More About Why Japan Will Do Fine
[This is related to my series on Thomas Malthus’ “Essay on the Principle of Population,” first published in 1798. You can find an overview of all my posts here that I will keep updated: “Synopsis: What’s Wrong with the Malthusian Argument?”]
In my last post, I developed a forecast for the size of the population of Japan, and also for fertility there. I have a small simulation model in the background that creates a whole population of a few ten thousand that I then let develop with certain dynamics. That also means I can calculate all kinds of other quantities. Basically, anything where you need age, number of children or siblings or relatives more generally.
I would like to present some results below that I found particularly interesting. That’s also the quantitative side to my post “Fallacies about “Aging Societies”” where I argued the case qualitatively, ie. with some handwaving, and only with an analysis in steady state.
— — —
To recap what I do: I start with a population of 10,000 people who are all born at the same time. They then have children according to a level for fertility. Since there is a momentum effect initially (for an explanation of the concept see here), I let this run with replacement fertility for 200 years to get close to a steady state. The population is then about 27,000 people strong. Then I suddenly switch the dynamics on that I suspect actual populations pursue to get to a target size for population, and I set this target to two thirds above the actual level.
As I have shown in my previous post, this leads to pretty high fertility early on that then falls as the target size is approached. Since there is population momentum, ie. growth goes on for decades despite low fertility, there is some overshooting. As a reaction, fertility goes to below the replacement level for decades. But when the target size is then breached from above and there is undershooting, a “baby boom” develops. The size of the population oscillates around the target size with decaying amplitudes over the longer run and eventually converges to it.
The first part of the development that comes out of my simulation matches the actual data very well. The starting point would then be roughly in 1935. As there was no sudden switch then from the replacement level to the assumed population dynamics, the reaction in the simulation is somewhat too drastic at first. That’s why I cut the first twenty years away in my charts below. In other words: I start in 1955.
— — —
As noted above, I can calculate all kinds of quantities for the population in my simulation that involve ages and how people are related. Although I don’t assume I will get a perfect match for the population of Japan from 1955 until now, my guess is that I come pretty close as I am also able to capture the dynamics both for the size of the population and for fertility astonishingly well. Beyond the current time, my results depend on whether the model is correct as a forecast. But even if you don’t believe this, I think the findings are relevant for any population that shows such a behavior: oscillations that die down around a certain population size.
Here is now the first analysis that focuses on a topic that you see a lot of commentary on in the media: Japan aging! What happens with the share of old people and what is the relationship with those of working age who have to pay for them? The funny part here is in my view that much of the usual doom-mongering is not only wrong, but even so wrong that the truth is more like the opposite.
Let’s start with the share of the population below age 20 (exclusive) and above 65 (inclusive) as well as the population of working age between these ages. Those are the green, the red, and the blue curves, respectively.
The share of the population of working age in steady state, ie. for a stable population, is the value the blue curve oscillates around. The level is about 55%. As you can see, the amplitude is not very large. The minimum is below 40%, the maximum is almost 70% (but that is somewhat an artifact, see below for an explanation).
The first surprising result might be that the share of working age was very low in the 1960s, much lower than it will ever be in the future. In other words, that was the time when things should have been at their worst. From there, the share climbs to a peak in 2000. Basically, you have the best of two worlds: few seniors from a population that was smaller in the past and few children with low fertility now.
It is true, that the share of working age will fall from the high. But the interpretation is quite different from how it is usually presented if you see this in perspective. It will take until 2040 to reach only the normal level for a stable population. The development comes from an unusually fortunate situation that cannot last. That can only happen with high fertility that then goes down to low fertility. This windfall is often called a “demographic dividend.”
In the second half of the century, the share of working age will keep falling to a low. However, that is slightly below 50%, or only somewhat more than 10% below the normal level, hardly a dramatic movement, also a very slow one that you can prepare for as a society. With the “baby boom” I predict, there will be another “demographic dividend,” though a smaller one when the share of working age goes to about 60%, about 10% above the normal level.
Afterwards, not a lot will happen. While all this certainly warrants some adjustments for the Japanese, any predictions of fast decline seem completely unfounded. Human populations have had to deal with such shifts since the dawn of time. And they have always managed to cope. So it is hard to see why this should bring a rich country like Japan to its knees.
— — —
The drama here comes in because commentators usually focus on the growing share of older people. That is the red line in the above chart. Note that the sudden kink in 2000 is an artifact of the simulation. I switched high fertility on in 1935, which means very many children were born at first. Those reach an age of 65 in 2000. In reality, the transition should have been much smoother.
It is true that the share over 65 will rise in Japan, but only from a low around 10% of the population to somewhat more than 25%. It is tempting to extrapolate this short-run trend, but that is wrong. Actually, the increase slows down and comes to a halt around 2040. That is so because fertility will rise again according to my forecast. Afterwards the share of those over 65 will fall to about 15% at the end of the century.
As you can see, you have the same oscillations as for all other quantities, and the level for a stable population comes out somewhat below 20%. A share of only 10% was a windfall for the Japanese that could not last anyway. Most of what you see is a reversion to normal levels. If you view it from this angle, what happens is far less dramatic than how it is often portrayed. There will be a rising share of seniors for some time, but not much more than for the long-run level and a stable population.
But then the common confusion is anyway to only take seniors into account and ignore children. Those have to be taken care of, too. According to an analysis by the USDA (why do they do this?), it costs between $170,000 and $390.000 to raise a child to age 18. Now, compare that to a senior. Life expectancy at age 65 is perhaps 15 to 20 years. In other words, raising a child is like paying a pension of between $10,000 and $20,000 to a senior on an annual basis.
Hence a society that has more seniors, but also fewer children has a net effect that is much smaller than the share of seniors would indicate. If the two shares for seniors and minors mostly cancel out, the impact could be quite small for a society. Figures may not be the same for Japan, but the general relationship should not be all that different. It is not correct to focus only on seniors as if everything else remained the same. Good to create panic, but that’s misleading, at best out of ignorance or at worst deliberately.
As you can see, the two sides cancel out to a large extent: More seniors means fewer children and vice versa. That’s why the effect for the share of working age is much more modest. If you also keep in mind that you should also net expenses for children and seniors, the whole issue seems much less dramatic than it often sounds.
— — —
Children have also another consequence: Someone has to take care of them. Realistically this shoots one person, and mostly a woman, out of the labor force for something like 10 years after a child is born. That means that the share of working age is not yet the right reference point. We should take the actual labor force instead.
There are also other people who for various reasons are not in the labor force although they are of working age. As far as I know, 90% is something of a maximum for those who are actually employed of those who could be. Hence I scale the share of working age down by 10%, and I also take all women out with children below the age of 10. If you don’t like my bias here, taking out men and women equally or men only would lead to the same result for the shares.
Here is what I get:
The blue line is for the total share of working age, the green line for those who are actually employed. As you can see, the green line comes closer to the blue line in the phase until 2000. That should be so because fewer women have children and also fewer of them when they do. So more will be employed. You can see the same result here as for other quantities. The unusual development is not what comes in the future, but what happened in the past. You have an extraordinary value around 2000, and then things normalize. The share in the labor force for a stable population would be much lower, around 45%. Note that the peak is somewhat of an artifact in the simulation.
To make this more transparent, I plot a chart for labor force participation next, which is the ratio of those in the labor force to those of working age:
As you can see, labor force participation is very high around 2000 and that will go on for about 30 years. While this is not perfect, it comes pretty close to what really happened. Japan indeed has very high labor market participation now. This could only happen because more women were in a position to seek employment with fewer children to take care of.
Compare this to the 1960s, a time with still high fertility where labor force participation is lower by about 10 percentage points in the simulation, but I guess also in reality. Even at the trough in the second half of this century, and that’s with a “baby boom,” it will not go as low again. The normal level is around 74%, and most of what happens is again a normalization from an extraordinary situation and modest fluctuations around a long-run level.
Some people comment on such developments as if you could pick and choose: How about having such a high labor force participation as in Japan AND very high fertility, too. But that is not independent. If you move one, the other has to follow. You can’t have it both ways. That’s also so with financing seniors and children. Both sides move in opposite directions. Sure, you can reduce expenses for seniors with higher fertility, but then you also have to pay more for children or vice versa with lower fertility. Once you realize this, many of the hot takes you read sound outright silly.
— — —
We can now connect the two results to calculate so-called “dependency ratios,” ie. how many people have to be taken care of on a per capita basis. The reference point could be people of working age or people in the labor force. Here is a graph:
The blue line is dependent people per capita for people of working age. Those are minors and seniors. The long-run level is about 0.8, that’s the normal state. Japan was again in a particularly fortunate situation around 2000, when the dependency ratio fell to below 0.5 (at least in the simulation). That will change in the future, but not dramatically so if you view it from a realistic perspective. At the peak, the dependency ration will top out at slightly more than 1. This may look high viewed from an extreme low, but only moderately high versus an appropriate comparison with a level of about 0.8.
But then people of working age are not necessarily the adequate reference point here. We could also take people in the labor force. That would mean the effect from relatively more women in employment softens the moderate blow even further. Of course, to a certain extent this is only a shift from unpaid labor to paid labor. But still it is something that can be tapped into for financing exactly because it is now paid labor.
Arguably, I should also include unpaid work in the analysis when seniors are taken care of. That also means more people not in the labor force. I could accomodate this in my simulations, too. But my point here is only to get a qualitative feel for what is going on. Mostly this would lead to only some modest adjustments and leave the qualitative result unchanged. Since I also take out 10% of those of working age as a matter of principle, dependency ratios per those in the labor force must be considerably higher than per those of working age. But then variation is also lower because there is an effect from more women in the labor force.
The normal level here is around 1.5. The peak is roughly 2, and the bottom 1. So the move from the low around 2000 to the high some time between 2080 and 2090 is by about 100% whereas relative to those of working age, it is somewhat more. But then I also predict a “baby boom” that cuts against extraordinarily high levels of labor market participation. Yet, what comes until the middle of the century is only a normalization, and then there is a modest increase by a third over about half a century beyond it. Think about it in this way: If there is only moderate economic growth over the next half century, that already pays for the extra burden versus a realistic level and probably easily.
— — —
Let me finally discuss what is perhaps the pet peeve with commentators about Japan: It is an “aging” society! First off, societies do not age like individuals. An individual can age to 80 years and then die, a society can’t. Mean age for a society can go up and it can also go down. Unfortunately, the latter never happens for individuals.
Here is now an analysis for mean age. I could also do median age as it is mostly framed, but I was too lazy, and the difference might not be all that large:
The blue curve is for the mean age of the population in my simulation. That comes pretty close to reality. Japan currently has a median age of 47.3 years, and that has been rising for some time, pretty much in line with what I get. However, here is the catch: The trend will not go on! Mean age will increase a little further, but only to a peak. Afterwards it is going to fall again, actually to the upper 30s at the end of the century in my forecast. Note that this conclusion depends on my prediction of a “baby boom.”
But then if you look at the level over the long run, ie. for a stable population, that is about 39 years, so this is a level that is the natural level for such a population anyway. In other words, a mean age above 40 years may look dramatically old compared to an exceptionally low mean age in the 1960s of almost 30. However, that is somewhat of an optical delusion. The high watermark comes across as much less impressive versus a realistic benchmark.
And it gets worse for doomsayers. The facile conclusion is that Japan has a senile workforce and that will do the country in. That’s why I have also calculated the mean age only for those of working age here. This is the green line. Now, if you look at the graph, all this handwringing looks even less impressive. The normal level over the long run is something like 41 years. And the peak will be 44 years, only three years more.
Such small movements that are even drawn out over decades should be hard to spot in a society. The idea that the workforce in Japan is or will be a lot of tattering people who show each other their new dentures is nothing but a delusion that stems from misreading the mean age of the population (or median age if you will). Of course, in a phase where the share of seniors goes up while that for children goes down, the mean for the whole population may shift up considerably. However, that does not matter all that much for those of working age. You don’t have many 80-year olds in offices at any time. And neither were there lots of little children jumping around in the 1960s when the mean age was so much lower.
— — —
All in all, the conclusion from my simulation is very unpanicky for Japan. You may not go along with my forecast, but the results up until our time are not too far away from reality. And even if you balk at my model, a comparison with a steady state for a stable population is perhaps not an absurd reference point while taking an extraordinary situation as normal is seriously misleading.
In general, all developments are very slow and work out over decades. There are no sudden leaps. Versus a normal level, all developments look exceedingly modest. Extrapolation of short-run trends is not only not warranted, but so misleading that it is hard to explain without a fervent wish to talk up a catastrophe. And then it all only results from a major buildup for population in the past century, which is a one-off that is already baked into history and works out over the next century.
Mean age may rise a little, but then it will top out and fall again. There is no unidirectional development here as for an individual where a society can only grow older and that is one year for every year that goes by. It is not like the Italians put it so cutely when someone yammers about how time flies: “Ogni anno ne passa uno.” (Every year one passes.) And then mean age is mostly irrelevant for the productive part of society. The share of working age only fluctuates modestly around a level, and that is even cushioned because labor force participation goes up when more women are employed with fewer children.
There will be more seniors relatively that need to be taken care of. But that will not go on forever and even turn around. And on the other side, much of the costs net with fewer expenses for children. Already very unambitious economic growth can make the extra burden pretty negligible on a per capita basis for those who work. And then this will also turn around. No society will be able to escape such a situation anyway over the longer run. It might happen sooner in Japan, but that does not make it different from what will also take place in other countries.
My conclusion from all this is that this whole panic about demographic developments is a storm in a teacup (or a “water glass” for Germans who are not as much into tea as Americans). The only thing that makes societies really miserable is the doom-mongering and talking temporary trends up as if they went to infinity. The only demographic catastrophe in my view is the nonsense you are fed about these things on a daily basis. Japan is a country with a great future ahead.
