Saturday, July 1, 2017

Automation and jobs

I am often asked to opine about whether automation will destroy all the jobs. Yes, we talk about tractors, which brought farm employment from something like 70% of the country at the beginning of the 20th century to about 3% today. And cars, which put the horse drivers out of business. And trains, which put the canal boats out of business.

A more recent case occurred to me. This is what offices looked like in the 1950s and 1960s:

Typing Pool. Source: Getty Images

This is a "typing pool." There used to be basketball-court sized rooms that looked like this, all over the place, staffed almost exclusively by  women.

Then along came the copier -- many of these women are copying documents by typing them over again with a few sheets of carbon paper -- the fax machine, the word processor, the PC. And that's just typing. Accounting involved similar roomfuls of women with adding machines. Filing disappeared. Roomfuls of women used to operate telephone switchboards, now all automated.

This memory lives on in the architecture of universities. All the old buildings have empty hutches for secretaries.

If you are prognosticating in about 1970, and someone asks, "what will happen now that women want to join the workforce, but office automation is going to destroy all their jobs?" It would be a pretty gloomy forecast.

What actually happened: Female labor force increased from 20 million to 75 million. The female participation rate increased from below 35% to 60%. Women's wages relative to men rose -- they moved in to higher productivity activities than typing the same memo over a hundred times. Businesses expanded. And no, 55 million men are not out on the streets begging for spare change.


Civilian Labor Force Level: Women

Civilian Labor Force Participation Rate: Women

I'm simplifying of course. And surely some people with specific skills -- shorthand, typing without making mistakes, and so on -- who could not retrain didn't do as well as others. But the magnitude of the phenomenon is pretty impressive.

Update. So did women just take all the men's jobs? As MC points out, the male labor force participation rate did decline, from 87.5 to 70.0. That's a big, worrisome decline. But it's 15 percentage points, while the women's increase was 25 percentage points.

But even if women are moving in and men are moving out of employment, that does make the case that you don't just look at who has what jobs now threatened by automation! The typing pool got better jobs.

Please (please!) keep in mind the point here. No, this is not a post about all the ills of the labor market, and "middle class" America, and all the rest. Yes, there are plenty. The narrow point is, will automation mean that all the jobs vanish. In this case, even combined with a large expansion of the people wanting to work, it did not.



Also the male labor force expanded from 45 million to 82 million. So the idea that there is a fixed number of jobs and if women take them men lose them is not true.


48 comments:

  1. I was *literally* just in the workforce as the typing pool disappeared. I had just started at Northern Telecom (later: NORTEL Networks, now dead) as a college dropout "technical proofreader" for required regulatory process docs. I had to take existing test documentation, *literally* cut it up and paste in new directions, etc. Then take it to the typing pool for production.

    I'd been there 3 months when I stole a IBM correcting Selectric typewriter off the desk of a fired secretary since that was gonna be faster for me. Then six months later I got a Monolithic Memories CPM/86 machine with Wordstar and a line printer. Mirable dictu!

    None of the typing pool ladies got fired, but is gone 18 months after I started.

    Soooo, I think the economic expansion probably did absorb a lot of the people who suddenly had jobs (if not skills) that were obsolete.

    I will also note that I was around when we went from typed "foils" for "overhead projectors" to Harvard Graphics and HP line plotters. So our presentations went from half a dozen typed slides to dozens of graphics laden heavily tweaked efforts.

    So standards also changed as to how much work had to go into tasks.

    -XC

    PS - hey, kids, get off my lawn!

    ReplyDelete
  2. My father was lawyer from the late 1940s through the 1980s. He once told me that he kept 2 secretaries busy. By the time he retired, our firm maintained a ratio of about 2 lawyers per secretary. I am retired now, but my friends who are still working tell me that there are many fewer secretaries now, and that junior lawyers seldom use, or even have access to a secretary.

    ReplyDelete
  3. What also actually happened... wages at the bottom and in the middle stagnated, inequality, household debt and house prices soared.

    But let's just look at the selected facts that support OUR case. It's all good.

    ReplyDelete
    Replies
    1. Usually when I see a discussion on stagnating wages I have concerns about how the new inflation adjusted wages are calculated. The effects of technological deflation, which have provided almost everyone I know with unparalleled access to information and capital, are almost always undervalued when discussing wage stagnation.

      Delete
    2. Can we dispel this fallacy that "wages at the bottom and in the middle stagnated"? https://piie.com/blogs/realtime-economic-issues-watch/has-productivity-outstripped-wage-growth

      Delete
  4. And this is the fault of office automation and women entering the labor force? Including house prices soaring? Let's try to have some vague sense of cause and effect, and just a little bit of sticking to the point.

    ReplyDelete
  5. You note that 55mn men aren't out on the street begging for change. But the male labor force participation rate did decline by 20% over this period (!). https://fred.stlouisfed.org/series/LNS11300001

    This isn't my area (am I even looking at the right series?), but do you think there is a less optimistic story?

    ReplyDelete
  6. "And no, 55 million men are not out on the streets begging for spare change." /Please support and subscribe

    ReplyDelete
  7. It's good to look at both price and quantity here. From WW2 to the early 70s, male median income trended with productivity. From the early 70s onward, male median income has trended with inflation at $30-35K/year. Below this level, it becomes more profitable to search for government support than work or even illegal to work for less (minimum wage constraint starts to bind), so it makes sense that the labor market is increasingly quantity-clearing (less participation) than price-clearing (lower wages).

    It doesn't seem farfetched to me to expect a continuation of the long run trend towards more government support (food stamps, disability, student loans, Obamacare) resembling a universal basic income, along with lower labor force participation, as leisure is more attractive than the job market.

    AI probably won't cause us to lose work all at once--the various fiefdoms inside corporations, universities, and governments will resist disruption as long as they can. However, gradually but surely, the machine is coming for us all. What value the machine has for us will be determined only by how much of an allowance Big Brother gives us.

    ReplyDelete
  8. Each time automation increased productivity workers had to increase their skill level. With the coming wave of automation the skill level is going to have to rise substantially. My question is whether or not the workforce has the ability to go where it needs to go? Suppose everyone needs to become software engineers?

    ReplyDelete
    Replies
    1. Automation often allows work to be done by less skilled workers. A typist needs to be skilled enough to type an entire page without an error - otherwise they need to start all over again to produce a perfect copy. A word processor operator can make a lot of mistakes - they just backspace and correct to eventually produce a perfect copy. Try using a hand plane to smooth a piece of wood rather than an electric plane - the former takes a lot more skill than the latter.

      Delete
  9. I think the fear of automation is ironic, since it is most often brought up by people who's value systems are oriented around equality of outcome. It seems to me that the clearest path to a world where people have similar degrees of freedom is one where the burden of necessary but menial tasks has been satisfied by technology. I recall reading a story by Isaac Asimov about a future where entire planets hosted one or two families and most tasks were carried out by robots. Setting aside philosophical debates about AI and consciousness, this sounds like the utopia everyone dreams about.

    On a more practical note, my elementary understanding of economics leads me to believe that an increase in the cost of labor favors an increase in the use of technogly. The decline in the male participation rate in the workforce might have as much to do with the additional costs, which have priced them out of the market, as it has to do with automation. I think these costs combined with the barriers which are placed on entry to new business might have created a more sluggish and less adaptable labor market. For example, if you need to be retrained from a typing pool to a more intensive role like becoming s paralegal, but have no skill set in that area, the inability to offer to work below a certain price floor really removes the most persuasive negotiating power you have.

    The ease of creating business and hiring employees appear to me to be the real drivers behind the structural unemployment which is laid at the feet of automation. To make matters worse, for an alarmingly high number of college undergraduates, there is a serious gap between what skills employers are looking for and the education young people receive. This gap is encouraged by higher education subsidies, which ask young people to make expensive choices, without any work experience or in many cases qualified advisors, about how much education is worth.

    These problems (higher costs of labor, higher start up costs for businesses, higher costs of educations and gap in skill sets) are not solely due to cheap technology, and some of them are not due to technology at all. The claim that automation will be responsible for driving people out of work or the labor market is spurious and, I think, elitist. It is founded on the idea that the most meaningful contributions humanity are those things which can be automated by robots.

    ReplyDelete
    Replies
    1. Almost certainly, you are correct on anti-business government action and wild regulation is the main source of unemployment, not automation. 150 years has entirely transformed the work being done and more people than ever are employed while 95% of the original jobs are gone.

      In the last 8 years, regulation and conformance has easily increased 50% as a time load for C-level executives for businesses under $50M, and under $5M, it has consumed them. All this time squandered has come at the expense of the number 1 job of C-levels', namely increasing business. This is the source of unemployment and under employment.

      Delete
  10. NO, no, no we won't be put out of work by machines, robots and innovations there will always be work to do and some of it more interesting and rewarding than what the robots and machines will do; but as real incomes continue to rise (growth rates of income may slow down but doubt it goes to zero) we may choose more leisure time - time with family and friends and travel and reading and music and art appreciation, and boating and sailing and tennis and golf. All this by choice not because of government programs- though we should be careful about any further increases in income inequality and hopeful that we can get less income and wage inequality - strengthening education may be a key - both so workers can handle the more interesting jobs and to appreciate and enjoy more leisure time activities. Even at 2 percent annual real income growth, future generations will be much richer than we.

    ReplyDelete
    Replies
    1. I think upward mobility is a larger problem than income inequality. Savings and work are not uniformly valued, so it's unfair to judge the distribution of wealth without considering effort.

      Delete
  11. Isn't it basically a truism that automation cannot, in the long run, reduce employment (aside from people choosing not to work due to increasingly generous unemployment benefits resulting from greater productivity)? It's really just the *pace* of automation that impacts short term structural unemployment (I guess short term here could mean as much as a generation, though).

    ReplyDelete
  12. Jobs v automation? This question was answered by Parkinson in 1955.

    ReplyDelete
  13. I know this article compares men and women's workforce because of automation, but we need both men and women in the workforce to help grow our economy (more people in workforce, better economic growth). Also we didn't start consider hiring women to workforce not until recently. So of course we will see women in workforce increase and mens were declining. But that does not explain women took men's job. This might be men do not like changes so they stay with old jobs (?) i dont know. but this reminds me of the movie "Hidden Figures". Middle of the movie, African American or "Computers" were working doing math and other type required job in NASA. But IBM came with machine that can do lot more and lot faster job than human can.They were verge of losing job, but the manager learned how to run the IBM machine and all of their "Computers" move from old job to the new one. This one clearly shows Automative machines took jobs but also created another. This only helped people who seek for new jobs and willing to learn new ways. It still happens today, doctors needs to re-new their license, many computer science major need to learn new technology and even car mechanics needs to learn new cars to fix. but requires less people to do job, so we might not see huge decrease in workforce in early but it will gradually reduce in numbers.

    ReplyDelete
  14. I know this article compares men and women's workforce because of automation, but we need both men and women in the workforce to help grow our economy (more people in workforce, better economic growth). Also we didn't start consider hiring women to workforce not until recently. So of course we will see women in workforce increase and mens were declining. But that does not explain women took men's job. This might be men do not like changes so they stay with old jobs (?) i dont know. but this reminds me of the movie "Hidden Figures". Middle of the movie, African American or "Computers" were working doing math and other type required job in NASA. But IBM came with machine that can do lot more and lot faster job than human can.They were verge of losing job, but the manager learned how to run the IBM machine and all of their "Computers" move from old job to the new one. This one clearly shows Automative machines took jobs but also created another. This only helped people who seek for new jobs and willing to learn new ways. It still happens today, doctors needs to re-new their license, many computer science major need to learn new technology and even car mechanics needs to learn new cars to fix. but requires less people to do job, so we might not see huge decrease in workforce in early but it will gradually reduce in numbers.

    ReplyDelete
  15. Economies are zero sum. Consumers are workers, workers are consumers. Eliminate workers and you eliminate consumers.

    Designating government as the consumer with people as proxies taking in taxes, or inflation, the difference in business revenue less wages paid (and spent) merely compensates for failing to cut prices down to labor cost (capital and production). In an economy with zero workers, prices must be zero.

    ReplyDelete
    Replies
    1. Economies are zero sum? So the planet's wealth has not increased in the last 100,000 years? We're not living in caves you know.

      Delete
    2. I am going to disagree that economies are zero sum. I would enjoy seeing an economical model which explains the last 10,000 years and is zero sum. I would touch on the idea that government is the consumer by proxy of people or that no workers equates no consumption but in the words of Wolfgang Pauli 'That is not only not right; it is not even wrong'.

      Delete
    3. Whether our economy is zero sum or not; the issue of declining overall employment participation (men ~ 70%; women ~ 60%) is a real problem, and the issue that automation more dire3ctly faces. Millions of middle pay jobs are going open and recent analyses suggest both low pay and high pay jobs are increasing, but the higher paid jobs are unfilled and middle class employees are being forced into low pay jobs. At the same time GDP is increasingly more capital than labor. Is this affecting demand now? and how low can labor participation go before a real calamity occurs?"

      Delete
  16. I would think that whether or not automation displaces many jobs will have to do more with demand rather than supply. If in the future there are lots of new types of things or experiences that people want, and it takes a lot workers to supply those things or experiences, those who are being thrown out of work by automation will find employment. On the other hand, if we all decide we're pretty happy with what we've got, or the creation of new things requires input from very few workers, there will be more unemployment.

    ReplyDelete
  17. The scary thing about automation is not necessarily jobs going away, it's that the new jobs are more cognitively demanding (and less physically demanding) than the old jobs. What happens when entrepreneurs simply don't know how to utilize low-skill labor anymore?

    ReplyDelete
    Replies
    1. Just as operating a word processor is more cognitively demanding than operating a typewriter.

      Delete
    2. That was exactly my issue above. I fear the increase in cognitive function required this time is much greater than previously. It's a lot easier to learn how to use a word processor than to become a software engineer.

      Delete
    3. Not everyone has to become a software engineer. Most new jobs are going to be in services, using computers and more and better and simpler apps, not programming them. In my experience most traditional secretaries did not make the transition. People type their own stuff, and most traditional secretaries never learned how to do simple html coding or things needed around offices today. University secretary hutches are empty, or filled with research assistants.

      Delete
    4. People do cognitively intensive activities everyday. Millions of people drive cars making split second decisions to balance safety and comfort. People also parallel park, which is a complicated task. Humans are adapt at reorganizing themselves to new cognitive tasks.

      Delete
    5. You are up against human wetware limitations. We evolved to defend ourselves from large predators, hunt, and gather. That's it. If in the primitive economy you were a crummy hunter, you'd still be a passable gatherer, so everyone would have a place. Evolution selects for survival utility.

      Now, our amazing brains do allow us to accomplish things that evolution did not select for. We didn't evolve to drive cars or write programming code. That is a happy evolutionary accident. But today, employment and productivity is dependent on the "happy accident" portion of our heritage. Things might work out where everyone has a productive job, *but it might not*. Especially when you see the targets of automation, almost always low-cognition physical labor. Throw in cutting-edge robotics and AI research, and it is easy to imagine a world where a large chunk of people (again, who evolved to hunt and gather) are completely obsolete.

      Delete
    6. Even a simplistic understanding of human cognitive abilities has to recognize the bell curve, whether this is a fundamental difference among people or a motivational one. Both are very difficult to alter. Talking to many auto workers who were laid of by automation, I am shocked by how few had even used the internet and computers. Those who chose education rather than low level jobs wound up earning less in the longer run. This may be an aspect of the bell curve. Re-training is not a certainty and skill limitations may b e crucial for the future of labor participation.

      Delete
  18. In 1969, I entered the then-nascent field of automated material handling: computer-control of conveyors and automated storage and retrieval machines. At that time, labor unions looked upon those in my profession (computer programming) as enemies, fearing the very situation that this article discusses -- the elimination of working-class jobs. There has been a huge loss of decent-paying working-class jobs in our nation but it has not been principally due to automation, as the unions feared four decades ago. Instead, it is more the result of the development of manufacturing capacity by countries whose workers were paid less than ours.

    In America today, there are over three million available jobs for which employers cannot find workers with adequate skills. People are not entitled to a job, only a fair opportunity to compete for the ones that are available. It is up to the individual to acquire the skills necessary to be hired. Our government may help to make training more available but, ultimately, it is the worker's responsibility to acquire the skills necessary to find employment.

    My profession did away with some warehousing jobs -- but others were created. The technology we developed resulted in the huge distribution centers that today fill business parks across America and employ millions of workers. These jobs required new skills and, somehow, people acquired them. I believe this process will repeat itself again, as it has throughout history. And this is Mr. Cochrane's premise.

    ReplyDelete
    Replies
    1. Personal anecdotes are always interesting and fun, but don't compare to broad analyses. There is a very persuasive graph from David Autor that shows a powerful set of job changes over the last decades showing loss of middle class jobs and higher paying jobs going unfilled and yet declining in number as well. Only lower paid jobs are increasing, and this augurs poorly for demand, except for luxuries for the rich.

      Delete
  19. Well it's true that it didn't in the past. But I would assert that it clearly will someday. Humans are just a special-case of a machine. The services and problems we can solve with our bodies and brains are not infinite, and incrementally we move forward to a world where we continue abstracting these away, and automating them. No one can predict with foresight how fast this happens, or where it will end up. But it's also a mistake to look at the 20th century economic process that generated this data, and claim it will let us predict how automation/labor will interact in the 21st, 22nd 23rd... centuries.

    ReplyDelete
    Replies
    1. When I was young my father had a picture of him as a child up on the wall. People would say 'you looked just like him when you were young' to my father. To which he would reply 'he looks just like me'.
      Its machines which are like humans, not the other way around.

      Delete
    2. It is already true that computers (we don't need to talk about Ai - AI is just another set of math and software on computers) can do as well as the best people more cheaply.

      Delete
  20. The problem isn't that automation destroys jobs, it's that the workers in the new jobs don't have their old bargaining power. It's a cycle. New technology, old jobs vanish, new jobs are created, but too many people wind up making way too little. The workers eventually get their act together - Chartist Movement, labor movement, New Deal - and get a piece of the rising productivity. Then newer technology comes in and the workers have to scramble for bargaining power.

    We saw this with the industrial revolution, the late 19th century modernity revolution, the 1920s electrification revolution and now again with the computer revolution. Each time the workers had to fight for a share of the productivity gain. It wasn't always pretty.

    ReplyDelete
    Replies
    1. Over the last few decades, workers have lost their share of productivity increases, and it has pretty much gone to those with capital or high professional skills. This is a profound change that recalls the financial barons of the early 20th century.

      Delete
  21. I’m working on this topic, though I’m not trying to answer whether all jobs will be replaced by computers. On page 1 and 2 of my paper (link below), you can find a little survey of current debates about the relationship between automation and our macroeconomy.
    The supply of robots has been, for some reason, speeding up after the crisis of 2008. Although most authors are looking for the effect of automation on the labor mkt, I take a different approach. And I find that…the risk of automation makes us worse off!

    https://www.jsie.jp/kansai/wp/wp-content/uploads/Tsuboi_Paper_170722.pdf

    Short summary (i.e. slides):

    https://www.jsie.jp/kansai/wp/wp-content/uploads/Tsuboi_Slides_170722.pdf

    ReplyDelete
    Replies
    1. Thank you for your link. The effect of increasing computer capabilities on people's motivation to improve their education and training when they see this rapid replacement of work by computers is undeservedly neglected by economists. It may be a powerful reason for the appeal of authoritarian policies in a democracy. It certainly has the power to destroy an economy.

      Delete
  22. More interestingly, suppose robots did the work, and wages for workers just ended. Who would consume the product -both goods and services - of machines?

    Answer: Equity owners. The speed of such a change probably depends on how fast people become equity owners. Equity results from invested savings and dividends. The government would need to generate forced savings (Social Security) as well as tax and budget reductions, and redirect expenditures toward activities with a positive return, e.g., public and individual equity. The private sector would need to use public and private savings fast enough to increase productivity, and growth, inventing new goods and services, to keep up with market demand, and the probable acceleration of the rate of savings. We have a long and exhausting road to run to keep up with this possible prosperity.

    ReplyDelete
    Replies
    1. In economics, we usually stop to consider whether there is some market mechanism for achieving a transition, and then whether there is some documentable failure of the market mechanism before running to government intervention. It is just remotely possible that some human and physical capital investment can occur without the government "genererating "forced savings (of which Social Security is the opposite BTW), "redirecting" expenditures towards activities that it decides have positive return (something it's not been terribly good at historically, see housing subsidies).

      We've been at this since Og invented the stone axe 200,000 years ago. New labor saving devices,... save labor.

      Delete
    2. A sovereign wealth fund sharing the physical and land capital of the nation with everyone is the only answer. Market failure is already obvious in the enormous inequalities that robots will only accentuate.

      Delete
    3. I let a few of these comments from "unknown" through to share with my readers. It looks like the capabilities of computers to emulate people and spam blogs has taken a radical unsurge!

      Delete
  23. Someone who copied manuscripts in the Middle Ages was actually a highly skilled artist. A good typist is an artisan. A word processor operator can make lots of mistakes and still produce error free copies. Each step in the progression of jobs can be performed by less exceptional workers.

    ReplyDelete
  24. Technological progress according to Mankiw is labor augmenting, it doesn't increase the amount of workers, but makes each worker more productive over time. The fact that the male participation rate decreased from 87.5 to 70, is most likely due to multiple reasons, not from women joining the labor force. We have seen throughout history that even though we increase in automation, say construction, agriculture, even transportation, new jobs will always be added and there will always be a need for a workforce that can maintain everything.

    ReplyDelete
    Replies
    1. No, we "assume" that technological progress is labor-augmenting, for technical reasons (warrant the "balanced growth", etc...), but not necessarily. If you want, you can have capital-augmenting technological progress, or "neutral" technological progress. In symbols (A: Technology/ K: Physical Capital / L: Labor), you can write: Y = F(K,AL) or Y = F(AK,L) or Y = AF(K,L). Of course, I simplify a lot, but these are points. If you want to talk about "technological unemployment," you have to write down your model! You may also want to check the term "creative destruction."

      Delete

Comments are welcome. Keep it short, polite, and on topic.

Thanks to a few abusers I am now moderating comments. I welcome thoughtful disagreement. I will block comments with insulting or abusive language. I'm also blocking totally inane comments. Try to make some sense. I am much more likely to allow critical comments if you have the honesty and courage to use your real name.