Robot Workers Stoke Human Fears; how we can still win
by John MacBeath Watkins
Finally, a headline that made me feel like I was in the 21st century appeared on the print edition of Tuesday's New York Times:
Rise of Robot Workforce Stokes Human Fears
The e-version of the Times, not restricted by the character count, had a longer headline that did a better job of explaining the fears of human workers as artificial intelligence makes machines more capable, but the headline in the print edition is what a time traveler from 1950 might have expected to see above the fold on the Dec. 16, 2014 New York Times.
In some ways, the future has been late to arrive, but as Louis Althusser noted, "l'avenir dure longtemps" (usually translated as "the future lasts forever," the name of his memoir.) There is plenty of time for the future to happen, and it will arrive in unexpected ways.
Let me first tell you my solution to the automation problem, then I will explain what it means.
We have been managing our economy for a "natural" rate of unemployment and an inflation rate of 2%, with inflation defined by the core consumer price index number. We have no real empirical knowledge of what the natural rate of unemployment is, nor have we any reason to assume 2% is the proper level of CPI growth. We should be managing the economy by some objective metric.
Even the core CPI is subject to external shocks. The inflation metric we should be using is wage inflation, and we should be managing it to match productivity growth. In fact, since we have not done so for a long time, we should be managing for wage inflation higher than that until it catches up to the pre-1980 trend. This would ensure wages don't, in the long run, exceed the structural capacity of the economy, and that they don't trail so far behind it.
One of the things that makes capitalism different from previous systems of market economies is that in most of them, the workers owned their tools, and land and labor were the keys to making a living (which is why if you wanted to be wealthy, you needed to get land or a mine.) The total amount of wealth was presumed to be finite. Capitalism is a system where the tools usually don't belong to the artisan, they belong to those who accumulate and invest capital. And investing capital in technology is often a way to increase wealth, meaning that capitalism isn't a zero-sum game.
The transition from a traditional market economy to a capitalist one is not great for those who used to be the skilled artisans, which is why the invention of mechanized weaving sparked the Luddite uprisings of the early 19th century. The Luddites recognized that while weaving made cloth cheaper, and enabled most people to buy more clothing and make it a smaller part of their budget, it also meant that the weavers, once high-skilled and well-paid artisans, were surplus to requirements.
Eventually, we found things for people to do. The key was to have enough economic activity that the increased productivity does not permanently unemploy workers whose careers have been disrupted by more productive technology.
Economists used to think that there was an inverse tradeoff between inflation and unemployment, a relationship called the Phillips Curve. This was displaced in the late 1960s and 1970s by a new concept, the natural rate of unemployment, championed by Milton Friedman and Edmund Phelps.
Friedman and Phelps argued that for there to be a permanent increase in employment, something would have to change in the real economy. Essentially, he argued, the Phillips Curve relied on an illusion, and when inflation expectations for wages and prices caught up with reality, this would leave unemployment unchanged.
The problem with managing the economy for the natural rate of unemployment is, there is no established way to know what the natural rate of unemployment is. Friedman and Phelps, it would appear, moved us from managing the economy based on an illusion to managing it based on a guess.
How's that working out for us?
http://en.wikipedia.org/wiki/Labor_force
United States Labor Force Participation Rate by gender 1948-2011. Men are represented in light blue, women in pink, and the total in black.
Labor force participation peaked in about 1998, which is about the last time we had much in the way of wage inflation, but really, labor force participation has been stagnant or declining since about 1990. That's because wages, like any price, are a signal, in this case, a signal to come one out and get a job. Take a look at the comparison between productivity growth and average real wages:
http://en.wikipedia.org/wiki/Real_wage
1990 is about the time real wages for the average worker fell below the 1970 wage level, and it's been there since. Never the less, female participation in the labor force has increased, and women's wages, while they have not caught up to men's, have at least been increasing.
Here is another set of data that most people have not incorporated into their analysis:
- See more at: http://marginalrevolution.com/marginalrevolution/2011/06/the-great-male-stagnation.html#sthash.5cU3F6M0.dpuf
http://marginalrevolution.com/marginalrevolution/2011/06/the-great-male-stagnation.html
While working men's wages fell a bit from 1980 to 2012, men found it harder to get a job. As a result, the decline for income among all men has been much worse than the situation among working men.
http://www.motherjones.com/kevin-drum/2012/07/working-class-mens-wages-have-plummeted-over-past-40-years
Much of the decline in male incomes has been at the median and lower end of the distribution of education:
Data: Hamilton Project
http://www.washingtonpost.com/blogs/wonkblog/wp/2012/07/31/wages-arent-stagnating-theyre-plummeting/
Dyland Matthews notes about the above chart (and I recommend following the link and reading his full essay)
Now, there's a good news/bad news situation here. Men's labor force participation has been falling since the 1950s, even though through the 1950s and 1960s their wages were increasing rapidly, so what's causing their decline in labor force participation isn't the integration of women into the work force. The bad news is that their wage gains started falling about the time women's labor force participation increased.
I'm not convinced that this is a case of post hoc, ergo propter hoc. The thing is, at about the same time, we had a major recession, then started managing the economy for the unknowable natural rate of unemployment. Median male income started declining about the time productivity increases and wage increases became de-linked. That's about the time we adopted two of Milton Friedman's big ideas, managing companies for shareholder value and managing the economy for the "natural" rate of unemployment. I strongly suspect that these two ideas played a strong role in removing the link between productivity increases and wage increases.
Since incomes for men have fallen most for the least educated, the indication would be that the kinds of work available do not play to male strengths, such as upper-body strength and a willingness to take jobs with lots of heavy lifting, risk, and inclement weather, which would indicate one possible reason the male workforce participation rate has been falling since about 1950.
Teachers used to say, "you want to dig ditches for a living?" but now, to do that you have to be a heavy equipment operator, and a lot fewer of those are needed than ditch digging humans for the same size ditch. The increase in productivity is in an industry were the demand for ditches is not particularly price elastic.
While many people might say, "that's a pretty good price on shirts, I think I'll get two," few people look at a contractor's rates and say to themselves, "well, I don't really need a new ditch, but at these prices..."
More and more of those risky, physical jobs in the open air have been mechanized. Now, artificial intelligence offers the opportunity to replace humans at inside work with no heavy lifting, which will affect both genders.
This is also an opportunity for those who own capital to take more of the gains from productivity. The question is, why should they? Those weavers who became Luddites were employable, and had they had access to new skills and a hot labor market, they might not have minded so much losing their work as artisans. While the early part of the industrial revolution created great misery and inequality as the capitalists took most of the money, eventually, we figured out how to redistribute the wealth and start the great era of the middle class.
We need to give people access to new skills. Right now, higher education and the student loan program are a mess. How we expect to get a skilled labor force with any spending power out of that is one of the great mysteries of our time. The acquisition of new skills is one of the great levelers for a society, and we've made it tremendously expensive and difficult.
We need to manage for a higher level of employment. The fact that so much of our productivity gains have gone to the top instead of the working class indicates that we are managing for such low levels of employment that there is no pressure for higher real wages. Employers have been able to manage for stagnant real wages across the board, and plummeting wages for the class of people -- white males -- who had been paid best.
Managing for a slack labor market has certain advantages for the capital owning class, and which includes me to a small extent, but is concentrated at the top end of American incomes. Companies can spend their money on dividends and stock buybacks instead of worker's pay. The speculative natural rate of unemployment isn't particularly scientific, but it has provided a rationale for managing the economy for sufficient slack in labor demand to suppress the rise of real wages. The NRU has continued in use not because it is economically useful or provable, but because it is politically useful.
Ah, you say, but if we manage for wage inflation related to productivity growth, what about asset bubbles? I submit that those have as much to do with changes in banking as they do with monetary policy. The real estate bubble accompanied innovations in the banking industry that allowed the sale of securitized mortgages, making far more money available than could be invested by savings and loans making mortgages based on deposits. The great age of corporate raiders was also the great age of junk bonds. The tech bubble actually happened at the same time as some wage inflation.
Asset bubbles may well have more to do with the deregulation of the banking system and the rise of the shadow banking system than with monetary policy, and should be dealt with on that basis.
Finally, a headline that made me feel like I was in the 21st century appeared on the print edition of Tuesday's New York Times:
Rise of Robot Workforce Stokes Human Fears
The e-version of the Times, not restricted by the character count, had a longer headline that did a better job of explaining the fears of human workers as artificial intelligence makes machines more capable, but the headline in the print edition is what a time traveler from 1950 might have expected to see above the fold on the Dec. 16, 2014 New York Times.
In some ways, the future has been late to arrive, but as Louis Althusser noted, "l'avenir dure longtemps" (usually translated as "the future lasts forever," the name of his memoir.) There is plenty of time for the future to happen, and it will arrive in unexpected ways.
Let me first tell you my solution to the automation problem, then I will explain what it means.
We have been managing our economy for a "natural" rate of unemployment and an inflation rate of 2%, with inflation defined by the core consumer price index number. We have no real empirical knowledge of what the natural rate of unemployment is, nor have we any reason to assume 2% is the proper level of CPI growth. We should be managing the economy by some objective metric.
Even the core CPI is subject to external shocks. The inflation metric we should be using is wage inflation, and we should be managing it to match productivity growth. In fact, since we have not done so for a long time, we should be managing for wage inflation higher than that until it catches up to the pre-1980 trend. This would ensure wages don't, in the long run, exceed the structural capacity of the economy, and that they don't trail so far behind it.
One of the things that makes capitalism different from previous systems of market economies is that in most of them, the workers owned their tools, and land and labor were the keys to making a living (which is why if you wanted to be wealthy, you needed to get land or a mine.) The total amount of wealth was presumed to be finite. Capitalism is a system where the tools usually don't belong to the artisan, they belong to those who accumulate and invest capital. And investing capital in technology is often a way to increase wealth, meaning that capitalism isn't a zero-sum game.
The transition from a traditional market economy to a capitalist one is not great for those who used to be the skilled artisans, which is why the invention of mechanized weaving sparked the Luddite uprisings of the early 19th century. The Luddites recognized that while weaving made cloth cheaper, and enabled most people to buy more clothing and make it a smaller part of their budget, it also meant that the weavers, once high-skilled and well-paid artisans, were surplus to requirements.
Eventually, we found things for people to do. The key was to have enough economic activity that the increased productivity does not permanently unemploy workers whose careers have been disrupted by more productive technology.
Economists used to think that there was an inverse tradeoff between inflation and unemployment, a relationship called the Phillips Curve. This was displaced in the late 1960s and 1970s by a new concept, the natural rate of unemployment, championed by Milton Friedman and Edmund Phelps.
Friedman and Phelps argued that for there to be a permanent increase in employment, something would have to change in the real economy. Essentially, he argued, the Phillips Curve relied on an illusion, and when inflation expectations for wages and prices caught up with reality, this would leave unemployment unchanged.
The problem with managing the economy for the natural rate of unemployment is, there is no established way to know what the natural rate of unemployment is. Friedman and Phelps, it would appear, moved us from managing the economy based on an illusion to managing it based on a guess.
How's that working out for us?
United States Labor Force Participation Rate by gender 1948-2011. Men are represented in light blue, women in pink, and the total in black.
Labor force participation peaked in about 1998, which is about the last time we had much in the way of wage inflation, but really, labor force participation has been stagnant or declining since about 1990. That's because wages, like any price, are a signal, in this case, a signal to come one out and get a job. Take a look at the comparison between productivity growth and average real wages:
1990 is about the time real wages for the average worker fell below the 1970 wage level, and it's been there since. Never the less, female participation in the labor force has increased, and women's wages, while they have not caught up to men's, have at least been increasing.
Here is another set of data that most people have not incorporated into their analysis:
- See more at: http://marginalrevolution.com/marginalrevolution/2011/06/the-great-male-stagnation.html#sthash.5cU3F6M0.dpuf
While working men's wages fell a bit from 1980 to 2012, men found it harder to get a job. As a result, the decline for income among all men has been much worse than the situation among working men.
Much of the decline in male incomes has been at the median and lower end of the distribution of education:
Data: Hamilton Project
http://www.washingtonpost.com/blogs/wonkblog/wp/2012/07/31/wages-arent-stagnating-theyre-plummeting/
Dyland Matthews notes about the above chart (and I recommend following the link and reading his full essay)
High school dropouts' earnings have fallen 66 percent since 1969, and people with some college - the median level of education in the US - have seen earnings fall by a third.
Now, there's a good news/bad news situation here. Men's labor force participation has been falling since the 1950s, even though through the 1950s and 1960s their wages were increasing rapidly, so what's causing their decline in labor force participation isn't the integration of women into the work force. The bad news is that their wage gains started falling about the time women's labor force participation increased.
I'm not convinced that this is a case of post hoc, ergo propter hoc. The thing is, at about the same time, we had a major recession, then started managing the economy for the unknowable natural rate of unemployment. Median male income started declining about the time productivity increases and wage increases became de-linked. That's about the time we adopted two of Milton Friedman's big ideas, managing companies for shareholder value and managing the economy for the "natural" rate of unemployment. I strongly suspect that these two ideas played a strong role in removing the link between productivity increases and wage increases.
Since incomes for men have fallen most for the least educated, the indication would be that the kinds of work available do not play to male strengths, such as upper-body strength and a willingness to take jobs with lots of heavy lifting, risk, and inclement weather, which would indicate one possible reason the male workforce participation rate has been falling since about 1950.
Teachers used to say, "you want to dig ditches for a living?" but now, to do that you have to be a heavy equipment operator, and a lot fewer of those are needed than ditch digging humans for the same size ditch. The increase in productivity is in an industry were the demand for ditches is not particularly price elastic.
While many people might say, "that's a pretty good price on shirts, I think I'll get two," few people look at a contractor's rates and say to themselves, "well, I don't really need a new ditch, but at these prices..."
More and more of those risky, physical jobs in the open air have been mechanized. Now, artificial intelligence offers the opportunity to replace humans at inside work with no heavy lifting, which will affect both genders.
This is also an opportunity for those who own capital to take more of the gains from productivity. The question is, why should they? Those weavers who became Luddites were employable, and had they had access to new skills and a hot labor market, they might not have minded so much losing their work as artisans. While the early part of the industrial revolution created great misery and inequality as the capitalists took most of the money, eventually, we figured out how to redistribute the wealth and start the great era of the middle class.
We need to give people access to new skills. Right now, higher education and the student loan program are a mess. How we expect to get a skilled labor force with any spending power out of that is one of the great mysteries of our time. The acquisition of new skills is one of the great levelers for a society, and we've made it tremendously expensive and difficult.
We need to manage for a higher level of employment. The fact that so much of our productivity gains have gone to the top instead of the working class indicates that we are managing for such low levels of employment that there is no pressure for higher real wages. Employers have been able to manage for stagnant real wages across the board, and plummeting wages for the class of people -- white males -- who had been paid best.
Managing for a slack labor market has certain advantages for the capital owning class, and which includes me to a small extent, but is concentrated at the top end of American incomes. Companies can spend their money on dividends and stock buybacks instead of worker's pay. The speculative natural rate of unemployment isn't particularly scientific, but it has provided a rationale for managing the economy for sufficient slack in labor demand to suppress the rise of real wages. The NRU has continued in use not because it is economically useful or provable, but because it is politically useful.
Ah, you say, but if we manage for wage inflation related to productivity growth, what about asset bubbles? I submit that those have as much to do with changes in banking as they do with monetary policy. The real estate bubble accompanied innovations in the banking industry that allowed the sale of securitized mortgages, making far more money available than could be invested by savings and loans making mortgages based on deposits. The great age of corporate raiders was also the great age of junk bonds. The tech bubble actually happened at the same time as some wage inflation.
Asset bubbles may well have more to do with the deregulation of the banking system and the rise of the shadow banking system than with monetary policy, and should be dealt with on that basis.
Comments
Post a Comment