[n00b101]

"Why aren't hedge funds compensated based on some sort of alpha, the difference from a benchmark?"

This is called a "relative return" strategy. Suppose the benchmark falls 70% and the hedge fund loses 50%. Your hedge fund manager has lost half your money, but has beat the benchmark by 20% (i.e. "positive alpha"). How does a fee based on alpha work in this case?

Hedge funds generally claim that they aim to achieve a positive return on investment regardless of whether markets are rising or falling (i.e. "absolute return strategy"). In theory, this means that hedge funds should have low correlation with the benchmark. This low correlation is attractive to investors because it provides a diversification benefit and (in theory) improves their efficient frontier. Low (or ideally zero) correlation to the benchmark is the main reason that institutional investors are willing to pay expensive hedge funds fees. However, if hedge fund managers have incentives based on relative returns to the benchmark, then hedge funds will be more correlated to the benchmark.

[Kadin]

> Suppose the benchmark falls 70% and the hedge fund loses 50%. Your hedge fund manager has lost half your money, but has beat the benchmark by 20% (i.e. "positive alpha"). How does a fee based on alpha work in this case?

Some percentage of 20%, the amount by which the fund outperformed the benchmark. This makes total sense to me, at least. Or rather, it's no more nonsensical than letting the fund manager take a cut of all the assets under management, win or lose, which is a common compensation strategy and has never seemed remotely reasonable.

You obviously need a way of compensating the fund manager during a bear market, or else you're going to have your staff leave as soon as the market starts declining, because they won't be getting paid. So any scheme that doesn't pay the fund manager for losing less money than the dead-hand benchmark is pretty severely flawed.

[philipodonnell]

There is usually a separate fixed percentage of assets fee for just this reason, typically 1%.

[escape_goat]

I'm asking for my edification, rather that putting forward a line of argument, but assuming for a moment that the trades are not so large as to have a direct impact on the market, there would appear to be a great deal of hindsight information available upon which to build a benchmark.

A hedge fund appears to be an investment strategy that compensates for having imperfect information. Why is it not possible to estimate an alpha on the range between the results of a completely naive Monte Carlo simulation ("no information") and the results of a search for the optimal hindsight strategy ("perfect information")? That is, the payoff for the manager will be fixed and proportional to the fraction of hindsight performance that he achieves.

You might not want to peg compensation directly to this, but rather to relative performance against the alpha (compared to other management strategies), but that's more of a salary negotiation detail.

This is the sort of idea that my brain comes up with when I try to think, except I somehow doubt that I am all that much brighter than the average hedge fund manager. I hope I'm not just wasting your time with the obvious, but why doesn't a system like that work?

[imcoconut]

These days a lot of the money in Hedge Funds is institutional. Typically institutional investors coming to funds already have a specific allocation to an investment style, market, or asset class that they are trying to fill. So they already have an idea of whether they want market exposure, which would be benchmark oriented and should have high correlation to the benchmark/market, or whether they want no market exposure, which would would have low to zero correlation to the market (they also could want exposure anywhere in between). So in long only (no shorting) funds benchmarks can be arbitrary (though are usually indexes), but they are fixed and agreed upon by manager and investor.

The mandate for the manager in benchmark oriented strategies is to track the agreed upon benchmark with a correlation (or more accurately Beta) of as close to 1 as possible, while still beating it by a certain margin. So if the benchmark is up 10, the fund should be up 15. If the benchmark is down -10 the fund should be down -5. Technically, this margin is often measured not as the difference between the fund and the benchmark but as the annualized standard deviation of the difference, known as the Tracking Error(TE). The TE is typically agreed upon between the fund and investor(s).

It's very common for a fund to charge only a management fee in benchmark oriented strategies. In a benchmark oriented fund with a target TE it would not really make sense to charge performance fee on the magnitude total fund performance since that will largely (or completely) be a function of the benchmark/market - remember correlation should be 1. It also would not make sense to charge performance fee on the excess return over the benchmark because that is incorporated into the TE and should be relatively constant over time (in the example earlier the fund should always beat the benchmark by %5.

So usually there is just the management fee. The reasoning is that if a fund can yield consistently high returns with a consistently low risk (as measured by vol/std. dev. of excess returns a.k.a. TE) then they have some skill and should be compensated accordingly. The excess returns adjusted by risk of excess returns is called Information Ratio (IR). IR is analogous to Sharpe Ratio(SR) in non benchmark oriented funds. In a long/short fund, the higher the SR the higher the fees usually. Similarly, in the long only benchmark oriented fund, the higher the IR the higher the management fee.

This is how fees usually are determined for funds that have benchmarks.

[ctrl_freak]

I'm not an investment expert, but wouldn't there be virtually unlimited upside in an optimal hindsight strategy? I.e. you could use derivatives and other financial instruments to gain an arbitrary amount of leverage and hence return, since there would be no risk.

[pyromine]

Expanding on this, this idea of looking backwards as a method determining what returns you could have made is actually the root of a lot of _bad_ investment advice.

People often look back and try to determine some type of optimal portfolio that they claim is the best in all economic environments because they found it to be the best portfolio in the certain timeframe they looked, but they had the benefit of checking hundreds of different potential sets of funds and if you were to actually calculate the probability that composition of funds is better than any other composition you'd find that it was just random chance it provided the best returns over that time frame.

Granted there is a lot to learn from looking back, but it's also very imperfect if you're not taking in to the benefit of hindsight.

[cocoablazing]

Compensating for this effect of training may be done by properly discounting the measure using the Deflated Sharpe Ratio or similar corrected SR's. I always ask any interviewee who cites experience producing these measures a question with this effect as the crux. Few come back with a correct answer.

[savanaly]

>This is called a "relative return" strategy. Suppose the benchmark falls 70% and the hedge fund loses 50%. Your hedge fund manager has lost half your money, but has beat the benchmark by 20% (i.e. "positive alpha"). How does a fee based on alpha work in this case?

I'm not who you're responding to, and I've never heard of "relative return" strategy, but it makes perfect sense to me and your hypothetical poses no confusion in my eyes. If the benchmark (agreed upon in advance of course) would have caused my money to go from M to B by the end of the year, and the hedge fund actually caused my money to go from M to H, then they get paid some percentage of H - B. It's completely irrelevant what M is in relation to H or B.