Risk Parity: Case study: ATP
Risk parity alone cannot do the trick of maximising risk-adjusted returns and minimising the risk of large drawdowns. Henrik Gade Jepsen describes the additional pillars on which ATP's investment approach rests
Risk parity has been a very important component of ATP's investment approach for the past five years. In the investment portfolio we focus on generating a return that is sufficient to preserve the long-term purchasing power of pensions - that is, to ensure that pensions can be increased in line with inflation while being able to cover the provisions to finance life-expectancy increases. At the same time, the risk of large losses on investments should be low, given that heavy losses could make it difficult to bear investment risks going forward - and thus affect the ability to deliver high returns in the future.
Accordingly, the investment portfolio is managed with the aim of delivering positive returns in most financial climates. The challenge is to maximise risk-adjusted returns while minimising the risk of large investment losses. The question is how one should approach this problem in a world characterised by fat tails, unstable asset class correlations and stricter solvency requirements.
Risk parity has been an important part of the solution but should be combined with a number of other components, described below. We define risk parity as an allocation exercise which aims to spread risk more or less evenly among risk factors in order to generate higher risk-adjusted returns than equity-centric institutional portfolios. Risk parity partly solves the dilemma created by the need for high investment returns and the imperative of remaining solvent at all times. In order to ensure that the return stream is as stable and as independent of economic trends as possible, ATP's portfolio is invested in five risk classes with very different risk profiles. The five risk classes focus on interest-rate sensitive bonds (the interest rate risk class), the ability of issuers to repay debt obligations (the credit risk class), corporate earnings (the equity risk class), general price developments (the inflation risk class) and oil prices (the commodity risk class).
Risk parity - a necessary but not a sufficient condition
On its own, risk parity cannot do the trick of maximising risk-adjusted returns and minimising the risk of large drawdowns. ATP's investment approach thus rests on three additional pillars.
First, to generate the highest possible risk-adjusted returns it is necessary to hedge all uncompensated risks. For ATP the main uncompensated risk is the interest rate risk on pension liabilities. Hedging is performed through a separate hedging portfolio consisting of very long-dated interest rate swaps and government bonds.
Second, in recent years, ATP has been focusing on protecting itself against ‘tail risks', such as sharp drops in the value of the investment portfolio: a 30% plunge in equity or commodity prices, for example. Tail risks can represent a significant risk to ATP, and so it is prudent to safeguard against heavy losses if the events were to occur. Tail-risk protection also makes leverage an acceptable way to balance risks in the portfolio as the potential larger losses from leveraging can be eliminated by buying downside protection.
Third, ATP cannot invest freely if the value of its assets approaches the value of the guaranteed benefits. Should this happen, the risk must be reduced to ensure that sudden losses do not further weaken the solvency ratio. This weighting is explicitly formulated in the dynamic rule, ensuring that the level of risk is appropriate at all times. The dynamic rule defines a risk budget based on ATP's reserves. ATP's board has determined that risk must not exceed the risk budget. Accordingly, high investment returns mean that it is possible to increase the risk further by purchasing risky assets while major losses result in active divestment. Reducing the risk budget is the last line of defence if diversification and tail hedges fail to mitigate the impact of large financial market declines.
Constructing a balanced portfolio
Instead of looking at the investment universe as asset classes, we allocate to five risk classes: rates, credit, equities, inflation hedges and commodities. Moreover, we allocate the entire investment portfolio risk according to risk parity principles, which probably distinguishes ATP's approach from that of most others' risk parity investments. The risk classes reflect very important underlying economic risk factors which are the primary determinants for return. Having a risk factor approach for the entire portfolio enables us to focus on the most important return drivers and avoid spending time on less relevant investment discussions.
We allocate risk between the risk factors to obtain the most effective diversification. The whole idea of allocating risk instead of cash is to avoid a single asset class (or risk factor) dominating the portfolio return merely because of the way the underlying investment is structured. One dollar invested in equity is much more risky than one dollar invested in Treasury bonds because equity is a leveraged investment (being the lowest part of the capital structure leveraged up by bank loans, corporate bonds, mezzanine debt and so on). Letting that determine portfolio characteristics would be like letting the tail wag the dog. Nevertheless, this is what happens in many institutional portfolios.
Merely allocating risk between the five risk classes reflects a passive investment approach. We have taken things a step further by making the investment approach dynamic by introducing other systematic risk factors that are important for the investment returns. For example, the risk class rates reflects changes in the level of interest rates, which is the most dominant investment return driver for bonds. But the steepness of the yield curve is also a systematic risk factor that is an important determinant for investment returns over time. If attention is not drawn to these other factors a high potential return can be lost. And since these risk factors reflect systematic market risk they should be addressed directly in the investment portfolio and not in an alpha overlay.
The starting point for us is to get the exposure within each risk class that gives the highest expected risk-adjusted return over time. On top of that we develop and implement strategies which dynamically change the exposure to other secondary - but important - risk factors within each of the five risk classes. In this way we achieve a portfolio that captures changes in systematic market risks.
For example, in the risk class rates we have a strategy for the systematic risk factor ‘steepness' that dynamically changes the exposure to the yield curve by switching the total government bond portfolio between bonds with short time to maturity and bonds with long time to maturity. Thanks to this strategy ATP's investment portfolio has gained more than 1% compared to being ‘passively' invested in bonds with 10 years to maturity.
We pursue similar strategies within the other four risk classes. For example, in the risk class commodities - where we only invest in oil futures - we use a similar ‘steepness' strategy to determine the maturity profile of the futures contracts in which we invest. While many investors in recent years have experienced high negative carry from investing in oil, our strategy has generated significantly better return characteristics. In the risk class inflation we allocate a part of the risk budget to illiquid investments like real estate, infrastructure, alternative energy and forestry. The aim of these investments is to give us stable inflation-adjusted cash flows in addition to the illiquidity premium. The liquid part of the portfolio is primarily placed in index-linked bonds and inflation swaps. For the liquid part we have curve strategies for the real yield and break-even curves and a strategy that switches the exposure between real yield and breakeven inflation.
The risk factor approach also ensures that we implement the portfolio exposures in their most ‘clean' form, thereby avoiding unwanted risks. For example, we have not had holdings in bonds issued by peripheral EU countries in the portfolio as these are adding spread exposure - in other words, credit risk. If we want to use these bonds for spread exposure they will belong to the risk class credit and will be scaled accordingly.
Long-term return results
The investment approach has delivered very satisfactory results. Since 2006 the excess return of ATP's investment portfolio (return after funding, tax and costs) amounted to almost €7bn - which is above the target of €6bn that is sufficient to adjust ATP's guaranteed pensions for inflation and longevity over the same period. Compared to a portfolio where each of the five risk classes are represented by a simple index, the implementation strategies have added a significant excess return of more than 5% over the last four years.
During times with very high uncertainty about economic development a balanced portfolio is even more justified, as it is not betting on one single outcome. We believe that a balanced portfolio is a very important building block to get high risk-adjusted rewards, but to get superior risk-return characteristics over time one should add a dynamic investment approach on top of it.
Henrik Gade Jepsen is chief investment officer at ATP