Scott Sumner  

Automation destroyed 20 million manufacturing jobs

Teaching the Minimum Wage... Three Themes of the Niskanen C...

I am becoming increasingly frustrated by all the articles I'm reading about how trade is supposedly decimating jobs in US manufacturing. I went to the FRED data set, and they have manufacturing output (real) going back to the first quarter of 1987. So let's start there. Their series shows total manufacturing output rising from 69.789 in 1987 to 129.129 in the most recent quarter. That's an 85% gain.

At the same time, manufacturing employment has fallen, from 17.499 million to 12.275 million. This represents a decline from 17.3% of total employment to only 8.5% of total employment. That's the figure that has people so upset. But the cause is not trade; it's automation.

What people forget is that an increase in imports tend to cause an increase in exports. It's true that imports may currently exceed exports, but that gap is not caused by trade, it reflects saving/investment imbalances. But let's say I'm wrong. Let's say trade does cause the current account deficit. In that case, how many jobs have been lost to trade since 1987?


Not because the CA deficit is zero; it's 2.26% of GDP in the most recent quarter. However in the first quarter of 1987 it was 3.19% of GDP. So the trade deficit has shrunk over that period. If you really believed that trade deficits caused unemployment (I don't) you'd be forced to conclude that, in net terms, trade has added jobs to US manufacturing since 1987. After all, the deficit has gotten smaller.

When people say they are upset about trade, I think that what really bothers them is that automation is allowing us to produce 85% more manufactured goods with far fewer workers. That transition has been painful for many workers, but it's not about trade---except in one respect.

Trade allows the US to concentrate in industries where we have a comparative advantage (aircraft, chemicals, agricultural products, high tech goods, movies, pharmaceuticals, coal, etc.) We then import cars, toys, sneakers, TVs, clothing, furniture and lots of other goods. It's likely that our productivity is higher in the industries where we export as compared to the industries where we import. So in that sense, trade may be speeding up the pace by which automation costs jobs. But probably only slightly; in previous posts I've shown that even within a given industry, such as steel, the job loss is overwhelmingly about automation, not trade.

Why do so may people blame trade? Cognitive illusions. It seems like imports would reduce aggregate demand, and that this would reduce employment. Those effects are highly visible. It's human nature to demonize foreigners. But even Paul Krugman rejects that argument, at least when we are not at the zero bound. The Fed would simply offset any reduction to AD due to trade. Even if you thought it was depressing output during the recent recession, the effect would have gone away once we exited the zero bound. But the jobs are still not there in manufacturing. The bottom line is that to the extent trade is a problem, it has nothing to do with aggregate demand. The real problem is frictional unemployment, the difficulty of transitioning from dying industries to growing industries.

Business Insider has an excellent article discussing a recent interview with Carrier (United Technologies) CEO Greg Hayes, by Jim Cramer:

The result of keeping the plant in Indiana open is a $16 million investment to drive down the cost of production, so as to reduce the cost gap with operating in Mexico.

What does that mean? Automation. What does that mean? Fewer jobs, Hayes acknowledged.

From the transcript (emphasis added):

GREG HAYES: Right. Well, and again, if you think about what we talked about last week, we're going to make a $16 million investment in that factory in Indianapolis to automate to drive the cost down so that we can continue to be competitive. Now is it as cheap as moving to Mexico with lower cost of labor? No. But we will make that plant competitive just because we'll make the capital investments there.


GREG HAYES: But what that ultimately means is there will be fewer jobs.

The general theme here is something we've been writing about a lot at Business Insider. Yes, low-skilled jobs are being lost to other countries, but they're also being lost to technology.

Everyone from liberal, Nobel-winning economist Paul Krugman to Republican Sen. Ben Sasse has noted that technological developments are a bigger threat to American workers than trade. Viktor Shvets, a strategist at Macquarie, has called it the "third industrial revolution."

One by one we are repeating all the mistakes of the Great Depression. We are falling prey to fallacies that were adopted in the 1930s, but rejected by the 1990s. Now they are all coming back:

1. The belief that financial crisis caused the Great Depression (rather than vice versa).

2. The view that the Fed was out of ammo.

3. The view that interest rates measure the stance of monetary policy.

4. The view that exchange rate depreciation is a beggar-thy-neighbor policy.

5. The view that fiscal stimulus is needed in recessions.

6. The view that a higher minimum wage could boost the economy.

7. The view that mercantilist policies are justified.

I spent much of my life studying the intellectual climate during the 1930s, including reading all of the New York Times from 1929-38. I can't tell you how depressing it is to see today's intellectual climate reverting back to the vulgar Keynesianism of the 1930s. To see us making all of the same mistakes.

But opposition to free trade might be the worst of all, as it's based on pure innumeracy. That's why even Keynesians like Krugman don't buy the argument. Unfortunately, the innumerates are probably in the majority.

PS. About the title of the post. The US did not even have 20 million manufacturing jobs in 1987. So how could we lose that many? I derived this figure by first assuming that our current manufacturing output was produced with the same level of productivity as in 1987---in which case employment would have risen by 85%. That would have required an extra 15 million workers. Then I noted that actual manufacturing employment has fallen by 5 million. The gap is 20 million. Is that a ridiculous comparison? Of course it is. But so are all the estimates of jobs lost from trade.

So what's all this really about? Perhaps the "feminization" of America. When farm work was wiped out by automation, uneducated farmers generally found factory jobs in the city. Now factory workers are being asked to transition to service sector jobs that have been traditionally seen as "women's work". Even worse, the culture is pushing back against a lot of traditionally masculine character traits (especially on campuses). The alt-right is overtly anti-feminist, and Trump ran a consciously macho themed campaign. This all may seem to be about trade, but it's actually about automation and low-skilled men who feel emasculated.

COMMENTS (28 to date)
Daniel Kahn writes:

Isn't part of the disconnect that when you and most economists say jobs you mean net jobs whereas trade protectionists are mad about specific jobs that were lost, some of them to lower-wage countries? Often the new jobs created pay lower wages than the old ones. I don't know if that's actually true but it seems to be an assumption of the protectionist crowd.

Mark Bahner writes:

Hi Scott,

In my lunchtime, I don't have much more time than to say that this is good stuff! By that, I mean it's dealing with an important topic, and it presents a clear and seemingly persuasive argument.

Your work on monetary analysis and policy is good stuff too. Keep up the good work!

Best wishes,

Nathan Taylor writes:

Excellent post.

Minor quibble/typo: since this is a post against innumeracy, maybe worth fixing the numbering scheme on 1930's fallacies.

Right now the numbers go 1,2,3,4,3,4,5. But I'm pretty sure the numbering convention goes 1,2,3,4,5,6,7. :)

Anyway. Fine post. Typo not that big of a deal I guess.

John Hall writes:

It seems like what you are arguing is that the people in favor of trade restrictions have the following model in their heads:

unemployment rate = a + b*trade deficit + error

where b>0. And since the trade deficit has declined, then the model implies that the unemployment rate should decline.

I worry that you are straw manning the protectionists.

I hear them talk a lot about how the new jobs are lower paying than manufacturing jobs. I hear them talk about how the employment to population ratio is low compared to history even if the unemployment rate is elevated.

They already know that the relationship between current account and unemployment is weak and are making other arguments.

MikeP writes:

Right now the numbers go 1,2,3,4,3,4,5. But I'm pretty sure the numbering convention goes 1,2,3,4,5,6,7.

I took it as a Freudian expression of a series of mistakes that repeatedly appear.

Scott Sumner writes:

Daniel and John, Agreed, but that's equally true of trade and automation. Both are simply a way of producing a given output with fewer American workers. Both are probably making wage income more unequal.

And many people do point to the trade deficit, including politicians like Trump, and academics like Autor.

Thanks Mark.

Nathan, That's what happens when you keep adding items, without re-numbering!

Chuck E. writes:

Sometimes it's not automation.
In the late 90's, tool & die jobs disappeared in the Chicago area. NAFTA was passed and there was a 30% difference in US dollar/Canadian dollar. All my customers bought their dies from companies in Toronto. Different zip code, different area code; but same quality at a 30% discount in costs & duty free. There was no way to make up that much difference with automation or belt tightening. Most owners closed up their shops and retired. Now we have lost a basic skill in this country, and this causes buyers to look over seas for suppliers. One wonders what would have happened if our government addressed the currency imbalance at the time?

Kevin Erdmann writes:

"When farm work was wiped out by automation, uneducated farmers generally found factory jobs in the city. Now factory workers are being asked to transition to service sector jobs that have been traditionally seen as 'women's work'."

Actually, it's the same problem as with the transfer from agriculture. The new service jobs are also in the city. We don't have free flow of labor, because the cities where the service jobs are are closed to new residents. And construction isn't exactly 'women's work', as it were.

All the recent work I have read about the trade problem treats the housing "bubble" as a temporary mask of the trade and manufacturing problem. But what if there were never too many houses in general? And what if there were especially never too many houses where the opportunities in non-tradable sectors are strong (the high income urban centers)? Then the anomaly has been the housing bust. We destroyed a bunch of construction jobs for no reason. Researchers on this topic explicitly note that employment was just fine until housing collapsed. They just mistakenly believe that the boom was the problem and the collapse was the correction.

There's a virus in our brains. We've got to get it out.

Effem writes:

I don't see this pertaining to a race or sex. No group likes when their livelihood declines. If the day were to come when robots could do "female" jobs (nanny, nurse, teacher) and we recommend they pick up male-roles like construction, policing, and sanitation I'm pretty sure there would be female revolt.

Seems to me like the main lesson of the Great Depression is that if you are going to unleash productive forces on one large segment of the population you'd better have a plan in place to help those people. Economics can change a lot more quickly than culture can. People only have one life...telling them to "get over it" because their grandkids will be better off doesn't count for much. Once we failed to learn that lesson, the rest was just dominoes falling (and continuing to fall).

Effem writes:


I certainly think lack of mobility is a problem, and i've enjoyed your work on this. It's not just housing costs but also the fact that two-worker households are far less mobile. This is a big problem.

Also, there is a massive cultural difference between rural and urban life. I live in NYC, there is very little chance that an unemployed welder from Pennsylvania is going to be accepted in this city. Quite the contrary - For all this talk of city-dwellers being "progressive" they sure seem to scoff at other versions of the American value system (e.g., guns, hunting, religion, cargo pants).

Kevin Erdmann writes:

Effem, I think you have to think on the margin on this. The average households doesn't have to move. The marginal household does. Naturally, the families that best fit the locality will be the ones to move. And, I would suggest that lifting the sharp obstructions to movement that have been in place would almost certainly have to end up with a better match between families and their locations than what we have now.

Clearly there is a massive demand for urban living. That's why the prices are so high. If apartments in San Francisco go for $3,000 per month, then the family that would have moved there for $2,990 must feel pretty comfortable about being there.

Bob Murphy writes:


Overall I like your post, but I think you are opening yourself up to a counterattack if you play the card of "the CA deficit is lower now than in 1987 so if anything, trade 'created' jobs."

That's just an artifact of you starting in 1987 (since that's apparently when the FRED series on manufacturing starts). If you look at the long-term graph of the current account deficit, you see that the standard story of "US workers are getting killed because of the rising trade deficit" fits the facts decently well, in the sense that it is entirely a phenomenon of the 1980s and onward.

There's nothing more I can say except that, people should click the link and look at the chart. Someone who believed the protectionist storyline would find Scott's contrast between 1987 and 2016 to be extreme cherry picking.

(To reiterate, I know it wasn't cherry picking and I don't think free trade hurts workers, but I'm just warning about that particular argument from Scott's post.)

Nathan Taylor writes:

On twitter, Noah Smith links to a 2016 paper by Pierce and Schott

Then says "Next time someone tells you the decline in U.S. manufacturing employment was all about automation, link them to Pierce & Schott (2016)."

Here's direct link to paper

Where I'm now confused is it seems like everyone agrees that more productivity decreases factory employment. At least per good sold. But that this productivity growth can come from within a country, or indirectly via trade. So I'm not sure if you disagree with Pierce and Schott. But Noah Smith seems to say that, which I can see why. So is this mostly a semantic/accounting issue, where everyone agrees that tariff decline boosted trade driven productivity and hence jumped up job loss, at least on the margin? Or is there a deeper, more substantive disagreement?

For helping you decide if you want to click through to the paper, here's a copy/paste of the abstract.

This paper links the sharp drop in US manufacturing employment after 2000 to a change in US trade policy that eliminated potential tariff increases on Chinese imports. Industries more exposed to the change experience greater employment loss, increased imports from China, and higher entry by US importers and foreign-owned Chinese exporters. At the plant level, shifts toward less labor-intensive production and exposure to the policy via input-output linkages also contribute to the decline in employment. Results are robust to other potential explanations of employment loss, and there is no similar reaction in the European Union, where policy did not change.

Mr. Econotarian writes:

We should be honest though that exposure to trade with low labor cost countries leads to enhanced capital investment in the country with higher labor costs and thus enhanced productivity per worker (and perhaps a change in the ratio of returns to capital and labor).

Brent Buckner writes:

Matthew Klein at FT Alphaville took a rough cut at attribution (automation vs. globalisation):
[site registration required]

I think his counterfactual analysis misses a significant element. By "Assuming no additional boons to productivity" I think he's assuming that there wouldn't have been additional automation in the U.S.

Brian Donohue writes:

The devaluation of brawn has been pretty relentless over the past 60 years. In 1955, 1 million 25-54 year old American males were 'inactive'- now it's 7 million. Activity rate has gone from 97% to 88%.

Not your (or my) problem yet, cuz we're smart, not brawny.

Scott Sumner writes:

Chuck, You said:

"One wonders what would have happened if our government addressed the currency imbalance at the time?"

What makes you think there was a "currency imbalance"?

Kevin, More construction jobs would help, but would not come close to offsetting the loss of manufacturing jobs.

Effem, You said:

"If the day were to come when robots could do "female" jobs (nanny, nurse, teacher) and we recommend they pick up male-roles like construction, policing, and sanitation I'm pretty sure there would be female revolt."

I'm not so sure. Huge numbers of female jobs were lost when telephone operators were replaced with electronic switchboard, and I don't recall any great protest movement from women.

Bob, Sure, you could argue that a growing CA deficit cost jobs between the 1960s and the 1980s, but not since then. On the other hand most people seem to think that the big loss of jobs to trade occurred since the 1980s.

(And of course you and I know that CA deficits don't cost jobs in the first place.)

Nathan, That abstract actually undercuts the argument, as the EU has also lost massive numbers of manufacturing jobs, despite them running a big trade surplus. It's a global problem, that has little to do with trade.

Again, look at my recent post on the US steel industry. We still produce lots of steel, it's just that it requires very few workers. Even China has a rust belt where manufacturing jobs are being lost at a rapid clip.

pyroseed13 writes:

Scott, I am confused by this post. Sure, if you stretch the time horizon long enough, you will find that factors other than trade have substantially reduced manufacturing employment. But if you stick to the period from 2000 to 2008, it seems hard to ignore that China was the main culprit in reducing manufacturing employment. Also, I'm not sure how the Schott and Pierce paper undercuts this argument. The authors say in the abstract that there was no similar reaction in Europe. You ignore mounds of evidence by claiming this "has little to do with trade."

I agree with points made in the post about benefits of trade and that mfg. job losses have been caused more by automation than trade with Mexico and China (US/China policy is public school and homeschool debate topic for the current school year).

I think though that the wider point should be made that we can't know how much manufacturing could have flourished in less regulated US economy. How many thousands of small firms have been tripped up by regulations? (I spent a dozen hours on the phone with Texas govt. officials trying to explain my few hundred dollars in Texas income as nonprofit for seminars and book sales shouldn't owe $2,000 in "estimated" sales taxes.)

The hard work of designing and making things, like the hard work of research, writing, and editing,is rewarding when markets guide the production process toward goods consumers value. We can only speculate on whether the U.S. would have another 10, 20, or 30 million people active as craftsmen and manufacturers.

Ron Swanson of Parks & Recreation advocates woodworking. Mike Rowe promotes training "for skilled jobs that actually exist." Prison reform and disability reform could open doors for tens of millions to return to productive work (where they are current blocked by government force or incentives). We have hundreds of nonprofits trying to help with job training in part because they are taxed and regulated more lightly than for-profits.

Consider the mention of comparative advantage in airplane manufacturing. At least a dozen innovative airplane companies have failed to gain regulatory approval, any one of which could have grown into large firms. Rich Karlgaard at Forbes has written over the years on aircraft innovation. And more recently we have federal regulations blocking Uber for airplane rides.

So if we consider firms across dozens of industries, thousands of small firms would likely grow and hire many people to design and make things (plus provide services) after local, state, and federal regulatory burdens are reduced.

When regulatory and corporate tax reforms allows overall economic growth to return to 3-4% or more from 2%, opportunities and jobs will open in aviation and other industries across the economy. Housing and health care deregulation would reduce rents, mortgage, and health insurance payments. Reduced regulations would allow new private medical schools to open, Skype and smartphone medical sessions, and more immigrant doctors and nurses. All could dramatically expand health care services creating jobs, and reducing medical and health insurance costs.

Okay, this is all stuff readers here know. I just though it would be worth emphasizing these reforms would create the jobs that less-informed people thing are lost due to trade with Mexico and China. .

I wouldn't be surprised if a big part of the negative perception surrounding imports is a widespread misunderstanding of the equation to calculate GDP:


Imports are subtracted to calculate GDP, so it looks like imports reduce GDP. This is an illusion though, because subtracting imports merely prevents the imports included in C, I, and G from being included in Gross DOMESTIC Product.

Adam writes:

Yes, good analysis, right down the center lane of logical and facts. But the conclusion it weirdly illogical: " it's actually who feel emasculated". LOL.

john writes:

re "Not because the CA deficit is zero; it's 2.26% of GDP in the most recent quarter. However in the first quarter of 1987 it was 3.19% of GDP." Doesn't % GDP cover up sector changes, such as more or less imported oil (or Ag products) & the price, such that the point of the argument (more or less manufacturing jobs) gets lost?

Glen Smith writes:

Automation has not been a threat to low skilled jobs until recently except in a more indirect manner. In fact, many industries offering low skilled (with the expected low pay) cannot properly staff those positions. Remember, the stuff that is pretty easy to automate is hard or dangerous for humans while many of the so-called low skill human tasks are harder to automate.

G Marshall writes:

The labor mobility issue is not just 2 income households making it more difficult to move. It is a 60% divorce rate. Dads are not willing to move because it means they cannot see their children. Moms are not even able to move because of shared custodial arrangements. You need a court order to move your children away from their other parent.

E. Harding writes:

"This all may seem to be about trade, but it's actually about automation and low-skilled men who feel emasculated."

-It's not even about that. It's about education. The sheer extent and broadness of Trump's gains relative to Romney (and Clinton's losses relative to Obama) with the non-college-educated of every race, location, and sex is something which cannot be explained by the loss of manufacturing jobs, racism, or any other facile and wrong explanation. Trump even made gains (percentage-wise, at least) relative to Mitt Romney in most counties in Texas, even as he got a smaller percentage of the vote than Mitt Romney nationwide (and in Texas, due to the cities shifting against him). Most counties in Texas have never had an interest in protection of manufacturing jobs. Almost all were solidly Democratic (pro-free-trade) in every sense until 1928. In the county of Indiana which went heaviest for Cruz in the primary, Trump in November got a higher percentage of the vote and a higher number of votes than Mitt Romney did in 2012.

Focusing on manufacturing jobs (proof: rural Texas), racism (proof: 1964 election, Trump's gains among Blacks and Hispanics), sexism (proof: Hillary 2008, Trump's gains among non-college women), or even primary vote (proof: Indiana 2016 primary), as an explanation for Trump's gains over Romney is a fool's errand. It's about education.

Bogwood writes:

An issue not under the "street light" ,difficult to define and measure is the life cycle cost of automation. In terms of energy, watts or BTUs it may be negative. It uses less calories per quart for the farmer to milk three cows by hand than to milk two hundred with machines. The borders on this type of calculation are fuzzy, what gets counted,so there is little data(do you count the farmer's children's energy consumption?). It is generally agreed that the automated food system uses ten oil calories for each delivered food calorie.

Then there is the tendency to automate the production of inherently evil products like four thousand pound cars designed by congressional mandates.

World trade at the current scale is an ecological dead end,literally. Both automation and world trade may be temporary artifacts related to artificially low energy costs subsidized by debt.

Larry writes:

Get it that the robots are surging.

What happens when the robots take the (steering) wheel. The #1 job of American males is driver. SGet it that the robots are surging.

What happens when the robots take the (steering) wheel. The #1 job of American males is driver. Somehow I don't think that monetary policy can do more than not make matters worse. What happens to them? They become nurses?

While I long for the era of NGDPLT, something bigger is happening. We can use things like wage subsidies to buffer the change, but where are we headed? SGet it that the robots are surging.

What happens when the robots take the (steering) wheel. The #1 job of American males is driver. What happens to them? They become nurses? somehow I don't think that monetary policy can do more than not make matters worse.

While I long for the era of NGDPLT, something bigger is happening. We can use things like wage subsidies to buffer the change, but where are we headed? Does monetary policy matter as much in the era of the Matrix, when nobody works? What is the replacement for work and raising kids as the basis of dignity Beer and video games (and worse)? Are we past dignity?

AFAIC you crushed it on monetary policy. I urge you to put that clear thinking brain of yours to work on this bigger question. Thanks!

Olga Baban writes:

Hi Scott,

I like your article. I agree with you that loss of manufacturing jobs is primarily related to automation and disruptive technologies like robotics. Factory workers become obsolete, and fewer of them are needed. Factories must adapt to the new environment and use new technologies to be competitive. This will eliminate even more workers.
On the other hand, trade is great, and we should use comparative advantage with other countries because it benefits the US economy and pushes prices down.

There is a short-run solution for job losses:
Factories, companies and firms are all in need of high quality modern workers. I think the solution is in education, training of human capital to become new gadgets testers, solar panel installers, oil industry equipment installers, smart city infrastructure workers, modern highway and railway construction equipment operators and so on. I think the government should fund this type of training and education programs for workers.

But in the long run, what do you think will happen if this tendency with automation continues? Will all manufacturing jobs, engineering jobs be destroyed? If robots/AI develop software, analyze data, shopping/banking/insurance goes 100% online, drones deliver goods, autonomous cars ride the roads, all cities become smart cities with very low maintenance, what will be left for people to do for a living?

Thank you,

Comments for this entry have been closed
Return to top