Ok Computer
Are the robots taking all the jobs? Are they going to? Are these even the right questions to ask?
If you are a subscriber to this newsletter, you may know that back in 2016 I published a book, The Wealth of Humans, which was in part a consideration of how technology is affecting the global economy. It presented as a book about robots taking jobs: a fact which got me labelled, among those who were aware of and perhaps even read the book, as a robots-are-taking-the-jobs sort of person.
A robots-are-taking-the-jobs sort of person? In the circles where people argue about these things there are basically two kinds of people: those who think the robots are taking the jobs—a group which includes Andrew Yang, who is currently engaged in a long-shot bid for the Democratic presidential nomination and who recently wrote an op-ed in the New York Times about how the robots are taking the jobs—and those who are fairly certain that the robots are not taking the jobs (and who sort of suspect that the robots will never take the jobs). In TWOH I tried to do something unusual, which is to be neither of those people. Readers, it didn’t work.
Let’s back up.
Around the beginning of this decade, there was a surge in interest in new technologies, particularly those associated with artificial intelligence. Researchers were making surprising progress on a number of challenging AI problems: like autonomous driving, voice recognition, machine vision and translation. These developments prompted excitement, but also concern: it was very easy to see how an AI that could converse with people or drive a truck could displace vast numbers of workers. Books began to appear about what was going on; for a lot of people, the real attention-getting moment was the publication of Race Against the Machine, by Andrew McAfee and Erik Brynjolfsson, which triggered an avalanche of writing about the possibilities and pitfalls presented by AI, which continues today.
I did a lot of that writing, much of it in The Economist (see this, and this). As I did, the flow of debates about technology became dismally familiar. Someone would write something expressing a concern that new technologies could have negative labor-market effects. Immediately, folks in the economics and journalism communities would trot out the same sets of responses:
People have been worried about technological unemployment for centuries and have always been wrong
If this time is different, how do you explain the absence of mass unemployment and high productivity growth
All this talk about automation is a distraction from the real labor-market problems, which are [insert favorite labor-market problems]
But as I spent time with the subject I realized that these standard responses were in many respects misleading. For instance, it has very much not been the case that technological progress always works out splendidly for workers, despite the naysayers. In the book I talk about the Luddites, who we basically write off as colossal idiots for thinking that technology could be bad for them. But the Luddites were right! Industrialization destroyed the livelihoods of many semi-skilled workers whose training was made obsolete by new machines. Those machines did create work for lots of other, unskilled workers, but whole generations of them derived little if any benefit from industrialization. For the first century of the industrial revolution real living standards grew slowly and fitfully when they grew at all.
Things did get better. But it was not the case that technology just kind of sorted everything out on its own. It took a lot of doing to transform industrial economies into places in which machines were not principally a means of exploiting unfortunate labor. Workers had to win the right to organize, which was a long and violent struggle. They had to gain political rights. Government had to insert itself into the economy in a massive and unprecedented way: by providing public education, regulation, and a social safety net. It is true that technological progress created the potential for humanity to enjoy sustained, long-run growth in income per person. But it would be wrong to conclude from economic history that technological progress is rarely or never bad for workers. It would be more accurate to say that technology is often very bad for workers in the absence of an aggressive social and governmental response to the problems that new technologies create. It’s at least a little funny, then, that the people warning about the labor-market consequences of new technology and advocating for action to mitigate them are the ones being told that they’re dead wrong.
What about the second point, then: that it is ludicrous to speak of worker-displacing technologies at a time when both unemployment and productivity are low? The logic behind these arguments seems sound enough. Productivity, economically speaking, is a measure of how much output you produce with a given set of inputs. If you’re producing more stuff with the same number of workers, or the same amount of stuff with fewer workers, then mechanically productivity ought to be going up. If you’re producing the same amount of stuff with fewer workers, then you’ve also let some workers go. And if the way automation were unfolding was that over a relatively short period of time, firms rolled out powerful new technologies which displaced lots of workers, then, other things equal, we would see both soaring productivity and a big jump in joblessness.
In practice things are not so simple. As an example, let’s consider one of the possibilities often warned about in discussions about automation: the development of autonomous trucks. Suppose that right now firms could go and buy a self-driving truck that could handle essentially any point-to-point hauling route. Assume that all potential regulatory headaches have already been ironed out. Would that mean the rapid end of work for human truckers?
No. New trucks are expensive, especially state-of-the-art ones outfitted with lots of new technology. In how many scenarios will it be immediately profitable to replace functioning capital equipment with a more-expensive replacement simply in order to reduce labor costs? There might be other financial incentives involved; insurers might push for investments in the new trucks if they are thought to be much safer than human-driven ones. On the other hand, there are risks to deploying any new technology, and a lot to be said for waiting a bit until kinks have been worked out and costs have come down. Delay would also allow firms to think hard about their business models and human resources, and how those might need to change, perhaps dramatically, in order for autonomous trucks to make financial sense.
There would be exceptions. We might expect driverless trucks to be adopted first by companies with extremely deep pockets, a very long-term outlook, an innovative mindset, a less-than-congenial relationship with workers (particularly those without white collars), and whose businesses are extremely reliant on shipping. Perhaps unsurprisingly, Amazon is among the firms most keenly interested in autonomous shipping technologies, from delivery drones to driverless trucks. Note that it is investing in these technologies even as it spends vast sums of money to increase its human-controlled delivery options: which is to say that it’s creating a lot of insecure work in trucking even as it prepares to make quite a lot of that work redundant.
The fact that roll-out of driverless trucks wouldn’t be immediate does *not* mean that workers would be unaffected by the arrival of this new technology. In the short run, the availability of driverless trucks would effectively shift bargaining power away from human drivers. Their ability to demand higher pay, better benefits and improved working conditions would immediately shrink. Potentially, the quality of the truck-driving workforce would decline (and perhaps productivity with it) as drivers with the best alternative labor-market options observe the writing on the wall, note the crummification of their working environment, and find jobs elsewhere. Eventually the story of driverless trucking would be one in which humans were largely displaced and productivity soared: because the same amount of or more trucking than before would be handled by many fewer workers. But you could have quite a long period in which conditions deteriorated for workers even as those workers remained very much on the job, all because of the emergence of a labor-displacing technology.
If you’re an economist, you think about the arrival of a new labor-displacing technology and you generally assume that it will lead to substitution of capital for labor with certain predictable effects. But you could instead describe the emergence of these technologies as an expansion in potential labor supply: a reservoir of quasi-labor which can be pulled off the shelf when market conditions are favorable. Maybe that translates quickly and cleanly into substitution of capital for labor, but maybe not. Maybe it does in some cases but not in others. Maybe that reservoir mostly depresses worker bargaining power and potential wage growth. Maybe, by depressing wage growth, it actually leads to the creation of more low productivity jobs. Depending how how substitutable one low-wage worker is for another, displacement in one part of the economy could lead to an abundance of available low-wage labor in another, which reduces whatever pressure firms were feeling to seek out and deploy labor-saving technology.
Or maybe firms don’t feel much pressure to automate for other, unrelated reasons: like the fact that they enjoy a lot of bargaining power in labor markets and not a lot of competitive pressure in their product markets. Maybe there are chronic demand shortages in the economy which reduce the pressure to expand production and contribute to a constant background level of labor-market slack. (Both sets of circumstances could in fact be linked to changing technology, as I explained in TWOH, but that’s a matter for another newsletter.) In those cases, automation might be a very real threat worth preparing for, even though it is not currently wreaking much labor-market havoc.
If there weren’t any indication that technology were improving in a way that might hurt workers, some of them at least, then perhaps we’d look at the data and say yeah, automation clearly isn’t a threat. But the data are at least somewhat consistent with the stories above, and we do have plenty of examples of new technologies which can now, or conceivably could in the not-too-distant future, displace substantial numbers of workers. It seems to me that the people who worry about automation are deserving of more than a haughty dismissal.
That is especially the case given economists’ recent record on any number of labor-market phenomena. It was not long ago that economists almost universally dismissed the possibility that trade could do substantial, long-run harm to workers. It wasn’t long ago that economists almost universally shrugged off the problem of regional economic decline. It wasn’t long ago that economists were pretty sure that inequality first rose and then fell as an economy developed (as opposed to rising, then falling, then rising again). It seems to me that those who are sceptical that job-stealing machines are a problem should come at the topic in a spirit of, “I have my doubts, but economies are complicated and I’ll keep an open mind,” rather than, “This isn’t happening. Look, my time series proves it”. Does it really? Are you sure?
***
The knee-jerk, extremely confident responses to automation discussions bother me. But you know what, I get it. I get it now in a way I didn’t get it three years ago when I was trying to get people to read my book. I get the logic behind the third bullet point above (“this is a distraction from the real problem”).
Being an economics writer or a wonk generally isn’t working in the salt mines, but it’s harder than you probably think. To do your job well, you have to take expert views and research seriously, but also understand that a lot of it is not very good. You’re operating in the modern media environment, which is an absolute hellscape. Vast numbers of PR people, unorthodox thinkers and charlatans fill up your inbox and twitter mentions with crazy ideas of all types. Maintaining your sanity means tuning a lot of that out. You simply don’t have the time to openly engage with many conventional-wisdom-challenging points of view. Maybe you pick a few that seem particularly interesting and plausible, and so long as peers you respect hold the line on others you keep your stock reply at the ready (nope, robots aren’t taking the jobs) because in most cases the CW-smashing view is wrong.
Moreover, you have a limited amount of time and space to devote to the writing for which you are responsible, and very little control over what happens to your ideas once they’re launched into the aether. That means, first, that you communicate something about the importance of a topic by choosing to write about it and, second, that you’re aware that engaging with dubious ideas can serve to amplify them in the public discourse, which you may not wish to do. So if you believe that weak demand is by far the most important near-term labor-market concern, and certainly the area for which relatively small tweaks to policy could make the most difference, then it makes a huge amount of sense to focus on that almost exclusively in your writing, and to dismiss pretty quickly and haughtily any efforts to redirect attention to other things.
As an author, you have to know your audience! But look, I don’t choose the areas where I have what seem like important and novel insights. Back in 2015 I was thinking the thoughts I was thinking, I was feeling like there was a good book in those thoughts (still do), and I was determined to write it.
I wanted to challenge the way people thought about technology and the economy. I wanted to get people’s attention. It seemed to me that framing the book as a contribution to the discussion about robots and jobs would be a good way to get publishers and then readers interested in it. But for lots of reasons the readers I cared most about were the writers and thinkers participating in big labor-market debates. And what I didn’t appreciate was that presenting the book the way I presented it would, for many people, trigger the intellectual response described above: sorry, don’t have time for it, not the most important thing right now, probably wrong.
In hindsight, I think I might have had better luck inducing engagement with the book had I pitched it differently. You’ll have to pardon me if this all feels a little self-indulgent. TWOH was the first “real” book I’d ever written (I’m proud of The Gated City, but it was only a Kindle Single). You pour all that cognitive and emotional energy, all those hours, into an endeavour like that and you can’t help but obsess a bit over how it’s received (or I can’t, anyway).
If I could do it over again, I would frame the book as being about the problem of abundance. That really is what the book is about, after all. For countless millennia, humankind struggled to deal with scarcity. Technological progress has slowly but surely transformed the economic environment we face from one of scarcity to one of abundance. But economic and social systems which have evolved to manage scarcity begin to misfire when confronted by abundance. That is particularly the case when the emergence of abundance is halting and uneven: such that some things remain scarce while others—especially labor—become increasingly abundant.
Labor is abundant, and is likely to become ever more so. The book explains why: because ever more of the world’s population participates in the global economy, and because technology becomes ever better at adding to effective labor supply. And as technology improves, and machines become ever better at imitating human capabilities, potential effective labor supply will approach infinity. That’s not where we are now, to be clear, but that is unquestionably the direction of travel. So what does that mean? Well, it means that the social and economic structures we count on to deliver reasonably efficient and fair outcomes in a context of labor scarcity start to spit out lousy outcomes as scarcity turns into abundance. And it means that over the long run, “how should people spend their time” will cease to be a question about the efficient allocation of scarce resources, and will instead become a judgment based on community values and the need to maintain social stability.
That’s the book! “Abundance is a good thing, but as it turns out it’s hard to manage.” The message is there: the title is, after all, the wealth of humans, which is wordplay, folks, gesturing among other things at the idea that labor is increasingly abundant. But it could have been foregrounded more. It would have been clarifying to have done it that way. It prompts you to ask, for instance: what is scarce? (Valuable social capital, the book answers.) It helps you to think about responses. (One of the off things about the book is the reluctance with which I turn to labor organisation as a way of coping with abundance; that’s a bit of intellectual baggage I have since thrown overboard.)
And maybe it would have lured in a few more readers, a few more writers. Or maybe not. But on the nights when reflections on the book are the thing keeping me awake as I lie in bed, this is what I think about.
***
This week’s column considers a running debate about economic development (and economics more generally) which has intensified in the weeks since the Nobel prize for economics was awarded.
As a post-script, I am aware that my newsletters have run a touch long. Is that annoying? Let me know and maybe I’ll try to adjust my habits, I don’t know, we’ll see.