Retirement age and life expectancy

Authors: Servaas Houben


Publisher, publication date: AENORM, 2015-12



The desire from the Dutch regulator (DNB) to reduce the number of Dutch pension funds from its current level of 300 to about 100, has been mainly inspired by reducing both the cost of supervision and the sustainability of supervision: too many pension funds may lead to an overstretch in which proper oversight might become problematic. Although this approach might suit the regulator’s interest, and even that of the pension funds (as larger funds will reduce their administration expenses), this approach might overlook important differences between policyholder sub-groups. Servaas Houben recommends the DNB to switch its approach from its focus on the number of pension funds, to an approach in which pension funds represent homogeneous groups in society as this will limit the excess of subsidizing effects between subgroups especially between the well- and not-so–well-educated.

Background and scope

Due to the increase in life expectancy, the focus of discussions on a suitable pensionable age, have been limited to an age-discussion: as longevity is setting in, everyone is expected to retire at a later moment during their lifespan. However, in the entire discussion, issues around the expected years after retirement have not been tackled: a well-known issue for pension funds that don’t insure a homogeneous set of policyholders is that the lower educated members with a lower expected life expectancy may subsidize their higher educated fellow policyholders. Thus, is the increase in life expectancy a trend which is applicable throughout the entire population or are there specific groups which mainly profit from this trend? How can pension funds and regulators take into account these differences between participants without harming the principles of risk-sharing, and collectivity, and without discriminating against certain sub-groups?

Gender dependent state retirement age within the OECD

In the Dutch pension system, no differentiation is applied for the accrual of state pension benefits between men and women. Although the state based pension (AOW) was in first instance only provided for men[1], the same rights apply for male and female from 1985 onwards. Also when in 2013 it was decided to increase the pensionable age stepwise to the age of 67, a similar approach was taken for males and females. However, this equivalent setup is not embedded yet entirely in all OECD countries (let alone the rest of the world). In the OECD 2012 dataset[2], 14 out of 34 countries applied a different retirement age for males than females: in all these cases the female pensionable age was less than the male pensionable age.

Official retirement age Males Females
58 0 1
59 0 0
60 2 3
61 0 2
62 1 2
63 2 0
64 0 1
65 20 10
66 3 2
67 3 2

Table 1: retirement age across 34 OECD countries

Besides a couple of Eastern European countries, which might still have to further develop their pension system after years of Soviet influence, there are also some more “developed” countries which apply a different pensionable age for male and female such as Australia, Italy and the UK. Furthermore, there does not always seem to be a very clear link between the official retirement age, the life expectancy at birth and the life expectancy at age 65. The OECD uses the latter as a benchmark for expected pensionable years, as most countries from the dataset still have 65 as their official pensionable age, although some countries have already been increasing their retirement age since 2012. When looking at the relationship between male life expectancy at birth and from age 65, and the official retirement age, the pattern looks as follows:

Retirement age and life expectancy_males and females

Figure 1: retirement age and life expectancy from age 65 for males

The male population shows a positive relationship between retirement age and life expectancy during retirement (0.61). The relationship between life expectancy at birth and retirement age is slightly weaker (0.51) which seems to make sense as choosing an appropriate retirement age mainly depends on the number of years in retirement. So apparently within the OECD countries, there is a fairly good connection between the male official retirement age and male life expectancy during retirement.

On the other hand, for the female part of the population there is a weaker pattern:

Retirement age and life expectancy_females

Figure 2: retirement age and life expectancy from age 65 for females

The female part of the population shows a lower correlation between both the retirement age and the life expectancy at age 65 (0.38) and between the retirement age and the life expectancy at birth (0.31). This result makes sense as some countries are currently catching up by increasing their female retirement age to the same level as the male retirement age.

However when combining the male and female data sets a rather different picture emerges:

Retirement age and life expectancy

Figure 3: retirement age and life expectancy from age 65

The data set shows a surprisingly consistent result: for males all the squared scatter points are below the circle ones implying that retirement age is relatively high compared to life expectancy at age 65. For females the relationship is less strong. The lack of a clear relationship between life expectancy at age 65 and retirement age is reflected in the correlation of 0.07 showing significant subsidizing effects between males and females. Some of the more extreme examples are shown in the table below:

  Pensionable age Life expectancy after age 65
Country Males Females Males Females
Austria 65 60 18.20 21.50
Chile 65 60 16.80 19.90
Italy 66 62 18.90 22.60

Table 2: pensionable age and life expectancy for a subset of OECD countries

Due to the overall weak link between the official retirement age and expected years during retirement, one can wonder if a system of substantial inter-group subsidy is both desirable and sustainable in the long run.

Legal limitations for insurance companies

Despite clear differences in life expectancy between male and female, on 1 March 2011 the European Court of Justice has ruled out any price differentiation by insurance companies based on gender. This ruling implies that from 21 December 2012 onwards unisex premiums are applicable for new insurance contracts for both men and women. Looking at the data in figure 3, it does seem rather odd that gender discrimination is not allowed for insurance companies, while a considerable amount of OECD countries, of which some are part of the European Union, are still applying gender discrimination in their state pension benefit scheme. Although this does not imply that differentiation in retirement age by pension funds is prohibited, it does not seem likely this will be applied anytime in the near future.

Life expectancy and SES[3] factors

Jan-Willem Bruggink has performed a considerable amount of research in the Netherlands on the effects of SES factors on life expectancy. Especially for the effect of education on life expectancy, several studies have been published. One of his studies is for the period between 2005 and 2008 showing the following results:

  Life expectancy
Level of education Males Females
Primary school 74.1 78.9
Secondary school – non advanced 76.5 78.6
Secondary school – advanced 78.5 84.9
Tertiary education 81.4 85.3

Table 3: life expectancy and education level

Both for males and females there is a substantial gap between life expectancy for the higher educated compared to the lower educated. Furthermore, when assuming a standard retirement age of 65, the gap is even more emphasized:

  Life expectancy from age 65
Level of education Males Females
Primary school 9.1 13.9
Secondary school – non advanced 11.5 13.6
Secondary school – advanced 13.5 19.9
Tertiary education 16.4 20.3

Table 4: life expectancy from 65 and education level

The OECD 2013 report[4] confirms the significant impact that education can have on life expectancy: for males the average OECD impact between primary education and tertiary education at age 30 was 7.8 years for males, with extremes up to 16.8 years, while for women the average was 3.8 with a maximum of 8.5. The level of education has hence a substantial influence on life expectancy. However, there haven’t been any pension systems yet, which have recognized its effect by making the retirement age dependent on the level of education.


Despite the quite significant difference in life expectancy between Social Economic groups, there are some severe practical restrictions implementing these theories in practice. Poor physical health or bad habits (such as smoking) result in lower life expectancy and hence from a theoretical point of view should result in a higher benefit payment. However when people are actually benefiting from bad habits, this might result in the wrong nudges[5] in which people decide to make decision increasing their pension benefits, but decreasing their health prospects and increasing their old-age costs to society. Furthermore, it seems rather harsh to consider aspects as education as a purely nurture choice of individuals instead of a nature element one is born with. So even when the SES factors do provide clear evidence of differences in life expectancy, there are several social economic constraints which limit these from being implemented in practice.

Homogenous risk groups

Throughout this article we have been made aware of the difference in life expectancy between different parts of the population. Gender, education, and social habits have a clear and material impact on the expected benefits one is expected to receive during his retirement. The legal framework and social preferences, may limit the reflection of these findings into the actual benefits paid to different groups. However, a more simple and practical solution would be to reflect the heterogeneous composition of the population into the scope of pension funds: although there are several Dutch pension funds that limit their scope to a certain part of the insured population (e.g. notaries, dentists, medical specialists), this is not the standard approach. Instead the two biggest pension funds (APG, PGGM) consist of a heterogeneous set of policyholders ranging from highly educated to people with a limited educational background. The current focus of DNB is to increase the size of pension funds to simplify supervision and reduce administration expenses[6]: the latter occurs as bigger funds can benefit from economies of scale and hence reduce their administration and investment expenses.  As (administration) expenses have a compounding effect on the funds funding ratio, this seems like a sensible policy, however this one-size-fits-all approach overlooks the important differences in clients that pension funds have, thus resulting in inter-group subsidy. Also in the creation of the Dutch general pension fund (APF) the focus has been on reducing administration expenses and creating subfunds based on risk preference, not the expected years of benefit during retirement.

To avoid the issue of excess subsidizing altogether, DNB could consider limiting the scope of pension funds on sets of policyholders with similar backgrounds, and hence similar life expectancy. While focusing on purely gender might be a social undesirable goal, as one does not choose its gender, creating sub-groups based on other criteria such as profession or educational background might be a more practical solution. While the creation of more homogenous pension funds will reduce the diversification effects within funds, and might make the funds more vulnerable for trends for specific sub-sets of policyholders, this will result in a fairer premium and expected benefit level for all as one is paying premium more in line with one’s expected benefit payment. Furthermore, pension funds themselves could consider creating subfunds consisting of policyholders with similar backgrounds. This would enable them to still profit from the economies of scale regarding administration expenses and investments costs, but would allow them to provide a better combination of paid premium to received benefit level for their clients. Nevertheless, the current tunnel vision of reducing the number of Dutch pension funds resulting from a pure cost perspective is not in line with the more heterogeneous nature of most pension funds, and hence the pension fund members’ interests.


Arie Slotje and Servaas Houben, Zijn pensioenfondsen too big to fail?, De Actuaris, 1 April 2015

Bruggink J.-W., Ontwikkelingen in (gezonde) levensverwachting naar opleidingsniveau, Bevolkingstrends 2009; 57(4):71-75.

Hans de Mik, Gelijke verzekeringspremies voor mannen en vrouwen, De Actuaris, March 2012

Kardal, M., B.J.H. Lodder en M.J. Garssen, 2009, Levensverwachting stijgt, maar verschil tussen laag- en hoogopgeleiden blijft groot, Nederlands Tijdschrift voor Geneeskunde, 2009, blz. 153:A689.

OECD, Health at a Glance 2013, 2013

Richard Thaler, Cass Sunstein, Nudge – Improving decisions about health, wealth and happiness, 8 April 2008

Stam, S., M.J. Garssen, M. Kardal en B.J.H. Lodder, 2008, Hoogopgeleiden leven lang en gezond. In: Hilten, O. van, en A. Mares (red.), Gezondheid en zorg in cijfers 2008, blz. 9–19. CBS, Den Haag/Heerlen.



[3] SES: Social Economic Status

[4] Health at a Glance

[5] Thaler and Sunstein

[6] Arie Slotje and Servaas Houben

About Servaas Houben

I am a Dutch actuary and worked in the Netherlands for the first 4 years of my career. Thereafter, I worked for 2 years in Dublin and 4 years in London. I am now heading the actuarial department of ENNIA in Curacao.
This entry was posted in Articles, Behavioral finance, English, Longevity, Pensions, VSAE. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s