In the Netherlands, sharply rising life expectancies have been an issue since 2009. Every pension fund is using its own numbers, and the Dutch Central Bureau of Statistics (CBS) forecasts diverge from the Dutch Actuarial Association's. André de Vos tries to sort the confusion

"We know one thing for certain," says Peter van Zijp, president of the life expectancy tables commission at the Actuarial Society. "The life expectancy forecasts we have now will not turn out to be correct." There has been much ado about the differing forecasts offered up last year by the CBS, the Netherlands' central statistical office, and the Actuarial Association. The two models still yield different outcomes, but the difference in anticipated remaining lifespan for 65-year-olds in 2060 is now only half a year. Van Zijp says: "With a margin of error of 4-5 years, those kinds of differences don't matter. It is just very difficult to do precise forecasts over such a long period."

It is striking that for a long time there was almost no discussion about rising life expectancies, until the CBS came out with life expectancy forecasts that were significantly above the level anticipated in 2008. Then, in August 2010, the Actuarial Association's numbers turned out to be sharply divergent from CBS's. Whose numbers were to be believed? And what is causing the increase?

Van Zijp sums up the state of play. "You have a couple of developments that are intertwined," he says. "We've only been projecting life expectancy trends into the future for the past couple of years. Before that, we simply documented historical observations. Another development is that the annual increase in life expectancy has accelerated since the turn of the century. And then you have the differences between the data in our tables and CBS's. CBS publishes the most up-to-date data, which we then process. Processing the data takes time, so we are necessarily running a bit behind. Our August 2010 forecasts, for instance, are based on observations from 2008, while CBS was able to use 2009 data for its December 2010 tables."

Since the CBS numbers are always more up to date, it might seem redundant for other organisations, notably the Actuarial Association and the Dutch Association of Insurers, to come up with their own. Van Zijp explains: "It depends on what you're using the numbers for. CBS distinguishes causes of death and looks primarily at illness. Its model is used for demographic forecasts. But we're interested in forecasts and longevity trends for all the different age cohorts, which is information pertinent to pension funds."

Until 2007, the association published periodic mortality tables based on historical trends once every five years. Since then, it has also produced forecasts of longevity trends and now publishes once every two years. The changes to the model meant, among other things, that the difference between the forecast tables for 2005-50 and 2010-60 was very substantial. Over just three years, life expectancy for a 65-year-old jumped by two years.

Since 2007, pension funds worked with forecasts based on the assumption that life expectancy will continue to rise. Insurers, which do not have the luxury of being able to hive longevity risk off on an employer or premium payer, started looking at these issues earlier and place far greater emphasis on forecast accuracy. That is also why they look beyond 2060, the current cut-off date for CBS and Association forecasts.

For decades, life expectancy grew, on average, by around 0.2 years annually. Since the turn of the century, the rate of increase has accelerated to 0.3 years. The reason the association's life expectancy forecasts initially lagged behind CBS's was because it only decided to incorporate these new developments into its predictions in August 2010. "We still think the rate of increase will fall back to 0.2 years. But until then you have to deal with a faster increase," says Van Zijp.

À la carte mortality figures?

Pension funds can choose which mortality tables to use. They can also compile their own. The Dutch National Bank (DNB) does not regulate this, even though lower life expectancy forecasts lead automatically to higher cover ratios. "You need to have a consistent policy when it comes to setting your life expectancy," says a DNB spokesperson. "You can't use CBS's tables one year and the Association's the next if that should happen to make your cover ratio look better."

Pension funds are also required to use the most up-to-date mortality tables. In December, the DNB sent pension funds a letter saying they had to use the Association's tables for 2010-60 from August 2010 for any recovery plan they were considering. Last year, there were numerous pension funds that did not use the most up-to-date tables, which made like-for-like comparison of cover ratios difficult.

Most pension funds use the Association's figures. ABP chose CBS's numbers in 2005, because they were the first that allowed it to take into account trends in life expectancy. "Furthermore, CBS always has the most up-to-date numbers," explains a spokesperson for the fund. The fund therefore had to absorb a five percentage point shock in its cover ratio last year from sharply increased life expectancies.

This year, the increase in the CBS figures is much less pronounced and the effect on the coverage ratio is limited to just over one percentage point. Pension schemes that have begun using the new association or CBS tables for the first time in 2011 are likely to have to lower their funding ratios by several percentage points.

White collars
Almost all pension funds adapt CBS and the Association mortality tables for their own population. The DNB even insists on "adequately prudent fund-specific realised mortality rates". This makes sense, says Van Zijp. "CBS's and our numbers are based on the Dutch population as a whole," he says. "Pension scheme members on average have higher life expectancies. And there are big differences between pension funds as well. The higher the level of education, the higher the average life expectancies."

For ABP, life expectancies for 65-year-old men in the fund's own population are 1.7 years higher than the CBS figures. For women, the difference is 0.8 years. The method by which realised mortality is calculated is not so advanced at all pension funds. Here, too, insurance companies are ahead because they have to deal with the consequences of calculation errors on their own. "It's not so long ago that the corrections made by pension funds were very simple indeed," says Arnold Jager, a specialist in the area of life expectancy at Aon Hewitt. "White-collar workers were reckoned to live three years longer on average, blue-collars one-year longer. In the meantime, much more advanced calculation methods have been devised. We ourselves have a model based on insights from marketing. We're getting very precise forecasts from that. But there are few pension funds in the Netherlands working with those kinds of instruments." But Jager notes that interest in good correction models has grown over the past year. "Low cover ratios and the huge impact of life expectancy have jerked trustee boards wide awake."

The historical facts cannot be changed: it is better forecasts for the future that are needed. For the Association of Actuaries that means they will release new numbers more often in order to avoid overly sudden increases in life expectancies in future. Furthermore, future models will also incorporate multiple scenarios. It remains crucial that forecasts be based on historical data and not only on experts' (optimistic) future expectations. "There is far more focus on our work," says Peter van Zijp. "We feel the pressure that results from the growing importance of our findings. We aren't going to change our methods because of that - we're independent. But I do think we need to be clearer about how we work and how we reach our conclusions. It could be made much more transparent."