For decades, the gold standard for portfolio construction has been rooted in mean-variance optimization and modern portfolio theory (MPT).
Starting with a few asset class inputs (mean historical return, standard deviation and correlation), this process produces a series of (allegedly) efficient portfolios that maximize expected return at a given risk level.
Like any widely-accepted standard, this approach has many benefits. It’s familiar, widely-taught and scalable. And when it comes to marketing firepower, what’s better than attaching your firm’s approach to a Nobel Laureate? (More on this later.)
To put it mildly, traditional portfolio construction has been challenged many times in the past 50 years. Financial markets have experienced drawdowns of a magnitude and frequency that MPT-based models would not have expected. And diversification has shown its tendency to vanish when it’s needed the most.
Even Dr. Markowitz Agrees. During the chaos of 2008, traditional diversification methods and modern portfolio theory became everyone’s favorite scapegoat. Everything we were told about what to expect from our portfolios seemed to have been turned on its head.
Dr. Harry Markowitz (Nobel Prize-winning father of MPT) addressed these reactions in a number of public forums, including an interview with Research Magazine in August, 2011 where he said:
“We believe that a financial advisor will use the risk-return tradeoff curve of modern portfolio theory more effectively if the advisor knows, in general, the assumptions behind MPT…”
“While it is true that the well-advised investor may encounter unanticipated hardships, the ill-advised investor courts almost certain disaster.”
So What’s the Problem?
This approach of building “efficient” portfolios falls short in three ways:
1. Simplifying Assumptions
Asset allocation models and traditional measures of risk (like standard deviation) are rooted in many false assumptions.
They treat upside and downside risk equally which is completely divorced from investor reality and causes an over-approximation of the incremental reward for additional risk.
They also assume that security returns and volatility follow constant probability distributions, meaning that there is always an equal likelihood of a positive or negative returns, regardless of recent history. In fact, security returns evidence serial correlation, in that positive (negative) returns are more likely to follow positive (negative) returns.
But the most egregious (and surprisingly, the most well-known) flaw in MPT models is the assumption that security returns follow a normal distribution. By now, I think we all know that this is not the case at all.
2. Reliance on History
When models rely on historical inputs and assumptions, their outputs are 100% contingent on one thing being true—that the future will look like the past.
And herein lies the problem. Financial markets are complex systems. Advances in mathematics and logical analysis now allow us to appreciate that:
Markets are often unpredictable and subject to large events.
They respond to both bottom up and top down influences in a non-additive form.
Markets produce amazing novelty: some good & some bad.
Complex systems do not behave in a predictable, stable manner. MPT models assume that asset classes will perform at their average long term returns (ignoring the path of returns) and that the relationships among asset classes will remain historically accurate. Neither are true.
In the words of Dr. Markowitz, “…the inputs to an MPT analysis are not supposed to be historical average returns, volatilities and correlations. Rather, they should be forward-looking estimates.” (Research Magazine, August 2011)
As they say, history doesn’t repeat itself but it often rhymes.
3. Arbitrary Constraints
Anyone that has gone through a mean-variance optimization exercise can attest that the outputs can sometimes appear odd, often suggesting an allocation that appears anything but “diversified.”
Optimizers commonly allocate too heavily to high return asset classes (e.g. emerging markets) or low volatility assets (e.g. real estate) And based on the average annual returns, standard deviation and correlations input, these would have been the optimal weighting if you could go back 30 years in time.
But no advisor will recommend that clients allocate 40% to emerging markets or 80% in real estate. Instead, practitioners apply arbitrary asset class constraints to produce a result that’s palatable to clients. Is the result an “efficient” portfolio? Possibly, but most likely it is not.
These points should not be dismissed as benign or academic. They are meaningful. Simplifying assumptions, unreliability of historical data and arbitrary constraints result in an underestimation of risk and results in portfolios that are at worst inefficient, with expected results that are surely unknown without hindsight.
An Alternative Premise?
Of course, diversification is critically important but it’s not the destination. Maybe portfolio efficiency is the wrong goal. Maybe instead of efficiency we should be aiming for resiliency.
In a future article we will address this notion of portfolio resiliency and share our take on why we believe it to be a far superior destination for asset allocators.
NOTE: The information and investment results stated herein are hypothetical and dependent upon the assumptions and conditions indicated, which may be superseded by actual market events or other unforeseen and unforeseeable intervening causes. The transactions proposed herein are informational only and do not constitute an offer to purchase or sell securities.
This information is not intended to, and does not relate specifically to any investment strategy or product that Arin Risk Advisors, LLC offers. The information contained in this blog is provided for informational purposes only to assist an investor’s own analysis and an investor’s own view on the topic discussed herein. Past performance is not a guarantee of future results.