No, technology isn’t going to destroy the middle class

October 21, 2013

Wages for ordinary workers in textile mills were stagnant for the first few decades of the Industrial Revolution. But as the technology matured, wages rose more quickly. (Photo by H.C. Williams)

Is technology killing the middle class? The George Mason University economist and well-known blogger Tyler Cowen thinks so. In his new book, "Average Is Over: Powering America Beyond the Age of the Great Stagnation," Cowen predicts a world in which 10 to 15 percent of the population are skilled at working with the smart machines of the future. He believes they will become extremely wealthy, while everyone else will face stagnant or falling wages.

To many people, such a gap between a meritocratic elite and most workers would be deeply unsettling. Some see extreme inequality undermining democracy; others think it would create economic crises and undermine economic growth.

But is the bleak world depicted by "Average Is Over" really around the corner? There are good reasons to be skeptical. People have been predicting that technology will kill the middle class since Karl Marx. They have generally been wrong. True, middle class wages do stagnate sometimes, as has been the case for the last couple of decades. But over the long run, technology has made large numbers of ordinary workers relatively wealthy. Thanks to technology, the average wage in the United States today is over 10 times what it was 200 years ago, after adjusting for changes in the cost of living. Given the poor track record of these past predictions, there are strong reasons to be skeptical about Cowen’s forecast.

Insights from games

The core of the book is Cowen's analysis of Freestyle chess, in which teams of humans and computers play against each other. Cowen looks in detail at how chess players actually work with machines, and this gives him rich insights into the possible qualitative nature of future work.

Computers have, of course, gotten very good at playing chess. In 1997, IBM’s Deep Blue computer famously beat Gary Kasparov, the best human player. Since then, chess programs have gotten progressively better. Nevertheless, computer chess programs are not perfect. Different programs have various strengths and weaknesses that human users understand and can exploit. It turns out that the most effective approach consists of computer programs, often multiple programs, teamed with human players. Humans can make the judgments about which programs to use at which times and to override the machines on occasion.

Cowen sees this teamwork as a model for the future workplace. As long as smart machines are not perfect and do not perform all tasks — Cowen is looking out over the next 20 or 30 years or so — then much valuable work will be performed by humans working with machines. Hence the valuable skills are those that work with the machines, enhancing them, coordinating them, and overriding them at times. This includes the engineering skills to design and program new machines.

But, importantly, those are not the only skills that matter. Cowen writes, “The ability to mix technical knowledge with solving real-world problems is the key, not sheer number-crunching or programming for its own sake.” This includes “understanding what kind of internet ads connect with their human viewers, or understanding what shape and color makes an iPhone attractive in a given market.” Marketing skills are important, as well as design. Some of the skills are learned through experience, for example, gaining an understanding of where the machines are weak.

For this reason, Cowen argues that “Our educational standards, and in the longer term our regulatory standards, will need to change.” A medical technician working with a smart machine to diagnose illnesses need not be a physician, but does need to understand the technology, how it works and where it is likely to fail. And those skills might be learned on the job more than in a classroom. Cowen cites commenters on his blog who argue that formal education is superfluous, unlikely to be of much benefit.

Where does the inequality come from?

These insights about the nature of work are rich, interesting and compelling. But then Cowen takes a leap. "I imagine a world where, say, 10 to 15 percent of the citizenry is extremely wealthy and has fantastically comfortable and stimulating lives, the equivalent of current-day millionaires, albeit with better health care," he writes. "Much of the rest of the country will have stagnant or maybe even falling wages in dollar terms, but a lot more opportunities for cheap fun and also cheap education."

But Cowen provides no justification for this view and no rationale for his 10 to 15 percent figure. Presumably only this elite has the skills that complement the smart machines. But why? New technologies have complemented some skills and substituted for other skills for at least 200 years without creating extreme inequality.

Indeed, history suggests that the rising income inequality of recent decades may prove to be a transitory phenomenon. The early years of the Industrial Revolution produced economic trends much like those we see today. A small elite of capital owners, mechanics and engineers — e.g. those who were first to capitalize on the opportunities created by steam engine, textile machines, and related mechanical technologies — saw their incomes rise rapidly, while the incomes of ordinary factory workers were stagnant. But as technology matured and skills became more standardized, the gains became more broadly shared.

We're likely to see the information revolution follow a similar course. So far, the gains have mostly flowed to the most talented and entrepreneurial workers. But as these technologies mature, we're likely to see increasing demand for moderately-skilled labor that complements the capabilities of computers.

To be sure, the new machines will be smarter, performing more tasks involving logic and inference than older technologies. But intelligent machines per se are not new. Some of the earliest machines used in the Industrial Revolution had “smart” features, for example, stopping a loom if a thread broke or changing the speed of a spinning machine as yarn was wound. Machines have become progressively smarter, and this changes the nature of the tasks needed to complement the machines. But there are plenty of examples of middle-income jobs today where humans work with relatively smart machines.

Take automated teller machines (ATMs), for example. These are smart machines that perform many tasks that were once the exclusive domain of human tellers. In the past, many people predicted that their introduction would dramatically reduce the number of tellers banks hired.

In reality, the opposite has occurred. The number of tellers has grown, and their wages have risen modestly in recent years. From 1999 to 2009, the number of tellers increased 123,000 to 577,000 despite the recession; their pay increased about 3 percent after adjusting for inflation. Tellers have survived because they have other important skills that enhance the value of the machine: tellers can perform complicated transactions that the ATMs cannot yet do and, more importantly, tellers form an important part of the “relationship banking” team that markets higher added-value services to bank customers. Customers are happy to withdraw cash from a machine, but if they want to pay off a loan, or to split a deposit between accounts, or if they're shopping for a mortgage or a retirement account, they want a human being who can explain the options and answer questions. Indeed, the latest ATM technology has gotten even smarter by adding a video link to a live human teller.

Just as with Freestyle chess, the combination of humans complementing machines can be particularly effective. The ATMs have reduced the number of tellers needed to operate a bank branch, reducing their cost. The banks, taking advantage of these lower costs to expand their markets and to compete more effectively, have dramatically increased the number of branch offices. And these new branches need the value-added services that tellers can perform. So while fewer tellers are needed per branch, the net effect has been to increase the total number of tellers. But these tellers now mainly perform higher value tasks.

Tellers are not the only example. Accounting software has taken over routine tasks, but the number of jobs for bookkeepers and accounting clerks has increased because they have other valuable skills; smart machines have replaced typists and switchboard operators, but the numbers of secretaries and receptionists has grown because those are the jobs with the skills that better complement the new office technology, handling more complex tasks and human interactions. From 1999 to 2009, the number of secretaries (excluding legal, medical and executive secretaries) grew by 216,000; the number of receptionists and information clerks grew by 64,000. These gains offset the loss of jobs for typists, word processors and switchboard operators (including answering service). The typing pool is dead, but the secretary who can make travel arrangements is very much in demand.

We don’t have to speculate about the future by looking at chess games. We can see emerging smart technologies in use in the workplace today. And what these examples show is that there are still plenty of mid-skill, mid-income jobs for people working with these new technologies. Based on current evidence, there is no reason to think that only 15 percent of future workers will be capable of complementing machines in ways that generate significant value.

Training a new generation of workers

Still, it is true that in recent years, not enough well-paying jobs have been created quickly enough to offset the loss of well-paying industrial-age jobs. But history suggests that this may prove temporary for a reason that Cowen has identified: the skills needed to work with new machines often cannot be taught in school, at least not initially. The workers who have mastered those skills are mostly self-taught, and only a minority of workers have the talents required to learn in this fashion.

But that doesn't mean the rest of the population can't learn these skills. Rather, it means that we need new institutions to train workers and then connect them with firms that need their labor. Until that happens, new machines will not profit many workers.

This is not new, either. During the Industrial Revolution, factory wages were stagnant for decades. Between 1830 and 1855, the pay of weavers in the cotton factories rose only 14 percent despite a doubling in their output per hour. But by the 1870s, weaving technology had become standardized, allowing workers to acquire skills and transfer them from one employer to another. A weaver with alternative job opportunities could demand higher wages. These newly professionalized workers began to earn better pay, allowing a blue-collar middle class to emerge. By 1900, weavers’ pay had more than doubled after accounting for changes in the cost of living.

Writing in Das Kapital in 1867, Karl Marx looked at the growing gap between poor factory workers and the wealthy owners of capital and concluded that the Industrial Revolution would progressively immiserate the working classes. In fact, the opposite was true; the Industrial Revolution was on the cusp of creating the modern middle class.

Today, Cowen is making a similar mistake. He sees a growing gap between the cognitive elite and everyone else, and concludes that these trends will accelerate in the next few decades. But history suggests a more optimistic possibility: as information technology matures, the economy will create a growing number of opportunities for moderately-skilled workers to develop skills that complement intelligent machines. Once that happens, today's extreme income disparities will begin to moderate, and the plight of America's middle class will stop looking so bleak.

James Bessen does research on technology and innovation at Boston University School of Law. He is the co-author of "Patent Failure: How Judges, Bureaucrats, and Lawyers Put Innovators at Risk." He is currently writing "Learning by Doing," a new book about technology and Jobs. You can follow him on Twitter.

Comments
Show Comments
Most Read Business
Next Story
Andrea Peterson · October 21, 2013