Financial Institutions and Franchise Value

Plus! SVB Updates; Hit Economics; AI Costs; Lending Against Regulatory Uncertainty; AI Assistants; Diff Jobs

Financial Institutions and Franchise Value

The financial services industry is fundamentally in the business of moving claims on future cash flows from one balance sheet to another. Some of this is done in a purely transactional way, whether it's Visa and Mastercard capturing basis points on small transactions, or investment banks and realtors taking a larger cut of larger deals.

And, of course, a big chunk of the industry is devoted to actually acquiring and managing assets. That asset-accumulation business is generally a form of intermediation between a demand for investments with certain characteristics, and a supply of investable opportunities that represent the raw material for manufacturing the exposure investors want. Which means:

What all these machinations create is a universe of equity-like assets that represent exposure to a portfolio of assets that also have some kind of market value. So there's a strong gravitational force pushing every publicly-traded asset-accumulating financial company to trade at book value. (For a quick primer on book value and its complexities, see this piece from our sister newsletter, Capital Gains.) Since a financial intermediary is typically investing in some category of assets, any time they can raise $1 from outside investors and turn it into, say, $2 of market value by putting it in the right category, that's exactly what they'll do. And this basic thinking also applies to entities like ETFs, mutual funds, and hedge funds. The fee for a managed asset that allows investors in and out at net asset value is basically an amortized version of the premium you'd pay to buy and hold it over a longer period.[2]

But some of these companies routinely trade at a premium to book value, mostly as a proxy for investors' views on their unique advantages. An insurance company is not just worth the sum of its financial assets minus the amount it will likely pay out in claims; if it can earn a superior return on those assets, or (more likely) if it has some structural cost advantage in sourcing new business, that gets reflected in its valuation. Back in the early 1980s, shares of George Soros' Quantum Fund were traded on the secondary market at a premium. Liquid products like oil futures, currencies, and equities can trade at a premium to face value if there's a good manager attached.

Similarly, a bank can be worth more than the sum of its assets when it has either proprietary access to good borrowers or proprietary access to good depositors. SVB, in its 2019 incarnation, had both: the company was the default choice for startups to deposit their cash, and was also able to offer unique financial services, like mortgages for the equity-rich but cash-poor, or lines of credit for investment vehicles. But there are microcosms of this in other parts of the banking system—such as tiny institutions that happen to have a great relationship with one category of depositor or that caters to an unusual geography or demographic. (For example, a US bank catering to a community that speaks a language other than English will limit its growth prospects, but will also have access to unique lending and deposit opportunities (Edit: Apologies, an earlier draft said "unique lending and deposit insurance). And indeed, startups speak a very different language than established companies, at least in terms of what kind of growth and cash flow profile makes them a good credit.)

The premium an asset-accumulator deserves to book value is a useful number to keep in mind, especially when the value of the company's assets has been impaired and its survival requires someone else to buy the business or otherwise shoulder the risk. This premium partly a measure of value-add, which, in a crisis, is useful to two different groups:

  1. For investors, it tells them whether a rescue needs to be evaluated based on its positive externalities or based on the direct profits from providing it. In SVB's case, that value has been seriously impaired—banking with them went from a decision one didn't have to think about to being the only thing people have been thinking about in short order.
  2. For the government, it contributes to the math around bailouts. If there's a company that's worth a discount to book value, bailouts are wealth-destroying because they throw good money after bad.[3] Bailing out a business that historically traded at a premium to book value is just one step in structural reforms, but it's the most urgent one—whatever risk got the financial system in trouble is typically the last risk anyone wants to take once that trouble arrives, so there's generally breathing room before the next crisis.

A good measure of the success of banking regulation is: what fraction of a bank's customers recognize the terms "10-Q," "Uniform Bank Performance Report," "Call report," etc. Banks create value in large part because they can create money, but "creating" money" means that deposits at that bank are always, unquestionably, worth exactly 100 cents on the dollar.

Banks, as financial intermediaries, are economically required to take some risks unless they get replaced with a narrow bank. The menu of risks they can take is constrained by regulations. And those regulations tend to get written in the wake of big losses. The Dodd-Frank rules did make the financial system more stable, but the problem they were designed to address was a liquidity crisis driven by wholesale funding markets seizing up in response to credit losses. It's been a long time since the financial system was more threatened by rates than by credit, so the rules were naturally more flexible there. So the rules were friendlier to companies that chose to take interest rate risk instead. But as it turns out, SVB's model was already taking rates risk, since their deposits were fueled by a strong equity market that was the last place investors could get high returns in a low-rates world. From a Diff piece in late February ($) (which, ironically, focused on how big a bank run would have to be to impair SVB, and on how they'd presumably raise equity if one started—simpler times):

What is interesting is that SVB's funding and its assets are now tied to the same risk: higher interest rates have a disproportionately negative effect on the cash balances of negative cash flow companies, because they decrease the financial salience of future profits relative to current losses and thus make fundraising harder.

One surprise in of all of this is that if they had invested in the CLO market, periodically the subject of dire warnings about financial instability, this specific problem would have been avoided: CLOs basically trade rates risk for credit risk, by returning floating rates but tying repayment to the financial fortunes of levered companies.

Agustin Lebron's The Laws of Trading makes the point that since traders are being paid to take risks, they should only take the risks they're being paid to take. If your edge is in predicting oil prices, and you find a way to express a view on oil prices that also involves credit risk (say, buying an oil company's junk bonds) or counterparty risk (doing an over-the-counter bet with a flimsy counterparty), you're basically volunteering your firm's capital for some unpaid charity work on the side. This is frowned on, for good reason. SVB had a comparative advantage in taking rates risk on the deposit side, which meant it had a comparative disadvantage in taking such risk on the asset side. One of the goals of post-crisis regulation was to make bank CEOs think more like bankers and less like traders, but at least in this case, a trader asking "what's my rates exposure, and what can I do to hedge that if I don't think I have an edge in predicting rates?" would have been nice.

  1. There's sometimes a safety tradeoff, too, but in general when you're looking for something equivalent to cash, you're not looking for something equivalent-to-cash-unless-everyone-needs-cash-including-you, even if it pays a slightly higher yield. Inevitably, after long periods of boring headlines, this criterion gets relaxed. ↩︎

  2. Which also leads to some really fun ideas. Suppose you're investing in a company that either is explicitly a closed-end fund, or that technically isn't one but operates like one. If you buy shares of the fund itself, and short everything it owns, your exposure is hedged—except that the difference between market price and book value reflects the expected future outperformance or underperformance of that manager. You can basically construct a bet on the market's perception of a manager's future performance! ↩︎

  3. As in the case above, there may be positive externalities. The SBA loan business ($), for example, amounts to a subsidy for loans that are too inconvenient to underwrite normally, but are socially useful since they provide capital to small businesses. ↩︎

A Word From Our Sponsors

This newsletter is brought to you by Tegus, your modern research launchpad.

Investors waste too much time looking for actionable information across disconnected sources. Tegus streamlines the information you need to move quickly with conviction and make better decisions. With access to the largest transcript library, expert calls with no markups, driveable financial models, and searchable SEC filings, you’ll be the smartest expert in the room.

Right now, readers of The Diff can trial the Tegus platform for free.


SVB Updates

Yesterday evening, after a day of struggling to find a mutually-agreeable buyer for SVB, the Fed, Treasury, and FDIC jointly announced that they were guaranteeing all deposits of Silicon Valley Bank, and of Signature Bank, which was also shut down over the weekend. (Signature shares were down 37% in the last week, initially more from Silvergate contagion than from SVB.) The Fed also has a new facility for lending against treasury bonds and mortgage-backed securities, valuing them at par. If this had been available a week ago, SVB's held-to-maturity book with a cost of $91bn and a market value of $76bn would have been treated as ~$91bn of collateral.[1] This is pitched as having no cost to the taxpayer, and for technical reasons it's good for that to be true. There is, of course, a cost to providing a backstop, even if it doesn't get used; if you were buying such a backstop from a third party, you'd certainly expect to pay for it. But in another sense, since financial crises are expensive, the net cost to taxpayers can be negative. There should, at least, be a wide confidence interval around the net cost/benefit, with positive and negative numbers both possible.

There are two important implications of a par value-based funding guarantee:

  1. This makes banks’ accounting more realistic. The original argument behind letting banks defer recognizing mark-to-market losses on held-to-maturity securities was that the banks didn’t have to sell. Lending against them makes it so—banks now have liquidity even when they don’t have mark-to-market equity.
  2. A more subtle consequence of this is that it constitutes partial monetization of government debt. If a bank can borrow money from the Fed based on the par value of its treasury bonds and agency securities, then the economic substance of buying or holding them is that it’s close to equivalent to holding cash, but with a higher interest rate. That’s close in economic substance to a world where the Fed creates money to buy these securities directly, though there’s an extra step (a bank needs to buy them and then borrow against them).

This is a good time to take a step back and ask what a financial crisis is and how it gets resolved. There are many ways to describe it, but one of the most powerful models is that it's a bull market in dollars. And, as with some other bull markets, only the best will do! In the 90s, "the market" as a whole peaked in March of 2000. But within that market, trends varied: tech, of course, did very well right up to the end, but the Dow, weighted more to older industries, was basically flat from May 1999 through January 2000. And then the Dow went down in the first few months of January, before staging a big rally as the Nasdaq collapsed. In other words, the late stages of the 90s equity bubble constituted a rotation away from some equities and into the growthiest ones, and a vicious reversal soon after.

In a financial crisis, we generally go straight to that terminal bubble phase, when only the genuine article will do. Specifically, as some cash-like assets turn out not to be cash after all, the ratio of available cash to cash needs goes down. Meanwhile, demand for cash goes up. That's an unpleasant situation.

The only thing that resolves it is returning the money market to equilibrium. This can be done through some combination of 1) just letting things work out on their own, which caused plenty of macro volatility in the 19th century and which, in our more levered and interconnected era, would be catastrophic, 2) supply new cash to offset the transformation of near-cash into not-cash (this is why charts of M2 look so scary; they don't show the disappearance of credit that performed the role of cash, just the growth of cash to offset that absence), or 3) magically transforming not-cash back into cash, either through merging distressed financial intermediaries with stronger institutions or by guaranteeing some or all of their deposits.

Guaranteeing deposits does, undeniably, create some level of moral hazard. But depositing money in a bank should not be viewed as a risky decision in need of careful underwriting; since larger banks are (correctly) viewed as still too-big-to-fail, a regime where there's significant uncertainty about bank deposits is one where there isn't an economic justification for any bank other than the biggest handful. If banks are going to have a government guarantee, it doesn't make sense to extend this guarantee only to the biggest.

Hit Economics

If you ever want to lose a lot of money quickly, make a not-especially-high-margin consumer durable that suddenly gets popular. The maker of the Instant Pot has hired restructuring advisors after a 50% drop in sales for the category ($, WSJ). The difficulty for this category is that they have to order inventory based on anticipated demand, and it's almost impossible to forecast when a trend reverses—especially for a small company that doesn't have direct experience with trends of that magnitude. At some point in every one of these growth stories, there's a moment where they have excess inventory that's hard to move, and selling it fast means hurting the brand with lower prices.

AI Costs

An open source developer wrote a port of Meta's LLaMA language model that runs locally on a Macbook. It's been surprising to see how quickly the cost of running AI models drops, even though the cost of building and running the newest ones continues to rise. This significantly weakens the bear case for AI articulated in The Diff earlier this year, that the high fixed costs make the economics look more like an old-line industrial company rather than a software business. If the computational cost for a useful AI, rather than a novelty, gets low enough that they can run locally, then some of the economic potential remains with the companies shipping new systems, but a lot of it moves to the companies that can effectively market AI-driven products before they're commoditized.

Disclosure: Long META.

Lending Against Regulatory Uncertainty

An industry that popped into existence on Friday and then disappeared yesterday was the provision of various kinds of bridge financing for companies that kept most of their cash in SVB and needed to pay bills and make payroll. This took two forms: Jefferies offered to buy claims on SVB accounts at 70 cents on the dollar ($, The Information). That was an aggressive price given the 95% recovery rate calculated by an analyst at... Jefferies. (The best bet is that these numbers were produced by two completely separate teams at two different times, and of course the midpoint of a valuation is not the same number as what a market-maker quotes in a rapidly-evolving situation.)

Meanwhile, Brex was offering emergency lines of credit, to be funded over the weekend. That plan obviously doesn't need to pan out right at the moment given other developments. And it's unclear what the rates would be, though Brex said they didn't plan to turn a profit on it. But it has two implications:

  1. On the credit side, even the existence of a willing partner who is making an effort to supply capital can reduce the risk of contagion. Financial crises happen through surprising nth-order effects, and promising a backstop decrements the N in question.
  2. It's fundamentally hard for financial intermediaries to differentiate themselves, since so many of the products are commoditized. But one time when money is not commoditized is when it's most needed. Brex is in the very fortunate position where they've been able to demonstrate that they're there for customers, but, assuming the raise didn't complete, Brex is not in the position where it's done a bunch of emergency lending deals with customers who will immediately regret it. (Jefferies, on the other hand, will have to weigh the relative benefits of making a 43% return over the weekend from buying distressed credit and selling government-backed securities against the long-term impact on their tech underwriting.)

Disclosure: I worked at a Jefferies subsidiary years ago, working on things that were fun but in no way related to distressed claims trading or bank analysis.

AI Assistants

LinkedIn is incorporating AI-generated prompts for conversations and posts. And it is, in fact, a reasonable place for them: good business communication is an art form, but the median business-related communication, especially at the start of a discussion, has high redundancy. The recipient knows how to filter it out, but it's still expected.

Microsoft, LinkedIn's parent company, has extensive experience with expediting the creation and dissemination of boilerplate. This older New Yorker article claims that when the team PowerPoint started work on the "AutoContent Wizard" feature, which built pre-outlined presentations around user-selected topics, originally got that name as an internal joke. It turns out that some kinds of communication are simple enough to be automated, but not superfluous just yet.

Disclosure: Long MSFT.

  1. We don't know the exact prices at which they bought, and it's possible that they paid a premium to par value, but for MBS in particular, the rate decreases that make simple fixed-income assets trade at a premium in turn lead mortgages to get rapidly refinanced. ↩︎

Diff Jobs

Companies in the Diff network are actively looking for talent. A sampling of current open roles:

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.