Risk Parity: Nice idea, awkward reality
The tweaking and adjustments managers force upon ‘risk parity' strategies betray the risks at the heart of the concept, writes Joseph Mariathasan
Whether they have used a mean variance optimisation approach or some other methodology, many European institutional investors would regard their asset allocation strategies over the past decade as a failure, primarily because of the large weightings to equities. During the 10 years to 30 September 2010, the annualised return (in euros) on the MSCI World was -1%. Moreover, in a typical UK- or US-style 60/40 equity bond mix, over 90% of whole portfolio volatility over the past 40 years or so would have been due to the equity component. Given the disappointing performance of equities, and the historical overweighting to them, institutions are grappling with the idea that there could be better alternatives to variations on 60/40.
That is where risk parity comes in. An intuitively attractive strategy, it has gained attention in the US and is now spreading to Europe. One reason is the fact that the traditional mean-variance optimisation process for asset allocation famously outlined by Nobel prize winner Harry Markowitz has always looked far better in theory than in practice. The efficient frontier of portfolios is produced by finding portfolios with the maximum return for a given level of risk, based on expected returns of each asset class, and the historical volatilities of and correlations between them.
The problem - as anyone who has tried to adopt this approach has found - is that the portfolios are very sensitive to exactly the information that is least reliable, namely the expected returns. Indeed, the sensitivity can be an order of magnitude greater than the sensitivity to changes in volatilities and correlations. This means that many asset classes may be ignored because their expected returns may be just a few basis points less than others, even though any sensible error margin would completely swamp such differences. One remedy has been to impose constraints (on maximum and minimum weightings, for example). But then portfolios end up being more a function of the constraints than the optimisation.
By contrast, a ‘classic' benchmark risk parity portfolio makes no assumptions about future returns - although it does assume that all assets included have the same Sharpe ratio (that is, the same ratio of expected excess return to volatility). As such, a risk parity portfolio is not a risk-minimising portfolio and can lie below the efficient frontier produced by mean-variance optimisation. The way risk parity portfolios achieve higher returns is through leveraging their exposures to asset classes with lower volatility.
Paul Goldwhite, director of research at First Quadrant, attributes the increasing interest in risk parity to supply, demand and innovation. "In terms of the supply of investments, there has been a decade of disappointing asset class returns," he says. "Active management has been disappointing, volatility has been higher than expected and diversification also disappointed as correlations turned out to be higher than expected. In terms of demand, the pressures are greater for both private and public pension schemes facing shortfalls. Finally, a wave of innovation has made risk parity approaches more acceptable: greater use of derivatives, shorting and investment in hedge funds has made it more acceptable to move away from conventional strategies."
Still the classical approach, with its agnosticism about expected returns, might limit the attractiveness of the approach, particularly in Europe. "In the US, there is more scepticism on active management, whereas in Europe it is very difficult for investors to believe in a process that is completely agnostic with respect to asset class returns," says Jean-Louis Nakamura, CIO of asset allocation for Lombard Odier. Moreover, portfolios will invariably have much higher weightings to bonds and cash-type instruments, which also raises issues, according to Nakamura.
Nonetheless, Lombard Odier is using an asset allocation strategy that employs risk parity concepts. And in Germany, CIO Harold Heuschmidt says that since 2008 Aquila Capital has attracted significant inflows to what it calls a risk parity fund, launched in 2004 and based on equalising the volatility components in a portfolio of just four asset classes, namely equities, government bonds, short-term interest rates and commodities. Its great attraction has been that its returns are robust and uncorrelated with equity markets. This is not surprising when the actual weightings are compared with the risk profile. As figure 1 shows, weightings to equities are much less than 10%, with nearly three quarters in short-term interest rates. As one can imagine, the return of such a portfolio would be low in the absence of leverage, and this is a key feature of risk parity approaches.
Ben Inker of GMO has questioned the usefulness of risk parity, pointing to three weaknesses. First is the identification of volatility with risk, a failing found in all value-at-risk approaches based on modern portfolio theory. Historical values of volatility and correlations are dependent on the time periods chosen and tend to ignore events that have not occurred frequently or recently. Second, having no view on future returns means that asset classes can be included that may have zero or even negative risk premiums. Commodities are a case in point; government bonds are potentially another example - Inker, along with many others, believes that the risk premium on bonds may be negative for an inconveniently long time. Third, and related to the incomplete view of risk represented by volatility alone, several asset classes exhibit negative skew - negative returns tending to be larger than positive returns. This is particularly true of credit: one can lose all of an investment, but the upside is limited to the coupons and return of principal.
When these three flaws are combined with leverage, Inker argues, it will give a result that probably looks well-behaved until the moment when it suddenly doesn't. The problem with leveraged strategies that are marked to market is that an unleveraged investor can generally afford to wait for prices to converge towards economic reality, but a leveraged investor may not have that luxury. This was indeed the case during the market crash, when hedge funds were forced to liquidate holdings on the basis of marked-to-market losses caused by a market panic, as against real defaults on their bond cashflows. Historically, the evidence appears to be that a leveraged risk parity portfolio would have significantly underperformed during the 1990s and significantly outperformed during the past decade.
"If equities had delivered 10% every year, risk parity would not have appeared," declares Goldwhite. It is worth bearing this in mind when trying to understand what role it might play within European institutional portfolios. Risk parity has some appealing aspects, but it encompasses a number of different approaches. There needs to be a distinction made between using a pure risk parity strategy to invest a complete portfolio, using risk parity as a benchmark for an active process and using risk parity funds as an additional component to a multi-strategy portfolio.
Most fund managers offering risk parity products are invariably combining a passive benchmark with an active overlay. Aquila's fund for example, despite its name, has had a pure risk parity portfolio for only 10% of its six and a half years, according to Heuschmidt. At the moment by far its most significant position is the nearly 75% of capital it has in three-month Euribor and Eurodollar interest rate futures - a position that generates a terrific carry as long as money market yield curves remain as steep as they are. Should interest rates rise to the extent that the curve inverts, it would start incurring costs. But, as Heuschmidt admits, the strategy's active overlay could then cut in and suggest alternative assets.
Lombard Odier also explicitly combines a risk parity benchmark with a tactical active asset allocation strategy. Its approach differs in using rolling nine-month historical volatility and correlation figures, rather than Aquila's figures based on 10 years or more of data. Nakamura's argument for such a short period is that it enables progressive changes to the portfolio as risks change - although, as he admits, the strategy cannot really adjust to sudden shocks. So, while it lags the marketplace, it nonetheless adjusts risks in a significant way rather than simply taking risk as a fixed figure.
Nakamura believes this approach can be applied to individual asset classes with very heterogeneous components such as the commodity markets and emerging market equities, and Lombard Odier is looking to launch funds on this basis.
The active risk parity strategies being marketed are seen by consultants as alternatives to diversified growth strategies, according to Goldwhite. They can look attractive when that comparison is made. "A typical diversified growth strategy has lots of asset classes but they tend to be highly correlated to equity markets," he says.
One area where they may have a role is as the default option for DC pension schemes, and this has been the background behind First Quadrant's launch of a pooled fund in the UK which aims to improve the risk/return profile of currently available default fund options by diversifying risk more equally across global developed and emerging equity markets, global sovereign bonds, US inflation-linked bonds, commodities and real estate. The approach has proved successful in the US 401(k) market.
Goldwhite describes three components. Global equities are accessed using futures for developed markets and ETFs for emerging. Individual countries are equally risk-weighted. The same applies with country exposures in the global sovereign bonds section. The third component consists of assets that aim to provide inflation hedges - index-linked bonds and commodities. On top of this benchmark allocation, First Quadrant adjusts the exposure to risky assets based on a dynamic assessment of risk that uses variables such as the VIX index (tracking the implied volatility of S&P 500 options), credit spreads and macro-economic indicators. Volatility is assessed on an even shorter time frame than at Lombard Odier, with figures averaged over one to three months.
Clearly, what are described as risk parity strategies can vary tremendously. Sometimes there can be only a loose connection to the idea of using a reference benchmark with equally weighted risk allocations. Even similar strategies can differ considerably because they may allocate to a different set of asset classes. Deciding whether US equities are a separate asset class to EAFE or whether both should be grouped together as global equities can end up the chief determinant as to whether the allocation to global equities is 5% or 10%.
Risk parity may have an intuitive attraction and a pragmatic utilisation of the concept may have value as a benchmark - but there is no fundamental rationale for it. As a result, investors wedded to liability-driven investment will see no benefit to another approach also advocating high fixed income exposures, while equity-focused investors can argue that returns to equity come predominantly from dividends, so that even if equity prices do not increase substantially, real yields represent at least an approximate match for pension-type liabilities.
Inker perhaps sums up the real problem with risk parity that will limit its appeal in Europe: "Whilst investors need to take advantage of risk premiums if they are going to have any hope of meeting the targets they have set for themselves, those risk premiums can neither be assumed into existence nor counted on to continue because they were there in some historical backtest." As such, he argues that no particular fixed weight benchmark is a good solution for all time or all environments. "Risk parity portfolios are no exception."
With a few notable exceptions, most of the fund managers offering ‘risk parity' products would probably agree. Indeed, that is precisely why those products so often employ some kind of active overlay. Ultimately, that should make us ask whether ‘risk parity' is just the latest marketing slogan to be applied to strategies that, one way or another, we have seen before.