Monday, July 07, 2008

Money Follows Latest PnL

Money is now pouring into commodity funds, and any other strategy that has worked for the past 18 months. Do these strategies have higher expected returns, or lower risk than 18 months ago? I don't think so, but I do think it plays off a common investing tendency to use the recent past as large indicator of future returns. This reminds me of a Gallup poll put out by Paine Weber in 1998, at the beginning of the stock market boom. They surveyed an expected return of 13% from investors. After back to back 20%+ returns in the S&P500, and the Nasdaq more than doubled, investors raised their expectations to 18% in February of 2000, right before the peak. Two years later, after a 50% correction, they anticipated only a 7% expected return. Steve Sharpe and Gene Amromin also looked at survey data, and found investors were insanely simple: when the market has risen, they raise their estimate of future return and lower the perceived risk. On this dimension, investors are stupid.

I remember pitching a strategy to a hedge fund guy in 1999 and he said, it worked fine, but not so much in the past year. Could I tweak my model to work that year too? 1999 saw the Nasdaq double, so anything less, especially negative, seemed positively lousy. Another instance in the 90's I was talking to someone in the equity index group at my bank, and said that, since low vol stocks outperform high vol stocks, you could generate a couple percent premium by investing in low volatility stocks and have low aggregate 'risk' as measured by vol or beta. He said, 'couple of percent? I can get 10%!'

Another peculiar bias is that no matter how long a backtest, people want the model to have worked in the past 12 months. Looking at the rise in Commodity ETF and commodity hedge funds, it seems that people merely following the money. I know a very smart and famous economist charged with coming up with a futures strategy because commodity futures have been going through the roof. And of course, all you have to do is rationalize a long commodity bias over the past 5 years and whatever strategy you have will have done well. Smart guy that he is, I'm sure he'll rationalize his model very well, but it's all pointless because there's an appetite to invest in what has worked recently. The only trick is some logic to convince yourself, or you backers, that this isn't so dumb as merely going long those commodities that went up over the past 5 years.

If you look at mutual fund inflows, you see a distinct inflow pattern that's highly nonlinear and dependent only on the past year's returns (Sirri and Tufano, 1998 JoF; Chevalier and Ellison, 1995), and convex payouts are more valuable the more risk you take (e.g., higher vol increases the option price). Risk metrics or benchmark returns don't matter AT ALL in mutual fund inflows, it's just past year's absolute return, because for someone who knew, there was no risk, and the strategist will swear there was no risk, he just knew. This is probably relevant to momentum, in that initially something works for real reasons, it's eventually accentuated by those putting capital into what has gone up, which causes positive autocorrelation.

I was reading the latest Bloomberg magazine, and it had an article by Kambiz Foroohar on recent successful stock market pickers. It highlighted the 'best' calls, meaning, analysts were highlighted for choosing the biggest winners. It didn't give average stats for the individuals, though it did show 'averages' by firm, which ranged around 50% for 'correct' picks. Also, it didn't count a pick correct if it was called too early, so if you were bearish on Bear Stearns in 2006, the fact that it failed in March 2008 doesn't count, even though the fundamental (weak housing market, too much housing exposure) was baked in then. Some statistics only targeted the 'highest volatility stocks', meaning, if you analyzed a low vol stock, you couldn't be right. The market seems to evaluate investors using a metric that clearly gives an incentive for merely taking big gambles on highly speculative plays, updating them frequently, and if one comes in, you ignore everything else and ride it. Given the nonlinear inflows or media attention to a 'big winner', professionals are very much like stupid retail traders, all stalking Peter Lynch's ten-bagger (something that can rise 10-fold).

I was talking to someone running a hedge fund recently, and he said they were looking for strategies that produced 'out-of-sample' Sharpe ratios of 2+, preferably 4. It was like a mutual fund manager saying he wanted people who outperformed the benchmarks by 10%, preferably 20%. It was a big fund. Most (all?) people who would surpass this filter are ignorant or deceitful, because such objectively good strategies are about as rare going forward as they have been in the past (the average fund, which is a composite of strategies, generally has a Sharpe below 1.0).

In sum, institutional preferences are for investments or strategies that 1) have huge potential returns or 2) work all the time. I think positive thinking is healthy, but at this level it's a little bit like candidate criteria that no one passes, so the choice is made for some highly unquantifiable reason because the alternative is to choose no one.

This suggests that stacking the odds in your favor is easy, if you just have a long horizon that avoids high fliers. You won't win any single tourney, and you may never get a chance to ply the strategy, because people with capital want someone to gamble their money on something that can win huge though rationalize it well. This preference isn't going away, so I say, invest in conspicuously boring assets (though, not as boring as 'cash').

7 comments:

Anonymous said...

Thanks for dumbing down this post so us non-finance people can understand it.

Is there a convenient way to invest in a low volatility portfolio? Wouldn't a "value" fund be pretty close? - Is there something even better? For the bond portion I can just buy AAA.

Eric Falkenstein said...

Well, be careful, there is the BBB-AAA spread that is 'too high', so BBB is the sweet spot in terms of bond investing (AAA bonds seem to have a liquidity premium). The annualized default rate for BBB bonds is only 0.25%, insignificant for a diversified portfolio, but you get an extra 120 basis points with miniscule extra vol.

As per low risk stocks, Value has a lower beta, so on that dimension it's lower risk, but fund managers tend to emphasize 'risky' value stocks. There aren't any good low risk index funds.

Pete S said...

Relevant to your point on why these things don't work in the future: Any decent modeler knows that if you overfit to the data, you are giving too much weight to the outliers. That will reduce the model's ability to predict.
One challenge with financial market data is that you don't usually have a "hold-out" sample, as you do when modeling some other data (say, supermarket scanner data), to test your estimates.

Anonymous said...

that spread is too high when you consider all the angles. possible legislative changes which will basically increase payments to BBB strips at the expense of AAA, incompetent servicing, hard to get leverage for players who can value these strips etc.

point is the forces that distort the markets are more difficult to quantify than simple historical default rates. I came across a post on overcomingbias following your post there.

http://www.overcomingbias.com/2006/12/the_future_of_o_1/comments/page/2/#comments

in the comments discussion someone explained very plausible (to me at least) why the markets were so wrong assessing a 6% probability of $100 oil price by 2010. talking about orders of magnitude...

Anonymous said...

peter, I would think not giving enough weight to outliers is a bigger problem. in sample testing is just not predictive, that's all.

J said...

"Momentum" should not exist in a rational market. But our monkey minds are wired to copy the successful and to repeat what once produced the desired outcome. This is so much so that it is possible to eke out a living in the market playing momentum. Interesting note.

Anonymous said...

Is there any pattern to the process reversing itself on an asset class level? In your example, when does money flow OUT of commodity funds - either when another asset class offers better returns or do negative returns beget more negative returns....