Autor, Dorn, and Hanson on the China Shock
By Scott Sumner
Noah Smith has an article in Bloomberg discussing a new paper by David Autor, David Dorn and Gordon Hanson (ADH):
In his recent book “Economics Rules,” Harvard economist Dani Rodrik laments how economists often portray a public consensus while disagreeing strongly in private. In effect, economists behave like scientists behind closed doors, but as preachers when dealing with the public.
Nowhere is this evangelism clearer than on the issue of trade. Ask any economist what issue they agree on, and the first answer you’re likely to hear is “free trade is good.” The general public disagrees vehemently, but economists are almost unanimous on this point.
But look at actual economics research, and you will find a very different picture. The most recent example is a paper by celebrated labor economists David Autor, David Dorn and Gordon Hanson, titled “The China Shock: Learning from Labor Market Adjustment to Large Changes in Trade.” The study shows that increased trade with China caused severe and permanent harm to many American workers:
Adjustment in local labor markets is remarkably slow, with wages and labor-force participation rates remaining depressed and unemployment rates remaining elevated for at least a full decade after the China trade shock commences. Exposed workers experience greater job churning and reduced lifetime income. At the national level, employment has fallen in U.S. industries more exposed to import competition…but offsetting employment gains in other industries have yet to materialize.
Autor, et al. show powerful evidence that industries and regions that have been more exposed to Chinese import competition since 2000 — the year China joined the World Trade Organization — have been hit hard and have not recovered. Workers in these industries and regions don’t go on to better jobs, or even similar jobs in different industries. Instead, they shuffle from low-paid job to low-paid job, never recovering the prosperity they had before Chinese competition hit. Many of them end up on welfare. This is very different from earlier decades, when workers who lost their jobs to import competition usually went into higher-productivity industries, to the benefit of almost everyone.
I have a lot of problems with this claim. First, even if China trade was bad for the US, it was almost certainly extremely good for China, which was a vastly poorer country in 1990. So I’m quite confident that economists are justified in supporting free trade. Whether they are justified in suggesting that Chinese trade is beneficial to the US is another question.
Second, this is just one study, and as we’ll see it’s far from convincing. We don’t abandon views held for 200 years, and supported by hundreds of studies, just because of a single study. I can’t speak for other economists, but I very much doubt whether economists are holding back some sort of “secret” information that free trade is actually bad.
The ADH paper does show a nice job of showing that Chinese exports have depressed some local labor markets. But unless I’m mistaken the paper doesn’t tell us anything about the macro effects of Chinese exports, which would require a macro model. Here are three possible ways that Chinese exports might hurt the aggregate economy:
1. It might depress aggregate demand
2. It might depress aggregate supply by reducing the long-term productivity of the US economy.
3. It might reduce aggregate supply by causing medium-term structural “reallocation” problems, as labor had trouble migrating to new jobs.
I’ll call these the AD shock, the AS/efficiency shock and the AS/reallocation shock channels. Even after reading the ADH paper, I am having trouble understanding which channel is relevant.
The easiest shock to address is the AD shock. EC101 students are sometimes confused by the GDP equation:
GDP = C + I + G + (Ex – Im)
This equation makes it look like a current account deficit would reduce GDP. However the CA deficit is exactly equal to the capital account surplus, which is I – S. So if we import more than we export, we also invest more than we save. Thus a CA deficit might boost investment, or if it reduces saving it might boost consumption. There is no “accounting argument” for the claim that trade deficits reduce aggregate demand.
There is a more sophisticated argument that CA deficits reduce AD, but only at the zero bound. Paul Krugman has suggested that when interest rates are zero, a CA deficit may depress the equilibrium interest rate, making monetary policy effectively tighter. Because we are at the zero bound, the Fed may not offset this shock and total AD may decline.
There is one big problem with this theory; it doesn’t apply to the 1990-2007 period considered in the ADH study. Interest rates were never at zero, and thus monetary offset clearly applied. Try to imagine a policy counterfactual involving a ban on Chinese imports. Unemployment was only 5.2% in June 1990, near the peak of the Reagan boom. By June 2000 it had fallen to 4.0%, near the peak of the Clinton boom. In the subsequent recession it never got higher than 6.3%, and then fell back to 4.6% in June 2007. Whatever you think about this data, the Fed clearly thought AD was adequate, or they would have eased policy further. Thus if a ban on Chinese imports did somehow boost AD, its effects obviously would have been offset by the Fed. I’m pretty sure that even Keynesians like Paul Krugman would agree with that claim. So I think it’s safe to assume that whatever the channel was by which China trade hurt the US economy, it was certainly not the AD shock channel.
The long run AS/efficiency channel also seems unlikely. Basic economic theory suggests that productivity and efficiency are highest when a country concentrates on producing those goods for which it has a comparative advantage. Thus the most likely channel would be the reallocation channel; something prevents workers who lost their jobs in one sector from quickly finding jobs in other booming sectors. And indeed ADH do frequently discuss the problem in exactly those terms, workers don’t seem to be able to easily reallocate out of areas hit by the China trade shock.
However there are also lots of puzzling observations that seemed to conflict with this interpretation:
In the German case, the impact of rising Chinese import competition between 1988 and 2008 was compounded by an even more rapid growth of imports from Eastern Europe following the fall of the Iron Curtain. Distinct from the U.S. case, German manufacturers sharply increased exports to these lower-wage countries, resulting in a more modest trade deficit with China and a trade surplus with the Eastern Europe. The unemployment gains related to these export opportunities roughly offset the job losses from import competition in the case of China, while actually raising German employment in the case of trade with Eastern Europe.
This makes no sense to me. Unless I’m misreading them, they seem to be suggesting that, unlike the US, the German labor market has not been hurt by recent trade patterns, because Germany runs a trade surplus. Thus job losses due to imports were offset by job gains due to exports. But that only make sense if the AD shock channel was operative, and as we’ve already seen the AD channel cannot possibly explain why China trade would hurt the US, on either theoretical or empirical grounds. I would add that German labor market data was far worse than US data during the 1990-2007 period, with unemployment rising from 5.6% in 1991 to 7.7% in 2000 to 8.6% in 2007. (I could not find 1990 data for Germany.)
The Economist also thought the ADH paper implied that trade deficits might be a transmission channel:
The costs of Chinese trade seem to have been exacerbated by China’s large current-account surpluses: China’s imports from other countries did not grow by nearly as much as its exports to other countries. China’s trade with America was especially unbalanced. Between 1992 and 2008, trade with China accounted for 20-40% of America’s massive current-account deficit; China imported many fewer goods from America than vice versa.
And the ADH paper says something similar:
The impact of China’s recent growth on the global economy is not just about the country’s long-run comparative advantage. Also important for the near-term labor market consequences of its trade expansion is that China’s trade surplus widened substantially.
But the Keynesian theory that trade surpluses overseas can cost jobs in the US (which only applies at the zero bound) also implies that the (roughly equal) German trade surplus was just as harmful as the Chinese surplus, even though Germany exports less to the US than China. The AD shock theory implies that overseas trade surpluses hurt us by increasing global saving and reducing velocity, thus bilateral trade patterns don’t matter. In that case the problem isn’t China, it’s everyone with trade surpluses.
If the AHD model were based on the AS/reallocation channel (which is the only channel that makes sense to me), then the German trade surplus would be irrelevant. It’s not obvious why it would be harder to reallocate German workers from import competing to export industries, than it would be to reallocate American workers from import competing to Boeing exports plus housing investment plus manufacturing aimed at the domestic market, as long as total AD was on target.
This also confuses me:
Two additional sources of linkages between sectors operate through changes in aggregate demand and the broader reallocation of labor. When manufacturing contracts, workers who have lost their jobs or suffered declines in their earnings reduce their spending on goods and services. The contraction in demand is multiplied throughout the economy, depressing consumption and investment. Helping offset these negative aggregate demand effects, workers who exit manufacturing may take up jobs in the service sector or elsewhere in the economy, replacing some of the earnings lost in trade exposed industries. Because aggregate demand and reallocation effects work in opposing directions, we can only detect their net impact on aggregate employment.
Again, unless I’m missing something, ADH seem to be suggesting an AD shock channel, which makes no sense.
They also don’t discuss all the jobs created in places like Seattle and Silicon Valley, as a result of China trade. Without exports, how would China be able to buy all those Boeing jetliners? In fairness, ADH do admit that their study may miss important linkages in high tech and business services, but this admission occurs only in a footnote:
A further issue is that measured input-output linkages may miss some positive demand effects from U.S. exports. Consider the iPhone, whose back panel states, “Designed by Apple in California. Assembled in China.” From its U.S. headquarters, Apple offshores production to Foxconn, which employs 300,000 workers in its iPhone operations in China. If productivity in Foxconn rises, iPhone sales may expand, thereby increasing demand for design services among Apple’s 50,000 US employees. However not all of Apple’s design exports to China may appear in U.S. trade data. For tax purposes, Apple may attribute some iPhone revenues to overseas subsidiaries. These revenues would not appear in the US current account until the earnings are repatriated, possibly far in the future. A similar logic applies to U.S. business services that may expand as a result of increased trade with China.
Elsewhere the anecdotes they use don’t even fit their hypothesis:
When looking within manufacturing, Tennessee, owing largely to its concentration of furniture producers, is far more exposed to trade with China than his Alabama, which has agglomeration’s of relatively insulated heavy industry.
The Atlantic interpreted ADH this way:
But that didn’t really happen. In Tennessee, few workers within the commuting zone of struggling plants moved away after their work prospects declined. Instead, unemployment rose both among manufacturing and nonmanufacturing workers, suggesting that the ill effects of increased trade had a spillover impact on the larger local economy. On top of that, average weekly wages declined. In general, places like Tennessee were very slow to adapt to the new economic reality–their elevated unemployment rates and diminished wages persisted for a decade, the paper’s authors estimate. The workers there are also saw a lower lifetime income.
That makes it sound like Tennessee’s economy really sucked during this period. In fact, the average unemployment rate was slightly below the national average:
Indeed the booming Tennessee economy has done far better than Alabama since 1990, drawing in workers from many other states. Population growth from 1990 to 2010 was 18.3% in Alabama, 24.1% in the USA and 30.1% in Tennessee. Why not choose Detroit as an anecdote? Perhaps because China doesn’t export many cars to the US. Indeed total Chinese exports were only 2.2% of US GDP in 2007. If Detroit has been hurt by trade, it’s been from Japan, Korea, Germany, Mexico and Canada, not China. (Although auto parts makers have been hit by China.)
Although Chinese exports are a small share of GDP, they are very important to the sort of low and middle-income workers who shop at places like Walmart. Trade with China has improved living standards for millions of Americans. These media reports seem way too pessimistic to me, unless I’ve completely misread the ADH paper. Tyler Cowen suggested:
This is some of the most important work done by economists in the last twenty years.
Maybe I’m missing something, but I don’t see that. We’ve always known that local labor markets can be hit hard by import competition, and thus ADH don’t really change our understanding of trade in any major way.
Interestingly, the Economist seemed to have a different interpretation from Noah Smith, arguing the ADH paper merely reduced the estimated net gains from trade, which remain positive:
But those benefits are only visible after decades. In the short run, the same study found, America’s gains from trade with China are minuscule. The heavy costs to those dependent on industries exposed to Chinese imports offset most of the benefits to consumers and to firms in less vulnerable industries.
On the other hand, there are also many good features in the ADH paper, as when they discuss how government benefit programs may slow the adjustment to trade shocks:
Workers eligible for TAA receive extended unemployment benefits of up to 18 months, as long as they remain enrolled in a training program, and may obtain allowances towards relocation, job search, and healthcare.
Overall, I had a lot of trouble making sense of this paper. To draw macro implications, you’d need a macro model, including assumptions about monetary offset to evaluate a counterfactual with no China trade. But I couldn’t find this model.
And who is the intended audience? The sort of economist who is most likely to be receptive to this message is not a free market supporter like me, but rather a left of center pragmatist. But people like Krugman and Summers were extremely skeptical of the claim that structural/reallocation theories explained high unemployment after 2008, and insisted that an AD shortfall was the real problem. I agree that AD was the real problem after 2008, which is one reason I am skeptical of this paper. But if you did dismiss the claims of people like Arnold Kling, that much of the unemployment was due to the difficulty workers had reallocating out of residential real estate construction, then why would you be receptive to the ADH paper?
I don’t want to sound too negative here. While I don’t buy the argument that trade is harmful in a macro sense, I do think ADH have done a good job of showing that labor reallocation may be harder than we assumed. This has lots of policy implications. We should be more skeptical of policies that slow reallocation, such as zoning restrictions on development and rent controls (both of which Steve Waldman recently defended), and extended unemployment compensation programs.
One of the things I liked most about the ADH paper was that they recognized the massive gains from trade to China. Thus even if the effects of trade on the US were slightly negative in net terms (which I doubt) the case for free trade would remain overwhelmingly powerful, at least unless you were a nationalist who opposed any sort of foreign aid, even aid that hugely boosted world efficiency.
The thing I liked least is the ambiguity about the model they were using. If they are right about the costs of reallocation, does it suggest that all “creative destruction” is bad, including technological progress? That would seem to be the implication, but I doubt they’d want to go that far. So why focus on jobs lost by the China shock, but not German exports or robots replacing workers? Early in the paper they suggest that China is special, as it was a once in a lifetime massive shock from a huge country, which hit us rapidly, but also that wages in China are now rising fast, so much of the adjustment is over. So even if they are right about China during 1990-2007, it probably has no policy implications going forward, as other types of creative destruction like rising German exports and robots tend to occur more gradually over time. And if I am too complacent about robots, then the policy implication of the paper is not anti-China (it’s too late to prevent that shock), but rather anti-robot, as robots might be the next massively disruptive shock.
Sorry to be so long winded, but I’m not seeing anyone seriously grapple with the implications of their research, if it is correct. I would appreciate any comments you might have.