The Danish pensions giant is renowned for cutting-edge work in liability-driven investment and separation of alpha and beta. But Martin Steward finds that the reality is different from what the buzzwords imply - and even more pioneering
It is a cold March morning, and I am trudging through misty woodlands, past the hard-frozen lakes of Hillerød in Denmark, on my way to visit the offices of ATP. That's right - one of the world's most sophisticated pension funds is located, not in Copenhagen, nor even a suburb of Copenhagen or a satellite town of Copenhagen, but in the suburb of a satellite town 30km north of the capital.
It is probably fanciful to ascribe ATP's pioneering spirit to this remoteness but, still, one of the (unprecedented) seven trophies the fund picked up at last November's IPE Awards was for innovation, and pioneers are rarely found hanging out in big cities. Similarly, the reality of life on the frontier often gets simplified or garbled as messages are passed back to the mainstream. That is why IPE recently interviewed ATP's three most senior investment professionals - to take another, closer, more detailed look at this strategic vision for the cutting edge of pension funding.
Freedom
Among the most recent innovations from ATP is ‘investment-driven liabilities', according to which the interest-rate return on member contributions is guaranteed at the level available to buy in the market.
It is not just ATP that claims that this structure, in place for two years now, results in higher guaranteed rates without affecting the freedom of the investment portfolio to optimise returns for bonus payments. "With the old fashioned guarantees, pension companies have to adjust investment prudence up or down, depending on the market situation," noted Mogens Steffensen, associate professor of actuarial mathematics at the University of Copenhagen, when the new model was introduced. "The terms of new guarantees fluctuate with the market. As a result, prudence is maintained at the same level. This means that the freedom to invest remains the same over time."
Indeed, it could be argued that the overall investment freedom of the bonus-paying portfolio is enhanced. As correlation risks between the guarantee hedge and the investment portfolio are eliminated, the purpose of the investment portfolio is purified. It must meet the LIBOR funding requirements of the hedge and the inflation-matching part of the bonus payments, but apart from that it can become a classical, efficiency-optimising absolute return portfolio.
"Like everywhere else, in Denmark there is a trend from DB towards DC, collective to individual - and we think that's a really bad model," says CIO Bjarne Graven Larsen. "We wanted to retain the advantages of a meaningful nominal guarantee and of collective risk-sharing, but combine them with the unconstrained risk-taking possible in individual DC schemes where there is no guarantee. We decided that guaranteeing 80% of the contribution and having 20% to put at risk was enough to let us implement a relatively unconstrained investment policy for the long term."
ATP does not swing for the fence with 100% in equities like a 30-year-old with a 401(k), but still, that "relatively" is nothing to do with ‘relative to liabilities' - it is constrained only by the solvency requirements of the bonus-paying portfolio, as defined by the Danish Financial Supervisory Authority's ‘traffic-light' system.
ATP calculates that the FSA's ‘red light' capital requirement fluctuates between about DKK15bn (€2bn) and DKK20bn (€2.7bn) of the fund's total reserves of DKK65bn. But on top of that, ATP implements a value-at-risk (VaR) limit: there should be no greater than 1% chance of hitting that ‘red light' level over any one quarter. (In fact, this is a conditional VaR limit, as the FSA's traffic light system is based on extreme scenario stress testing). Factor that in, and ATP reckons its minimum capital requirement is about DKK30-35bn. It estimates that other Danish funds will have to adopt this extra capital buffer as well, once Solvency II is active.
Against this backdrop, dynamic risk budgeting rules adjust overall investment risk to the size of the bonus-paying portfolio relative to that CVaR limit - and thus ATP's ability to bear that risk. "Given that surplus of DKK30bn, you could argue that we have too little risk in play - and that's what the dynamic rule does," Larsen explains. "It allows us gradually to ‘step-up' our risk budget for the following months."
Fundamental
But let's not get carried away. ATP really doesn't swing for the fence, even when it's pitched a sitter. Amid the post-crisis clamour for more tactical asset allocation, this innovator's approach to risk taking can sound curiously old-fashioned: "We want a portfolio that is relatively independent of whatever economic scenario plays out," as Henrik Gade Jepsen, the CIO for beta, puts it.
The difference between ATP's portfolio and the standard diversified portfolio optimised according to the historical volatilities of different asset classes is hinted at by Jepsen's reference to economic scenarios. Modelling a range of portfolios against 100 years of market data, ATP sought to optimise for absolute return through the four basic economic regimes: high growth, high inflation; low growth, high inflation; high growth, low inflation; low growth, low inflation.
This framework of research led naturally to breaking down the portfolio according to economic sensitivities. "We don't look at asset class correlations, because those correlations are a function of the underlying fundamental factors," Larsen observes - which, although ATP has been doing it for years now, still looks cutting edge to much of the rest of the pensions industry. The five risks that ATP identifies - equity, credit, nominal interest rate, inflation and commodity - are not a million miles away from the standard ‘asset classes', to be sure. The key, however, is the way the premise changes the portfolio-construction question being posed.
"Our analysis revealed that we did very poorly in the ‘high inflation, low growth' regime," Larsen recalls. "We were low on assets that did well with higher inflation and found that a lot of equities underperformed when growth was low. That helped us to move towards rebalancing some equity-related risk into inflation-linked securities."
Ultimately, the long-run average risk (not asset) allocation to each factor falls out as 35% equity, 25% inflation, 20% interest rate, 10% credit and 10% commodity.
If all of that sounds somewhat static, do not fret. That fundamental misunderstanding of what ATP means by its well-known ‘alpha/beta separation' strategy (implemented since 2005) is one reason why it is changing some of the terminology it uses. Jepsen will not be ‘beta CIO' for much longer. ‘Beta' carries too much suggestion of ‘passive', so that part of the strategy will be called the ‘risk-seeking portfolio' instead.
And indeed, it is anything but passive. The active element is not restricted to, or even primarily about, taking tactical bets with risk allocation. In fact, ATP goes out of its way to stress that it doesn't rely on any forecasting or market-timing ability. When it took out put-option protection on its equities in 2007 it was not because internal discussions about the emerging sub-prime crisis led to a prediction of the broad financial collapse that was to follow. "Our conclusion: we just couldn't know how significant this thing might turn out to be," Larsen recalls. "We asked different questions: Do we fear the risk? And if so, can we protect against it? We decided that we did fear the risk, but because the rest of the market didn't fear it at that time we were able to buy options 27% out-of-the-money. So we thought, why not?"
Still, this was an unusual move for unusual times, as was the decision to load up on Danish bonds at distressed prices at the beginning of last year. On average, the long-run risk allocations are meant to be strategic and they do not fluctuate very much. The active element of the risk-seeking portfolio is focused less on risk allocation and more on risk implementation.
Implementation
That means that, within the strategic equity risk allocation, the people in what used to be called the ‘beta' team actually spend a lot of time pursuing what the rest of us would call traditional active management (the premise being that most active management is really only enhanced implementation of systematic risk-taking). In other parts of the risk portfolio Jepsen indicates that "getting the shape of the yield curve right is an important factor". The interest rates portfolio delivered an explosive absolute and relative return in 2008, he notes, "not because we predicted rates, but because we are always managing the shape of the yield curve well".
The equivalent in commodity risk - managing the futures term structure - is run just as actively. Look at ATP's commodity portfolio at the end of 2009 and you see a capital allocation of DKK15bn but a net exposure of just DKK8bn. You might be tempted to conclude that ATP runs its commodity portfolio like a diversified trend-following managed futures programme, leveraging some commodities and shorting others. You would be wrong. Plenty of that goes on in ATP's ‘alpha' portfolio, but in the risk-seeking portfolio there is currently nothing except oil (with gold on the discussion agenda). The short exposure is from various points on the futures term structure of that single commodity.
"Most people are trying to forecast the spot price of oil," says Jepsen. "We think it's much more important to manage the yield from backwardation or contango actively - because if you don't you can lose 40-50% of your potential return. We have three strategies: for backwardation, a simple long position at the shorter end; for contango, either a simple long at the longer end, or long/short carry. The investment committee approves changes between those three strategies. This is ‘beta' for us, harvesting a market risk premium."
As noted, this is far from ‘passive' - and the risk/return payoff is far from marginal, as figure 1 shows by comparing ATP's actual returns with a naïve implementation of its strategic, long-run risk allocation. But the most important thing to note is that it was this freedom to be active in the ‘beta' portfolio, rather than any ambition to purify the ‘alpha' portfolio, that was the primary objective of ATP's alpha/beta separation project.
"The alpha/beta structures you find elsewhere are typically driven by the desire to find better structures for alpha," observes Fredrik Martinsson, CEO of ATP Alpha. "That's a ludicrous starting point. It should be about providing flexibility for the beta side - after all, that's of much greater importance to a pension fund. We need to get away from the traditional structure with all these prima donnas sitting on asset classes trying to protect their stupid benchmarks."
In other words, separating alpha from beta leads naturally to placing things like traditional active management against equity benchmarks, active duration management against bond benchmarks and active roll-yield management in commodity futures where they belong - in the ‘beta' or systematic risk-taking portfolio - rather than pretending they are a source of true alpha.
Idiosyncratic
Once again, the way that ATP's terminology is used in the wider industry is causing it to re-think that terminology. We have already seen LDI become ‘IDL' and beta become ‘risk-seeking'; the question now arises whether ‘alpha/beta separation' really describes what ATP does at all. In a way, the key decision was not to split alpha and beta, but to split hedging and risk seeking. Once you have cut the risk-seeking activities loose and broken them down into your fundamental economic risk factors, it becomes clear that if alpha is to play any part at all, it is as the sixth distinct risk factor - not separate from the other five, but an equal alongside them.
"If we went back to 2005 today, we'd probably describe a hedging portfolio and a risk-taking portfolio which had one extra, ‘idiosyncratic risk' class, rather than implying that alpha and beta were two equal parts," Martinsson suggests. "So Henrik's role is risk-taking, broadly defined, and you could say my role is to be responsible for a very small part of that."
To re-cap, ATP's approach does not begin with the objective of defining alpha more strictly so that it can strip out, through engineering, the correlation risk of systematic beta exposures within its alpha portfolio. "If you turn this process into an academic engineering exercise the only thing you'll achieve is to slow down all of your activity," Martinsson observes. "We can think about how you try to make sure that alpha is really pure, but to overlay definitions of beta, alpha and structural beta turns it into an academic discussion - and even the academics remain unclear about whether structural beta is really beta. Look at the hedge fund replication engines out there. They essentially work by curve-fitting - give me enough variables and I can explain whatever you want me to explain. We know a couple of groups that have gone down the more academic route to alpha/beta separation, and they end up being hurt by implementation issues and get nowhere."
There is no doubt that ATP arrives at the same place - that is, purer sources of alpha - but it does so in a more practical and robust way by turning the question around. If alpha is conceived as one of six distinct risk exposures in an all-weather diversified portfolio, then, by definition, it has to be as distinct as possible. For ATP's alpha team, therefore, idiosyncrasy and non-correlation is the starting point.
"We emphasise independent teams following independent strategies, but obviously taking account of and dependent on the activities in the risk-taking portfolio," explains Martinsson. "That means thinking through things like hidden beta at the design stage. If you look at the net long exposure in the external long/short equity space, for example, it tends to fluctuate between 20-50%, whereas our equity long/short strategies have much, much tighter maximum net long position limits."
There is nothing resembling traditional long-only active management coming from Martinsson's team. Most equity management is market neutral, and even those teams hired for their market-timing and stockpicking skills are required to do so within a constrained net-long regime. "If a team has a long-term proven ability in those skills, I'm happy to take it," he says. "But that has to be absolutely clear, ex-ante. What I can't accept is people who view themselves as fundamental stockpickers having huge chunks of notional equity risk premium in their portfolio. We invest a lot of time in that ex-ante design process - much more than when pension funds invest through funds of funds - precisely because we can influence that process."
Having a big internal alpha operation is clearly the major determinant of ATP's ability to incorporate alpha into its portfolio in this way. Is it possible for the wider pensions industry to achieve something similar, jostling with thousands of other pension funds, endowments, private banks, family offices, funds of funds and high net worth individuals to define their alpha allocation according to their specific needs? Even ATP itself, which eventually hopes to have 20% of its alpha coming from external managers, has yet to make a single external allocation.
Of course, that's largely because most of the hedge fund world provides stuff ATP either doesn't want or doesn't need: "We are trying to build 10-12 internal teams complemented with a handful of external managers," says Martinsson. "We want depth of skills rather than breadth of coverage. We alreadt have strategies like global macro and CTAs, so we will not be anything like a mainstream pension fund when we go out to find hedge funds."
Nonetheless, other pension funds that do not have the same internal input at the design stage will still find it difficult - if not entirely impossible - to find ready-packaged alpha that is as pure as that put together within ATP. They might find that they have to go down the route of purifying the alpha as much as they can themselves; or they may decide that that is a fool's errand, and the best they can do is recognise that they cannot acquire pure alpha, stop allocating to separate ‘alpha' or ‘hedge fund' risk buckets as if they can, and start incorporating those exposures into their systematic-risk buckets.
After all, with IDL, alpha/beta separation, fundamental risk allocations, and much else besides, ATP is a true pioneer. Everyone else eventually follows where pioneers lead, and there is no shame in adapting the lessons we learn from them to our own limitations and constraints. But to do so successfully it is important to understand properly how the pioneers' thinking led to the position they occupy today. In ATP's case, that thinking process has not necessarily been as simple as is implied by some of the buzzwords we now associate with its strategy.
