Susan Houseman on Manufacturing
Oct 21 2019

mfg-200x300.jpg Economist Susan Houseman of the Upjohn Institute for Employment Research talks about the manufacturing sector with EconTalk host Russ Roberts. Houseman argues that the data surrounding both manufacturing output and employment have been misunderstood and misinterpreted. In particular, she argues that conclusions about the growth of manufacturing are driven overwhelmingly by computer production while the rest of manufacturing has been stagnant. She also argues that productivity has a small role in reducing manufacturing employment. Trade has been the main cause of employment reductions. These claims go against the standard narratives most economists have been telling for the last 20 years.

David Autor on Trade, China, and U.S. Labor Markets
David Autor of MIT talks with EconTalk host Russ Roberts about the fundamentals of trade and his research on the impact on workers and communities from trade with China. Autor's research finds large and persistent effects on manufacturing jobs and...
Adam Davidson on Manufacturing
Adam Davidson of NPR's Planet Money talks with EconTalk host Russ Roberts about manufacturing. Based on an article Davidson wrote for The Atlantic, the conversation looks at the past, present, and future of manufacturing. Davidson visited an after-market auto parts...
Explore audio transcript, further reading that will help you delve deeper into this week’s episode, and vigorous conversations in the form of our comments section below.


Brett Ferrell
Oct 21 2019 at 8:16am

Listening to this week’s episode, and I’m curious how this is actually done.  For example, in this link you’ll see the various legal entities for the Procter and Gamble company.  I can assure you that not everyone involved in “manufacturing” their goods resides in one of the entities.  For example, the “Paper Products” company sure includes their paper facilities, for both line workers and support staff.  I’d be curious how this is classified.

Richard Sprague
Oct 21 2019 at 4:56pm

Yes, I’m curious too.

When I look up the SIC codes for a few tech companies, I see:

Apple, Inc:  SIC CODE 7372 – Prepackaged Software and a bunch of others.

Amazon: SIC CODE 7374 – Computer Processing and Data Preparation and Processing Services

Microsoft: a combination of Software Publisher and others.

From the guest’s comments it sounds like the classification is based on the physical building where the business is performed. But I could see a lot of ways that could go wrong. Amazon’s AWS hosting is performed in largely people-free data centers, but the real work happens remotely.

David Jinkins
Oct 22 2019 at 3:25am

They used the term “establishment” a few times, and I wish they would have explained it because it has a specific meaning in the context of national economics statistics.  Here is the BLS definition:

“Establishment: The physical location of a certain economic activity—for example, a factory, mine, store, or office. A single establishment generally produces a single good or provides a single service. An enterprise (a private firm, government, or nonprofit organization) can consist of a single establishment or multiple establishments. All establishments in an enterprise may be classified in one industry (e.g., a chain), or they may be classified in different industries (e.g., a conglomerate).”

As a rule of thumb, I think of an establishment as a plant or a store. A firm might control many plants or stores, and they might do different things.  According to my understanding, support staff, say the janitors, at a manufacturing plant are counted as manufacturing workers by the BLS.

a curious cat
Oct 25 2019 at 10:26pm

It gets even more complicated, there are also NAICS codes that are similar to but different from SIC codes. NAICS codes are used by insurers to classify businesses and these codes inform the insurance rates the business gets from their insurer.

Patrick Chase
Oct 21 2019 at 1:26pm

My intuition is that Houseman’s work could explain some of the decoupling between unemployment and inflation discussed in this article from the Economist. I’d be interested to hear any insight someone might offer related to that. I find the apparent lack of relationship between the two mildly disconcerting if only for the fact that it doesn’t really seem to have any reliable explanation.



T Freeman
Oct 22 2019 at 6:13am

Does anyone have any insight on how declines in manufacturing relates to  Baumol’s cost disease? “The rise of salaries in jobs that have experienced no increase of labor productivity, in response to rising salaries in other jobs that have experienced the labor productivity.”

A GP doctor from the 1800s doctors likely sees a similar amount of patients – the same for an orchestra who has a similar number of performances.  What does it mean when the auto plant goes from 5,000 down to 1,000 – all while producing more cars – if the staff have not harnessed their increased productivity with wage growth?

At one point Russ and Susan compare manufacturing workers income to other workers – shouldn’t it be the other way around? Much of the service sector is relationship based with limited scope for productivity growth – a hairdresser can’t perform 150 haircuts in a day, nor can a doctor see 250 patients.

If our economy (I’m Australian BTW) is increasingly service based, with hollowed out manufacturing, what will lift the  productivity tide to buoy rest of the workers in the economy?

Brandon Hull
Oct 22 2019 at 8:33pm

This was a tough interview to follow, one of the hardest in memory.  The kernel of the narrative: consensus wisdom on US manufacturing productivity misinterprets the BLS statistics due to massively lopsided contribution of the ‘computer’ sectors.   These distortions are difficult to see, because they come not from an increase in computer sector manufacturing share, which has remained flat, but from the massive ‘price reductions’ represented by the sector as tech keep getting better, very quickly.   Russ says that this has rattled his assumptions that automation and commensurate productivity gains reassure us that temporary trade imbalances are tolerable.   But Susan’s work concludes that automation gains were far smaller than we thought, and trade/job exports account for a much larger portion of the manufacturing job loss and dislocations roiling society.

A pretty provocative thesis and worth plowing through an interview that doesn’t flow very well!


Oct 23 2019 at 1:20pm

I agree this discussion was very hard to follow, which is a shame because the message is interesting and important.  Frankly, I think Russ as moderator was guilty of the “curse of knowledge” from his prior reading of Houseman’s work: Russ knew the conclusion ahead of time, so felt comfortable immediately diving into the supporting data.  Houseman didn’t get a chance to outline her full narrative until well into the podcast — so we, as listeners, didn’t know how to mentally organize all the data that came before.  Houseman does a good job telling her own story in this video on c-span.

Oct 25 2019 at 11:08pm

Yeah, I’m with you. It was oddly difficult to follow this discussion. As you said, it didn’t flow. I believe I got the big point, but I feel like I need to listen again to understand all her points.

It also felt like an important episode, as it’s not every week where Russ starts to doubt one of his core ideas.

Duncan E
Oct 23 2019 at 12:28am

Wow so is she pro Trump trade policy? Basically the speed of trade adjustment is the cause of working class pain. Its not the robots after all. Its Chinese pseudo slave labor maybe?

I would be interested to see Don Boudreaux’s take on this?

And even more cheerfully its happening to R&D next!?

Oct 30 2019 at 12:24pm

I think that you are making a big jump here from the statement that “the speed of trade adjustment is causing working class pain” which I think that both Russ, and Susan Houseman, and many others on the left and right would agree with, to the conclusion that Trump’s trade policy of tariffs is the answer.  I might have missed it, but I didn’t hear any specific policy recommendations offered.

Hudson Cashdan
Oct 23 2019 at 9:36am

Fascinating podcast around a provocative and complex thesis. Russ and Susan did a good job of explaining the mismeasurement of inflation, Hedonic Adjustment, and the outsized role of electronics in the manufacturing story; but perhaps Russ, Susan, or someone in the comments can elaborate on this part?:

“in some sense, we are, in effect, understating our imports in real inflation-adjusted terms; we’re overstating the value added in this country–how much output we’re producing–or we’re overstating productivity growth.”

I will try: because we have used the wrong inflation adjuster, we are actually importing a higher volume of product than we thought (widget volume= total$/price). Since the total output of finished goods is more accurately measured, we are actually using a higher volume of inputs per unit of finished good production than we previously thought. Ergo, we have been overstating productivity growth.

The total production side is a bit less clear. Houseman says that total production is GDP. GDP is the difference between real (inflation-adjusted) output of finished and real inputs of raw and intermediate goods. “Real” means the aggregate volume of goods. Houseman is saying that the volume output is being properly measured (or close), but the volume of inputs is being under-estimated. We see [output – input] growing, which we call GDP growth and assume we are producing more. But really our margins just got fatter due to cheap imports.

Public data seems to support this thesis: since the mid 90’s, S&P 500 Industrials operating profit margins have expanded from ~6.5% to 11.5%.

Thanks for another great podcast, Russ.

Rob Wiblin
Oct 23 2019 at 4:45pm

Fascinating interview that forced me to update my views about various things.

But I find interviews about this topic frustrating and unconvincing when they don’t consider the welfare of all the Chinese/Vietnamese/Bangladeshis/etc affected by US trade policy, or their right to freely sell products to Americans unless there an incredibly compelling reason to deny it to them.

Considering the effects on US residents, but not the poverty reductions or human rights of foreigners, misses what is of greatest moral importance here.

Oct 23 2019 at 9:42pm

Rob, consider the morality of propping up oppressive genocidal communist totalitarian governments, while eviscerating the manufacturing capacity and knowledge of the worlds largest democracy. Maybe there is a small moral win over the last 10 years getting Chinese farmers into factories, but consider the long term cost of billions of people oppressed. Or a future where China is the one world power.

Oct 27 2019 at 3:12pm

If the argument is that we should have kept China poor because they would be more likely to democratise (wrong in my view), or would have given the CCP fewer resources to promote autocracy globally (more plausible) — and in the long run that outweighs the direct humanitarian costs of poverty in China — than that’s interesting, and would make for an engaging and provocative podcast episode.

But I’ve never heard someone actually make that case! And I don’t think it would stand up that well.

Oct 30 2019 at 12:28pm

I think that most of the evidence supports the view that free trade leads to more freedom in other areas of life as well.  Certainly China is a much more free place now than it was in the past.

Mayfair Holmesburg
Oct 23 2019 at 9:55pm

Great podcast. But they discussed trade and technology separately, never really blending the two together.

The explosion in global trade was enabled mostly by technology – both communication hardware and software, translation software, Internet, Ecommerce, ERP, Supply Chain, GPS tracking etc all make it seem like the Chinese factory is right next door.

Supply chain software with the ability to predict demand across bill of materials, lead time calculations and order volume mitigate the longer lead times from overseas.


Oct 24 2019 at 1:44am

Shouldn’t it be way easier to measure CPI today? Will all the e-commerce website, you don’t have to actually go to Walmart to get many data, just run some python program and you get many  datayou need, even for Walmart.

Oct 24 2019 at 2:06pm

I’m always amazed when someone finds huge gaps in our understanding of relatively simple things such as this.
What I missed was how wholesale input pricing is compared to retail outputs to determine manufacturing. Sure, sometimes the output is also wholesale but not always. More importantly, the finished output that is sold retail often has a much bigger intellectual property component than the input parts. How would that be sepeated to measure manufacturing output?

Second, how are natural resources considered in manufacturing. For example, to produce pure copper, gold, etc., which often come from the same ore, there is a material value and a manufacturing cost but I don’t see how to separate those using an output value measure.

Ben Riechers
Oct 25 2019 at 11:01am

I’m reminded of the observation that bad data is worse than no data.  One of the big challenges with any analysis is the extrapolations made by agenda-driven people (usually not connected to the analysis).  As I and many others have said, you can draw a line in any direction from a single data point.

We are once again responding to actions from policies developed from an oversimplification of a complex world, including the world economy.  Without the simplification, inaction by the government would be more likely and would almost always produce a better overall result.  Where was Hayek’s quote, “The curious task of economics is to demonstrate to men how little they know about what they imagine they can design” when it comes to the simple-mindedness of our free-trade platitudes?  How big does a market have to be to gain 80% of the comparative advantage of trade?  I’ve read that President Lincoln believed in free trade and defended free trade as a lawyer within the U.S.  I remember scoffing at his limited understanding.  Oops.

Dallas Weaver Ph.D.
Oct 27 2019 at 1:35pm

The most profound concept in this discussion was the observation that how you do the accounting details determines the interpretation of the results.  It is a bit of a DUH but accounting systems from the past can give an irrational view of the present when the technology changes.


The conclusion that technology shifts in the world aren’t responsible is partially based upon a lack of understanding of how technology is shifting the rules and decision making in manufacturing along with other areas of our society.     By saying that most of the change in manufacturing was in the electronic area and dismissing fundamental change in other manufacturing sub-sectors indicates little real-world experience on the factory floor.


As science and technology have become so specialized that the general person and especially those trained in the Social Sciences/humanities areas are really clueless and still depend upon the industrial Union views of manufacturing as just plugging in interchangeable people on an assembly line.   The reality of science/technology and associated businesses becoming increasingly specialized, workers are no longer interchangeable.   The man who fixes the robot or computer-controlled machine tool may even be an outside contractor or an independent.

Most of the manufacturing sectors showing little growth are often facing direct or indirect competition from the electronic sector or just people only being able to use so many appliances, etc.    These internal shifts in the manufacturing sectors as it adapts to competition don’t say anything about foreign trade.

The increase in specialization and rapid technology shift does say a lot about opportunities to make “intermediate” goods using advanced technologies and specialization.   With science and technology open to the whole world combined with the evolution of the regulatory state in the US made it desirable to build many new “intermediate” goods outside the US where a new greenfield factor could be created in 3 months with 500 employees.   The delay times of regulators were often more significant in decision making than labor rates.   Could Apple have obtained permission to assemble the i-phone in California with a 100,000 employee new manufacturing facility in less than a decade?


Old industries like primary steel who wouldn’t or couldn’t innovate and keep up with the technology (note that US steel destroyed its world-renowned steel R&D operation in the ’60) lost to innovative mini-mills (using increasing amounts of recycled steel) and blamed their failures on foreign imports  when their businesses were loosing to technological innovation.   Blaming imports is good politics for rent seeing,

Comments are closed.


This week's guest:

This week's focus:

Additional ideas and people mentioned in this podcast episode:

A few more readings and background resources:

A few more EconTalk podcast episodes:

TimePodcast Episode Highlights

Intro. [Recording date: September 4, 2019.]

Russ Roberts: My guest is economist Susan Houseman. She's the Vice President, Director of Research at the Upjohn Institute for Employment Research. She is a labor economist whose recent work focuses on domestic outsourcing, offshoring, manufacturing, and measurement issues in economic statistics. And our topic for today is what has been happening to manufacturing in the United States. Susan, welcome to EconTalk.

Susan Houseman: I'm happy to be here.


Russ Roberts: Until I read your work, I thought I understood what's been happening to manufacturing in the United States. Manufacturing employment has been falling pretty steadily as proportion of total employment since the end of World War II. In the first decade of the 21st century, even the absolute number of manufacturing employees, not just the proportion, but the absolute number fell by over five million jobs. At the same time, manufacturing output has grown steadily, more or less. Obviously, there's ups and downs, but the general trend is strong and upward.

So my story--my narrative about this sector of the economy, and it gets a lot of attention for a lot of different reasons, but my narrative has been, and I'm not alone in telling the story--that over the last 50 years, manufacturing as a source of employment, has become much less important. But, manufacturing as a source of output is healthy. We're not being hollowed out. It's not true, "We don't make anything anymore." What's happened is that we're making a lot more than we used to, but with fewer workers. And that's because U.S. manufacturing workers are a lot more productive. Due to the use of technology and the production process, we just don't need as many workers as we once did to produce even more than we've been producing in the past.

And the least productive, the most manual-type jobs in the manufacturing sector have gone overseas. So, that leaves the United States with a smaller number of manufacturing workers doing more high tech stuff, producing a lot more. So that's my old understanding. Old, till about a couple weeks ago.

Then I read a number of your papers, and I realized that while there are numbers that back up my story, the truth is a lot more complicated and revealing. And there are a number of different aspects of this we're going to explore, but in one paper of yours that you wrote with Christopher Kurz, Paul Lengermann, and Benjamin Mandel, you wrote, "The aggregate numbers are unrepresentative of the trends in most manufacturing industries."

That's one of the findings. That's quite interesting. We're also going to discover that how we just do the basic counting of how big the sector is and how many workers work there, those two things would seem to be pretty straightforward is more complicated than you might think.

So, I'm going to start with a really basic question. And we're going to say that even the simplest of questions is not so simple. And by the way, I want to let listeners know, although our focus is on manufacturing and it's a really important sector of the economy, for political and other reasons, we're going to apply some of these insights, I hope, to other areas.

So, the first question I want to ask is, how do we count manufacturing jobs in the data in the United States as we typically see it? You're reading USA Today, or you're reading The New York Times and you see a little chart and it shows manufacturing workers either as a proportion of total employment, but let's say we're looking at just the absolute--we're just counting: How many workers are there in the manufacturing sector? How do we do that? How's that done in the United States?

Susan Houseman: Well, in the United States, as in really all countries, what gets counted as a job, employment, in the manufacturing sector are people who are employed by businesses that are classified as manufacturers.

So, you have to be a payroll employee of a business that's a manufacturer to be counted in the manufacturing sector. Basically, what happens is, is that the Bureau of Labor Statistics [BLS] sends out a survey to businesses that asks them how many employees they have. That's reported in the Current Employment Statistics [CES]; that's kind of linked to some other administrative data, and it's wrapped up. And so, we gathered the number of people who are employees of manufacturers.

And I'm really stressing the word 'employee' because we also know that many people who are needed for manufacturing jobs or sometimes who actually work in factories are not employees, and they don't get counted.


Russ Roberts: Yeah, we've talked about that phenomenon in the program. We're going to come back to it. Let's start with an even more basic question. Obviously, a business has to be classified as in manufacturing or not, and I suspect that some of the things that are called manufacturing, some folks might go, 'Well, is that manufacturing?' So obviously, there's a definitional issue just to start with, right? Is there anything interesting in that?

Susan Houseman: Well, I think that's pretty straightforward, because the classification occurs at the establishment level. So, generally, we would not be getting sales forces that were headquartered in corporate offices as counting in manufacturing. There would have to be a substantial production element going on in that establishment for it to be counted in the manufacturing sector. So--

Russ Roberts: Somebody told me--I don't know if it's true: I didn't have time to check it before we talked--that journalism is part of, or was part of manufacturing. Obviously, printing newspapers is manufacturing. Are journalists counted as manufacturing workers?

Susan Houseman: They shouldn't be, unless they are working at the printing press itself. So, everybody who's in that establishment is counted, and not everybody in the establishment will be like a line worker in the factory.

Russ Roberts: Right. So obviously, within a factory--and I've been in a few, not a zillion, but a number of factories--there are people that we would obviously recognize as manufacturing workers, the equivalent of people tightening bolts on assembly lines or steering an engine into the body of a car as it's being manufactured.

But of course, there are many people in that building who are not in the direct line of manufacturing. They're the receptionists, the assistants in the offices related that are part of that factory. There are the janitorial crew. They might be the workers who repair certain pieces of equipment, including a copying machine in the office. Correct?

Susan Houseman: That's absolutely right. And I want to add that some of the challenge in measuring manufacturing employment, as will be the case in most sectors is that many of the workers may be contracted out.

So, for example, the manufacturers often use temporary help workers to staff lines in core production jobs. Because those workers are supplied by temporary help agencies, they're not employees of the manufacturer. They get counted in services, not as part of the manufacturing workforce.

There is some evidence to suggest that that phenomenon of contracting out to other companies or using independent contractors who are counted as self-employed is growing. So, to the extent that that's happening--that, being outsourcing of jobs to other firms such as temporary help agencies or using independent contractors--those people are not employees, they're self-employed. To the extent that that's happening and that is growing over time, then we're kind of, in some sense, under-counting the number of people who are involved in, and needed for, the production of manufactured goods.

Our estimates suggest that on the order of half of all jobs required to produce goods--a good, manufactured goods--are counted somewhere outside of the manufacturing industry.


Russ Roberts: Which is kind of extraordinary. We're going to come back to that, but it's just an accounting, a measurement issue, that if you are a factory that has some part of your workforce that is not working for you but working for an employment agency, you're not a manufacturing worker. That's kind of shocking to me, but that's the case.

Now, let's move to a second--and we'll talk about its significance in a few minutes--but I want to move to a second very basic idea, which once you start to think about it, it starts to keep you up at night, I think, if you're in this business. But, it's, on the surface, quite straightforward, which is: output.

So, when we count manufacturing output, once you start thinking about it for a second, you realize it's harder than it might sound. You can count the number of airplanes that Boeing produces, or the number of nuts and bolts that a nuts-and-bolts company produces. But of course, from year to year, those numbers are going up. Some are going down. Is one airplane the same as one bolt? No. So, how do you count manufacturing output to measure the overall significance of the sector in the output side?

Susan Houseman: Right. So, let me make a few distinctions here when we're thinking about output. So first, there are various output concepts, but the main one that we use is basically Value Added. So, let's think of a manufacturer. It purchases, let's say, a steel producer purchases coal, iron ore, and other inputs. And then, in the steel plant, it turns out steel bars or whatever.

We think of the value added of that steel producer, what its labor and its machines--capital--produce, less its purchased inputs. So, that's value added. That's the main output concept used in our country and most countries. It's also sometimes called GDP [Gross Domestic Product]. GDP, value added are the same thing.

The second concept is that we want to adjust for price changes. You talked in the beginning about, well, we want to kind of--this notion that we want to count how many bolts are produced or how many steel bars, I-bars are produced or whatever, but we can't do that. How do you add a bolt up with a nut and products [?]--

Russ Roberts: You just weigh it. It's easy. If the weight has gotten bigger over time. But, of course, that's not significant.

Susan Houseman: Right. You could do that on a very narrow product basis, but that's not the way it's done. In statistical agencies, that's not how the statistics are generated--the ones that you read, other economists read, the media reads and so forth. Everything is counted on a dollar basis. Okay? They count up--they would survey manufacturers or survey business instance[?] and say, 'What were your shipments? What were your purchased inputs?' Subtract one from the other; voila, you've got value added. Okay?

So, knowing that we do this kind of on a dollar basis, the other challenge is that we want actually to measure in some sense, whether the quantity produced has increased, decreased, or stayed the same. Right?

For example, suppose we see that output, value added, has grown by 10% over the last year. Well, maybe that's because prices are 10% higher, and we're producing the same amount. Or maybe prices have stayed the same, and we're producing 10% more. Or maybe--and usually, this is the case--there's some combination of changes in prices and changes in quantity produced.

So, what the government does--this is primarily done out of the Bureau of Labor Statistics--is it conducts very extensive price surveys of manufacturers and other businesses to adjust for inflation levels.

I'm just going to throw this in there, you know, kind of a thought. We can come back to it perhaps later in the conversation; I think we will come back to it. But that is: There's a real challenge to measuring changes in prices over time, especially--well, particularly when kind of the products or the models are changing over time.

Russ Roberts: Yeah, it's a similar issue to what we've talked about here before, which is when we're trying to measure consumer prices, we understand that a TV that's, say, the same price as it was 10 years ago, but has a sharper picture, never breaks, gets a thousand channels instead of three, and can play video games, and surf the Internet and so on is not the same TV. The price has actually gone down. It's bigger too, by the way, often.

And so how do we count for that? And the obvious answer is--I mean, there's two problems with this. One is, how do you account for that? But the second is, the BLS, the Bureau of Labor Statistics, is inevitably less than instantaneous in how it observes these changes in prices and quality.

So, in your work, that we'll talk about in a minute, if domestic manufacturers are substituting foreign inputs, they're importing intermediate goods into the manufacturing process like the coal or the--it's probably not coal, but say other types of either manufactured goods themselves or raw materials from overseas. And the mix of domestic and foreign inputs is changing, because the foreign inputs are cheaper. You're going to mismeasure the size of the inputs and the process that you're describing that the Bureau of Labor Statistics tries to do.

Susan Houseman: That's right. Those are actually two separate issues. And if I could just spend a minute--

Russ Roberts: Yeah, go ahead.

Susan Houseman: explaining both. So, one is: One of the classic problems in measuring inflation--price changes--is when the products themselves change. Here is the example that you used--it's a great one--of TV sets. If the TV set is the same, but consumers--sorry--if the price of the new model of TV is the same as the old model of TV, but consumers value it twice as much--they would be willing to pay twice as much as the old TV model--then in some sense, the price has dropped in half, because you're just getting more than what you used to. So, that's what we call in--what the price index literature calls is quality adjusting your prices.

The second issue that you just alluded to is: what happens when the manufacturer, for example, substitutes for a cheaper input? It was using a producer somewhere that was providing its input widgets for $1 a widget. And now it's found another producer, say, in China, that supplies it for 25 cents. That price has gone down by 75 cents. But, what happens is, is that you might assume that somehow the statisticians would be able to measure that. And that we would somehow know that the quantity, say, that that manufacturer of widgets' inputs it was using stayed the same. But actually, it's another classic problem in price index literature: When there is a substitution from one source to a cheaper source, you miss that price drop.


Russ Roberts: So, let me just give the consumer side example of that, because it's--again, as you say, there's a literature there that it might be a little easier for listeners to understand. When Walmart comes into the world and lowers the price of a lot of goods for consumers, if the BLS, the Bureau of Labor Statistics, isn't sampling Walmart's for prices, which they don't initially because it's brand new, or if Walmart starts carrying a product it didn't carry before, and therefore the BLS misses it, then it will look like prices are higher than they otherwise actually are that consumers face, which means you will understate the purchasing power of consumers. You will understate their standard of living. You'll overstate inflation and you'll understate the growth in material progress. And it's a huge issue, I think, in the measurement of standard of living in the United States. But what's fun--that's not really the right word--but what's fun is that a very similar issue is going on here.

So again, to use your two points, it's not just the TV is getting bigger: I'm also buying it at a cheaper outlet that isn't might not being sampled by the statistical agency that's collecting the data. So, that combination is not good.

Susan Houseman: Right. So, actually, the problem is slightly different, in fact. It may be that the statisticians are slower or BLS is slow to pick up new outlets. But when that switch occurs, the price drop is not captured. Just really quickly, here's why.

The way prices are measured--and it's exactly the identical problem when a bias in indexes and mismeasuring productivity and output when a [?] shifts to a low cost foreign supplier as it is when our consumers shift from purchasing most of their goods at mom and pop shops to Walmart.

The way the price indexes are constructed is they go to the mom and pop shop, and they measure period to period to period to period the price of a very specific good, say Crest toothpaste. Okay? And now they suddenly go to Walmart, and it's, say, 10%, 15% lower.

The way the indexes are constructed at Walmart is it measures the price of, say, Crest toothpaste from period to period to period to period at Walmart. That price drop when you shift your sourcing, isn't captured. We probably don't want to get too much into the weeds as to why that's the case, but it--

Russ Roberts: I didn't realize that. That's dark; that's so depressing.

Susan Houseman: Yes. And that is the fundamental problem. So, if you think that there's a lot of dynamism going on in the economy whereby consumers are constantly trying to shift to--you know, there's a lot of innovation, say, in retail and--

Russ Roberts: Amazon, yeah.

Susan Houseman: Amazon and so forth. And so consumers are shifting their purchases over from a higher-cost retail outlet, say, Walmart now to Amazon that's offering it at a lower price, that price drop just intrinsically, by the way our price indexes are constructed, is not captured.

The identical thing occurs when you shift--when a producer shifts the sourcing of its inputs. That price drop, when it shifts to a lower cost supplier, generally isn't captured when that shift occurs. And so, if you think that a lot of that is going on--and we think that a lot of that has gone on, at least since the late 1990s with the growth of global supply chains and the move towards lower cost areas of production--then in some sense, we are, in effect, understating our imports in real inflation-adjusted terms; we're overstating the value added in this country--how much output we're producing--or we're overstating productivity growth.


Russ Roberts: So, I want to make sure I understand the point you're making about the construction of the index, and I thought I understood this. I guess it's worse than I thought. So, what you're saying is, is that if I've been buying my books at the local corner bookstore, and then Amazon comes along, and discounts books dramatically, and I shift from paying the price at the local corner bookstore, and I start paying a much lower price online at Amazon, eventually, the BLS realizes that Amazon is there and they start sampling Amazon prices. But they're not literally sampling Amazon prices. They're sampling the increase in prices over time. So, that initial drop doesn't ever get registered in the CPI data? Is that what you're telling me?

Susan Houseman: Right. In order to do that, they would have to, for some basket of books, they'd have to average the prices across outlets and--

Russ Roberts: weighted by the sales ratios of those different places in theory, right?

Susan Houseman: Exactly. Now, there is some movement towards doing that, but that absolutely is--and people have looked at scanner data where you're able to get very good prices for identical products across outlets and worked on how much of a bias that is. But, that's not done systematically and it's not done in, certainly for imports or generally for purchases that businesses make.

Russ Roberts: But the BLS does try to adjust for the quality issue in some dimension.

Susan Houseman: Yes, absolutely it does.

Russ Roberts: The first thing we talked about. It may be slow to do it. Right? And of course, some of the ways--I mean, the most obvious example of that kind of challenge we're talking about here is that if your cell phone goes from $600 to $700 to purchase a new smartphone, but now you don't have to have a GPS or a stereo, or etc., a music player, or you name it, that's not a straightforward analytical challenge to correct for that. It's not just that the TV got 12% bigger and I'll say, 'I'll call that a 12% drop in the price,' if the price stayed the same, because it got 12% bigger. When you start getting at the level of changes or brand new products that didn't exist before--like a cell phone to start with, the whole concept of a cell phone--is a bigger intellectual analytical challenge.

It's not just that the TV got 12% bigger and I'll say, "I'll call that 12%. drop in the price, if the price stayed the same," because it got 12% bigger. When you start getting at the level of changes or brand new products that didn't exist before, like cell phone to start with, the whole concept of a cell phone, is a bigger intellectual analytical challenge.

Susan Houseman: Right. Challenge. Absolutely. This problem has been widely recognized and there are methods for addressing and adjusting for quality changes in products. The biggest issue is that it's extremely expensive to do that. The best way arguably to do this is to run what's called hedonic adjustment models. We're not going to get into the weeds on that one, but it just requires a lot of data. Understanding how the different attributes of the product change. And really what they're trying to, what you want to get at is not whether the TV is 25% bigger, but how much do consumers value being 25% bigger.

Russ Roberts: Yeah. Good luck with that, but there are techniques for trying to do that. By the time this episode airs, I'll have a video out called, I think it's going to be called, "Let's Party Like it's 1973." And I talk about the challenges of trying to measure the change in standard of living in the United States for the average American, say, since 1973, over a long period of time.

You go back to 1973 Honda. Honda Civic existed in 1973, just like it exists now. It actually sold for $1,973. A very clever marketing technique by Honda. But if you sat in that car, and I've seen pictures of the interior of it, it would be a little bit jarring. It's basically a metal box with a steering wheel and not much else. There's a radio, but there's no way to charge your cell phone. There's no way to play your phone through the stereo, the speakers. There's no anti-lock brakes. There's no airbags. And, there's a lot of things that are different. And you're right, the real challenge is trying to measure how much people would pay for a modern Civic versus that car.

And, you can't do that for every car at every product. But my understanding is the BLS does do some corrections for areas--for products like cars--that clearly have changed a lot and are important parts of consumer expenditure. Is that correct?

Susan Houseman: That's right. So specifically for autos, it has run a survey of auto manufacturers for a long time that tries to get at how much of the price increase is due to quality changes.


Russ Roberts: So, it's hard to do but they try for some things. They can't do it for all things. And we could stop this conversation here. It's sufficiently educational for me so far, Susan, but we're going to go on because actually, this is just the tip of the iceberg. There's a couple more shocking things to discover.

I want to go back to the quote I read a minute ago from your work where you said the "aggregate numbers are unrepresentative of the trends in most manufacturing industries." Now, we understand it's possible to drown in a lake that has an average depth of a foot. We understand that averages are not representative, necessarily, of the entire population. And in this case, it was surprising to me and really quite a revelation to understand that almost all of the gains in manufacturing output in recent years are due to a single sector, which is computers and electronic products.

We all understand that those have been revolutionized. They're more powerful, dramatically more powerful than they were five years ago, certainly when they were 10 years ago, 20 years ago and so on. But, to realize that the bulk of the output gains--and you'll give me some of the specifics--are due to one sector, what that means is, is that the aggregate number for manufacturing is not capturing how the whole sector is doing.

It's kind of like, if I have--if Warren Buffett walks in the room, the average income goes up a lot, but that average isn't related to anybody's income in the room. It's an enormously larger amount than most of us were making, and it's quite a smaller amount than Warren Buffett is making if it's a big room.

And so, it's a beautiful example, but alarming, about how aggregate data can be misleading, because you're treating it as sort of representative. It's not in this case. Explain.

Susan Houseman: Sure. No, absolutely. I don't know. Maybe we can just back up a little bit and talk about how this became such a salient issue in recent years. As I'm sure your listeners rather know, what was happening to manufacturing became a very prominent issue in the 2000s. It was strongly featured in our last presidential election. Arguably, our current president was elected in some part because of him saying we needed to bring manufacturing back.

And his story was that the tremendous declines in employment that manufacturing experienced in the 2000s were the result of trade. Just to give you--because I think it's important to set this up a little bit about how the manufacturing numbers were so misleading and widely misinterpreted by economists.

Manufacturing employment was, as you mentioned in your introduction, had been shrinking as a share of employment in the country overall for some time, but employment numbers have been relatively stable. There had been some decline in the early 1980s, but those were concentrated in steel and apparel textiles, and represented about a 7% decline. Manufacturing employment was pretty stable through the 1990s. And then in the 2000s, some people referred to it as the collapse of manufacturing employment. Between 2000 and 2007 manufacturing employment declined by over three million jobs, or about 20% just in seven years. Those are both business cycle peaks. And then the Great Recession hit in 2008. It declined by another couple of two to three million. Sorry, it declined by about two to three million. Yeah.

And so in just a decade, manufacturing employment fell by a third. That's completely unprecedented. Those declines in employment had absolutely devastating effects on many regional economies. And there's just been a lot written about those effects. There has been some recovery in recent years, but we are still down by between four to five million jobs--by about 25% since 2000 in employment.

Now, that gets into what was going on. What caused that decline? We heard from President Trump and others, such as Bernie Sanders were among the most prominent proponents of this view: It's trade. They really emphasized trade. And there was some reason for emphasizing trade here. The 2000s, the dollar was appreciating. We had a surge of imports, particularly from China. Our trade deficit increased to 4% of Gross Domestic Product in the country, which was a very high level. It remains about that size today.

And so there are a lot of, kind of, evidence in the trade statistics and so forth that suggested that this just sudden decline, collapse of manufacturing, the 2000s, was linked to trade.

But the other narrative that you started off with, the segment end, was really that this was just part of a long-term trend. Manufacturing employment as a share of aggregate employment that have been falling for a long time. It is also the case--many people don't appreciate this--that manufacturing as a share of our Gross Domestic Product had been falling almost commensurate with the fall of employment share.

But, as you also pointed out in the beginning of the segment, this real output growth--that is adjusted for inflation--of the manufacturing sector had been largely keeping pace with that of the aggregate economy for many decades. So, that presented a bit of a puzzle, right? We've got falling employment shares, falling share of manufacturing and the economy. But once you adjust for inflation, it looks like manufacturing is just doing great.

Russ Roberts: [?] along.

Susan Houseman: It's keeping up with the rest of the economy. So, when we look at the real terms, the sort of quantity measure, we're just doing great.

How do you reconcile these two apparently conflicting sets of descriptive evidence, descriptive statistics? Well, most economists would point to two things. First, if the share of manufacturing in GDP has been falling, but it's real growth--that is, adjusted for inflation--is keeping pace with the aggregate economy, it must be that manufacturing prices are rising more slowly than those price levels overall. Inflation is less in the manufacturing sector. So, that's one thing that's happening.

Russ Roberts: And the argument there was that the application of technology to the manufacturing process could have that--that was a reasonable story.

Susan Houseman: Right. And then the second thing that must be the case is as: if the employment share is falling, manufacturer's employment share is falling in the economy, but output, real output growth is keeping pace with that of the aggregate economy, it must be the case the productivity growth is much higher. Much higher. So, this--

Russ Roberts: Not unrelated to the claim that there's more technological innovation in the manufacturing process; that's keeping inflation down in manufacturing relative to the other sectors, and it's making workers where they are more productive.

Susan Houseman: Exactly. And this story was embraced by many, many, many economists. It worked its way into some speeches by President Obama. You know, the notion that American manufacturing workers, it's sad that we're losing all these jobs, but they are the victims of their own sector success. We're just doing great.

Russ Roberts: Yeah. I've told that story myself many times.

Susan Houseman: But that was really a great misinterpretation of the numbers, and now to your question: Which is that it turns out that the apparent robust, real output growth and fantastic productivity growth in the manufacturing sector in recent years, as well as going back at least to the late 1970s was largely attributable to one relatively small industry--computer and electronic products.

So, computer and electronic products actually has accounted for just 15% of output in U.S. manufacturing for decades. However, when it is output in computer and electronic products, we'll just call it the computer industry. It's mostly--the big products in the sector are computers and semiconductors. When those products are adjusted for the tremendous improvements in quality, the prices are falling and sometimes they're falling really rapidly.

Just for example, during the decade from 1997 to 2007--I happen to have these figures in front of me--the computer electric products industry was growing on an annual basis--again, once you've quality adjusted for the improvements in products--at 20% to 25% per year. Many other industries were actually falling. Their output levels were falling, and about half or more were just roughly stagnant. You had a few that were growing in the low single digits annual rate. It was an enormous outlier.

And when you take that one sector, that big sector out, or that relatively small sector in terms of as a share of manufacturing--when you take computer electronic products out of the mix--you find that over the 1979 to 2000 period, real output growth, while it's keeping pace with aggregate GDP, when it's included falls by over half. So, it's about 45%. The output growth rate is about 45% of that of the aggregate economy.

Russ Roberts: So it's only growing half, roughly half as fast. It looked like it was growing at a similar rate because of computers. Once you take them out, it's dramatically slower, the growth rate.

Susan Houseman: Exactly. When you take computer and electronic products out in the 2000s, manufacturing output is pretty much stagnant. It's not growing.

Russ Roberts: So instead of growing by--

Susan Houseman: Yeah.

Russ Roberts: the same rate as the economy more or less.

Susan Houseman: Right. Right. The numbers are so low, but output fell during the 2000s in the economy overall. But it's on the order of 12%.


Russ Roberts: Here's what I don't understand. So, the story you're telling is that aggregate manufacturing output, using that as a measure of, say, the health of the sector is grossly misleading, because there's a single sector, part of that sector, the computer part that's getting all the--it's an enormous outlier. And we know it's an enormous outlier in terms of price change, for sure, because most of us in our personal lives have experienced some version of that.

Here's what I don't understand. How does it remain a small and, I thought you implied, a somewhat steady fraction of the entire sector? That can't be right. It has to grow dramatically as a proportion of the sector itself, right? If computers in 1980 or even 1990 were x percent of total manufacturing, by today they must be a dramatically larger percentage, because you're telling me they're growing rapidly and everything else is either stagnant or growing slowly. Is that correct?

Susan Houseman: Yeah, in quantity terms, but not in--and remember, their prices are falling, so as a share of nominal GDP, the share has remained roughly the same. The other thing: so, remember the story about productivity growth causing all of the employment losses in manufacturing or the decline each year was really a story about automation. Right? We're thinking about machines coming in and replacing--

Russ Roberts: Computerizing assembly lines, robots.

Susan Houseman: Exactly. But that's not what was going on in computers, in the computer sector. Not at all. What was driving that real output growth and productivity growth were quality-adjusted prices: the improvements in the quality of the products. Where is that happening? It's happening in research and development. It doesn't have anything to do with automation.

And in fact, what's really ironic is you might sort of step back and say, 'Well, the computer industry is an important one.' And, 'Look, it was going great guns. This has to be the real American manufacturing success story.' But there again, the story is more complex. Because, all of this is coming. It's largely coming.

It's largely coming from improvements to product quality. The improvements to product quality could be driving this apparent, enormous growth at the same time that the United States was losing market share in these key sectors. The locus of production during this period was of computers and semiconductors were shifting to Asia.

And so, yes, we still have a strong semiconductor industry in the United States, but it's lost share at all levels to Asia. And when I say at all levels, all kind of different types of chips. My coauthors, Tim Bartik and Tim Sturgeon, and I have a 2015 paper that really uses some industry data that shows the shift in production in the semiconductor industry over to Asia.

Computers, what are we manufacture in computers anymore? Not much. What we do is a lot of defense and some government computers. That's where the industry is concentrated. We don't do consumer production of computers.


Russ Roberts: But as you point out, the contribution of R&D [Research and Development] is enormous. The value added there, if you think again, going back to the smartphone: the smartphone is produced in Asia, manufactured in Asia, but the value added is manufactured--which is not really the right word, because it's not a manufacturing sector, but in everyday English, the word is 'manufactured'--in U.S. design laboratories, design shops, and research and development shops at Apple and elsewhere.

And so there's a footnote to your story. It's true that the computer industry--it's a great point and I did not fully appreciate it. It's a great point that the changes in the price of computing isn't done via technology applied to the production process. It's through technology in the design of the product from scratch. Or, I don't know how to--R&D is the right way to say it. And that's okay. At one level, I don't care where it comes from. But I do care if I'm worried about, say, the employability of less educated workers who can't work in that design shop or in the lab that's creating the new chip or the lab that's designing the newest iPhone screen. They're more likely to work in the production of the screen; and that is shaped outside the United States. And, their opportunities are not so obvious, where they--it's certainly not where they used to be.

Susan Houseman: That's exactly right. So that is, in part, one of the factors--not the only one, but one of the factors that is likely driving the growth of inequality. That, some of these formerly good-paying jobs in the United States are now offshored. So, it's important that we do, that the United States for the strength of the economy, for national security purposes, and other reasons, that we retain very strong research and development capabilities. So, that's a good thing. Right? If much of the research and development is still occurring in the United States, that's important.

But, if the policy issue that we're addressing is, why did so many people in the manufacturing sector lose their jobs, it would be misleading from these statistics alone, which most economists were doing, to jump to the conclusion that because productivity growth was so high in the manufacturing sector, that reflected automation. And it was just the march of technology that was displacing people.


Russ Roberts: Now, I want to come back to the policy implications and inequality in a second, but I just want to make sure we have a chance to get to the point that you make in a different paper on that we've alluded to already, but I want to let you bring it home now and give us the punch line, which is that: If American manufacturers are increasingly using foreign inputs that are cheap, that price drop is not measured in the data. Which means that we've understated the increase in foreign inputs; which means we have overstated the productivity measure of U.S. manufacturing when it does take place in United States. Is that a correct summary?

Susan Houseman: Yes, that's right. When, in the paper that you're referring to, which was published in 2011 in the Journal of Economic Perspectives with several Federal Reserve Board co-authors, we looked, and we tried to get some bounds on the size of that in manufacturing, owing to purchased materials inputs. So, that was just a very partial look at the size of that effect. And it did seem to lower things by a few tenths of a percentage point. Not huge, but big enough to care about. We also underscore that we're only looking at, you know, just purchased materials inputs. The problem is, as the Internet grows, as we have growing trade in services inputs to the degree that oftentimes labor is cheaper overseas than it is here--say, an engineer in India is considerably less expensive than an engineer in the United States--we're already seeing quite a lot of growth of offshoring of services in this country. And that is something to watch out for or just to pay attention to. There's kind of a policy issue, but the statistical issue is that we're going to be underestimating in our data in real terms--real terms: that is, inflation-adjusted terms--how much of services we're importing.

Just to give you an example, think about something: Let's take imports, the tremendous growth of imports from China. Suppose we start buying some consumer electronics from China that are on the order of, you know, a third less. Those purchases come in, and what they're displacing is in terms of what is manufactured in the United States, is actually dollar-wise quite a lot larger than what they show up in our import trade statistics.

And because we don't adjust for the price drop, we're not--we're somehow kind of understating the impact that, you know, picking up $100 million worth of increased imports from a lower cost country, what kind of impact that has in the United States, because maybe it displaces $150 million worth of production here because of those price differences. People often assume that that's somehow captured when we price-adjust imports. When we look at the growth of real imports. By and large, it's not.


Russ Roberts: So, this--a lot of what you're talking about, what we're talking about together, is discussed in a previous EconTalk episode, which I encourage listeners to listen to, with David Autor, and his work with coauthors on what sometimes called a China shock. And your work is kind of an exclamation point on top of that to my ear. It's basically saying that--you know, my story would have been that the drop in manufacturing was a mix of technological change, innovation, and trade. And that that--I don't know the exact mix. You're saying it's overwhelmingly trade, not innovation, not automation, not the application of technology.

And the question, then, is: What's the--why is this a concern relative to past examples of economic change? And one obvious answer, and I'm going to accept this for sure, is that the speed of it is unprecedented, as you mentioned. The speed of which this transformation took place, of the substitution of foreign manufactured products, for U.S. manufactured products is very fast.

At the same time, these large drops in prices that you're talking about--and you can think about these across all the issues we're discussing: the availability of less expensive outlets such as Walmart, and then Amazon, online retailers, the availability of foreign suppliers for both manufacturing and consumers to consume them, and to use them in the production process if you're a manufacturer in the United States--these lead to usually really good things. That, people now have more money available, and more resources available to spend on other things. And those things expand, and we don't see those. It's my critique of David Autor's work and others in this area is that, well, of course, manufacturing was hurt by Chinese trade in 2000, the increased access that American consumers and manufacturers had to Chinese produced goods. But, that led to some unseen, difficult-to-measure benefits in other sectors. Employment didn't collapse--total employment didn't collapse--by five million.

Now, the actual numbers in employment are masked somewhat by changes in labor force participation. I don't think we have a good feel for what happened to the pay of the jobs that people who used to work in manufacturing now are earning elsewhere. They are presumably lower though--I assume, but I don't know that. The story is a little more complicated, I think, given a bipartisan spin, than the Trump-Sanders story of 'we let all the jobs go away,' and in return for cheap goodies, a lot of poor people bought those cheap goodies. It's a little bit more complicated. Or do you disagree?

Susan Houseman: No, I completely agree. And I want to also maybe restate where I stand on all this, which is: I was, in my writing, I was pointing to a widespread misunderstanding of the statistics. Misinterpretation of the statistics. That was really leading to a story, a narrative that really, trade is a very minor part of the job losses, and that it's really fundamentally all about automation.

I think that story, certainly--well, that story certainly is not supported by the descriptive evidence that is widely cited: that manufacturing, output growth is so large and productivity growth is so large, much larger than in the aggregate economy. And this implies that automation has been largely this whole story. And to quote one prominent economist, 'It's not even close.' Okay? That is not supported by the evidence that widely is cited, if you understand what the data say.

I would be the last person to argue that technology has no effect on jobs. It certainly has. Very often, technology and trade kind of work together. Right? And so they both displace[?]. I actually have argued that the simple decompositions that researchers and policymakers often want to do such as, 'Well, technology accounted for 70% ; and trade 30%,' is not a valid--it's not possible. You can't do it. They're interactive. There's no simple decomposition that can be done. And if you think carefully about the way research is conducted, it's usually addressing a much more narrow question than what share of the job loss was accounted for by trade versus technology.

So, just to--that's my position. They're both important. They're often interactive. But the strong narrative that it's really almost all about technology isn't supported, and it really isn't supported by the data going back to the 1980s with the decline in employment share. It's just more complicated.

So, then to get to your other thing, and it's certainly the case that consumers benefit from a strong dollar and a lower cost import goods and services. It's a double edged kind of sword. If we thought that--and, there are always, we know as economist, that there are winners and losers to trade. Right? Some people benefit, some people lose.

The point that I think isn't appreciated enough or maybe brought up enough in policy circles is that when you have very, very large shocks to a sector, like the manufacturing sector experienced in the 2000s, those are fundamentally different than the kinds of changes over longer periods of time, on the margin, to what's done here versus what's done overseas. When you have employment in manufacturing dropping by a third, the number of factories closing by--I think it was on the order of 20%, in a matter of years--that has, you have a tremendous destruction of both physical capital and human capital that can send regional economies into a downward spiral that can take a generation or more to recover from.

Those sorts of things need to be taken into account when we're developing trade policy.

I think that the biggest problem, from my perspective, with the strong form narrative that I'll call it, that it was really all this job loss and for decades and including mostly in the 2000s was just about automation. Right? The problem with that is that it kind of stifled an intelligent debate about trade and trade policy that we needed to have long ago. And what the consequences of trade is for an economy. It's not whether you want to trade, whether you want to close the borders or not. Of course, there's benefits. There's benefits to everyone.

I think it's more a matter of the pace. And when things happen so fast that they cause massive disruptions, that's where the problem can be. And when we didn't diagnose that correctly, we didn't have an important policy conversation that we needed to have.


Russ Roberts: Well, what's interesting about that is that people on the Left and the Right now have decided that trade was an enormous or has been an enormous punisher of certain types of skill levels and certain geographic areas. Whether they misinterpreted the manufacturing data or not, at this point, I don't think it's so important.

There's a growing consensus, I think, somewhat correct, but somewhat wrong--I'll try to outline which is which--that expanded trade with China and the rest of the world has been very damaging to certain areas and certain types of people. Now, part of that obviously is true, and part of it is exactly for the reasons you said I alluded to earlier, which is the speed of that adjustment was quite large. But I want to go back to what you said, and I want to give you my version and let you react to it.

You said there are winners and losers in trade. I would say it a little differently. I would say there are winners and losers from economic change in the short run. That's been true since human beings began to have any dynamism in their choices in life. It's a part of capitalism. It's part of creative destruction.

And so for me what's important, among other things, is the fact that over time, the standard of living rises steadily from the two things we're talking about today--trade and innovation, trade and the application of technology. The challenge is, what do we do about the fact that not everyone gains at every point in time?

And there's a growing feeling among folks that we have to do something about that. Either dial back how open our economy is to trade, which would also imply I think dialing back something about how innovative our economy is and how much technological change we allow. Because, even if I accept your story, Susan, which I think I do, that the technological gains in manufacturing are not as important as I may have thought, they are happening elsewhere. Software is eating the world in many places, many parts of the economy.

So, I don't want to dial that back. I don't want to dial back trade. What I want to do is make it easier for people to transition to something different. And that may be a fantasy. That may be easy for me to say, because I have a good life; but I think the way we have made it hard for people to move to cities where there's often more employment from rural areas that had vibrant manufacturing sectors and now have lost those, I think it's a terrible tragedy. But we've compounded it by making it hard for people to find new opportunities. So, what are your thoughts on that?

Susan Houseman: Right. So, you're getting into a whole other line of research that takes place here at the Upton Institute for Employment Research, which is whether it's coming from trade, technology, or something else. There is a sense that change is occurring rapidly, and that the workforce needs to have continued education and skills development, lifelong learning as it were. And how do we set up systems to deal with that? So, I think that's right. I largely agree with you on that score. We can't bury our heads in the sand and dial back to 1950 when the United States was so dominant in manufacturing coming out of World War II. That's not going to happen.

We do, I think--just to reiterate something I said earlier--we do want to think about how quickly we adopt policies, or adopt policies that may lead to very rapid, disruptive change that may overwhelm regional economies and so forth, or at least have systems in place, programs in place, supports in place to facilitate that adjustment. Because as I said earlier, one thing that we do know from economic history, one of the most robust results in labor economics is that when you have large scale displacements, it can take a regional economy a generation to recover. So, if we're in a world where there is more rapid change, we need to figure out systems to deal with it.

I want to bring up two quick points, though, about trade that I would be remiss not to mention. One is that there's an ongoing debate about how much you can retain research and development in this country when you're doing manufacturing overseas.

A few years ago, that was the subject of a university-wide study at MIT. There was great concern that the loss of manufacturing and the skills would actually have quite adverse effects on our ability to do R&D, because oftentimes, not always, but oftentimes, R&D needs to be co-located with manufacturing. So, that raises big concerns about the vibrancy of our ability to innovate in this country if we don't have a strong manufacturing sector. It also raises, of course, national security concerns.

And then the second thing that is part of the ongoing debate and I don't want to put in my own opinions about this too much, but just raise it, flag it, is that not all of our trading partners, of course, always adhere to the free trade model.

Certainly, when you're negotiating trade deals and so forth, obviously, it is a concern to think about what your trading partners do--from allegations which were widely accepted in the 2000s of exchange rate manipulation to depress certain country's currencies, not just trying to butt other countries in order to provide favorable export markets for local producers, to restrictions on operating in other countries who've heard a lot about that recent years in China. Those are all other factors that we have to take into consideration, of course.


Russ Roberts: Well, I don't agree with you, of course. I wouldn't take them into account. [phone call dropped and reconnected, 1:08:57-1:09:32--Econlib Ed.]

I have to disagree with that 'of course' part. I don't care so much whether our trading partners follow free trade. I do care if they steal our intellectual property. I think that's probably a little bit different, but if they're more protectionist than we are and harm their own people at the expense of--and benefit their own manufacturers--and in turn give us cheaper stuff, I don't see that as a problem.

It could have some distributional consequences in the short run. I do think the big challenge of policy in this area is--it's a little bit weird. I forget who pointed this out to me, and I apologize. Somebody pointed out that when you integrate a billion people in China into the world trading system, there's this enormous shock, but you're not going to get a second one.

So, for better, for worse, that shock is--it's done. We did maybe make a mistake in the speed of allowing that to happen quickly. We can debate that. I don't think it's easy to measure what the full costs are. You can argue we had policy responses that were inadequate. But going forward, which is what we really care about, there is this growing unease with trade and economics change generally, and a feeling, which I think is somewhat--well, I think it's definitely overstated--that somehow the elites have rigged the system to benefit themselves from taking[?] of cheap goods, and everyone else is going to have to pay the price of not having, being able to find work. And I just think that's an incredibly dangerous populist mantra. And, I don't think it's true. So, maybe we can close on that.

Susan Houseman: Right. Yeah, certainly, that perspective is out there. One of, you know, my perspectives is, is that if we had acknowledged the effects earlier in policy that trade was having on many communities that more could have been done, perhaps. And maybe we wouldn't be in this extreme situation now.

It think that's--it's--I do--we'll go back to my original conclusion, which is: No matter where you stand, I think we did the whole conversation about what was trade doing to the economy, what were the implications for job losses, and so forth, was really stifled when so many politicians and economists were saying it's just automation. And not really acknowledging that issue. So, does that--

Russ Roberts: I agree.

Susan Houseman: Here we are today in a middle of a trade war. Do we need to get here? Most people think that's very disruptive. Maybe if we had had a calmer, more intelligent conversation and acknowledged what was happening earlier, that, that acknowledgement had been more widespread, perhaps we wouldn't be where we are today.

Russ Roberts: Well, I want to confess and say that I certainly suspect that my willingness to accept the narrative of the automation story, well, was supported by the data that I was aware of. I suppose part of my willingness to accept that was that I did not want to believe it was trade. I'm a big pro free trader.

Russ Roberts: And I want to confess that your work has really jarred my thinking on this. I'm not going to say I was wrong and you were right. What I'm going to say is I am very concerned that I was wrong and that you are right.

Obviously, measuring these changes and trying to account for the phenomena that you're talking about--your work has, I'm sure, some challenges of measurement as well, but it's deeply disturbing to me in terms of my previous narrative, and I'm readjusting. So, I deeply appreciate that.

I'm curious if you would just say how other economists have reacted to your work. What kind of response has it received?

Susan Houseman: I think it's been pretty positive. Again, I'm not--some people may think I'm going out there and saying it's all about trade. I'm not doing that. I'm just saying that the standard narrative is a misinterpretation of the data. And there's really not a lot of debate about that.

Once you point it out, it's--people who understand anything about statistics can see the problem: that you had one industry that was really driving the apparent robust growth and productivity growth; that it wasn't about automation; and that it was skewing the statistics and masking a great weakness in many underlying manufacturing industries. So, there really hasn't been a lot of debate. I've been gratified that a few economists have come back and sort of actually stopped telling that story. That's been most gratifying.

Russ Roberts: I'm one of them now. You can add me to the list.

Susan Houseman: Yeah. So, that's been most gratifying for me. But, there's still a lot of really good work, research that needs to be done to better understand impacts of trade, as well as technology on the workforce and what the policy implications are moving forward. But, we don't want to start with a simplification and misinterpretation of the statistics.

Russ Roberts: My guest today has been Susan Houseman. Susan, thanks for being part of EconTalk.

Susan Houseman: Thank you very much. It's been fun.