Modern Portfolio Theory Needs Updating, Badly
Modern Portfolio Theory has informed the science of investment management since its introduction over 60 years ago by Harry Markowitz.
Why? Well, for the most part, it works.
But in the market crash of 2008, MPT’s fault lines were dramatically exposed. It has since become clear that the intentionally simplistic assumptions of MPT — so handy before high-speed computing — don’t reflect the reality of today’s capital markets. Specifically, more sophisticated and realistic inputs to the three principal assumptions in MPT — returns, risk and relationships — are needed to make continuing use of the theory.
“Quiet-” and “Turbulent-Time” Returns
MPT requires that we forecast the future returns for the asset classes we are considering as part of a portfolio. But historical averages for this purpose can be misleading. A broad average of the returns in past periods simply lumps together data reflecting very different market and economic environments. The result is that the all-inclusive average does not accurately represent any particular scenario.
Additionally, returns are not static. They are dynamic and influenced by current asset valuations, stage of the economic cycle, geopolitical balance and other factors. The key is to identify useful patterns in historical returns. We recommend parsing market data into "quiet" and "turbulent" regimes to better understand how assets perform under different environments.
In order to do so effectively, it is important to develop methods for distinguishing between regimes and assessing which regime the markets are currently in or heading into. Developing a dashboard of market and economic data can be the cornerstone of this effort.
Look at Risks Through Clients’ Eyes
Risk has long been measured as standard deviation or the amount of dispersion around the average return. We believe it makes more sense to look at risk the way our clients typically do, as the chance of significant loss, or of not meeting a financial objective that matters most to them. We recommend that various metrics be used to capture this view, including “shortfall probability” — the probability that the return will be below a certain user-defined target.
To fully capture the dimensions of truly catastrophic risk, however, the investor wants to know not just the chances that the return will be below a certain threshold, but also how bad things may get in that case. One way of calculating this is to use “conditional value at risk.” Long a staple of the actuarial profession, CVaR can be a key metric in constructing portfolios that seek to avoid catastrophic losses.
Relationships via Copulas, Not Correlations
MPT calls for analyzing the relationship between asset classes using correlation. This measure is simple, but no longer relevant to portfolio design.
A key flaw? MPT assumes that the relationship between two assets is linear and the same under all circumstances. This is clearly not the case, since correlations change over time, especially in turbulent periods. Asset classes that may have weak correlation during normal markets tend to have strong correlation in highly stressed markets — when you are especially eager for them not to correlate.
The relationship between any two assets alters depending on how the assets are performing. Instead of a single number (such as correlation) to capture this relationship, it makes sense to do so using copulas rather than correlations. A copula allows you to describe the dynamic relationship between two assets across their full range of behavior, including how assets behave relative to one another in different markets. This can be a powerful tool in designing a portfolio that is less vulnerable to contagion.
A modernized MPT is one step toward bringing investment risk management into the 21st century. Modernized MPT, along with dynamic asset allocation and enlightened tail-risk hedging, could form the framework for the next generation of investment-risk management.
The market turmoil we’ve seen recently makes updating MPT an imperative for the industry. If we don’t upgrade our tools, clients who have forgiven us for exposing them to the contagion of 2008 may not be so generous the next time.