Aaron Saenz at Singularity Hub had post a while back with some great videos showing the extent to which manufacturing has automated. I’ve included a couple of the videos below; you can see the rest here.
As you watch these, keep in mind that automation is going to rapidly penetrate nearly every employment sector—not just manufacturing. If you have a “knowledge-worker” job, automation will not require investment in expensive machines or robots: it will only require a sufficiently sophisticated software automation or specialized artificial intelligence application. (By “specialized artificial intelligence” I mean the type of expert systems that currently land jet aircraft and perform thousands of other tasks. These types of systems will get better and better and will be able to take on a much broader range of jobs and tasks. Automating the vast majority of jobs—even those that require college educations and high skill levels—will not require the “general” or human-like artificial intelligence that we see in science-fiction movies.)
As Aaron says:
As robots continue to step in and increase factory productivity, new and cheaper goods will become available. If we plan it right, that will mean we can spend more time being creative, relaxing, and enjoying the fruits of our mechanized labor.
Yes, but the problem is that if you don’t have a job, then you don’t have an income. In the U.S. you don’t have a right to health care, or the right to a home. So how exactly are we going to get to the point where we can begin “enjoying the fruits of our mechanized labor?” (“We” meaning the majority of the population—not just the top few percent.) That’s the question that I try to explore in The Lights in the Tunnel (which you can now download in PDF format for free).
We keep hearing that the unemployment situation is going to begin turning around any minute now, and previous disappointing results have been dismissed and excused on the basis of weather and other factors. Now the latest weekly report shows an increase of 25,000 in new unemployment claims—instead of the decrease expected by economists.
Clearly, the economy continues to struggle with job creation, and I think that automation is playing a significant role. The fact is that businesses continue to invest in labor-saving technologies during downturns and then do not need to re-hire as many workers once recovery begins. I think there is a very real risk that unemployment will continue to outpace expectations and that the resulting impact on consumer spending will threaten long term recovery.
For the last two years, the weak economy has provided an opportunity for employers to do what they would have done anyway: dismiss millions of people — like file clerks, ticket agents and autoworkers — who were displaced by technological advances and international trade.
Of course, the article accepts the convention wisdom which says that this story is only going to occur in select areas of the workforce (file clerks, travel agents, etc.) and that it will be temporary as most workers will eventually adapt by re-training, etc. What if it is not temporary? What if the trend accelerates and begins to impact a broader and broader segment of the workforce as advancing technology makes even highly skilled jobs obsolete?
As the article suggests, very often the job of last resort in cases like this is a part-time position at Wal-mart. What happens when Wal-mart begins to automate? Do people really believe that will never happen? Wal-mart management was already starting to think about it back in 2005.
Ryan Avent has a post and nice graph at the Economist‘s Free Exchange blog about rising labor productivity over the past 25 years or so:
The red line is the annual change in output per hour, and the black line is the ten-year moving average. There’s a story here relating to technology. And one of the interesting subplots to that story is how the revolution in computing and communications technology is impacting different occupations differently. Some are more or less unaffected (typically non-routine manual tasks, like janitorial work). Some have become far more remunerative as a result (typically non-routine abstract tasks, like development of products that can be sold globally). And some have been destroyed by the shift. Routine professions in manufacturing and business have been laid to waste by improvements in computational power, and the resulting effect on automation and offshoring.
As I’ve pointed out here many times, there is no reason to expect that this trend won’t continue–and there are probably good reasons to anticipate that it will actually accelerate as technology continues to advance. I think it’s very likely that within the next decade (or perhaps two at the outside), full automation technology will begin to encroach on the work capability of most average workers—including those with significant training and education.
The Economist also has an article on “The Polarization of the American Workforce,” which takes the conventional view that automation and offshoring are destroying low and middle skill jobs while creating plenty of new jobs for the highly educated and skilled. That may have been the case so far—but I don’t believe it will continue to be true indefinitely. Automation technology (as well as offshore workers equipped with increasingly sophisticated information technology tools) is going to increasingly move up the food chain and and go after those high paying skilled jobs. As the years and decades pass, the number of people in protected occupations is going to continue to shrink.
Fed Chairman Bernanke’s speech in Atlanta earlier this year focused a lot of attention on the debate about the forces that led to the current crisis. Was it primarily Greenspan’s low interest rates, or as Bernanke suggested, was a lack of regulation and oversight a more important issue? Nearly everyone seems to agree that the crisis was brought on by some combination of these factors. In other words, the current situation is viewed almost entirely in financial terms.
While there is no doubt that the financial meltdown was the proximate cause of the current recession, is it possible that a more fundamental cause exists? What if the housing bubble and the financial crisis were merely symptoms of something even bigger—perhaps of a structural shift occurring in the broader economy?
I’ve argued previously that advancing technology is likely to result in structural unemployment in the future. In fact, I think this is a trend that is already well underway. The last decade has been characterized by substantial advances in information technology and fairly dramatic increases in productivity. Average workers have seen stagnant or even decreasing real wages, while health care costs have been exploding. Until the onset of the current crisis, official unemployment numbers were low, but those statistics fail to capture underemployment, such as workers who are forced to work multiple part time jobs with no access to benefits.
Globalization, of course, gets much of the blame for the plight of average workers, but the reality is that advancing technology has a larger impact. Jobs are not just moving to China—they are being automated away completely. This is happening not just in the United States but in low wage countries as well. And it isn’t just in manufacturing; as I’ve pointed out previously, service sector and knowledge worker jobs are increasingly subject to automation as well.
As the dual forces of technology and globalization progressed over the past decade, I suspect it became pretty clear to most average workers that holding a job at the prevailing wage offered little hope for getting ahead. Recognition of that reality certainly played an important role in the politics that led to the creation of subprime lending programs. You can make a pretty strong case that the housing bubble was caused not simply by low interest rates but by widespread recognition that investing in a home represented perhaps the only viable hope for a typical American family to achieve any measure of prosperity.
The last decade also saw a massive shift away from consumer spending supported by wages toward spending supported by debt (much of it anchored to inflating housing values). That debt-enabled spending drove economic growth not just in the U.S. but, of course, in China and in the rest of the developing world as well. Was all this really caused by the Fed’s policy? Or was it fundamentally caused by advancing technology (as well as globalization) driving discretionary incomes down to a level where broad-based consumer spending became unsustainable without reliance on debt?
I think that is a very important question. Virtually all mainstream economists are focused on a financial solution to the current crisis involving some combination of monetary policy, additional stimulus and re-regulation. If it it turns out that the root cause of the crisis is really technology, rather than finance, those solutions will simply not be sufficient in the long run. Technology is relentless. Automation will never stop progressing, and no conventional financial or monetary policy can, by itself, address that issue.
In my book, The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future, I make the case that a basic shift is occurring in the economy. Technology is becoming autonomous, and job automation will invade virtually every employment sector. The result will be structural unemployment and declining wages for all but a tiny (and shrinking) elite. I think it is very possible that the beginning of that trend underlies the current crisis to a significant extent. I suspect that very few people will agree with me on this, but if I’m right, the implications are scary: it means that virtually everyone is focused on solving the wrong problem.
Ericsson is sponsoring a new website called “2020 — Shaping Ideas.” The site will feature short videos that present the ideas and predictions of twenty significant thinkers regarding the year 2020. Areas covered include technology, economics, social trends, media, etc. Some of the people featured include Jeffrey Sachs on fighting poverty, Carlota Perez on economics and Ian Pearson on future technologies.
I was asked to comment on some of the videos, and I thought I would start with a fascinating presentation on the the “Rep Rap” machine — a personal-scale 3D printer that is cabable of replicating itself. (A 3D printer is a device that is able to automatically fabricate solid objects by building up successive layers of material.) The Rep Rap machine is the invention of Adrian Bowyer of the University of Bath in the United Kingdom.
Dr. Bowyer’s vision is that we all might one day own a personal 3D printer and use it to fabricate products. If the technology were to indeed follow that path, one can imagine a world in which physical products become much like digital products today. In other words, all the value would be in the intellectual property—presumably people would need to pay for a license to the design specifications, but the cost of actually creating the product would be minimal. Many technologists believe that if you carry that tend far enough into the future, advanced nanotechnology would come into play, and it might be possible to literatally assemble materials and objects through molecular construction or self-replication. That could conceivably result in something like the “materializer” on Star Trek.
Personally, I’m a bit skeptical that things will go that way, and almost certainly not in the near term. I suspect that technologies like 3D printing are more likely to have a high impact in the commercial arena for the simple reason that small, inexpensive machines are unlikely to be competitive with leading edge, commercial grade machines. There is, of course, also the problem of assembling the parts into a working machine once you have fabricated them.
Nonetheless, I think technologies like 3D printing are likely to have a transformative effect on manufacturing, especially when combined with other forms of automation, such as robotic assembly. In the video, Dr. Bowyer asks, “What happens if we destroy all of manufacturing?” While I don’t think that’s a major concern in the foreseeable future, I am a lot less sanguine than Dr. Dowyer about the potential impact on employment. As technologies like this continue to advance relentlessly, there is clearly a point at which the participation of human workers in the production process will essentially be eliminated. The economic impact of that transformation is difficult to overstate—especially in low-wage developing countries heavily geared toward manufacturing.
In my book, The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future, I argue that technology is likely to transform manufacturing and ultimately perhaps even reverse globalization for manufactured goods. As factories begin to approach full automation and labor costs fall to near zero levels, the savings that come from locating in a low wage country are likely to be outweighed by the transportation costs associated with moving products to market. That will be especially true if, as many analysts predict, oil production peaks and energy costs escalate significantly in the future.
If that scenario unfolds, it seems quite possible that manufacturing will see a “jobless repatriation” where highly flexible automated production facilities are built close markets in order to minimize transportation costs. That would have dire consequences for employment, especially in countries like China.
Most economists, like Dr. Dowyer, don’t seem to be espeically concerned about this trend. They assume that if manufacturing jobs disappear, workers will find jobs in other areas—most likely in the service sector, where jobs “can’t be automted.” The problem with that is that jobs in the service sector can and will be automated. After all, a sufficiently advanced automation software or artificial intelligence application is essentially a “3D printer” for the service sector—and for many jobs it may actually represent a lower technical hurdle. If you really think through the implications of advancing technology it becomes pretty clear that there really aren’t many employment sectors that will be protected in the long run.
I’ve been arguing here that advancing technology is likely to lead to a serious structural unemployment problem in the long run (and perhaps we are seeing the beginning of that trend now). Marketwatch reports that a new study makes exactly the opposite prediction: as the workforce ages, there’ll be a shortage of workers and we may see the beginnings of it within the next 2-3 years.
How did the report’s authors arrive at this conclusion? First, government analysts expect 14.6 million new nonfarm payroll jobs will be created between 2008 and 2018. Including self-employed workers, family members working in family businesses and workers in farming the total hits 15.3 million new jobs. Read the Bureau of Labor Statistics’ employment projections.
Next, given the government’s projected population growth and current labor force participation rates — and assuming no major changes in immigration — there will be about 9.1 million additional workers over the same time period. Taking into account multiple job holders, the total number of jobs expected to be filled is 9.6 million.
Finally, subtracting the projected number of filled jobs from the expected number of new jobs results in a range of 5 million to 5.7 million vacant jobs. However, using projected labor force participation rates – baby boomers are not expected to retire at as high a rate as earlier cohorts of older workers — there would be 3.3 million to 4 million vacant jobs, according to the report.
While I don’t doubt that we will see shortages in certain occupations (can we really expect people to line up to be a “home health aid?”), I’m very skeptical of these numbers—and of any approach that attempts to predict the future purely on a numerical basis without taking into account the structural and technological shifts that are going to continue (and very possibly accelerate) in the coming years. I wonder how they arrived at the “14.6 million new jobs created” number.
I think it’s very likely that estimates like these significantly underestimate the pontential impact of both job automation technology and offshoring. Automation—in the form of robotics as well as software automation and focused artificial intelligence applications—will gradually penetrate into nearly every employment sector and threaten both skilled and unskilled workers. In many cases, higher paying jobs, especially knowledge worker positions, will be among the most vulnerable.
Where automation is not yet capable of taking over a job, management will increasingly turn to offshoring. As software continues to advace, the result is likely to initially be tools that enable low wage offshore workers to perform increasingly complex, high value jobs. Eventually, those software applications will, in many cases, advance to the point where they become autonomous—eliminating the job entirely. I also think it is likely that both automation and offshoring will become less expensive and easier to access—in other works, they will become more accessible to the small businesses that drive most job creation in the U.S.
While demographics may gradually reduce the number of available workers (relative to the total population, not in absolute terms), technology has the potential to rapidly advance in a disruptive way—possibly eliminating entire job categories that include millions of workers in a very short time. Yes, advancing technology will also create new industries, and therefore new jobs, but history has shown pretty conclusively that new technology industries tend to not be very labor intensive—they are capital intensive and do not create large numbers of jobs.
As new industries employ only a few people and older industries get disrupted by advancing technology, things may get ugly. Compare Google with McDonalds. What happens if McDonalds automates and someday begins to look more like Google in terms of the workforce it requires?
While, I feel confident that these trends will evenually develop, it is very difficult to project the exact timing. My guess is that we’ll begin to see significant evidence within the next decade and that by 2020 there will be significant doubt surrounding the historical assumptions about job creation. As I noted earlier, the San Francisco Fed recently published a study suggesting that assumptions about the relationship between econonic growth and job creation are changing. In any case, numbers such as the ones from the study cited above are likely to be meaningless.
The Work Design Collaborative, an organization which conducts research into how the nature of work is changing due to advancing technology, a changing workforce and new workplace designs has reviewed The Lights in the Tunnel:
If you care at all about the future of the economy, the future of work, and the future of society, you will find this a very provocative read.
The Federal Reserve Bank of San Francisco just issued a report showing how Okun’s Law—the historical relationship between changes in GDP and unemployment—is starting to break down. In other words, rising productivity is resulting in less job creation than has historically been the case.
In 2009, strong growth in productivity allowed firms to lay off large numbers of workers while holding output relatively steady. This behavior threw a wrench into the long-standing relationship between changes in GDP and changes in the unemployment rate, known as Okun’s law. If Okun’s law had held in 2009, the unemployment rate would have risen by about half as much as it did over the course of the year.
In my book, The Lights in the Tunnel, I argue that the heavy emphasis on econometrics (statistical economic data analysis) in modern economics is problematic because it is entirely focused backward on historical data. Economists generally refuse to accept any ideas that are not backed by hard data and quantitative analysis. However, all the “hard data” is OLD—often years or even decades old. If, in fact, the exponential acceleration of information technology is going to have a dramatic impact (as I believe) in the future, that is going to be very hard to pick up looking exclusively at past data! I think the only way to see what’s coming is to think through the implications of the actual technologies likely to be developed in the coming years and decades.
Nonetheless, it looks like the economists at the SF Fed have started to pick up something in the data. I think as time progresses there will be more and more quantitative evidence to support the theories that I am advocating here. The problem, of course, is that by the time the PAST data is unambiguous and everyone has no choice but to acknowledge the problem, the future is going to look very bleak indeed. Perhaps the economists should spend a little less time looking at their historical data and a little more time looking at the computer they are using to analyze it—and how the capability of that machine and the software that runs on it is changing over time.
I’ve arguing here that automation technology (as well as offshoring and globalization) is likely to depress wages and lead to significant structural unemployment in the coming years. One of the most common criticisms of my argument is that I am “not thinking like an economist” and that I’m viewing things in terms of dollars, rather than in terms of purchasing power.
Here’s part of a comment that James D. Miller, an economics professor at Smith College, made in response to one of my previous posts:
Non-economists (even when they are very smart and well-studied) get in trouble when they consider trade issues in terms of dollars (“everyone could work for a dollar an hour?”) It’s better to think of wages in terms of what you could buy. In a world with hyper-productive robots you could buy lots of stuff if you could work at a task for say 1,000 hours that saved a robot 10 seconds of time.
So the basic idea here is that although automation may result in very low wages in dollar terms (as well as high unemployment, since we do still have a minimum wage), things won’t be so bad because the efficiency of production will increase dramatically and everything will be really cheap.
To see the problem with this, view this graph at Visual Economics showing how consumers spend their incomes. The graph makes it immediately clear that consumers spend the lion’s share on their incomes on things like housing, insurance, health care, transportation and food.
“Hyper-productive robots” are not going to lower anyone’s mortgage principle, and interest rates surely cannot go much lower. Nor can rents adjust too far downward without threatening the landlord’s mortgage. The same is true of insurance. The reality is that the most of the average consumer’s budget is based primarily on asset (and debt) values—and not directly on how efficient the economy is at producing goods and services. Food and energy prices are likewise unlikely to adjust downward. Expenditure categories that might see falling prices as automation progresses, such as apparel, entertainment and miscellaneous represent a tiny fraction of the average budget, and in many cases prices have already been minimized by globalization.
The only way to have expenditures fall in line with wages so that consumers could maintain their standard of living would be to have asset and debt values collapse. And that, of course, would be catastrophic for the financial system. Asset values in the United States reflect the basic assumption that we are going to continue to have a vibrant mass-market economy and a first-world living standard. You cannot have third-world wages with first-world asset values. That is the reason that countries like Thailand prohibit foreigners from buying property and driving values beyond the reach of their population.
As wages fall and unemployment rises, the average consumer is going to be squeezed by the fixed costs that cannot adjust downward. Mortgage defaults would soar and discretionary consumer spending is likely to plummet.
A recent article in U.S. News noted that spending is already heavily concentrated among high income consumers:
The top 10 percent of earners account for 22 percent of all spending, for instance, according to Moody’s Economy.com. The top 25 percent of all earners account for 45 percent of spending. The bottom 50 percent of earners, by contrast, spend just 29 percent of all the money in the consumer economy.
As job automation (and globalization) drives down wages and creates structural unemployment, these numbers will become even more concentrated. At what point does this become unsustainable? In an environment with extreme financial stress due to loan defaults and falling asset values, can the wealthy few really drive consumer spending indefinitely? Recent history shows very clearly that when fear is pervasive, rich people stop buying as well. So where will consumer spending come from?
Martin Ford 