Monday, 20 February 2017
Last updated 2 days ago
Jul 16 2014 | 11:02am ET
By Peter Hecht, Ph. D.
Vice President, Senior Investment Strategist
Evanston Capital Management
Over the past few years, risk parity has become a component of most investors’ lexicons and, possibly, portfolios. Risk parity products have also become more common in many asset management firms’ offerings. However, even with all of the attention on risk parity there is still a significant amount of confusion surrounding it.
It is not uncommon for investors to view risk parity as a partial replacement for, or supplement to, a hedge fund/absolute return program. Let me be frank here: this makes absolutely no sense to me. It’s like comparing apples and oranges, as they say.
Perhaps part of the confusion is driven by the fact that some traditional hedge fund managers offer risk parity products. Nonetheless, the goals and risk/return properties for standard risk parity approaches are completely different than hedge fund or absolute return strategies. Standard risk parity is about delivering long only, static, risk-balanced exposure to various traditional betas without incorporating explicit expected return views. In my opinion, this approach sounds exactly like a public market strategic asset allocation solution.
In contrast, most hedge fund strategies aim to deliver an ample amount of idiosyncratic alpha and non-traditional beta in their return streams utilizing dynamic, proprietary expected return views. Since most hedge fund strategies have large idiosyncratic risk components, utilize short selling, and/or are partially hedged, they have a better chance of delivering more consistent returns across various market environments when compared to risk parity or other long only, public market portfolios. This is not a criticism of risk parity or long only, public market portfolios.
A long only, public market, traditional beta portfolio can only do so much from a risk/return perspective. Risk parity might be more consistent than a 60/40 stock/bond portfolio over the long run, but it can’t avoid the standard risk problems associated with long only, public market, traditional beta portfolios.
Risk parity has gained traction based on the belief that, when implemented, it can deliver a long-only portfolio with a higher risk-adjusted return over the long run, even without an explicit stance on expected returns. However, standard investment textbooks and classes emphasize the importance of both expected risk and return assumptions in the formation of an optimal portfolio. As a result, it has been natural for investors to ask about and seek to understand the set of assumptions in which risk parity coincides with the textbook optimal portfolio solution.
For this discussion, we will define optimal as a portfolio that maximizes expected excess return (in excess of the risk free rate) for a given level of volatility (risk). In other words, a portfolio is optimal if it has the highest Sharpe Ratio. In general, risk parity does NOT deliver an optimal portfolio with the highest Sharpe Ratio. Risk parity only coincides with an optimal portfolio if each asset has an identical Sharpe Ratio and each asset has identical pairwise correlations.
The identical Sharpe Ratio assumption makes intuitive sense. If each asset has the same return per unit risk (Sharpe Ratio), then it makes sense to allocate equal risk to each asset, although the second requirement, identical correlations, is still critical. Correlation is a measure of diversification benefit. If each asset has the same return per unit risk AND the same diversification benefit, then it is logical that the optimal portfolio would allocate equal risk to each asset. Many explanations of risk parity ignore the importance of the equal correlation assumption or incorrectly state that correlations must be zero, which is a special case of the equal correlation assumption.
Because risk parity only coincides with the optimal portfolio under a set of restrictive assumptions, when the prospective risk and return parameters are known, classic Markowitz mean-variance optimization will, in general, outperform risk parity on a risk-adjusted basis – a point that is missed or underappreciated in typical risk parity communications. However, in practice, risk and return parameters are not known. They must be estimated, and they are estimated with error. Unfortunately, it’s very difficult to estimate expected returns with any precision. Thus, using a mean-variance portfolio optimizer with “noisy” expected return inputs can lead to the classic “garbage in, garbage out” problem. In contrast to expected returns, public market volatilities generally can be estimated with more precision. As a result, people began to think of ways to form “reasonable” portfolios that relied on volatilities, not on explicit expected return assumptions. Risk parity does just that since it requires no expected return estimates. Additionally, although theory does not necessarily predict this, people felt comfortable believing that assets offer similar Sharpe Ratios over the long term.
This belief about long-run Sharpe Ratios gave people comfort that risk parity portfolios were somehow “quasi optimal” from a strategic asset allocation perspective. Naturally, people thought, “Why would I want more risk in one asset over another if all assets offer the same return per unit risk?” Note, however, as mentioned before, risk parity portfolios are not theoretically optimal unless correlations are identical too (diversification matters!). In sum, although theoretically inferior to mean-variance optimization, once estimation error/implementation risk is taken into consideration, risk parity is a reasonable “starting point” for an investor’s public market, strategic asset allocation. To the extent that an investor is confident in his or her expected return assumptions, the mean-variance framework still offers a superior way to incorporate that information into a model portfolio.
Risk parity approaches might lead to a different portfolio relative to the standard 60/40 stock/bond portfolio, but the underlying concepts are far from new. Classic 1950s mean-variance portfolio theory entails (1) identifying the highest Sharpe Ratio portfolio (also known as the “tangency portfolio”) and (2) using leverage (or T-bills) to hit a target return or risk level. The highest Sharpe Ratio portfolio is identified using a form of risk balancing in order to achieve the lowest possible volatility for a given level of expected excess return. In fact, by definition, the highest Sharpe Ratio portfolio has implemented risk balancing optimally. The highest Sharpe Ratio portfolio might produce different weights than risk parity, but, as discussed before, that is because the highest Sharpe Ratio portfolio will also take into consideration explicit expected return assumptions.
A natural question arises: if risk parity concepts are not new, how did we end up with the 60/40 stock/bond portfolio – a portfolio with approximately 90% of its risk in stocks? The answer is that many investors are unwilling to use leverage. Thus, if an investor has a high target expected return (e.g. 8%), risk balancing is sacrificed by having concentrated positions in higher returning asset classes (e.g. equities). Under the constraint of no leverage, the high target expected return could not be achieved without sacrificing risk balance.
Risk parity portfolios tend to have large dollar allocations to low risk assets, such as fixed income. Given the current historically low interest rate environment, this property of typical risk parity portfolios has been criticized heavily by many in the investment community. In my opinion, this critique of risk parity is misguided. As discussed before, risk parity is a strategic asset allocation solution that is “quasi optimal” when Sharpe Ratios are identical across assets. If one believes interest rates are going to unexpectedly rise, then the Sharpe Ratio for bonds is going to be negative or low relative to other asset classes. With this tactical view, one would hold fewer bonds than the strategic asset allocation implied by the risk parity portfolio. This in no way disproves risk parity. Again, think of risk parity as a “starting point” for one’s strategic asset allocation. If one has a tactical view (i.e. current risk/return view is different than the equal Sharpe Ratio assumption implicit in the strategic asset allocation), then adjust the risk parity portfolio weights accordingly. This applies to any strategic asset allocation approach, whether it is risk parity, constrained mean-variance optimization, or some other alternative.
Risk parity approaches to asset allocation have become a component of most practitioners’ investment toolboxes. However, the theoretical and practical uses as well as the advantages and disadvantages of risk parity relative to alternative approaches are commonly misunderstood. This short piece has tried to provide additional clarity on some of the more important risk parity topics.
Peter Hecht is a Vice President and Senior Investment Strategist at Evanston Capital Management, LLC. Prior to joining Evanston Capital Management, Mr. Hecht served in various portfolio manager and strategy roles for Allstate Corporation’s $35 billion property & casualty insurance portfolio and $4 billion pension plan. He also had the opportunity to chair Allstate’s Investment Strategy Committee, Global Strategy Team, and Performance Measurement Authority. Mr. Hecht also served as an Assistant Professor of Finance at Harvard Business School. His research and publications cover a variety of areas within finance, including behavioral and rational theories of asset pricing, liquidity, capital market efficiency, complex security valuation, credit risk, and asset allocation. Mr. Hecht previously served at investment banks J.P. Morgan and Hambrecht & Quist, and as a consultant for State Street Global Markets. Mr. Hecht has a bachelor’s degree in Economics and Engineering Sciences from Dartmouth College and an MBA and Ph.D. in Finance from the University of Chicago’s Booth School of Business.