Why everyone wins with z-scores
The publication of the first performance test of the Dutch industrywide funds has put the spotlight on the Netherlands’ unique system of pension fund rating – known as the z-score.
The test is an assessment of how funds have measured up to their investment strategies in the years since the system began almost four years ago. All 65 mandatory industrywide pension schemes covered by the system passed the test in the sense that they did not fall below the minimum score prescribed by government. The significance of this minimum score is that if an industrywide scheme does fall below, employers who are members of the scheme can opt out and look elsewhere for a pension.
However, questions remain about what z-scores have achieved in broader terms. The scoring system has been both praised and criticised: praised because it is the first attempt to create ‘apples with apples’ comparison; criticised because it does not fully reflect the performance of asset classes, like real estate, that do not have a benchmark. Its critics say it has made pension funds too cautious, and too ready to hug the index. Its supporters say it has made them more aware of risk and the management of risk.
This matter will shortly be resolved. Following a survey of its members, the Dutch Association of Industrywide Pension Funds (VB) has presented a detailed account of the shortcomings of the z-score system to the Ministry for Social Affairs and Employment and is asking the ministry to consider making changes.
However, the VB emphasises that the system needs repairing rather than replacing. Overall, the consensus is that the z-score system has achieved the objectives it was set when it was introduced.
The z–score system was introduced by government legislation in 1998. The then Minister of Social Affairs wanted to retain the system of mandatory industrywide pension schemes but also introduce some market activity to provide incentives for them to perform well.
One option was to bring the industrywide schemes into line with the corporate pension funds. Roughly half of second pillar pension assets in the Netherlands is in corporate schemes such as Shell, Unilever, and Philips. Because these funds are controlled by the companies themselves, there is a strong interest in running them professionally.
However, the industrywide funds are different. Their members are predominantly employees on low incomes, and the amount invested by each employee is relatively small (E15,000). Keeping pension fund costs down is therefore important.
Roland van den Brink, one of the architects of the z-score system, and currently managing director of the E12.6bn Pensioenfondsen Metaalindustrie PMI/SVM says the issue of costs was crucial in the government’s decision to introduce z-scores: “If the government chose to go for completely open competition like the insurance companies this would mean that there would be profits and promotion costs to pay. This was not considered a clever idea.
“The government said that a far better approach was to make the market transparent and to punish poor results – by allowing employers to leave if the funds fell below certain levels. It would also enable economists and statisticians to analyse the results. Both forces will ensure that the funds will become more and more professional.”
The z-score system does not look at total returns. Instead, it considers returns relative to policy return - how the fund would have performed had it invested passively in its strategic policy weights and selected benchmarks. This allows funds with different liability structures to be measured in a comparable manner.
The system works in two stages. The first stage is the annual z–score. Each year, a fund’s board of management specifies a standard portfolio. The standard portfolio is an asset allocation model (ALM) that indicates the percentages of fund assets to be placed in various asset classes
The board then chooses an independent index that has a similar allocation of assets. A common example is the MSCI–Europe index. At the end of each investment year the yield achieved on investments is compared with the yield on the standard portfolio. The comparison allows for administration costs. The costs of the standard portfolio are 0.15% of assets
The second stage is the performance test, a multi-year average based on the sum of successive annual z–scores divided by the root of the number of years.
If the test result falls below minus 1.28, an employer can ask to be granted exemption from mandatory membership of the scheme.
The first performance test, for the period April 1998 to year-end 2001, covered 3.7 years. Next year , it will be extended to 4.7 years. Thereafter the test will be over five years.
So far no pension fund has performed below standard level. Part of the reason is the formula of the performance test. The yearly z-score is divided by the root of the number of years rather than just the number of years. The effect of this is that the formula becomes more demanding after five years than it is after three years. Therefore, the VB says, some funds can be expected to fail in 2003.
Another reason that funds may fail is that, under current conditions, the standard deviation – the extent to which the performance of an industrywide pension fund may deviate from the performance of the standard portfolio over a five year period – is too tight. The standard deviation was set using performance data of the Dutch pension funds over the years 1992 to 1996, a period of low market risk. The VB has therefore suggested that the formula should be changed to take account of changed market conditions.
So what conclusions can be drawn from the average performance test and the average annual z-scores? Have managers hugged the benchmark to avoid falling below the minimum performance level? Or have they pursued active management strategies in a bid to raise their annual z-score, and so improve their five year performance test?
Critics of the performance test as it now stands say it presents an excessively high statistical risk of undeserved failure. And in its current form the z–score rewards a passive investment policy. For example, Roderick Munsters, executive director at PGGM has suggested that the score does not reflect the PGGM fund’s performance because the fund invests in assets for which there are no standard indices – notably private equity and real estate.
The industrywide schemes appear to have added value by active management after cost, according to van den Brink, and the absence of standard indices appears to have helped rather than hindered the funds’ performance. “The positive averages for 1998, 1999 and 2000 are mainly due to the fact that for the illiquid investments, where benchmarking is difficult, funds can use TRR fixed income index plus 1% as a benchmark. In these years particular investments such as real estate and private equity did well compared with fixed income. In 2001 fixed income performed well, so there was a drop in the overall z-score.”
The statistics show that there is no correlation between the different years of the z-scores, he points out. “Out performance or underperformance is not consistent over the years. In fact, correlations between years are modestly negative. There is also no correlation between the asset mix and the z-scores.”
However, some funds show a large fluctuation in their yearly z- score, suggesting a high tracking error. “It can be supposed that before the z-score these deviations were even greater,” van den Brink says.
The overall conclusion is that funds are more risk aware, he says. Analyses by CEM - an extensive database with 300 funds and a seven year history - show that funds with the highest added value are those that take modest risk, typically with a 20% active and 80% passive portfolio.
He concludes; “Z-scores do not discourage risk-taking because the legislation says you may do what you want, so there is no obstacle to do active management. But you must take responsibility for it. The z-score system forces people to think twice before pursuing an active strategy. It’s not something you can do just for fun.”
The encouragement of a responsible attitude to risk is perhaps the chief achievement of the z-score system. Alain Grisay, managing director of F & C Netherlands investment managers, says “One could debate whether the technical formula was the right one, but in the big picture the z-score system been a plus. It has increased the responsibility of the funds’ boards of directors. Before z-scores there was little interest in terms of investment style. Now there is a growing focus on finding the asset allocation that will best fit the objectives of the fund.”
There is also more interest in risk management, he suggests “In the past there has been quite a bit of focus on the liabilities side. Now there is greater use of ALMs to establish the appetite for risk. There is also more emphasis on the definition of risk and the levels of acceptable risk. This was not often discussed not so long ago. Now people are willing to measure the sort of risk they are willing to take. It has bring some transparency to the measurement of performance.”
However, Grisay suggests that z-scores may have encouraged funds to choose a passive style of management at a time when only active management is likely to produce the necessary returns. “A large slice of pensions business is indexed, partly to fulfil z-score requirements. But if you talk to any manager of a pension fund in the Netherlands they will tell you most of the expected equity returns are going to be lower than they have been in the past. So hugging an index may be all right in terms of the z-score but it will create problems in terns of asset liability matching. You can’t fall asleep at the wheel.”
Could z-scores provide a model for pension funds in countries other than the Netherlands? Its proponents certainly believe so. They say that, with its strong emphasis on responsible investment and risk management, the z-score system is well-fitted to play a key part in any future regulatory system for European pension funds.