In the previous blog I asked why real wages stopped growing in the 1970s. A host of explanations has been discussed by economists and political commentators (although they tend to focus on the related issues of income and income inequality). David Leonhard, for example, listed 14 possible causes for the income slump during the last decade. A similar list was earlier discussed by Timothy Noah in a series of articles on why income inequality has been growing since the 1970s.
As I said in the last blog, the usual approach is to consider each possible cause in isolation. But that is not how we can weigh their relative importance. Furthermore, societies are complex dynamical systems, in which different causes are interrelated through a web of nonlinear feedbacks. Translated into plain language, this means that everything can potentially affect everything else, and this should be taken into account. Finally, to understand the dynamics of wages and incomes we need to go beyond economics.
To answer the question of why real wages abruptly stopped growing in the 1970s we need a unifying theoretical framework, which can be confronted with data. In short, we need Cliodynamics. Here’s my stab at disentangling the thicket of economic forces that may help explain the dynamics of real wages. (I will discuss the more complex question of household incomes in a later blog.)
The obvious starting point for thinking about wages is to consider how much total income is produced by a society; then divide it by the number of people. A common measure is gross domestic product (GDP) per capita. This quantity is often referred to as ‘per capita income,’ but such language is misleading because the income of a typical person (or a household) can diverge pretty far from GDP per capita.
Still, the real GDP per capita today is 5.6 times higher than in 1928 (for reasons explained later I begin my data series in 1928). Surely such a huge increase in the gross national income should have an effect on how much individual workers earn.
And it does. But it is only part of the story. In 1978 the per capita GDP was 3.2 times higher than 50 years before, but the compensation of production workers increased 4-fold over the same period (again, both GDP and wages are expressed in inflation-adjusted dollars). Even wages of unskilled workers grew faster than GDP per capita (3.5 times between 1927 and 1978).
After 1978 GDP per capita continued to grow at a good clip (in 2012 it was 70 percent higher than in 1978), but the production workers wages stagnated and the wages of unskilled workers actually declined, by 10 percent. As I said, trends in real wages can diverge substantially from trends in the per capita GDP.
What forces cause such deviations? We can divide them into two broad groups: market (economic) forces and non-market factors. The second group includes influences of social norms and institutions, as well as the balance of political (or even coercive) power that various players can wield. I will deal with non-market factors in the next blog.
The first group reflects the operation of market forces. Human labor can be thought of as a commodity, whose cost (i.e., real wages) is affected by the interplay between demand and supply. If demand for labor exceeds its supply, real wages should go up. Conversely, when there is an oversupply of labor, real wages should decline.
Can we put some numbers to this quantity? To estimate how demand changed over the years I take the GDP for a particular year and divide it by current labor productivity. The Bureau of Labor Statistics has measured the average output per hour per worker since 1947, and less precise estimates are available for previous periods. Full time employment (40 hours per week) translates roughly into 2000 hours per year, so hourly productivity multiplied by 2000 gives us productivity per year. Dividing GDP by this number we obtain an estimate of how many workers are required to produce the goods and services in that year.
For labor supply I use the labor force data from the Bureau of Labor Statistics. These numbers include both employed and unemployed, looking for work. (I am skipping a lot of technical details here; in a few days I will post a manuscript that will provide such details.)
When we put the estimated trends in labor demand and supply on the same graph, here’s what we see:
At first the demand curve grows faster than the supply curve, so during this period market forces favored wage growth. During the late 1960s, however, the supply curve accelerated and after 1970 outpaced the growth of demand. One reason for this acceleration was the reversal of immigration policy in 1965 that facilitated the arrival of workers from overseas. By the early 2000s, one in six American workers was foreign-born. However, the initial rise during the late sixties and the seventies was mostly due to the second factor, internal demographic growth. The generation that reached marriageable age during the Great Depression and World War II had fewer babies than the post-war generation (the parents of baby-boomers). When baby boomers started entering the job market in massive numbers after 1965, they quickly drove up the supply curve above the demand.
The gap between the two curves was also affected by developments on the supply side. 1975 was the last year when the US enjoyed a trade surplus. Since the 1970s the trade deficit has been ballooning, exceeding 5 percent of the GDP in early 2000s. Trade deficit subtracts from the GDP; having more imports than exports means that demand for labor is satisfied by foreign workers.
One startling development is that around 2000 the demand curve stopped growing altogether. This remarkable occurrence was due to a combination of sluggish economic growth together with rapid gains in labor productivity, which put a lid on the number of workers needed to satisfy the demand for labor. Notice that the plateau occurred before the Great Recession, and it provides one possible explanation for the ‘jobless recoveries’ following the last two recessions.
Overall, the trends in demand and supply curves appear to yield interesting insights into the forces shaping the dynamics of American real wages. However, before we can quantitatively estimate the relative importance of the effect of the demand/supply ratio on wages, we need to quantify the dynamics of extra-economic factors (in the next installment of this series).
A final note: the demand/supply approach is a very parsimonious way of summarizing various market forces that affect the trends in real wages. Thus, immigration (both legal and illegal) enter on the supply side – by increasing the supply of labor they should depress its price. Foreign trade enters on the demand side. A trade surplus increases demand for labor and should have a positive effect on real wages. Trade deficits do the opposite. We can even assess the relative effects of immigration and trade, even if crudely, by expressing trade deficit in worker units (using average annual productivity of US workers).
To Be Continued
Robotization will drastically reduce the need or desire to hire humans, as explored in the story of Metropolis. Theorists of many kinds ignore the implications of drastic overpopulation in the face of technology’s increasing ability to make humans irrelevant. Few will see any increase in real wages – ever.
In fact, many have commented on the effects of robotization, or automation more generally. It has already affected large swaths of manufacturing. So blue-collar workers have already been exposed to its effects. Next are white-collar workers. Here, take a look at this article by Randall Collins:
politconcept.sfedu.ru/2010.1/05.pdf
Every one seems to want to stimulate the economy. I think the economy is too stimulated already. This country is enormously productive, maybe 20 times or more so than 200 years ago. There is no reason why we can not go to a 6 hour day or a four day week. That much work would provide us with all the necessities and all the comforts we need, and most of the luxuries. I suspect it would even provide us with gambling casinos, mansions, monuments, and even million dollar bonuses to parasites. If some of the last did have to be cut back on, I would not be distressed.
A mandatory 6 hour day or 4 day week with time and a half for overtime in corporations would solve the unemployment problem almost over night. With the extra time people could enjoy themselves. There is no divine law of the Universe that says we must work exactly 8 hours, not a second more or less, each day. Indeed it is said that Kellogg cut the work week to 30 hours without loss of productivity [Schwartz]. The important matter is for everyone making money, no one destitute. A fringe benefit would be a drastic reduction in unemployment benefits that plague the government.
If there was not enough production that way to make gambling casinos, no problem. People could gamble by card games in their own homes if they wished. Of course it would be necessary to HAVE their own homes for that. Foreclosure could be prevented by the federal or state government paying the full amount immediately (thus bank liquidity solved) provided the owner agreed to pay back by an agreement to pay 20% or so, or whatever was enough, additional income tax, which would include principle amortization, 6% or so interest, and cost of life insurance to cover the remaining principle. That would eliminate most foreclosures. It would also eliminate the federal loss currently incurred arising from multi billion gifts to our nobles in the financial sector channeled through subsidies, bailouts, low interest rates or etc, with citizens having a whole life time to pay back if necessary.
Sincerely, Charles Weber
Schwartz M 2009 Quick study: The future of work. Reader’s Digest (June) p 120-123.
A long time ago John Maynard Keynes predicted that by now we would be having short working weeks, just as you describe. This would result from increasing productivity. The productivity growth fulfilled the promise, but it did not result in shorter working hours and more leisure. Well, there is more leisure for the unemployed, if leisure is the correct term.
Some years ago France shortened the work week to 35 hours, and I remember how they were criticized in the press at the time. Do they still have a 35 h week? It did not seem to hurt their economy in a noticeable way.
Thanks for the article – I’ll be sure to check it out.
Dear Peter Turchin,
You are welcome.
Perhaps you will find some health articles even more useful for someone you like or love as listed in http://charles_W.tripod.com/index.html . They would be very useful for our society as well if instituted because heart disease, rheumatoid arthritis, gout, and high blood pressure (potassium), and herniated discs, hemorrhoids, and aneurisms (copper) cause enormous losses.
Sincerely, Charles Weber
I think the increase in the “supply” had more to do with the entry of women into the labor force on a massive scale, rather than the easing of immigration restrictions. Between 1948 and 2011 the number of women in the labor force increased by about 315%, while the number of men by about 90%. For between 1975 and 2011 the corresponding figures are about 92% and 46%, respectively.
I don’t know off the top of my head how many foreign born workers there were in 1948 or 1975. Today there are about 24.4 million. Assuming for the sake of argument that there were 0 foreign born workers in 1975, that 24.4 million should be compared to the increase of 26 million in “all men” and the increase of 35 million in “all women” (those numbers of course include the foreign born, of which about 60% are men)
More generally the effects of immigration on wages is a bit more complicated because foreign born workers are not perfect substitutes for native workers. In fact, depending on industry, level of education and other characteristics they may even be complements. This would imply that the entry of new immigrants into the labor market would increase the wages of native workers and lower the wages of… previous immigrants. But it is a bit more complicated than that. The economic impact of immigration though probably has more to do with reshuffling wage levels across industries/skill classes rather than affecting the overall average (or median) wage level of native born workers.
I have not gotten to the remaining parts of this blog post so it’s possible I am addressing a side issue. The standard economic explanations for the phenomenon you’re describing are “skill-biased technological change”, “competition from abroad” and “decline in union/worker power” with most economists, at least up until early 2000’s putting the most emphasis on the first one, with the other two being seen as relatively minor (there are some dissenters though)
(great blog btw)
And now I see that this is directly addressed in the last post of the serious. My apologies.
The best work on the effect of immigration on wages is by George Borjas, in my opinion. There are several unexpected results. It’s worth a separate blog to explore these issues, however.
I have an idea for illegal aliens. Give every alien a hundred or two dollars for every year they have been here and send them home. It would be the best deal we ever made. It would cost that much or more to apprehend them otherwise and they would not arrive home destitute. You may not like them here, but most of them are not criminals and do not deserve vicious treatment. That would only cost taxpayers 30 or 60 dollars each ( the income tax their employers would have had to pay otherwise) for each year. Sounds like a good deal to me. On top of that you greedy Americans would have received cheap labor from them, a profit probably well over their mustering out pay. Then a constitutional amendment making babies acquire the citizenship of their parents put in place.
If I remember correctly Borjas’ key contribution back in the day (1980’s) was accounting for how workers allocate themselves in response to wage differentials in the estimates of immigrants’ impact on wages (actually David Card might have been the first one to do it) and for distinguishing between the skills of “old” and “new” migrants. But that was some time ago and his work on the topic suffers from some severe drawbacks.
First, it’s almost all (except the most recent stuff) partial equilibrium – it doesn’t take into account feedback effects (for example increased demand for goods and services by migrants, which would affect labor demand as well). Another aspect of this is that most of it doesn’t take into account the increases in capital stock which results from immigration (basic economic theory tells us that in the long run the capital-(effective)labor ratio is constant and this accords with observed data). In some of the more recent work Borjas does acknowledge that the long run impact on natives’ wages is about zero.
More importantly, in most of it, the assumption that natives and immigrants are perfect substitutes is imposed – this pretty much forces the estimated impact of immigrants on natives’ wages to be negative (and that only in the short run). Once this assumption is dropped the impact differs across education cohorts and for many workers, the effect is actually positive. There’s still some debate as to the exact size but at least in my opinion Borjas is on the loosing side here (at least in the sense that whatever the sign of the impact, the magnitude is minuscule).
Some papers on this are:
Ottaviano and Peri: http://www.nber.org/papers/w14188
Borjas’ (and Grogger and Hanson) response to, I think an earlier version of O&P, http://isites.harvard.edu/fs/docs/icb.topic803549.files/Week%207-October%2020/borjas_imperfect.pdf
This I think is the one which accounts for Borjas’ response: http://onlinelibrary.wiley.com/doi/10.1111/j.1542-4774.2011.01052.x/abstract
Here is David Card, who’s findings are pretty much inline with O&P
http://www.nber.org/papers/w14683
Here is another Peri paper on immigration increasing productivity and capital intensity
http://www.nber.org/papers/w15507
Here is Brad DeLong’s comment on the debate
http://delong.typepad.com/sdj/2007/06/ottoviano-peri-.html
(Apologies in advance if that’s throwing too much stuff out there at once)
Radek, no problem about the links – in fact, thanks for them (although the spam filter software was in doubt, and I had to tell it it’s OK).
On the substance, Borjas is quite aware of all the pitfalls in the analysis. He explicitly focuses on the short term. The problem with the long term is that nobody knows how to analyze or model the contribution of immigrants to the growth of GDP and eventually labor demand. I’ve been asking around with my economist friends, and they tell me that the processes underlying economic growth are extremely contentious. In short, they don’t know. That’s why in my model I don’t attempt to model GDP growth, I simply take the actual trajectory. There are reasons to believe that immigrants, especially illegal immigrants, may not contribute much to the GDP growth. Their consumption rate is pretty low, and they export a large part of their wages to their home countries.
Also, the ‘short term’ may be 5-10 years. That’s a lot of time.
Borjas and coworkers, judging from their papers, are quite aware about ‘perfect’ versus ‘imperfect’ substitutes, so this is, I feel, a red herring.
I just finished fitting a regression model to the data which takes into account autocorrelated errors, so I now can assign standard errors to the estimates associated with various factors. Both labor demand/supply ratio and the proxy for cultural factors are highly statistically significant.
Your supply-demand analysis appears to be tautological. Productivity is output per hour worked, or GDP / (W*Hw) where W is total number employed and Hw is average hours worked per employee. You multiply this value by a nominal 2000 hrs worked/yr, divide it into GDP and call this demand:
demand = GDP/2000*Prod = GDP/[2000*GDP / (Hw*W)] = (Hw/2000) * W
Now W is total employed which is labor force (LW) mutliplied by 1-U where U is the unemployment rate. This gives demand as:
demand = (Hw/2000) LF (1-U) = R LF (1-U)
here R is the ratio of the actual hours worked to the nominal 2000 hours. You defined supply as LF. Let’s detrend the data by dividing by LF. Then supply becomes a constant equal to one and demand is given by
demand = R(1-U)
Thus demand relative to supply would should a secular trend given by R with oscillations about the trend that vary with the business cycle. If we use a moving average to smooth out the cycles you can then approximate U as a constant equal to the average unemployment rate.Thus your graph is an expression of R. Before the passage of labor legislation in the 1930’s they was no national 40-hour work week. Working class people typically worked 50 hour weeks,with no overtime premium. Even salaried workers had longer work weeks (in those days the New York Stock Exchange was open on Saturdays). With the introduction of the 40 hour work week, R would fall, making demand look weaker.
Today a full time salaried worker officially works 40 hours a week no matter how much actual time is spent at the office. Employers can work their salaried workers 60 hours week and get 60 hours of output, but this only registers as 40 hours resulting in an apparent decrease in worker demand by a third This is an artifact of the way these things are calculated.
And so you see demand fall off a cliff after 2000.
As for the shift you see in the late 1970’s part of the answer is what William Greider called “The October Revolution in Federal Reserve policy” in 1979. I talk about this in this article:
http://www.safehaven.com/article/11230/stock-cycles
The other issue is not just the number of workers, but the quality. I hire people all time as the co-owner of a ~ 5 mil annual sales business. Obviously, I want the bast caliber employee I can find, and am competing against other employers to get them. When supply/demand was balanced, a median job likely attracted a median quality worker, and crap got crap. Now, the least desirable job in the can land a ~35th percentile worker in education/attitude or whatever else the employer values, and and the median employer can likely get a 60-70th percentile worker. Under this system, “enough” is never enough. If there are only enough jobs for 20% of the workforce, great, open the flood-gates. Means all my employees can be 80th+ percentile. May cause societal problems, but in the short term more is always better, even if I have no real use for the surplus.
Sorry for the typos. Thought I’d be able to edit before it posted. 🙂
So Peter Turchin, are you going to reply to Mike Alexander’s comment ? Especially the first 2/3 of it ? Your measure of “labour demand” is the {ratio of actual average hours worked per worker-year to 2000 hours} multiplied by {the employed share of people seeking work}. It’s a very strange measure.