Dutch Decision Time

As pressures on pension fund solvency increase, Dutch giants ABP and PGGM have chosen divergent strategies. Nicholas Dunbar reports

Niels Kortleve, head of the integrated balance management committee at Dutch pension fund PGGM, likes to use a parable to describe the dilemma currently faced by his fund. "I want to go on holiday to the Far East, and I plan my trip," he says. "Two days before I'm due to leave I get a phone call saying my mother-in-law is in hospital. You can reach the conclusion that you have to change your plan due to short-term risks; you have to deviate from your long-term plan."

The parable goes to the heart of a debate taking place across Europe: how should funded pension plans, charged with a mission of paying long-term retirement benefits to employees or taxpayers, reconcile these long-term goals with the short-term constraint of solvency in a mark-to-market environment? Until recently, many pension funds felt free to completely ignore short-term considerations in their asset allocation decisions, cosseted by non-market valuation methods that permitted them to ignore cold reality. However, under regulatory and market pressure, this has been changing. The triple threat to solvency from equity underperformance, declining interest rates and increasing longevity became impossible to ignore. It was the 'phone call' of Kortleve's parable, forcing a reappraisal of a much cherished plan for the future.

Nowhere is this debate more acute than in the Netherlands. Unlike the UK, whose brazen culture of occupational pension deficits and defaults makes continentals shake their heads in disbelief, the Dutch pension funds are widely respected for their professionalism and fully-funded status. Yet the introduction of mark-to-market discounting of liabilities has exposed a worrying vulnerability in the system, which alerted regulators at the Dutch Central Bank (DNB) and led to a new financial assessment framework, FTK.

The new climate has already transformed the pensions debate in the Netherlands. It forces a decision: whether to stick to long-term asset allocation, or swerve from it and reduce short-term risk using liability-driven investment or derivatives market solutions. Some are looking carefully at Denmark, whose move to mark-to-market in 2001 prompted a massive use of interest rate hedging instruments, and some Dutch funds, notably Hoogovens, have already taken this route.

There are two pension funds that dominate the Netherlands because of their sheer size. There is ABP, which provides pensions to government employees and currently has EUR170 billion in assets. Then there is PGGM, which caters to the healthcare and social services sectors, and reported assets of EUR64 billion. Both of these powerful Dutch institutions have resisted the introduction of FTK, but with their underfunding risk levels above the regulatory limit, both are being forced to come to terms with the new climate.

The way they do this will be closely watched, and not just by fellow Dutch pension funds. With around EUR200 billion in liabilities between them, a move to Danish-style interest rate hedging by the two giants would be likely to have a significant impact on long-term euro bond yields and swap rates.

Not surprisingly, ABP and PGGM have said little up to now about their risk management plans. But facing increasing regulatory scrutiny, both have started to talk. Meanwhile, their Danish equivalent, the supplementary state pension provider ATP, with assets of EUR40 billion, has been talking about its experiences in 2001 and its evolving approach to risk. Are there any lessons for ABP or PGGM?

The traditional starting point for asset allocation at pension funds is asset-liability management (ALM) modelling. Typically, ALM takes a pension fund balance sheet at a given point in time, and by simulating its behaviour into the future under a wide range of scenarios, attempts to find the optimal asset and liability mix for the fund. The process is intrinsically long-term in its assumptions, explains Kortleve. "In ALM we do a best estimate of what could happen to the world, long-term," he says. "We make assumptions of what are the most sustainable levels of economic growth, interest rates and equity market returns over 10-20 years, and we run Monte Carlo simulations around these assumptions."

ALM attempts to encompass the three big moving parts that impact the assets and liabilities of a fund over such timescales. For Dutch pension funds this includes not only investment decisions but also the level of contributions and salary level indexation added to the fund's nominal legal liabilities to members. When used properly, ALM defines the type of pension product that a fund is capable of offering.

Kortleve explains how it works. "We have to see the fund through difficult periods, but to what extent can you lower indexation if the funding ratio is low, for instance, and to what extent are you able to raise contributions?"

When ALM models first appeared in the mid-1990s, they seemed to offer a readymade risk management framework for pension funds. However, because of the computationally onerous multi-period stochastic optimisation techniques and econometric modelling required, ALM could only be deployed infrequently. For all except the biggest pension funds, it became the domain of specialised consultants. ABP was an early adopter of ALM, and as Tom Steenkamp, the fund's chief investment officer for allocation and research proudly points out, it built its own model. "It's entirely our own," he says. "We feel that big pension funds like us need to have the expertise ourselves." And, true to its promise, ABP treats the model as a risk management tool. "Our main risk allocation system is effectively the ALM model," Steenkamp says.

For all the promise of ALM, the long-term horizons and infrequent modelling cycles seem designed to filter out short-term information. And ABP, which only makes asset allocation decisions at three year intervals, likes it that way. Steenkamp comments: "What the model says is that the long-term risk characteristics of assets are different to the short-term ones. That's very important."

Some observers contend that this is a weakness of ALM, as it reinforces the idea that liabilities do not fluctuate in market conditions, in contrast to assets. Steenkamp refutes such suggestions. "We already modelled a mark-to-market environment," he says, "and in our original ALM work we had a market valuation of our liabilities. Real risk means that the liabilities have a market value with an index-linked discount rate."

Market valuation applied to internal ALM modelling is one thing, but its use in pension fund disclosures is another. ABP is a case in point. In its 2004 annual report, the fund reported a historical funding ratio for 2001-2003 based on the real market-based discount rate. For those three years, ABP was attempting to treat discretionary index-linked liabilities to its members as if they were the legally enforceable nominal liabilities. In 2001, the funding ratio on this basis was 112%, but in the following two years it had plunged to 86%.

Had the indexed liabilities actually been legally enforceable, as would be the case for a UK defined benefit scheme, ABP would have been EUR25 billion in deficit in 2003. But Steenkamp insists that this is a meaningless, short-term interpretation of the figures. "To conclude that a long-term investor has a deficit of EUR25 billion, only has value when you immediatly sell your liabilities and assets on the market, assuming this is possible," he says. "Given the long-term financial indexation, contribution and investment policy of ABP, there is no deficit at all."

Does that mean the calculation is only of intellectual interest? Steenkamp denies this, pointing out that ABP's market valuations were taken seriously enough by management to influence the fund's contribution and indexation policy. But regarding the impact of mark-to-market valuation on investment policy, a dividing line appears. For an example of such impact, look no further than Denmark's ATP. Along with other Danish pension funds, ATP offered defined contribution pensions with a 4.5% minimum interest rate guarantee and discretionary bonus policy. As long-dated interest rates fell in the late 1990s, concerns grew about the liabilities lurking on pension fund balance sheets.

Henrik Gade Jepsen, ATP's head of fixed income, takes up the story. "Back in 1999 we had realised that we needed to focus more on liability issues," he recalls. "We had built a full-scale ALM model to handle these issues, and we also knew that marking-to-market might come over time. So we ran simulations and looked at various scenarios going out into the future so we knew where we would stand." The result of the ALM simulation was alarming, Gade Jepsen continues. "We found that if we did nothing to hedge this interest rate risk coming into the liabilities there was a huge risk of us becoming insolvent. Something like 25% of scenarios would end up with insolvency. It was really scary. The risk of doing nothing was simply too high and it swamped all other risks including equity risk in the portfolio."

Convinced that the problem was more than an intellectual issue, Gade Jepsen and his colleagues took action. "We called an extraordinary board meeting," he says. "No matter which way interest rates were going, this was going to have huge consequences for ATP, so the board needed to be involved so they could understand and approve it."

Late in 2001, following the adoption of mark-to-market methodology by Danish pension regulators, ATP entered the derivatives markets for the first time, hedging some EUR9 billion of liabilities with interest rate swaps. Eventually, the fund would almost completely immunise its exposure to interest rates, with its 2004 report listing a swap portfolio with a market value of EUR3.7 billion and notional size of EUR29 billion.

Although their ALM studies had not led them to a change in risk management or asset allocation, market events eventually forced ABP and PGGM to revisit the contribution and indexation aspects of their policies. For example, the emergence of a deficit within its index-linked market-based framework led ABP to firmly emphasise in 2004 that its ambition to index pension payments to salary growth was not a guaranteee.

Steenkamp explains: "If we really promise indexation, we need to discount our liabilities against a real interest rate. That's a problem for us, so we don't promise indexation any more. Each year, if our solvency position is insufficient to give indexation, then we do not."

In 2004, the indexation offered by ABP was only 0.12%. For PGGM, which "wants to provide a highly indexed pension benefit in the long run" according to Kortleve, it was zero. The situation is likely to be repeated in 2005. Both funds point out that this reflected salary freezes in their respective sectors, and therefore was not solvency-related. But the fact remains that Dutch consumer price inflation for the year was not zero.

Thus, the 697,000 and 175,000 pensioners at ABP and PGGM respectively would have faced a decline in the real value of their pensions. Steenkamp points out that in the long run, wage-indexed pensions outperform price-indexed ones, but obviously this depends on indexation not being dropped due to solvency reasons.

Also controversial are attempts to tweak the other moving part of the pensions picture, namely the contributions by active employees. Between 2001 and 2005, ABP increased contributions from 11.8% of salary to 21.6%, while the old-age provision component of PGGM contributions has risen from 5.1% to 15.5% over the same period. Even ATP, with its early hedging advantage, had to increase contributions in 2004, which now stand at about 1% of average Danish pay. It has also cut its once generous minimum guarantees.

But while the ability to change contributions and benefits should in principle allow pension funds to take more risks than insurance companies which lack this freedom, risk taking has its limits. The biggest risk for pension funds with defined or guaranteed benefits is interest rates. When ATP hedged in 2001 and 2002, long-dated euro swap rates were comfortably above 5%. Over the last two years, they have inexorably declined to below 4%.

And there is another risk facing such pension schemes: increasing longevity. ATP is one of the few pension funds to spell out the consequences of this risk in its annual report, stating that a 10% improvement in Danish mortality would cut the fund's bonus reserve by 25%. This is equivalent to an effective 0.25% cut in interest rates, according to ATP's chief actuary Chresten Dengsoe. "We spend a lot of time thinking about longevity," he says.

These two factors have whittled away the solvency position of the Dutch giants, and in the absence of hedges, they have undone the effect of positive investment returns. On a nominal mark-to-market basis, ABP's funding or solvency ratio declined from 125% in 2003 to 121% in 2004, while PGGM's ratio fell from 123% to 116% over the same period.

Neither fund has disclosed interim solvency figures for 2005, but some clues can be found in the figures released by ATP for the first half of the year. Declining interest rates increased ATP's liabilities by EUR4.3 billion, while improving mortality added another EUR509 million. Luckily for ATP, this was offset by investment returns of EUR5.1 billion, almost two thirds of which was due to mark-to-market gains on interest rate swaps. Both ABP and PGGM are believed to be largely unhedged. It is this period which has seen the emergence of FTK. Much debated in the Netherlands, the FTK framework requires a minimum capital requirement for pension funds in the form of a 105% minimum solvency ratio, and a maximum one-year under funding probability of 2.5%, measured using nominal liabilities.

Although officially delayed, FTK is now a reality for Dutch pension funds. Needing a funding ratio of over 130% in order to get the probability below the 2.5% limit, both ABP and PGGM are squarely within the DNB's sights. But as Kortleve explains, the impact comes not so much from the more gently enforced 2.5% rule, but the fear of draconian sanctions if the minimum 105% limit is breached. "The law states that if you don't have that 130% level, then you have 15 years to get back to that level. But we always look to the 105%, because the pension law says that if you fall below that, within eight weeks you have to come up with a short-term recovery plan which implies one year to recover."

This contingency effectively closes off the traditional freedom to increase contributions, Kortleve continues. "If this new pension law applies, then any underfunding below 105% must be recovered within 12 months. The pain would be very substantial if you have to cope with that using contributions only. It would lead to a 10-20% increase in contributions just in one year. We are working for the social welfare and healthcare sector which has very tight budgets and that really would be an issue for them."

This unpleasant dilemma has led PGGM to make a decision to break with tradition. He explains his role. "My department is responsible for setting up a financial policy committee at PGGM. The first part of that is ALM which is long-term, but I'm also responsible for what we call integrated balance management (IBM), to try and find out whether we can better model the systems to manage the short-term risk, and look at solvency for the intermediate term. IBM is solvency management." Kortleve makes clear that this has been driven by FTK's probability limits. "What the IBM project is doing," he says, "is calculating those kinds of numbers: the likelihood of getting below 100% funding at the end of the year. What we do is take the current asset allocation and we look at how much the duration of the fixed income portfolio deviates from the duration of the liabilities."

For Kortleve's team, this is only the first stage of the analysis. "Then we have an economist and strategist from the investment department looking at the financial markets, picturing what will actually happen, and looking at the risks". Armed with this understanding, the financial policy committee considers a range of hedging instruments and considers the implications, says Kortleve. "Assuming that we intervene to increase duration, the committee could say, what is the impact on the probabilities of getting underfunded, and the distribution of that? We show the consequence of interventions on the risk in the balance sheet."

Although he will not disclose specifics, Kortleve insists that this is far from being an intellectual exercise. "The duration of the benefits is 16 years, and the fixed income portfolio used to have an average market duration of about five years. You can see that this in itself would lead to a substantial gap." The logical way to manage this duration risk would be to have a swap overlay capability, and possibly use interest rate and equity options as well. Asked whether PGGM has done this, Kortleve declines to comment.

Unlike PGGM, ABP is sticking with its traditional ALM approach, and its internal solvency model serves as an intellectual defence for its position versus the DNB. Steenkamp comments: "The regulator's model is solvency-based, and it allocates risk to different asset categories. We use stress-based parameters, so there's no correlation between hedge funds and equities and fixed income, for example."

Rather like a sophisticated life company, ABP hopes to exploit correlation modelling to substantially reduce its economic capital. "We think our internal model is especially different from FTK with respect to correlation," Steenkamp says. "For example, we treat hedge funds and commodities as very diversifying assets within the model. So we might give them a correlation of zero, while in the FTK model they have a correlation of one. That's a huge difference."

The bold use of correlation assumptions may prove a challenge for the DNB, but Steenkamp is unapologetic. "Diversification is the most important investment paradigm, and if you don't get rewarded for it in the FTK rules, it's sensible to do it in the internal model."

Steenkamp is dismissive of the current vogue for liability-matching solutions. "In general, for ABP the interest rate risk is far more important than it used to be," he concedes. "But I think that in the short-term our asset allocation will not change that much." The reason for this stance is that ABP remains attached to its traditional long-term policy, "directed towards the trade-off between real return and real risk" as Steenkamp puts it. "And we don't believe in real liability matching because it leads to high contribution rates," he adds.

Matching nominal liabilities is a distraction, continues Steenkamp, because it ignores inflation. "It is not very optimal in an inflation-matching sense," he says. "Why? If inflation increases, it is not a very good strategy. The second reason is that we think interest rates will increase, so the fair value interest rate in the long-term is higher than the 3% now." Are pension funds being irresponsible in hoping to be bailed out by rising interest rates or inflation? This question goes back to the heart of the holidaymaker parable. Kortleve appreciates the dilemma. "Our long-term ALM model concludes that on average, equities will help the fund to have a higher return, or that current bond yields are too low. But we also realise that equities can have poor performance or that bond yields could go even lower. We might think we can handle the risk over the long term, but in the IBM framework we come to decide whether we can bear the risk or that we need to take extra measures to limit the risk."

But Steenkamp is strongly critical of short-term solvency risk management. "I don't think it is sensible, because I think we are already looking too much at the short term. I try to avoid this. I don't think it is sensible for the financial planning committee to look every month at our coverage ratio. It leads to panic, bad decisions and pro-cyclical investment policies."

Is this criticism of risk-based asset allocation fair? At ATP, Gade-Jepsen concedes that Steenkamp's argument has some validity. "When you use the standard type of dynamic asset allocation approach, reducing risk when reserves are under pressure and increasing risk when you can afford it again, we found that in a crisis, you will always end up selling too late, at the bottom of the market."

But rather than give up at this point, ATP believes it can improve things, Gade Jepsen explains. "What we're doing is to try and be much more forward looking, to do things at early stages and in small steps, to avoid the negative effects of being forced sellers. We set up a model that looked forward in time in assessing the risks on a three-month basis, and once we go above a risk threshold we will reduce risk, just a little bit. At the same time, if risk is too low, we will increase it - but gradually. We did all kinds of ALM studies and the simulations showed that if you did it this way, the risks of selling at the wrong time were very small."

Having hedged its risk early, ATP can face the future with some degree of comfort. The Danish fund is now considering introducing market-driven minimum return guarantees. Meanwhile, with ABP and PGGM apparently set on firmly divergent paths, the pensions community will watch closely the kinds of risk management solutions they adopt, and just as crucially, whether or not Dutch regulators approve of them.