Andrew J Dudley explains how Putnam’s dynamically allocated risk parity approach balances risk contributions
The ultimate goal of asset allocation strategies should be to improve investment outcomes with higher returns and reduce performance volatility. In other words, asset allocation investment strategies should achieve both higher returns and a greater efficiency of returns measured as higher returns per unit of risk. At Putnam, we believe that the asset allocation investment strategy defined as risk parity can be utilised to help reach these goals. What’s more, we think that actively managing allocations within the context of risk parity can achieve better return potential than a more static approach. This paper seeks to define the underpinnings of a risk parity concept relative to traditional balanced portfolios, reflects on what historical return data says about the strategy, and outlines Putnam’s approach.
What is a ‘balanced’ portfolio?
The purpose of traditional asset allocation is to achieve better returns with less risk by mixing asset classes with different risk premia and risk factors. Determining appropriate allocations begins with an analysis of the performance and risk derived from the raw risk premia of individual asset classes. A typical starting point for the effort to achieve portfolio ‘balance,’ for example, is the choice to move from a portfolio of 100% equity to a portfolio of 60% US equity and 40% US bonds in order to reduce the volatility associated with equities. Investors moving to this allocation do so with the knowledge of the trade-offs involved in trying to smooth out performance: Long-term returns will likely be lower, but an allocation to bonds will reduce the overall variability. Simply stated, because the long-term risk premium from bonds is less than equities, portfolio returns and variability will both decline.
We can examine the effort to find balance with a simplified example using two asset classes, stocks and bonds, as represented by US large-cap stocks and intermediate-term US Treasuries, through the prism of the past 40 years of return data. As expected, compared with a 100% equity portfolio, a 60% equity/40% bond portfolio reduces overall risk and returns.
Next, we can examine how the move from 100% equity to a 60/40 split affects the efficiency of returns. Despite having a lower absolute return, the new portfolio has an improved efficiency of returns, as measured by the Sharpe ratio, which is equal to the excess returns to a risk-free investment like Treasury bills divided by the volatility of those returns. There are two reasons for this outcome. During this period, bonds had a higher Sharpe ratio than stocks and so, all else being equal, the allocation to bonds produces a higher combined efficiency. But even if the Sharpe ratios of stocks and bonds had been the same, the low correlation between these asset classes created a diversification benefit that also contributed to the portfolio’s improved efficiency.
It is important to recognise that these factors alone do not fully explain portfolio risk. Investors should also consider the composition of portfolio risk, which is the breakdown in the sources of total risk. Defining risk in terms of its composition of variability reveals that the 60/40 portfolio actually behaves very similar to the all-equity portfolio. Namely, even at a 60% weight, more than 90% of the contribution to variability still comes from the equity portion of the portfolio. In short, while the 60/40 allocations reduced total risk, the risk contribution is still concentrated in equities. By this measure, risk has not been ‘balanced’ at all.
Investors need to consider risk contributions
The first major step in moving from traditional asset allocation towards a risk parity strategy is to re-define the allocation decision with a new metric. Instead of making portfolio allocations by percentage of capital, the investor calibrates the allocation in terms of contribution to portfolio risk. Although one can question the reliability of prospective asset class risk estimations, investors can still utilise the historical data as an initial guide.
What we find is that, in our simple example with two asset classes, reducing the equity risk contribution to 50% of the total during the past 40 years would have required a 25% stock/75% bond allocation.
Why choose a 50% equity risk contribution in this two-asset-class example? After all, in this particular 40-year period, bonds had a better Sharpe ratio than stocks (0.41 versus 0.34). If efficiency were the only consideration, it would argue in favour of allocating more of the risk contribution to bonds. The reason for 50/50 risk contributions is that the past 40 years might not be a representative period. In the fullness of time, we are not confident that bonds inherently are more efficient return instruments. We believe it is just as likely that Sharpe ratios are more similar across asset classes over long periods of time than structurally different. In other words, it is better to balance risk contributions than to concentrate them.
Establishing risk contributions with an approximately equal balance between stocks and bonds also further improves the efficiency of returns relative to both the all-equity and the traditional balanced portfolios. However, despite (or perhaps because of) the effort to balance risk contributions, this allocation still puts us at a lower expected return point. How can we make up the return difference?
The next step in building a risk parity strategy is allowing the use of a modest amount of leverage. Using leverage creates ‘equity-like’ alternatives to our equity exposure. Leverage allows the investor to generate greater performance from the balance of risk contributions in a way that either targets a specific return (if one has a view of prospective risk premiums) or a specific overall risk point (using historical risk estimates as a guide). In our simple two-asset case, if we substitute a greater proportion of bond risk for equity risk, it will require a greater US dollar exposure to bonds. With leverage, investors are not required to give up the potential return of the equity exposure when allocating to an asset class with lower absolute returns and lower volatility.
As an example, we can return to our original traditional balanced 60/40 portfolio to obtain our desired risk level. We then analyse performance data of the past 40 years to estimate the leverage necessary to build a risk parity portfolio with the same risk level. The result shows that the appropriate portfolio allocation is 42.5% stock/127.5% bonds, and would involve using a reasonably modest amount of leverage of 1.7x. With these parameters, we can build a portfolio that has the same overall risk as the 60/40 portfolio, but has a better return profile. We can also maintain the better Sharpe ratio when moving from the unlevered portfolio to the levered portfolio.
Other asset classes/risk factors
Of course, in the full application of risk parity, we draw from a broader set of asset classes. The full menu of options is more substantial and more diverse than our simple two-asset example. It includes foreign equity and debt markets, credit-sensitive assets, and inflation protection - or real return - assets. Adding these exposures creates further potential benefits from diversification across a larger set of less correlated choices. Incorporating these markets in a way that both accounts for their risk overlap and makes use of their differences can contribute additional return efficiency beyond the benefits of risk parity.
The Putnam Policy Portfolio
The outcome of this analysis is the Putnam Policy Portfolio, which defines the starting point for our risk parity strategy. Generally speaking, the Policy Portfolio maintains a balance in the risk contributions from equities versus the three other major risk factors: interest rate-sensitive fixed income, credit and inflation protection. The portfolio employs 1.5-to-1.0 (50%) leverage, which is slightly below the level suggested by the dual-asset example. This more conservative stance acknowledges the potential concern around dependence on significant leverage in the long term. The allocations lean in favour of asset classes that have performed strongly in periods of economic prosperity while reducing the overwhelming dependence on equities that characterises traditional balanced portfolios. Comparing the Policy Portfolio with the previous alternatives, we see a return profile that is better than the 60/40 portfolio, at a risk point that is lower than the traditional 60/40 balanced portfolio. Most importantly, it carries the highest Sharpe ratio of the portfolios considered.
Dynamic allocation is essential
However, Putnam’s investment process does not stop at the allocations and leverage points of the Policy Portfolio. We do not pretend that a set of returns from any arbitrarily chosen historical period necessarily provides an accurate measure of either expected risk premiums for specific asset classes or their future risk profiles. Stated a different way, we believe that a static application of a risk parity portfolio strategy that has performed well historically would be dangerous. Instead, we believe that ongoing management and dynamic asset allocation is necessary. We regard risk parity as an active and flexible process that must incorporate shorter-term valuation measures and dynamic estimates of statistical relationships used to adjust allocations towards more attractive exposures. As mentioned, we explore these and other issues in a separate paper titled ‘Risk disparity’.
There are operational considerations in the application of a risk parity strategy as well. The use of leverage raises the question of how to achieve market exposures in a way that balances the risks of leverage and the need for portfolio liquidity. We find that over-the-counter and exchange-traded index derivatives have evolved to become a liquid investment option for taking market risk. More than that, they have greatly simplified the task of achieving modest leverage. These derivatives have dramatically reduced the challenges of achieving leverage through traditional borrowing and lines of credit, which can come with unwanted risks. Although index derivatives do not entirely eliminate these concerns, the year 2008 was an interesting test for these instruments and for their application in a risk parity strategy. Many ‘unfunded’ derivative instruments, somewhat ironically, traded in a more liquid fashion than their cash market counterparts during the credit crisis and provided sufficient market depth during that challenging period. We are comfortable that such market instruments can provide substantial liquidity to the Policy Portfolio and make risk parity strategies a strong potential complement to other less liquid ‘alternative beta’ strategies. We intend to explore this and other issues in a forthcoming paper appraising the advantages and limitations of leverage and derivatives.
The continuing debate
Despite the successes of risk parity strategies, an industry dialogue continues in an effort to understand this success and the outlook for the strategy in changing market conditions. After all, strategies that allocated more to fixed income and other non-equity risk factors during the past 10 years — the ‘lost decade’ for equities - might not find the coming decade to be nearly so friendly.
We offer the following observations. First, there can be environments when risk parity strategies underperform - most notably, when equities outperform in a risk-adjusted fashion over other classes for a sustained period. As they did during the late 1990s, risk parity strategies as currently constituted would be hard pressed to keep pace. Second, we question the confidence that a number of practitioners place in the stability of risk premiums and risk factors over all time periods. From this confidence comes a commitment to more static allocations that equally weight three to four risk factors, and requires a higher degree of leverage, as much as 3x. We do not see a strong basis in evidence either for these equal weightings or for this consistent use of a higher level of leverage. We prefer a more flexible approach.
We see stronger evidence for approaching risk parity as an active and flexible process incorporating shorter-term valuation measures and analysis of dynamic statistical relationships. At Putnam, we try to strike the right balance in our design between exposure to equity risk, leverage, and the operational considerations of the strategy.
Andrew J Dudley is an investment director for Putnam Investments’ Global Asset Allocation team.