It's all just algorithms.

Photographer: Ben Hider/Getty Images.

Are Hedge Funds Fake?

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.
Read More.
a | A

Here's the seven-factor Fung and Hsieh model:

We start with Fung and Hsieh (2001) who argue that, theoretically, a payoff of an actively managed fund that can perfectly predict price trends resembles a payoff of a lookback straddle. They call the optimal trend-following strategy of buying lookback straddles the “Primitive Trend-Following Strategy” (PTFS). The authors construct PTFS factors for stocks, bonds, interest rates, and currency and commodity markets, and they show that returns of commodity trading advisors (CTAs), a popular hedge fund strategy, are well explained by these factors. Furthermore, Fung and Hsieh (2004) combine three PTFS factors (for bonds, currencies, and commodities) with four conventional asset-based style (ABS) factors and propose a pricing model for any hedge fund. This model is widely known in the literature as the seven-factor Fung and Hsieh (FH7) model.

That's from this paper by Mikhail Tupitsyn and Paul Lajbcygier of Monash University in Australia, which finds that most hedge funds' performance has only linear exposures to factors like, um, those seven factors. Matt Yglesias summarizes the result as "most hedge funds don't appear to be doing any hedging or active management at all": 

To reach that conclusion, the authors took a large set of hedge funds that deploy a variety of investment strategies and researched whether the funds exhibit linear or nonlinear returns relative to broader markets. The math behind this gets pretty funky, but the basic idea is that linear returns are simply proportional to what you could get with an index fund, whereas nonlinear returns aren't correlated with a simple index.

Now that would almost be true, if you could get an index fund made up of lookback straddles on stocks, bonds, interest rates, currencies and commodities. But obviously your S&P 500 index fund is not made up of lookback straddles. A lookback straddle is not the same thing as just buying and holding the broad stock market. You can't even buy a lookback straddle from Vest. A lookback straddle is only tenuously a thing at all. It is a way to replicate certain hedge fund strategies mathematically, not a product you buy at the supermarket.  

Tupitsyn and Lajbcygier's paper is called "Passive Hedge Funds," and the first sentence of the abstract is "We show that most hedge fund managers are passive, not active." This is, I hope, meant to be read in only the very most metaphorical sense. Specifically, in the sense that "Active management should be manifest through nonlinear exposure to the systematic risk factors that drive hedge fund returns," which is the second sentence of the abstract. Tupitsyn and Lajbcygier find that a majority of hedge funds have only linear factor exposures, which they count as "passive." But if you read only the first sentence, you might well conclude that hedge funds are literally passive, in the sense that they buy and hold broad market indexes, and therefore overcharge investors for something that they could get cheaply from Vanguard. This is the conclusion that Yglesias, and Hamilton Nolan at Gawker, seem to have drawn.

Of course it is possible that hedge funds as a group underperform the market net of fees, that hedge fund fees overall are too high, that most hedge fund investors rue the day they ever heard of "hedge funds," etc. It is certainly possible that some hedge fund managers just put all the money in index funds and play solitaire at work. Here is a much longer paper, by Mila Getmansky, Peter Lee and Andrew Lo, with the weirdly bland title "Hedge Funds: A Dynamic Industry In Transition," that rounds up the evidence on hedge fund performance and how much of it can be attributed to identifiable systematic factors rather than idiosyncratic skill. It is not, shall we say, an unalloyed case for the wonderfulness of hedge funds.

But these attempts to reduce hedge funds to a list of factors are not quite the same as proof that hedge fund managers are passive, or slacking, or lying about what they're up to.  Of course many hedge fund strategies are correlated with the stock market or with credit spreads. In the long run, everyone is correlated with the economy. If the economy does well, stocks do well, people get rich, hedge fund managers are happy, everything is great. If the economy does poorly, I mean, sure, some hedge funds do great, though probably fewer than expect to. But, despite the name, it would be pretty odd if hedge funds in aggregate did better as the economy got worse. In the aggregate, a collection of strategies that own assets will do better as asset prices go up. You don't see a lot of fat happy hedge fund managers in "Mad Max."  

Papers like this are attempts to describe and quantify and explain those connections between the actions of hedge fund managers and broader systematic factors. They are about proving that hedge funds can be reduced to a series of rules and heuristics and equations. This should not be surprising, because economics is about reducing human behavior to a series of rules and heuristics and equations. Hedge fund managers are humans; as a rule of thumb they tend to be more rational and economically motivated than the average human. So it would make sense that, if the operation of an economy of hundreds of millions of entrepreneurs and managers and factory workers and bartenders and bloggers and Uber drivers and Etsy crafters could be modeled, even partially and imperfectly, in a series of equations, then the operation of a few hundred investment firms run by mathematicians in fleece vests could be modeled at least as well. And since the job of economists is to do that modeling, they do that modeling. A paper titled "We Don't Know What Hedge Funds Are Doing, It Is Crazy Unpredictable, We Won't Even Hazard a Guess" would not get published.

The other thing going on here is that the modern world has a tendency toward automation. A lot of people want to build robots that can replace hedge fund managers, or at least do most of their jobs reasonably well. It is utterly, utterly plausible that they will eventually do this. They're building self-driving cars! Global financial markets are unpredictable, but they are much more predictable than Park Slope pedestrians. And if your hedge fund crashes, it probably won't kill anyone.  Building a robot that can replicate the performance of a hedge fund manager, but for a lower price and with less shouting, is an attractive and reasonable and potentially lucrative thing to attempt. And so lots of people are working on it. Two of those people are Peter Lee and Andrew Lo, authors of that "Dynamic Industry in Transition Paper," who "are affiliated with an asset management company offering a hedge-fund beta replication mutual fund." If you can build a robot that acts like a hedge fund manager, someone will buy that robot. 

Saying that most hedge funds are "passive," in the Tupitsyn and Lajbcygier sense, doesn't mean that they are literally "passive" in the way most people use that word. It doesn't mean that they don't do any active management, or that you could replicate their performance by buying broad indexes. It just means that there are models that describe their performance in relatively straightforward ways. It means, if the paper is right, and if many hedge fund managers are linearly and predictably exposed to known investable factors, that it wouldn't be too hard to replace them with robots. If you built the robots right. The robots would probably be cheaper than the managers, but they wouldn't just be buying and holding index funds.

But the fact that a hedge fund manager can be approximated by an algorithm is no reason for shame, although it may in the long run be a reason for lower fees. There's a famous paper that built a robot to approximate Warren Buffett. The robot can replicate his performance fairly well, in-sample, at least as regards public stocks. "Bring me a cheeseburger and a Cherry Coke," the robot cries, between games of ping pong. People are partly predictable.  Modern science is always getting better at explaining them: We can sequence the genome and predict elections and teach a computer to play Jeopardy. In many ways, this increased understanding is good. But it does tend to take the mystery and magic out of things.

  1. Here's Fung and Hsieh's 2001 paper originally describing the model.

  2. Plus other factors. In addition to "the seven factors of Fung and Hsieh (2004b)," there are "the six factors of Hasanhodzic and Lo (2007)" and "an extended set of Hasanhodzic and Lo (2007) variables augmented with another eight factors identified from the literature."

  3. It's even nonlinear! The returns have linear exposures to lookback straddles, but lookback straddles have nonlinear exposures to the underlying assets.

  4. I mean, as stylized facts, those all seem pretty popular. There are counterarguments. Comparing all hedge funds -- even those that don't trade equities, and those meant to be uncorrelated or inversely correlated with equities -- to the S&P 500, as many critics do, seems silly. And broad performance is notoriously hard to measure, as there is in a sense no such thing as "the hedge fund industry." There is a wide variety of hedge funds, and some are good and some are bad and some are even identifiably and predictably good, and public indexes only imperfectly capture any of those dynamics.  

    The best argument that hedge funds work is that people keep paying for them: As Cullen Roche says, "There’s a certain hilarious irony in seeing efficient market cheerleaders bash an industry that is growing by leaps and bounds as if to argue that the market is actually inefficient." Though that is an argument that proves too much (not everything that costs money is good!), and of course there are agency-cost reasons that might explain some hedge fund asset gathering. I don't know. I don't think "Are Hedge Funds Bad?" is a question susceptible to a definitive answer, certainly not in this post, though the Yes camp scores some points.

  5. Nor is a linear exposure to the factors necessarily a bad sign. "Linear" does not mean "bad." A fund whose returns equaled the S&P 500 return, plus 10 percent a year, would have a linear exposure to the S&P 500. And yet it would be wonderful. And impossible to replicate cheaply (or at all) by just buying the S&P 500.

  6. More subtly, many hedge fund strategies are correlated with lookback straddles, because a lookback straddle is a way to describe a certain trend-following strategy. You may or may not want that strategy, but if you do want it, a lot of hedge funds will give it to you, and an index fund won't.

  7. Though there is a Robert Harris novel about a hedge fund whose computer kills people. That is really what that novel is about. Spoiler alert I guess.

  8. Here's a footnote!

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author on this story:
Matt Levine at mlevine51@bloomberg.net

To contact the editor on this story:
Tobin Harshaw at tharshaw@bloomberg.net