Can artificial intelligence reverse the productivity slowdown? My long-read Q&A with Nicholas Crafts

Why has productivity slowed in the
past half-century? How can we kick-start innovation in the economy? Could we be
on the cusp of a fourth industrial revolution powered by artificial
intelligence? Nicholas Crafts joins Political Economy to discuss these and
other questions about the productivity slowdown.

Nicholas is a professor of
economics and economic history at the University of Warwick, where he studies
economic growth, the Industrial Revolution, and the history of general purpose
technologies. He recently gave a fascinating presentation to the Bank of
England, titled “AI
as a GPT: An Historical Perspective

What follows is a lightly edited transcript of our conversation, including brief portions that were cut from the original podcast. You can download the episode here, and don’t forget to subscribe to my podcast on iTunes or Stitcher. Tell your friends, leave a review.

Pethokoukis: To start out, why did
we see this productivity slowdown in the early ‘70s? And why has it continued
for so long — except for that brief surge in the late ‘90s and early 2000s?

Crafts: Well, when the productivity slowdown started in
the ’70s, I think the story would be that we’d have the so-called Second
Industrial Revolution in the late 19th or early 20th century that delivered us
a lot of very important technological change. Eventually those technological
changes — which included electricity, the internal combustion engine, aviation,
all those sorts of things — gradually started to work through that potential
and we needed new things to replace them, to drive productivity growth forward
through technological change. I think there was a hiatus, if you will, until
information and communications technology really gave productivity that boost
you mentioned towards the end of the 20th century.

Previously, we’ve had Robert Gordon on here, who’s written about how America’s economic growth was driven by these “great inventions” from the 1870s to the early 1970s, and then it just kind of petered out. So… is that the whole story? Are big innovations the biggest driver of growth, and they just haven’t been driving it much lately?

I don’t think that is quite the whole story, no. My own
take on economic history is that great inventions — a number of which I just
mentioned — don’t completely dominate. What’s important is that you have an
economy which can achieve lots of small productivity improvements as well by
becoming more efficient and eliminating businesses that aren’t really pulling
their weight.

I think if we look in the later 20th century, and
particularly the early 21st century, there are probably some reasons to think
that the US economy lost some of its earlier dynamism. Exactly why that might
be is controversial. But the two obvious candidates are that, particularly in
the last 20 years or so, competition has weakened considerably — and
competition is an important driver of productivity growth — and perhaps
regulation has become somewhat more onerous.

via Twenty20

Talking about that period of fast growth, (and most people probably would be surprised to hear this) the 1920s and the 1930s are considered periods of exceptionally fast productivity growth, even though we had a depression. There’s the work by Alexander Fields on this issue. Is that kind of a settled economic science, why those decades are very fast growth — and were they as growthy as what we think they were?

The story about the 1920s and 1930s, I think, is it was a
period of fast growth. I think the American economy actually did a little bit
better after World War II, but it’s within the boundaries, perhaps, of
measurement error. Yes, I think it is now settled that (particularly surprising
to the layperson, I suspect) the 1930s was a period of very strong productivity
growth. I think some of that is the American economy getting rid of low-productivity
producers during the depression’s cleansing experience. You can see it in
things like automobiles.

But it’s also a period when research and development was
pretty strong and pretty productive. This is the aspect of the Second
Industrial Revolution which goes along with a more scientific approach to
technological progress, the heyday of the R&D lab, that kind of thing.

Since growth dropped off so soon
after the 1950s and ‘60s — which are considered sort of this golden age of the
US economy — I wonder if the ’50s and ’60s really weren’t that productive, and
they were sort of living off the gains of earlier decades. Back then, we had
very high tax rates, and there wasn’t a lot of foreign competition spurring
efficiency. So is the golden age of the ’50s and ’60s a little overrated?
That’s my theory.

I think in terms of performance at the time, it was very
good and it probably was the halcyon period. I think you may be right to say
that it wasn’t perhaps storing up as much in the way of future technological
progress as one might have hoped. Yes, you can perhaps say the glass was half
full or half empty, depending on which perspective you take.

Incidentally, I think you’d be fair to say that we don’t
entirely understand the slowdown of the ’70s and ’80s. There clearly are some
unpleasant shocks, things like oil prices, Japanese competition starts to grow,
as I think you mentioned. I think we possibly entered the era where measuring
GDP starts to get a bit harder. Your older listeners will surely remember the
so-called Boskin Commission of the mid-1990s, which pointed out that inflation
was probably not being measured very well, and therefore real productivity
growth was subject to bias as well.

We’ve discussed this issue from time to time on the
podcast, and one thing that’s also come up when looking at that slowdown was
that starting in the ’70s, maybe late ’60s, we started focusing less on growth
and maybe started valuing some other things like the state of the environment.
What do you make of any of the research blaming environmental regulations or
other kinds of regulations for slowing down growth and maybe putting weights on
the legs of the economy?

Yeah, I think there is some evidence that regulation was a
little bit of a drag. I think it’d be difficult to say that it really accounts
fully for the slowdown. Environmental regulations are part of it, but there are
all sorts of things which just interfere a little bit with the productivity
generation of the economy. One that got some attention in the recent past is
the strong growth over time of occupational licensing, for example, which
eventually starts to restrict the mobility of labor.

Another theory: Did we start
taking fast growth for granted after the 60s? Were we so optimistic about
growth that we stopped making it a key focus of policy?

I think that would probably be a fair comment about, say,
the later 1960s and on through the 1970s. It took people a while to realize
that the slowdown in growth was permanent and not just temporary shocks hitting
the economy. But both in this country and in the United States, it’s hard to
see the 1980s as a part if that complacent era. On the contrary, I think we were
seeing, at that point, Reagan and Thatcher pushing back against the
encroachment of the state. And they were doing so in the context of feeling,
“Ah, things are not going so well.”

Former British Prime Minister Margaret Thatcher (R) shakes hands with former U.S. President Ronald Reagan as she participates in a birthday salute celebrating Reagan’s 83rd birthday, Via REUTERS

Yet the slow productivity growth has persisted, except for
in the late ’90s and early 2000s. Now, a partial explanation for the
continuation is that we’re in between invention waves.

I think so, yeah. I mean, I think one could be optimistic
that things like artificial intelligence and robotics will improve their
contribution to productivity growth in the near future. I’m not myself
persuaded that all the great inventions are in the past or that R&D will
increasingly fail to deliver productivity improvements, but I think we have
passed the peak of what in Europe we would call the ICT boom. In a sense,
there’s an element that part of what’s going on is a lull between its major
effect and the major effect, perhaps, of the next general purpose technology,
which could easily be 10 or 15 years away.

I’m glad you brought up AI —
that’s the technology so many techno-optimists think will bring back fast
productivity growth. What do you see as the potential value of AI? Is it the
next great general purpose technology, like steam or electrification? Or, as
you’ve been writing, is it a general purpose invention that will help us find
the next great ideas? I suppose… maybe there’s value in both.

I think there is value in both. I think what I was trying
to do in my recent presentation was not to decry the idea that machine
learning, the use of AI in production, will be valuable. I think it will. I
think we will probably see increasing applications over time. That’s the normal
history of a general purpose technology.

But I do think this one should perhaps to be thought of
also as the invention of a method of invention. I think it was Joel Mokyr who
said AI might end up being the best research assistant ever. We see harbingers
of this. The MIT guys discovering Halicin, the new antibiotic, using this kind
of machine-learning methodology is perhaps a straw in the wind.

Going back to Gordon’s great
invention theory, there’s a related idea that all of the low-hanging fruit has
already been picked, so it’s harder to find good ideas than before — you need
more money and manpower in order to innovate. Could AI help us go up that
ladder and get more fruit? Or could it help us discover there’s a lot more
low-hanging fruit than we thought?

It could be both, but certainly there are, or should I say
have been, signs that the potency of research and development has been weaker
in the recent past than it was in that halcyon period. It seems to take more
inputs to create inventions, people make their great discoveries later in life,
so perhaps there’s a great deal more homework that needs to be done. That’s one
way of thinking about it.

But at the same time, there’s an awful lot of knowledge
out there. A lot of technological progress comes from hybrid inventions —
putting together two things and making some new third. Searching that stock of
knowledge is a very fruitful way of making advances. I think the argument might
be that there are more trees in the orchard than there used to be, but there are
so many trees in the orchard that you and I can’t scan them effectively. We
need some help. The help comes, I think, potentially with things like AI. That
would be the big potential change, the big potential advantage. That really
goes to the way, I think, quite a lot of economic historians think about the
idea of an industrial revolution: It’s that the nature of invention changes.
And we’ve seen that several times in the past.

Marjon Blondeel, AI R&D developer wears VR goggles at the AI Xperience Center at the VUB (Vrije Universiteit Brussel) in Brussels, Belgium February 19, 2020. REUTERS/Yves Herman

Well, I think that’s super interesting because in a recent
paper you wrote that one way to think about the Industrial Revolution — whether
it’s the first, the second, or the third — is by looking at the inventions and
how that affected those economies. It’s also a story, again, about how people
went about inventing things, even going back to the initial Industrial

I think that’s right. The Industrial Revolution — the First
Industrial Revolution in England — I think involves essentially the culmination
of an Enlightenment approach to trying to find out more about the world. It’s
the beginnings of what we now think of as scientific method — a more
systematic, empirical approach to finding knowledge about the world. That is
then built upon, and in a much more scientific, technologically trained world —
that period at the end of the 19th century and beginning of the 20th century — when
we move through to that so-called Second Industrial Revolution. Thomas Edison
had an industrial research lab. James Watt didn’t.

We’ve had Carl Benedikt Frey on the show, and he argues that one reason we saw these great advances in the 19th century and beyond is because government allowed them to happen by actually siding with the disruptive innovators, which was rare at the time. My concern today is that I’m not sure government will continue to be on the side of creative destruction, given all of the concerns we hear about robots taking away jobs. Should we be worried?

I think that is a worry, and I think it’s a worry partly
in the context of what I referred to earlier: We are likely to become an
increasingly regulated society. Environmental concerns have grown, not reduced.
But also I suspect the shakeout from the current virus problem is probably
going to be, in the UK at least, a society which starts to care more about
distributional issues than it has done and a bit less about what we might call
efficiency or productivity.

There is sometimes a perceived conflict, and I think it is
important the way politicians play that. But I would add and go back to
something I said a few minutes ago: I think a really key feature of allowing
disruption to happen is competition. If you weaken anti-trust, you weaken
competition policy, you undermine those processes. I think there’s quite a bit
of evidence that this really has been the story of the United States in the
last 25 years or so.

It seems like a lot of talk about that seems to focus on
the big tech companies. As you’ve seen the research, do you think it’s mostly
those companies? I mean, a lot of this economy is housing and a lot of it is
healthcare. So I wonder about the competition more broadly in the economy,
rather than just focusing on Amazon or Google.

I really do not think this is just Amazon or Google. I
think we’re talking about this more general weakness in the dynamism of the
American economy. We’re talking about too little new entry, too few startups.
We’re talking about situations where the market seems to respond only slowly to
opportunity. Barriers to entry seem to, in one way or another, have grown.
Obviously there are big concerns about Big Tech, but I think it’s more than
just that.

I’m also concerned that, because
we’re funneling a lot of fiscal stimulus right now, we’re creating zombie
companies and losing that “cleansing aspect” of the 1930s. Without an invisible
foot kicking the failing companies out, we won’t get the new ones coming in.

Yeah, I think that’s absolutely right. I was focusing a
moment ago on problems of entry. You’re talking, I think rightly, about
problems of exit. That is partly, at least in the UK and I suspect in the US, a
little bit of a side effect of the interest rate policies. We’ve had zombie
companies, which under normal interest rates wouldn’t survive.

In a more European context, traditionally I think we’ve
had problems in allowing exit because we’ve tended to subsidize losers. We’ve
tended to employ what we call industrial policies, which slow down exit. I’m
sure there’s going to be a lot of that in the UK, in the context of people
losing their jobs as a result of coronavirus and government sort of trying to
stop the bleeding. It would be bad if the US got into that in the same way as
Europe has done in the past.

I think there’s a temptation to
look at China, think that their bureaucrats and planners are smart enough to
manage the economy and finance the right sectors, and wonder why we don’t do
that here. You know, maybe we also need to be subsidizing 5G or advanced
batteries, because China’s found a better way of doing things.

There can sometimes be cases for subsidizing new
industries, but I think one should, first of all, note that China has wasted a
huge amount of its investment over the last 20 or 25 years. If you’re investing
40 percent of GDP, you ought to be in some ways doing better than they have.

People wearing face masks, following the coronavirus disease (COVID-19) outbreak, walk past the CCTV headquarters in Beijing’s central business district (CBD), China, July 16, 2020. REUTERS/Tingshu Wang

The thing that I would go back to, seeing it from a
British perspective, is that we used to think that governments could pick
winners. We found out that losers could pick governments. What that means, of
course, is that losers come along and lobby. They are the interest groups,
which are more successful, generally speaking. Once you go down the path of a
lot of intervention, the danger is that it ends up slowing exit, rather than
succeeding and promoting dynamic productivity, improving sectors. If you can
pull off the trick of promoting the positive part of the economy, very good,
but the politics of it is pretty difficult.

To finish up, based on what we’ve just talked about, can
we grow an economy faster than what we’ve been doing? Do you think with the
right policy — whether it’s interventions such as more scientific investment,
or removing barriers, or maybe regulatory reform — can we speed up
technological progress such that it also speeds up productivity growth and then
economic growth more broadly? Or should we be happy with two-percent growth, at
least in United States, going forward?

No, I think we could hope to do a bit better than that. If
we’re just really talking about labor productivity growth, I think it would be
reasonable to think you could add half a percent to that with better policies,
so two becomes two-and-a-half. If you are also perhaps riding the back of a new
general purpose technology, we might see three percent is possible. I think the
thing, though, that is paramount to remember is that productivity growth is
really central to all sorts of things like government finances simply because
the US, the UK, Europe generally, are just not going to have very much in the
way of employment growth. Growth is going to come from labor productivity
increasing, not labor inputs increasing.

Yeah. This is my sort of second worry — that people have
grown more skeptical about the value of economic growth. They feel like it only
goes to the top, and in the process we’re ruining the environment. I sense a
lot more growth skepticism than perhaps was the case a generation ago.

Yeah. I think those environmental concerns have been there
for a long time, but I do think they have grown. The striking feature of the
last few decades has, particularly in the United States, been trends in the
distribution of income. Growth has been much less inclusive than it was, say,
at the time in the mid-20th century. Some of that, again, goes back to
competition policy, anti-trust and so on. It seems to me that that would be
something to strengthen, both in terms of thinking about income distribution
and greater efficiency.

My guest today has been Nicholas Crafts. Professor Crafts,
thanks for coming on the podcast.

Thank you, Jim.