This job is ‘getting old’

David Autor interviewed by Romesh Vaitilingam, 27 March 2009

Unfortunately the file could not be found.

Open in a pop-up window Open in a pop-up window


Download MP3 File (6.49MB)




View Transcript

<p><em>Romesh Vaitilingam interviews David Autor for Vox<br />
<br />
January 2009<br />
<br />
Transcription of an VoxEU audio interview []</em></p>
<p><strong>Romesh Vaitilingam</strong>: Welcome to Vox Talks, a series of audio interviews with leading economists from around the world. My name is Romesh Vaitilingam, and today's interview is with Professor David Autor, from MIT, the Massachusetts Institute of Technology. David and I met at the American Economic Association's annual meeting in San Francisco in January 2009, where we spoke about a paper he had just presented at the conference, which looks at changing trends in the labor market.</p>
<p><strong>Professor David Autor</strong>: This paper is called &quot;This Job is Getting Old: Using Changes in Age Structure to Measure Changing Job Opportunities.&quot; And the starting observation and the kind of motivating fact of the paper is that in the United States and, actually, across the EU, you see this kind of hollowing out of the distribution of occupations. So, essentially, there's growth in the tail of high wage, high skill occupations: attorneys, doctors, scientists, finance until recently. And then, on the left hand side of the distribution, there's growth of, basically, in person, service occupations, so janitors and cleaners and restaurant workers and childcare workers, personal care, security, firefighting, police.<br />
And what's disappearing is a kind of middle skilled swath of things that look like kind of clerical, repetitive production, accounting, a lot of kind of information processing tasks that are very susceptible to automation, and to some lesser extent, probably to outsourcing as well. But this was discovered sort of simultaneously in the US and the UK, and now has been documented throughout.<br />
So, for example, there's a paper being presented by Maarten Goos and Alan Manning and Anna Salomons that shows that 14 of 16 EU countries they look at, you also see this polarization of the occupation structure, where the growth is at the tails and the center is hollowing out.<br />
And the work that I've done with Frank Levy and Richard Murnane sort of identifies a likely cause of this as non neutral technological changes basically replacing these what we would call routine tasks, these things that can be readily kind of codified in terms of a set of procedures or steps or rules, and once they're codified in that way can be automated due to the declining power and increasing sophistication of computer software.<br />
And so you have a lot of sort of, basically, capital labor substitution, and a specific swath of tasks. And these are neither very high nor very low skill tasks, right? They're actually kind of moderately educated activities. Obviously, it's hard to automate the core activities of an attorney or a scientist or a doctor.<br />
Ironically, it's also very hard to automate the core activities of a janitor or a restaurant worker or a truck driver, just because those things require essentially environmental adaptability and a little bit of spoken language that has proven expensive, and tricky engineering problems, much more so than playing chess, for example, on a computer.<br />
So that's the kind of motivating observation of the paper. So then we say, well, let's see if we can learn about these kind of changing occupational opportunities by looking at age structure and across space.<br />
So the first thing we say is, well, it's probably the case that when occupations contract, they tend to get old. The reason being that if there's any sort of occupation specific skills, as an occupation contracts, the older workers will have an incentive not to leave, and the younger workers will have an incentive not to enter because they don't want to make investments in a kind of declining field.<br />
And sure enough, we show that that's very clearly true, that basically this U shaped growth of employment, growth on the tails, is mirrored by a kind of upside down U of age structure. So middle skill jobs are getting old, no question.<br />
So then we say, OK, well let's relate that, then, to their specialization in what we would call these routine activities. We're using a measure called the Dictionary of Occupational Titles. It has measure on job tasks. And we look at jobs that are specialized and basically procedural, set limits and tolerances and standards, specialize in those relative to these what we call non routine tasks.<br />
And sure enough, jobs that are specialized in routine activities are getting old, also. And it's not identical to the jobs that are getting smaller. There's independent variation in both. But they're clearly getting old.<br />
OK, well, now we've sort of proven the concept. Now, what can we do with this?<br />
So we'd say, well, let's look across local labor markets in the United States. And by local labor markets, what we mean are commuting zones, or 722 clusters of counties in which, essentially, commuting patterns are pretty strong within the clusters and pretty weak outside. And so we think of them as quasi independent labor markets. And let's line them up by their specialization, initially, in routine, intensive occupations.<br />
So we take the third of employment that's most intensive in routine activities, and we identify commuting zones according to this routine intensity measure. And what we would expect to be the case is, if they were specialized in those activities in 1980, over the subsequent 25 years, employment would have differentially hollowed out in those areas because those occupations would be in decline. And so we'd see a movement leftward and rightward to the sort of tails of the skill distribution of occupations.<br />
So the first thing we show is, sure enough, places that were initially specialized in these routine activities have a very substantial decline in those occupations. That decline is concentrated among non college workers in those occupations. So, basically, you can imagine the college workers doing the more sophisticated accounting and clerical work and stuff that requires more decision making and problem solving. And the less educated workers are doing the more rote activities, and those are the things that are being supplanted.<br />
So, interestingly, you see this movement out, but the movement out is concentrated among the young, as you would expect. So those occupations differentially get old in those areas. And then the young are moving out.<br />
We said, OK, well, now where do people end up? So we say, OK, let's take the other two thirds of occupations that don't fall into these categories and let's just divide them in half according to the mean wage in the occupation.<br />
So we take the half of employment in the higher paying, and we call those the high skill, non routine activities, and we take the other half, that's the low skill, non routine activities. And those are quite diverse. So the high skill, non routine would include attorneys, but it would also include skilled construction workers. It would also include some production and designer and so on. And then the low skill, non routine things are cleaners and janitors or truck drivers.<br />
We say, OK, let's look at this differential hollowing out and say, people are moving towards the tail, who's moving where, breaking up by education and by age. So the first thing you find is in these areas where we're initially concentrated in these routine occupations, the predominant movement of employment is towards the left, in other words, towards the lower skill non routine work. So the middle, unfortunately, it&rsquo;s not being equally redistributed.<br />
Second of all, the leftward movement is much more pronounced among non college than college workers. Moreover, the leftward movement is more pronounced among the old than the young. In particular, what you see is among non college workers in all age categories, they&rsquo;re all moving into these basically lower skill non routine occupations. Among college workers, the young college workers are moving rightward, moving into the higher skilled non routine occupations. Interestingly, increasingly you find older college workers in these activities that would not be considered historically college jobs. As it holds out, it seems like only the young and well educated are shifting rightward.<br />
Now, I want to make one clarification. I say the young and the well educated, I don't mean to imply that these are individuals moving. Actually, we're looking at cost and age brackets with changes over 10 and 25 years. So increasingly if in 1980 you find older college grads in the middle of the distribution, 25 years later you find older college grads in the left hand tail of this skills distribution or occupation distribution. Those are not the same people. It's just that the jobs they would have occupied when they got to that age are no longer there, and the question is, what jobs do they occupy instead?<br />
So it presents a striking and potentially somewhat concerning picture. If there&rsquo;s really this decline in the middle activities, things where high school graduates and some college people did a lot of their work, these are basically moderately paid, moderately well educated jobs, and they're going away, you wonder if there's going to be a potential sort of missing rung in the occupational ladder. You are either doing in person service work, which isn't highly skilled but it is hard to automate or offshore, or you're doing very highly skilled abstract work that pays well, it is a great job, but of course it requires a lot of education, it is not accessible to everyone, the skills are a bit rarefied.<br />
So, I think we are only beginning to get a sense of the shape of the changes and what they also imply for the wages as well.</p>
<p><strong>Romesh</strong>: I was about to say, what are the implications of this polarization for inequality?</p>
<p><strong>David</strong>: In the United States in particular we do see polarized wage growth. Essentially we have sharp wage growth above the 70th percentile distribution, and then we have relative wage growth at the 30th percentile and below, and the middle is actually comparatively flat. Now, that is not as true across the rest of Europe. In the US, basically the shape of employment and the shape of wage growth look like the same picture. That has not been confirmed for continental Europe, though a little bit more for the UK. So I don't want to act with uncertainty, but the next step in this project and I should mention that my coauthor on that project is David Dorn, a graduate student from the University of St. Gallen who is on the job market this year -- the next step in that project will be to look at if these movements to the left and to the right for these age and education groups is paralleled by movements either down or upward in the expected wages. Do we see in some sense as the center bifurcates, that people end up earning less than we would have predicted based on their age and education 20 years ago, or more if they are moving rightward?<br />
So I think we will better understand. But my suspicion is that those occupation shifts are not neutral for earnings. There's a change in job quality, stability and benefits, and probably in monetary compensation as well. So I suspect that it does not augur improved opportunities for those groups, in fact just the reverse.</p>
<p><strong>Romesh</strong>: This seems like a rather different story from the standard one of technical change creating more demand for high skills. Presumably it has some implications for the education policies you might want to take from that.</p>
<p><strong>David</strong>: I have been arguing against the simple skill bias technical change story for some years, so this really goes back to a paper that I wrote with Frank Levy and Richard Murnane in 2003 published in the Quarterly Journal of Economics called &quot;The Skill Content of Recent Technological Change&quot;. We basically posed this point that we don't think that technological change, or even the current era of technological change, is always and everywhere skill bias. Basically there's a set of tasks in which current technology has a comparative advantage, and these are things that are procedural and rules based, and therefore can be codified in software and automated. But there's only a certain set of tasks that we know how to do that for. We don't know the rules for being a lawyer or a doctor, nor do we know the rules for being a janitor, because the locomotion and object manipulation actually is at present very difficult. So the set of activities that are most subject to this replacement are these middle scale routine tasks. We think there is strong complementarity between those routine tasks and the activities of professionals - clerical and accounting, these are support occupations to the professions. And so plausibly declines in the cost of performing those support occupations would raise the productivity of people doing the other ones.<br />
But these non routine manual activities at the lower end are somewhat orthogonal. They are not directly substituted or complemented. They just hang around. But what that implies for the wages in those occupations partly depends on the complementarity in consumption between the outputs of these goods in tangible occupations and the outputs of these very tangible services. So I think that the simple, monotone skill bias technical change, we have been saying for some years that that's probably not the most constructive framework for thinking about this. I hope this one is a little better.<br />
Implications for education policy are very tricky. It presents a little bit of a picture of a Latin American country where you have an elite and then a service sector that services them. That's not a very attractive model in terms of mobility. At present I would be very loath to change. The correct answer for the last hundred years has always been always and everywhere more education is better than less education. On the basis of a couple of years of research, I wouldn't change that. I guess the question is, what are the set of skills that one can impart that are very broadly applicable. And the last thing you want to do is limit people's opportunity to move up. If the bottom is larger and larger, you want to give people the tools that are going to allow them to potentially escape, or at least rise as high as their capabilities permit.</p>
<p><strong>Romesh</strong>: David Autor , thank you very much.</p>

Topics:  Labour markets

Tags:  occupational distribution

Ford Professor, MIT Department of Economics


CEPR Policy Research