When I analyze this investment “performance”, it's pretty scary to realize that had I been a portfolio manager from 2002-2007, I'd have done pretty well for my clients, and great for myself -- even with no investment skill. I simply was the beneficiary of a decision to over-invest in segments that got hot for 5 years. I'd have been amply rewarded for that move from 2002-2007, but not “punished” when the approach backfired in 2008. Great deal if you can get it.
As noted in last week's blog, Performance Measurement as conceptualized by investment science seeks to untangle the causes of portfolio performance, separating manager skill from other factors such as the direction of the market and the amount of risk taken. And though I didn't realize it at the time, it turns out I was making choices on how much risk to assume for my investments. By buying small and value, I had inadvertently ratcheted up the risk, so the additional rewards were to be expected in the long run.
The Capital Asset Pricing Model (CAPM) is an important foundational theory in investment science that purports to explain differences in returns of individual stocks. CAPM posits that individual equities are provided a premium based on their degree of exposure to market or systematic risk. The more risk a particular equity assumes relative to the market, the higher the variation or volatility in its returns. The relationship between individual portfolio performance, overall market returns and risk takes the form of a linear model for which each unit of systematic risk the portfolio assumes is rewarded with a unit of “excess” return over the risk-free asset (T-Bills).
The CAPM is now deployed extensively to measure portfolio performance, using returns data to distinguish manager skill from risk assumed and overall market performance. Algebraically:
returns from portfolio(j) = alpha(j) + beta(j)*market returns.
Returns from portfolio(j) and market returns are inputs for calculating alpha(j) and beta(j) with linear regression. Beta(j) is a measure of risk and covariance for portfolio(j) in contrast to the overall market. A beta value of 1 indicates average risk, in line with the market; a value > 1 is aggressive, with higher volatility than the market; and < 1 connotes a non-aggressive, lower-than-market volatility. Alpha, on the other hand, is interpreted as a measure of skill. A positive value of alpha(j) suggests portfolio performance in “excess” of the risk assumed and the direction of the market; a negative value connotes sub-par performance.
CAPM can readily explain my investment success – and failure. My portfolio returns mirrored the overall market, but more dramatically because of the excess risk/volatility I assumed. My beta was > 1, so I beat the market when times were good from 2002-2007, but underperformed on the downside in 2007-2008. My alpha was either zero or negative, reflecting an absence of skill. The good fortune was due to the positive movement of the market and the additional risk I assumed, not, unfortunately, to any portfolio management skill I possessed.
CAPM-like portfolio performance thinking has applicability for BI. Returns from portfolios represent the performance of the enterprise or unit we're looking to assess. Market returns are “common causes” that hold for all participating in the economy, segment or industry. To enjoy common cause success, it's only necessary to be “in the market” – to show up, to not screw up. Think of market returns as headwinds or tailwinds affecting all competitors. Beta is the risk assumed to get results. Higher betas provide excess returns above the market index in good times, but more significant losses in bad. And alpha -- what is left after common cause and risk returns -- is viewed as a measure of skill. Positive alpha is the holy grail of performance measurement with CAPM. Wouldn't it be nice if BI could generally distinguish common cause/market risk from skill components when assessing business performance?
Steve Miller also blogs at miller.openbi.com.