Welcome to the free edition of The Diff. This newsletter goes out to 31,219 readers, up 514 since last week. In this issue:

• $1 Trillion Worth of Risky Loans Packaged Into Highly-Rated Securities is not the End of the World

• Diff Jobs

• Netflix's Explore/Exploit Strategy

• EVs and Platforms

• Third-Party Sellers

• An Entrepreneurial Country

$1 Trillion Worth of Risky Loans Packaged Into Highly-Rated Securities is not the End of the World

A good way to look at financial history is that every post-crisis period overcorrects whatever the previous crisis got wrong. After 2000, rates were cut aggressively, and growth companies were viewed skeptically, so a lot of leverage flowed into a) financing buyouts, and b) financing mortgages. After 2008, policymakers decided that extensive stimulus, fiscal or monetary, could easily ignite the next bubble before the effects of the previous one had been fully worked off, so we ended up with a long era of subpar growth and low real rates. In 2020, we basically reversed this: remember the brief and wonderful period where it looked like recessions were a policy choice, and that it was entirely possible to spend enough money to actually reduce the poverty rate during a recession?

This instinct is healthy: making mistakes is bad, but making the same mistake twice is just egregious. But it does mean there's an excessive amount of pattern-matching. And one place this shows up in is analysis of the leveraged loan marketplace. Run through a list of its characteristics and you'll get serious déjà vu:

• Borrowing has risen at an unusually fast pace for an extended period, this time from the corporate sector rather than households.

• Higher borrowing has led to deteriorating credit standards. More loans are made at higher levels of leverage: in 2011, most most leveraged loans were made to borrowers whose total leverage was under 5x their EBITDA, and right now a majority are to borrowers above that. As of 2018, about a third of leveraged loans were to companies borrowing at least 6x their EBITDA, and more recently the average has been around 6x, with tech multiples even higher.

• Loans used to have strict financial covenants, but starting in 2012 this was dramatically relaxed, and now most loans have very limited ones.

• These loans are being packaged into structured products, CLOs, where a bundle of junk-rated corporate debt backs a product that's mostly rated AAA. And these AAA-rated CLO tranches are being bought by banks, insurers, pensions, and specialized funds.

Here we go again? That is the thrust of some media coverage, like this WSJ piece from April looking at European banks' exposure ($), and this much more alarmist piece from the NYT in 2019 (helpfully illustrated with a big image asking: "Another 2008?"). And it is true that there are parallels, and that synthesizing triple-A-rated assets out of junk sounds like witchcraft.

On the other hand, it's easy enough to overfit. After all, there is a very safe, highly liquid asset class that you're told never to worry about, but that's backed by a complex mix of products that are a) all less liquid, and b) all substantially riskier: if you have a bank account, you, too, are participating in the alchemy of risk reduction. It's entirely healthy: having a low risk and liquid store of value is useful, transmuting savings into investment is also useful, but doing this in a simple one-to-one way means creating an unstable financial system.¹

And for the corporate borrowing market as a whole, there are a few stylized facts worth remembering:

• The number of AAA-rated bond issuers has declined from 60 in 1980 to two. AAA-ratings are indeed quite predictive, unless something really bizarre happens almost overnight.

• Banks have broadly gotten more cautious about lending, especially to smaller and less creditworthy companies.

• Meanwhile, roughly 70% of corporate borrowers are below investment grade.

• Banks' leverage has declined dramatically since the 2007-era peak. In 2007, Bear, Lehman, and Morgan Stanley were all levered at or over 30:1, now the biggest US banks are levered 12 or 13:1.

So banks are quite a bit less likely to blow up lately, and they aren't especially excited to take idiosyncratic risks. Enter the CLO: highly-rated tranches are most of the pool, but the typical structure is that there's an equity layer that takes the first loss, a mezzanine layer that takes losses after that, and only then do the AAA-rated slices take a hit. This useful BIS paper from a few years back shows the typical structure as around 5-10% equity, 25-30% mezzanine, and the rest investment-grade. And that is a lot of collateral: the underlying loans need to lose, on average, 37% of their value before the highest-rated slices take any losses. And those losses would have to be very severe indeed to eat into the banks' own equity.

Essentially, the structure of the CLO market has enabled two kinds of outsourcing. First, banks that buy CLO debt tranches are outsourcing a lot of their underwriting to the managers of the product. Someone is still figuring out which loans need to be made, but buying a diversified pool of these assets converts the loan from an idiosyncratic single-company bet into a macro bet on corporate credit. Second, banks are outsourcing some of their capital requirements to other investors. Typically, asset managers and hedge funds are overrepresented as buyers of mezzanine and equity tranches, while banks, insurers, and pensions buy most of the highly-rated pieces. For banks, this means that they're taking a loan that would be risky to hold on their books, shifting most of that economic risk to someone else, and keeping a lower-payoff slice on hand.

The structure of these loans is also well-tailored for banks. They're floating-rate rather than fixed, which has a few effects. First, it means that their returns line up with the cost of short-term financing, so banks aren't taking long-term interest rate risk when they buy them. If a CLO pays some premium to the risk-free rate, and banks' deposits pay some discount to it, they'll collect roughly the same spread regardless of where rates go. Rising rates were the proximate cause for a few financial crises (the first round of the S&L crisis in the late 70s and early 80s, Europe's debt problems in 2011), and have been at least a partial factor in many more.

For borrowers, exposure to interest rates can be a problem: their borrowing costs continuously reset as rates change. All else being equal, the more dubious a borrower the more their ability to pay correlates inversely with rates. Rates rise when the economy is strong and drop when it's weak. (Depending on how you slice the data, sufficiently low-rated bonds can have negative duration; higher rates reduce the value of future cash flows but that's more than offset by an increase in the probability of receiving them—for more, see this duration explainer.) All else is, sadly, never equal: if rates only move in response to growth, this is completely true, but if rates also move in response to exogenous inflation, then those borrowers run into some problems. Some of them hedge, and it's a reasonable bet that the most aggressive hedgers are the companies that have the least confidence in their pricing power. But this does mean that we're at a point where the model will be tested more than it has in the past.

And, as basically always happens in finance, that test comes at a time when assets have never been higher, and when their composition has changed substantially. One limiting factor for bank lending is that the cost of underwriting a loan doesn't scale much with the size of the loan, so if fixed costs rise or if banks get broadly more reluctant to lend, that shows up as an increase in the minimum viable size of the loan (or, equivalently, as higher spreads for smaller loans). But there is one good way to manufacture new borrowing at scale: a leveraged buyout. A company that borrowed $20m/year for growth might struggle to find a compliant lender, but if that same company grew a bit more slowly, and then got bought out in a deal requiring $200m in debt, that would be a chunky enough transaction to justify the time and effort required to understand the company.

So CLOs have largely been an asset class built on the leveraged buyout boom, which also means they've been an asset class that reflects where that boom happens: CLOs have gotten increasingly concentrated in tech deals. This is partly a reflection of tech market dynamics over the last decade. For any company that broadly screens as a tech business, the two exit paths are acquisition or IPO. As long as a company seems on track for one of these, it can raise money and keep growing. But every round before the growth stage creates "valley of death" risk: the business might scale, but reach a point where it clearly won't IPO. If it can't find a strategic acquirer, it's stuck. What private equity firms have found is that 1) the cash flow profile of a software company that gives up on growth can be very attractive—lots of recurring, high-margin revenue, and much less sales and marketing or R&D expense needed if the company isn't seeking many new customers, but 2) the sorts of founders who aim to create IPO-track companies that grow at 50% a year, mostly through new customer acquisition, are very ill-suited to running a company that tries to maximize cash flow while hiking prices 10% a year and bleeding 5% of its customers each time. So PE can simultaneously cash out the founders and investors (at less than they were hoping for but more than they'd get elsewhere), drop in new management, and create a big chunk of debt.²

Zooming way out, this is the financial system doing what it's supposed to do best: building financial structures that suit companies' economics, and matching resources to where they're needed—which includes giving employees a nice exit package and a swift goodbye if they signed up for a growth company and now work at a business that isn't going to grow.

Another big risk factor, though, is from the financial system doing something it does but ideally shouldn't: making money off of zero-sum technicalities, and aiming for fee maximization over returns. This is harder to quantify, and some parts of it are debatable; one party's "that's an obvious loophole" is another party's "a deal is a deal and you should have checked, first." There are some abusive transactions, all Matt Levine bait, in which a set of lenders initiate some transaction that transfers wealth to them from other lenders (and generally pays the borrower a sort of commission). This kind of behavior is good for returns, at least for the winners. And it's especially good for companies that have streams of income related to capital markets and consulting in addition to asset management: they can make money from opportunistically rejiggering a company's financial structure, then make even more when that rejiggering also requires the company to issue new debt or otherwise adjust its finances. One problem this creates is that it makes some loans default while others pay off; activities that involve a zero-sum but unpredictable change at the economic level are negative-sum at the creditor level, because the creditor only has full exposure to the downside. Lenders are short variance.

The other issue is that it makes diversification a harder sell. If there are risks buried in the fine print of some deals, then a more diversified portfolio is a portfolio with more fine print risk! And if CLOs are diversified buyers buying from more concentrated sellers, then the sellers probably have a relative advantage in assessing these risks.

This risk is another one in the category of things whose scope we'll only know in the aftermath of the next big recession. There's a lot more upside in getting a subset of loans paid off at 100 cents on the dollar when more loans are worth less than that. But the PE industry has one nice limiting factor there, which is that an increasing amount of private equity assets under management are held by large, publicly-traded companies, which are more sensitive to reputational risk because they see it reflected in their stock price. Most of the people who do these deals don't think they're doing anything particularly wrong, but a real-time barometer of their popularity means that they have to focus on behaviors that don't even look like they're wrong.

That doesn't fully mitigate the zero-sum refinancing risk in CLOs, but it does mitigate it somewhat. And while there have been some interesting transactions that shuffled money from one set of borrowers to another, it has, in the aggregate, just not been a big driver of overall defaults.

CLOs look eerily familiar because they follow some of the steps that got us to 2008. But those steps were largely errors of degree, not kind. The thing that turned 2008 into such a nightmare wasn't the initial round of defaults, but the seize-up of liquidity that led companies to suddenly collapse. CLOs firewall some of that liquidity risk; they're already illiquid enough that few investors, especially systemically important ones, are building models that assume they can quickly exit their portfolios when things go south. The risk-bearing tranches are much bigger than they were for subprime-backed CDOs, and they're being held by entities that know they're taking risks, and have a good sense of what risks they're taking. The market’s even getting more transparent: there’s a new, tiny exchange-traded fund that gets investors exposure to the non-investment grade tranches of CLOs (it’s not that meaningful because of the size, but notably the borrow rate on this ETF is currently 8%—somebody out there wants to short CLOs but doesn’t have access to the credit derivatives market). So the leveraged loan market ends up being one of those walls-of-worry that markets have to climb: yes, there's leverage, and yes, there's complexity, but the degree of both is far smaller than it was the last time things really went south.

Thanks to Christian Champ for some very helpful discussions of how the industry works.

Share

Diff Jobs

• A high-growth startup in the food space is looking for entrepreneurial people who can help launch businesses from scratch in new geographies. If you've ever wished you'd been at early Airbnb or Uber/Lyft, this is worth a look.

• A social commerce company is looking for a senior data engineer with strong Python skills. (US, remote)

• A growing startup in a strongly countercyclical sector is looking for both experienced full-stack developers (Ruby and AWS are important), and junior engineers. (US, remote)

• A fintech startup is looking for a senior hire with capital markets experience—someone who is used to finding ways to fund debt or other capital-intensive business models. (Bay Area)

• A company building a marketplace for PE secondaries is looking to add a 3rd cofounder to lead research and deal sourcing. (US, remote)

• A company making the VC fund closing process easier is looking for an early product manager. (NYC or Philadelphia)

• A fintech company helping retail investors create systematic strategies is looking for someone (with a flexible title depending on the candidate) but experience with order execution and other operational processes for an investech product. (US, remote)

• And we have many other open roles in the Diff network, on the engineering, product, and business ops side.

Find a Role With Diff Jobs

Hire With Diff Jobs

Elsewhere

Netflix's Explore/Exploit Strategy

Vanity Fair looks at the sudden reversal of both Netflix's status in Hollywood and its strategy with producers:

One Netflix showrunner, whom I’ll call Showrunner X, remembers the streamer’s early pitch being a golden dream: “We’ll leave you alone, and we’ll order your show straight to series. You don’t have to go through all the bullshit of the pilot process and take a gazillion network notes. We’ll give your show a lot of support, we’ll give it the time and space to find its audience, and we’ll give you almost total creative freedom.” Recently, though, that dream has been unraveling for some showrunners. “Well, we actually are going to interfere with it,” Showrunner X continues, still channeling a Netflix exec. “We actually are going to start canceling stuff pretty expeditiously, and we actually are going to give you a gazillion notes based on our data. And we actually are going to appeal to the lowest common denominator audience. But we’re also not going to pay you as much as network television.”

One way to read this is that initially, Netflix was trying to get as large a sample size as possible. That's a naturally unconstrained business: you want to try as much weird stuff as possible, whether it's pursuing unusual concepts ("A show rife with 80s pop culture nostalgia made by twin brothers who grew up in the 90s?") or paying unusually large sums to attract name brands. Netflix needed to measure how different variables impacted signups and viewership, and didn't necessarily know what those variables were. But once the game moves to retention instead of user acquisition, they move towards the lowest common denominator. That metaphorically recapitulates how people used the original DVD product: adding foreign dramas and highbrow documentaries to their queue, but ultimately watching lighter fare. The fluff is less informative than the risky, high-quality stuff, but the economic point of taking those risks is to learn what works and then to keep on doing it.

For an earlier Diff look at Netflix, see this collaboration with Evan Armstrong of Every ($).

EVs and Platforms

GM has been relatively slow to roll out electric vehicles, but part of the reason for that is that they're building a more modular platform that can cut their costs for a variety of new vehicles. Larger companies face more risk from industry pivots, because they have more business to cannibalize and more fixed costs to worry about. But they have the advantage that they can move slowly and thoughtfully; if GM really is going to stop producing gasoline-powered cars by 2035, they need to map out exactly how to get there. And part of that map means building a scalable way to iterate as they go.

The Diff previously covered EVs, with a nod to GM's modular approach, at the end of Autos Week earlier this month.

Third-Party Sellers

Amazon sellers, particularly the highly levered rollups, are being simultaneously squeezed by slowing demand from consumers, higher costs from Amazon and for shipping, and less access to capital markets. Part of the pitch for the third-party rollup business was that it was a way to arbitrage different valuations: diversified consumer product companies sell at a much higher multiple than single-product/single-platform ones, so building the former by acquiring the latter looks like free money. Ideally, these companies would be throwing off so much cash flow that they'd be able to survive, and perhaps pick up some affordable Amazon retailers in the bargain. But it turns out that one of the big risks in this business was that it can get a lot more capital-intensive: supply chain issues simultaneously mean that they're paying more for inventory, that they need to hold more of it, and that more of their inventory is stuck in transit for longer. Any business can be vulnerable to a liquidity shock, but a business that borrows and diversifies, but can't diversify out of that specific risk, faces even more of it.

Disclosure: I own shares of AMZN

An Entrepreneurial Country

In recent years, China has seen a record pace of new business formation despite public crackdowns on the private sector. This looks like a leading indicator of the country's slow shift to a services-driven economy: service companies can be smaller, or even sole proprietors, but if most of the growth is in exports and heavy industry, then the median firm size will be high and new companies will be fairly rare. Growth in the service sector fits in with some of China's economic priorities, and is a natural side effect of higher median income, but it also means that more economic activity is in places with lower productivity growth. So China is becoming a richer and more economically normal country, but that also means it's downshifting to slower growth.