The Chairman (Mr C. Patel, F.I.A.): We have developed an insatiable appetite for wanting to learn just about everything about how we live, how we die, where we live, where we die, and what we do with ourselves in between. In order to feed that appetite we have today two leading experts on mortality and longevity. They are going to give us something more to think about, this time from an international perspective.

Our speakers are Dr Terry Sithole, who is at Kingston University, and Professor Steven Haberman. Professor Haberman, familiar to most of you, is Professor of Actuarial Science and Director and Deputy Dean at Cass Business School. We therefore have two very eminently qualified authors to present to us this paper on an international mortality comparison.

I now pass over to Professor Haberman to make some introductory remarks.

Prof. S. Haberman, F.I.A. (introducing the paper): Terry and I are pleased to be here. Richard Verrall, who is the third member of our authoring triumvirate, sends his apologies for not being able to be present.

I will say a few words of introduction and then Terry is going to go through the main results in the paper. Then we look forward to the discussion.

I should like to start with some acknowledgements. This is an update of a piece of work that we carried out in 2005–6, which was then funded by the UK Profession as well as some consulting firms. The update was commissioned by the IAA Mortality Task Force because new mortality tables had been adopted in many countries since that first study and, of course, population mortality has itself moved on. This work has again been funded by the UK Profession through the Mortality Research Steering Group.

The most difficult part of the study was actually getting the data, so a lot of work has been expended in getting the relevant data from the right people in the right country and then going back and asking questions and ensuring that it is up-to-date and that we really do understand the way the data are being utilised. I should also like to thank the people in the various countries who have been supplying the data to us.

This is an update of the study that we carried out five years ago. What we are trying to do is examine the assumptions used in the valuation of pension liabilities in a number of different countries as they appear in financial reporting. This can be different to the liability values used in pension funding.

We consider these results in relation to the underlying population mortality rates in order to see the strength of the mortality assumptions adopted for financial reporting purposes in the different countries.

The motivation for the study is that, as you know, corporate pension liabilities are a significant item in the balance sheets of many companies. They are of interest to regulators and analysts. Indeed, the Financial Times recently had a leading article addressing this point. Current accounting rules require transparency regarding these assumptions. As we have moved into an environment of much lower real interest rates, the importance of the mortality assumptions has become more significant. These assumptions are really quite material in the calculation of the pension liabilities.

The countries that we have included are as shown in Table 1 of the paper. Our methodology has been to take the EU and add to that a series of countries for which data were readily available.

We recognise when we look at different countries that the mortality assumptions are governed by two key things: the base level of mortality and the assumption regarding the future trend of mortality rates. So any differences are going to reflect differences, either in the underlying base population mortality and/or the trend assumptions themselves.

We focus on the assumptions used in each country, compare these with the underlying population mortality and try to illustrate the differences between the countries. We look back to 2005 and see if anything has changed. We also clarify what assumptions are mandated by the authorities in these different countries - in terms of both the base mortality and mortality improvement. We illustrate the impact of the mortality assumptions on a simple liability calculation. As you will see, some countries take a different attitude to mortality improvements from the attitude that we take in this country.

I now hand over to Terry, who will summarise the main results of the study.

Dr T. Sithole (also introducing the paper): As Professor Haberman has already explained, the bulk of the work pertained to getting the data.

Consider Figure 1 of the paper, which graphs the expected future lifetime from age 65 of various countries. We note that the country with the lowest expected future lifetime is Denmark, where a 65-year-old is expected to live for a further 16½ years. I should explain that we are concentrating on a 65-year-old male in this introduction, although the sessional meetings paper, and the more comprehensive background paper, includes other ages: 50, 60, 65 and 70.

Figure 1 shows the difference is about two years between Australia, with the highest expected future lifetime for a male from the general population, and Denmark, with the lowest. The UK is somewhere in the middle with a difference of about one-and-a-half years less than the equivalent in Australia.

Figure 2 in the paper shows the expected future lifetime for a 65-year-old male member of a pension scheme, based on the tables applicable to members of the pension scheme. We have a difference of more than 12 years between France, which has the highest life expectancy, and Denmark, which has the lowest. However, France is rather peculiar in the sense that the tables that are assumed for males are in fact the female mortality tables, and this explains the exceptionally large difference between France and the rest of the countries shown. If we exclude France, then we see that the difference between Denmark and Spain, the country with the highest life expectancy, is about eight years.

For the countries shown, the five countries with the lowest life expectancy - Sweden, Norway, Switzerland, the Netherlands and Denmark - did not provide an update of the tables that are used for pension-related benefits. Hence for comparative purposes, we have assumed the mortality tables currently in use in these countries are still the same mortality tables as the previous study, which might not be the case.

If we exclude those five countries, Belgium has the lowest life expectancy for a 65-year old male member of a pension scheme. We still have quite a large difference of five years between the highest and lowest life expectancy, compared with a difference of two years for the base expected future lifetime of the population.

Figure 3 in the paper shows the differences in years between the population life expectancy and the tables assumed for pensioners. We note that for France the difference is just over 9 years. In other words, it is assumed that a male member of a pension scheme in France will live more than nine years longer than a male member from the population in general.

In fact, you will notice that the countries with the highest future life expectancy tend to have the largest difference between the base (population) expected future lifetime and that for a pension scheme member. For example, if we consider France, it has the largest difference and also the highest expected future life expectancy for a male member of a pension scheme.

Excluding Norway, Netherlands, Switzerland and Denmark, we see that Japan has the lowest difference. This is because the tables assumed for pension benefits are based on the population tables; in other words the tables assumed for pension benefits are essentially population tables with some slight modification and hence the differences are small.

We also compared, in Figure 9 of the paper, the current expected future lifetime for a male aged 65 from the population and the results of the 2005 study. Here absolute differences are not that meaningful because, while the current population mortality experience relates to the experience in either 2006 or 2007, in the 2005 study the population mortality was based on different years for different countries. Canada is a typical example. In the previous study, the population mortality for Canada actually related to 1995, which was the latest available population mortality experience at the time.

A more useful depiction of this information is given in Figure 10 of the paper which highlights the annual percentage increases in expected future lifetime of the male population since 2005. So, we see that, on average, the Netherlands has the highest annual percentage increase in mortality at 2.6% and the least is Canada at 1.3%. Hence, if differences in mortality assumptions used for pension benefits were simply due to the underlying population mortality, we would expect the Netherlands to show the greatest improvement in mortality.

Figure 11 in the paper shows a comparison between the 2005 study and the current study of the difference between observed population life expectancy and assumed life expectancy for pension benefits. In some countries, we see that the difference has narrowed, whereas in others it has widened. In particular, the difference has widened in France, Austria and Germany, where we see that the difference is now greater than it was in the 2005 study. On the other hand if we look at Italy, for example, we see that the difference between the expected life expectancy based on the tables assumed for pensioners and the population life expectancy was higher in 2005 compared to the current situation. For Denmark, Switzerland and the Netherlands, where the assumed mortality for pension benefits is the same as in the previous study, we have a negative difference.

Having looked at the expected future lifetime, we now look at annuities. Here the discount rate assumed in determining the annuity rates is 3%. The annuity rates include a widow's pension on the assumption that the widow's pension is 60% of that of the male member's pension.

Considering the general population first in Figure 15 of the paper, we note that the picture is similar to what we had previously for the expected future lifetime, with Japan having the highest annuity rate of 16.3 and Denmark the lowest of 14.7.

Again, when we look at the annuity rates based on the tables assumed for pension benefits in Figure 16, we have a pattern that is very much the same as for the expected future lifetime. We have France with the highest annuity rate of 19.6 and Denmark with the lowest of 13.9.

Figure 17 of the paper shows the percentage difference between assumed male mortality table annuity and that of the observed male population. Looking at France, as an example, the difference is about 23%. As we noted before, the difference for Japan is the lowest after excluding Norway, the Netherlands, Switzerland and Denmark.

We then look at pension liabilities, comparing what would happen if we take the UK as a base, assuming that we have a liability of £1,000 million in the UK. What would be the comparative liability in the other countries? Figure 18 provides the answer. What we find is that in France, as we would expect, the liability would be a lot higher at £1,149 million because the assumptions for France are stronger. A liability of £1,000 million in the UK would be equivalent to £814 million in Denmark.

I should explain that the mortality tables that have been assumed for the UK are the tables for self-administered pension schemes, and we have assumed a long-term mortality improvement rate of 1%, and also of 1.25%. The results are not very different when we assume a long-term improvement rate of 1% p.a. or 1.25% p.a. As an example, when the mortality improvement rate for the UK is assumed to be 1.25% we have a liability in Denmark of £811 million compared to £814 million assuming a UK long-term improvement rate of 1% (see Figures 18 and 19).

Taking the UK as a base, we next ask the question: if you assume that the mortality assumptions for pension benefits are the UK mortality assumptions with a discount rate of 3%, what would be the equivalent discount rate required in order to get the annuity rate that we have for each of those countries?

So, for example, assuming a 1% long-term mortality improvement rate and a discount rate of 3% for the UK, for France, we would need a discount rate of 1.86%, whereas for Denmark we would need a discount rate of 4.81%. Figure 20 in the paper sets out the full set of results. There are similar results for an assumption of a long-term mortality improvement rate of 1.25% for the UK, see Figure 21.

In brief, these are the results that we have from the study. Detailed results are given in the paper and in the more comprehensive background paper.

The conclusions that we have drawn are similar to those drawn before. The practice varies considerably across the EU and the variations in mortality assumptions used for pension benefits are not really explained. It does not appear that the differences in, say, the base mortality, which is the population mortality, is a good explanation of those differences. Hence we feel that the differences merit further investigation.

There are some countries that allow for mortality improvements but the majority of the countries do not. Hence, to some extent, we are not comparing like-with-like. However, as these are the assumptions that are being used, and we are comparing annuity costs across countries, we do need to examine those differences and look at why there should be such huge differences.

Mr G. L. Jones, F.I.A. (opening the discussion): The paper is remarkably simple in concept. It sets out evidence of population mortality and simply compares it to current practice in the valuation of pension liabilities in a simple update of an exercise done six years ago. The paper notes the changes over that time. The summary paper distils around 300 pages of detailed analysis on those 18 countries, including Australia and Japan, into a much shorter document.

The first observation I would make is that the evidence on the ground is moving far faster than regulatory practice, and possibly actuarial practice. That I find disappointing. The linkage between evolving longevity and the renegotiation of the intergenerational transfers embedded in pensions systems, certainly across Europe, could hardly be more topical. In fact, today – 20 June 2011 - the UK Parliament debated a pensions bill that included a possible change to the retirement age.

For us as practitioners it is very helpful having this information collected into a single place. I see positive value in updating the previous paper.

But ultimately the interest resides in the implications that are not directly spelt out. What this paper highlights is a huge lack of consistency in practice. The most obvious ones are the allowance for future mortality improvements - both means of calculation and regularity of update - and, secondly, the lack of timeliness with many liabilities being valued from tables based on data dating back, to pick the US as an example, 20 years. One can contrast this to the 2007 sessional paper, Two dimensional mortality data: patterns and projections (BAJ 14,1, 1–18), where a common approach was applied to evidence across a number of countries, and that consistent viewpoint led to at least comparable conclusions.

The difficulty in establishing what current practice is, and often its lack of a relationship with any clear evidence, has led many of us to advocate far greater transparency and disclosure in the calculation and publication of pension liabilities. That lack of consistency has been highlighted by the authors in the apparent anomalies where present day evidence shows differences of a couple of years in life expectancy for a male at 65 but a far greater spread in terms of apparent regulatory practice.

The treatment of future improvements, to pick a particularly significant item, is very varied. In many cases, the practice is to provide some form of interpolation between short-dated improvements and some long-run assumption. What is remarkable, given the commonality of that approach, is the variation in outcome, particularly since the historic evidence shows a remarkably consistent 2 to 3 months per annum increase in life expectancy in each region.

The difficulty in obtaining the underlying data is a challenge. All of us who work in the area are familiar with that. In fact, both the Netherlands and Denmark have changed the basis from that described in the paper, with the more recent basis in the Netherlands increasing life expectancy of the order of three years from what is shown in the paper, and a similar move in Denmark. Certainly, there is a very wide spread in valuation practice, and I question what interests at all are served by marking liabilities at the extreme ranges shown.

The timeliness in reporting that we see in regulatory practice is an issue. The format of this paper is to set out a table at a specific point of time with collated data. I suggest that there is interest in maintaining an update of regulatory practice across countries but it may be, in the electronic age, that a database can update the underlying assumptions without the necessity of publishing them at a single point in time.

What do I do as a practitioner? I attempt to set the cost of what is a global and systematic risk in a consistent way. It is hugely frustrating to have to communicate with corporate and insurance companies who have completely inconsistent ways of dealing with valuing their liabilities and who therefore do not have the means to understand the liability that they carry and the associated risk, other than purely as an accounting rule or a funding target.

Organisations, such as the Life and Longevity Marketing Association, are attempting to introduce a consistent infrastructure to how we calculate population indices and potentially how a market might arise in this risk. They list the following qualities as desirable in such an index: tradeability; transparency; robustness; objectivity; simplicity; clear governance; timeliness; consistency; and universality. I suggest that these are attractive qualities for any liability valuation framework, and that is far from the case in what is currently set out in the various regulatory regimes.

I would also note that material longevity risk transactions do take place in the UK, but not in significant quantity in many of the other markets that we are talking about, which highlights one of the issues that inconsistencies can bring – they may be an obstacle to transactions. From a different point of view, the difference in valuation of liabilities provide an opacity to anybody who was interested in, for example, cross-border acquisition.

This paper can be viewed as simple record of how longevity liabilities are valued across countries. However, I see the value of this paper in the challenge it presents to respective actuarial professions and accounting practices charged with valuing such longevity liabilities. I thank the authors for so diligently laying the facts out.

Mr M. F. J. Edwards, F.I.A.: I wonder if you could give us some idea of the extent to which the differences you bring out in the summary paper relate purely to the difference in improvement rate, as opposed to the differing base levels. Also, can you give an idea of the extent to which the differences in improvement rates or bases seem to relate to any form of evidence?

Unknown contributor: I would just like to update the information in the paper on Norway and Sweden, where I do some work.

In Norway, the mortality tables have actually been updated in 2006: the latest tables are referred to as the K 2005 Tables. There are three static versions and one dynamic version which allows for mortality improvements. So in Norway at least there has been a fair amount of progress in this area.

In Sweden the mortality tables were revised in 2007. The tables are actually used for local accounting GAAP purposes, which also then indirectly leads to the taxation computation for the companies. The tables do not have mortality improvements but they were generational in the sense that there is a different table for every decade in which people are born - so very different tables depending on when you were born in the 1920s and so forth. These tables are currently being revised as well, with the latest tables being released on 1st July this year.

Mr A.G. Sharp, F.F.A.: I, too, would like to thank the authors for updating their earlier very worthwhile paper. As Professor Haberman said, in anything to do with mortality and longevity, data is everything, and getting up-to-date data is very important.

It is disappointing, despite the authors’ best efforts, that not everything is up-to-date in the paper. That observation comes on to supporting the opener's suggestion of more of a live database. Perhaps this is something which a Task Force of the International Actuarial Association (IAA) might be interested in taking on.

Where else is this paper is going, aside perhaps from the IAA? I would have thought that it is very important for each of the professional bodies in each of the component countries that were looked at to consider the findings. That might then facilitate greater international cooperation on this.

I wonder whether in perhaps one or two countries there are relatively small differences between assumed pension mortality and population mortality that, perhaps, is a function of the degree of coverage of pensions in those countries. The authors have explained the position in France as having a very different population. I wonder whether somewhere, like Australia, has much more commonality of population between the two sets of figures.

The Chairman: The idea of a live data base is a good one. But it is only as good as the people who contribute to it. You have to find people to contribute to it. I am reminded of the pensioner SAPs investigation where it was so difficult to sell what seemed such an easy idea to pension scheme managers, heads of consultancies, and so on. They needed a lot of persuasion that this was going to be worthwhile before they signed up to it.

Doing the analysis is the easy bit. Getting people to contribute sensible data, and regularly, is the difficult bit.

Mr C.G. Luckraft (student): You stated that there is not enough research done into mortality studies and mortality improvements. Have you done any sort of correlations between the number of studies that are done in other countries and how this differs to that of the UK?

Mr C.T Haines, F.I.A.: There is not much published that actually compares international mortality tables. So I do appreciate this paper's publication, although I also express my disappointment that it has not used any of the more up-to-date tables in places like Norway, Sweden, the Netherlands and Switzerland. They have all been updated in the last five years and there has been quite a lot of public debate on those changes. If the paper is going to be sent to other professional bodies, you may just want to caveat that point accordingly.

I spend a lot of time looking at tables that are used around the world. My firm does a lot of research in looking at mortality tables used by FTSE 100 companies around the world, and also the world's largest multinationals in the FTSE Global 100 Index. So there is data out there which can potentially be used for studies.

I wish to draw out two key points. You make the point that underlying population rates do vary by country. One table shows there is a difference, but actually not a lot of difference. It shows life expectancy for retirees is between 16½ and 18 years in most countries, which is a pretty small gap. The UK is almost right in the middle. So general population suggests many other countries have higher general population life expectancy than the UK.

So to the big question - which I often get asked by clients and journalists: why is it in the UK we are assuming that pensioners will typically have much longer life expectancy than those in pension schemes in almost every other country?

The studies we have done of assumptions used in other countries typically show that UK life expectancy is around 87 years for a male aged 65. In other countries it is around 84 to 86. We are doing some research at the moment which shows some recent jumps in the Netherlands. But actually the UK, when it comes to company accounts, is actually leading the way in terms of how long pensioners are expected to live.

However, is it really going to be true that UK pensioners will outlive pensioners in other countries? I have never seen any academic paper reach this conclusion.

If it is going to be true that UK pensioners will outlive pensioners in other countries, why is it true? Is it just because somehow UK companies avoid employing the people who are not expected to live for a longer time, whereas in other countries they are recruiting a more unbiased subset of the population? I do not think that is the case.

My hypothesis is that UK companies are typically using assumptions which are more prudent than around the world. In the UK we have big funded pension schemes. We have pension increases and we have spouse's pensions - two things not typically found in the US. We have trustees and we have the Pensions Regulator. Over the last five years there has been a gradual trend for trustees, pension funds and companies to adopt longer longevity assumptions and that is actually pushing up pension costs and causing, as a result, a lot of companies to change their strategy. Schemes are closing to future accrual, de-risking and transferring pensioner liabilities to insurance companies. This is even more so now that the cost of insuring can be close to that used for funding.

I think in the UK there has been a lot of awareness about mortality changes. As a result, we can probably say that our tables are more up-to-date and more debated than in any other country.

But the question to be asked is: are pension liabilities in other countries massively understated if calculated with the same sort of debate and rigour and external pressures of trustees and regulators as we have in the UK?

You mentioned the point that we have not looked at longevity trends and how much longer people will be living in 20 years’ time. I actually would commend such an investigation by this group or another party because life expectancy in 20 years’ time is another key assumption. Indeed, future improvements in Germany now are typically much higher than we see in the UK. I think it is probably imperative to have such a study to put all these results into context.

Let me make one final point. I am not too sure what is meant by “regulatory practice” as used by the authors. Are they assumptions that are mandated for local funding use? Are they minimum assumptions? Are they assumptions for converting lump sums to pension, or vice versa? Are they assumptions in countries where pension schemes are particularly defined contribution and there is very little defined benefit, unlike the UK?

Many large companies actually use their own assumptions because they have to make a best estimate, so these assumptions may not follow those required under regulatory practice. I think we need to be careful in comments we give to bodies like the International Accounting Standards Board.

I must say that it has been really good to have this paper, an update of the 2005 study. I think that there is a lot more work that can be done, not just by this group but also by the International Actuarial Association to bring all these strands together and answer the question: are UK pensioners really expected to live 2 to 3 years longer than pensioners in any other country?

Dr L.M Pryor, F.I.A.: I should like to thank the authors very much for this paper. It is always good to have more information available.

I should like to pick up a couple of points that have come along. The last speaker asked whether UK pensioners are really expected to live for two or three years longer. The answer may or may not be “yes”. But, remember, just because you are out-of-line does not mean you are wrong.

Possibly the UK assumptions reflect the expected UK pensioners’ experience better than those of other countries. But it is not only mortality assumptions which go into the calculation of the total pension liabilities. Obviously, there are many other assumptions being used: inflation, discount rates, and so on. One of the really interesting points made by the paper was how little difference the quarter per cent difference in mortality improvements made to the overall liabilities. It is important to keep things in context.

It is always important to remember the purpose for which you are making assumptions. This paper gives information about mortality assumptions used for pension liabilities in company accounts. If you are working for corporates internationally, there may be many other purposes for which they want estimates of pension liabilities. This paper gives one data point, if you like. Just because a mortality table is identified in this paper does not mean it is the right one to be used.

So I urge actuaries always to think from scratch, take all the available information into account, but to remember that they are themselves responsible for coming up with the right assumptions for the purpose for which they are going to use them.

The Chairman: Perhaps someone who practises on the international side might want to make a contribution, particularly in the context of mergers and acquisitions. There was a point made earlier by Colin Haines that, in the UK, awareness is at a pretty high level purely because of what we have had to go through over the last five or six years. That may or may not be true in other countries.

In an M&A situation you only get one chance and you have to work with it. So what do people do when they are buying subsidiaries, where there is a large workforce in different countries in Europe? Do they just accept these numbers as they come through, or do they have the benefit of other information, and what is that information?

Mrs D.R.C Siedentopf, F.I.A.: One of the saving graces is probably that not all countries have flourishing defined benefit schemes.

As somebody who advises companies on their reward strategy, it strikes me that there are two real questions here. First, are defined benefit pension plans really a good thing to have in a reward strategy? Second, in the emerging countries will the mortality trend of employees of big multinationals in secure jobs have a different experience in 20 or 30 years’ time when they reach retirement age from those of the indigenous population as a whole, because they will have different lifestyles and different ways of accessing medical cover?

When we look at morbidity statistics I think we would also see a big difference in different countries, which might have a spin-off effect on the life expectancy of retirees.

The Chairman: I see before me a mix of pension and insurance actuaries. Let me put into the discussion one or two more thoughts.

We have had a number of speakers commenting on the problems with the consistency of data. It seems to me that the insurance industry needed to tackle much the same problem recently and I wonder how they dealt with it.

I ask this question because in the run-up to Solvency II a great deal of time has been spent by the regulators and the insurance industry in trying to get the discount rates right to the last few basis points. Whether or not you include a liquidity premium may be a matter of fine tuning the discount rate by another 10 or 20 basis points, and I know that much time and effort has been devoted to acquiring and challenging the data for this. Consider this in the context of the uncertainty in the underlying mortality data, particularly in Dr Sithole's last chart which showed, even after taking out the extremes, what seemed to be a range of 150 or so basis points between different countries. That translates to something like a 30% difference in technical provisions! It is an observation that is just as relevant to insurance as it is to pensions. The insurance industry must have had to deal with it in their tightly drawn Solvency II measurement framework. It would be good to have some views on how they settled on this one.

Mr G.R Hibbard, F.I.A.: Just to give our insurance friends a bit more time to think about that question, going back to your earlier statement about M&A or international work, certainly a technique I have used often in due diligence is to apply UK tables to other countries but then to adjust by a rating effect.

You can get population data tables from the OECD which are very useful. So if you see, for example, that Dutch population longevity is half a year different, on average, from the UK population data, you then adjust that back through your model. It is well worth doing. You can get some very interesting comparative results and start to see those effects between country tables that allow for projected mortality and those that do not, and the degree to which that is actually going on within the tables.

Prof Haberman: I thought I might respond to some of the questions that have come up from members of the audience, particularly on the mortality aspect rather than the calculation of insurance or pensioner liabilities.

There was a very helpful comment about Norway and some other countries where the data is more up-to-date. It is very heartening, although frustrating, to hear this news. We can show you the e-mails from chosen nominees who refused to cooperate with us in those countries.

Someone asked whether we could say how much of the differences are due to base mortality differences and how much are due to differences in mortality improvement. We can answer this question in a descriptive sense. What would be interesting, however, would be to try to model that and be able to say what percentage of the difference is attributable to mortality improvements being present or absent. We can also comment on the extent to which the tables are based on solid evidence, and that is done in the 300 page extended version.

Then there was a comment on studies on mortality improvements in various countries. A lot of material has been published on mortality improvements in many different countries, including many of the countries that are covered in the current study. Many of those studies are academic in nature but they are very thorough. They cover for example Japan, Sweden and the US. The results are useful but deal with the full population, in each case, rather than a particular sub-set. But those studies all deal with population mortality. There is very little research that has been carried out and published on the relevant population subsets, that is, the populations that might have pension coverage through some private arrangement.

I guess that is the point, is it not? The general population is easier to define. These indices for the general population are calculated at particular points of time and hence are period-based indices.

When we look at the pensioner assumptions, they pick out a subset of the population. That is one element. They include or do not include mortality improvements. They are calculated on a period or a cohort basis. So the differences between the pensioner populations are much more complex.

Obviously, you have the nature of the pension provision, UK versus France being an example that we highlight in the paper. Then, as one member of the audience has mentioned, the difference between a best estimate and a prudent estimate is a further overlay on these assumptions. So it is quite complex to unpick all of these elements.

So, the differences between the populations themselves are expanded, with this rather narrow range of two years in life expectancy for males in the general populations (in Figure 1), to a range of over 12 years in the pension scheme sub-populations (in Figure 2) because of all of these other factors. We cannot say to what extent each of these factors is important without further analysis.

The Chairman: Would anyone like to offer some views on what some of the implications might be once you have gone past the data consistency problems? If you leave out some of the outliers in the analyses before you, are you seeing some useful messages?

I now invite Brian Ridsdale to say a few words on behalf of the sponsors of this project.

Mr B.P Ridsdale, F.F.A.: The International Actuarial Association Mortality Working Party expressed an interest in a comparative study of mortality tables used in defined benefit pension plans, and studied the 2006 paper by Verrall et al. The big differences at that point between the tables prescribed for use in some countries and the population mortality in that country was cause for concern, but more so was that some countries made no provision for improving mortality. It was also clear that the paper was getting out of date. Therefore the IAA Mortality Working Group asked for an update.

I brought the request back to the UK Actuarial Profession and was delighted that the Cass team was willing to update and extend the 2006 paper; and that the UK Actuarial Profession's Mortality Research Steering Group was willing to commission its production with a grant.

On behalf of the IAA, I should like to express gratitude to the authors and to the UK Actuarial Profession.

The Chairman has asked me to give you a brief update on the activities of the IAA Mortality Working Party.

It was set up in 2008 in order to compare notes and approaches to mortality internationally. It set itself the task of preparing papers, often of a review nature, and initially for the International Congress of Actuaries in 2010.

The areas we have covered so far include: sources of mortality data worldwide; the Society of Actuaries’ International Mortality Experience Study; mortality trends and uncertainty; pandemics; social and demographic stratification in mortality; and mortality analysis by cause of death.

It has to be recognised that these papers and any future papers will sometimes, as in the case of this paper, be state of the art; and sometimes they are aimed more at actuarial readers in countries that are not at the leading edge of developments in a particular area.

Work is progressing on studies bringing together other work on what to us are new areas: projection techniques; mortality of disabled people; mortality-related financial products; and underwriting.

We have also started the development of a website that will provide information on areas of interest in mortality. This can be accessed at www.actuaries.org/mortality.

This is at a very early stage and we are delighted to receive recommendations for papers that are of interest to international actuaries on the areas that I have just mentioned, and others. Contact details are on the website.

All UK actuaries are automatically members of the IAA. If you have not used the IAA website before or recently, then I suggest you log in. If you log in then you get a lot more data and you do not know the data is there if you do not log in.

Returning to the conclusion of the paper, despite the authors’ caveats, it does seem clear that the startling differences between mortality assumptions between some countries and others must lead to substantial discrepancies when you are attempting to put a value on a company where pension fund liabilities are a significant part of the whole. This is clearly of interest internationally.

The Organisation for Economic Cooperation and Development (OECD) last year set up a project to look at how pension funds and annuity providers and the regulatory framework incorporate future improvements in mortality and life expectancy. It aims at developing policy recommendations and best practices. In particular, it attempts to answer the following questions:

• 1. how do pension funds and annuity providers currently provision for longevity risk?

• 2. what mortality tables are legally required, and to what extent do they incorporate improvements in mortality and life expectancy?

• 3. how do liabilities of appropriate provision for longevity risk compare with the liabilities from actual practice and liabilities when applying the regulatory framework?

Today's paper will of course provide much helpful information to the OECD. Pablo Antolin of the OECD was disappointed that he was not able to accept our invitation to attend the meeting this evening.

I have noted the suggestion from Gavin Jones and Gordon Sharp regarding a live database, and also the caution from our Chairman, Chinu Patel, about some of the difficulties that could involve. I am wondering whether I might pass that on to the OECD as a suggestion.

In conclusion, I thank the Profession for putting on this research event this evening.

Mr S.L Wasserman, F.I.A. (closing the discussion): I join everyone else in thanking the authors for the paper. It is a very useful update.

The cost of pensions is a matter of major concern for most companies these days. Most companies are looking at ways to either control or to cut the cost of pensions as they appear in the balance sheet of the companies. That is really where the rubber hits the road, when potential investors and analysts look at their results. Against this background, the current study is a very welcome update to the original study first conducted in 2005-6.

As Professor Steven Haberman and Dr Terry Sithole have said, the authors’ aim was to review the relative differences between national mortality statistics and the mortality used in the calculation of pension liabilities within the accounts. They state that they had expected differences in mortality by country across Europe, and indeed we do find that to a very limited extent. If we pick up the expectation of life for a male aged 65 from the population statistics, we do find there is variation. We are looking at a range of 16½ to 18 years for males, roughly speaking, which is really not that big a difference, and a range of 20 to 22 years for the expectation of life of a female at 65. So this is a pretty narrow range for populations.

However, when it comes to the same country and the longevity for the same age for the purposes of the accounting results, the gaps widen. If we ignore France – and I will come back to France shortly because I think it has potential lessons for us – you get a range of 16½ to 23 years for males and about 17 to 27 years or more for females.

Based on these figures I should emigrate from the UK to Australia, because I will probably live for an extra couple of years. Better, I should emigrate to France because I can expect to live another nine years.

There are differences, and the differences are much greater when it comes to the financial results than when we are looking at the populations themselves.

As Steven Haberman has also said, we need also to be cognisant of other factors, such as the relative importance of the mortality assumption. For example, in France it is much less critical for the accounts for various reasons, than it is in the UK. So maybe the way they use mortality tables does not matter as much as it would do over here.

The fact that some countries will routinely adjust the mortality tables to reflect what is normal practice I do not think is picked up by the paper.

Then we also need to consider, coming back to France, the dangers posed by the recent Test-Achats ruling banning gender differentiation in insurance contracts and how this might impact annuity rates. Strictly speaking, it does not apply to non-insured annuities. But there are concerns that it is the thin end of the wedge and that it may only be a matter of time before it is applied to non-insured situations. Indeed, from what we see in France the population expectation of life at age 65 is 18 years, and that becomes 27 years. It becomes 27 years partly because of gender equalisation and partly because there is a margin between the female population longevity and that assumed in the accounts.

So it highlights the potential danger we could face if we lurch into gender equality and there is not a great deal of thought put into exactly what the impact will be.

In a slightly tangential vein, I notice as well that the paper is referring, inevitably, to mortality and longevity under defined benefit (DB) pension schemes. That is because that is what appears in the corporate accounts. We maybe should consider whether we should be extending this to include defined contribution (DC) schemes. I realise that the way things stand at the moment, DC schemes does not create volatility in corporate results. However, there are some “buts” here. The DB samples that we have of the population will get progressively smaller as the world moves towards an environment in which DC is much more prevalent.

We are also likely in the UK – and probably in other countries – to see the emergence of non-insured annuities when someone rather than an insurance company, either the member or the employer, is taking a bet against mortality in some way and maybe has some sort of flexibility. That could well end up having an impact on corporate results. Also, hybrid systems, such as cash balance, for example, blur the edges. They blur the differentials. Should we include cash balance or not?

Finally, Solvency II is a big unknown. Will Solvency II have an impact on pension schemes in the UK in the same way as for insurance companies? If it does, will it affect DC pensions in the same way that it affects DB pensions? If it does, could that mean that all of a sudden DC pension schemes will have an impact on corporate results? You will be holding some sort of solvency buffer reserve which may well have a mortality component contained within it.

So, we should give consideration to whether it is right to stay with just DB pension schemes or whether we should be looking in a wider context.

It is clear that in general this paper is touching on some very real issues for multi-national companies, where prudence against population mortality varies significantly by country. As the authors point out, and as various speakers have pointed out, it is a valid question to ask as to whether the discrepancies - or the differences in prudence - are really justifiable across Western Europe.

To my mind it is clear that there is need for further research, not into the statistics of mortality as much as into why there are these differences. Are these differences justifiable and should they remain or should we be looking to link the assumption of mortality within pension schemes to the population mortality in general?

Dr Sithole pointed out that mortality improvements averaged 1.4% per annum in the UK since the last study while the median was about 1.7% per annum. So, if the past is repeated, the assumptions used, as put forward by the Pensions Regulator, of an annual rate of improvement of about 1¼% to 1½% per annum actually do not seem very wide of the mark. The big question is whether the past is a good indication of what we should expect in the future.

As the Chairman commented, liabilities in a low inflation environment do vary considerably from country to country. 150 basis points I think was mentioned. That is a huge difference. It is a difference that is very difficult to believe can be justified.

Several speakers touched upon the question of the difficulties of obtaining data. I agree it is difficult. Maybe there should be a live on-line database, maintained by co-operation between bodies. If so there probably is a great deal more data that can be acquired quite easily without some of the pain that the authors have had to go through.

As Gavin Jones said, events are moving very quickly. It is really quite amazing thatiIf we think back about 25 to 30 years, mortality at later ages was seen pretty much as something that was slow moving and pretty static. But over the last period of time it has become clear that there are rapid changes. This study has really honed in on that, and it emphasises just how quickly things are changing. It is a challenge I think to get the right balance between the speed of reaction, not missing the fact that mortality tables are changing, versus taking long enough to make sure that our observations are robust and do show a genuine trend.

That leaves us the real question: are the differences between countries real or are they just a function of practice that has emerged in countries without any real basis in statistics? That surely is something which is worth exploring in more detail.

In conclusion, I think that the authors are definitely to be applauded for continuing this important research and for focusing our attention on an issue that has a very real impact for multi-national companies.

Prof Haberman (replying to the discussion): On behalf of Terry Sithole, I should like to thank those who have spoken for their helpful and very positive comments about what is quite a simple study. The suggestions for future work, both in terms of research and in terms of how this study could be extended into a database, have been very helpful.

The key thing, it seems to me, is allowing for mortality improvements and recognising that mortality is a process that is dynamic. The CMI came up with this idea in about 1924. So it has been around for 87 years.

I think that the UK Profession, without being complacent, needs to proselytise just a bit more and persuade sister organisations that this really is a very important factor that needs to be allowed for. I say “without complacency” because this trend that we see in the UK has been around since the early 1950s. It is a very consistent trend in the improvement of life expectancy. Gavin Jones quoted some well-known statistics about the pace at which life expectancy is increasing. We might not have paid much attention to it in the past because we have seen discount rates of 15%, when mortality trends are swamped. When the discount rate is 1% or 2%, mortality trends become rather important. These improvements have been there for some time and I think that they will continue and that they need to be allowed for.

The Chairman: Thank you very much. If I might add some comments of my own or, better, a wish list given that this is very much work-in-progress which is going to feed into work that the IAA is doing, and possibly into that of the OECD.

I would like to flag three areas. The first is consistency in measurement and reporting. It seems to me that this calls for better understanding and education among the providers of the data. It is an international educational effort that is needed and actuaries are the key players here. So I wonder if the IAA, working with the OECD, are the right bodies to pick up this challenge.

The second relates to the potential impact on regulation. If you recall my earlier example, there is vast variation between countries of between 100-150 basis points depending on where you draw the line on outliers. Is this reflecting underlying reality? This question emerged from the 2005 study and still remains unanswered. There is an urgency in answering it if only to steer the architects of risk-based solvency standards away from translating this uncertainty into large capital requirements.

Finally, we need to understand better what lies behind the surface once we have peeled away the uncertainty and the data inconsistencies. We need to learn more about the causes of different mortality. Why is there a difference between Japanese and UK mortality, for example? Is it the lifestyle and diet? Is it due to cultural differences? Is it due to geographic and climatic differences, or is it due to the social and welfare policies on things like health care, smoking, etc.? The point here is that in an era when we are breaking down all the barriers to globalisation, it would be natural to ask whether any of the factors for mortality improvements are transportable. If they are, then the future of mortality may yet be rather different from that suggested by our domestic investigations.

It remains for me to thank the speakers. Thank you very much for the splendid effort. Despite all those problems of trying to get hold of the data, you have made a very useful contribution to help us at least understand what the problems are and to get us to think about moving forward. I declare the meeting closed.