Did Tariffs Make American Manufacturing Great? New Evidence From The Gilded Age

>> John Taylor: Thank you so much for joining us.

>> Speaker 2: Patterson.

>> John Taylor: Chris is coming here this morning and going back to Davis tomorrow tonight. So it's quite remarkable. So thank you very much for being here and spreading your wisdom. The title of your paper is Did Tariffs Make American Manufacturing Great?

Again, you have a suggestion for a title, New Evidence from the Golden Age. So thank you so much for being here and spreading the word. We appreciate it very much.

>> Christopher M. Meissner: Thanks, John. Thanks for the invitation to come and talk. It's good to be back ever so momentarily.

Two disclaimers before I start. One is the joint work with Alex Klein. Two, as much as it may seem, the title is a little bit in the mix of role of politics, but rest assured, no one paid us to write this paper and to say any particular thing about the effect of tariffs of the 19th century.

So, just our first look at the record on tariffs between roughly 1870 and the early 20th century in the United States. So to warm up, the top line is the average tariff rate in ad valorem equivalent terms on products that actually had a tariff. You can see that never really dips below 40%.

So, United States had pretty high tariffs relative to its peer economies at the time. The lower line is the overall tariff rate and ad valorem equivalent terms on imports spit lower because over time a number of products were put onto the quote, unquote duty-free list. Middle line is representing our average tariff rate in the data we have, which is basically at the manufacturing sector.

And that goes from a high of about 40% early on and it declines towards 30%. But overall it's a high tariff economy. And of course this economy is growing quickly and productivity is growing as well. The United States is a really good economic performer at the time. And so you might ask, and it has been asked indeed before, but we're going to try to add something to our knowledge base here.

What's the connection between American economic growth and productivity growth and these high tariffs in this period? Now, we're going to try to examine this question from a causal interpretation perspective. And my view of literature, and it's broad, but my view is that, you know, very few opportunities present themselves to attempt this kind of causal interpretation.

Rourke Juhaas has a famous paper now about the blockades and the Napoleonic wars, which tries to show that that promoted the reorganization of French industry in the early 19th century. It's kind of a natural experiment. Other work that we're aware of looks at liberalizations or something like that, and earlier economic history work in the areas kind of cross-country and low N and low T at that.

So, not a lot of information from the historical perspective in our view. So we are going to treat you to an exploration of this question with an instrumental variables approach. And notwithstanding the comments from my co author who said that's a dangerous thing to do these days, we're going to go ahead and do that.

And I'll tell you how we're gonna do that.

>> Steven Davis: Dangerous to do it, dangerous not to do it. Well, also wondering, I mean, in general, tyros also have a lot of other de effects, right? And sometimes, even if they had a positive effect on productivity growth, there are gonna be distortions of consumer prices, losses of consumer surplus.

Are you gonna try to put any productivity effect in the context of the broader kind of welfare evaluation of the tariffs, or are you really just interested in the impact on productivity growth?

>> Christopher M. Meissner: Sure, this is again, thanks, Steve. This is gonna be a pretty limited study. We're gonna be looking at productivity growth.

We'll be looking at the extensive margin gross output, average firm size growth in the workers in these industries. So pretty limited. It's not a general equilibrium approach. It's not a structural approach at this point. Whether the data would allow that, it's a good question, something to think about.

So we are going to present instrumental variables approach which builds on some great work by Andy Greenland and John Lopresti at North Carolina State and William Amary. And I'll show you how we do that momentarily. But basically because of the way tariffs were structured at this time allows us to get a little leverage on exogeneity and the tariff level at the industry level.

So we're gonna go for that. Just before the barrage of questions begins, I'll get my bottom line in. And really it's that there's no evidence of a positive relationship between tariffs and labor productivity at this time in the American economy. Despite what many might offer as opinion, the data don't support the idea that tariffs promoted productivity levels or growth as we will see.

>> Speaker 5: Negative relationship or are you saying there's no relationship?

>> Christopher M. Meissner: Well, we'll see if you buy my instrumental variables approach and if you do, we will find statistically significant negative effect in a single equation system and then we can go from there. But the safest thing is to say this, sure.

You might ask how tariffs matter and probably a lot of ways they matter. And just thinking about it, might think that tariffs protect industries in their quote unquote infancy when they're learning how to do things. You might think that rent seeking and lobbying and politics plays a role, and we're gonna find some evidence consistent with that fact in this period.

And limit competition if they inhibit imports, which they may or may not do depending on the elasticities. Doug Irwin's always been a fan of the idea that an import tariff is also an export tax due to what he calls what is called learner symmetry. So that possibility, and Steve already mentioned welfare and the possibility that prices go up on imports and other products as well in general equilibrium.

Hurting welfare of consumers at the helping producers at the same time, but hurting consumers. So the data we're gonna bring here are new historical data which is basically hand collected. It's pretty granular, we're gonna have sic 3 digit industry data at the state level over a number of years, census years.

We have tariff line data which is ultra detailed I would say in historical terms and maybe even by today's standards. And those data, official trade data we're using also allow us to measure the tariff rates both in ad valorem equivalent terms and just we have the statutory data too, but not gonna make great use of that.

>> Steven Davis: There are many other relevant policies at the same time that affect the manufacturing sector, such as non-tariff areas or production subsidies. Could be a matrix of other policies that might. Correlated with tariffs across the industry.

>> Christopher M. Meissner: Yeah, I think at this time it's pretty safe to say there aren't a huge number of industrial policies or interventions by the government.

There are a few non-tariff barriers, some of which we discuss and they come here and there. Shipping was heavily promoted by the government and protected. But other industries, not so much directly like we would think of in the modern period, I would say-

>> Steven Davis: Data on policies and our trading partners?

>> Christopher M. Meissner: No, yeah.

>> Steven Davis: And international capital flows and exchange rates?

>> Christopher M. Meissner: Somewhere in the world these exist and somewhere on my laptop they exist, but not in this paper. So I mean it's worth bearing in mind the US pretty closed economy. Import and export shares of gross production are low in the range of 3 to 7%, so not very high. But in certain industries-

>> Speaker 6: Margin is probably higher.

>> Christopher M. Meissner: Indeed, yeah, so we're gonna take the US just in isolation in this paper.

>> Speaker 7: Why do you think tariffs are usually levied on imports rather than exports?

>> Christopher M. Meissner: Well, sometimes they are levied on exports. In the US case, I can't think of too many examples of that in this period.

It's a good question. I don't know the answer off the top of my head, but you could start to think about who the winners are, the corporate interests, the business interests. Who the losers are, the consumers, collective action problems and that sort of thing. But I don't know if that gets to the precise question you had asked.

But I don't think it's relevant at this point in time given the historical record. So we'll move on. As far as I could see, mostly just import tariffs, right? So what this project is part of is a bigger project to collect data. Data were mentioned. We do have also on this hard drive that I'm clutching here with my version of the presentation.

Over 2 million trade lines and bilateral product level data waiting to be utilized, not utilized in this paper. And then an NSF grant with multiple CO PIs across the country, looking at different periods. But collectively we're putting together every tariff line in the history of the United States all the way up to the 1970s.

We're gonna meet up with the modern trade people's product level trade and tariff data for the US. So in about a year or two or three, maybe you will be able to access this database at the tariff line, beginning of the country, all the way up to present, I guess.

>> Speaker 8: Historical gravity model?

>> Christopher M. Meissner: No gravity model in this paper, this is gonna be just about product. In the bilateral stuff, lots of fun you can have with that checking on elasticity. Checking on the emergence of different industries and consumption patterns. That's kind of all in the background.

And the tariff, I'm sure that we can find lots of ways to get in trouble too, but we're looking forward to putting that together. Okay, so it's kind of stylized facts. It's not a very open economy. Beginning of the period, the United States is pretty much an agricultural commodity exporter.

Main exports being cotton and wheat, several other commodities, and certainly one of the net exporter of manufactured products. But by the end of the period, that is no longer the case. The US is now in that export of manufactured products. Our degree of elaboration varies, but eventually it will become a world manufacturing powerhouse by the first few decades of the 20th century, and it's well on its way here.

Diversifying its product space, its partner with whom it trades or with which it trades. And despite this all, it's behind a very restrictive trade policies. You've seen the tariffs, tariffs are comparatively high. US doesn't like to sign trade treaties at this period, even though the Europeans are promoting trade treaties, especially around the 1870s period.

Many sectors are protected. And there's a partisan political thing going on in the background here. You probably recall, generally speaking, the Democrats were in favor of free trade or lower tariffs. The Republicans were generally aimed at protectionism, although even Grover Cleveland vetoed lower tariff bill in the early 1890s, himself a Democrat. So it's not as clear cut, yeah.

>> Kenneth Judd: Now, this partisan Democrat versus Republican, how important is it that it's south versus North? The Democrats controlled the south at the time. So the Democrat versus Republican labels are not terribly informative, things change. But was this a North versus South kind of thing, that the South wanted free trade and the North wanted protection?

>> Christopher M. Meissner: Yeah, largely, but we got to consider what's called the west at that point or the Midwest. And there we see some variability and emerging manufacturing sectors around Chicago and other places.

>> Kenneth Judd: I just think that'd be a more informative-.

>> Christopher M. Meissner: Yeah, sure.

>> Kenneth Judd: Description of the political-

>> Christopher M. Meissner: Sure.

>> Kenneth Judd: Preferences.

>> Christopher M. Meissner: Sure, yeah. I mean, there's a lot to say there, right? I mean, not everybody's in the industry in the North and not everybody's an AG in the South. And so, their sectoral interests and the parties kind of line up geographically like you say.

But for the most part, it is a partisan issue talked about in this way. We're building off the work of Doug Irwin, who's done massive amount of work on US trade history. And has come to the conclusion that tariffs were very unlikely to promote aggregate productivity growth in US manufacturing.

Now Doug's evidence is not based on micro level data. It is more kind of looking at the aggregate outcomes and reasoning from first principles, I guess, if I'm being fair to him. I hope I'm being fair to him and I'm not. If he's in the Zoom room or wherever he may be, he might stop me.

But this is what I got from reading his body of work, yeah.

>> Speaker 10: To your knowledge, has anybody written a sentence which just says we're not an important factor, we're irrelevant to, the growth was so strong you can't draw any real conclusion? Has anybody written any sentence like that in the whole literature?

>> Christopher M. Meissner: Yeah, I mean, Doug Irwin, one of his famous lines, if I remember correctly, was, The US grew despite tariffs, or something like that.

>> Speaker 10: Somebody, the high import tariffs were irrelevant. It was not a negative factor in driving late growth.

>> Christopher M. Meissner: Right, well, he also-

>> Speaker 10: Just refuting that proposition.

>> Christopher M. Meissner: Refuting, I mean, Doug has himself, I think, argued that in any case, US wasn't very open. And so how would tariffs matter if the international economy was such a small part of US activity? But as might not be on the margin, and particular industries might reveal more interesting dynamics.

So maybe in the interest of time, I'll skip the literature review but we can talk about it as we go along. Very little has been done at the disaggregated level in US economic history as far as we are aware. Maybe we'll go back one, and there's been some general equilibrium modeling by Yoon in particular which calibrate a general equilibrium model using aggregate data.

Came to the conclusion largely consistent with ours, or vice versa, that tariffs didn't really matter for productivity. They might have brought some resources into the industrial sector. Learning by doing wasn't a very big factor either in this period. So that's nice to have there, we're gonna go more disaggregated than his work.

We're gonna build a lot off Greenland and Lopresti and what we do in our empirics and I'll tell you about what they did and how we modify their approach in just a moment. Our baseline estimated equation is going to look like this. We're gonna be interested in the level of value added per worker in industry k, in state s, census period t, so it's a panel, an unbalanced panel.

We also have information, obviously, on worker gross output establishments and average establishment size measured as output per establishment. We're gonna relate something on the left hand side in the first case labor productivity to measure of tariffs. So we're gonna take the log of 1 + AVE. AVE is defined as the ad valorem equivalent of tariffs, and we'll call that the log of gross tariffs.

We're gonna lag that variable by 10 years, which is a census period. So our outcome will be, for instance, in 1880, and our main control variable will be lagged 10 years. And that's pretty consistent specification with the growth literature that we're familiar with. And then we're gonna have a set of fixed effects in these models.

One of those fixed effects mu is going to be industry state fixed effects and then the second of which is going to be census period fixed effects in an error term.

>> Speaker 11: Why do you just disaggregate by state when the variation driving variation is national?

>> Christopher M. Meissner: Why not?

Why not? There's a treatment out there, and some are going to be more exposed than others. We'll see how it works out. But I mean it's not an uncommon approach I think in the applied literature to use a national policy or a local policy that covers many workers, for example, use the workers as observations.

I don't think that's breaking any universal rule. We're gonna take care to note that our observations may or may not be totally independent and we'll have to deal with different correlations.

>> Speaker 12: But you're not gonna say that something in the middle of the country is less exposed cuz of market access or something.

>> Christopher M. Meissner: I'm gonna go in that direction, but it's something worth thinking about. But we want to have a good way to operationalize that. So, okay, here's the meat of the procedure and forgive me for going into this, but I think it's worthwhile to see how we do this.

This is building off the work of, as mentioned, Greenland & Lopresti. We're gonna modify their approach just ever so slightly to make it work for us. You can imagine, first, ad valorem equivalent tariff of a product c. We're gonna start with product data and build up to the industry here.

So that's where we're going. So in the upper left, we have a formula, an identity, really, for the ad valorem equivalent of a statutory tariff on a commodity c in any given year. And that's the sum of two parts, the ad valorem tariff, which may be expressed as kind of like 25% for this product.

And then there's a second part, which you may be less familiar with, which is a specific tariff. And that is levied in terms of dollars or fractions thereof on a physical unit of a product c. So f will be something like $0.05 per gallon. The ad valorem equivalent of that specific tariff obviously depends on the unit value of that product, or the price, P.

So if the price goes up, then the ad valorem equivalent goes down, and vice versa. So really that's the heart of our kind of instrumental variable there, that price. Those prices are gonna be able to, as we'll see, derives from the variation in tariffs and allow us to identify this thing.

So moving forward, the STS variable is going to be the share of tariffs in specific revenue. So each product, to the extent it has or has not a specific tariff, will have a share of total revenue collected from specific tariffs that ranges between 0 and 1, okay? And you can rewrite it like this and you put those two together and an approximation to the change in the log of the gross tariff is given by three parts.

The unit price change, the initial share of the specific tariffs, and a ratio of the ad valorem equivalent of tariffs in the leg period, yes?

>> Steven Davis: I'm wondering, I mean, related to where you're going with the instrument. I was just wondering about the kind of political economy of tariffs at the time.

What was the sort of stated justification? Were they viewed in terms of sort of enhancing productivity growth or national security? And then also the decision to do specific versus ad valorem and whether to change specific right? Or that there's gonna be a looking at the economy behind that process, right?

>> Christopher M. Meissner: Yeah, good questions, all relevant. So there's not really a national security thing, except for in shipping, as far as we know. So the shipping industry is heavily protected, but it's not really gonna be in our data. So restriction is the word of the period according to Doug Irwin, who's written the thousand page work on this.

And the idea is restrict international competition. You wanna think of a period where industry is pretty heavily concentrated. Mentioned the Gilded Age already. And this just increases and helps those business interests reduce competition. So restriction is probably the first thing to think about. How do you decide whether to have an ad valorem or a specific tariff or some of each?

Well, there's lots of reasons, and this is gonna be crucial to our identification strategy. Older work by Taussig and others who looked at the history of US Tariffs suggested that specific tariffs were a solution to the problem of undervaluation at the border. So if you brought in a boatload of silk cloth, you, you might tell the border guard, well, you could pay them off and have them look the other way.

Or you could just say, well, this is only a $5 boatload instead of 100,000 worth of product and the tariff is going to be much lower. But if you can count the product, you can just. Levy the tariff on the physical volume, with the repercussion that the revenue collected will vary by the price of that product.

Actual protection an industry gets is, it's going to be variable, and that is somewhat known, I think. But that share, we will argue isn't correlated with very much and is established in early part of the Civil War with the moral tariff. And it's pretty sticky over time, and we're going to argue it's predetermined, and not something to worry about too much.

>> Speaker 13: Are your time periods going to be decades that match the decennial sense of manufacturing?

>> Christopher M. Meissner: Right, decades, yes.

>> Speaker 14: So we've got two depressions in here, and you've got argument about bimetallism going on.

>> Christopher M. Meissner: Yeah.

>> Speaker 14: How does all that play out?

>> Christopher M. Meissner: Okay, so we do have a period of price declines, Greenland-Lopresti, for example.

Look at 1900 to 1940 where the price swings are pretty massive and aggregate for many products. But our price declines and changes aren't going to be nearly as big. But price movement, to the extent it's driven by monetary force, is gonna be helpful. But we're gonna be looking at the product level.

So we're going to look at idiosyncratic changes in prices and something to worry about when we go back and estimate this thing. Let's think about reversion and have a specification which kind of deals with that. But is there another question?

>> Speaker 15: I was just gonna mention that. Protecting high wages was an important part of McKinley's rhetoric for the tariffs.

There was an interesting book by Karl Rove about the McKinley campaign strategy. People might find that interesting. He was reaching out to this immigrant working class Catholics mainstream. How much clout that line actually had? I don't know whether it was there.

>> Christopher M. Meissner: Yeah, I mean, I guess, along with the high prices came high wages, which leads went to infer that import taxes and export tax, and in general equilibrium, everything just gets bid up.

So that's part of the story, for sure. Thanks for that, yeah.

>> Speaker 16: Do you have information on revenue collected? How important is that, relative to total revenue?

>> Christopher M. Meissner: Like almost 100%.

>> Speaker 16: Yeah.

>> Christopher M. Meissner: There's a few excise taxes. There's no real income tax in this period, so-

>> Speaker 16: Land sales that weren't-

>> Christopher M. Meissner: Yeah, but this is like, I've got the number, I think, in the paper I looked up, it's well above 90%.

>> Speaker 16: Didn't we use to give away land?

>> Christopher M. Meissner: Different things happened. Different things happened. But yeah, this is the vast majority of American government revenue. And so, one of the drivers of tariff revenue or tariff policy, as well as the federal budget situation.

That's noted by Taussig especially in the 1870s coming out of a financial crisis, where government revenue really lags and they gotta do something about it.

>> Speaker 17: Paying off the debt in the Civil War, too.

>> Christopher M. Meissner: There's also that.

>> Speaker 17: Paying the veterans.

>> Christopher M. Meissner: Yes.

>> Speaker 15: Wouldn't that be a critical element in thinking about why you have tariffs and it changes?

>> Christopher M. Meissner: Sure, it would, I think that explains the aggregate level of tariffs, the average level. Like we're gonna be using inter industry variation across industries and over time. So I think there's something left there even after we think about the overall level of tariffs. So there will be substantial temporal variation, cross industry variation here where the lobbying and maybe other things come into play here as a factor.

Yeah, so there's people back here, I can't see if you put your hands up, but just jump in, right? So, what we get from the middle line is that there's a component here which we think may be an exogenous driver of average tariffs and the ad valorem equivalent anyways at the product level.

And we'll call that RP, for realized protection. And the change in realized protection is going to be a function of the change in the price of a product. So log change in the price of a product, and how much that product relies on specific tariffs. We don't rely on specific tariffs if STS is 0, RP is not affected by unit value changes, that's fine.

But then that's kind of like a control group if you want. And then there's a treated group and they're treated. The dose depends on how far they go in terms of specific tariffs. Some industries have full on specific tariffs, some are more compound and have both kinds of ad valorem and specific tariffs.

But the key driver here is gonna be the unit price change. And keep in mind, these are import prices. So we're gonna view the United States as a price taker in these products, that's gonna allow us to leverage as price changes is exogenous. So that's one of the assumptions we'll make that essentially this instrument that you're looking at, RP can be driven by these exogenous changes in unit prices.

And second, that the shares are probably also pretty exogenous and determined by these considerations about fraud at the border. And factors that were relevant in 1861 when they reinstituted specific tariffs and the Morrill tariff to help fund the Civil War.

>> Speaker 18: It could be driven by common technology shocks or common price shocks, right?

>> Christopher M. Meissner: Yeah, I've heard that one before, that's possible. But you know as well as I do, yeah, no, it's something that kept me up at night for a while. But when you read the literature on different industries, and you know as well as I do, probably, right? That industries use different techniques on this side of the Atlantic than they did in Europe.

Technology wasn't always translatable. I mean, you just pick up Clark's work or Taussig's work, and you just find that the technologies had to be adapted to American circumstances and relative prices. So I think it's a little bit less of a concern. But if you're really, really worried, then take STS as exogenous.

And I think the econometricians tell us, we're good if we have one of these, at least, exogenous. If both are, it's great, too, but just-

>> Speaker 18: And goods that were not subject to ad valorem tariffs. What gave rise to reasonable quality data on import prices? We needed to level a specific tariff or no tariff, so.

Well, I'm just wondering why we should trust the quality of the import price data.

>> Christopher M. Meissner: Sure.

>> Speaker 18: Aside from the ones that were subject to ad valorem tarrifs?

>> Christopher M. Meissner: Sure, that's an interesting question. I mean, those are on the bill of lading when it's the price at the port before they sent the stuff out to the US, doesn't include any freight or insurance or anything.

And that's the declared value. Yeah, I don't know, really, the answer to that one. We can look into it a little bit more, once it's levied on a unit basis, I don't necessarily see. And the tariff's fixed, I don't see a reason for the importer to engage in fraud.

But possibly-

>> Speaker 18: These prices only come from custom authorities. Is that where they come from?

>> Christopher M. Meissner: Well, they come from the,

>> Speaker 18: Somebody bought.

>> Christopher M. Meissner: An exporter would declare the value and the shipper would, as I understand it, hand the bill of lading to the customs person that would see the declared value and unit values of these products, yeah, yeah.

So okay, so RP is essentially gonna be the building block for an instrumental variable. And we're not quite there, we got two stages instead of, well, we have three stages instead of two. And so in our like initial stage we're going to take that middle, that first equation there in the middle of the slide at the product level.

We're gonna try to predict the log change in the ad valorem equivalent of a tariff at the commodity level using that RP, that exogenous variation. And we're gonna include time dummies and that's about it. And then we're gonna use the information from that regression, basically, the predicted change in the tariff.

We aggregate up to the industry level, so take all the commodities and industry k to get a predicted change in the log of the gross tariff. So in the spirit of kind of GMM, we're gonna be using kind of changes to predict levels. And if you go back to our estimating equation here, we're gonna predict that level of tariff with predicted changes.

That's basically the instrumental variable strategy, and then we're going to get the data from official sources. The United States had pretty detailed and rich data. We're gonna have access to that. We collected data because it's so voluminous at 10 year intervals to match up with the census periods, 1870, 80, 90 up to 1900 for the trade data.

That trade data includes import dollar value tariff revenue quantities, the statutory tariff rates and item descriptions which are quite complex sometimes and a bear to deal with. So we've been cleaning that for a while. Looks like this just for fun this morning. If you zoom into the middle, cotton and manufacturers of, there's some pretty complex textiles descriptions here.

So this is highly detailed data in some instances, in other cases not so detailed. And you can see in the second column the rate of duty. Some of those cotton products, two cents per square yard, those are specific. And then some have compound duty $0.03 per square yard and 10% and so on it goes for well over a thousand lines per census year.

Okay, what do the data look like? There is inter industry variation in 1870. Maybe you can or cannot see it. Paper is the least protected according to this tariff measure. Tobacco is the highest, it's pretty surprising when I looked at this graph and I thought of what's going on in tobacco.

Which is over time essentially monopolized by Duke and you're like why would he need a tariff that's almost 100% that he's so productive? Produce a lot of cigarettes at low unit cost. Why would you need that? Well, more is better I guess.

>> Speaker 18: As much you're waiting for your answer.

>> Christopher M. Meissner: I don't know, I looked up that, I tried to find a good answer. I didn't find much about this operation. I just, yeah, I mean there is that, right? Thank you for that, right? I mean, tobacco and alcohol and some of these other more inelastic products would have gotten hit.

But it's all too tempting to think something going on with political power here. Other industries are,

>> Speaker 19: Free rider problem.

>> Christopher M. Meissner: All going to one. Yeah, so that's kind of just emphasizing here that there's inter industry variation and if you look real closely and squint or get the paper, you'll see that industries move up and down the ranks.

These are two digit industries but we're actually gonna be looking at the three digit. So there's even more variation.

>> Speaker 20: That's right, we had 100% tariff would mean doubling the price that an importer would have to pay twice as much as the export.

>> Christopher M. Meissner: We have some information on inputs, ask me about it later if we have time if you're interested, I won't dwell on that.

The manufacturing data comes from the census of manufacturers. After cleaning the data we're gonna have about 80 three digit industries in an unbalanced sample. Work with value added output, value added per worker. So labor productivity, gross output, number of establishments in an industry, number of workers. To deflate the output data, the best we could do at this time is two digit deflators we have to apply those to three digit industries.

Was not ideal but that's part of the problem in economic history I guess. But I think it works okay.

>> Speaker 20: Do you know anything about plant and machinery or capital and so on? I'm thinking this is an error of kind of transition from water to steam power and potentially a lot of changes in capital stock as well.

>> Christopher M. Meissner: Yeah, the sense of manufacturers does have other information, right? It does have a measure of capital, it does have a measure of materials is useful for getting value added. We use my co author Alex, I'm gonna take to be one of the leading experts on these sense of manufacturing data suggests we do not use the capital stuff.

Other people have different views and have used it before. You get mixed reactions when you think about TFP. And I've heard from experts we may not want to do that and so we're not going to do that. We'll just look at labor productivity. But people have done it and one could do it, but we don't wanna go there, maybe referee two will ask us for us, but

>> Speaker 20: There are all sorts of problems measuring the capital stock and it even works back then.

>> Christopher M. Meissner: That's right, yeah, it's hard, right? So I'll put up the summary statistics. Some people look at the very first chart I put up and said there's no time variation in your tariffs. So get out by faith case, we disagree. We look at the standard deviation of the log change in the gross tariffs.

It's not huge, but the second column at the bottom row, about nine log points. So it's nine significant, there's an interdecile range there too. And so there is a pretty substantial variation given the average is something like 27, 30% ad valorem equivalent. Okay, let's look at some results.

That initial stage predicts the change in tariffs at the product level, pretty well realized. Production is a good predictor of changes. So price variation is gonna predict the change in the log of gross tariffs at the product level, yeah.

>> Speaker 19: Your percentage points on the horizontal scale.

>> Christopher M. Meissner: Yeah, yeah.

>> Speaker 19: Okay.

>> Christopher M. Meissner: Well, no, it's a log change, right?

>> John Taylor: So minus 30 to 20 or is it 3 to 2?

>> Christopher M. Meissner: Yeah, minus 3 would be a gigantic change. A big swing in your unit value, right?

>> John Taylor: So, right, I'm trying to get the sense of the scale.

>> Christopher M. Meissner: Yeah, well, the bulk of the, the distribution is going to be between minus one and one and you know, those are even still.

>> Speaker 20: One log point or 100 log points?

>> Christopher M. Meissner: 100 log points. Yeah, so those prices are highly volatile.

>> John Taylor: Okay.

>> Christopher M. Meissner: Yeah, but this is the product level.

And if you have been paying real close attention, you'll notice, or if you look in the paper, we aren't only gonna do this like it's a repeated cross section. We don't have a panel of products over time. We have to do this for every year. And the reason we wanna do that is because the product is exactly the same one year after the other in the tariff data.

But if you go longer than that, it's gonna be hard to match the product.

>> Speaker 21: Almost a three log point change in price makes me worry there's classification there.

>> Christopher M. Meissner: Yeah.

>> Speaker 21: Depending on how well you have harmonized.

>> Christopher M. Meissner: Well, some of these things are commodities. They come from different places, different shocks hit.

I mean, the price of salt might go up by 200%. It might, I don't know, it depends. You could count these up, there's only 30 or 40 commodities out of 1500, right? So, yeah.

>> Speaker 21: You should just show us what those extreme examples are.

>> Christopher M. Meissner: Okay, yeah, I looked some of them up and was thinking the same thing.

But some of them are just, I mean, you import stuff from Britain and then you import it from, I don't know, Southern China. And maybe it's totally different quality of good this year you didn't get the British stuff, you got the South China Sea stuff. It's totally different quality.

Things like that can happen, but they all have the same tariff and the realized protection would change. Okay, so just kind of looking at the raw data on our outcomes, final stage, just relating labor productivity to tariffs here. Maybe the x-axis scales a little bit more manageable, ad valorem equivalent of tariffs ranging from about 20% to 0.2%.

Once we control for all this stuff, we got a bunch of fixed effects in here. Okay, so that's why we get negative tariffs. But the partial relationship is decidedly downward. And we can take that to something a little more scientific and get you standard errors and do something more conventional.

A little hard to see here, but OLS regressions on the top IV in the middle and IV with extra fancy fixed effects on the bottom panel. Those extra fancy fixed effects are state by year fixed effects. And in the middle we'll just have industry by state fixed effects and census period dummies.

Okay, so LS provides a negative coefficient in the first column and that's minus 0.34. So the negative impact of higher tariffs on value added per worker labor productivity, once we run the IV regression, that effect is much, much bigger, which is troublesome and worrying. But if you size this, right, what you're getting is about a one standard deviation decline in tariffs would lead to about a little bit less than 1 standard deviation rise in labor productivity.

So it is definitely sizable. The interesting thing to note here, right, you might ask why is OLS so biased downward? I think we've convinced ourselves that one plausible explanation here is lobbying. And the story goes like this. OLS provides a biased estimate of the impact of tariffs on productivity.

If the true effect of tariffs on productivity were to be negative, right, and then the bias, right? So you think about what the bias is adding by not doing IV. Well, the bias is a two part term, a product one is a function of tariffs related to what we might call capability to get high tariffs.

Big companies, highly productive companies, fast growing companies, our industries might get higher tariffs. And then the correlation between capability and tariffs is there. And the second part is also positive, which biases the OLS up towards zero. So that's our story on that. And it seems pretty consistent with the historical literature that says tariffs were driven in large part by lobbying of powerful industries.

So our results are not inconsistent with that. Going across the row in the middle to look at other OLS results. Tariffs bring down average establishment size. They improve or raise real value added. They raise but not statistically significantly gross output, real gross output in the industry. They bring in workers and are associated with a larger number of establishments.

So it's almost textbook kind of theory here. A higher tariff would bring in resources to the domestic producers, raising domestic outputs and got to have workers to produce this stuff. So they bring in workers. And at first blush, it seems consistent with kind of a Melitz's view of the world, where if firms are heterogeneous in terms of their productivity and we protect the industry more.

That on the margin we're gonna be bringing in less productive or smaller firms. And that's consistent with our results in Column 2 there, where tariffs bring down the average establishment size.

>> Speaker 22: Surprise at the difference between the OLS and the instrumental regressions is so large and often even the opposite signs.

And, I mean, you gave an explanation, but might be nice to really tease that out and see where the endogeneity, I mean, does it make sense? Can you see that in the data?

>> Christopher M. Meissner: Yeah.

>> Speaker 22: In some other way, I mean, it's huge, right? And usually one likes the IV to kinda correct the OLS a little bit, but.

>> Christopher M. Meissner: Yeah, it's pretty big. I don't know how much further we could go in terms of this identification strategy or what we could find here, but it's worth thinking about more for sure. Manufacturing data from the censuses? No, that's not gonna be available.

>> Speaker 19: It's okay, if that's the case, what I'm thinking about is that different parts of the country had access to railroads at different times.

And under some stories you would expect to see the lower portion of the distribution of labor productivity disappear once you had a railroad that serviced that region. And that's essentially a story about competition. It's not tariffs, but it's kind of the same.

>> Christopher M. Meissner: Yeah.

>> Speaker 19: It's kind of the same mechanism that could be in play.

>> Christopher M. Meissner: Yeah, yep?

>> Speaker 19: You could get big productivity effect from the extra competition through that mechanism.

>> Christopher M. Meissner: I agree. Thanks for bringing up Rick's data. We don't have it. The manuscripts do have establishment level data, but it's voluminous and very, very complicated to digitize it. And Rick and others are working on that.

But we don't have it. But it would be fun with unlimited resources, and maybe the world of AI will bring us into that stage where we can come back 10 years from now and answer that question.

>> Speaker 23: Part of the distributional consequences of the tariffs as well, in terms of industry wages or measures of quasi rent, that would also be the other classic prediction, right, of a specific practice type story?

>> Christopher M. Meissner: Yeah, we do have a way to Bill in this data, we haven't made great use of it. Small complication in the data, the workers are, later on there's workers and managers. So they have two parter there. And then early on it's just like workers, I think. So we stick with workers all the way across and the wage bill only gives us an aggregate for all kinds of workers.

It's kinda hard, but maybe those data are out there in more disaggregated form in the manuscripts too. But, so we don't have any information on that.

>> Speaker 22: So imagine lowering tariffs, right? And imagine you have an industry that is kind of, you know, monopolistically competitive, let's say. So there are less productive and more productive companies.

Now you open that industry to trade, right. And that means that all the less productive firms and all the workers lose their jobs and so forth. It's true that productivity for the remaining companies goes up. But for the workers that have been displaced, that would be bad news rather than good news.

>> Christopher M. Meissner: Yeah.

>> Speaker 22: This one suggests the opposite result, right? You open it up to trade, then all of a sudden all these workers become more productive, isn't that wonderful? So I'm just trying to think of how to think about the results that we're seeing.

>> Christopher M. Meissner: Okay, I think if you believe these, I think you're thinking that tariffs increase concentration and reduce competition and reduce innovation and make workers less productive.

Brings resources in, yes. And I think the link between market concentration or industry concentration and tariffs is ambiguous. There's a paper we cite which suggests that tariffs go up and there's greater concentration, then you get higher productivity. But I don't see that argument. I would think oppositely.

>> Speaker 22: It's closing down less productive firms.

>> Christopher M. Meissner: If they were to liberalize-.

>> Speaker 22: Yeah.

>> Christopher M. Meissner: I think so. I think so. I think that's consistent with our argument. We have a lot of case studies at the end where we talk about different industries and we'll see if we get to it, but-

>> Speaker 22: What is it that you bring in?

I don't know, tools from abroad that make these workers, stuff in a better way that they couldn't do before. I mean, these are two very different stories. And I think what's important to know what's going on.

>> Christopher M. Meissner: There are legends that the tariff system at the time, you know, lowered the relative price of capital.

Work by Jeff Williamson suggested that maybe not a lot of stuff. We imported some in the import data I found, but it's not very specifically defined and matching it industry is difficult. So I don't know how far we'd get on that. What I will say, the woolen industry, for example, gets a shock of liberalization in the 1890s.

And despite super high tariffs prior to that, imports were pretty high because we just couldn't compete with the quality and goodness of European producers. And after this moment of liberalization in the mid-1890s, I have a working paper on this. But it's in early stages and the imports just don't come back after that period of liberalization, which is consistent with, we got a little shock here.

We learned how to do things better and there's more competition now and we learned how to survive. But other stories could be told, I suppose, it's not locked.

>> Kenneth Judd: Could the growth of unions is starting in about the 1880s have any effect? They would want greater tariffs.

>> Christopher M. Meissner: Yeah, it's complicated story.

I mean, in the period, tariffs were a bugaboo. They were associated with concentration and monopoly and the populace aren't gonna be too down with that. So, unions, to the extent that they have some power, maybe, but maybe in play quite yet.

>> Speaker 24: Some numbers between 1870 and 1900, federal debt fell by half.

>> Christopher M. Meissner: Yep.

>> Speaker 24: Okay, so through that whole period, the history of the country was a balanced budget norm. And after every war we paid it all back down. So that debt as a share of GDP fell from about 10% in 1870 and ended up at about 6% in 1900, that's nominal.

>> Christopher M. Meissner: Okay.

>> Speaker 24: So really, the government was rolling in dough. No, seriously.

>> Christopher M. Meissner: Yeah, sure, yeah.

>> Speaker 24: And when you think about it, rolling in dough, you're not going to worry too much about getting every little item taxed, you know, just not that important. The main point is, is that the numbers are overwhelming spending, so there's no pressure.

And 90 to 95%, rarely less than 90, as high as 95% on custom duty per your share of federal revenue. So just overall growth was so strong, sucking in imports, that even with lower rates and all that, it's just, you've got this remarkable period of 30 years that the economy is just roaring.

You know everything is roaring and so forth. So I think under that situation, slippage, I'm not sure what do I want to say. These numbers strike me as relatively small, but there may just be rounding errors in the bigger data picture and, and you can't really make too much of them.

You've really dug down here, but at the end of the day, you're talking about 1 or 2 or 3% of what's go. I once asked chat TDP how many variables drive economic growth. And I said it was 25, I said it was 25 more, and I could have gone on.

>> Christopher M. Meissner: Yeah.

>> Speaker 24: 40, 50, and how many of those are in there in this period that you don't know are really overwhelming the system? So that you're missing variables may be way more important than getting at productivity value added. So you have some fixed effect. But it's a tough business to deal with economic growth.

>> Christopher M. Meissner: Yeah, it's tough. And going down the instrumental variables path is even tougher. But hopefully we get some exogenity here and we want to look on the margin and we're not general equilibrium territory yet and we'd like to be later, but we're not quite there yet. And so, it's a good question, right?

But the budget was on people's mind in the 1890s, a pretty severe financial crisis and turmoil. And Congress has called back and revenue is gonna be a concern in this early 1890s period cuz banks are failing and the currency is in question, which is real tough.

>> Speaker 24: You're going to have this.

But taking looking at it decade by decade-

>> Christopher M. Meissner: Hindsight is amazing, hindsight is amazing, but I don't think they had it at that point, so it's tough. Okay, so where there is a bright spot in terms of thinking about benefits of tariffs maybe is when we look into heterogeneity.

And in particular, if we look at some special industries that we might classify as second industrial revolution. So we think about advanced chemicals, consumer durables, bicycles and others. These are emerging industries. I don't know if I would call them infant industries, they're emerging industries and they're emerging everywhere.

But so it's not quite an infant industry story, but here the evidence is that in riv regressions, tariffs are more strongly associated with positive productivity. And here, the effect is ginormous. The difference between OLS and IV is even bigger and one standard deviation rise in tariffs gonna give a two and a half standard deviation rise in labor productivity, which is hard to stomach.

But there it is, and the results on labor and establishments and average establishment size are all very mixed in the IV regressions. It's a much smaller sample, IVE doesn't work as well. The F stat is not as high. So I don't know what to make of these results but we checked for it.

>> Speaker 15: Getting back to the political mechanism, are there data that enable you to see how tight the correlation was with labor productivity growth and wage growth at the industry level? For sure that's what you're doing.

>> Christopher M. Meissner: Yeah, well I don't think we have the good wage data that you know, you would want to do that study.

We could double check that. I could double check that. But again I think it's a little mixed up between the. I think it's a little mixed up between what we're actually getting and the changes over time. So we want to be worried about that.

>> Speaker 15: What's causing people to support high tariffs, or higher tariffs?

>> Christopher M. Meissner: Well, I don't think people, I don't think-

>> Speaker 15: 96 could have been a lot about views about tariffs and wages.

>> Christopher M. Meissner: Yeah.

>> Speaker 15: And obviously, this big junction between how manufacturing is fair and how farmers are faring.

>> Christopher M. Meissner: Yeah, typically I would have thought of William Jennings Bryan and the contest between him and McKinley as a monetary thing and about greenbacks and gold backing.

But certainly industrial concentration and was on people's mind as well. So, but I don't know if people reaction depression and farm prices. Yeah, sure, 1890s unrest, yeah. So, my view is right. People aren't necessarily supporting getting their say here, that it's all done in Congress and they're concentrated lobbying interests.

So it's hard for people to express their opinion as well as you might like at the ballot box. But that 1896 election is interesting one I think. Interesting and possibly strange thing we found when we went a little bit further here and maybe you can tell us your opinions.

I'd love to hear it. We decided to break the regressions down by two digit industry and related two digit industries. Sometimes there's a couple in there. So we go textiles, wood, metals, instruments, consumer goods. And we have another two pages here. And the results on productivity are largely a bit mixed actually often finding negative coefficients on tariffs and productivity.

However, one odd thing here, I would say thing that we haven't quite put together is that here we have two digit industries. Or groups thereof and almost every single extension margin coefficient is positive and highly significant here. So tariffs in these five groups bring in workers, they raise output and bring in establishments.

But, we didn't choose to do it this way. This is just what fell out of the data. We go to the next group of three and it's basically all negative on the extensive margin. So in these industries, tariffs reduce the number of workers, reduce the number of establishments and reduced gross output by and large.

Not every time, but that's the signal we're getting from this group of industries. And then there's a final group where there's no statistically significant relationship. My first thought was well, it's got to be related to industrial concentration. It's not totally obvious here. Petroleum is in the top group in this last table, highly concentrated.

But chemicals maybe aren't glass, I don't know. Machinery and electric goods, I don't know. Food and tobacco, highly concentrated, definitely paper and publishing. I've got a little interesting story to tell at the end, if we get there about that batch of industries. Transportation also got a little interesting anecdote towards the end here.

But not necessarily highly concentrated coefficient very stable across those different groups of industries.

>> Speaker 25: Some kind of non monogeneous there or something about the first stage.

>> Christopher M. Meissner: Yeah, interesting question. I should look into that more carefully, I don't know off the top of my head. I mean the strength of the instrument is pretty good throughout, but the exact coefficient and the first stage.

Yeah, I'll look and see.

>> Speaker 25: Could tariff choices by the exporting countries have something new with that? So, I mean imagine one industry where the US raises tariffs, but the other country doesn't cause this industry, where they know the other country retaliates and it's trade of heterogeneous goods.

So that could explain some of the differences and the reactions.

>> Christopher M. Meissner: Yeah, I don't think the European powers thought too much about us unfortunately, at that time, and certainly didn't as far as I know, think the US was the only market to worry about. So they weren't necessarily reacting to us in this period, maybe in selected industries at certain points in certain places.

But on average I think they're worried about what their trade treaty partners, mostly European nations are doing. And Britain after all doesn't move its tariffs, it's got free trade and a number of other places do too. So the main trading partner out there, Britain is almost oblivious to Almost half of.

Yeah, another big chunk with Canada, Canada might worry about what we're doing and they took action in a certain regard. So that, that might be something to worry about. But the Europeans where a lot of the action is not a big, big thing. Yeah, sorry.

>> Speaker 26: I don't know if it explains it, but one thing you could look at is labor versus capital intensity of various industries.

I don't know whether that would. I mean, just eyeballing it, except for metals, this looks more labor intensive than the ones on the last slide. So again, I can't think of immediately think of the causal mechanism why that would account for the differences.

>> Christopher M. Meissner: Yeah, I mean if it's capital intensive at the time you're talking large concentrated sectors or markets or industries.

Certainly food and tobacco lives up to that and eventually transportation like automobiles will. But early on it's a little more heterogeneous. One is this one. Well, chemicals is a mixed bag, petroleum definitely is, whereas metals here Metals. Yeah, metals would be right.. That's in the first group. This is the first group again.

So why you might think of US Steel at this time, I mean there are a lot of smaller producers. And there are heterogeneous products in here which are more specialized consumer and producer goods. So it's not just raw pig iron or steel, although, it's a big chunk of production.

But it's not all that, I've become very friendly with ChatGPT.

>> Speaker 11: Okay, and I have to say four is a dramatic upgrade over 3, 5. And there's a number of economists, I'm not gonna mention names, you could take this paper and feed it in.

>> Christopher M. Meissner: Okay.

>> Speaker 11: And then you would say okay, do a search on every data and paper out there that you've got access to and match it up against my results.

But you may wanna do a conversation, keep pushing it. So I asked DVD about tariffs for this particular period but I didn't really push it very far to get the pro answers and negative answers and comes out all in the wash basically. And the one thing it sort of says is in the startup culture they're healthier, helpful.

If you keep them past the startup culture, they're harmful. And so, you may have some good results and you might call those the startup culture. And you might later, if you were to keep running them, well, we'll get rid of the tariffs at some point with the income tax, it might not be so good.

So the rate at which data is being added, and ChatGPT is not one of the most serum comprehensive, but I just recommend that to everybody around the table. I don't know if you all do that or you don't like to do it but just have a conversation, feed your data into it and you'd be amazed at how much it knows.

And so I mean I did that and I, I think it comes out with pretty much what you came out with. Yes, four of one, four of the other and four we don't know. But I think that would change time period to time period, so anyway that's all I wanted to add to the conversation.

>> Christopher M. Meissner: Sure, I did a lot of studies on liberalization, just I'm not ChatGPT in any instantiation or anything, but the papers were showing a mixed bag. The one that's most consistent with us is Amity's work with co authors that showed pretty negative relationship between tariffs and productivity from a liberalization in Indonesia I think it was so that was pretty good data.

It's a bit older study, but it really depends on where you're looking. Other work-

>> Speaker 11: I'm gonna say you have to push the damn AI because it doesn't wanna give you what you want on the first pass.

>> Christopher M. Meissner: Yeah.

>> Speaker 11: After three or four passes you'll get a better handle on what you want out of it.

Anyway, that's been my experience, it takes a few passes through to get the balanced information on what you're comparing it with.

>> Christopher M. Meissner: Yeah, much will be done, I am sure with these kinds of tools, it's great. And hopefully to help us digitize even more of these real heterogeneous data sets, individual data sets, really a new world.

Other literature, I mean, great studies by Ian Key and co-authors show that tariffs and selected industries in Canada work to promote productivity. So we're at odds with them and their results. But they had more of an infant industry story and maybe in Canada that's a little bit more plausible.

Okay, where are we at here? I have a couple more tables and a couple of anecdotes should we have time. This table is about adding in input tariffs, right? So there's an old literature on effective protection which, well, requires assumptions. But it seems like modern literature just looks at input tariffs and kind of end user tariffs or downstream tariffs and runs this horse race.

We don't have great input output data. It's kind of at the two digit level this period. So it's not something to really stand on too hard here. But if you throw input tariffs in, those are gonna be negatively associated with productivity, we call these final goods tariffs. But in the IV regression is never statistically significant.

So, do we have enough power in our IV in first stage? Probably not, and it's a lot of measurement error, still IV might not be dealing with too well. So we went there and it goes okay, but not as well as we hoped. You might worry a little bit about dynamics.

>> Speaker 15: Yeah, how are the input tariffs living at that time?

>> Christopher M. Meissner: Okay, so-.

>> Speaker 15: They're specific and they may have been more general.

>> Christopher M. Meissner: So I went fast. I didn't define what I meant by an input tariff. So there's a range of products on which we have tariffs.

We have to find out what the inputs are. To do that we took an input output table from the ONTF, I think that's for the 1920s. And we said, well, industry six uses industry one, someday we use a lot of cotton. What's its share? And then we just said, this is the tariff on cotton industry. 40% weighted by those cost shares. That's as good as we were able to do here.

>> Speaker 15: We never.

>> Christopher M. Meissner: Of the input shares, yeah. Cost shares really on gross output, it's not ideal, it's not as good as the data are today and even those have issues. But we did what we could there to look into it.

But there may be more to do certainly over time. Just anecdotally, right, the government actually does kinda get it in their head. That, yeah, probably it's good to reduce tariffs on key input products, and you the 1890s goes in that direction a little bit. The first chart I showed you showed you a divergence between the dutiable goods and the all goods. And that in large part was due to later in the period due to this recognition that inputs might want to have them come in a little bit cheaper than the fine goods.

>> John Taylor: California still taxes business to business capital investment.

>> Christopher M. Meissner: Well, we have our issues, yeah.

>> John Taylor: Sacramento.

>> Christopher M. Meissner: Yeah, the message doesn't always get through to politicians necessarily, right? Okay, so one, I guess final table here about dynamics. Maybe two tables we looked into long changes instead of levels of productivity. And so we just take the long difference of our estimated equation and think tariffs affecting.

You have average growth and the answer is pretty much the same, it's a negative relationship. Bit more in the spirit of kind of like convergence and dynamics here. We're gonna look at growth of productivity, average growth rate over 10 year periods here now so the coefficients are a bit smaller.

And connection between that growth rate and tariffs at the industry level. And we're gonna include the leg level of the dependent variable in the first column, that's gotta be labor productivity.

>> Speaker 16: Why not show impulse response where it's not t minus 10 but t minus 1, t minus 2, t minus 3 and so forth. And then you'd be.

>> Christopher M. Meissner: Yeah.

>> Speaker 16: So productivity won't stop for that?

>> Christopher M. Meissner: Yeah, okay, so we're working on getting the annual tariffs. We'll never have annual.

>> Speaker 16: I see.

>> Christopher M. Meissner: Okay, right now we don't have annual manufacturing productivity or output. We can get it for selected industries and something I wanna do.

And the tariffs like the annual data are coming, you know, we're working on them. As fast as can be and they will be available. So we will be able to do that. But right now we're having constrained without the annual tariffs. Related to Harold's question, you could do some kind of placebo type specifications, right?

>> Steven Davis: If you run a sequence of cross-sectional regressions, you can regress current productivity growth on future tariffs, right? And check about the kind of validation check that current and past tariffs drive productivity with future tariffs. Depends a little bit on the serial correlation of the changes. It seems like there's some scope for some placebo specifications.

Yeah, event studies, you know that kind of leading leg stuff.

>> Christopher M. Meissner: Yeah, I'm fully on board. Not available yet, but hopefully soon. There are some great policy experiments here too in the period which we hope to exploit. There's across the board cut in tariffs in the 1870s of 10%.

There's the shock of the 1890s when the Democrats retake Congress and control the administration and they lower tariffs for about three years and they go right back up when the Republicans come back. So there's that experiment, there's probably others product level. So we're looking forward to all that and hashing this out more.

But the only dangerous thing about this despite that, in addition to the IV is the fact that we have these fixed effects in here. Once we throw the lag dependent variable and we're trouble with potential nickel bias and out worried us, this was-

>> Speaker 27: Over the IV one last time.

>> Christopher M. Meissner: Sure.

>> Speaker 27: I get nervous when I see that IV blowing something up like that, yeah. And it makes me think that the thing that's being the real protection here is unrelated to the tariff, is what you're saying?

>> Christopher M. Meissner: Okay, now I realize I probably didn't go as far as I should as a.

Good to apply to concentration in terms of motivating this instrument, right?

>> John Taylor: Five minutes.

>> Christopher M. Meissner: Yeah, five minutes. Five minutes, okay, that's not gonna work. But I mean I think it is definitely relevant, right? Definitely predicts changes in tariffs. So I think your question is more about the exogeneity, right, I understood.

Unit values, the US is a price taker, okay? What have we done seriously to deal with this? Okay, we can do a regression of these unit values on some information we have, legged imports, lag changes in imports, SIC industry by census period or year fixed effects. None of those things explain changes in unit values.

We have no explanatory value in observable lagged imports or sic industry fixed effect which proxies for industry supply and demand shocks. So we think we're on pretty Solid ground there. Doug Irwin in his book proclaims US prices to be exogenous, but he doesn't have data here we got the data.

And to back that assertion up, those shares of specific tariffs, it is argued that those are predetermined by Greenland and Lopresti. We haven't looked into it too much, but again we also have, we also have regressions where we try to explain those shares as a function of observables and there's not much going on there, right?

Here we try to explain the share of revenue at a product level or the industry level, I guess in this case as a function of observables. And you can't really explain that either. We think those things are exogenous, but you know, it's not obviously the last word. There are other things to do and other things to check and hopefully we've done enough here to convince you.

But I realize people are always skeptical of these things, especially in our case where the completion's blowing up so much. I'm I remaining like one and a half minutes or so. What time do we end?

>> John Taylor: 1:20.

>> Christopher M. Meissner: 1:20, okay, yeah. Yeah, there were a couple anecdotes, I find them relatively amusing.

Paper and publishing, so this includes final products and raw material like paper stock. Paper stock tariffs were pretty low in these industries. But we did put tariffs on Canadian imports of raw materials like paper, which were low cost. So final goods in this two industries could have done better with these lower tariffs, probably with cheaper inputs.

But the final goods producers had a card up their sleeve from the 1890s. The Platt Simmons act said they had non-tariff barrier. They said, well you can publish here as long as it's printed on in America. So told the foreign publishers they had to print in America and with American publishing houses.

So even if the tariffs were low, there was a non-tariff barrier here in play. And productivity may have grown but probably could have been better with these cheaper inputs and without the protection. This one was really startling to me. According to James Foreman Pack, a leading business historian, he has paper on the automobile industry and this is in the category of transportation, pretty high tariffs over time.

Ford hasn't really come up with his big stuff yet. So it's a bit early, but the Europeans are making serious advances. And according to Foreman Peck and according to, well, other business historians, which he cites. US engineers and designers were in the Stone Age compared to European engineers and trying to solve problems they'd already solved in Europe.

And I find that just kind of consistent with the idea that innovation is stifled by having high tariffs. Later on Ford will make advances and the US Auto industry will come to dominate. Maybe that's coincidence, maybe you could argue it's due to tariffs, but I find this earlier story a little bit more consistent.

So, yeah, I guess. Yeah, sorry. Yeah, conclude, yeah. So, as been said here several times, tariffs don't seem to be associated with productivity growth in United States industry on average. We got some new data to show that. We got a new kind of instrumental variables approach that's on the table here using these exogenous kind of shocks to values.

What's the interpretation? Well, we think Tara's probably limited competition and without that competition, you're not getting the innovation you need. More competition would drive leaner industry and higher productivity through innovation. And oppositely probably a lot of the story here on American tariffs for why they were so high was due to endogenous tariffs and lobbying.

And there's no shortage of anecdotes about that. So could US manufacturers have grown because of tariffs? We think not. I think the evidence is in the fact that US Manufacturing exports, which require you to be productive to meet your international competition, weren't so hot early on. And to the extent that they became more important and grew over time was probably due to agglomeration economies, the patent system and intellectual property rights.

The natural resources that people in this room have written about as a big boost to American productivity and the influx of immigration, which all helped the cause. But tariffs sure didn't help. Corey.