In the past three decades, the number of jobs for knowledge workers has never been rising as quickly as it is right now.
As recently as the mid-1980s, you could categorize American workers into roughly three equal-sized groups of about 30 million people each. About 31 million people had nonroutine cognitive jobs, what is often called “knowledge work,” consisting of varied intellectual tasks such as professional, managerial or technical occupations. Just under 30 million people had jobs that consisted primarily of routine manual work—on assembly lines or in warehouses, doing physical tasks day after day. About 30 million people had jobs consisting of routine office work—bookkeepers, filing clerks, bank tellers and so on—work that doesn’t involve much physical activity but is highly routine and doesn’t necessarily require high levels of knowledge. A fourth, smaller group, did nonroutine manual tasks, such as many service occupations.
But over the past three decades, almost all job growth has come from the two categories of work that are nonroutine. Meanwhile, routine jobs have been under a lot of pressure, especially during periods of recession.
Why is this happening? Economists believe it’s because jobs that are highly routine are the most susceptible to being replaced by automation and technology—Excel spreadsheets replace bookkeepers and sophisticated robots replace people on the assembly line. Especially during periods of recession, many workers in routine occupations lose their jobs and aren’t hired back as the economy recovers. (The definitions of these categories come from the Federal Reserve Bank of St. Louis.)
While routine jobs have gone nowhere over the past three decades, the number of people in knowledge work jobs has more than doubled, and there are no signs of that trend slowing. This strongly suggests that even though technology is eliminating some jobs, it’s creating even more in different fields.
In fact, knowledge work occupations have been adding more jobs than any other year since the 1980s—about 1.9 million per year. The other categories are growing too, but only by about 100,000 to 250,000 per year.
The phenomenon describing these trends is known as labor polarization, where growth is concentrated among high- and low-skilled jobs but the middle shrinks. The topic is heavily explored in the work of economists like David Autor of the Massachusetts Institute of Technology.
The exact classification scheme is far from perfect—even within occupations, or within the same employer, some people are doing more routine work and others are doing more variable work. It’s a reasonable bet that routine tasks are disappearing even faster than the data suggests because it’s not capturing the decline of routine jobs in knowledge-work occupations, nor the stability of the most complex jobs in otherwise routine occupations.
But quibbling with the precise definitions is missing the point. In previous research on the topic, Henry Siu at the University of British Columbia and Nir Jaimovich from Duke University have tried different ways of grouping workers and consistently found the same pattern: Different ways to separate routine and nonroutine jobs all show the same pattern of decline for routine work and growth for nonroutine work. For people considering if their job will be at risk from technology, the most important question is whether most of the job could be automated and not where it lands in this particular classification.
There is no doubt that machines are getting smarter, faster, more powerful and more dexterous—and potentially capable of doing more and more of the tasks that humans do. It’s easy to find warnings of the imminent risk of a jobless future. Most dramatically, a group of researchers at the University of Oxford warned three years ago that technology was on the cusp of destroying nearly 47% of U.S. jobs in coming years. It’s only been three years since that prediction, but so far new knowledge jobs are easily eclipsing the jobs that are disappearing.
Even as machines get smarter, many jobs have critical components that are social, emotional, creative or relational. These are overwhelmingly likely to be classified as non-routine types of jobs. The prospect of robots or automation replacing all of them remains remote. In other words, there’s good reason to think knowledge work will continue to grow.