Two years ago, Bill McGreevy had analyzed mortality data of WBG retirees for an article in the Quarterly. After a gap of two years, we decided to analyze data on mortality of World Bank Group (WBG) retirees in calendar years 2022 and 2023. To our surprise there has been a notable shift in the age at death of WBG retirees by gender. Previously Bill had reported in the Quarterly that in 2021 the average age at death was 82 both for males and females compared to the average life expectancy for all U.S. residents which was 76 for males and 81 for females. Bill had reported that COVID-19 had reduced the averages for residents of the USA but the pandemic had no notable impact on WBG retirees. Male WBG retirees lived significantly longer than the average for U.S. male residents, while female WBG retirees were slightly above the average for U.S. females.
In 2022 and 2023, total deaths were recorded of 261 male retirees and 128 female retirees. The average age at death of male WBG retirees increased further to 84.6 compared to U.S. male life expectancy of 76.1. The average age at death of female WBG retirees decreased slightly to 81.2 bringing it almost at par with U.S. life expectancy for females of 81.1.
The 2021 report had surmised that “Better health, nutrition and well-being among WBG staff explain the higher survival rates of Bank retirees.” We know that currently 10 retirees are over 100 years old and, according to Bank data shared at the 1818 Society Annual Meeting, five (5) percent of retirees and pension beneficiaries are aged 90 years or more. That is the good news. However, the gender gap has grown males achieving better outcomes than females.
We would rather not speculate on the reasons for the increasing gender gap in mortality rates of WBG retirees in 2022-2023. Suffice it to say that this runs which run counter to global demographic trends. We wonder why?
Comments are welcome!
Disclaimer
Member’s blog posts reflect the views of the author(s), drawing on prior research or personal experience. Freedom of expression is an essential part of the 1818 Society’s culture. The 1818 Society® is a nonpartisan, independent organization and does not take institutional positions.
KEYWORDS longevity, mortality, retirees, WBG
This is a test reply only.
Hi Anis,
many thanks for these interesting observations – even though a comparison of Bank staff post-retirement mortality rates with the mortality rates in the higher income brackets of the US would have been more meaningful than with the average mortality rates, also in the earlier eanlyses. The difference in mortality in the US between the highest and the lowest income brackets is well over 10 years as it were, more for women than men (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4866586/).
I am wondering whether there was someone who could take a deeper look at the data – to debunk-or-confirm two very persistent myths – (a) the the older you are when you retier from the Bank, the shorter you live (resulting in advice to retire earlier), and (b) women at the Bank have a higher rate of (breast) cancer and thus die earlier (during service and after retirement). These are stories that I have heard across the Bank throughout my entire time there – and I would love to see data on it…
Greetings from Europe!, Monika
Hi Monika, Interesting questions which would need access to HR data before we can do further analyses. Let’s see what we can get within the limitations of the Bank’s newly formulated Data Privacy Policy. Anis
Nice work, but the universes might be more appropriate if you compared life expectancy at retirement ages (say 65) for the US and World Bank retirees. Life expectancy in the US includes the (very premature) deaths of gang-bangers, thugs, military types, and other not-so-peaceful professions that World Bankers do not usually undertake. Paul Meo
Great to see the old WBG analytical juices flowing. I am sure all that mental stimulation will promote longevity, And skew the actuarials upon which the pension plan is founded. 😊
That said, I think that we need a larger and better dataset in advance of more hypothesizing.
I would be supportive of doing the analyses you describe if you had indeed established that there is a difference in mortality rates between male and female retirees. The aggregate average age at death is a meaningless indicator for establishing mortality differences when one compares two populations (male and female retirees) that have very different age structures. A higher mean age at death for males could mean higher mortality rates for males, higher mortality rates for females, or no difference. It is the age structure of the two populations that drives this as much as the mortality rates.
To determine what the mortality risks for male and female retirees are, you would need to do a simple calculation of the probability of dying by age. We do have the numbers to answer this. You already have the numerator (number of male and female retiree deaths by age) that you used in your analysis, so you need to compare these with the population at risk (number of male and female retirees at that age). I assume the 1818 Society has the number of people who are retired (by age and sex), otherwise Pension Administration would, I assume happily provide this to the 1818 Society. That will give you a close approximation of the risk of dying. My hunch is that the rates will not be in favor of male retirees.
The female advantage in mortality is quite strong at higher ages. For example, Black females in the US, a particularly disadvanted socioeconomic group, over the age of 70 have lower mortality rates than White males. White females have about the half the mortality rates of White males over the age of 60. These are from CDC life tables for 2020. You may also want to consider that the Staff Retirement Plan for 2022 on the 1818 website by the actuaries of WTW reflect substantially lower mortality for female retirees, based on the UN Joint Staff Pension Fund. So I would be surprised if male Bank retirees would be doing better than female retirees, but if they do, that would definitely be very unusual and a reason for looking further into this.
I am amused by your resistance to analyzing gender differences. We do not have access to the data on ages of retirees so making further analyzing contingent on satisfying your statistical benchmarks is neither possible nor necessary. We will however, share your comments with the new team that is likely to examine these data much more carefully over the next year or so. Maybe that will lead us to walk back the questions we raised, or we might have more robust analysis and more data to answer these questions more rigorously. Happy Holidays!
Anis, I dont think Ed’s comment shows any resistance to analysing gender differences, rather he offers methodological guidance for doing so in a demographically meaningly way.
Joy, Ed is not convinced that there are any differences in mortality rates between male and female retirees and he raises methodological objections to the variances from a demographer’s perspective. Indeed, we too expected female mortality rates of retirees to be lower than that of males. There are, however, several social (and economic) variables which disproportionately affect female retirees and may render them more vulnerable beyond the age structure of retirees. A more robust demographic analysis would of course have to be part of the analysis to check if indeed there are counterintuitive gender variances in mortality rates of retirees. But that alone will not be adequate unless other variables — grade levels of male and female retirees, age at retirement, marital status, etc. that may be making females more vulnerable — are also added to the inquiry. Anecdotal evidence seems to suggest those too are not equal across gender. A multidisciplinary analysis may be more useful both to investigate mortality rates but also to get at these underlying causes of vulnerability.
Anis,
Thanks for your kind wishes. I am following your comments to Eduard Bos with interest. I think recruitment historically would make a fascinating research project. There may be differences in the profiles of women and men recruited during earlier periods, in terms of access to health care and wellbeing ‘cultures’ relating to exercise, diet etc. both before WBG recruitment and after retirement , affecting mortality. Complicated, but interesting part of WBG’s gender legacy.
Regardless of the statistical relevance of the results, do you think there may also be differences relating to the country or origin / retirement and work status of women and men. Historically in the WBG women were far more likely to come from the global South (c.f. health and wellbeing) and also to work in administrative posts. The comparisons being made are with US citizens. Does this relate to WBG retirees who remained in the USA after completing their time in the Bank. Just some sociological observations which may well be qualitative, but you asked for comments on the results.
Hi Caroline,
Good to hear from you. And good questions, as usual.
It is pretty meaningless to compare life expectancy in the US with the very different measure of average age at death of WB retirees. At age 60, say, when many WB staff enter the population of WB retirees, some 17% of US males and 10% of US females of the life table cohort are already dead. WB retirees don’t become “eligible” to die until they retire, so no wonder we do better than the US population as a whole.
I agree with a previous comment that the sample size is too small to conclude much about trends. A few early deaths among female retirees could have thrown it off. Eduard Bos
I am not an actuary, but I question the comparator here.
For the sake of argument, let’s assume that WBG retirees retired at age 62. Some retired at the mandatory age, and stayed retired. Some took early-outs, and did not take other work. Others retired, and came back as STCs or took other employment.
The US life expectancy numbers are at birth, whereas WBG retirees had already made it to retirement age. A better comparator might be the Social Security life tables (https://www.ssa.gov/oact/STATS/table4c6.html), which estimate life expectancy at particular ages.
The latest available are from 2020 (i.e. the height of Covid mortality). For 2020, life expectancy at age 62 was 19 years for men and 22 years for women. For 2019, pre-Covid, the numbers were 20 and 23, respectively. So, absent Covid, a 62yo male could expect to live to age 82 and a female to 85.
In that context, the WBG numbers of 84.6 and 81.2 are much less impressive. The male numbers are – pardon the expression – dead normal, considering socio-economic factors. The female numbers are alarming.
Which brings us back to sample size. Either that, or working at the Bank is disproportionately hard on women.
John,
I find your reasoning sound. If we accept that premise that male WBG retirees do marginally better by 2 percentage points than their US counterparts, and women WBG retirees do notably worse by 4 percentage points. While these variances may well be a consequence of the small sample size I wouldn’t be as dismissive about this variance as some are suggesting. Maybe we need to dig deeper going back a few years, and also monitor results going forward. Perhaps a question for the new Gender TG Chair to consider.
One issue to consider is that, unlike life expectancy at birth, mean age at death is not age structure free. In other words, it matters how the population at risk is distributed by age. In high fertility, high mortality countries the mean age at death can be decades lower than life expectancy at birth, and the opposite applies in low fertility, low mortality countries (but not by as much).
This may very well explain the difference (and trend) between male and female mean age at death for Bank retirees (but one can’t tell for sure without looking at the data). Consider a hypothetical example. Imagine that because of past employment practices in the Bank, there were no female retirees over the age of 80 yet. Then even if female retiree mortality rates were very low, the mean age at death would always be below 80. Men, on the other hand, would typically be dying at higher rates, but because there are also retired men over 80 (and 90 and above), the mean age at death would be pushed up. The proportion of female employees at the Bank has been increasing, and I think so has the proportion of all retirees that are female, with the highest growth in the younger retired age groups. While it is highly likely that their mortality rates at every age is below that of men, the mean age will be pulled down by the higher proportion in the younger age groups.
The point is that mean or median age at death is a poor measure to analyze mortality risks. The proper way to do that would be to construct a life table for Bank retirees (not easy, given the different ages at which Bank staff become Bank retirees, but it could be done).
Eduard, One of the purposes of writing this blog was to generate debate around a factoid that is unusual and counterintuitive given the data and conventional wisdom that women outlive men. That purpose has already been served.
The point you make about variances depending on the country context cannot be the explanation since WBG employees come from a wide range of countries including high fertility, low fertility, high mortality and low mortality countries. What we see here (albeit from a small dataset that may not hold up as we get more data over time) is the uncomfortable possibility that this may be a consequence of WBG hiring practices. Caroline Moser makes an important point. Not only has the proportion of female employment increased, the profile of women in WBG employment has also changed from a preponderance of secretarial/admin staff to professional staff, and now to managerial and even senior management staff. But this change is relatively recent and there is probably a legacy of more women from junior grades in senior age cohorts compared to male retirees. So yes, we would need more systematic data but it is not just the numbers that count. We would also need to factor in economic variations between men and women, and institutional culture.
One could hypothesize that male retirees have lived more privileged lives and this would be more acutely true among senior age cohorts before the Bank introduced gender reforms. This may well be showing up in the age at death of males and females. Could be a fascinating research project.
The reason I brought up high/low fertility and mortality countries was only to illustrate how poor the “average age of death” measure is of mortality conditions, and to indicate how biased the average age of death statistic is because of age structure differences. I am sorry if this confused you. This has nothing to do with Bank retirees.
But my point still is: you can’t use average age of death data to show differences in mortality rates (or life expectancy) when the age structures of the populations at risk are different.
What your analysis shows is an interesting example of Simpson’s paradox.
You are missing the point. Setting aside any comparisons between WBG retiree age at death and life expectancy (since we agree they are very different measures) there is still the unanswered question of the possible underlying drivers leading to gender differences among WBG retirees. Yes, we can play the statistics game to say we don’t have the numbers to answer them but the uncomfortable reality is that statistics alone is incapable of explaining that differential. We would have to look at hiring practices and grade differentials, which would translated into economic benefits accruing to female and male retirees as quite likely driving some of these trends, besides family composition. Unless we start exploring the qualitative drivers behind these patterns no amount of WBG retiree data will yield statistically significant results. And I refuse to accept the argument that because statistics as a science is incapable to examining these questions we shouldn’t look elsewhere. On the contrary, all the more reason to look outside the statistical box.
Interesting. But perhaps someone with better statistical skills than I could say whether the differences – across gender and years – are statistically significant or not? The numbers are quite small.
I am pretty sure they are not statistically significant. Nor are the data on age at death strictly comparable to life expectancy figures. But they cannot be dismissed either. All previous reports on age at death have shown that WBG retirees tend to outlive the average age of U.S. residents, which should not be surprising given the higher economic wellbeing of WBG retirees. But t his is the first time that we observed the age of male retirees at death exceeding that of female retirees at death. It could well be a temporary aberration but is worth watching.