Markets can crash at any time; October is a peculiarly unlucky month, thanks to 1929, 1987, and 2008, but you'll sometimes get New Year, Newly-Poorer You (Japan in 1989, US SaaS starting in early 2022), or a springtime panic (2000 Nasdaq peak, Covid in 2020). The sample size of big market moves is too small to effectively generalize about such things; get enough data to reach statistical significance and your dataset will have crashes mediated by carrier pigeons as well as by HFT algorithms.

Crashes will occasionally have a clear, direct cause; it's very understandable that an investor might question the net present value of the future cash flows of the S&P 500 in March of 2020, for example, though at this point that particular question has been answered. But usually, they happen with either a minor catalyst or none at all, just a general reset after a period of rising valuations, growing leverage, and increasing nervousness.

It's that last piece that contributes to a fun kind of crash: the summer slump. In 1997, 1998, 2007 ("quant quake"), 2011, and 2015, July and August featured extreme market gyrations, some of which had meaningful causes, and some of which, like 2007, seem kind of random. If you look at a chart of a broad market index for that period in 2007, you won't see much excitement, but that hides lots of dispersion—quantitative strategies started losing money all at once, on both legs of the trade—so someone betting on momentum would find that the declining stocks they shorted rallied while the rising ones they bought were tanking, and a systematic value investor would see that low price/book stocks were sinking and high price/book ones were rising. So the market looked placid overall, but was actually going crazy.¹

The mechanism for these summer losses is, at least in part, a matter of tracking down the person who's in a position to make an authoritative decision. That person is just a lot more likely to be out of the office and hard to reach in July and August. Consider 1998: Long-Term Capital Management was losing money on an astonishing variety of relative-value trades—betting on the convergence between off- and on-the-run treasury bonds, betting on yield convergence in sovereign debt, swap spreads, shorting equity volatility, and more. They tried to arrange a bailout, and actually got an offer from Warren Buffett: he was familiar with the general trades they had on, and had seen plenty of levered investment vehicles collapse. So he knew that there was an opportunity for easy money. What he also knew was that his bid would establish a floor for the valuation of the portfolio, and would get shopped around. So he lobbed in an offer with a tight deadline, and then went on a long-planned vacation to Alaska, where he'd be a) busy, and b) hard to reach by phone or fax. That deal did not go through, though LTCM did end up getting bailed out.

Sometimes the big issue is time zones. There are some problems where the optimal approach is a marathon negotiating session, typically kicking off after the market close on Friday with the expectation that a completed deal will be announced before the open on Monday. But many logistical factors can impede that—mismatched time zones make deals across different continents more constrained (and if you're in a conference room in New York trying to hash things out with London or Tokyo, you're presumably working on a bigger-than-average problem). Religious observance also means mismatches.²

There's also a bandwidth problem. There's a reason financial companies went back to the office faster than other ones, or created a sealed, quarantined compound in Palm Beach: in-person communication is high-bandwidth, and the more information you're dealing with at once—when the market's collapsing and you can't figure out why, for example, or when the market isn't collapsing and you really can't figure out why your portfolio is—then it's at a higher premium.

A stylized version of a summer collapse follows the classic model where risk is reduced within one or two standard deviations of the mean, but at the cost of fattening the tails. Suppose you're at a fund, and your managing director has just jetted off to a vacation somewhere nice. You're sitting there, watching your screens, and there's a steady drip of losses from one day to the next. Nothing big enough that you'd flag it as a one-day emergency, but not the right direction. At some point, your issue flips from "this is not a big enough problem to interrupt someone during their vacation" to "this is a big enough problem that the vacation is over, and I definitely should have highlighted it first." And when the boss finds out, either they're going to suggest immediately cutting risk or they're going to hop on the next flight—adding another period of dicey communication and stress and, prior to the rise of in-flight wireless, another few hours where events can happen while the decision-maker isn't available to react to them. As the selling happens, other people's vacations get ruined, liquidations get sloppy, and suddenly the market is diving on little to no news. And how do people react in this case? Just as a rough guide, institutional investors tend to assume that either someone knows something they don't or that they face career risk for holding on to a losing trade when they clearly have incomplete information, so they keep pushing prices in the same direction. Retail investors don't have that much career risk, though the next brokerage statement will sting, and they do tend to buy the dip.

The slow-then-fast nature of a summer slump means that central banks do, typically, have enough time to react. They certainly rode to the rescue in 1998, for example. And the different behavior of retail and institutional investors means that these are typically a within-industry wealth transfer from people who take vacations to people who don't, and outside of that a transfer from institutions to retail investors.

Timing crashes is hard, and timing these is harder still, because the catalyst is either so peripheral as to be irrelevant (investors pulling out of subprime-backed CDOs over the course of early 2007, forcing multi-strategy funds to cut their gross exposure in other strategies as well) or completely out of left field (Russia defaulting on ruble-denominated debt). So, you're reading this because it's summer, not because this particular summer is uniquely risky. Then again, VIX is low, stocks are at record highs, and just a few months ago people were asking if the dollar was finished as a reserve currency. So if you do go on vacation, don't put your phone in do-not-disturb mode when the market's open.

Here at The Diff, we like thinking about market crashes. Some highlights:

• Modern financial history starts in 1998

• Correlations go to 1 in a crisis ($), or: finance is always a form of time travel, reflecting information about future cash flows in present prices. And one instance of this is that when people start losing money, every bad outcome happens at once.

• Finance as a Scale-Invariant Global Computer

• What Would it Look Like if The Pod Shop Model Blew Up?

• The Man Whose Bank of England Broke ($), a look at the Soros pound short from the other side.

Join the discussion!

1  And if we assume that quants are identifying and exploiting behavioral patterns that are common to all other investors, it's even worse than that—everyone else has to be slightly positioned against the quants, so what the median market participant saw was that stocks were doing fine and they were doing a bit better—that stock they'd sold after a run-up was sinking once more, and that company they'd bought into at a nosebleed valuation was still rising.

2  Hotel companies will often call out the impact of religious observances that shift based on the Western calendar, not because the total share of their customers who observe is necessarily high, but because if you need a meeting of N people to happen, the odds that one of the N will be unavailable due to the High Holy Days or Ramadan means that those N room-nights will all be shifted either earlier or later.