Risk Parity: Simple idea, complex questions
Brendan Maton asks what the track records of risk parity strategies can tell us about their suitability for new economic environments, and how they might fit into the thinking of European pension funds suitability for new economic environments, and how they might fit into the thinking of European pension funds
Here's a teaser for pension fund managers. Take three countries: Brazil, Japan and Australia. Select the best performer in government debt over the past 10 years.
On raw performance, the answer might be Brazil. But in the paradigm of risk parity, Japan comes top. Risk parity practitioners are a different breed: they prize stability above return because the latter can always be magnified by leverage - you just need to pick the right instruments. Thus 10-year JGBs returned only 2.7% over the last decade, but with a standard deviation of 4.5%, delivering a Sharpe ratio of 0.61. For Australia's bonds, the Sharpe ratio is just 0.29 over the same period.
This puzzle, posed by a couple of risk parity practitioners, is just one way into their unconventional world where portfolio allocations are measured by risk rather than capital value, and where total money weighting, or exposure, is typically 250-300%, and can rise above 1,000% - in the order of 20-times leverage - to specific sources of risk.
On first acquaintance, such leverage might dissuade the average pension fund trustee. "All risk parity products are leveraged," says Michael Mendelson, a principal and portfolio manager at AQR. Last year AQR's co-founder Cliff Asness co-authored a paper that attempted to instate leverage as a key component of optimal portfolios in modern portfolio theory (MPT). In MPT the tangency portfolio is the one with the highest possible realised Sharpe ratio. Asness, alongside co-authors Andrea Frazzini and Lasse Pedersen, claimed that risk-tolerant investors should use leverage (borrowing at the risk-free rate rather than investing at the risk-free rate) to invest more than 100% of their capital in the tangency portfolio.
It is worth noting that, in their back-tests which use market data to 1926, the tangency portfolio was 88% bonds and 12% equities.
Why doesn't this happen outside the world of risk parity? The authors note that in the capital asset pricing model (CAPM) the market portfolio, which is the value-weighted portfolio of all assets, must be equal to the tangency portfolio. But in the real world, where there is leverage aversion, the tangency portfolio and the market portfolio are different because the tangency portfolio holds more ‘safe' assets like bonds and less in ‘risky' assets like stocks. The market portfolio has a greater capital allocation (and a far greater risk allocation) to stocks, and stocks have not realised the significantly higher Sharpe ratios relative to lower risk assets that they would need to justify that allocation. In fact, the reality has gone the other way - stocks have actually had a slightly lower realised Sharpe ratio than bonds. The authors conclude that since some investors choose to overweight riskier assets to avoid leverage, the price of riskier assets is elevated or, equivalently, the expected return on riskier assets is reduced. In contrast, the safer assets are under-weighted by these investors, and therefore trade at relatively low prices.
There is clearly no emphasis here on forecasting returns, which tend to provide most investors with their first deep impression of any strategy or asset class. Risk parity practitioners, according to Mendelson, but also Ed Peters, co-head of global macro at fellow quant specialist First Quadrant, try to do something very useful but much easier: forecasting risk. Hence the appeal of Japanese bonds in risk-parity weighted bond portfolios.
This all seems pretty schematic: bonds and equities, rebalanced to have equal risk weighting rather than equal (or the traditional 60/40) capital weighting. But one way practitioners can be distinguished from one another is by the number of risk factors to which they have exposure. Most houses have upwards of 50 different elements ‘under the bonnet' which help spread risk not merely between asset classes but also across differing economic environments. These elements would not be entirely new or esoteric asset classes - not least because risk parity depends on abundant liquidity - but would typically include a clutch of sovereign bonds, equity indices and a wide spread of commodities.
One pension fund currently exploring the strategy said it would prefer a provider with multiple risk exposures rather than just a few, to help it preserve capital under stressful conditions when correlations can become extraordinarily high. The striking observation about this comment is that it suggests some risk parity practitioners are narrower than others.
It is not easy to probe further. In the absence of a developed system of classification by the major fund ratings agencies, this strategy is without a universally agreed definition. Such a classification is unlikely to materialise before longer track records are built and we see major asset inflows to pooled products and thus far, even the likes of State Street Global Advisors are running only segregated accounts. To make the job of comparison harder, practitioners employ a range of interventions to what is essentially a quantitative process. Houses such as Hamburg-based Aquila Capital promise their system is 100% systematic - the only interventions are in the improving the instruments used to gain exposure. Others such as First Quadrant do employ human override in extreme conditions.
At Bridgewater Associates, often cited as the originator of the risk parity portfolio, co-CIO Bob Prince categorises its flagship All Weather fund as passive, with static exposures to market betas that never have to be touched. He even jokes that the firm initially baulked at the idea of charging for the strategy, whose main client hitherto was a dependents' trust of the firm's founder, Ray Dalio. Maintaining those exposures, however, is not as easy as ‘passive' suggests. For Prince, the skill rests on understanding the movement of markets relative to two factors: economic growth and inflation. But that movement Bridgewater interprets in terms of risk. "All I care about is whether I can assess the risk," says Prince. He puts little faith in predicting asset correlations.
For AQR, a lot of the publicly disseminated validation of risk parity (like the paper by Asness) comes from simulations. Using a simple three-factor model of bonds, equities and commodities based on market data, AQR finds risk parity has delivered a superior Sharpe ratio than any of these asset classes in two of the last four decades and is second twice. Overall, the returns to the simple risk parity strategy have been consistently positive due to the broad diversification risk.
AQR has done further analysis month-by-month under scenarios such as rising inflation versus declining inflation and declining productivity versus rising productivity (see table). Although the simple risk parity strategy had a negative Sharpe ratio when analysed under a simple recessionary regime, it is no surprise to find that a plain 60/40 equity/bond allocation did, too.
Likewise, Peters at First Quadrant has taken its Essential Beta strategy back to 1988 and stripped out the contribution from leveraged bonds. Annualised over 20 years, return falls from 7.6% to 3.06% over LIBOR, which remains a lot more attractive than the return from a 60/40 equity/bond allocation. The smaller figure is earned from volatility management and commodities, which are the other factors in the Essential Beta strategy. Peters emphasises that these figures are derived from a combination of simulation and real track record.
Simulation is always taken with a pinch of salt by prospects. Peters valiantly defends the practice for quantitative strategies by pointing out that they are more suited to modelling than qualitative strategies. A stronger argument might be that those risk parity funds in the vanguard all have decent real records; they are merely shorter than the backtests.
The issue with real records, which are always more surprising than simulations, is precisely that they may contain the human interventions; but also, as we all know, that past performance is not necessarily a guide to future returns. Sceptics worry that, with sovereign debt yields currently at rock bottom, it is a bad time to be entering the strategy.
"Risk parity poses a tricky problem at the moment as it relies on leveraging bonds," explained one UK-based in-house manager selector. "Currently bonds have 0% yields and can only lose capital value from here. So, the worry is about the entry point currently for risk parity strategies."
Phil Tindall, a senior investment consultant at Towers Watson in London, acknowledges the ‘regret risk', but notes that that is not exclusive to risk parity. "Trustees sometimes do not give a new strategy enough time if it underperforms in the early years," he says.
This is the conclusion reached in one of the few independent studies on risk parity, by Greg Allen of the Callan Institute. Although it found a most satisfactory performance by a hypothetical risk parity strategy over the last two decades, numerous rolling five-year underperformances during the bull equity run of the 1990s led Allen to warn that "a successful investment policy must be designed to persist, not only through market cycles and unexpected investment outcomes, but also through turnover at the staff, board, consultant and the investment advisor levels".
In their own way, the practitioners all seek to allay fears that risk parity is not prepared for the coming decade when a secular bull run in highly-rated sovereign debt is all but impossible, except in the nightmare scenario of all-out depression (and perhaps not even then).
"If we had a swift and large hike in interest rates, that would be a shock," says Dan Farley, senior managing direct at State Street Global Advisors. "But if it was incremental, then exposures to diversifying holdings such as real assets would kick in and reduce the negative effects."
To be fair, the worst conditions for risk parity would not suit many other strategies either. If nothing else, at least risk parity's questioning of the preponderance of long-only equity and the myopic measurement of portfolio risk by capital allocation value fits snugly into a general reduction in equity risk exposure by pension plans.
PKA, the DKK153bn (€20bn) umbrella fund for Danish healthcare workers, mulled a move to risk parity when it started its recalculation of investments based on risk premia and other market ‘effects'. For now, the fund has restricted its evolution to the equities portion of its portfolio. Yet reducing volatility (through minimum variance and other strategies) is one of the major components in the new approach to equities, according to Michael Flycht, financial analyst at PKA. So there is a desire common with risk parity of smarter, more holistic risk management. Perhaps what is most different between PKA's project and risk parity is that the reduction in risk afforded by the recalibration of the Danish fund has gone not into leveraged bonds but absolute return strategies, says Flycht.
But then, PKA was an early mover into hedging out interest rate risk 12 years ago, which is a form of leveraged rates exposure: as such it has benefited sweetly from the latter half of the global bond bull market and evidently feels that it does not need to replicate risk parity across the entire fund when its liability-matching portfolio is so strong.
So there may be a connection between risk parity and total portfolio risk management but few are touting it, least of all risk parity practitioners.
"Risk parity is not part of a liability matching strategy but it is often used in the growth allocation of liability-matching funds," says Chris Palazzolo, AQR managing director. "While the logic might seem to dictate that risk parity is an alternative to traditional allocation, practical capacity constraints make it really just an addition."
Mendelson adds: "There aren't enough of the right assets in the world for it to be a broad based replacement." In the US it has been an equity proxy, says Farley at SSgA: "Funds like equity-like return with lower volatility and less likelihood of severe drawdowns." And Peters agrees that risk parity is "much better" than the typical diversified growth fund (DGF) for the return-seeking component of LDI: "DGFs still have circa 90% exposure to equities, versus about 57% in risk parity."
But will it catch on, even on those terms?
"I don't think the average trustee understands the leverage component," says Ciaran Mulligan, head of investment manager research and selection at consultancy, Buck Global Investment Advisers, in London. "We have concerns that some managers themselves new to this space aren't sufficiently familiar with leverage either. This is one of the top three risks for us in risk parity, especially given the connection between returns from fixed income and collateral management itself."
He said trustees might prefer to see leverage employed elsewhere, which comes on to another major risk for Mulligan: that risk parity funds fare badly in an environment where asset classes other than fixed income outperform.
Tindall and Mulligan both draw comfort from the longer track record of established risk parity practitioners. Tindall's observation on leverage is that it is new to most fiduciaries. Some may have permitted its discretionary use by a single-strategy hedge fund. Adopting risk parity, however, means the fiduciary board itself accepting far greater responsibility themselves because leverage is pretty much a given in the strategy in all its forms.
Generally, both consultants are attracted to risk parity. Tyndall goes so far as to say that Towers Watson does apply aspects of the theory for clients. "Allocating by risk rather than dollars makes more sense and fits in with our general risk management principles," he says before the adding the caveat that each client has different needs.
There is certainly room in the market for a plain vanilla risk parity fund. Mendelson reckons that in spite of all the theoretical work AQR has conducted with an equal risk-weighted, three-leg strategy, he does not know of a real product obeying the process. If one does launch, then at least risk parity will have a transparent standard from which practitioners and their prospects can contrast and compare.
Aquila Capital - risk reduction at 1,300%
Hamburg-based Aquila Capital's AC risk parity 12 Fund can include exposure in excess of 1,000% to short-term interest rates. Here, senior fund manager Armin Gudat explains its place in risk reduction to Brendan Maton
Let's start with a ‘traditional' way of looking at allocating a sum of money in equal weights to four asset classes (equities, bonds, commodities, short-term interest rates). Each asset class would get 25% of the money, but this does not tell us at all what the true risk contributions are. They are, of course, not 25% each as some asset classes are riskier than others. Indeed, the risk allocation with equal money as above is such that of this portfolio's risk, indicatively, about
• 50% come from the (25%) commodity allocation,
• 40% from the (25%) equity allocation,
• 8% from the (25%) bond allocation and
• 2% from the (25%) allocation to short-term interest rates
Secondly, as we want to achieve equal risk contributions, we have to change the money allocations. Such an approach would lead to a money allocation of about
• 3% to commodities
• 4% to equities
• 19 % to bonds and
• 75% to short-term interest rates
While we have now achieved a portfolio of equal risk, the overall volatility of such a portfolio is, naturally, very low. It is about 1%.
In order to construct a desired portfolio with an overall volatility of, say, 7% or 12%, we have to, thirdly, increase individual exposures, mainly to those asset classes that have the lowest individual risk contributions. This is done by using prudent leverage: we do not borrow money. Such ‘good' leverage is a normal part of our investment process.
Whilst initially a high percentage of exposure to short-term interest rates (through futures, in percentage of the portfolio) might lead to an impression of high riskiness of the portfolio, looking at the true dynamics shows that this is not the case.Since inception in 2004, the AC Risk Parity 12 fund's worst month has been -5.93% - well within stated maximum loss constraint of 7%. Annualised return has been 10.62%; Sharpe ratio has been 0.93.