Note: This is the once-per-week free edition of The Diff. Enjoy!
Hyman Minsky has a powerful descriptive model for leverage-driven bubbles. The process goes like this:
- At the bottom of the credit cycle, assets are cheap, liquidity is scarce, and volatility is high. Anyone willing to extend credit gets paid well to do so, although returns will be bumpy. Average returns are great if you time the bottom of the cycle, but the variance in unlevered returns also goes up: some obscure asset classes get deeply distressed as the market basically shuts down. In 2008-9, that included complex structured products (where there was basically no bid), share class arbitrages (the spreads between economically equivalent classes of stock exploded), and some unusual corners of the market like auction-rate securities. If you bought stocks at the bottom, you probably paid 60 or 70 cents on the dollar, but some credit products traded at 10-20 cents on the dollar.
- As the economy recovers, returns are driven more by price appreciation than by yield. Liquidity returns, investors who were happy to earn a 20% spread between high-yield debt and treasuries in late 2008 were even happier to see that spread drop below 6% in 2010. Liquidity improves, and volatility declines.
- As the market returns to normal, returns start to drop, but volatility drops faster. Now there’s an anchoring problem: the investors who performed the best were the ones who took risk at the right time, but their future returns are likely to be lower than their recent returns. One option is to accept a lower return, but the other is to lever up and get the same return net return by putting more borrowed money to work.
- This rise in leverage keeps asset prices rising, and leads to more liquidity. As long as everyone keeps making the “lever up to produce the same net return” trade, the trade works; there are always incremental buyers, those buyers boost the value of the underlying collateral, and the combination of more general credit availability and the wealth effect of higher prices keeps GDP growth humming. Until there’s some sort of shock—the “Minsky Moment”—and everyone either tries to sell at once or has to sell because of margin calls and redemptions. Sometimes, this produces a brief panic, and sometimes it wipes investors out and leads to cascading sales. In the latter case: go back to step one.
This story is well-understood in capital markets. Minsky himself was somewhat obscure for a while, but started to get popular again just before the peak. (How quaint to see the “quant quake” of August 2007, which is barely visible on a long-term chart, described as “market turmoil… rocking investors around the globe.”) It certainly got a lot of attention in 2008.
The naive story of the financial crisis is that banks made bad loans, and that’s why they lost money. The more sophisticated version is that banks were bad at borrowing; they owned portfolios of complex, hard-to-value securities that were only understood and traded by a few counterparties, and they funded these portfolios by borrowing overnight. The problem with mortgage defaults was not the loss itself but the uncertainty: mortgage delinquencies rose from 1.6% in early 2006 to 6.6% by the end of 2008 and hit 11.5% by 2010, but to someone lending against mortgage-backed securities, a high overall delinquency rate meant uncertainty about the value of the collateral. Some mortgage-backed securities were fine, some were zeroes, but basically all of them went down because funders didn’t know which was which, and they weren’t getting paid enough to find out.
You can think of yield spreads for various asset classes as a sort of minimum wage fund fundamental analysts. If something pays you tens of basis points above the risk-free rate, you simply can’t afford a ton of due diligence; if you’re getting paid tens of percentage points above risk-free, you can’t afford not to do a ton of research.
But Minsky Moments don’t just happen in financial markets. They happen in any scenario with liquidity and maturity mismatches—any time an activity needs a constant supply of short-term inputs to fund something with longer-term, potentially-uncertain outputs.
Consider a car company. It buys parts from various suppliers, assembles cars, and ships them to customers. Those parts suppliers sometimes have to buy their own components, and some of those suppliers do, too, and on and on until you get to the level of metals being dug out of the ground. In the 60s, Toyota realized that they’d get a higher return on investment, and force themselves to run a more efficient process, if they reduced the amount of raw materials and in-production inventory they had on hand, and ensured that their inputs would arrive exactly when they were needed.
It’s not a coincidence that this started in Japan, where banks and regulators coordinate behavior across adjacent firms. GM’s suppliers might have said no to something like this, but Toyota’s suppliers might not have had an option. Minksy again—Japanese industrialization was heavily bank-funded, and a company that couldn’t roll over its loans would quickly go bankrupt, so Japanese industrial companies effectively had their strategy set by their bankers while their executives focused on implementation.
Just-in-time is a good way to conserve capital, meaning it’s a great way for a capital-constrained company to grow. If the banking system is compliant and has sufficient reserves, it’s worth the risk—in the event of a supply chain disruption, banks can extend enough credit to keep the company alive until shipments resume.
It creates systemic vulnerabilities, though, and they’re eerily similar to the 2008 funding story: if part of the supply chain breaks, some suppliers can’t get the intermediate goods they need to produce their outputs. The more complex the supply chain, the more uncertainty there is in both directions. If some car component is unavailable, should parts manufacturers still sell to GM? If GM might need to cut orders, should that manufacturer’s suppliers sell to them? They’d presumably still do it, just on harsher terms—so uncertainty ripples through the supply chain in the form of smaller orders and more aggressive terms for paying receivables.
Apple is hyper-sensitive to this, both because they have a ludicrously complex, China-centric supply chain and because they run a very tight ship (their receivables turnover in the last fiscal year was 11.3x, and inventory turnover was 63x(!)). But if Apple has a shortage of phones and spare parts today, how many companies with a larger inventory buffer will discover that they’re facing a shortage next week, next month, or next quarter?
Reducing working capital needs is, in effect, a way to raise a company’s leverage. Less capital at work for the same output, or the same capital at work for more output. But when traditional credit is so cheap, the just-in-time form might be relatively overpriced. Despite the explosion in corporate debt this economic cycle, retrospectives might say that companies should have issued a few more bonds, and kept a little more inventory on hand.
Just-in-time is all well and good, but it comes at the expense of just-in-case.
The Pension Not-Even-a-Bailout
STUMP, the single best source I’ve found for pension news, covers a wimpy bailout for multi-employer pensions. The problem: optimistic return assumptions, exacerbated by shrinking union membership, is putting union pension funds at risk, which will exhaust the funds of the Pension Benefit Guaranty Corporation. The proposed solution: lend the pensions money. Note that this doesn’t solve the fact that they’re underfunded, or that they don’t have enough workers paying into the fund to support retirees. It just lets them keep writing checks. By 2027, the PBGC will be insolvent, at which point all pension recipients will face the actual, rather than theoretical, risk that they won’t get paid.
I strongly suspect that there will be a federal pension bailout at some point in the next decade. This would involve two steps: 1) subjecting pensions to much more stringent return assumptions and stricter controls on what they invest in, and 2) printing enough money to make the pensions whole at their current obligation levels. I wrote a while back about point 1, that even today’s pension funding numbers assume unrealistic returns:
Right now, pension funds’ assumed rate of return is about 7.4%, compared to 8.0% in 2002, according to NASRA. Or, to benchmark that to a low-risk long-term investment: in 2002, they were expected to earn about 3% more than 10-year treasuries. Now, they’re expected to earn 4.8% more. However, nobody is explicitly calling for pension funds to take more risk, and nobody (as far as I know) believes that pension fund managers, in the aggregate, are adding 180 basis points more alpha per year than they used to.
Printing money is always scary, because it sounds inflationary. As the response to 2008 shows, it’s not inflationary when it offsets deflation elsewhere. Crucially, a full bailout of pensions is inflation-neutral: people paying into plans act as if the plans are a money-good asset, i.e. they save less than they otherwise would. Pension recipients also act like the pension is money-good. If we made it so, it would not have a direct effect on consumer behavior, except by pushing asset prices up; lower expected returns and less aggressive investment strategies for pensions would, in fact, be modestly deflationary, by raising the required savings rate of people currently paying into pensions.
In a sense, underfunded pensions are an incipient balance sheet recession. Workers paying into pensions are paying 100 cents on the dollar for obligations whose real value is well below that and declining fast, and the sooner the gap is recognized and addressed, the less painful the adjustment will be.
RIP Good Times, Redux?
Sequoia Capital shared a timely, deeply pessimistic message about macro trends with their portfolio companies—in 2008. RIP Good Times was a classic, and it included a call-out to an even earlier memo, six weeks after the Nasdaq peaked in 2000, exhorting portfolio companies to raise money as fast as they could. The new memo is a whole lot less apocalyptic than RIP Good Times, emphasizing “turbulence” and “disruption,” even the nightmare scenario of “a few poor quarters.” If anything, this is more optimistic than consensus. The S&P is down 10.7%, the VIX is at 40, and Sequoia is asking companies to prepare for the possibility of a brief recession.
In other Coronavirus news (is there any news that is not, in some sense, Coronavirus news?): US hotel occupancy was down 1.7% Y/Y in the week ended February 29th, partly offset by higher prices. It’s an interesting read on the gap between corporate caution and leisure traveler indifference: “We saw declines in airport markets like Newark, Chicago, Denver, San Francisco and New York, while markets with a lot of domestic traffic like Orlando, Dallas and Atlanta were actually up for the week.”
KIDS Act, Section 230
Two new stories on regulating consumer Internet companies: the Senate is considering a bill (The “Kids Internet Design and Safety (KIDS) Act”) that would ban autoplaying videos after a video ends and prevent video sites from recommending unboxing videos to children. Meanwhile, the DOJ is working with big tech companies to reduce child exploitation on their platforms. In both cases, the obvious platforms at risk—Facebook and YouTube—are also better-positioned than smaller competitors to comply. As Mark Zuckerberg pointed out in an internal meeting, “Twitter can’t do as good of a job as we can. I mean, they face, qualitatively, the same types of issues. But they can’t put in the investment. Our investment on safety is bigger than the whole revenue of their company.”
Looking at Scale, Backwards
Two research pieces that both ultimately address the same topic: the gap between large and small companies is growing and large tech companies acquire younger companies, often in the same industry. The common factor is that big tech companies have monopolistic economics, and a comparative advantage in predicting the immediate future; this makes them unusually skilled at early acquisitions—while most acquisitions don’t ultimately work out, YouTube, Instagram, and Booking.com-level wins subsidize the entire process. This also explains a detail from the first piece: returns to scale seem higher in the US. That’s a feature of globalization—companies that can service the US domestic market first, can hire American tech workers, and tap into US VC and equity markets are the default winners. (China would be more competitive, but thanks to general cultural hegemony and lax IP enforcement for video, English-language sites are way easier to use than Mandarin sites in countries where neither language is the main language.)