Sunday, July 19, 2009

Global Population Ageing - What Do We Know?

Update 1: I have added a comment below by Warren Sanderson who is one of the co-authors of the paper discussed below. It was first posted over on my personal blog where this entry has been up for a couple of days.

Update 2: Just to remind our readers (if they had not discovered it themselves) that there is a very interesting discussion unfolding in the post by Aslak about the correlation between higher overall fertility and non-marital births.


In this enty I am going to plug a recent paper on global ageing [1] by Wolfgang Lutz Warren C. Sanderson, and Sergei Scherbov published in the brand new journal, Journal of Population Ageing, edited by the team of demographers, Sarah Harper and George Leeson, at the Oxford institute of Ageing (Oxford University). Also, this would naturally be a nice occasion to ever so slowly reveal what I am arguing in the context of my own research; research which is naturally heavily affected by the likes of Lutz, Malmberg, Lee, McDonald and other of our time's great (econ)-demographers. [Graphs are made by the author and not taken from Lutz et al.]

With respect to Lutz et al they set out to grapple with one of the most difficult issues in the context of demographics and ageing in the form of our ability to forecast the future trends of ageing in the context of the global economy and different regions. The main conclusions indicate a rapid and, in some cases, accelerating process of population ageing in a global context in the next decades as well as a the very likely outcome that global population growth will cease to exist as a lingering phenomenon in the century which follows. Here is their abstract;

Population ageing is, in the first instance, a demographic phenomenon; although its consequences go far beyond demography. But the future trends of ageing are not yet known and many of the consequences of ageing will depend on the future speed and extent of ageing. Here we summarize what is already known and what is not yet known about future ageing trends in different parts of the world. We do this through the means of new probabilistic population forecasts. The section ‘New Regional and Global Probabilistic Population Forecasts’ presents the results of those forecasts. They confirm the earlier finding (Lutz et al., Nature, 412(6846), 543–545, 2001a) that it is highly likely that the world’s population growth will come to the end during this century. The following four sections present results for proportions of populations 60+, old age dependency ratios, proportions 80+ and average ages. In the section ‘New Measures of Ageing’, we analyse a new measure of ageing that takes life expectancy changes into account.

The first question which immediately springs to mind (or at least it did to me) is first and foremost what kind of techniques the authors are using in their endeavors to actually make forecasts on a variable as complex as ageing and so far into the future. Well, this is also where water gets muddied since it is important to realize, a priori, the amount of uncertainty which are attached to the kind of exercise Lutz et al. embark on. As most of you will know ageing is driven by the joint process of declining fertility and increasing life expectancy (and to some extent migration) and once we use this break down to operationalize ageing it gets dreadfully difficult to forecast although there are tentative conclusions which can be drawn based on existing evidence. In this way, it is important to dwell at the main conclusion, namely that the world as a whole is set to age rapidly before heading off into more specific forecasts.

The formal technique used by Lutz et al. is probabilistic forecasting which is basically a technique of collecting random draws from existing data on fertility, mortality and migration in order to produce a series of potential distributions and paths which can ageing can take as we move forward. Lutz et al. center on five simulations. The first in the context of total population size, the second on the proportion of people aged 60+, the third focuses on dependency ratios, the fourth turns the attention to the proportion of people aged 80+ and the final simulation centers on average age.

In relation to the first simulation which focuses on population growth it appears certain that the global population will continue to grow up until 2030-2050 depending on the path you look at. Beyond that point the simulations become extremely insecure ranging from a continuation of population growth to a decline in population growth which will restore the global population level at its 2000 level in the year 2100. Needless to say that such exercises are largely pointless from the point of view of inference about e.g. economic effects. However, what seems to survive as a main conclusion is that it is very likely that the tendency of population growth will come to an end some time during the next century, something which is significant in its own right in a world where the discourse on global warming, in my opinion, risks to engender policy advice to emerging economies which are essentially ill informed.

With respect to the paper in general I take away the following points in random order.

  • The potential importance of migration. In this regard, it is worthwhile to have a look at the simulated path for population growth (and thus in some sense the change in age structure) for Eastern Europe and Sub-Saharan Africa. As many of you will have guessed the former shows a steady decline (rapid ageing) from 2010 and onwards whereas the latter is set to increase (in all probability) towards the year 2050 (and beyond). Although uncertainty remains it is undoubtedly true that in a world of complete equalization of factor returns and with no barriers (of any kind!) to labor flows a lot of the detrimental impacts from ageing could be mitigated. Having said this though, I still want to emphasise the importance of managing the transition in an emerging market context since all evidence points towards a much more rapid move towards below replacement levels once economic development sets in. In short, population managment policies need to look at both sides of the coin.
  • A rapid increase in the proportion of 60+. Perhaps this is the most striking aspect of the coming process of ageing and whether we concur with the actual forecasts, there is no doubt that the change in age structure measured as an increasingly large share of, in this case, people aged 60+ is going to be extraordinary. According to Lutz et al. by 2040 the probability of the entire of Europe (CEE+Russia), Japan/Oceania as well as China having more than a third of its population aged 60+ is over 80%. This raises some extremely important important questions which researchers are only now starting to focus on. What happens to the labour supply of people in the 60s-70s? How does the life cycle of the young adjust to the skewed age distribution and what about that of the "old"? How should life expectancy be calibrated into the issue of raising retirement age for this growing age group? What happens to the productivity profile of society? And of course, my favorite; how will these changes be transmitted through capital flows both in a time series perspective (i.e. for the individual economy) as well as of course in a cross sectional perspective where the fact that all surpluses and deficits must add up act as a decisive binding constraint on the ability of some economies to smooth consumption and investment optimally.
  • The analysis on old age dependency (60+/20-59) ratio mirrors the points above to a large extent, but leafing through some of the tentative conclusions e.g. in the context of Japan and China I am pretty amazed, if not scared, by the projections. Especially, in the context of the former some of the projected paths would quite literally mean that the market economy (with associated welfare structures) will cease to exist. The interesting thing about the latter (China) is of course that for all the hopes about China ascending to take over the baton of the US as the sole global leader, most people are missing the fact that the country, in all likelihood, is going to catch up (and surpass) large parts of the OECD in terms of ageing as soon as the 2020-2025 mark.
  • Turning the attention to the forecasted share of people aged 80+ my own opinion is that this is pretty useless. I mean, I salute the effort but as the authors themselves point out; once we incorporate the long time horizons (up to 2100!) and most importantly the extreme uncertainty surrounding future mortality evolutions for this age group (both in terms of life expectancy at birth as well as on a cumulative basis for the people alive today) the uncertainty takes on huge proportions. What we know naturally is that the number will increase substantially, but also that it is likely to do so following a concave function of time as one would assume the extra gain in life expectancy gets smaller. [2].

Now, as Lutz et al never tire of pointing out throughout their excursion into the world of forecasting population ageing, this is all very uncertain and for two reasons primarily. Firstly, there is the inherent uncertainty associated with forecasting and essentially mapping changes in fertility, mortality, and migration. Secondly, there is the effect from moving the goal posts, or in this case; the reference points in relation to ageing. The simple point here is that it is not certain that old will mean the same thing tomorrow as it does today. This is especially the case of course if mortality continues to decline.

If this concludes the look on the paper by Lutz et al. I think it is worthwhile to produce something of an apology. Consequently, I have on several occasions pointed towards the futility (and in some cases stupidity) of narrating demographic changes in the context of what might and might not happen once the calendar shows 2050. Yet, here I even venture something about the state of affairs in the year 2100. An explanation is in order I think.

First of all, I maintain my view that the most important issue with respect to population ageing is the period one or two decades forward in time. Even if we might be able to plausibly say something interesting about ageing beyond this point the economic effects from ageing are correspondingly almost impossible to map. One good example here is the supposed asset meltdown scenario which is supposed to hit us in 2040-2050 as the weight of older age cohort's dissaving pushes up real interest interests and floods the market with assets (i.e. disinvestment) for which there is not a significant amount of takers. Not only do researchers only scantly understand what is actually meant by dissaving, there is also reason to believe that although optimal from an individual's life cycle perspective dissaving is not optimal from the point of view of an ageing society writ large [3].

With this general point in mind it is also worth recognizing a fundamental pre-requisite for focusing on ageing, namely the fact that the demographic transition is not over. It is ongoing and given the fact that the coming age of ageing which was initiated somewhere in the late 1960s (in the context of the OECD) it means that fertility will not stabilize at replacement levels. Despite the obvious reality of this point and the subsequent need to adjust models, views and analyses accordingly many still assume for example that global fertility, by some form of magic, is imbued with a drift parameter that will take it to replacement levels in the year 2100. Let me state as clearly that I can that such assumptions are completely useless. Interestingly and as a short digression, demographers have been focusing on this for quite some time and although the people at the UN, arguably, are getting better I would still recommend anyone interested in this to go back to the seminal volume edited by Jones et al and look up the chapter by Demeny where he ties the UN Population projections, of the time, up in knots. He essentially rams home the point that replacement fertility constitutes, in his own words, an implausible endpoint of the demographic transition; a point John C. Caldwell made already in the beginning of the 1980s. Whoever made the point first, this is extraordinarily important to take aboard.

In fact, I would take all this a step further.

In most modern accounts of the demographic transitions the traditional process is amended by an additional phase, or transition if you will, called the second demographic transition Van de Kaa (2002)[4] (or phase of ageing Lee (2003)). Now, I tend to take a more drastic approach. Quite simply I think that the transition need to be revamped all together and that a focus on ageing and age structure should be adopted at the offset. In this sense I think that Malmberg and Sommestad (2000) is a very important starting point since this the most comprehensive contribution which maps the whole transition in the context of ageing (specifically using Sweden). Working from this I believe that we can narrate the demographic transition in a way which makes it much more likely for us to use it to make inferences on economic processes which is ultimately my goal even if the study of demographics is a fascinating area in itself.

With respect to the question implicitly posed by Lutz et al. and thus how much we actually know about the global and regional process of ageing the initial answer has to be that we know quite a bit. Most importantly, you have to remember that ageing is not a new phenomenon and as I would argue a lingering aspect of the entire transition. The key is what happens next to mortality, fertility and migration and here uncertainty is vast although I should stress yet again that idea of convergence towards homoestasis in which these parameters are constant is not a desirable way to look at demographic processes. If we want to model this and if we want to make solid economic inference we need to take into account a myriad of feedback loops as well as the path dependency of the transition. With this end point it is perhaps apt to recall Socrates who reminded us of the fact that knowing that one knows nothing or very little is actually knowing quite a bit.


For more on population ageing, the Economist had an interesting briefing on the issue a couple of weeks ago which touches on some of the same points as above and provides some nice examples and cases. Finally, I also think that I should plus Georg Magnus' book the Age of Ageing which is really also a nice introduction for the intermediate laymen. In terms of academic papers that offers a general introduction to the issue of modern population dynamics and ageing Lee (2003) is a must, but also this paper by David S. Reher is very good (I discuss the paper here). Hopefully, I will have my own contribution to offer soon which sets the demographic transition in the context of economics and how it should understood in order to best make the connection to the study of economic phenomena.

List of References

Harper, Sarah and Howse, Kenneth (2009) - An Upper Limit to Human Longevity?, Journal of Population Ageing, Vol 1, issue 2.

Jones, Gavin W; Douglas, Robert M; Caldwell, John C; and D'Souza, Rennie M (1997) - The Continuing Demographic Transition, Clarendon Press Oxford; chapter 5, Replacement Level Fertility: The Implausible End Point of the Demographic Transition, Paul Demeny.

Magnus, George (2008) - The Age of Ageing, Wiley

Malmberg, Bo and Lena, Sommestad (2000) Four Phases of the Demographic Transition, ”Implications for Economic and Social Development in Sweden 1820-2000” Arbetsrapport/Institut för Framtidsstudier; 2000:6. The paper was presented at the SSHA meeting in Pittsburg. October 2000.

Lee, Ronald (2003)The Demographic Transition – Three Centuries of Fundamental Change, Journal of Economic Perspectives vol.17 issue 4, pp. 167-190 fall 2003

Lutz, Wolfgang; Sanderson, Warren C; and Scherbov Sergei (2008) - Global and Regional Population Ageing - How Certain Are we of its Dimensions, Journal of Population Ageing vol 1, issue 1. '

Reher, David S (2007) - Towards long-term population decline: a discussion of relevant issues, European Journal of Population

Van de Kaa, Dirk J. (2002) – The Idea of a Second Demographic Transition in Industrialised Countries, National Institute of Population and Social Security, Tokyo, Japan 29 January 2002


[1] - The paper is walled for non-subscribers, and I can't of course upload a version here, but mail me and we can discuss the paper "further" if you like ...

[2] - Although it should be noted here that in a recent literature survey on human longevity by Sarah Harper and Kenneth Howse evidence is provided, from Japan, that mortality shows no sign of "compression" at older ages which indicate a linear rather than a concave function.

[3] - I will have much more about this in my next posts where I respond to this.

[4] - The idea of the SDT dates back to the 1980s and earlier works by Van de Kaa as well as Boongaarts.


CV said...

This is a comment by Warren Sanderson:

Dear Claus,

Thanks for discussing our paper. It is part of a larger research program at the International Institute for Applied Systems Analysis in Laxenburg, Austria. You can see more about our work at and clicking on the tab to the left that says World Population.
There you will find a new dataset that deals with the history and future age and sex distributions by level of educational attainment. There is also an educational transition that has to be taken into account, when thinking about economic growth and capital flows.

In addition to producing probabilistic population forecasts, we have created new measures of aging for all countries of the world from 1955 to 2045. These can be found by going to and searching under the names Sanderson and Scherbov.

Further, we are in the midst of new research that links demography, human capital, and economic growth. I would be delighted to be in contact with you or others who are interested in any of these topics.


Safdar Ali said...

Nice post with great details. I really appreciate you for sharing this detailed information with us through this blog. Thanks. world population