In the primary article on this series, we discussed the necessity for clear communication within the early stages of the investment process. We began with purpose and goals as the premise for making fundamental decisions about investment strategy. In this second part, we discover the communication challenges related to traditional investment decision frameworks and risk concepts comparable to standard deviation.
So what’s improper with traditional investment decision frameworks?
Most large institutional investors hire advisors to assist the parties involved communicate and evaluate the trade-off between risk and return. Most use a mean-variance optimization (MVO) framework to assist investors make decisions.2 In an MVO framework, the goal return is the “mean” or return of a portfolio and the usual deviation is the “variance” or risk. MVO makes the investment strategy decision easy and stylish: each goal return corresponds to an “efficient portfolio” with a risk defined by a normal deviation.
However, standard deviation doesn’t characterize risk in a way that’s meaningful to most investors. It measures the upward and downward fluctuation of portfolio returns. But most investors don’t view rising portfolio values as a risk – they’re concerned with losing money. They often have a look at returns in absolute terms and are inclined to agree with the adage that you may’t eat up relative returns, i.e. returns in comparison with a benchmark. And although many investors realize that they will expect a decline in portfolio value, especially in crises of any kind, the most important risk of their eyes is avoiding what they consider is the utmost allowable loss, also generally known as risk capability or “loss”. becomes border.”
Only by likelihood would an investor’s loss limit ever equal the usual deviation of an MVO portfolio. The following graphic shows a mean-variance frontier with the best expected goal returns and corresponding standard deviations for 2 portfolios. For the general public foundation with a goal return of 6.75%, the usual deviation of the mean-variance efficient portfolio is roughly 13%. In practice, an advisor could convert a normal deviation of 13% right into a loss level that has a 5% probability of occurring, or about 1.65 standard deviations, which on this case is 15%. But what if the investor’s loss limit is 10%? What if it’s 25%? And what if 5% is just too high or the probability of losing 10% or 25% is just too low?
Mean-variance efficient portfolios
If the loss limit is 10% and a 5% probability of that loss is appropriate, the inspiration’s mean-variance efficient portfolio has a normal deviation of about 9.7% and a lower expected return of 6% (−10% = 6 % − 1.65). × 9.7%). This is a totally different portfolio. Without conversion for the investor, the probability of achieving 6.75% for this lower-risk portfolio is unknown. This makes compromises when using this framework difficult at best, especially for non-investment professionals.
In any case, it seems that standard deviation doesn’t fully describe realistic potential portfolio outcomes and the potential paths to those outcomes, so MVO excludes essential decision-making information. In particular, the potential for very large losses in the worth of the portfolio (tail risk), smaller sustained losses in the worth of the portfolio (sequence risk) and the exhaustion of the portfolio (depletion risk) over an investment horizon is ignored.
Tail risks come into play more often than MVO realizes.3 The following chart shows potential portfolio values (outcomes) under normal and realistic non-normal asset return assumptions for a $100 million private foundation portfolio with a goal return objective of 8.04%. The portfolio’s strategic asset allocation is 30% US stocks, 30% non-US stocks, 30% US bonds and 10% broadly diversified hedge funds. The five-year investment horizon results for each distribution assumptions reflect the Foundation’s strategic allocation and investment activities over the five-year horizon, including quarterly expenses, fees and asset realignment. The averages of the outcomes are shown by the vertical lines.
Distributions of portfolio outcomes, net of outflows and rebalancing
The differences in results are significant, especially relating to potential losses. Any decision that eliminates this potential for loss can lead to regret, forced selling, unexpected costs, lower than planned compound annual growth rates, and exhaustion.
The following table shows the everyday standard metrics used to explain portfolio risks for every resulting portfolio distribution. Decision makers face the challenge of interpreting these metrics. If we assume non-normality, is 14% too high a normal deviation? What confidence level is acceptable for Value at Risk (VaR)? In general, such standard metrics don’t convey sufficient meaning because they lack context – the precise information decision makers must make informed decisions about risk.
Standard metrics for investment risk
| Normal | Not normal | |
| Annualized standard deviation | 10% | 14% |
| Five-year value-at-risk (ninety fifth percentile) | 29% | 44% |
| Five-year conditional value in danger (ninety fifth percentile) | 33% | 51% |
| Average drawdown | 11% | 13% |
| Average maximum drawdown | 21% | 29% |
Given this discrepancy between standard metrics and investor context, institutions naturally prefer to offer only vague or no reference to risk of their investment policies. They offer statements comparable to the next: “Achieve 5% growth over the investment horizon plus inflation and costs,” “Maximize long-term returns consistent with prudent levels of risk,” “Achieve adequate returns at acceptable levels of risk,” or “ Beat the policy benchmark by 2% over rolling three-year periods.”
The bottom line is that an MVO approach has serious shortcomings relating to risk and standard metrics have little meaning. Most importantly, these metrics can result in poor investment decisions and cause regrets.
In the ultimate article on this series, we’ll explore an alternate approach to enabling decision making between competing goals.
Footnotes
1. is published by the Investments & Wealth Institute®.
2. The MVO framework determines the utmost expected return that corresponds to a given portfolio risk level. Typically, risk is defined because the volatility of a portfolio of assets. The framework relies on Harry Markowitz’ seminal work from 1952.
3. Financial market data exhibits abnormal behavior, including volatility clustering, autoregression, fat tails, skewness and asymmetric dependencies. For a summary of stylized facts describing price changes and their impact on securities, asset classes and portfolios, see “Many risks, one (optimal) portfolio, by Cristian Homescu.
If you enjoyed this post, do not forget to subscribe.
Photo credit: ©Getty Images / aluxum
