Risk Parity: Achieving targeted returns with an ‘All Weather’ asset allocation
Bob Prince and Paul Ross discuss the approach of Bridgewater Associates, founder of risk parity
All sophisticated investors understand the basic principle that a diversified portfolio is better than an undiversified portfolio. And yet, nearly all institutional investors hold an undiversified portfolio that is concentrated in a single asset class - equities. When we surveyed the strategic asset allocation of nearly 200 US pension funds we found that 98% of them were more than 75% correlated to a single equity index, the S&P 500. Given the sophistication of these investors and the degree of oversight of their activities, it seems obvious that this pervasive concentration of risk is intentional. Why would such capable investors, who know the benefits of diversification, intentionally hold an undiversified portfolio?
The traditional approach to asset allocation forces investors to trade off diversification for high returns. Most institutions are long-term investors and believe they can absorb short-term risk in exchange for higher long-term returns. And higher long-term returns reduce the cost of funding future liabilities. So most believe it to be economical to target high returns and bear the additional risk of having an undiversified portfolio that is concentrated in equities.
But there is a dangerous flaw in this thinking. Assets have short-term risk as well as long-term risk. Every asset class, including equities, will go through sustained periods of underperformance that can last for decades or more. And any portfolio that is concentrated in a single asset class will similarly go through sustained periods of underperformance. Such underperformance jeopardises the ability of investors to meet their obligations. Today's public pension funding crisis is a direct result of this flaw.
We believe there is a better approach to asset allocation. We believe that investors can have high returns and limit their risk through diversification by holding an ‘All Weather' risk parity portfolio. By holding a diversified portfolio that has the same expected return as the existing undiversified portfolio, both short-term risk and, more importantly, long-term risk will be reduced, and the cost of funding future liabilities will be the same or less.
In this article, we discuss why long-term risks exist in a concentrated portfolio. We describe the rationale and methodology for what we believe is a reliable alternative approach to strategic asset allocation, which produces the same high expected returns as the existing approach but with less risk, or higher returns with the same risk.
Since 1996, we have applied this All Weather risk parity approach through significant bull and bear markets in equities, two recessions, a real estate bubble, two periods of Fed tightening and Fed easing, a global financial crisis and periods of calm in between. Throughout these varied environments the All Weather asset allocation mix achieved approximately a 0.6 Sharpe ratio, consistent with its performance in simulated back tests through the Great Depression and across a variety of other countries. At the normal 10% targeted risk it has outperformed stocks, bonds and the conventional asset allocation portfolio, with much less risk.
Balancing your beta: the ‘All Weather' approach
Twenty years ago we asked ourselves the following simple question: What mix of assets will perform well across all economic environments? We knew we couldn't get to an answer through the traditional approach because that approach relies on correlation and volatility assumptions. Correlations are unstable and unpredictable, particularly when things go badly. Similarly, asset risk is difficult to predict and when things get bad, risks tend to spike higher. And most measures of risk do not adequately reflect the potential for sustained adverse environments that produce sustained poor returns.
So to answer our question we started with a blank slate, a state of ‘not knowing', and from there built up the most basic elements of asset pricing. This blank slate approach led us to recognise two fundamental characteristics of asset pricing which we think are universal truths we can count on to hold true in the future: (1) asset classes outperform cash over time; and (2) asset prices discount future economic scenarios. We believe that these two conditions form the primary basis of all asset pricing because they reflect the essential ingredients that investors require from an investment transaction.
Regarding the first, an investment is simply an exchange of money today for money tomorrow. When you make that exchange, you transfer liquidity from your pocket to someone else's, and you need to be compensated for that transfer because it carries risk (ie, giving up liquidity today creates the risk that you lose an opportunity to put that liquidity to work tomorrow). And generally, the more risk you take, the more compensation you require. More broadly, the existence of a risk premium is essential to the functioning of the capitalist system. Without an adequate risk premium capital would not be transferred, and the system would seize up. The amount of risk premium required changes over time, but in order for the system to function, the risk premium must reach a level that allows for the transfer of capital from those who have liquidity to those who need it.
Regarding the second, the pricing of an investment will reflect a discounted set of cash flows, and these cash flows will be impacted by future conditions such as the level of inflation, earnings growth, the probability of default and so forth. As conditions and expectations change, the pricing of assets change. For example, if inflation rises, expectations of the value of money tomorrow changes, and this change in conditions will be priced into the value of assets today.
Given these two structural elements of pricing, the returns of assets will be driven by how conditions unfold in relation to what was discounted and by how discounted conditions change, plus an accrual of the risk premium. From this fundamental understanding, we built up a new approach to asset allocation which we call the ‘All Weather' approach. We call it that because it's geared to weather whatever environment is thrown its way.
Key conceptual underpinning
Each principle above is connected to a key concept which we explore before we explain how to implement the approach in the next section.
Given the first principle, which states that ‘assets outperform cash over time', we are confident that we want to hold assets versus cash. Of course, we don't know which assets will outperform the others, so we want a balanced mix of assets. And the best mix of assets - the one that will most reliably capture the risk premium - is the one that will have the lowest asset cross-correlations. But unfortunately we don't know what the correlations between assets will be.
This brings us to the second principle, that ‘asset prices discount future economic scenarios', which reveals the fundamental driver of the variability in asset returns: how actual conditions transpire in relation to what was discounted and changes in these discounted conditions.
Because assets discount future economic conditions, and because equities and other investment assets have a long duration, their long-term risk is much higher than is normally captured by most measures of risk. Sustained underperformance in assets results from sustained shifts in actual conditions in relation to what is discounted. For example, in 1960 long-term bonds were discounting low inflation far into the future; instead, inflation trended higher for 20 years, causing long-term bonds to perform poorly during that period. In 1990 the Japanese stock market was discounting decades of strong growth. But instead Japan experienced a depression and 20 years of weak economic growth, which produced two decades of poor Japanese equity performance. As a result of the long duration of investment assets and the normal tendency for sustained shifts in the economic environment, the true long-term risk of an asset is many times larger than what is commonly understood.
A portfolio that is concentrated in a particular asset class runs the risk of sustained underperformance which jeopardises its ability to meet its future funding obligations. By understanding what causes these sustained periods of underperformance, you are able to balance the portfolio's exposures to achieve the required return more reliably over shorter and longer periods of time.
It is because assets have both long-term and short-term risks that you need diversification, even if you are a long-term investor. Instead of balancing assets based on unknown correlations, we believe the best asset allocation weights can be achieved by balancing the knowable drivers of asset prices. While there are many fundamental drivers of asset returns, we believe the most important are growth and inflation. And that a portfolio that is balanced to changes in discounted economic growth and inflation will capture nearly all of the potential diversification available to a strategic asset allocation mix, because these two conditions dominate the cash flows of asset classes and are therefore the primary drivers of variations in asset class returns. Other important factors, such as changes in credit risk and monetary policy, are derivative factors. For example, credit risk rises when cash flows fall because growth underperforms expectations. Similarly, while monetary policy has a clear impact on asset prices through its influence on the interest rate structure, it is driven by changes in growth and inflation (for example, easy monetary policy comes in response to weaker growth and/or lower inflation, all else being equal).
The best way to illustrate this concept is through a simple example. Consider the correlation of stocks and bonds in light of how they discount future economic conditions. Figure 1 shows how shifts in economic growth and inflation structurally influence the returns of stocks and bonds.
As shown, falling inflation is generally good for both stocks and bonds and vice versa. So if inflation was the only thing that mattered, you would think that stocks and bonds would be positively correlated. But economic growth is also an influence. Strong economic growth is good for stocks and bad for bonds and vice versa. So if economic growth was the only thing that mattered, you would think that stocks and bonds would be negatively correlated. Given this, what will be the future correlation of stocks and bonds? You really can't know without knowing the future economic environment, which is a problem if you are trying to build a portfolio for all environments.
Because of these fundamental connections, we believe that shifts in the relative volatility of economic growth and inflation cause understandable but not predictable shifts in the correlation between assets. For example, the dominant volatility of inflation and inflation expectations in the 1970s and 1980s led to a positive correlation between stocks and bonds. And in the 2000s the dominant volatility of economic growth in relation to inflation expectations led to a negative correlation between stocks and bonds. Given their structural pricing characteristics and the instability of the economic environment you really can't make a reliable assumption about the future correlation of stocks and bonds. Figure 2 shows the radically different correlation of stocks and bonds in the past two decades.
While this instability of asset cross-correlations would pose a problem if the approach relied on an assumption of the future correlation of assets, it is not a problem because the approach does not rely on such assumptions. Instead, by balancing the portfolio to the drivers of the volatility of returns, you could balance your allocation across lowly or negatively correlated asset returns without having to predict which assets those will be at any particular time. This happens through the natural cause-effect linkages of each asset to any given environment. The particular blend of lowly correlated (or negatively correlated) assets will shift over time based on whether changes in economic growth or inflation are the dominant influence on asset returns. You naturally get the most diversification where you most need it.
For example, if economic growth is the primary source of the volatility of market returns, then assets exposed to shifts in economic growth will tend to be the most volatile and the least correlated to one another. And the more volatile economic growth expectations are, the more negatively correlated those assets should be. And if changes in inflation and inflation expectations are the primary source of the volatility of returns, assets exposed to these shifts will tend to be the most volatile and the least correlated. And the more volatile inflation is, the more negatively correlated those assets will be. Through this process, you don't have to predict the future correlations of assets or the future environments in order to achieve diversification. It happens naturally by balancing our exposure to the fundamental drivers of asset returns. And this is true both for short and long time frames.
In short, the All Weather approach seeks to maximise the diversification of a portfolio by balancing the allocation of risk across structurally unrelated asset classes so that their environmental exposures offset one other, leaving the accrual of the risk premium as the dominant element of returns. This environmental balancing is the key to performing well across all environments and producing the highest risk-adjusted returns possible.
The All Weather balancing process occurs in two essential steps. First, you can increase and decrease the risk levels of all asset classes so that they have similar expected returns and risks. This provides you with many asset classes that have similar expected returns and risks. Second, you can balance these assets against one another so that the portfolio doesn't have any bias to perform well or poorly in different economic environments. This is accomplished by holding a similar risk exposure to assets that do well when (1) inflation rises, (2) inflation falls, (3) growth rises and (4) growth falls.
Because higher risk assets have higher returns and all risk assets are higher returning than cash, you can borrow or lend at the return of cash and adjust the risk of these assets to any level, and when you do, you also raise or lower their returns. Through leverage you can adjust the risk of any asset to any desired level, giving you the full range of asset choices at any level of risk. Figure 3a shows commonly held expectations of return and risk. Using these expectations, when you adjust these assets to a common risk, for example to the risk of stocks, the expected returns of the assets are all about the same, allowing you to choose among them on the basis of their diversification potential (figure 3b).
This approach can be illustrated through a simple example. Consider the case of stocks and bonds. Once you adjust bonds to have the same risk as stocks, they have had almost exactly the same returns over the past 40 years. But the asset classes made and lost money at different times. Why would you pick just one? It makes a lot more sense to have both in a balanced portfolio, and if you do, you experience returns that lie between the two. You get the same ultimate return much more smoothly.
Figure 4 shows this historical picture with relevant statistics. You can see that stocks and bonds diversified one another well in relation to shifts in economic growth. Stocks made most of their money when growth conditions were good, and bonds made most of their money when growth conditions were poor. But because they both benefited from falling inflation and were hurt by rising inflation, they both made most of their money when inflation was falling and did poorly when inflation was rising. They were not good diversifiers when changes in inflation expectations were a dominant influence.
So, balancing equal-risk stocks and bonds achieves a more reliable return pattern than either asset alone. But because both asset classes are hurt by rising inflation, to have a more fully balanced portfolio, you need to hold assets that do well when inflation rises - such as commodities and inflation-linked bonds. In order to achieve a more fully balanced portfolio, the portfolio allocates an equal amount of risk to assets that do well when growth is strong, growth is weak, inflation is rising and inflation is falling (figure 5). And in all cases we are referring to strong or weak in relation to what is discounted in prices. This balancing of risks is where the term ‘risk parity' comes from.
A balanced portfolio is not necessarily a low-returning portfolio. In fact, by raising the risk-adjusted returns of the portfolio you will generate higher returns at the same risk or less for the same return. Remember that a portfolio's return will roughly equal the average of the returns of its component assets. And the return of each asset can be adjusted to any reasonable level by borrowing or lending at the risk-free rate. So a balanced portfolio can be adjusted to any level of return that an investor desires. For example, a balanced portfolio can match the expected return of a portfolio that is 100% invested in equities, but will do so at a much lower level of risk. It is useful to compare a balanced All Weather portfolio with a 100% equity portfolio because many investors believe that a portfolio that is fully invested in equities would be their highest-returning portfolio but would have too much risk.
Figure 6a shows the cumulative return of a balanced portfolio and the cumulative return of a portfolio that is 100% invested in global equities. The balanced portfolio achieved the same return as equities with about one-third of the risk. The same returns were achieved with much smaller losing periods, and these losing periods passed relatively quickly rather than lasting for many years. And because today's conventional institutional portfolio is heavily weighted toward equities, the diversified All Weather asset mix produced the same return as the conventional asset mix with much less risk - around half. As mentioned, the benefits of the higher risk-adjusted returns can be realised through a lower level of risk at the same return or a higher level of return at the same risk.
In order to understand how much of the potential benefits of diversification you could capture through the All Weather approach, we compared the All Weather asset allocation return with a hypothetical ‘optimised portfolio' which had the benefit of perfect foreknowledge of correlations from 1970 to 2010. The historical Sharpe ratio of the All Weather asset mix was 0.7, very close to the 0.8 Sharpe ratio of the hypothetical optimised portfolio, which was achieved without making any correlation assumptions.
This indicates that nearly all of the asset class volatility and correlations over the past 40 years were driven by shifts in economic growth and inflation, and that asset returns corresponded in a logical way to these changes. By balancing the portfolio's exposure to economic growth and inflation, it derived a diversification benefit from holding offsetting amounts of whichever assets happened to be negatively correlated without having to predict which assets those would be. And this diversification advantage was not limited to a particular time frame. Longer-term risks derived from sustained shifts in the economic environment produced sustained outperformance in some assets which offset sustained underperformance in others, minimising the time span of losing periods.
We believe that our approach to asset allocation is based on reliable fundamental truths about asset pricing and is realistically grounded in ‘not knowing' what will happen. We are very pleased to see that this approach is now gaining in popularity through the growing field of risk parity. Though we are not comfortable with some approaches to risk parity (particularly those based on unreliable assumptions about the behaviour of returns), we think that the basic idea of balancing risks represented by risk parity is right. We believe that as this approach is increasingly adopted, it will have a radical, beneficial impact on institutions and the people they serve.
Bob Prince is co-CIO and Paul Ross is senior investment associate at Bridgewater Associates. Bridgewater Associates, LP advises certain private investment funds and institutional investors only.