Martin Steward discusses the philosophy behind the risk-factor allocation approach adopted by Danish pension fund ATP with CIO Henrik Gade Jepsen - and the impact of a new risk environment on its strategy

IPE: Pension schemes pursuing swaps-based LDI programmes are, in a way, already holding risk parity-like portfolios because of this leveraged exposure to interest rates. If they were to separate the growth portfolio from the hedging portfolio and make the growth portfolio an independent risk parity portfolio, the overall leverage in their asset portfolio - via swaps and bond futures - would be huge. As an LDI investor yourselves, how do you think about this leverage?

Henrik Gade-Jepsen: We split the portfolio into two because the first thing we do is make sure that we can pay the pensions that we have promised to our members - and our guarantee is the nominal long-dated interest rate at the time they pay in their contributions. That guarantee is handled in one, separate portfolio. In addition, we do have some capacity to take risk to try and provide a pension in excess of that which is guaranteed. But our starting point is that this risk-taking portfolio should be diversified on its own terms, not taking into account anything that has already been done in the hedging-instrument portfolio that is guaranteeing the pension promises. That's the way to achieve the highest risk-adjusted return and thereby use the risk budget that we have to generate bonus pension payments in the most intelligent ways for our members. It is worth taking the extra leverage to achieve that result because it gives us a more balanced and robust investment portfolio and at the same time enables us to hedge our pension liabilities.

IPE: Like many risk parity practitioners, ATP follows a portfolio construction methodology that focuses on fundamental economic risks, and on the relative volatility contribution from its five risk classes. However, its portfolio does not have ‘risk parity' as an objective. The strategic risk allocation is 35% equity risk, 25% inflation risk, 20% interest rate risk, 10% credit risk and 10% commodity risk. You have described an optimisation process, rather than an allocation tailored to deliver the highest return while respecting ATP's specific risk tolerances or liability profile. How did you arrive at this risk allocation, rather than an equal-weighted one?

HGJ: We did a lot of number crunching on a range of portfolios against 100 years of market data in order to come up with something that did well, both on average and in many different economic scenarios, including low-growth scenarios and high-inflation scenarios. What you end up with is a broad range of risk-based allocations to different economic risk factors - equity, credit, nominal interest rate, inflation and commodity risk - which are not necessarily equally-weighted. When we did that back in 2005 we had no idea that there was any such thing as a ‘risk parity' portfolio - so we didn't start with that assumption and that might explain why we don't allocate an equal amount of risk to each factor.

IPE: Pure risk parity is at least agnostic about any kind of optimisation for economic scenarios. How did you avoid the danger of curve-fitting in your process?

HGJ: You are right that you can only get part of the way by crunching the numbers. Then there is an element of common sense, experience - the art rather than the science of optimisation. Having said that, we feel that, while each risk factor does not have the same risk allocation, we have constructed a portfolio in which no single risk dominates. The biggest is just one-third of overall risk, and the smallest, at 10%, is not negligible. Furthermore, we believe that ATP's portfolio is more resilient against changes in the economic environment than an equal weighted risk parity portfolio.

IPE: Asset classes are very often bundles of economic risks and sensitivities, and it can prove difficult or costly to unbundle them. I know that you do not attempt to swap out the interest rate sensitivity of your credit portfolio, for instance, because you don't consider it to be a significant part of the risk. But ATP has removed peripheral euro-zone bonds from its rates portfolio because they are now primarily driven by credit risk.

HGJ: In the past we have assumed that we could put any OECD government bonds into the risk class that we call rates. But removing anything that has the slightest trace of credit risk has been very important for us, because otherwise the risk class would start to behave very differently - perhaps even like an equity risk exposure. So we came out of peripheral euro-zone bonds at a very early stage. We can put them into the credit portfolio, if we want, but not the rates portfolio.

IPE: But now, arguably, you could point to credit risk as a big driver of some of the core euro-zone yield curves. Some, like France's, have become higher-yielding, and while others, like Germany's or the UK's, have become safe havens, that in itself is a credit sensitivity, not a rates sensitivity, isn't it?

HGJ: That's true, but safe havens are good diversifiers in the same scenarios as interest rates. So you can say that yields can fall because economic activity is depressed, or because investors are looking for safety. For all the different drivers of those yields, the point is that it is a similar economic scenario behind each one. This was the part of the portfolio that did well for us in 2008 and the second half of 2011, and turned out to be the only really important diversifier.

IPE: Similarly, ATP has some private equity in its equity risk portfolio, and the inflation portfolio includes infrastructure and real estate. These asset classes bundle up a bunch of economic sensitivities, but then they all have illiquidity risk on top of that, too. The last few years has shown that illiquidity risk can overwhelm every other risk from time to time: isn't there a case for ATP adding ‘illiquidity' separately to its existing five risk classes?

HGJ: There can be short-term differences in the way private and public equity is priced, but in the long run they should be driven by the same fundamental risks. I completely agree that there are times when the liquidity risk becomes dominant - and that is a factor that we take account of across all of our risk classes. Risk parity is not sufficient at all times: liquidity is not an issue until everybody needs it. So while we have some illiquid investments we have always remained very liquid, at least compared to many other diversified investors, both at the overall fund level and also at the level of individual risk classes, to retain as much flexibility as we can at all times. And our put option overlay is also a liquidity provider at a time when the market is demanding liquidity in large volumes.

IPE: This refers to your tail-risk hedging programme of equity index out-of-the-money put options. Does this explains how, when world equities lost 18% in Q3 2011, ATP's equity portfolio managed to lose only 4.6%?

HGJ: We do some tail-risk hedging, and the value of that hedge increased. But in addition, we consider private equity to be part of the equity risk bucket, and that has exhibited a somewhat more stable price development through 2011, balancing out the losses in listed equities.

IPE: I'd like to ask about how ATP rebalances its risk portfolio. At the end of H1 2011, the inflation-risk capital allocation had grown to 29% of the portfolio. That was back to 16.8% at the end of Q3, and down to 15.6% at the end of the year - despite positive returns to the risk class. That suggests regular rebalancing. Similarly, a volatile asset class like commodities, which can only contribute 10% to strategic portfolio risk, must have to be rebalanced frequently: the capital allocation went from 3.7% to 2.2% over the course of Q4 2011, for example - which would be either a rebalancing, or a significant change to the risk contribution to the portfolio through a performance-related change in capital allocation.

HGJ: In fact, while we try to keep the risk allocations fairly steady, we are not rebalancing all the time. The reason you see quite a big movement in 2011 is that we de-risked quite substantially. Some of the exposure we took out was to inflation-linked bonds: in risk terms it wasn't a big move, but it was substantial enough in money terms. This was a discretionary decision, but I would say that we de-risked in a balanced way: we didn't take risk out of one or two risk classes - we took risk out across the board, so that we more or less kept the distribution of risk - if not the distribution of capital - at the same level.

IPE: So this is not so much about reducing risk by decreasing allocations to ‘risky' assets - because under the LDI two-portfolio model the assets in this portfolio are all considered to be risk assets. It's more about reducing market and risk exposure across the entire portfolio, while maintaining the balance between the five economic risk classes. You are not relying solely on the diversification of risk classes to protect you from market turmoil in the way a pure risk parity practitioner, who maintains the same level of leverage, would be. That seems important, given today's environment: the extreme valuations that we see in bonds and the market inflexion-point that we are experiencing might both be expected to feed into meaningfully higher bond volatility. Moreover, in an environment of low growth, sticky inflation and increasing concern about sovereign creditworthiness, we might expect considerable, extended downside correlation between bonds and equities - the sort of thing we saw, for a short period, in 1994 and 1995. That would be terrible for any kind of portfolio, of course - but a risk parity portfolio would underperform simply because it is a leveraged exposure to the same directional risk as that held in the traditional, non-leveraged portfolio, wouldn't it?

HGJ: There are so many moving parts, but overall I think we can safely say that instead of facing a normal distribution of potential outcomes for the world economy and markets, as we got used to during the 1990s and early 2000s era of stable growth, we now face a situation of multiple equilibria. Maybe we muddle through - growth doesn't get much worse, we avoid deflation, and perhaps we even see a inflationary environment in a couple of years' time. But then again maybe we see a deflationary shock emanating from one of the crisis zones in the global economy and financial markets. Maybe we'll end up with 1970s-style stagflation. To make it even more complex, we may find very different situations in two or three different parts of the global economy. There are fat tails in that distribution, volatility will be high, but returns won't necessarily be higher to compensate for that extra risk. When we are discussing risk parity it's easy to become fixated on interest rates, but I really think that this pertains to all asset classes - risk-adjusted returns could be meaningfully lower in the immediate future. But the uncertainty is not just around volatility, it's also around correlations, as you say. I'm sure that correlations between government bonds and equities will not be the same in each of those possible scenarios.

Taking all of this into account, we feel that you should probably focus more on defence than offence. So how can you play the defence? One way we have already been playing it is through the tail-risk hedge that we use - but it's rather expensive to increase that at the moment. So the other option is simply to lower the amount of risk in the portfolio until you have greater clarity about where we are heading. I agree that if you are highly-leveraged at a time when yield curves flatten or even invert, that could be very expensive. But the fact is that we are very liquid, which means that we can reduce leverage very rapidly within a few days - and we are not leveraged that much compared to some other funds out there. And, of course, if you are low-risk and fairly liquid, you leave yourself the option to be a buyer if we should end up with risky assets being much more cheaply priced.