It seems like the United States economy is enjoying more innovation than ever before. At the same time, statistics show the economy suffering from its slowest growth in decades.
Analysts often try to resolve this by arguing either that conventional statistics aren’t properly measuring the value of innovation or else that the apparent speed-up in innovation is actually an illusion and progress is slowing down. Another possibility is that things are exactly as they seem: Rapid technological innovation is real, and so is slow economic growth. In fact, in a sense the innovation is causing the slow growth.
Call it the productivity paradox, and recognize that it explains a lot about the current state and the future direction of the American economy.
Progress is real, but uneven
Most of the progress in recent decades has involved making cheaper and more convenient versions of products that already existed. Phone calls and photographs, for example, have been around for a long time, but over the past 20 years we’ve seen a revolutionary decline in the cost and massive increase in the convenience of snapping photos or making long-distance calls.
As innovation has pushed down the cost of certain types of products (mostly durable goods such as televisions, furniture, and clothing), Americans have used the savings to spend more on other things — especially education, health care, child care, and housing — where productivity growth has been much slower.
Over time, low-productivity sectors have become a larger share of the economy, while high-productivity goods production has become a smaller share. And an economy dominated by industries with low productivity growth is going to grow slowly.
Slow growth sounds bad, but the future implied by the productivity paradox isn’t actually so terrible. It means that in the future a small minority of people will produce the world’s material goods and automated services, while the rest of us are focused on providing personalized services to each other. It’s a future of material abundance and plentiful jobs.
Indeed, one way to think about it is that middle-class Americans are getting close to enjoying as much material comfort and convenience as it’s possible for any society to provide for ordinary people. Accumulating even more stuff isn’t going to make us much happier, so we’re devoting more and more of our incomes to personal services that don’t see rapid productivity growth but do a lot to make our lives better.
Productivity growth can make an industry shrink
Two things can happen when a given industry enjoys soaring productivity — it can expand, as new production techniques lead to a surge in output and consumption, or it can shrink, as a smaller and smaller number of people are needed to serve a fixed market. The history of the textile industry provides examples of both dynamics.
Americans started to mechanize the production of cloth in 1814. Productivity per worker steadily improved, and at first these gains powered a rapid expansion of the textile industry. People in the 19th century owned very few clothes, so as cloth got cheaper, people bought more.
But by the 1950s, this process had reversed. People already had plenty of clothes, so as prices continued to fall, people just spent less on clothing and pocketed the savings. As a result, apparel spending as a share of the typical household’s income has fallen steadily for the past 60 years.
Agriculture tells a similar tale. In 1900, farming was the biggest industry in America, employing about 40 percent of workers. A century later, farms were more productive than ever, but they employed less than 2 percent of the workforce. And because people only need so many calories, spending on food has plummeted.
In the mid-20th century, consumers took money they were saving from cheaper food and clothing and used it to buy a wide range of other manufactured goods that had recently been introduced: telephones, electric lighting, cars, radios, washing machines, refrigerators, televisions, air conditioners, and so forth.
Most of these products experienced a life cycle similar to the one I described for clothing. Cars went from nonexistent to a rare luxury item to a mass-market consumer product that employed a vast army of factory workers. The same happened for vacuum cleaners, telephones, refrigerators, and other household applies. But eventually, we reached a point where almost every home had a refrigerator and there was little room left for quality improvements. At that point, further productivity gains in manufacturing mostly meant that refrigerators got cheaper and consumers spent less on them.
This process can take a few decades, which means the many significant inventions between 1900 and 1940 allowed the manufacturing sector to continue growing through the 1970s.
The past few decades have been different, as economist Robert Gordon has argued. The list of major new inventions in the past 40 years is pretty short, and it’s dominated by computing gadgets — PCs, gaming consoles, DVD players, smartphones. In most other areas of our lives, Americans largely buy the same things we bought 20, 40, and even 60 years ago.
Innovation in manufacturing hasn’t stopped. American factories now produce about twice as much per worker as they did in the 1980s. But we’ve mostly gotten cheaper goods or modest quality improvements — not the invention of major new product categories. As a result, the manufacturing sector is winning a smaller and smaller share of consumer spending.
And as the chart above shows, this isn’t just an American phenomenon — and it’s not primarily about jobs moving overseas. Manufacturing’s share of the global economy has been shrinking, just as it has in the US.
When thinking about manufacturing’s declining share of the economy, it’s easy to think there must be something wrong with the manufacturing sector. But the truth is closer to the opposite: Manufacturing industries are victims of their own success. In recent decades, the manufacturing sector has consistently enjoyed higher productivity growth than the economy as a whole. Manufacturing is shrinking relative to the broader economy precisely because it has continued to get more productive even as demand for manufactured goods plateaued. That’s the productivity paradox.
Automating services could boost growth
With few new manufactured goods to spend money on, consumers have devoted more and more of their income to industries where productivity growth is slow or non-existent. These tend to fall into two big categories. Some industries, such as health care, education, and child care, suffer from low productivity growth because they are labor-intensive and difficult to automate. Others, notably housing in affluent areas, have become more and more expensive due to natural and legal limits on supply.
So to deliver high growth rates over the coming decades, one of two things would have to happen. One possibility is a series of major new inventions — perhaps flying cars, space tourism, holodecks, or nanorobots that cure cancer — big enough to entice consumer dollars away from low-productivity service industries. This seems unlikely to me — it would take some really impressive breakthroughs to reverse a 50-year trend — but it’s impossible to rule it out.
The more likely option is to figure out ways to automate industries that are labor-intensive now. An obvious example would be taxi and truck drivers being replaced by self-driving vehicles. If this kind of thing happened across a range of other industries, economic growth would soar. And many people think artificial intelligence software is about to make that happen.
I think that’s unlikely no matter how sophisticated AI software gets. The reason is that the productivity paradox operates within industries as well as among them. To see how this works, consider the case of coffee shops
When human labor is a luxury good
In 2012, the Wall Street Journal reported that “Starbucks baristas are being told to stop making multiple drinks at the same time” as a result of “customer complaints that the Seattle-based coffee chain has reduced the fine art of coffee making to a mechanized process with all the romance of an assembly line.”
A Starbucks barista in Minnesota griped that the new rules had "doubled the amount of time it takes to make drinks in some cases."
People don't go to Starbucks simply to get a cup of coffee — after all, there are lots of cheaper and faster ways to get a cup of joe. People go to Starbucks because they want a cup of coffee and the “romance” that comes from getting personal service from a human being.
This means that even if Starbucks invented a vending machine or robot that could make and sell coffee as well as a human barista, it wouldn’t make sense for Starbucks to lay off its human workers. Baristas aren’t just an expensive way for people to get the coffee they want; they’re essential to Starbucks’s strategy for distinguishing itself from lower-cost options like making coffee at home or the office or buying it from McDonald’s or Dunkin’ Donuts.
A lot of other service industries work the same way:
- Downscale restaurants often make customers order at a counter and bus their own trays. Fancy restaurants hire people to take orders, refill water glasses, take people’s coats, and so forth.
- Wealthy customers tend go to in-person fitness classes (or even hire a personal trainer), while more frugal customers buy exercise videos they can watch at home.
- ATMs have proliferated over the past 30 years, yet banks still keep human tellers on hand to answer customer questions and sell more profitable banking services.
- Amazon offers a fast, cheap, and convenient impersonal shopping experience, yet boutique shops and farmers markets have proliferated.
- TurboTax provides automated tax preparation services. But most people prefer a human tax preparer who can answer questions, provide advice, and spot opportunities for saving money the software might have missed.
- The startup Redfin tried to save real estate customers money by automating much of the home-buying process. But customers hated this early, barebones product because they actually wanted to talk to a human real estate agent. So over time Redfin has been forced to raise its prices, hire more agents, and become a lot more like a conventional realtor.
Automation treats human labor as a cost to be reduced or eliminated. But this attitude misunderstands the value of the human workers in these industries. The opportunity to interact with other human beings is a big selling point for fancy restaurants, farmers markets, and in-person fitness classes.
If we ever figure out how to automate aspects of education, health care, or other major labor-intensive industries, something similar is likely to happen.
If people develop online educational technology that works better than traditional lectures, tests, and so forth, that could be offered as a separate product for frugal students. But parents who can afford it are likely to prefer traditional universities. Traditional universities can always adopt educational technology for aspects of the college experience where it’s really better. But they can also offer personalized services — like faculty mentoring and face-to-face interactions with other students — that no online service can offer. So online learning will always be a cut-rate alternative to a four-year university, just as an exercise video or app is a downscale alternative to a fitness class or personal trainer.
Similarly, we may eventually invent software that can diagnose medical problems as well as a human doctor, and that would provide a low-cost option for people who can’t get time with a human doctor. But doctors do a lot more than diagnose diseases — they perform physical examinations and surgeries, answer patient questions, coordinate patient care, and so forth. So people who can afford it will prefer to talk to a human doctor — who may integrate the latest diagnostic software into their practice — just as most people prefer a tax accountant over TurboTax.
That’s what I mean when I say the productivity paradox works within industries as well as among them. As society gets wealthier and manufactured goods get more affordable, people spend a larger and larger share of their income on upscale, labor-intensive alternatives within any given industry.
Why service workers won’t get left behind
An important consequence of the productivity paradox is what it does to prices and wages. When a particular industry gets more productive, it tends to benefit both workers in the industry (who may get raises) and customers (who enjoy price cuts).
You might expect the converse to hold for low-productivity industries: that if productivity doesn’t improve, then wages and prices won’t change either. But that’s not how it works.
If you run a small-town restaurant that pays waiters $10 per hour and the factory in town announces it’s giving entry-level workers a raise from $11 to $13 per hour, you’re going to have to pay your waiters more too or risk having them quit for factory jobs. And because they won’t be getting any more work done than before, this likely means you’ll have to raise your prices.
This is a general economic principle: When some industries enjoy high productivity growth, industries with slower productivity growth tend to raise wages and, therefore, prices. A barber today can perform about as many haircuts as his predecessors 100 years ago, but barbers today make a lot more money than barbers did a century ago. As a consequence, haircuts cost a lot more, in inflation-adjusted terms, than they did a century ago.
In the economics world, this is known as Baumol’s cost disease. It’s named after William Baumol, the economist who first described the phenomenon in the 1960s. Baumol was trying to explain why performing arts institutions kept getting more expensive to run (he observed that playing a string quartet took exactly the same amount of labor as it had in the 19th century), but the principle he identified applies quite broadly:
This chart shows how prices in various industries have changed since 1978 — relative to the overall price level and adjusted for quality improvements. You can see that manufactured goods like cars, clothing, furniture, and toys have steadily gotten cheaper.
Meanwhile, medical care and college tuition have been afflicted with Baumol’s cost disease, as hospitals and schools have had to pay more to attract skilled workers to be doctors, nurses, professors, administrators, and so forth. The trend lines look similar for other service industries, including veterinary services and child care — costs have soared in recent decades.
Still, the name “Baumol’s cost disease” is unfortunate, because there are actually two sides to this coin. From the customer’s perspective, rising prices amount to a troubling “cost disease.” But if you work in a service job with low productivity growth, you’ll be happy about the phenomenon Baumol described: that workers in low-productivity-growth industries tend to get raises whenever their peers in high-productivity-growth industries do. We might call it “Baumol’s wage bonus.”
Baumol’s wage bonus is a big reason why we shouldn’t be alarmed by the prospect of more and more of our economy — and, therefore, our jobs — being focused on providing personal services. Many people believe that because services workers like teachers, nurses, barbers, and police officers tend not to become more productive over time, they will inevitably lag further and further behind manufacturing jobs in terms of pay.
But that’s wrong. Baumol’s work showed why it’s wrong in theory, and the data shows that it’s wrong in practice:
Until 2006, workers in the manufacturing sector did make a bit more than workers in the service sector. But in the past decade, the trend has actually reversed, with manufacturing wages lagging a bit behind service sector wages. But regardless of which sector is ahead at any particular point in time, the more important point is that manufacturing wages are unlikely to dramatically diverge from wages in the service sector. The reason is simple: If a persistent gap opened up, young workers entering the workforce would flood into the higher-paying sector until they were brought back into balance.
People often dismiss service sector work as burger flipping, but that’s a mistake for two reasons. The service sector isn’t limited to low-paying restaurant and retail jobs. Doctors, college professors, financial advisors, real estate agents, and the like are all in the personal service sector. The service sector offers work up and down the wage scale, just as the manufacturing sector does.
But the more important point is that factory jobs used to be awful. That changed slowly and painfully over the course of the late 19th and early 20th century as society developed institutions like labor unions that helped ensure ordinary workers were treated fairly.
There’s probably nothing we can do to stop the growth of service sector jobs. What we should be doing instead is taking this shift seriously and thinking more about how to make service sector jobs better. Depending on your politics, you might think that would mean stricter enforcement of labor laws, stronger union organizing, lowering of barriers to entrepreneurship, better worker training, etc. A serious debate about those alternatives would be a lot more productive than complaining about the decline of manufacturing — a long-term trend that can’t and shouldn’t be reversed.
The future will have plenty of jobs
At this point, it should also be clear why I think people are mistaken when they predict that automation will lead to a jobless future. Automation will certainly eliminate many jobs, just as it has done for the past 200 years. And some economists worry that the premature decline of manufacturing in developing countries will stunt their long-term growth. But a wealthy society has a basically unlimited demand for workers to provide personal services.
Most parents would like to send their young kids to day care options with fewer children per adult, and their older children to schools with smaller class sizes. People would like to provide their elderly parents with better elder care services with more human interaction.
People would like their doctors to have more time to talk to them. They’d like to go out to more fancy dinners and take vacations at fancier resorts. They’d like to have personal fitness instructors and life coaches. They’d like to go to more concerts, plays, and comedy clubs. They’d like to have people renovate their kitchens and bathrooms.
Demand for these services will always outstrip supply because each worker only has about 2,000 hours of work to offer to the market each year, and there’s a lot more than 2,000 hours of work each of us would like to have other people do for us. Most of us can’t afford all the human services we’d like to consume, so we buy a Roomba instead of hiring cleaners, buy frozen dinners instead of eating out, and so forth. But if automation made us richer, we’d spend more on these services and employ more people as a result.
In 1930, the economist John Maynard Keynes wrote a famous essay called “Economic Possibilities for our Grandchildren,” in which he speculated that the workweek could continue falling to 15 hours over the next century.
It doesn’t look like that’s going to happen, and our demand for personal services helps to explain why. Americans with above-average incomes could work a lot less and still support their families. A blogger named Mr. Money Mustache brags about how he retired at age 30 after living frugally as a software engineer during his 20s. Lots more people could do this if they really wanted to.
But most of us don’t want to. We’d rather work more and enjoy more luxuries. And while luxuries can take a variety of forms — with expensive housing being a big one in coastal cities — the most expensive ones are increasingly the ones that are the most labor-intensive.
So I think the future will look like the present — but even more so. A tiny minority of the population will produce the world’s clothing, smartphones, cars, household appliances, and other material goods. Almost everyone will work providing each other with personalized services — and these services will consume a growing share of our incomes.