Your research looks at key labour market trends since 1980, which include a decline in mid-skill, mid-wage work and a 30 per cent increase in low-skilled, low-wage service jobs. Why is this happening?
In the last 20 years, computerization has become very effective at prescriptive, procedural and quantifiable tasks, and as a result, many jobs have been completely transformed. For example, the job description for a secretary today is utterly different from what it was 30 years ago. Back then, secretaries handled paper and typed things up and filed them away; now it’s all about organization and tracking down information and coordinating things, and these are very different tasks. This shift also applies to blue-collar positions, where most repetitive activities have been automated. If you walk into a modern auto plant, you will be amazed by how few people work there, because so much of the work is done by robots.
However, this trend has had a limited impact on knowledge workers—people who do creative problem-solving and ideation-type jobs, particularly managing and solving novel problems. In fact, arguably, these jobs are actually complemented by information technology, because it increases the amount of work people can accomplish. If you’re a lawyer reviewing cases online, or a doctor accessing medical data, you can accomplish a lot more on a typical day than your counterparts could 30 years ago. Technology not only gets rid of a lot of drudgery for these workers, it plays to their specific comparative advantage.
As for service work, the reason for its significant increase is that most of this work is not conducive to automation. A lot of the work in, say, construction or restaurants involves visual and motor flexibility. It also requires adaptability, in terms of answering questions, giving people directions or taking orders. These are not high-skilled activities by conventional standards: adults who are literate and have their physical faculties can do these jobs; but robots are horrible at these things.
As a result of these three shifts, what we see today is a ‘U-shaped’ labour market curve, with the middle-skill jobs at the bottom of the U, and the knowledge workers and service workers bringing up each side of it. This pattern of ‘employment polarization’ is not unique to the United States: similar polarization by skill level has been underway in numerous industrialized economies in the last 20 to 30 years, including Canada.
What are the implications of this ‘hollowing out’ of traditional middle-wage jobs for the middle class?
In terms of generational mobility—the probability of moving up or down the income distribution ladder across generations—many people, myself included, would have suspected that it would be decreasing; but the most recent data does not suggest this. It indicates that mobility has not actually changed very much in the last 20 years—at least in the U.S. One way they measure this is to look at the likelihood of young people from top-20th percentile income households and lowest 20th-percentile income households to attend college. Apparently, if you compare kids born in 1991 with those born 20 years earlier, there hasn’t been any significant change in this regard, which is quite surprising.
That does not mean that there is nothing to worry about with respect to the middle class, because we have seen a very substantial fall in the real earnings of non-college educated workers, and alongside that, falling employment rates. You don’t have to care about inequality, per se, to be concerned about large chunks of the population seeing downward economic progress over time. The fact that people are dropping out of the labour force says one of two things: either employers have no use for them, or they have no use for the jobs that are being offered at the wages they can command. It’s hard not to see either as a very bad indicator. Work is what structures adults’ lives: it gives us purpose, focus, a set of responsibilities and an identity. So when people are not participating in the labour market, all sorts of other things often start to go wrong.
To what degree are factors such as declining manufacturing employment and offshoring contributing to labour polarization?
My coauthor David Dorn and I looked across local labour markets and compared them on many different measures, and the one thing that stood out as being highly predictive of polarization at the lower level of the labour market was the share of employment in routine task-intensive jobs—those that are most susceptible to automation. I am not saying that the factors you mention are irrelevant. The decline in manufacturing has had a very important effect on employment among certain groups, and it has also contributed to falling wages for those groups; but it doesn’t appear to be a main driver of occupational polarization.
As for outsourcing, the measurement of it is so weak that we don’t really have a good sense of it. Much of the measurement has to do with ‘outsourcability’ rather than actual outsourcing, which is not ideal. I don’t want to dismiss these factors out of hand, but in our analysis they did not appear to be significant in terms of causality.
As indicated, there has been much faster growth in service occupations in the last decade, but wage growth in these occupations has not followed suit. Why is this?
[This article has been reprinted, with permission, from Rotman Management, the magazine of the University of Toronto's Rotman School of Management]