollo, uh, hasten your demise?) Also it’s not really a quick buck; you’d have to get a physical and fill out forms. The underpriced option that insurance companies sell to customers is just comprehensively too hard for the customers to monetize, which means that it can keep being underpriced.

Those are my favorite examples, but surely the most important example is the 30-year fixed-rate mortgage prepayable without penalty. In the US, if you want to buy a house, a bank will probably lend you 80% of the value of the house at a fixed interest rate of, say, 6.65%. If interest rates go up, you don’t care: You pay 6.65% for 30 years. If interest rates go down, though, you can refinance at any time: You can go to a bank, take out a new 30-year fixed-rate mortgage at 4.5% or whatever, and use the money to pay back the old mortgage. You have a valuable interest-rate option: You have locked in a maximum interest rate for 30 years, but you have no minimum rate. In some approximate sense, you’ve got a floating-rate mortgage with a rate cap struck at today’s rates. 

And the bank [4]  will sell you that option quite cheaply: That 6.65% rate you’d pay on a 30-year mortgage is lower than the yield on Meta Platforms Inc.’s 30-year bonds, [5]  even though (1) Meta is a giant AA- rated company and you’re just a person and (2) Meta can’t prepay its bonds without penalty: If interest rates go down and Meta wants to refinance the bonds, it has to pay bondholders a make-whole payment that essentially captures the value of the interest-rate savings. 

Why do banks underprice this option? The basic answer is “because people do not optimally exercise it.” For one thing, when rates go down, people often do not refinance (or otherwise prepay) their mortgages. They are busy! They have a lot going on, refinancing takes time and is a pain, they do not trust mortgage bankers or want to spend a lot of time dealing with them, mortgages are confusing and it is hard to understand whether refinancing is a good deal. They are not checking in on interest rates every day, waiting for the optimal time to refinance. They’re getting tons of junk mail from mortgage companies saying “The Optimal Time to Refinance is NOW,” but that sounds fake and they throw it away.

Here’s “Why do borrowers make mortgage refinancing mistakes,” by Sumit Agarwal, Richard Rosen and Vincent Yao (2013), which gives a flavor of the problem [6] :

The decision to refinance a mortgage optimally requires solving a complicated system of partial differential equations. This can prove to be problematic because significant cognitive ability often is needed to properly make optimal financial choices. … 

Refinancing a mortgage requires not only that a borrower select an interest rate at which she is willing to refinance, but that she take the actions necessary to refi (such as contacting a broker or bank and completing paperwork). Agarwal, Driscoll, and Laibson (2012) argue that borrowers do not actively monitor mortgage rates and, even if they notice that the mortgage rate has reached their “trigger rate” for refinancing optimally, they may not immediately refi because they are too busy. …

They do not always monitor mortgage rates closely. Borrowers are faced daily with many complicated, time-consuming choices. Given a binding time-budget constraint, distracted borrowers may only be able to make certain decisions at stochastic intervals – or, put less formally, when they have a spare moment.

For another thing, when rates go up, people often do prepay their mortgages, even though that is suboptimal for them and a windfall for the bank. If interest rates move from 6.65% to 8%, a lot of people with 6.65% mortgages will nonetheless pay them back and take out new 8% mortgages. People move. If you get a new job in a new state, or if you have children and need more space, you might move to a new house. That will normally mean selling your old house, paying off the mortgage, buying a new house and taking out a new mortgage. If rates have gone up, you pay the higher rate. You’re optimizing things other than your mortgage rate.

Mortgage rates have gone up a lot since 2020, which has led to complaints and proposed solutions — like assumable or portable mortgages — that would allow people to avoid suboptimal prepayment. But if people could easily avoid suboptimal prepayment, mortgage rates would be higher. Right now, banks will give you a mortgage with a cheap prepayment option (that is, a relatively low rate), because they know that lots of people will exercise it suboptimally (because they move). But if that problem were solved — if you could keep your old low mortgage rate when you moved — then the banks could no longer sell you the option cheaply. The fact that people regularly move, even when mortgage rates have gone up, keeps mortgage rates lower than they otherwise would be.

Anyway Bloomberg’s Jack Trapanick and Scott Carpenter report:

Lenders say artificial intelligence will help homeowners whose mortgages are ripe for refinancing secure a new, cheaper loan far faster. The result could also squeeze investors in the $9 trillion market for mortgage bonds.

When interest rates fall, only about a third of homeowners who could save substantial sums by refinancing actually do it, according to research from Morgan Stanley. That’s because candidates don’t know they’re eligible or don’t want to go through the notoriously drawn-out and tedious process.

The number of takers is likely to rise, though, as mortgage lenders embrace AI to churn out approvals in a fraction of the usual time. Rocket Mortgage says a borrower can get from application to rate lock in just 30 minutes, and it’s aiming to cut that to 10 minutes. Rival United Wholesale Mortgage says initial approval can take as little as 15 minutes. Better.com, another digital lender with a small slice of the market, claims the firm can do it in only two.

Faster turnaround could double the percentage of eligible homeowners who refinance to perhaps 60%, according to a report from Morgan Stanley strategists including Jay Bacow, co-head of securitized products research. If that happens, they wrote, AI could make the 30-year mortgage seem like something “closer to a floating-rate instrument that only floats down.”

The Morgan Stanley analysts estimate that, if this happened, mortgages would “become more costly as investors demand extra interest to compensate for the added risk — perhaps one or two tenths of a percentage point.” “A floating-rate instrument that only floats down” should pay a higher rate than a 30-year-ish fixed-rate-ish instrument. Right now, US mortgages are more like the latter; in this imagined AI-assisted future, they’d be more like the former. (People would still move, though.)

This is, perhaps, a big deal on its own: There’s like $15 trillion of US residential mortgages outstanding, so a 10-basis-point cost increase would be $15 billion a year. But one could imagine a more general story like “broad AI adoption will lead to consumers exercising options more optimally.” Like: If you have three credit cards, an AI plug-in in your browser could automatically choose the most rewarding one for each online transaction. Or: Instead of “solving a complicated system of partial differential equations” to decide whether to refinance your mortgage, you just tell your AI “hey AI let me know when I should refinance my mortgage,” and the AI solves the equations and pings you when it’s time. Or: You could set up an AI agent to browse high-yield savings account offerings each day and move your money to the best one. When some US regional banks ran into trouble a few years ago, people attributed the scale and speed of the problem to the rise of the internet and social media: With social media, you could quickly learn rumors about a bank’s instability; with online banking, you could quickly move your money from an unstable bank to a safer one. “Game’s the same, just got more fierce,” said the vice chairman of the Federal Deposit Insurance Corp. Agentic AI could make it fiercer.

A lot of the consumer financial industry is based on consumer irrationality and inattention. Consumer financial products are built, and priced, for a world in which rationality and attention are scarce. AI could create a world in which rationality and attention — not human rationality and attention, but some bot that can search the web and do math — are abundant. What will that mean for credit cards and life insurance and mortgage rates?

Historically, private equity was a lumpy business. You and a couple of your buddies started a private equity firm, you went around looking for companies to buy, and occasionally you bought one. When you did, you’d need to write a big check. You spruced the company up for a while, and then you sold it again and received a big check. You might do this for 10 years, writing a dozen big checks to buy companies and receiving a dozen big checks for selling them. You might go months between checks.

Where did you get the big checks to buy the companies? Well, you had some investors, some limited partners who agreed to provide the money to buy the companies. These investors were the sorts of people who were willing to lock up their money in risky investments for years to earn higher returns: endowments, pension funds, sovereign wealth funds, super-rich individuals. You could imagine a system where, each time you wanted to buy a company, you called a handful of investors you knew and said “hey want to buy this company with me?” You’d cobble together money from investors for each deal, and then pool their money to buy the company. This exists — it’s called the “independent sponsor” model — but it’s a bit cumbersome; if you want to buy a company, it’s good to have the money lined up in advance.

So the more normal approach in private equity has been to raise a thing called a “fund,” which is maybe not exactly what it sounds like. (A more specific name is a “drawdown fund.”) A fund is like: You call some investors you know, you say “hey want to buy some companies with me,” and they commit a certain amount of money to you. You raise, say, $10 billion of commitments from those investors, and then you use that $10 billion to buy companies over a few years. You don’t have to raise money for each deal; the money is already there. Well, not quite. It’s not like the investors give you $10 billion and you put it in the bank and use it to occasionally write big checks. You don’t write that many checks; you don’t need to keep all that money in a checking account. Instead the investors give you commitments: They promise to give you that $10 billion when you need it, and then, when you do find a deal, you call on their commitments. You find a company to buy, you need $1 billion of equity, so you call 10% of each investor’s commitment. They are contractually obligated to wire you the money, and they do; you pool their money together and use it to buy the company. You do this each time you buy a company, and each time you sell a company you return some money to them.

For your investors, this is probably better than giving you all the cash upfront to put in a checking account: They can probably earn more money on their cash than you’d get in your bank account. It is also good for you, because your investors traditionally measure your performance based on your internal rate of return, measured from when you call capital to when you return it. Raising money from investors and parking it in a checking account earning 1% for two years lowers your IRR, and in fact there is a whole business of delaying capital calls to increase IRRs.

Still it is somewhat annoying for your investors. An investor that puts $1 billion into your private equity fund doesn’t wire you $1 billion that day. It wires you portions of that $1 billion, in lumpy unpredictable increments, over some period of years, as you find companies to buy. In the meantime, it has to do something with the money. It can’t invest it long-term; you might demand it any time on somewhat short notice. It has to keep an eye on its inbox for your capital calls, and when it gets one it has to move money around to send to you.

This is the traditional story, but modern private equity is a bit different, in two related ways. First of all, it’s giant. Private equity is not a niche business done by a couple of buddies for a handful of sophisticated clients; it owns perhaps 20% of the economy. It is still a lumpier business than, like, high-frequency trading, but it’s not like the giant modern alternative asset managers that evolved out of early private equity firms are going months between writing checks. If you’re deploying a trillion dollars, you’re doing a lot of deals.

Second, it is marketed to retail investors: If you’re deploying a trillion dollars, some of that comes from ordinary people’s retirement accounts. And ordinary people can’t really invest in drawdown funds. You can’t go out and raise, like, $5,000 commitments from thousands of retail investors, find a company to buy, and send each investor a capital call for like $500. It’s just too much of an administrative pain, and some of them will have moved and your capital calls will bounce back, it’s a mess. If you are going to sell private equity products to retail investors, you have to take their money upfront: If you raise a $1 billion retail fund, that means collecting $1 billion in cash and putting it somewhere until you need it. Maybe in a checking account, but not necessarily. If you’re constantly doing deals, and you have lots of existing portfolio companies and lots of other investors, you can probably find a more private-equity-ish use for their money. Put the $1 billion into some of your existing companies to pay out old investors or whatever. The money doesn’t have to sit idle for long.

Similarly, it’s not that convenient to go raise $1 billion from retail investors all at once: There are so many of them, and you can’t contact them all in a month to raise a fund. It’s better to have an “evergreen” fund where people can put in money whenever they want, so that if an investor gets an inheritance or wins the lottery or gets a bonus at work, she can immediately bash some of the money into your fund. Again, if you have a very lumpy business where you only deploy money every few months, this is annoying for you; the money coming in has no real correlation to the money being spent. But as your business gets bigger and more complicated, it’s less lumpy; if someone bashes $5,000 into your fund on a Tuesday you can probably find somewhere to put it by Friday.

And so your retail offering is “put money in whenever you want, however much you want; we’ll take it immediately and find something to do with it.” (It might also allow investors to take money out, at least some of it, at least some of the time: As money comes in from new investors, some of it can be used to cash out old ones, and retail investors love liquidity.) Meanwhile the institutional offering is “commit a fixed amount of money for years, and then we’ll call you for some of the money at times and in amounts that are convenient for us, and eventually we’ll pay you back when we decide to sell.” Which is … you can understand why it’s that way, but it’s kind of worse customer service, no? Like the retail product is kind of better?

Anyway the Financial Times reports:

Institutional investors are putting money into Blackstone and KKR vehicles set up to woo wealthy individuals, a move by the traditional backers of private equity that could ultimately upend the sector’s traditional 10-year fund model.

Evergreen funds, which allow investors to withdraw funds at regular intervals rather than lock up their capital for long periods, have become increasingly common as a way to make private equity and credit accessible to wealthy individuals.

But the funds, which tend to charge lower fees and target lower returns than the industry’s traditional closed-end funds, have also started to draw in institutional investors.

Blackstone’s evergreen funds for individuals had raised a “small percentage” from institutions, the firm’s head of wealth Joan Solotar told the FT. Interest had increased in the past year, she added. “It will continue to evolve and grow.”

KKR’s head of client solutions, Eric Mogelof, said it had recently launched institutional share classes in its buyout, credit and infrastructure evergreens to meet “growing demand”. …

Evergreens, meanwhile, take all of an investor’s commitment as cash on day one, removing the need for complex cash flow management that small organisations can find cumbersome. The funds’ need to keep cash on hand for redemptions generally leads to lower returns.

“We are at the beginning of a trend, but it’s going to continue to happen as the evergreen market evolves,” said Hugh MacArthur, chair of private equity at consultants Bain & Company. “Institutions are going to want the same . . . conveniences as individuals.”

It’s weird when the individual product is better than the institutional one. [7]  Part of the explanation is that it isn’t really better — evergreens “tend to charge lower fees and target lower returns” — but part of it might be that the institutional product was invented first, and the technology has advanced since then.

Arguably the way jobs work is:

1. You work at a job, they like you, they pay you.
2. The amount they pay you is, generically, enough to keep you. 
3. Therefore, to get more money, the main thing you have to do is to credibly demonstrate that someone else will pay you more.
4. The simplest way to do this is to go out and get another company to offer you a job at a higher salary, and then bring that back to your current job and say “see?” And then they give you a raise.
5. The other company in this scenario has done you a very valuable service: It has gotten you a raise at your current job.
6. What does the other company get out of it? I guess the answer is “some possibility of actually hiring you”: Maybe you’ll like the other company so much that you’ll actually change jobs, or maybe your current company won’t give you a raise and you’ll kind of have to change jobs. If it makes a lot of these offers, sometimes it will hire people!
7. But, in many cases, the outcome — in some sense the expected outcome — is that you stay at your current job and get more money, and the outside company gets nothing. It has given you something of value — a bid on your services that you can use to extract money — for free.
8. Inefficient!
9. It should charge you. Sign a contract like “sure we’ll interview you for this job, but if we give you an offer and you just use it to extract a raise from your current employer, you have to give us 10%.”

One could quibble. (Doesn’t the signal value of the outside offer go down, if the outside company is mostly making the offer to get a share of your raise?) Anyway:

Millennium Management is suing a Dymon Asia Capital employee for millions of dollars in a Hong Kong court after she accepted and later reneged on a job offer, said people with knowledge of the matter.

Millennium, a $92 billion hedge fund firm, is seeking HK$19.7 million ($2.5 million) from Hong Kong-based Dymon portfolio manager Tang Lin. That amount is intended to cover so-called “liquidation costs,” expenses that Millennium incurred while preparing for her to join the firm, said the people, who asked not to be identified discussing private information.

It is the latest example of the world’s biggest multi-strategy hedge funds fighting back at the practice of “gazumping.” The informal term, originally from the British real estate market, refers to when a seller initially agrees to an offer but then accepts a rival offer before the deal is done. Tang ultimately chose to remain with Dymon, people said.

I don’t know enough about her pay package to speculate, but it’s conceivable that paying Millennium $2.5 million for its participation in the sequence of events that led to her staying at Dymon could be a bargain for her.

Jane Street’s growing pains. Private Credit Investors Prefer to Be Trapped Than Take 26% Loss. Todd Boehly’s Insurer Vows to Slash Pile of Collateral Loans. Nvidia Sees AI-Fueled Demand Boosting Sales 70% Next Year. Nvidia Agrees to Buy Open Source AI Platform Hugging Face For $12.9 Billion. Bitcoin treasury companies shed $80bn in value as business model unwinds. KKR Agrees to Pay Record $250M Penalty for Serial Violations of Federal Premerger Review Law. Wells Fargo Steps Up Wealth Hiring After $1.5 Billion Revamp. Single Stock Leveraged ETFs Not Good for Japan, Regulator Warns. Harvard Business School explored European outpost after Trump’s visa threats. Greer Suggests US Needs to Consider Bans on Canadian Goods. The $2 Billion Brawl Over a Ruinous Wood-Pellet Trade. Junior consultants called back to office as AI increases need for human skills. Miami’s Wealthy Are Beating the Traffic With Helipads and Private Docks.

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!