Risk Parity: The truly balanced portfolio
Martin Steward spoke with Ray Dalio of Bridgewater Associates, the pioneer of alpha/beta separation and risk parity, about strategic diversified beta portfolios
Investors who want at least part of their asset management program to be an efficient, strategic exposure to global markets that requires minimal forecasting or tactical asset allocation ability face a knotty problem.
Finding two types of risk that respond differently to the two fundamental economic environments - growth and recession - is the easy part. Fixed income assets like bonds will do well when things slow down; growth assets like equities will do well when they heat up again. Achieving a balanced portfolio should be as simple as holding both and rebalancing regularly.
But achieving this balance isn't easy, of course, because bonds and equities exhibit different risk/return characteristics. Split your portfolio 50/50 and, while 50% of your asset allocation is in equities, those equities account for about 70% of your allocation to risk, because they are twice as volatile as bonds. 50/50 is really 70/30. Try to achieve a 50/50 risk allocation, and you end up with something more like a 35/65 portfolio. Which is great - except your gains will be paltry because bonds earn about half the long-term return of equities.
For some time two basic solutions to this problem have been proposed. The most widely implemented has simply extended the modern portfolio theory principle that underpins the initial equity/bond split (combining two assets with low correlation with one another can reduce risk more than it reduces return) into the equity side of the portfolio. By combining ‘diversified growth' asset classes, the theory suggests that an investor can maintain the return of a growth portfolio while bringing its risk closer to that of a bond portfolio. Reality, particularly the kind of reality we got in 2008, suggests that much of that diversification can evaporate from time to time; and during these times, not only is the risk not lower, it is usually higher because volatility rises as downside directional correlation increases.
The second basic solution has come to be known as a ‘risk parity' portfolio. As that name suggests, this also tries to bring the relative volatility of fixed income and growth assets into parity, but the key emphasis is less on diversifying the growth part than on leveraging the fixed income part (usually via bond futures). The last few years has seen a flurry of interest in risk parity, particularly among US institutional investors, but it was pioneered 16 years ago by Ray Dalio, president, CIO and founder of $120bn asset management giant Bridgewater Associates - and recent honouree in Time magazine's list of the top 100 most influential people in the world alongside the likes of Warren Buffet, Hillary Clinton and the Duchess of Cambridge. Founded in 1975 as a provider of economic research and advice, a fixed income and currency hedging manager for corporates and, later, institutional investors, by 1991 it had launched a hedge fund, Pure Alpha, which at $71bn is now one of the world's largest.
What is a hedge fund manager doing experimenting with dull stuff like strategic beta solutions? Well, Dalio pioneered risk parity in part because he had also pioneered ideas like alpha/beta separation, based on his conviction (now much more widely-accepted) that it is impossible to manage either one efficiently as long as they are being mixed together. Once you have decided that you should manage the two separately, it makes sense to make them both as efficient as possible on their own terms. Moreover, Dalio had a personal reason to develop an efficient strategic beta solution. "In the mid-90s I started to accumulate some money that I wanted to use to establish a family trust, and for that trust I wanted the right asset allocation mix," he recalls. "That's when I created the All Weather portfolio, which now accounts for virtually all of that family trust money."
When Dalio set out the All Weather process in an article a number of Bridgewater's clients decided they would like to allocate to the strategy, too. "They generally set up pilot programmes that represented 1-5% of their overall portfolios, which then increased through time as we monitored how the All Weather concept worked," he says. "Typically it has settled at 10-20%; in some cases it has evolved to 100%, where clients have gone on to implement it themselves."
So this is the essential idea behind risk parity: as Dalio puts it, once you have taken the steps to make all asset classes exhibit approximately the same risk, "you can begin to diversify for all economic environments without giving up expected returns". Or, to put it another way, your search for diversification need no longer be constrained by fear of its impact on your long-term returns.
The first thing to observe is that All Weather's approach to diversification differs from classic modern portfolio theory in that it is fundamental and qualitative rather than quantitative. All asset classes are priced according to what an investor would pay for the future cash flows upon which it is a claim, according to Dalio. "That's how a bond, a stock or a piece of real estate compete," he says, "so the most important driver of return is when the expectation of that income stream changes."
Given that, strategic diversification through the economic cycle is achieved through a balance between asset classes whose fundamentals are best suited to different parts of that cycle, defined as rising growth (good for equities, credit, commodities and emerging market debt); falling growth (good for nominal and inflation-linked bonds); rising inflation (inflation-linked bonds, commodities, EM debt); and falling inflation (nominal bonds and equities).
That has the advantage of recognising potential correlations between different asset classes - "any portfolio that contains corporate bonds and credit should put them in the same bucket as equities because they have the same environmental bias", as Dalio observes. On the other hand, it retains the assumption that pricing rarely dislocates from these fundamentals for long. But don't we have half a millennium of bubbles, manias and panics to prove otherwise?
"That is consistent with neither logic nor the evidence," Dalio insists. "In all of my time watching markets I have come to recognise that it is not at all easy to find mispricing - there are very few no-brainers. The market can get temporarily dislocated or out-of-whack for liquidity reasons and so on, but the essential proof of concept is the behaviour of the All Weather portfolio and the returns of each asset class, backtested all the way back to 1925. Imagine how much stress testing I must have done on this - it has virtually all of my money in it!".
But even if we accept that version of the efficient market hypothesis, we then have to consider the problem it poses to the other pillar of the risk parity strategy, because it requires us to believe that, while different economic environments will affect the strength and directionality of returns to asset classes differently, they will have no effect on their relative volatilities. That is crucial as it will determine the extent to which the leverage that we employ introduces a new, unwanted risk into the portfolio.
Consider what it is we are gearing-up in a risk parity portfolio: the volatility of assets with relatively low volatility. So our next question should be, ‘What causes one asset class to be more volatile than the next?'
Interestingly, Dalio offers two explanations. First - and as we have seen, this seems to be the main theoretical basis for All Weather - he makes a duration argument: "If the income stream of an asset is longer then we assume that it will have structurally higher volatility." In other words, because most bonds have a set maturity date but cash flows from equities are potentially perpetual, the risk (and therefore volatility) associated with equity cash flows is structurally higher.
"By borrowing cash, the first thing you do is raise the expected return of the item you are leveraging to a higher level than the item you are borrowing," Dalio explains. "The yield curve is normally upward-sloping, so bonds tend to yield about 2% more than cash over time, so when I borrow cash to lever bonds 1:1, I add another 2% yield by picking up the spread between what I'm borrowing and what I'm buying with the borrowed cash. It's an increased duration risk. So the question is simply, are these asset classes going to outperform cash? That's why we stress tested this portfolio through the Great Depression and Japan's depression: sure enough, it underperformed cash - but still radically outperformed the traditional 60/40 portfolio".
Risk parity makes the most sense if we believe that the duration is the key determinant of risk premiums, because under that assumption pricing across most asset classes would share a common delta (in the form of duration). As a result, one could expect the volatility regimes of different asset classes to remain proportionate and correlated through time, and this is important because it removes the big risk that leverage might otherwise introduce - the risk that the volatility of (say) bonds increases by a much greater proportion than that of (say) equities for a significant period. That holds even in the case of a severe spike in volatility - as long as it spikes proportionately across all asset classes.
But as an additional explanation Dalio points out that most higher-returning asset classes are already leveraged. "The average public company has a debt-to-equity ratio of 1:1," he observes. "If a law passed tomorrow that prevented companies from borrowing, the risk and return of equity would be less. That's where the equity risk premium comes from." If this holds any water, our conclusions about relative volatilities must be very different: first, we would expect equity volatility regimes to synchronise with the credit cycle, as corporations expand and contract their balance sheets; but more importantly, we might expect the volatility regimes of government bonds and equities to be negatively correlated, as public debt expands to fund automatic stablisers during recessions and contracts thanks to an increased tax-take during the good times.
Common sense would suggest that risk premiums are determined by both of these factors, along with a host of other economic, sentiment and behavioural inputs. But while Dalio acknowledges the importance of debt in determining risk premiums, and the folly of assuming that "the volatility of the recent past is representative of future volatility", the All Weather strategy is squarely based on "one assumed level of volatility" for each asset class, determined largely by its duration.
"Changes in the volatilities of different asset classes are significantly positively correlated," says Dalio. "That's good because it maintains the diversification benefits of the two assets alongside one another. The reason is that the same fundamentals are driving the volatility across all those markets: so in 2008, equity prices shifted from one level that discounted more income for the future to another level that discounted less income for the future, while the same influence was translated into bond price movements. We feel that the returns from these asset classes since 1925, and the performance of All Weather, suggests that these co-movements at different points in time are very reliable."
Figure 2 tests some of these assumptions using Bridgewater's simulated total return streams for nominal bonds and equities since 1970. Figure 2.a shows how variable the spread between equity and bond roling annualised monthly price volatility can be - and it is perhaps surprising how much of the line is below zero. Again, that's not necessarily a problem for risk parity as long as the correlation between the two volatilities remains positive. But figure 2.b also shows that rolling two-year correlation between annual equity and bond volatility can swing from almost perfectly positive to more than 70% negative, and that lengthening the sample period does almost nothing to reduce these extremes (correlation for the five years to February 2003 was -0.72, and to April 1994, +0.97). Correlation over the full 18 years comes in at +0.76; figure 2c shows the coefficient of determination at 0.07, and since the turn of the century that has only risen to 0.26.
As Dalio observes, such dislocation as there has been over eight decades has not been enough to de-rail the All Weather portfolio. We might observe that low correlation appears to have coincided with periods of low-volatility in equity markets, and high correlation with high volatility: that would account for the lack of impact that the dislocations have had on the portfolio. But the idea that the co-determination is stable is questionable, at least, raising the prospect of dislocations coinciding with periods of high equity volatility in the future.
Risk parity strategies have an intuitive plausibility and All Weather, in particular, has an impressive set of real and simulated numbers to confirm that intuition. Like most risk parity strategies, All Weather, which now managers $50bn, set out its stall very effectively during the turmoil of 2011: while US equities finished the year flat and a conventional 60/40 portfolio returned about 1.6%, the risk parity portfolio finished the year up just short of 19%.
But investors who consider this approach for their core beta portfolios might also consider the implications of what may be its internal contradiction: the idea that it is possible to diversify economic exposure to the credit cycle, and yet leverage parts of that exposure differently on the assumption that the credit cycle has no affect (or a uniform affect) on asset volatility. Fundamental diversification is famously the only free lunch in finance. One can leverage that free lunch by gearing-up the entirety of a fundamentally-diversified portfolio. There is no reason to think that gearing-up parts of that portfolio amounts to the same thing.