What alternatives are made of

In 2008 the American stock market fell thirty-seven percent. Hedge funds, which had been sold for a decade as the thing that would protect you from that, lost eighteen percent on average, their worst year on record. Funds that called themselves market neutral lost money. Funds that arbitraged convertible bonds lost a third. The only strategies that made money were the ones that bet on trends and the ones that bet against the market outright. A lot of families discovered that year what they had actually been holding.

The five papers in this chapter are the tools for finding that out in advance. Two of them take hedge funds apart. Two of them build a ruler for private equity and use it. One measures venture capital properly for the first time. Read together, they answer a single question: when you pay two and twenty for something you cannot see inside, what is in the box.

I have paid those fees for years. Some of the funds earned them and some did not, and for a long time I could not tell which was which. These papers gave me the tools to tell.

Fung and Hsieh, 2001

William Fung and David Hsieh started from a problem. Hedge funds report their returns, and if you run the usual regressions on those returns, the way you would for a mutual fund, you get nonsense. The exposures come out small and unstable. The reason is that many hedge fund strategies do not hold a fixed mix of assets. They hold a position that changes shape as the market moves, and a strategy like that has to be described by the shape of its payoff, by what it does in a big move up and a big move down, rather than by a list of what it owns.

They took trend followers as the test case. A trend follower buys what has been rising and sells what has been falling, across dozens of markets. Fung and Hsieh showed that the payoff of that strategy looks almost exactly like a portfolio of a particular kind of option, one that pays you the difference between the highest and lowest price a market touched over a period. Such an option costs a little every month in quiet markets and pays a lot when a market moves a long way in one direction. That is the whole return of trend following. It bleeds in calm years and gets paid in crises, which is why it made money in 2008 while everything else fell.

Once you can draw a strategy’s payoff, two things become possible. You can price it, because options have prices. And you can judge the fee, because the same shape can often be built for far less than two and twenty. Fung and Hsieh founded the discipline of asking, for any alternative, what is the cheapest way to buy this shape.

For a family whose real risks are a stock market fall and a margin call, the convex payoff is one of the few things worth owning that pays in those moments. The lesson is to buy the payoff from the cheapest systematic vehicle that produces it, and to stop paying for mystique.

Asness, Krail and Liew, 2001

Clifford Asness and two colleagues asked the question in the title of their paper, do hedge funds hedge, and answered it with a trick that has been used on every illiquid asset since.

Hedge funds hold positions that are hard to price. Some trade rarely, some are marked by the manager. So when the stock market falls in a month, a fund’s reported value falls only part of the way that month, and the rest shows up over the following months as the marks catch up. If you measure the fund’s exposure to the market using only the same month’s returns, you miss the part that arrives late. The authors added the previous months’ market returns to the regression and summed the coefficients. The industry’s exposure to the stock market roughly doubled, from around four tenths to over eight tenths. Categories sold as market neutral turned out to carry substantial exposure. Most of the industry’s claimed alpha disappeared, and the smoothing had also flattered the volatility figures, so the risk-adjusted returns had been overstated twice.

The point is mechanical. Stale prices are mostly a fact about illiquid assets, and nobody needs to be dishonest for them to appear. The problem is what people do with the smoothed numbers: they build risk budgets and justify fees on figures that understate the exposure. Seven years after the paper, 2008 arrived, the marks caught up all at once, and the industry’s true exposure to the market was on display in a single year.

The rule that follows is one I use on every private fund I hold, not just hedge funds. Before you accept a reported volatility or a reported correlation from anything that is marked by hand, undo the smoothing. Assume that the true correlation to stocks in a crisis is close to one, and size the position on that. The risk of an illiquid asset is all there. It just shows up late.

Kaplan and Schoar, 2005

Steven Kaplan and Antoinette Schoar built the ruler for private equity. Private equity funds call your money in over several years and return it over several more, and the industry reports the result as an internal rate of return, a number that can be lifted by returning money early or by borrowing to delay calling it. Kaplan and Schoar asked a plainer question. Take every dollar the fund called and every dollar it returned, and ask what those same cash flows would have earned had they gone into and out of the stock market index on the same dates. The ratio of the two is the public market equivalent. Above one, the fund beat indexing the money. Below one, it did not.

Three facts came out. The average buyout fund, net of fees, roughly matched the index, with a public market equivalent just under one. Performance persisted: a manager whose last fund was in the top quarter was far more likely than chance to be there again, something that had never been found for mutual funds. And money flowed into private equity after good years, and the funds raised in those years did worse.

The framework is permanent. The facts moved. With better data, the historical buyout result improved, as the next paper shows, and the persistence largely faded after 2000 as the industry grew and institutionalized, surviving mainly in venture capital, where the persistent funds are closed to new money anyway. What remains is the ruler and the third fact. Measure every private fund against the public money you did not invest, and commit the same amount every year regardless of how you feel, because the funds raised in the good years, when everyone wants in, were the ones that did worst.

Harris, Jenkinson and Kaplan, 2014

The buyout question was settled with better data. Robert Harris, Tim Jenkinson and Steven Kaplan used records sourced from the investors themselves rather than from the managers, which removed the biases of the earlier databases, and found that American buyout funds raised before 2006 had beaten the stock market by about three percent a year after fees, a public market equivalent of roughly one point two. Venture capital had done spectacularly in the 1990s, collapsed after the technology bubble, and recovered only partly.

Then the sobering part, confirmed by the same authors in later work. For funds raised after 2005, the buyout advantage shrank toward nothing as capital flooded into the asset class. The more money raised in a vintage year, the worse that vintage did. The mechanism is the one Berk and Green described for mutual funds, operating here at the level of a whole industry: an edge attracts money until the money eats the edge.

Two habits follow for anyone still committing. Treat the pre-2006 record as a different world, and underwrite new commitments assuming a public market equivalent of one, unless you can say in a sentence why your access is different. And read every internal rate of return with suspicion, because subscription lines and early distributions can inflate it; ask for the public market equivalent and for cash returned as a multiple of cash invested, and look at those.

Cochrane, 2005

Venture capital had the worst measurement problem of all, and John Cochrane fixed it. You only see the value of a venture investment when something happens to the company: a new round of funding, an acquisition, a public listing. Those events happen because the value went up. The failures sit quietly at their last mark, or vanish, and never enter the return statistics. So the average return people quoted for venture, close to a hundred percent a year in the raw data, was an average of the winners.

Cochrane modeled the return process and the probability of observing it together, and corrected for the selection. The average return, measured properly, fell to around fifteen percent in log terms, with volatility near ninety percent at the level of the individual company. The shape of the distribution is the finding. Most companies fail or go nowhere. A tiny fraction return ten or a hundred times the money. The arithmetic average that gets quoted is made entirely of that tail.

For a family this governs every venture fund and every angel cheque. The expected outcome of any single investment is close to zero. The return of the asset class exists only at the level of a wide portfolio that catches the rare outlier, which means diversification is a condition of survival, and a few good angel outcomes prove nothing about skill. With ninety percent volatility, they prove variance. When you review a venture book, use cash returned against cash invested across the whole portfolio since the beginning, because memory keeps the winners and forgets the rest.

What I do

Before I pay an alternative fee I try to draw the payoff. If I cannot describe what the strategy does in a crash and what it costs in a quiet year, I do not own it. Where the shape can be bought cheaply from a systematic vehicle, I buy it there.

Every private fund on the balance sheet is carried at a volatility about double what it reports and a correlation to stocks that I assume goes to one in a bad year. Every private equity commitment is measured against the public market equivalent, computed every year, and the commitments are paced evenly across vintages. In venture I follow Cochrane to the letter. The expected return of a single angel cheque is close to zero, so I invest only through venture funds, where the portfolio is wide enough to catch the rare winner, and I make no direct investments and no co-investments. The sleeve is judged on cash out against cash in, across everything, since day one.

Three habits

Draw the payoff before you pay the fee. If nobody can show you the shape of the return in a crash and in a calm year, the fee is for a story.

Undo the smoothing before you believe the numbers. Any asset marked by hand reports too little volatility and too little correlation; correct both, and size on the corrected figures.

Measure private funds against the public money you did not invest, and keep the pace steady. The public market equivalent is the only honest yardstick, and the years you most want to commit are the years the evidence says to hold back.

Gala: when someone offers you something you cannot see inside, ask them to draw what it does on the worst day. If they cannot, the fee is for the story, and you already know how stories end.