The Evolution of Alpha

The concept of alpha is the bedrock of financial markets. Without getting into the nuances of the active versus passive debate, investment managers aim to provide alpha to investors for which they charge a fee for their services. At its core, alpha is generally defined as excess return relative to a benchmark. The excess return can come from return enhancement, risk mitigation (downside protection), or a combination of the two. In the early days of defining alpha, returns from an investment fund were broken down into two sources: general and specific. The “general” source of return was an investment’s exposure to the broad market (benchmark) and called beta, while the “specific” source was derived from the decisions of the portfolio manager and labelled as alpha.

Since the 1960s when these two sources of return were defined, the formula to compute alpha has evolved. The seminal work from Michael Jensen was famously expanded on by Gene Fama and Kenneth French, who found additional factors outside of just broad market beta that influenced investment returns. The original formula, which was simply market beta plus a residual (alpha), was transformed into a model which included size and value factors to describe financial market returns more accurately. Researchers continued the segmentation of factors which contribute to the return of an investment, and in later years such factors as momentum, quality, and low volatility have been added to the factor factory. This article from Morningstar examines the evolution of alpha into beta over time and posits that once a factor has been scientifically proven as a driver of market returns, it ceases to be alpha and instead is turned into commoditized beta. If you can show quantitatively that buying stocks with low price-to-book multiples (simplistic example of the value factor) will generate a premium relative to a stock portfolio weighted by market capitalization, then any manager utilizing a value factor approach should have their performance compared to a value benchmark as opposed to a market capitalization weighted benchmark (e.g., S&P 500) when determining if the manager is adding value.

Even though previously unexplained alpha has morphed into one of the many sources of market returns, this isn’t to say that there isn’t a premium there that can be collected. However, for factors with low barriers to entry, the premium may only arise through periods of underperformance where weak hands are shaken from the market and the trade becomes less crowded. Effectively, the alpha gets commoditized to beta and the premium is then reliant on an investor coping with their behavioural biases more effectively than other market participants to realize that premium. Value investors know this cycle all too well.

The author of the Morningstar piece then goes on to provide his new definition of alpha: “alpha is the result of decisions that cannot be captured by any factor model, no matter how intricate the model, because the insight that underlies those decisions has not yet become public knowledge.” We would agree, but also go one step further and acknowledge that even if a certain alpha becomes public knowledge, there may be a structural reason that keeps the factor from getting commoditized into a component of market beta.

The point of all this preamble is to consider how investment portfolios can be constructed where betas (whether fundamental or economic) are efficiently balanced, and if there are opportunities to harvest alpha where others cannot. This is exactly what we believe a risk parity strategy can accomplish. Not only can a risk budgeting engine construct more robust portfolios, but we also believe there is a structural long-term alpha advantage that isn’t easily arbitraged away into a component of beta. First, we’ve shown in various research notes (here, here, and here) that there is a long-term advantage to diversifying uncompensated concentration risk. The insidious effect of volatility drag reduces the ability of a portfolio to compound in future periods and leads to a lower terminal value. Diversification reduces the effect of volatility drag and increases a portfolio’s Sharpe ratio (excess return per unit of volatility).

The challenge for many investors is that although volatility is reduced through diversification, the expected return of the portfolio is also lower. In practice, investors eschew too much diversification because it lowers their expected return to a level that won’t accomplish their investment goals, despite a more material reduction in volatility (i.e., higher Sharpe). Therefore, if an investor requires a higher expected return to meet their investment goals, the allocator (individual, investment advisor, institution, etc.) ends up having to capitulate and increase exposure to a higher expected return asset class (to increase the overall expected return of the portfolio), thereby increasing concentration risk and decreasing portfolio efficiency (lower Sharpe). Instead of increasing concentration risk to an asset class that has a higher expected return (i.e., equites), we believe that volatility scaling is the way investors can “eat the Sharpe” and harvest the benefits of diversification without sacrificing a required return target.

As everyone in financial markets knows, there is no free lunch (other than diversification), but we believe that aversion towards leverage impairs the ability of investors to harvest the diversification premium, which in turn will result in long-term persistent alpha. Utilizing a judicious amount of leverage on a diversified portfolio where a variety of betas are optimally balanced is much different than using leverage in a concentrated fashion such as for real estate or investing in small capitalization equities (as a simplistic example), and we believe that this approach to portfolio management should result in superior outcomes. While we won’t be able to scientifically prove a risk-allocation engine and volatility scaling is statistically significant given there aren’t enough unique live data points over long enough holding periods, simulated data shows that we can expect there to be harvestable alpha utilizing this approach. At the very least, using simulated data to confirm our hypothesis helps to provide confidence that our out-of-sample analysis will be closer to the ex-ante expectations, with the upside being the potential for long-term structural alpha.