By now, there is a pretty widespread feeling that in many developed countries, getting a job now is harder than it used to be during the postwar boom (approximately 1945–1973).
This feeling is backed up by data. For example, here are the average annual hours worked per worker from 1950–2017 for the G7 countries (data for US and France start in 1950, with other countries starting later):
Over the past seven decades, in every G7 country, the average worker has on the whole been working fewer and fewer hours. The greatest drop in this group is seen in France, where the average worker has gone from working 2,214 hours in 1951 to 1,514 hours in 2017. This is equivalent to a drop from 42.5 hours of work per week down to 29.0 hours per week.
Since work is the primary way most people make money, work scarcity leads to many economic problems. Employers become more powerful, while workers lose bargaining power. Wages stagnate or fall. Poverty worsens, as more and more people are unemployed or underemployed. Consumers buy less, weakening the economy and making employment prospects even worse.
Work scarcity can also lead to social problems. With no steady income, more families fall apart. Poor job prospects make it harder for young people to establish themselves in society. Support for the existing order drops, which can lead to political instability.
Why has this been happening? Is it because there is less work to do now than before, or because there are more workers now than before? In other words, is it because of decreasing labour demand, or increasing labour supply? The answer to this question affects our search for solutions for it.
In his work, Peter Turchin (University of Connecticut) has shown how societies, both ancient and modern, experience cycles of rising and falling labour demand relative to labour supply. When labour is in short supply, wages are high, the gap between rich and poor is small, and politics are generally stable. But when labour is abundant, the reverse happens: wages are low, economic inequality rises, and politics become increasingly unstable. These cycles, which he calls secular cycles, are very long, with one full cycle spanning anywhere from 150 to 300 years.
In particular, he has found evidence of secular cycles occurring in the US, which he details in his book Ages of Discord: a Structural-Demographic Analysis of American History. In it, he shows that labour demand in the US was outgrowing labour supply before World War 2, but began to lag behind it sometime in the late 60’s:
Turchin has a great blog post in which he discusses this graph and its significance. Here, I’m interested in the way he measures labour supply and demand. He takes labour supply to be the size of the labour force, including those who are employed and those who are unemployed but looking for work. Meanwhile, he measures labour demand by the number of full-time equivalent workers actually employed. This is given by the total number of hours worked in the US divided by 2000, the approximate number of hours in a year of full-time work.
This methodology is built on an assumption: that how much workers work is a reflection not of how much workers want to work, but of how much work there is available. The decline in average work hours in recent decades has often been interpreted as an increase in leisure time and an improvement in our quality of life, as in this CNN article from 2013. But as the same article explains, many people’s work hours are reduced not because they can afford to work less, but because full-time work is becoming more scarce. Instead of working full-time, more people are instead working part-time (which generally pay less), going back to school or staying in school for longer, or retiring earlier.
There are two other observations that are better explained by this interpretation of declining work hours as work scarcity rather than as a greater preference of leisure. First, it explains why, as the CNN article mentions, people with higher incomes work more hours, not fewer. Second, it explains why, as we will see in the data in this article, work hours decline the most during recessions, but rebound when the economy recovers.
In short, the only thing worse than work is not having enough of it.
Using Turchin’s methodology as a starting point, I decided to look at data compiled by OECD (Organisation for Economic Co-operation and Development) for six developed countries: US, Canada, UK, France, Sweden, and Japan. Of these, five are members of the G7. Since OECD data for Germany and Italy do not go back as far as the other countries, they are replaced by Sweden, another European country with a smaller but equally strong (if not stronger) developed economy.
Specifically, for each country, I looked at five measurements:
- Average hours worked per worker per year, as a measure of how scarce work has become.
- Total annual work hours, as a measure of labour demand. This is calculated by multiplying the average hours worked per worker per year by the total number of employed workers.
- Working age population (ages 15–64), as the first factor affecting labour supply.
- Employment rate, as the second factor affecting labour supply. Driven by the increase of female participation in the workforce, the employment rate has increased significantly in many developed countries, and is therefore an important factor to examine. I also compare male and female employment rates.
- Total number of employed workers, as a measure of the amount of labour actually being used.
Among the six countries, those that are close geographically show similar patterns: the US and Canada share a similar pattern; the UK, France, and Sweden share another; and finally, Japan shows a unique pattern all on its own. We can therefore consider the data in these three geographical groups.
Since the 50’s and 60's, both the US and Canada have seen significant declines in the average number of hours worked (per worker per year):
Since the 50’s and 60’s, the average hours worked per worker have declined. The US has gone from 2003 hours of work per year in 1955 to 1780 hours in 2017 (a decrease of 9.6%), while Canada has seen an even bigger drop, from 2059 hours in 1961 to 1695 hours in 2017 (a 17.7% decline). For reference, a full-time job with a 40-hour work week for 50 work weeks per year is equivalent to 2000 hours of work per year.
In the US, the largest decline in the average hours worked began 1965, while in Canada the start of the decline dates back to at least 1961. In both countries, the slide lasted until the early 80's. Then, both countries saw a slight rebound larger in the US), which held steady until the early 2000’s. There has been a slight decline in both countries since then, but much smaller than the huge slide in the 60’s and 70's.
For both countries, this decline in the average work hours per worker is evidently not because there is less total work available. The total number of work hours has been increasing quite steadily for both the US and Canada since the 50’s and 60's, with only slight dips during recessions (the 2007 recession being more noticeable):
But while the total number of work hours has been growing, the total number of employed workers has also been growing:
The ascending lines representing the total number of workers look similar to the ones representing total hours worked; the correlation between total hours worked and total employees is 0.999 for both countries. However, the decline in average work hours shows that the number of workers has been growing faster.
The total number of workers can be broken down to two factors: the working age population (all those aged 15–64), and the employment rate (the percentage of the working age population that is actually working). First, the working age population, which has been rising smoothly and steadily for both countries since the postwar boom:
To see how much this increase in the working age population alone has effected the amount of work available, we can see what would have happened if the working age population had not changed — that is, if only the employment rate varied, while the working age population stayed at its value in the first year of data. Here is what would have happened:
With only the employment rate changing, average work hours would have increased drastically. In the US, the average worker would have ended up working 3684 hours per year by 2014, compared to 2003 hours in 1955. Similarly, in Canada, the average worker would have worked 3720 hours by 2014, compared to 1879 hours in 1961. (As a reference, 3500 hours a year is equivalent to 70 hours of work per week, assuming 50 weeks of work per year). Of course, this is not a realistic scenario, since not many people would want to work 70-plus hours a week. These numbers merely serve to reflect the effect of the working age population on the labour supply.
Next, we look at the second factor: the employment rate. Like the working age population, the employment rate has also generally been increasing, though in the US it has stagnated and reversed somewhat of late. Since the actual employment rate figures provided by the OECD do not go back very far, I calculate the employment rate by dividing the total number of employed workers by the working age population:
In the US, the employment rate grew from just above 60% in 1955 to almost 73% in 1990. However, it then basically held steady until 2007, when it dropped quite significantly during the recession. Despite this drop, the last employment rate figure was still 69.2% in 2014, significantly higher than in 1955. Meanwhile, the employment rate in Canada has increased even more significantly, from 54.4% in 1961 to 73.4% in 2014.
The increase in the employment rates of both countries have mostly been driven by the increase in female labour participation. The graphs show that while the male employment rate has declined somewhat, the female employment rate has almost doubled in the US, and more than doubled in Canada. Because official OECD data on gendered working age populations do not go very back, male and female working age populations are each assumed to be half of the total working age population:
To see the effect of this increase in the employment rate alone on the availability of work, we can consider a second hypothetical situation. This time, we allow the working age population to change, but hold the employment rate constant at initial levels:
This time, with the employment rate held steady, the work available for each worker is still more than in reality, but is much less than in the scenario where the working age population was held steady. In the US, if the employment rate had remained unchanged, the average worker in 2017 would have worked an average of 2033 hours per week, only slightly more than the 2003 hours in 1955. Meanwhile in Canada, despite large fluctuations, the average hours worked per worker would have generally risen over the entire 1961–2014 period. By 2014, the average worker would have worked 2301 hours, an increase from 2059 hours in 1961.
Comparing the two hypothetical scenarios, the increase in average work hours is much greater when the working population is held steady. This means that in both the US and Canada, the growth of the working age population has done much more to decrease work hours than the employment rate.
In fact, we can quantify this difference. In 2014, the actual average hours worked per worker in Canada was 1706. With a steady employment rate, each worker would have worked 34.9% more hours, but with a steady working age population, each worker would have worked 118.0% more hours. This means that between 1961 and 2014, the increase in the working age population erased 3.4 times more average work hours than the increase in the employment rate. In the US, the difference is even greater; the increase in the working age population erased 7.6 times more work hours than the increase in the employment rate (an increase of 106.5% compared to 14.0%), between 1955 and 2014.
The data reveal a common pattern for both US and Canada: a steadily-growing labour demand, accompanied by a labour supply that has been growing even faster. In both countries, the growth of the labour supply is primarily due to the growing population, but also, though less significantly, due to the rising employment rate.
As in North America, average work hours have also generally been declining in the UK, France, and Sweden:
In all three European countries, the average hours worked per worker declined the most during the 1970’s, as they did in Canada and the US. In the UK, they declined significantly from at least 1970 (the first year in the UK’s OECD data), until 1981. This slide was then followed by a brief rebound in the UK that lasted until 1988, but the slide then resumed until 2010, after which it rebounded slightly again before flattening. Meanwhile, in France, the decline was almost continuous from 1950 until 2002, after which average work hours have stayed relatively constant. Finally, in Sweden, the largest decline in work hours occurred from 1950 until 1981, though this was followed by a long but weak rebound that, with the exception of a sizeable drop from 1997 to 2003, continued until 2010, when it began to flatten until 2017.
In contrast to North America, however, where total work hours have increased steadily, total work hours in the three European economies have seen periodic declines, with the most severe declines occurring in France. Here are the total hours worked for the UK, France, and Sweden:
In all three European countries, there were decreases in the total amount of work performed, starting from about the mid-60’s until around the early 80’s. The drop in France was particularly drastic, sliding from 41.1 trillion work hours in 1966 to 34.4 trillion work hours in 1985, a decrease of 16.2%. The UK also suffered a comparable drop of 14.3% from 1973 to 1983. Sweden’s drop in labour demand was the mildest but nevertheless significant, losing 7.9% of its total work hours from 1963 to 1978.
These declines were then reversed in all three countries, until 1990, when total work hours saw a sudden decrease. Although recovery began within a few years, it would be another 14 to 15 years before the 1990 peak would be reached again. Total work hours have been increasing fairly steadily in the last decade or so, with the 2007 recession causing only a slight dip (though the effect lasted longer in France).
Out of the three countries, France appears to have had the weakest growth in labour demand overall. While total hours worked in both the UK and Sweden have now well surpassed what they were in 1970, in France they surpassed 1970 levels only in 2014, and have yet to even surpass the peak in 1965.
As total work hours increased hesitantly in Europe, the number of total employees have nevertheless risen more steadily:
With the exception of a few dips, the curve for total employees in all three countries have grown fairly steadily. The steady increase in France’s total employees is particularly striking, in contrast to its V-shaped graph of total hours worked.
Notably, as total work hours dropped sharply in the 70’s in all three countries, the total number of employees rose. This explains the sharp drop in average hours worked during the 70’s. Dips and periods of stagnation in total employees can also be seen during the 80’s and 90’s, but these generally match the drops in total work hours in the same periods, thus having little effect on average hours worked.
The differences between trends in total hours worked and total number of employees are reflected in their correlation coefficients, which are lower than those for Canada and the US: 0.888 in the UK, 0.805 in Sweden, and just -0.073 in France. The almost negligible correlation in France is a stark illustration of how employment has been woefully inadequate there.
We can now examine the contributions of the working age population and the employment rate to the number of total employees. First, the working age population:
In all three countries, the working age population has been growing fairly smoothly, though there are several periods of slower growth. Again, the steady growth of the working age population contrasts with the much more start-and-stop pattern of increase in total work hours.
The increase in working age population relative to total work hours is significant enough that, if the working age population were held steady, the average hours worked per worker would be higher than their 1970 levels in both the UK and Sweden, and would be close to matching it in France.
Overall, however, the average hours worked in this hypothetical scenario would still have dropped during the 70’s, and would only have recovered to pre-1970 levels or just slightly exceeded them. This contrasts with the steady ascent of the same hypothetical scenario in both the US and Canada.
Next, we look at the employment rate:
Unlike the employment rates in the US and Canada, which have increased steadily over the last decades, the employment rates in the three European countries have been more of a roller coaster. There were sharp declines in all three countries during the 80’s and 90’s: in the early 80’s and 90’s in the UK, in the mid-80’s in France, and in the early 90’s in Sweden. The drop in Sweden was particularly steep, sliding from a peak of 81.5% in 1990 to 70.2% in 1994, erasing almost three decades worth of gain in just four years.
As in Canada and the US, the increase in employment rates in the UK, France, and Sweden have largely been driven by the rise in female labour participation. Like in North America, the male employment rates have decreased in Europe, and even more markedly so, especially in France. The female employment rates, have more than compensated for this drop, with the exception of France; however, these increases have not been as sharp as in North America.
As a result of the smaller increase in the employment rate, the effect of holding the employment rate steady on average hours worked per worker would not be nearly as significant for the European countries as it would be for North America:
In this hypothetical scenario, the average hours worked would still show a general decrease in each country over the entire time period. In particular, the drop in France would have been even more severe, since the employment rate there is now even lower than it was in 1955.
It appears that the UK, France, and Sweden show historical labour patterns that are distinct from the North American pattern. In contrast to the steady growth of labour demand seen in the US and Canada, labour demand in Europe has been much weaker, with declines that only ended in the 80’s and 90’s, after which it has shown stronger growth. The labour supply, however, has been climbing much more steadily, though not as steadily as in North America. As in North America, it is the working age population that has contributed significantly more to the increase in labour supply than the increase in the employment rate.
The three European countries are, however, not as similar to each other in their historical labour patterns as the US and Canada are to one another. Among the three European countries, France has suffered from especially stagnant labour demand. The UK and Sweden have fared better, with Sweden alone showing a general increase in average hours worked since the early 80's.
On the surface, Japan’s steady decline in average hours worked per worker appears typical of the countries we have already seen in North America and Europe. However, as we will see, the factors that has produced this decline in Japan is quite unique.
As in most of the other countries, the decline in average hours worked has been relentless. However, the timing in Japan has been different:
Japan saw a sharp decline in the first half of the 70’s, but this was followed by a period of almost 15 years when average hours worked almost held steady. Then, deviating from the West, Japan experienced another sharp decline starting in the late 80's. The rebound after the Global Recession has been weak.
A look at the total number of hours worked reveals the reason for this general decline:
With the exception of a decline in the early 70’s, total work hours in Japan grew steadily until around 1990. Since then, it has been shrinking almost as fast as it had been growing before. Following the global recession in 2007, there has been a slow but steady rebound beginning in 2009.
The fast decline in average work hours that began in 1990 is all the more remarkable considering the fact that the total number of employees was not only not growing around that period, but was actually in slow decline:
The total number of employees had been increasing steadily from 1970 until about 1992, when it began to level off quite suddenly, and continues to plateau through 2013. As in France, the total number of employees shows quite a different overall trend from the total of hours worked, with a correlation of just 0.108. This decorrelation reflects the failure of the labour demand to match labour supply.
Much of this decline in the total number of employees can be explained by the well-known decline in Japan’s working age population, which began to shrink after reaching a peak in 1995:
Out of the six countries considered here, Japan is the only country with a working age population that is shrinking over a prolonged period of time. This shrinkage also appears to be accelerating.
However, this decline in the working age population has yet to be great enough to cancel out its earlier growth. As a result, if the working age population were held steady at its 1970 level, average hours worked would still be greater than actual current levels:
In this hypothetical scenario, the average worker in 2013 would have worked 1914 hours, or 180 hours more than the actual figure of 1734 hours.
Interestingly, while the working population after 1990 has been falling precipitously, Japan’s employment rate during the same period has been rising:
After having stagnated since 1970, the employment rate took off around 1988. It then leveled off until 2002, but then has been rising ever since, with only a brief interruption during the global recession of 2007.
Japan’s male and female employment rates show a distinctly different pattern from the five Western countries. Specifically, instead of falling, the male employment rate has risen slightly over the past decades. The female employment rate, however, has risen somewhat more. This means that in Japan, too, increasing female labour participation has been an important factor in the rising employment rate there:
If Japan’s employment rate had held steady at its 1970 level, the average hours worked per worker in 2013 would be higher than it is currently, at 1951 hours. This is slightly higher than the average hours worked would have been if the working age population had held steady, which means that in Japan, the employment rate has contributed slightly more to the decline in average hours worked than the working age population. Out of the six countries examined here, Japan is the only country where this has happened.
In Japan, then, work scarcity has been created by both a drop-off in labour demand and a steady growth of the labour supply. Before 1988, the labour supply grew primarily because of population growth. Then, after a brief period of growth in both the population and the employment rate, the labour supply stopped growing for more than 20 years between 1992 and 2012. This, however, was still not enough to offset the steep drop in labour demand, resulting in the drastic decline in average work hours during that period. Since then, the working age population has been in free fall, but the labour supply has nevertheless grown because of an increasing employment rate.
We have seen the employment data for six developed countries in three different regions of the world. The data appear to support two general observations:
- Labour demand can behave independently from labour supply.
While labour supply and labour demand are sometimes correlated, they can also sometimes be severely mismatched. This has been especially true in France and Japan.
In Canada and the US, both labour demand and labour supply have been growing, though labour supply has grown faster. The labour demand in Britain and Sweden has grown more and absorbed more of the growing labour supply. In France, however, this has not been the case. Notably, the 30-year period from 1955 to 1985 saw a steep drop in labour demand, but the labour supply rose throughout. In Japan, too, the labour supply grew just fast enough to meet the labour demand from 1975 to 1990, but from about 1990 to 2010 the labour demand fell steeply even as the labour supply held steady.
2. The labour supply has grown because of both a growing population and a rising employment rate (largely driven by the increasing proportion of women in the workforce), but population growth has generally been a more important factor.
In five of the six countries studied here (Japan being the exception), the growing working age population has contributed more to the growth of the labour supply than the employment rate. Whereas the working age population has grown fairly steadily (Japan again being the exception), the employment rate has not risen as reliably. The employment rate appears to be much more susceptible to recessions and other temporary shocks. Sweden’s employment rate, for example, dropped more than 10% in the four years between 1990 and 1994, and had still yet to return to its 1990 levels by 2013. France’s employment rate has not increased at all over the period studied, being even lower in 2013 than it was in 1955.
Japan is the exception to this pattern. Japan’s working age population began to decline after reaching a peak in 1995, and this decline has been accelerating. But during the same period, the employment rate has been increasing. This has made the employment rate in Japan a slightly more important contributor to increasing work scarcity than the working age population, between 1970 and 2013.
The increase in employment rate over the past decades has largely been driven by the rising participation of women in the work force. Among the six countries studied, all except Japan have experienced a declining male employment rate; in all five Western countries except France, this decline has been more than compensated by a rising female employment rate. In Japan, the male and female employment rates have risen together, but the female employment rate has still increased more overall.
It is no surprise that labour is a fairly inelastic commodity — that is, labour supply and labour demand do not always match each other. A labour shortage may lead to wage increases and an uptick in the birth rate, but newborns take more than 15 years (usually significantly longer) to mature into productive workers. A faster response is to increase immigration, though this can sometimes be a politically unpopular solution.
The problem of labour oversupply has even fewer solutions, since governments cannot easily get rid of surplus workers. In the past, labour surpluses have eventually disappeared, but only through population decline brought on by war, disease, and a general decline in popular well-being (Turchin documents this thoroughly in his book Secular Cycles). Better alternatives are clearly needed. But governments has no legitimate way of reducing the working age population or the employment rate.
The common government response is to try to stimulate the economy and increase labour demand. That, however, is easier said than done. Labour demand is tied to economic growth, which in turn is tied to factors such as population growth, technological development, availability of natural resources, and international trade relations, none of which can be entirely controlled with policy alone.
Rather than trying to manipulate labour supply or demand, it may be easier to decouple income from labour altogether. Indeed, this is already being done, through employment insurance, tax credits, and other government benefits. Recent calls for Universal Basic Income also follow a similar line of thinking.
As population growth slows across the developed world, it is only a matter of time before the current labour oversupply becomes a labour shortage. In the meantime, however, our ability to navigate chronic labour oversupply, and the social stability it creates, will be a major test for the developed world in the coming years.
Screenshot of a scene from the drama Nigeru wa Haji da ga Yaku ni Tatsu (逃げるは恥だが役に立つ) taken from Twitter account とむ (@rom_0129).