Risk Parity: Risk parity for a single asset class
Pure risk parity just about works in a multi-asset class context. For individual asset classes, Joseph Mariathasan finds that it needs to be constrained and adapted - but still offers a useful corrective to the biases of the cap-weighted portfolio
Risk parity has gained some popularity as a strategy for multi-asset class portfolios that, in theory, allows investors to remain agnostic about expected returns by instead focusing on positioning to weather perhaps very different future economic scenarios.
For example, Lombard Odier Investment Managers' CIO for asset allocation, Jean-Louis Nakamura, describes three scenarios under which its strategy is expected to perform: growth without inflation (which requires equities); deflation or disinflation (government bonds); and an unexpected inflationary shock (where commodities and inflation linked bonds are the key asset classes). The textbook risk parity approach takes the asset classes best-suited to each of these scenarios (or a variation on them), and weights them so that they each contribute equally to the portfolio's overall volatility.
But Nakamura argues that it is not wise simply to equally risk weight equities, bonds and commodities to produce a well balanced portfolio - because as well as volatility, the portfolio also carries the risk of changing correlations between each asset class. Since 2008, commodities have shown increasing correlation with equities, he observes: "If [the asset classes] were given equal risk weighting, then 67% of the portfolio risk would be equity related." As a result, the portfolio is 40% risk weighted to equities, 40% to bonds and only 20% to commodities.
Lombard Odier's need for adjustments to a pure risk parity split in its multi-strategy fund reflects the fact that the fundamental requirement for success in any risk parity approach is to be able to diversify across investments that have low and ideally negative correlations with each other. Does risk parity therefore make any sense for investment within a single asset class? The answer appears to be a function of how little correlation there is within an asset class and also the liquidity of individual components.
"Most commodities don't have a relationship with each other at all so they have very low correlations and you get a big benefit using a risk parity approach," says Ed Peters, co-director of global macro at First Quadrant. "In contrast, most commodity indices are production-weighted, which means they are heavily skewed towards energy, for example, and not gold."
In the case of bonds, a capitalisation-weighted approach to portfolio construction can be a recipe for disaster and has no inherent value beyond convenience. "A risk weighted approach is better than cap weightings for bonds," Peters concludes.
Applying risk parity within equity portfolios is certainly a possibility considered almost from the inception of the methodology. Bridgewater Associates CIO Bob Prince actually addressed the question for a large public pension fund in 2003. But as Prince explained then: "The big theme that came from our work is that there is a lot of common, systematic risk across all sectors of the US equity market, no matter how you cut it. The diversification potential within domestic equities is limited. As a result, the potential to create a portfolio that produces much more consistent performance than a simple cap-weighted index is limited."
The US, like any single domestic market, can be cut in three different ways: small companies versus large; value versus growth; and industry by industry. Bridgewater found that you have the greatest diversification potential when you break down the market by industry.
"This gives you more pieces to work with," Prince explains. "It also allows for tighter logical connections between economic conditions and equity performance than we find with large versus small and value versus growth. Furthermore, a significant amount of the difference between large cap and small cap is attributable to differences in the sector composition of these indices."
First Quadrant has just launched a global equity portfolio based on risk weighting across global sectors. But within each sector, the correlations between stocks may be too high for risk parity approaches to add much value. "Once correlations get above 85%, the benefits of risk parity approaches get smaller," explains Peters. Instead, the approach is active: "We are betting within sectors and also taking country bets and large versus small company bets. It is only in the sector weightings that we use risk weighting."
Adopting a sector approach to risk weighting in developed markets can make sense. Lombard Odier combines 10 sector groups and three regions (Asia, Europe and North America) to produce 30 ‘clusters' that are relatively uncorrelated with each other. Global or regional sector weights are not so relevant for an emerging market equity portfolio, as emerging market equities are characterised by having high correlations to their country indices as against their sectors (in contrast to developed markets which usually exhibit higher sector correlations). Lombard Odier's approach to emerging market equities recognises countries as the key diversifier by categorising the universe into 21 countries.
"To achieve theoretical diversification each country is allocated an equal amount of risk," explains Nakamura. "How much capital we allocate to each country depends on its risk (volatility) and its diversifying features (correlation). As a result of risk parity, we allocate more to good diversifiers and less to poor diversifiers."
As Peters at First Quadrant points out, the MSCI EM index is dominated by China, South Korea and Taiwan, leading to a pronounced East Asia tilt overall. "Cap weightings may have a theoretical justification, but it allows an investor to invest too much into a country going into a bubble phase," he says. "Japan for example, became 60% of the EAFE index in the 1980s. Risk weighting is one way of reducing this risk."
While risk parity may produce a theoretically better risk/reward trade-off than cap-weighted approaches, the problem with emerging markets is that liquidity is not adequate in many of the 21 or so countries within the MSCI universe for a rigid risk parity approach. "Theoretical risk parity is not economically sound as some countries only trade a few liquid stocks," admits Nakamura.
Lombard Odier therefore modifies the risk parity ideal by targeting a risk contribution proportional to the available liquidity within the country. What that means is that the theoretical allocation of 4.8% to countries such as Morocco, Egypt, Colombia and Hungary is reduced while the allocations to Malaysia, China, Taiwan and South Africa are increased. This results in a portfolio that still looks very different from the MSCI index: China's weight goes down from 17.2% in the MSCI to 8.2%, whilst Malaysia's goes up from 3.4% to 9.2%. Russia and India are broadly unchanged at not far off 6.5% each under both approaches.
Within each country, Lombard Odier has also adopted a passive approach. It first constructs an index within each country with stocks weighted inversely according to their volatility, so low-volatility stocks have the highest weights. It then targets stocks to replicate, as closely as possible, the target risk parity portfolio (aiming for an ex-ante tracking error below 0.50% annualised), taking into account trading costs. This produces a diversified unleveraged portfolio, composed of around 700 stocks. "We believe that this approach allocates risk budgets efficiently and reflects dynamic changes in volatilities and correlations in the market," says Nakamura.
First Quadrant's solution to the illiquidity problem is to bypass equity markets altogether. In emerging markets, the correlation between equities and FX can be 85%, Peters observes, so his firm uses the FX markets rather than equities within their multi-asset strategies for emerging markets. "From a risk weighting standpoint, Asian countries are highly correlated," he says. "Within Europe, the relative volatilities are different and the correlations across Eastern Europe are less than those found in Asia. Poland has a higher volatility than Slovenia so Poland has a smaller capital weight."
Risk parity is an interesting idea that has provided a counterweight to the pressures to adopt market cap-weighted benchmarks irrespective of valuations. But like blind approaches to any idea within finance, successful investment strategies are based on understanding the limitations of the ideal as well as exploiting its strengths.