The CXO Advisory Group blog has a great post about the indicators of persistence in hedge fund performance. CXO examines a July 2010 paper, “Hedge Fund Characteristics and Performance Persistence”, by Manuel Ammann, Otto Huber and Markus Schmid, in which the authors investigate hedge fund performance persistence over periods of six to 36 months based on portfolios of hedge funds formed via double sorts on past performance and another fund characteristic, which might include the following:
- size;
- age;
- relative funds flow;
- closure to new investments;
- length of withdrawal notice period;
- length of redemption period;
- management and incentive fees;
- leverage;
- management personal investment; and,
- a Strategy Distinctiveness Index (SDI) defined as a strategy-normalized form (ten different strategy types) of one minus the R-squared of monthly returns regressed against an equally-weighted strategy index over the prior two years.
CXO reports the authors’ findings thus:
Using characteristics and groomed performance data for a broad sample of hedge funds over the period 1994-2008, they find that:
- Based on past performance alone:
- Persistence of raw returns is economically meaningful up to two years (but statistically significant only for six months).
- Persistence of multi-factor alpha is both economically and statistically highly significant up to three years. The difference in monthly future alpha between the top and bottom fifths (quintiles) of past alpha is 2.8%, 2.3%, 1.6% and 1.0% for rebalancing frequencies of six months, one year, two years and three years, respectively.
- Based on double sorts that divide each past alpha quintile into a high and low half for some other fund characteristic, SDI is the only characteristic that systematically improves prediction of future performance.
- The difference in monthly alpha between the tenth of funds with highest past alpha and higher SDI and the tenth of funds with the lowest past alpha and lower SDI is 3.0%, 2.6%, 1.8% and 1.0% for rebalancing frequencies of six months, one year, two years and three years, respectively.
- These results translate to annualized alpha improvements of about 4.0% and 2.3% for annual and biennial rebalancing, respectively.
- Results are robust to different factor models for calculating alpha, changing the order of double sorts, different demarcations for initial sorts (fifths, fourths, thirds, halves) and alternative definitions for SDI. However, the prediction enhancement of SDI disappears during the 2008 credit crisis, indicating that high-SDI funds have large idiosyncratic risks exposed by crises.
In summary, evidence indicates that hedge fund investors should focus on funds with the best past performances and the most distinctive (uncorrelated) strategies.
While past performance may be no guarantee of future results, it seems that past performance is a pretty good indicator of future performance.
In studying for CFA a lot of this type of analysis is done and I like the breakdown of risk elements, but I agree that many times the fact that stocks are cheap or expensive is the 700 billion dollar gorilla in the room — hard to advocate QE2 now that they say they are doing it to boost stock prices when as a value guy its tough to find much worth owning up here — QQQQ is at the 2007 highs, but we are spending money to make overvalued stocks go up…. anyways, its not about politics just that I agree that academics do not value analysis in the form of studying holistically a business from the ground up and evaluating its prospects and financial statements in order to determine an intrinsic value and a price you would be willing to assume risk in buying the asset hopefully at a significant discount. In the textbooks, however, it seems much more emphasis is put on beta, correllation, indexing, diversification, etc… I am all for some of these strategies, like adding gold a few years ago or looking for value in companies in the farming space due to rising commodity prices, but I mainly feel that long term business value and occassionally large discounts to tangible book value can make the best investments from a risk/reward standpoint.
LikeLike
I’ll look forward to it… hopefully you find something interesting.
LikeLike
interesting Joshua, this should be a blog post on its own. I’m certainly going to look into this more as you’ve raised some though provoking insight. awesome.
LikeLike
Forgot about this paper that inspired me to test it on more recent data.
———–
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1636408
R2: Does It Matter for Firm Valuation?
“A considerable amount of research has been devoted to why R2 differs across firms or markets, although little attention has been paid to the consequences of this difference. We fill this gap by investigating how differing R2 affects investors’ assessment of firm value. Using a sample of 90,111 firm-year observations from 1970 to 2004, we find that higher R2 leads to higher firm valuation and that, on average, high-R2 firms experience significant underperformance in the long run. These results suggest that high-R2 firms tend to be overpriced.”
———–
As far as I understand it, the general consensus on this is that stocks with a high r-squared/correlation are less costly to hedge (via futures or index puts) so you get a discount for taking this risk. I’m not really a fan of the whole risk premium theory. I think its a convoluted game of semantics academics play so as to not deal with all the empirical evidence that yes, stocks do become over and under valued.
LikeLike
Well, I think a big part of it is just math. By going long a bunch of stocks that are uncorrelated to their index, their correlations with each other are probably low as well. The stocks that are bought mostly have r-squares near zero which of course is different than stocks with the lowest correlation. Since correlations can be negative going long those stocks could at times just be a bet on gold miners and the like.
So this simple strategy is “harvesting volatility” off of a group of stocks that are zigging and zagging in relatively random directions. With the SDI factor in that paper there is no portfolio re-balancing effect since hedge funds with high SDI go on to outperform and they’re evaluated one-by-one, not as a portfolio, which lends credence to the idea that they do have distinct strategies. What I’d like to know is if a stock with a low r-squared goes on to outperform like the hedge funds do, so as to separate the re-balancing premium from any “unique business model” premium this might be capturing.
What I didn’t mention in the previous comment is that returns increase monotonically as you narrow the size of your basket, with a max return of %467 with about 20 stocks. After that results get pretty unstable.
Testing this on the Russell 2000 doesn’t do quite as well, with the bottom 1% (about 20 stocks) returning 57% versus the R2000’s 33.6% over the same period, but with more volatility.
LikeLike
thanks for sharing, that is very interesting Joshua….. any logical reason you think this is so?
LikeLike
That SDI factor is interesting… I’ve been playing around with a backtest on Russell 1000 stocks with the lowest r-squared for weekly returns against the Russell index. If one had bought the 5% of Russell 1000 stocks with the lowest r-squared, rebalanced every four weeks, for the past 9 years they’d have a 251% return vs. the Russell’s negative 4%. Transaction costs not included however. The 5% of stocks with a high r-squared underperform the index slightly, returning -10.7% over the same period.
LikeLike