Extending this further, Jensen, in '68 recognized that CAPM needed an additional factor (alpha), which some consider the manager's skill in finding performance beyond beta, generally known as Jensen's alpha and one of the many risk-adjusted return measures that are available and often used by asset managers. But if CAPM is truly dead, shouldn't Jensen's alpha be, too?
The Journal of Performance Measurement® is scheduled to include an article by Fama and French in its summer issue which does an exceptional job of describing the CAPM and addressing many of the tests which have been performed. It provides a very concise and fairly intuitive discussion of the many issues surrounding this area of investments finance. Perhaps if more in the industry understand the model's failings its constituents will be dropped from common use... but perhaps not.
We are planning a webinar in the coming months as well to discuss the model in what we hope will be an easy to comprehend manner. More details to follow.
.png)
Like most things in life, there are a few simple ideas that make good sense of things in the long run. The "randomness" of short term noise (short term trading, short term performance measurement...) is of course impossible to evaluate and quantify because it is, well, noise. Unfortunately, much investment activity (and fee generation) revolves around such noise. How does Beta fit into all this? And why does Beta survive as a meaningful statistic? Simply because IF you have an appropriate benchmark that describes the market exposures of an investment strategy, then Beta will indeed explain the majority of the performance over the long run, with the unexplained residual being rather small and entirely due to "idiosyncratic" or active effects. This can be easily demonstrated for equities, and even more strongly shown for bonds using the comparable statistic of Duration.
ReplyDeleteOf course, getting the benchmark right is the key, and it's the part where most "practitioners" get it wrong. Of course characteristics such as size and style (i.e. value vs growth) matter tremendously. So, you have the wrong benchmark if you do not get the style right. But once the benchmark is right we find that beta does a good job of evaluating performance.
What about beta as a predictor of return? Well, the idea that you can earn excess returns after adjusting for risk is contradictory to the basic principles of investments. We all understand that you are only compensated for market risk; if you diversify away idiosyncratic risk then you eliminate the return as well - there is no "free lunch." Unfortunately, we find that idiosyncratic risk is more difficult to diversify than most think. The outdated notion that 20-40 stocks form a diversified portfolio has long been disproven, yet many cling to this flawed idea when they compare such a portfolio to a broad market benchmark, or when they use a market benchmark to evaluate the predictive value of a Beta statistic. It's not the statistic that's flawed, it's the benchmark that's wrong. Yet so many scholarly articles are written (and good math is wasted) in explaining why a statistic fails after trying to use it to analyze performance against the wrong benchmark.
In the end, better benchmarks make for better analysis. If you get the benchmark wrong, then ALL of the resulting analysis will be wrong: the excess return, the attribution of excess return, and especially the relative risk. But, if you get the benchmark right, then you will produce relevant analysis with a few simple (and yes, still relevant) statistics such as Beta. Want proof? Just plot long term performance of various investments against their style-adjusted benchmarks and see how closely they "hug the line." The small residuals are the noise of idiosyncratic effects - and hopefully they are slightly positive to compensate for the efforts and expenses of active management.
Steve, thanks for your insights.
ReplyDeleteYou're no doubt aware of the many studies that Fama & French conducted where they compared CAPM to "markets" which comprised only U.S. stocks; I'm unaware that they conducted a study where the "market" was simply the appropriate benchmark. Some empirical research into such an evaluation would go a long way to justifying what you suggest. If the manager is truly managing against the S&P 500, then would beta, in fact, identify the bulk of the source of the return with a minimal residual remaining?