Is the decline in manufacturing employment due to trade competition? Insights from a decomposition.
Figure 1: Employment change from 1997Q1 (black line), change attributable to output change (blue bar) and attributable to productivity change (tan bar). Calculations based on log differences. NBER defined peak-to-trough recession dates shaded gray. Source: BLS Employment and Costs release, NBER, and author’s calculations.
This decomposition works off of the identity:
h ≡ y – (y-h)
where h is (log) hours in manufacturing, y is real value added output in manufacturing, and (y-h) is value added output per hour.
The graph shows that, while the employment loss during the 2001 recession is due to output reduction — in a Keynesian framework, a decrease in aggregate demand. However, the subsequent job loss is due to rapid productivity growth.
Does this mean that the job loss is not due to trade? Not necessarily; labor productivity growth is not exogenous. Increased import competition might induce accelerated productivity growth. In addition, increased offshoring as specialization breaks up value chains should increase productivity (that is, there is specialization in tasks).
So, while I can’t dismiss international trade as the key reason for reduced manufacturing employment, the decomposition is suggestive that one has to be careful about attributing the bulk to international trade.
Off topic, but a continuation of yesterday’s comments about yield spreads.
First of all, thanks to Prof. Chinn for posting the link to Shiller’s historical interest rate data.
Based on that, I did go back and take a look at Shiller’s annual 10 year Treasury bond data from 1871-1900, and compared it with his “1 year interest rate” data for the same years, as well as the NY commercial paper data, which I averaged annually as well.
In all but three years during the entire period, 1 year interest rates were higher than 10 year Treasury rates. The same was true with only one exception in the comparison with commercial paper interest rates. The reason for the latter is presumably that we are comparing corporate paper with government paper, and government paper is almost always going to command a lower interest rate for obvious reasons having to do with the likelihood of default. I suspect Shiller’s 1 year interest rates are also corporate bond rates, but I don’t have access to the book that he cites as his source.
I won’t bore people with the spreadsheet, but although very little can be gleaned from annual vs. quarterly or monthly data, one thing that does stand out is that the spread between commercial paper and long term Treasurys did increase, usually sharply, in the year just before or the first year of recessions during that era. So while we apparently can’t use a rubric of inverted vs. normal yield spreads, the *relative* yield spreads still had forecasting utility going back into the 19th century (and have continued to the present, albeit noisily).
Apropos of which, as of last month, the spread between corporate bonds and 10 year Treasurys was at its lowest level since the pandemic.
Corporate spreads are, indeed, quite narrow, by some measures the narrowest since before the housing crash:
https://fred.stlouisfed.org/graph/?g=1D8NU
Corporate spreads mostly reflect perceived payment risk, and logically ought to track measures of actual payment problems. Here are corporate spreads against non-performing loans:
https://fred.stlouisfed.org/graph/?g=1D8Qd
Pretty good match. Here are corporate spreads against a couple of broad measures of financial stress:
https://fred.stlouisfed.org/graph/?g=1D8R0
Also a good match, though perhaps suggesting spreads have gotten too narrow.
There is one measure of payment problems which has recently diverged from NPLs, credit spreads and broad measures of financial stress, and that’s bankruptcies. Corporate bankruptcies hit a 14-year high in 2024:
https://www.spglobal.com/marketintelligence/en/news-insights/latest-news-headlines/us-corporate-bankruptcies-soar-to-14-year-high-in-2024-61-filings-in-december-87008718
I’m sure somebody out there knows what’s driving this disconnect, but I haven’t figured it out yet. I’d provide a picture showing the divergence, but FRED doesn’t seem to track bankruptcy filings.
We might also want to consider exports as a share of GDP; maybe we’ve lost jobs to foreign competition in overseas markets?
https://fred.stlouisfed.org/series/B020RE1Q156NBEA
No. Exports as a share of GDP have more than doubled since 1970, with progress made through much of the period of increasing trade deficits.
Since 2015, there has been a drop-off in exports as a share of GDP, apparently associated with dollar strength:
https://fred.stlouisfed.org/graph/?g=1D8Mh
Dollar strength has been associated with relative U.S. economic strength in that period
One can always argue that we’ve lost exports as a share of GDP relative to some counter-factual in which overseas demand grew as it has, but overseas production did not, but there are some obvious objections to that argument.
We might also want to consider NET exports as a share of GDP; maybe we’ve lost jobs to foreign competition in overseas markets?
https://fred.stlouisfed.org/series/A019RE1A156NBEA
Net exports are a reasonable measure to consider, but involves additional factors. I intentionally chose just exports to keep things simple.
The original question was what caused factory job loss. Menzie showed that productivity gains account for most of the loss. My point was that exports have risen, so a fall in exports, by itself, cannot account for lost factory jobs. I tossed in “as a share of GDP” as a bonus, since GDP is also expanding.
What’s left to consider, once we bring net exports into it, is imports. Have we lost factory jobs because we import stuff? That’s your question, once the factors already discussed have been addressed.
OK, so net exports, and the current account overall, are driven by the national saving balance. We spend so much – save so little – that we run a current account deficit. Does spending a lot cost us jobs? Seems unlikely. Recession involves spending less, and a loss of jobs, while expansion means more spending and more jobs.
I dosee how it works the other way around. There’s some missing step between spending less (on imports) and having more (factory) jobs, a classic “underpants gnomes” problem.
Personally, I’m a fan of spending less to reduce imports. We are poisoning the world, and paying our trade partners to poison it, through our spending. I’m a fan of taxing the well-off as a way of reducing spending. Taxes would improve national saving, all else equal, and that would lead to a smaller current account deficit, mostly through reduced imports.
Off topic – wherein I make the idiotic mistake of disagreeing with Paul Krugman:
Krugman is thinks that financial market pricing shows that investors have stopped worrying about Trump’s immigration and tariff threats, even though Trump is likely to carry through with those threats.
According to Krugman, Trump “plans to impose 25 percent tariffs on Canada and Mexico on Feb. 1 unless they stop massive flows of migrants and drugs across their borders. This will be a hard demand to meet, because to a large extent those massive flows exist only in Trump’s imagination.”
https://paulkrugman.substack.com/p/voodoo-trade-policy-comes-to-america
I think this ignores a part of Trump’s method. When he claims that a problem exists that actually doesn’t, he can just as easily turn around and claim that the fictional problem has been solved – thanks to Trump. If reality doesn’t matter, then reality doesn’t matter.
That doesn’t mean Trump won’t go ahead with stupidly high tariffs. He might. One lesson of the first Trump term was that we should expect Trump to do what he says he’ll do, sometimes.
Another thing we learned the first time around is that he bullies and threatens to get what he wants. Merely threatening tariffs, if the threat gets him what he wants, is as good as imposing tariffs. That actually built into his threats against Canada and Mexico. Ya listening, Denmark? Panama?
Yet another lesson is that he listens to those who flatter him…one after another after another. Some sycophants will ask for some tariffs to be avoided. Some of them will get what they ask for. How far that will gi to prevent an increase in tariffs is anyone’s guess.
Expelling immigrants may be harder to fake. Only a few of us will ever know more about tariffs than faux news tells us, but those danged pet-eating immigrants are right out where we can see them. Can Trump lie-finesse his way through deportations they way I think he can through tariffs? Dunno.
It’s intriguing to see how productivity improvements can lead to job losses. However, could productivity growth bring back jobs in a different capacity through technological advancements? An identity can’t tell you what causes what. For that, one needs a model.
CuriousReader: I would be wary of imputing causality. I’m saying that for a given amount of change in employment, a portion can be attributed to change in output, and a portion can be attributed to the decline in hours of work per unit output.
Along with a model, we can look at data. Here are total employment in telecommunications and in information:
https://fred.stlouisfed.org/graph/?g=1D96c
Both added jobs, then shed them, and neither is back to it previous high. Y2K accounts for some of that pattern. Information jobs look like they got a boost from the Covid recovery, but maybe tech advancement has a limited shelf-lofe as a job creator?