The CFA Institute recently published a groundbreaking paper — “Ensemble Active Management (EAM) – Taming Toxic Tails” — that is gaining widespread attention among institutional CIOs because it introduces a new approach to active management that reflects an insightful integration of Machine Learning, a subset of artificial intelligence and active investment management.
But as the authors state, “EAM is not a strategy, algorithm, or overlay.” It is not another attempt to use AI to screen 10,000 stock variables a day to improve stock picking. Rather, it is a new approach to portfolio construction, and appears poised to transform the entire active investment management industry.
Why should institutional CIOs care? Because the need to improve investment returns without taking on higher, or exotic, risk has never been greater. A study by Milliman from October of 2017 showed that the aggregate Funded Ratio of public pension plans as of June 30, 2017 was estimated to be only 70.7 percent, and this represented a “sharp rebound” from 67.7 percent in 2016 (Source: Rebecca A. Sielman, Milliman, “2017 Public Pension Funding Study”, October 2017). As we know, there are only two ways to improve the funded status of a plan: improve returns or increase contributions. Since the latter option is such a terrible outcome, any fresh ideas to improve returns is a welcome conversation to have with plan sponsors and their advisors.
In a cogent and concise manner, the authors of the CFA Paper provide insight into a pervasive portfolio design conflict that has bedeviled investment managers for decades. Risk management protocols, which are necessary in today's environment, have impaired the industry from delivering on its mandate to outperform the market after fees. Amazingly, this design conflict has hidden in the blind spot of the industry for years.
The 4‐point message of the paper is simple and intuitive:
Point One: Portfolio managers create their potential for outperformance through their high conviction best ideas. These are the 15‐25 stocks that the manager has identified as having the best opportunity for outperformance. Managers allocate a portion of their portfolios to these top holdings while also allocating meaningful benchmark positions as a risk management tool. The article refers to these top holdings as the “Alpha Engine.”
Point Two: The very act of selecting stocks with the intent to outperform creates a risk factor that the manager is required to address: the potential of relative underperformance. A severe relative underperformance event is referred to as a “Toxic Tail” event in the article. Toxic Tails put both the end investor and the manager in harm’s way, and need to be successfully eliminated.
Point Three: The industry’s standard risk management tool to eliminate Toxic Tails is to build out a large number of stocks that are not intended to drive outperformance, but rather for relative tail risk management. Research shows that the average number of holdings for U.S. equity portfolios is close to 100, implying that these risk management positions are roughly 75 stocks (the holdings in excess of the high conviction best ideas). They are referred to as the “Beta Anchor.” Research also shows that the Beta Anchor is typically 60–75 percent of an active portfolio.
While the Beta Anchor is a highly effective risk management tool for Toxic Tails, it also comes with a performance penalty. As the Beta Anchor grows, it squeezes down the portion of the portfolio available for the Alpha Engine. For example, a Beta Anchor at 75 percent of a portfolio means that Alpha Engine is only 25 percent of the portfolio, and therefore three‐quarters of the alpha potential of the PM’s top stock selections have been permanently lost.
Point Four: Ensemble Active Management does what the industry could have targeted years ago — function as a replacement risk management tool for Toxic Tails. It potentially eliminates the risk of Toxic Tails on par with the Beta Anchor, but does not carry the performance penalty of the Beta Anchor. In fact, the research and theory in the CFA article suggests that it actually enhances performance.