Financial Bubbles are the Gnostic Heresy: The Voegelin-Minsky Synthesis

“Don’t immanentize the eschaton!” I felt a tinge of envy every time I heard this catchphrase. It’s a double-barreled term — you’ve got your fussy Latinate “Immanetize” and your stately Greek “eskhatos,” and you get to dismiss something as both woolly utopianism and a threat to civilization as

“Don’t immanentize the eschaton!”

I felt a tinge of envy every time I heard this catchphrase. It’s a double-barreled term — you’ve got your fussy Latinate “Immanetize” and your stately Greek “eskhatos,” and you get to dismiss something as both woolly utopianism and a threat to civilization as we know it. So I finally Googled it, and that’s when I started reading Voegelin.

Voegelin’s historical model looks at the interplay between society and representation — who we are, how we see ourselves, and how these forces shape each other.[1] Voegelin argues that modernism stems from the gnostic heresy: the idea that there’s some special, transcendent knowledge accessible to a chosen few, and that possessing this knowledge gives someone the right to impose their idealized utopia on the rest of us. If you’ve ever been told you’re on the wrong side of history, you might be talking to a gnostic.

The actual gnostics were a heretical early sect of Christianity, but Voegelin demonstrates that the mindset animated the Puritans, and thus played an important role in American progressivism. Harvard has been training people to be holier than thou for centuries, with only a few details of the nature of holiness changing. There have always been utopian idealists, but in a Western context they’ve ended up borrowing, either consciously or unconsciously, from Christianity. It’s very easy to see the role that sin and heresy, for example, play in modern political discourse, and how important it is not to upset the clerisy in their wealthy monastery on the Charles River.

Voegelin was writing in the context of the mid-twentieth century, when the old European order, an order that represented itself as having some sort of divine legitimacy, had been extinguished by three kinds of popular legitimacy: populist nationalism (most notably Nazism, which relied on race as its source of legitimacy), communism (legitimacy through the will of the proletariat, as interpreted by a dictator), and liberal democracy (representation through the will of the people). Given the usual recruitment process for historians (step one: be on the winning side), his audience would have viewed this as a victory of good over evil. Voegelin wants us to see it as the lesser of two evils successfully trouncing the third. Yes, there are worse outcomes, but it’s not like the result is good.

He was also writing in reaction to scientism and the death of the transcendent. Since the enlightenment, reason has acted as a universal solvent. Apply it to engineering, and you get the steam engine. Apply it to economics, and you get free trade and long-term growth. Apply it to biology, and you discover evolution. Apply it to politics, though, and you get — the Holocaust, Holodomor, the Great Chinese Famine, and other historic horrors. If you have any sense of prudent paranoia, you’ll look at these phenomena and try to figure out what they have in common. You can comfort yourself by pointing to advances in technology — maybe we just applied Taylorism to industrialized genocide! But you don’t want to go all in on one bet like that when the consequence of error is annihilation.

The Voegelin view is, basically, that there is a religion-shaped hole in the legitimacy modern political movements, including fascism, communism, and even the liberal democratic order. Political science is desperately trying to fill it without making a pilgrimage to Rome, the hajj to Mecca, or making an offering to Wotan. Attempting to bootstrap a new source of legitimacy, Voegelin argues, has been tried. And, he says, it’s always failed.

I recommend Voegelin unreservedly as a refreshing view: a look at the twentieth century from the point of view of the side that lost World War Two, and didn’t even make the semifinals. (Voegelin himself had to flee Vienna after the anschluss.) Human history always bounces between approaching utopia and approaching apocalypse, and Voegelin’s key claim is that we’re like inexperienced pilots in a bad storm with a broken altimeter: we feel the change in altitude just fine, but drastically misread the direction.

There’s never been a Voegelin Vogue, but now is a great time for him to be back in fashion. Sometimes I’ll read something he wrote and feel briefly disoriented. Did he write this in 2019? Because it’s on point. But no, he’d identified something fundamental and important by the 1950s, and when you read him you’ll see it, too.

One thinker I don’t need to introduce is Hyman Minsky. Minsky’s views got quite famous in 2008, because he built a very simple model of how bubbles and crashes happen:

  1. Investors find a good idea.
  2. They bid it up beyond all reason, and borrow money to do so.
  3. The steady inflow of borrowed money doesn’t just push up asset prices; it also reduces volatility, which encourages more leverage and thus higher prices.
  4. Eventually, reality sets in and prices crash.

Minsky is a fun guy to cite because he’s always relevant: if asset prices are crashing, you can say it’s a Minksy Moment and sound quite sophisticated. If they’re not crashing, you can say we’re setting up for a Minsky Moment, and you’re golden: if people agree with you, you sound smart, but if they disagree, that’s exactly what the theory would predict.

Minksy turns out to be a special case of Voegelin. The good idea is the gnosis, the special knowledge available only to a select few. The bidding process is immanentization. And the eschaton is the hypothetical future where the market has perfectly discounted the bubble speculator’s wildest dreams, transferring vast wealth and power to the traders who got it right.

Bubbles as a Search for Representation

It seems grubbily capitalistic to take Voegelin’s deeply erudite reading of history and apply it to something as prosaic as investing. And it is! But modern economics has benefited from the prevalence of natural experiments (any time an intervention is applied somewhat randomly — the military draft, college wait-listing, lotteries for charter schools). And the market is a natural experiment in measuring opinion and collective opinions-about-opinions, which is what Voegelin is talking about.

Is it wrong to use great philosophy as a framework for understanding markets? I don’t know. Is it wrong to notice that the model for Venus de Milo was pretty hot? Is it wrong to leave Eleven Madison feeling full? Is it a sin to use elegant mathematical tricks to make the blood spatter in a video game more realistic? There isn’t an obvious answer. Sometimes art serves a practical purpose. It would be suspicious for there to be art that nobody enjoyed and that nobody could apply.

Markets are the single best way laboratory for learning about human emotions. They won’t tell you much about interpersonal stuff.[2] But for mass phenomena, there’s no better way to see them in action. A market is a machine for synthesizing individual opinions into aggregate beliefs, then reflecting those beliefs back to the participants. They’ve got it all: hope, fear, the hope-fear cocktail of FOMO; nationalism and parochialism, exoticising the Other; overconfidence, underconfidence; laziness, hyperactivity.

There is a looking-for-keys-under-the-lamppost problem here; what if there’s a social phenomenon better-expressed outside of financial markets, and I’m just using markets because people keep assiduous records? I don’t see that as likely: markets are not so much the lamppost as they are a powerful searchlight. Ultimately, any disagreements about values turn into disagreements about value, and markets are an ongoing attempt to resolve such disagreements by transferring wealth from people who are wrong to people who are right.

Every investor is torn between hedging against the world as they fear it is and subsidizing the world as they’d like it to be. But since investing is about setting aside capital now and spending more of it later, it’s fundamentally optimistic, so wishful thinking dominates fear. A market rally is driven by the same instinct as a pep rally, or the Nuremberg Rally: at last, we all know our team — the Good Guys — will be victorious.

So we can use markets as a test case for Voegelin’s theory: what happens when investors become hyper-confident in their vision of the future? They can’t impose it by force (unless you really stretch the definition of “force” to include the way the IMF operates), but they can act as if they’re right and move asset prices in the direction they choose.

As it turns out, the gnostic framework is a fantastic way to understand bubbles, and bubbles are a beautiful demonstration of the gnostic process. The life cycle of a bubble is that investors see asset performance and try to create an analytical framework that explains it, then they respond to that framework accordingly. Over time, more assets are produced that fit the framework, and the assets that don’t tend to drop out of the market. So, over a period of time, the real world responds to the bubble by producing the reductio ad absurdum of the bubble’s justification. Just as a gnostic political system ultimately wobbles because it’s not tethered to any transcendent values, a gnostic bubble topples over when its simplified model of reality collides with the real world in all its multidimensional, fat-tailed glory.

In the 90s:

Investors, 1995: The Internet is a big deal; Netscape’s IPO pop told me so, and my broker agrees. Tech companies will take over every industry.

Entrepreneurs: Here, have roughly 1,400 tech-related IPOs! That’ll be a bit over a hundred billion dollars.

Investors, 2002: Hey, where’d my hundred billion dollars go?

In the 2000s:

Much more sophisticated investors, 2003: I fed the last hundred years of real estate performance data into a Monte Carlo simulation, and it tells me that assuming the future is exactly like the past, there will never be a nationwide decline in housing prices that lasts more than a year or two, as long as I diversify a bit, buying this overcollateralized mortgage-backed security is basically free money.

Those same investors, 2006: Hm, these bonds are still making money, but spreads have declined. On the other hand, defaults are quite low despite a huge increase in outstanding debt and anemic economic performance. I should lever up this trade a bit more; I can’t afford not to.

Home “owners,” circa 2006: Until my rate resets, living here is cheaper than renting and I’ll be able to sell this house for 50% more by then, anyway. Given the size of my down payment, that’s a roughly infinite return on investment.

Mortgage investors, 2009: These houses are so worthless that the only part of my collateral with any value is the copper pipes…

Mortgage investors, several minutes later: … which have been stolen.

A Taxonomy of Bubbles

A long, long time ago, I claimed that there were two kinds of bubbles: an equity bubble characterized by hope for the future, and a credit bubble defined by trust in the status quo.

That dichotomy doesn’t perfectly apply by asset class: a bubble in a developing market might happen mostly through people borrowing in a low-interest currency like USD or Yen, and lending money out in the developing market’s higher-interest currency. This has fixed-income characteristics, but the return profile is a bit more like equities: the principle way people get upside from a strategy like this is currency appreciation; the actual interest rate is just compensation for the risk of devaluation.

A better split might be qualitative versus quantitative bubbles, which you might also call story vs history bubbles.

You can think of discretionary investing as an effort to find good stories without getting sucked into story bubbles, and quantitative finance as an attempt to learn from history without overfitting to it. Overfitting is particularly dangerous because there’s so much data available today, and because humans can always fit a narrative to their story. Here’s Voegelin, in The New Science of Politics, writing about positivism but actually explaining the dangers of backtest-driven trading strategies, in 1951:

The use of method as the criterion of science abolishes theoretical relevance. As a consequence, all propositions concerning facts will be promoted to the dignity of science, regardless of their relevance, as long as they result from a correct use of method. Since the ocean of facts is infinite, a prodigious expansion of science in the sociological sense becomes possible, giving employment to scientistic technicians and leading to the fantastic accumulation of irrelevant knowledge through huge “research projects” whose most interesting features is the quantifiable expense that has gone into their production.

This gets at one of the fundamental divides in quantitative finance, and finance generally: inductive versus abductive reasoning. Inductive reasoning tells you that if markets underperform on rainy Tuesdays, you should trade based on the weather forecast. Abductive reasoning requires you to start with a theory (rain makes investors gloomy), test it out (do markets underperform every day it rains?), and only act if the facts you accumulate fit into some theoretical body of knowledge. An abductive investor wouldn’t bet on the Tuesdays-only thesis, unless it turned out that the only umbrella salesman at the NYSE always took Tuesdays off.

To use abductive reasoning, you need an underlying theory: why should these opportunities exist in the first place? A purely inductive approach leaves the fundamental question un-asked: Why would trading generate a profit? And that’s an important question to answer, because your theory of markets gives you a sense for the scope of your opportunity, and also tells you whether you’re a beneficial economic actor or a social parasite.

Inductive approaches treat the market as though it’s mostly noise, with meaningless patterns that somehow persist. But it’s hard to reconcile that with the fact that markets are right up there with Bach, Picasso, Hemingway, and Euler in the game of compressing a staggering amount of complexity into something very simple. Whenever something good or bad happens — a scientific discovery, an election, a war, a ceasefire — it’s instantly refracted across portfolios and prices. That’s a beautiful system, in a way: a farmer in Iowa shouldn’t spend a lot of time thinking about what victory against ISIS means for his business (oil down -> ethanol down -> profits down would be my guess); he just has to look at the latest quote. That implies that markets are full of meaningful patterns, and that understanding those patterns requires understanding the meaning behind them.

Abductive reasoning can go too far as well. The abductive approach is to form a theory, and update it with data. When the data doesn’t match the theory, you throw it out.

There are two problems with this, the obvious and the subtle:

The Obvious: All human beings are fiercely resistant to ever changing their minds, even in the face of mountains of blatant evidence that completely refutes their position. Republican readers, think of the Democrats you know; Democrats, think of the Republicans you know. Members of third parties and obscure post-political sects, think of literally everyone you know, especially people who claim to be part of the same belief system as you but are completely wrong about why they should hold their views.

The Subtle: Some economic stories are so good they make themselves true, or at least make themselves seem to be true. Think of a conference, like CES. The business model for a conference is: you lease some space in a conference center, and you convince everyone in the industry that everyone but them has already agreed to show up, then they all bid against each other. CES would be a bubble, except that the agglomeration effect of putting every end of a trillion-dollar supply chain into one big building, then blasting a firehouse of free liquor at them, turns the fib into reality. This also happens with two-sided markets: as Dan Wang notes in his year in review, Moore’s Law was a claim that may or may not have been valid when it was made, but it coordinated people up and down the supply chain (equipment manufacturers and fabs on one end, OEM buyers, software companies, and consumers at the other), and made itself happen. The Apollo Program is another such case: it was only feasible because everyone took it seriously enough to work on it.

But these are the exceptions. In 1961, JFK said we’d land a man on the moon by the end of the 60s, and we did it in the summer of 1969. Nixon committed to curing cancer (not there yet). In 1977, Carter set a goal of reducing US gasoline consumption by 10% by 1985 (we got it down 5%, but by 1987 it was back to where it had been at the time of the speech). When George W. Bush proposed a mission to Mars in 2004, it was seen as a joke. If you’d told someone in 1969 that the President in 2004 was proposing a mission to Mars, they might have wondered why we weren’t there already.

The power of coordination doesn’t get us as far as it used to, but it does provoke some changes. The popularity of the Internet as an investing vehicle did get people to spend time online, and the easy money from venture capitalists gave startups an ad budget they spent on more established sites. So, for a while the Internet looked like an industry where there were lots of viable businesses, some early-stage and still investing, some later-stage and making money. Really, it was an industry full of completely unworkable models, or models that would only function once everybody had a smartphone.

But in 1999, a dollar of venture capital might have led to $0.25 of ad revenue for the big portals, and if the portals were trading at 10x sales, that meant $2.50 in market value — so for a time, the bubble’s logic was self-fulfilling. More money did lead to higher market values, which was a recipe for even more money flowing in.

A key difference in the two types of bubbles is in exactly how they reach their extremes. For equity bubbles, we turn to a discounted cash flow analysis. A typical DCF will calculate the net present value of a few years’ exact cash flows (I’ve seen models with five years and models with fifty), and then a “terminal value” that assumes a steady growth rate for the indefinite future. If you look at a ten-year DCF for a fast-growing company, you’ll typically find that the terminal value is the majority of the total valuation, and is highly sensitive to your assumptions.

Suppose your model assumes that at the date of the terminal valuation, your company is doing $100m a year in free cash flow, your discount rate is 8.5% (this is Damodaran’s estimated average cost of capital for an Internet company), and you assume 5% perpetual growth. Your terminal value is ($100m)/(8.5%-5%), or $2.9bn. Suppose you bump your growth rate up to 6%. Now the terminal value is $4bn. Terminal valuations are incredibly sensitive to long-term growth prospects, and since the long term is the point in time about which we have the least information[3], it’s the most up for debate. And as that terminal growth rate edges up, the valuation necessarily gets more sensitive to it. So the more extreme an equity bubble gets, the more extreme it can get.

Very few investors take discounted cash flow models on faith, but they’re the tool people use to sanity-check their assumptions. Because of simple leverage, the more a bubble inflates, the fewer additional assumptions you have to make to get further inflation.

Historical bubbles have a different dynamic; nobody can build a bond valuation model whose returns asymptotically approach infinity. But they can build a model where a given investment opportunity’s potential to absorb capital. The saying about equities is “No tree grows to the sky.” For debt, the analogous view might be “No tree has roots so deep it’s immune to a sufficiently bad storm.”

Tour de Bubble: Fifty Years of Immanentization

Pure data acquisition is a recipe for spurious correlations, but pure theorizing is, in the words of Peter Woit, Not Even Wrong. We need to see where the rubber meets the road, and take a drive through the last fifty years of bubbles. I’ll skip a few — everything you need to know about the Beanie Baby Bubble is in the book, and as far as I know it didn’t lead to a recession — but this should cover all the major bubbles in domestic markets in the last half-century.

The 60s: Conglomerates and the “Great Garbage Market”

There were two parallel equity bubbles in the late 1960s: conglomerates, and small stocks. In a way, they were part of the same general phenomenon: increasing professionalization of investors and managers, coupled with the end of the Depression hangover.

Postwar Wall Street may have been the single best period in history for stock pickers. Investors had been scared off by the depression and hadn’t come back, but any business that had avoided bankruptcy in the 30s would probably be viable for a long time. So making money was a matter of digging up companies that had not just survived, but thrived. There were two avenues for this: an investor could buy a piece of them, or a company could buy all of them.

The latter strategy was made a bit difficult by aggressive antitrust enforcement: by the 60s, it had become effectively impossible for large companies in the same industry to merge. As a result, some dealmakers came up with a new model: instead of being a steel executive or an oil executive or something like that, they’d be an Executive, period, and apply their superior management skills to everything under the sun.

The canonical example of this is Ling-Temco-Vought. Ling was James Ling, who took his electrical contracting company public, and merged it with, in rapid succession, a speaker company, an aircraft company, a meatpacking company[4], and eventually one of the largest steel companies in the US. Why did this work? Private companies were cheap, public stocks were expensive, so if you sold equity to the public and used it to buy private businesses, you immediately made money. And electronics companies were accorded higher valuations than dowdy industrial companies, so as long as you put “Altec” in the parent company’s name and kept “Wilson Meatpacking” out, you’d keep that high multiple.

But every acquisition juices growth for just one year; the next year, proportionately more of your business is slow-growth industrials rather than rapidly-growing electronics. So the model only worked through continued growth.

An LTV investor could look at a succession of annual reports and see a strong upward trend in earnings, but that got harder to report every year. And LTV had imitators — Rapid-American (initially platemaking, lithographs, and dollhouse furniture; eventually watches, cosmetics, various retailers, whisky, and the Riviera casino), ITT (Levitt houses, Scott’s fertilizer, Wonder Bread, Who’s Who), Litton (defense, shipyards, typewriters, microwaves, Stouffer’s frozen foods), and more.

Eventually, everything cheap had been bought, and everything that could be bought was expensive. As conglomerates’ market values started to reflect this, the model completely fell apart. Most of the major ones would be slowly split up, and LTV, shorn of everything but its steel assets, would be the largest bankruptcy in US history.

A conglomerate bubble wouldn’t have happened if there hadn’t been value in the original idea. There is such a thing as general management talent; supply chains are hard, accounting is counterintuitive, and a smart tax lawyer can save pretty much any business a lot of money. At a time when ambitious people went to work for big companies, small companies tended to be mismanaged, and their market values reflected this. So the initial model made a fair amount of sense. But misleading accounting and blind extrapolation turned what could have been a short-lived trend (American management got bad in the 40s and 50s, and then improved) turned into a disaster for investors. Acquirer behavior got ahead of the underlying logic of acquisition, shareholders rewarded companies that stuck to a no-longer-viable playbook, and ultimately investors paid the price for immanentizing the scientific-management eschaton.

As small companies got vacuumed up into giant conglomerates, the surviving public companies in the late 60s started to get attention. They didn’t appeal to even the voracious LTVs and ITTs, but they did appeal to other investors. Richard Jenrette called 1969 “The great garbage market”:

[A] market in which the “leaders” were neither old blue chips like General Motors and American Telephone nor newer solid stars like Polaroid and Xerox, but stocks with names like Four Seasons Nursing Centers, Kentucky Fried Chicken, United Convalescent Homes, and Applied Logic.

When is “garbage” popular? Statistically: it’s always too popular. In Expected Returns, Antti Ilmanen notes an interesting kink in the risk-reward payoff graph: in any category, the most risky asset has the worst payoff, and the least-risky the best. So among investment-grade bonds, the riskiest do relatively poorly, while the least-risky bonds that are below investment-grade outperform. In stocks, there’s a general tendency for the biggest companies to do a little worse than the smallest (albeit with less volatility), but the very smallest cohort performs abysmally. Even in government bonds, investors get paid for taking duration risk (i.e. your expected return is higher for buying five-year bonds compared to the two-year), but the thirty-year doesn’t pay for its risk nearly as well.

Investors like lottery tickets. If you tell an investor that the investable universe is anything that’s less than a 5 on a 1–10 scale of riskiness, expect his assets to cluster at around the 4.9999 mark. People like to take risks, and they’re confident. But if everybody’s confident, then in the aggregate they’re overconfident. Thus, garbage.

But when does garbage do especially well? It’s after everything else has already gone up.

If you look at the “solid stars” Brooks mentions, stocks like Xerox and Polaroid, two things jump out. First, these were the great growth winners of the 1960s: the companies the new generation bought when the Long Generation finally retired. But second, these companies have a razor-and-blades model: Xerox wants you to own a copier so you’ll buy paper and ink; Polaroid needs you to keep buying film. So these are businesses that show high margins after their growth has peaked. (Exactly what investors look for in enterprise software today.)

Applied Logic didn’t fit that mold. They were a time-share computing company: they bought big DEC mainframes and sold time on them. Basically cloud computing, back before desktops. This business looked like a tech business (computers!), but the reason Applied Logic was hot was that it was a smaller version of Leasco, and Leasco was definitely not a technology company: Leasco also bought computers (IBM mainframes in their case), and leased them to companies. It was, in fact, entirely a financial engineering play: Leasco booked a tax credit when they bought computers, and they depreciated them more slowly than IBM did, so Leasco’s business had better reported earnings (but, presumably, worse cash flow) than IBM’s. As a result of this aggressive accounting, and some smart M&A, their stock went up 55x from 1963 to 1968.

Leasco would eventually try, and fail, to take over Chemical Bank (I told you: they were a finance company with a tech cost of capital!), and Applied Logic went bankrupt in the 70s.

I don’t know why nursing homes were a hot sector, but the fact that two different nursing home companies were on Brooks’ list is indicative: there wasn’t anything company-specific so much as there was a broader trend that investors were cottoning on to.

And KFC! That one’s easy: the interstate highway system was one of the great platform plays because, like the Internet, it was government-owned but available to private enterprise, so all the economic surplus accrued to businesses and consumers. If a KFC works in one city, it’ll work elsewhere, so it’s easy to sketch out a long-term growth plan. There were more established growth companies that had use the same model (the 100th Golden Arch went up in 1959, and the 1,000th in 1968, for a compound growth rate of 29%), so, once again, investors could apply the same framework — could immanentize the same fast food eschaton — by bidding up shares of KFC.

What investors didn’t realize about the fast-food bubble — the real-world problem that interfered with their pure gnosis about the eternal growth path of the fast-food business — was that it depended on cheap oil. US oil production rose throughout the 60s.

Then it peaked. Then the embargo hit, and suddenly an extra road trip was a big deal.

The 70s: The Nonfinancial-Assets Bubble

The 1970s were a terrible decade for savers of all stripes. The S&P was at 85 in January of 1970 and at 114 in January of 1980, but consumer prices more than doubled. So, overall, investors got all the risk they’d come to associate with the stock market, but none of the reward. With dividends, real returns were roughly 2% per year. Fixed-income investors did even worse, and faced two years of returns-destroying inflation, coupled with high rates, in the early 1980s. Really, the best form of asset allocation was to be a hoarder: own hard assets that appreciate with inflation, borrow money you can pay down with depreciating dollars, and just try to survive.

In theory, a business should be able to adjust for inflation: an increase in the CPI, after all, is an increase in revenue. In practice, cost structures are sticky, and price changes are unequal: as it turned out, many companies’ models were predicated on low, stable oil prices, and oil got quite expensive.

Corporate concentration, economic growth, and inflation in the 1960s left a structurally difficult legacy in the 70s: workers at heavily-unionized companies could push for higher wages to counteract inflation, and they generally got it. And companies responded to higher inflation by raising prices.

What took this from a one-time effect to an accelerating cycle of wage and price inflation was the second-order impact. Reification struck again: consumers and companies saw that inflation was high and rising, so they realized that the exchange rate between goods and money would be getting less favorable. Result: spending money to stockpile goods.

I read a personal finance book from the era that suggested, among other things, buying toilet paper in bulk as an inflation hedge. If you’ve ever wondered why inflation is self-reinforcing in the short term, now you know. (At least the author of that book was responsible enough not to suggest that dollar bills would be a suitable TP substitute in the near future.)

That’s a process that can’t continue forever, but it’s an inherently destabilizing one. Every round of hoarding produces an upward inflation shock: hoarders get their views validated, savers panic as their savings decline in value, and the recipients of hoarded dollars have an incentive to spend them quickly, too, both to build up their inventory back to normal levels and to stockpile to ward off higher inflation.

Normally in a bubble you expect people to make money, but the same self-reinforcing bubble dynamic can happen in situations where people are trying to lose money as well. In this case, the gnostic knowledge was that the dollar was on track to permanently lose value, at a faster and faster pace; consumers and companies immanentized this regression from currency to barter by stockpiling.

If inflation was a bubble, it was one of the few bubbles successfully popped by the Fed; Volcker raised rates aggressively, and signaled his willingness to let the country go into recession rather than let the CPI rise. To someone betting on a higher CPI, especially betting by way of leveraged ownership of hard assets, this was painful, but ultimately inflation ticked down from the low teens in 1980 to a 2–4% range by 1983.

The 80s: Drilling for Oil on the Floor of the NYSE

A theme in these bubble cycles is the accumulation of some kind of potential energy, followed by its sudden and violent release. Stock values in the 40s were a coiled spring, that uncoiled and sprung too far in the 60s and 70s; by the early 1980s, heavy accumulation of assets and painful financial market performance had created a new kind of undervaluation: companies that owned too much and were worth too little. The economic disruptions of the 80s also pummeled the unions; high interest rates are good for the dollar — the US dollar index (comparing against a basket of currencies like the Deutsche Mark, the Yen, the and the Pound) rose from 89 in early 1980 to a record 164 in early 1985. By weakening exporters, particularly exporters whose cost structure was fixed by union contracts, this eventually shook up labor markets and weakened the coupling between consumer prices and wages.

This was not necessarily a good trade, especially if you like having a strong manufacturing sector, but at least it slaughtered inflation. And as a second-order effect, policies that are good for the dollar and bad for exports helped make the US a better destination for risk-averse international capital: knowing that the Fed is on the side of UST holders rather than factory workers is exactly what a German pension fund or a Middle Eastern sovereign wealth fund manager needs to hear.

This shouldn’t be read as a conscious decision to de-industrialize; it was a conscious decision to solve one problem at a time. Over the long term, an inflexible manufacturing sector might be even less competitive than a manufacturing sector facing currency headwinds, especially if the cost of capital is elevated due to high and volatile inflation, so in a sense Volcker’s choice was between putting a bullet in the least competitive American factories or just letting them slowly bleed out.

There were two related forces that made the 80s a great decade for investors: first, the demise of some of the forces that made the 70s bad was a direct benefit to asset prices. If rates were high because inflation expectations were high, then a reduction in inflation expectations, driven by both Fed policy and an easing of the oil supply situation, is an immediate benefit to bondholders. Second, the companies that had hoarded assets while financial markets declined now had low market values coupled with high asset values. Corporate America was worth a lot to a private buyer, but the investing public didn’t accord it the same valuation. If price levels double and then inflation ratchets down, a $10m building still has a $20m replacement value, and if its stock price doesn’t reflect that, someone will take notice.

The infrastructure for dismantling undervalued conglomerates was actually put in place during the 70s, when high-yield bonds started to attract more investor interest. When Michael Milken joined Drexel in 1969, the bonds he was trading were “fallen angels”: companies that had previously been investment-grade, and had fallen on hard times. But Milken and company realized that there was room for new issues in the high-yield space; a bond backed by a growth company may not be a safe credit, but at a sufficiently high interest rate it performs more like a stock. While companies with poor credit ratings are vulnerable to a recession, their bonds are less vulnerable to higher interest rates; the high rate on the bond shortens its duration compared to a low-risk asset. So high-yield bonds made a certain amount of sense for a fixed-income investor or an equity investor, and Milken, who was an excellent analyst as well as a good salesman, was able to intelligently market bonds to both.

Once he had a network of bond buyers, Milken (and a few imitators, who didn’t do nearly so well) could provide capital to acquirers who didn’t merit an investment-grade rating. As long as the company they were acquiring had assets that could be quickly liquidated, a superficially risky bet was actually quite cheap. T. Boone Pickens’ Mesa Petroleum might have been a dicey credit, but when it levered up to borrow the much larger Cities Service, it was bidding on collateral it could use to pay down the debt. (Cities Service ended up selling to somebody else, but Mesa had purchased stock as part of their bid, so they came out ahead.)

If an investor looked at the history of high-yield bonds and leveraged buyouts in the mid-1980s, here’s what they saw: they got paid to take risk, but the risks largely didn’t materialize. Acquiring companies, selling down assets, and paying off debt was an established strategy, and it worked because companies’ share prices exceeded the replacement value of their assets.

As Larry the Liquidator says in Other People’s Money:

You know, at one time, there must’ve been dozens of companies making buggy whips. And I’ll bet the last company around was the one that made the best god-damn buggy whip you ever saw. Now how would you have liked to have been a stockholder in that company? You invested in a business and this business is dead. Let’s have the intelligence, let’s have the decency to sign the death certificate, collect the insurance, and invest in something with a future! ‘Ah, but we can’t,’ goes the prayer. ‘We can’t because we have responsibility, a responsibility to our employees, to our community. What will happen to them?’ I got two words for that — ‘Who cares?’ Care about them? Why? They didn’t care about you. They sucked you dry. You have no responsibility to them. For the last ten years, this company bled your money. Did this community ever say, ‘We know times are tough. We’ll lower taxes, reduce water and sewer.’ Check it out: You’re paying twice what you did ten years ago. And our devoted employees, who have taken no increases for the past three years, are still making twice what they made ten years ago. And our stock — one-sixth of what it was ten years ago. ‘Who cares?’ I’ll tell ya — Me.

I’m not your best friend. I’m your only friend. I don’t make anything? I’m makin’ you money. And lest we forget, that’s the only reason any of you became stockholders in the first place. You wanna make money! You don’t care if they manufacture wire and cable, fried chicken, or grow tangerines! You wanna make money! I’m the only friend you’ve got. I’m makin’ you money. Take the money. Invest it somewhere else. Maybe, maybe you’ll get lucky and it’ll be used productively. And if it is, you’ll create new jobs and provide a service for the economy and, God forbid, even make a few bucks for yourselves. And if anybody asks, tell ’em ya gave at the plant. And by the way, it pleases me that I’m called ‘Larry the Liquidator.’ You know why, fellow stockholders? Because at my funeral, you’ll leave with a smile on your face and a few bucks in your pocket. Now that’s a funeral worth having!

Larry is a caricature, but a good one: he’s boiled a business down to the balance sheet, he’s run the numbers, and he knows it’s time to pull the plug[5]. Larry’s not a bad guy for trying to liquidate; he’s just the bearer of bad news. In the early 1980s, the bad news was that US manufacturing wasn’t as competitive as it had been, and we needed to scale back. And a great deal of money was made in that scaling-back process.

But eventually we ran out of failing manufacturers, underpriced oil plays, and overfunded pensions. And yet that was the time when the private equity companies and corporate raiders could point to a long record of profitability. The question was: can someone who made his fortune buying at 50 cents on the dollar and selling at 80 cents on the dollar make another fortune despite paying full price? The logic of business said no, but the logic of trailing five- and ten-year returns said “Absolutely.”

This, of course, is when investors really piled in. Their special knowledge: publicly-traded America was poorly-managed; by paying better managers, we can increase the value of their assets, and by borrowing against those assets, we can ramp up our returns. An investor who believed this in 1980 would have been lonely but right; an investor who said it in 1987 would have been comfortably within the consensus, and dead wrong. The 80s eschaton was financialization: any stream of cash flows that could be represented by a blue-chip stock with some AAA-rated debt could also be represented by a tiny sliver of high-return equity and a mountain of junk bonds.

As it turns out, the fundamental thesis was generally correct. There is nothing special about the line between BB and BBB credit; the risk-reward is actually better for BB because of the lottery-ticket dynamic that makes the dodgiest of investment-grade credits a more fun gamble for bond managers. And the private equity model does work for many companies: public ownership has benefits, but it has costs as well; the public markets are good at punishing companies for a bad quarter, and very bad at tolerating necessary restructuring. But when you’ve run out of necessary restructuring and have “restructure this!” in your job description, momentum and compensation will tend to drive you to take worse and worse risks. The late-80s private equity deals were notoriously painful for buyers and the borrowers who lent to them; by the 90s, the market demanded a different story.

The 90s: Globalization and Technology

It got two: technology, and globalization. The technology story mostly played out in the public equity markets, so it’s a better-known part of the 90s equity boom, but globalization played a major role as well. When the globalization thesis blew up, in 1997 and 1998, it affected assets most investors didn’t trade (foreign sovereign debt, equity volatility, various credit derivatives), and led to rate cuts that bailed out equity markets, which they did notice. So to a public equity investor in 2000, 1997-1998 looked like a blip; to a bond trader, it was the (temporary) and of an era.

The purest globalization story in the 90s was Coca-Cola.

Coca-Cola had made money from a simple set of bets: people like carbonated sugary water, and they want to trust brand names before they put something in their mouth. Coasting on their exquisitely tuned knowledge of taste buds and advertising, Coca-Cola could have contentedly ground out low single-digit increases in cases shipped in the US indefinitely. But there’s a big world out there, and a lot of it’s hotter than here. A Coca-Cola executive once said:

“When I think of Indonesia — a country on the equator with 180 million people, a median age of 18, and a Muslim ban on alcohol — I feel I know what heaven looks like.”

He was right. Their small BMI is Coca-Cola’s big TAM. The only question was valuation. Coca-Cola was a solid growth company, and a metonym for great stock performance. (In their 1999 horniness anthem The Bad Touch, The Bloodhound Gang — not a band known for their frequent references to LIBOR or dividend payout ratios — name-checks Coca-Cola’s stock price performance.) At year-end 1997, Coca-Cola’s market cap was just under $150bn, earnings were a hair over $6bn, and they’d grown topline by 1% in the last year and 3% the year before. I don’t know what hell looks like, and we’ve established that heaven is Indonesia, which must mean that waiting for a single-digit grower to grow into justifying an above-market P/E ratio is purgatory. (Coca-Cola was trading below their early 1998 valuation ten years later.)

LTCM was another globalization bet. In their case, a bet on general convergence across global capital markets. A number of LTCM’s strategies made this bet directly: betting on Italian bonds, shorting treasuries. Some made it indirectly: LTCM apparently made a lot of money applying standard options-pricing models to the Japanese warrants market. And often, LTCM’s bet was just taking two similar instruments, one of which was less liquid than the other, and betting that the less liquid one would rise relative to the more liquid one.

As a general thesis this sounds… very 90s! If you had asked someone in Greenwich what the world would look like in the year 2028, they would have at some point theorized that there would be hedge funds just like theirs in Jakarta, Moscow, Johannesburg, and Istanbul. That these hedge funds would be staffed by smart and diligent people who generally took the right side of every bet. That there wouldn’t be systematic distortions in asset prices. In other words, that the world would get its act together. The world had been doing so for a long time: the US system of representative democracy and free-market capitalism had been tested by Fascism in the 20s, the Cold War, Japanese export prowess in the 80s — and every time, the good guys had won.

And since the US had a model that other countries could copy, you’d expect them to do so, which meant that any bet on America staying pretty American and other countries getting more American was a solid bet.

LTCM was always diversified, across currencies, asset classes, different kinds of derivatives; they hedged, they tracked value-at-risk, they even stress-tested their portfolio (“How would we do in a monster rates cycle like the early 80s? How would we do in a rerun of OPEC?”) As it turns out, the stuff commentators said they should have done was stuff they were already doing.

The problem was that despite lots of different trades and lots of different positions, LTCM was actually a global macro fund that was putting on exactly one trade: the more-like-America trade. They saw the world through a Whig History lense, where the fact that the past is less like the present implies that the telos of history is to become like the present only more so. You can’t stress-test that view, because it’s invisible to you. A kind of historical color-blindness.

The crisis that did LTCM in was Russia’s decision to default on their debt in 1998. LTCM didn’t actually own the debt, but the people who thought like them did. What could be a better globalization-and-homogenization trade than to lend money to the superpower we’d been aiming nukes at a decade earlier? When Russia defaulted, investors backed down from the entire Whig History trade, and LTCM, as a heavily-levered institution, quickly burned through their capital, forcing a bailout.

There were rumblings of doubt even before Russia, though. In 1997, an earlier bull market in globalization and convergence suddenly fell apart: East Asian countries like Indonesia, South Korea, and Thailand had been hot markets for years, with high GDP growth, generally fueled by exports. Investors looked at what had happened to Japan — and export-driven boom that transformed into a real estate bubble, eventually leading to a devastating crash. And, like Louis Creed in Pet Sematary, they thought: Maybe this time it’ll work, if I’m just a little faster.

It didn’t, of course. Different times, different details, same mistakes.

You often see foreign capital flows dominating a bubble, especially near the peak. It certainly happened with the East Asian bubble and crisis in the 90s, and loans to Latin American countries wrecked the US banking system in the 80s. It was even a feature in the 2000s bubble, when German landesbanks and Middle Eastern sovereign wealth funds invested in mortgage-backed securities (for more on this, check out my piece on the macroeconomics of fracking. In The New Science of Politics, Voegelin concludes by noting that the gnostic heresy did the least harm in places where it started the strongest — Puritans gained ground in England and then played a crucial role in America, but the general intellectual movement they were a part of found its fullest expression, in Voegelin’s view, in Nazi Germany and Soviet Russia. Voegelin argues that early gnostic revolutions were weak, and the institutions they fought were strong, so gnosticism got more virulent as it spread.

This is a sound argument; humans did the same thing. The only place where humans coexist with big land animals is Africa, because we coevolved; as we migrated out of Africa, the first thing we did was exterminate every dangerous or delicious animal we could find. (Look at Chavet cave, where ancient French people painted 30,000-year-old French lions!)

History Bubbles often morph into Story Bubbles in a foreign country where the people with capital don’t speak the language. The cycle works like this:

  1. Stocks in, say, Mozambique have a good run.
  2. Somebody who is vaguely familiar with the Mozambican market makes themselves available for interviews, and generates a Theory of Mozambique that explains why they’ll keep growing. Since very Americans speak Portuguese and almost none of them know anybody from Mozambique, there’s no check on this theory. And since the theory is there to explain why stocks went up, it’s a generally flattering theory.
  3. Capital flows in, assets appreciate. Unless the local government is extremely crafty, this will lead to a property boom (money has to find a home somewhere, and real estate is basically an index fund for economic growth). Property booms tend to grow the size of the construction and finance sectors, accelerating GDP growth. The construction sector can hire unskilled workers with a high propensity to spend, and the finance sector gets more willing to lend due to classic Minksy factors: accelerating growth lowers defaults, so they start to relax their standards.
  4. Something goes wrong, the market collapses in local currency terms and the currency collapses too, burning outside investors. The Mozambique expert’s books sell for $0.01 plus shipping on Amazon. He rebrands himself as an expert on cryptocurrencies (if it happened in 2017) or legal marijuana (2018), and soon he is once again booked on TV. The Mozambicans, many of whom lost their savings, are forgotten.

This is not to say that country-specific theories aren’t important. They are! If you don’t understand the incredibly durable English tradition of muddling through what would, in other countries, be a constitutional crisis sparking a civil war, you won’t be able to guess what will happen (or won’t) with Brexit. If you don’t understand the differences between Northern and Southern European work cultures, you won’t really get their productivity statistics. (According to the OECD, people in Greece work 2,018 hours/year, compared to 1,356/year in Germany. According to this guide for expats, German business culture is comparatively all-business. Prompt meetings, fixed agendas, direct communication, and minimal discussion of nonwork issues at work can raise the percentage of time-at-work that’s devoted to work.)

Everyone’s stuck threading the needle. You need a thesis that’s specific enough to be valid, but not so specific that it explains every single deviation between measured results and smooth exponential line screaming upward.

The danger of a vague globalization thesis is best illustrated by the slightly globalized tinge to the tech bull thesis in the 90s. While the Internet boom has been covered to death, by me and by others, it’s interesting to tie it into globalization. To justify a tech company’s long-term valuation, there are three things you can say:

  1. They’ll invent something new. (This is hard.)
  2. They’ll charge more for what they have. (They will, but this gets harder and harder.)
  3. This product, which is popular in America, will eventually be equally popular in the other 95% of the world. (Now we’re cooking with gas!)

To someone perched at the end of history in the late 1990s, this was actually the cautious point of view. Instead of arguing that the richest, most advanced country in the world would go through a period of hard-to-predict socioeconomic change, you could just argue that the change we’d already gone through would be recapitulated, much faster, everywhere else. Investors didn’t want to think about empty malls, unemployed travel agents, and the dismembering of the media monoculture. Easier to imagine the sound of a dial-up modem being heard in every time zone and at every latitude.

As we all know, the 90s ended on September 11th, 2001: a deeply symbolic moment when people with very un-globalist, very un-90s opinions murdered thousands of Americans. That was the day when you had to ask whether American culture is really taking over the world, or whether we’ve just found a way to get billions of people annoyed at how pushy we are about imposing pop music, blue jeans, elections, and free trade on parts of the world that don’t particularly care for such things.

Before 9/11, you could be skeptical that globalization should happen, skeptical that the benefits were as big as were being touted or that the costs were as low as everybody said. But afterwards, you had to ask: was it really happening? International trade has been rising over time, but it’s peaked and receded before. In 1913, for example. With that analogy in mind, you can look at globalization, not as a triumph of progress, but as a peak of hubris: the time when countries that have done well overestimate their capacity to share their ways with the world, and the world’s willingness to accept them. What if a global communications system creates a robust marketplace for ideas, and the ideas with the most virulence aren’t even compatible with the system that made them popular? (As Voegelin notes, every authoritarian ideology — communists, Nazis, Puritans — took advantage of a free press to get into power, and then immediately suppressed all discussion of their own philosophical foundations. A corollary to this: if you’re allowed to debate something in public, it’s probably not important.[6])

Growing up, I fully believed in the narrative of progress, in the theory that the rest of the world would keep getting more like us, and that this was a good thing, but if I’m honest that belief should have collapsed along with the towers.

Real Estate: The Mean-Reversion Bubble

I’ve written about the mortgage bubble at great length, both from the demand for capital from US homeowners side and, indirectly, the supply of capital from energy exporters side. But let’s talk about the social element.

Suppose you’re a reasonably smart mortgage-backed securities trader in 2004, and you realize the mortgage market is in a reflexive bubble: the thing that determines housing prices is mortgage availability, and the thing underpinning mortgage availability is the notion that housing prices never fall. It’s a tautology, so you should bet against it, right?

Maybe not. A tautology, after all, is always true.

If you’re paid an annual bonus, and your bonus is based on the performance of your mortgage-backed securities portfolio, you have two options:

  1. Be cautious. Only buy highly-rated securities. Maybe even bet against the garbage.
  2. Let the good times roll. Collect the spread between your cost of funds and the returns on subprime mortgages.

In the event of a good year for mortgage-backed securities, choice number one gets you fired, and choice number two gets you a bonus. (You probably won’t use it to buy a house.) In the event of a bad year, choice number two gets you fired, but choice number one probably does, too; you’re not the only mortgage trader at your firm, and a blowup will be big, so even if you happen to make money, that just takes the bonus pool from zero to… zero.

Every year, some people made some variant on choice number one, and some made some variant on choice number two. And every year, until 2007, choice number two was the winning option. So every year, the optimists got smugger, while the amount of capital pouring into their trades went up — which, of course, meant that their mark-to-market profits were higher. Within the industry, 2006 was by definition a year of peak confidence; the market can’t keep going up if people start getting doubtful.

There were a few clever structured bets a mortgage trader could have made.[7] But because they were clever, few traders made them. And cleverness still costs you reputational capital; even if you achieve the same return profile as your competitors, and you know you’re taking less risk, you’re taking the career risk that your results will deviate from the average over the short term. As long as bonuses happen more often than crashes, the incentive structure pushed traders to surf the wave.

And of course it crashed, vast sums were lost, people lost their houses, lots of people lost their jobs. Once again, the eschaton of smoothly-rising home values as an implicit proxy for smoothly rising growth was swallowed by its own contradictions: if every kind of growth raises housing prices, why bother producing wealth when you can just lever up and buy a stake in it? But if everyone’s just speculating in property instead of building real wealth, what supports those property prices?

The 2010s: What About Crypto?

When you talk about bubbles and reach for recent examples, Bitcoin naturally comes to mind. You have all the ingredients: wild optimism, crazy predictions, massive price swings up and then down, and of course a price unmoored from any sensible valuation.

But that’s by design: a currency is meant to be a bubble. Money is a bubble, and any bubble that never pops is money. (I don’t endorse the conclusions in that article, but as a description of money it’s top-notch.)

This doesn’t make Bitcoin underpriced, or overpriced, or fairly valued, or any of that — it just means that the analysis has to be different. Normally investors buy an asset with the expectation of some future return due to underlying economic activity. But they need some category of asset that they buy purely because they expect to sell it to somebody else in the near future.

I, for example, mostly use an asset known as the US Dollar. I have very good reasons to expect that tomorrow, if I want to buy a product, the seller will want to sell it to me for some US dollars. What gives that seller confidence is the collective judgment of all the other dollar-buyers, and what gives them confidence is that there’s a voracious buyer with a vast appetite for dollars, willing to accept them in exchange for something I and many others find quite valuable.

That buyer is the IRS, and every single year they sell me the right not to go to prison for refusing to pay my taxes. It’s a pretty sweet deal.

When you don’t have a committed buyer like our friends at Internal Revenue, you expect to see more volatility, but the fundamental dynamic remains in place.

Bitcoin does give its most ardent fans an eschaton to immanentize: they can hope for the collapse of the global financial system, just like our friend John Law linked above. However, it’s not required for Bitcoin believers; they can just hope for a version of gold or the Swiss Franc that’s a little easier to transfer, safer to secure (if you know what you’re doing), and that comes with a built-in script for basic smart contracts. The fact that it’s a bubble asset doesn’t mean it has to go to zero, but the fact that money is a bubble has never prevented any particular kind of money from going to zero.

The 2010s: Diversification as a Negative Externality in the Age of Discrete Volatility

Calling a market a bubble in retrospect is easy; saying it’s one now is a little too easy, and it’s bad form. You could have easily marshaled enough evidence to say that technology was in a bubble in 1997 — Cisco was trading at 7x sales and 46x earnings! A few years later, earnings had tripled and the stock was trading for, at one point, over 200x earnings.

It’s rude to call something a bubble now, because a bubble is a time period where people who believe in the bubble generate excess returns, and by definition it peaks when either the last optimist buys in or the last pessimist covers a short position.

Really, the only socially appropriate way to call something a bubble is if you do so in a way that makes it hard to bet against. So I’ll argue that we’re living through a distant echo of the 1970s, and diversification is, once again, the enemy.[8]

This time, the gnosis begins with Eugene Fama’s Capital Asset Pricing Model, the insight that any asset’s expected returns can be modeled as the sum of the risk free interest rate and the asset’s volatility to the market multiplied by the market return minus the risk-free rate. You think Consolidated Widget is a company with a solid track record, a clever CFO, a CEO who’s charming but needs to work on executing that cost-cutting plan, and a good shot at getting most of their customers to upgrade to the new 9300X model next year. Fama would just tell you it’s a stock with a beta of .9, so its risk will be slightly lower than the market, and its return commensurately so.

Of course, that’s too simplistic, so Fama’s disciples added other factors: now we can model returns based on a company’s size, its valuation, and other variables.

Where the model really takes off is when it looks at broad types of assets, measures their correlation with one another, and identifies a portfolio whose aggregate performance is satisfactory and whose risk (as determined by those offsetting correlations) is as low as possible.

For example, if you look at the last couple decades you’ll see that equities and bonds tend both produce positive returns, but correlate inversely: stocks down, treasuries up, and vice-versa. So if your risk tolerance calls for equity-level volatility, you buy some equities, buy some bonds, and lever up until your volatility reaches the market’s vol and your expected return is a bit higher. That strategy works great if you backtest over the last few decades, not so great beforehand. As it turns out, both equities and bonds face a risk from unexpected inflation, and when inflation spikes they both take a hit.

So you need to go further.

Add more assets and you can add more leverage: highly-rated corporate bonds, high-yield bonds, private equity, venture capital, timberland; they all produce different streams of returns, some of them uncorrelated or slightly correlated with one another, some negatively correlated.

Eventually, what you end up with is the least cross-correlated subset of the global investable portfolio. You are long a single meta-factor, called “uncorrelated growth,” and the assets that are left behind are the ones whose returns don’t justify their correlation to global asset performance.

This is theoretically optimal. It has two problems:

  1. The backtest problem: Our correlations are based on historical data. For some assets, we have prices going back a century or more. For UK bonds, you can construct a reasonable time series, with just a few gaps, going back to the reign of William III. Other datasets are harder. Good alternative asset data isn’t available because the industry was invented a few decades ago, and in other cases the numbers are spotty; one driver of real estate’s low volatilty is that, during a serious recession, banks are praying you won’t sell because then they’ll have to admit that their loan won’t be fully recovered. It’s easy to reduce your volatility if you can just close your eyes and ignore the most painful quote. And even this doesn’t quite capture the problem, because the other issue with finite backtests is that risks have an infinitely long tail. Run a backtest for ten years, and you’ll be able to hedge out any risk that shows up once a decade. Run a hundred-year backtest, and you can hedge out once-in-a-century risks. But a quick look at recent news will tell you that once-in-a-century events have a funny habit of happening more often than that. Humans are always inventing new ways to add chaos to the system.
  2. The other backtest problem: The other backtest problem is harder to model and harder to solve: what is the effect on markets when people are choosing asset classes, not companies? Nobody knows. It hasn’t happened before. Every time passive investment as a percentage of total investment ticks up, we are deeper into uncharted territory. If passive takes share from active, it means active investors have to sell their positions, so in the aggregate, the stocks stock pickers own will go down relative to the market (the ones they’re shorting, if they short, will perforce go up, both from active managers exiting positions and from indexers adding to them). This means that the “passive factor” exhibits positive returns, which accelerates the transition. If your model captures 99% of reality but determines 100% of prices, the remaining 1% will be the only source of unexpected variance.
    This creates a separate dynamic: discrete, rather than continuous, volatility. Volatility used to be modeled as a bell curve, then a bell curve with fat tails. But now, many factor strategies practice de facto mean reversion (the value and size factors are mean reversion, and rotating into factors that have underperformed recently is, too). Mean reversion traders reduce the amplitude of swings, but this means that if they leave the market, swings get wider. So the bell curve shrinks towards the middle, but fattens more at the tails.

Not all asset allocation follows this model, and not all attempts to follow this model pour money into passive strategies. Endowments invest in hedge funds, venture capital, and private equity, all of which indirectly bet against passive (hedge funds do it directly; PE does it by buying undervalued companies and selling them at better prices, and VC does it just by selling). But all the historical models of asset class-level performance — of small stocks, big stocks, Swedish stocks, Bulgarian stocks, bonds, preferreds, vol harvesting, etc. — all of them assume that the passive investor is a price-taker.

They put an investor in a position where exposure to factors replaces ownership of businesses, but exposure to factors only works if there’s somebody analyzing the business, somebody else disagreeing with them, somebody else who realizes that they’re both completely missing the point, and so on. We used to live in a world where individual investors and institutions made decisions one company at a time. Now we live in a post-Marxist world where the bourgeoisie are alienated from their capital.

If passives are a price-maker, nobody’s really in charge, and when that happens we’d expect the classic Minsky moment: the assets stock-pickers like get perversely undervalued, and the ones they don’t like get overpriced; eventually, the gnostic marginal trader overwhelms everybody else, and markets achieve the eschaton where statistical factors are perfectly explanatory and the specifics of companies don’t matter.


Cross-disciplinary work is a good intellectual party trick, and it works for a good reason: many academic disciplines find reasons to model behaviors that are echoed in financial markets, although sometimes it goes the other way around. Einstein discovered the math behind Brownian Motion in 1905, while thinking about atomic particles; Louis Bachelier scooped him by five years while writing about stock options.[9]

Voegelin was speaking at a time when the US had an obvious enemy in the Soviet Union, but he talked about an invisible problem. The insidious gnosis is always invisible. If it gets noticed before it takes power it’s suppressed, while after it takes over it tends to suppress discussion. While Voegelin was talking about an obscure problem, his alternative was even harder to see: faith in the transcendent is not exactly something you can measure with your transcendoscope.

It reminds me of another discussion of the unknown, from 1955. The stock market was gyrating wildly, and Congress demanded answers: why were stocks so high!? They’d finally passed the 1929 peak. Among others, they interrogated Warren Buffett’s mentor, Benjamin Graham.

Chairman: One other question and I will desist. When you find a special situation and you decide, just for illustration, that you can buy for 10 and it is worth 30, and you take a position, and then you cannot realize it until a lot of other people decide it is worth 30, how is that process brought about — by advertising, or what happens? (Rephrasing) What causes a cheap stock to find its value?

Graham: That is one of the mysteries of our business, and it is a mystery to me as well as to everybody else. [But] we know from experience that eventually the market catches up with value.

No linear regressions, no confidence intervals, no backtests or monte carlo simulations. Just… mystery. At some point, you run out of things that can be accurately modeled; you’re left with the kind of risk you can’t easily model, either. At that point, you can keep adding in additional variables, controlling for more, trimming or expanding the interval of your backtest — or you can concede that underneath

Voegelin could be wrong. His theory is suspiciously non-quantitative, even for a social scientist. But if you dismiss him, you won’t be the first person whose systematizing model left no room for error and thus lacked any real foundation. Before you had it all figured out, Voegelin had you all figured out.


Normally my stuff doesn’t get an acknowledgements section, but normally I don’t hit ten thousand words. This piece is actually a birthday gift from my wife: she asked what I wanted and I said “A full weekend day when you watch the kids and I just write all day,” . That’s what I got, and this is the result. Thanks! 32 is great so far.