Buffett pulls ahead in wager against hedge funds
finance.fortune.cnn.com
finance.fortune.cnn.com
Which I am going to pimp shamelessly here, as I wrote the code for it a decade ago. We have bets that go out many years. There's one that won't be decided until 2150. There's even one about how the universe will end.
If you would like to make a similar long-term bet, either use the site or let me know; I'm glad to facilitate. As long as the bet is more than 2 years out and the stakes are more than $200 each, you're golden.
We'd love to see the site used more to call out people making bold/crazy predictions. E.g., is somebody saying that NoSQL databases will rule the earth in 10 years? Do you think they're wrong? Tell them to put up or shut up! We'll put you both on record.
The predictions intentionally haven't been updated. With bets the adversarial bet definition process produced clear, ajudicatable criteria. Predictions, though, were intended to be looser. We just wanted to record them, and didn't want to be in charge of deciding them.
If I had it to do over again, I'd be tempted to solve the problem in a crowdsourced way, where people get to opine on the outcome of the prediction endlessly.
I'll check out PredictionBook; it looks interesting. And if you want to discuss this stuff with me or the Long Now staff, feel free to get in touch.
Just imagining what 'long now' code looks like!
I picked Java because even then it was obvious that Java was the new COBOL. Which had its drawbacks, but that meant high stability and longevity.
I avoided complex dependencies; it ran by keeping all data in a single process and writing everything out to a set of XML docs on a filesystem. There the bet was that Moore's Law would progress faster than the data set size or the usage volume; that turned out to be true. The only real dependencies were on the core Java and Java Servlet stuff, plus an XML library.
That simple architecture made it relatively sysadmin-friendly. Data was backed up by a simple script that just checked everything into CVS every 15 minutes. When they needed to move it to a new machine there was just setting up a standard Tomcat container, dumping in the WAR, and pointing it at the data. So there was never any significant downtime.
I also wrote a bunch of unit tests, so that in theory it would be easy for another developer to pick up. Only a couple of other devs worked on it, though, so it's hard to say how much that helped. But I was really happy to have the tests a few years later when we reved it.
Eventually it got rewritten in whatever their in-house dev was using so it could be more integrated with the existing accounts. The original code base lasted a decade or so, though, so I'm satisfied with how most of the initial choices worked out.
One source that looked at 12,980 hedge funds with $1.8 trillion in management: http://faculty.baruch.cuny.edu/tbali/BaliBrownCaglayanJFE201...
To anyone who is familiar with how money management works, the conclusion in the paper is obvious. Exposure to beta will increase returns in good periods and decrease returns in bad periods.
The research looking at "average hedge fund performance" is also misleading because while the analysis is technically correct, the conclusions in those papers are obvious as well. If you hedge out most of the beta, you will get zero return on average. Once you subtract expenses, you will by default perform negatively the market. (in reality, hedge funds do not hedge all beta to zero, they hedge it pretty close to 1 so they track the market, but they probably under track it. This also leads to negative alpha)
Where hedge funds play a role in investment management is when individuals can select hedge fund managers they believe will outperform. Professional investors, including fund of funds believe they can do this, and some will do this, some won't and the total will net to less than zero.
[[Added in response to additional point below]]
The paper cited is using an econometrics model which is using factor analysis to determine where the "biggest" source of performance comes from.
Because beta is a factor, it will overshadow alpha automatically if beta is large enough. Alpha between funds will be destroyed in the model so the model will output zero alpha with enough funds. While beta will be persistent since by definition beta is a performance magnifier (for lack of a better term.)
In other words, the conclusion in the paper is obvious, and not worthy of discussion.
[[Added in response to Jesse]]
Yes, if you are aggregate all money invested in hedge funds the average return will be lower than the benchmark that Beta is based off of. ETFs may or may not beat the benchmark depending on the ETF. Some have tracking error which causes them to not beat benchmarks.
Also, you may notice that I am not arguing whether the OP is correct, I am generally agreeing with them, but the point is that his assertion is obvious and the cited paper is obvious and provides no real value.
If the returns are actually from "dumb" beta, then the hedge fund fees are not appropriate for what you're getting. You're getting market returns appropriate for the amount of risk that's being taken, rather than outperforming, and then losing most of that to fees. Yes, exposure to beta can be a smart choice, but it can be done without excessive fees.
>Where hedge funds play a role in investment management is when individuals can select hedge fund managers they believe will outperform
Yes, this is the common refrain in mutual funds too: Buy the top performing managers, not the average fund. But the research shows that hedge funds that outpeform do not persist in outperforming. This suggests that the managers are not being particularly skilled in their selection of risk. Most fund of funds do not perform significantly better either -- there's little evidence that hedge funds are providing investors real value. Hedge funds are a status symbol for the wealthy, not a smart investment.
I realize that there are benefits in diversification, but travisp is generally correct.
I'm not really disagreeing with you either: I agree that the conclusion of the paper is mostly obvious, but it's valuable because it is important to show that beta is overshadowing alpha significantly in hedge funds. I could be wrong, but I would be surprised to see top hedge fund managers claiming that their results come from beta, not alpha. If they did, it would be harder to justify the amount they are paid during periods of good performance.
I also do think that some manager skill exists in all areas of the market to a limited and varied extent. But, I suspect that even the hedge fund managers who do "provide alpha" capture most of it with their fees, leaving their clients with market returns or worse.
Its not sexy to say "take all your retirement money and put it into three funds: international stocks (eg. VEA), domestic stocks (eg. VTI), and domestic bonds (eg. AGG)" and rebalance it once a year to ratio based on your age but the research shows this is the most effective method of investing because it takes the emotion and costs out of it.
https://personal.vanguard.com/us/funds/vanguard/TargetRetire...
You are missing the rest of the world and smaller stocks. This implies that you are missing some diversification that would reduce your risk.
Three basic index funds each with very, very low fees compared to actively-managed funds. I'm at 60/30/10% respectively in those funds.
The companies in the fund of funds, presumably don't have the same "blue chip" ratings of the S&P companies and their management may still be undergoing their personal tests of courage.
In the end, in the absence of a extremely hot market - which could be a good or bad thing given the experience level of senior leadership, it's probably a better bet to go with the tried and true.