What Beta Is and What Beta Isn’t
Beta is a metric, a measurement of one return stream as it relates to a relative benchmark.
Beta is not “exposure to a market.”
Beta is not “correlation to a market.”
Beta is not a measure of “volatility.”
If one takes a return series and performs a linear regression against a relative benchmark for that return series, with the benchmark as the independent variable, one finds a regression line defined as slope times benchmark plus intercept plus error term. The slope is the beta; the intercept is the alpha.
If the benchmark used is the dividend-adjusted S&P 500 return, we can easily ascertain some facts about beta, through review of backtested system results.
* Trading systems that never use U.S. stocks can have beta very close to 1.00.
* This doesn’t mean the systems’ correlation to the SPY is high; the error term could be large.
* In the case of a system with a large error term, volatility is likely much higher than that of the SPY, even though beta is close to 1.00.
* Trading systems that exclusively use U.S. stocks can have beta that is rather low (like 0.585 on monthly periods and -0.046 on annual periods), even though they are 100% exposed to the U.S. market, 100% of the time.
* Such systems might have correlation coefficients “R” that are rather low (like 0.416 on monthly periods and -0.024 on annual periods), even though they are 100% exposed to the U.S. market, 100% of the time.
* Volatility for such systems may be extremely high on a monthly basis, even though beta is low.
Please … help stop the abuse. If you want to discuss an allocation to an asset class like U.S. stocks, or the use of a known factor model leading to outperformance in stocks, or the volatility of a portfolio, please, please, PLEASE just use the correct terminology, and leave my poor buddy beta alone.
August 13th, 2008 at 6:41 am
Bill,
Great post…….I hate the way people throw the word “Beta” around, usually in the wrong context.
Jeff
August 14th, 2008 at 10:07 pm
Very true for Beta in a CAPM context, but many people also refer to beta in an APT context. In that case, they are talking about exposure to some kind of common factor or variable. Such exposure is calculated in the manner you describe, but their are usually multiple variables, and multiple betas, and they do not necessarily have to be associated with benchmarks (although the factors or variables usually can be expressed as benchmarks in some way)
However these are APT betas, and Beta with a capital B is usually referring to a single-variable/benchmark beta in a CAPM form
September 4th, 2008 at 6:57 am
[…] discussions is the assertion that the strategy will have “beta of 1 to the market” - this is “beta bullshit” that I have already addressed in another post. Beta could be high or low, depending on the factor models […]