The question of whether trade with China has inflicted lasting harm on the United States is the subject of some much-celebrated research by three distinguished economists: David Autor, David Dorn, and Gordon Hanson. They argue that import growth from China cost the United States about 2.4 million jobs over a dozen years. To give an idea of this work’s reception, economist Tyler Cowen referred to it as “some of the most important work done by economists in the last twenty years.”
I strongly disagree. The work provides evidence that U.S. labor markets do not adjust quickly or easily to shocks. When a manufacturing worker loses a job, not only can it take a long time to find another, not only might that new job have lower wages, but the worker’s entire lifetime stream of earnings may suffer. But we already knew this.
In a presentation in Washington on Thursday, Hanson said the authors hoped to make the point that adjustment to trade shocks may be neither quick nor easy, as some trade models predict. Yet this has been part of the trade canon for at least 45 years, since the “specific factors” work of Ron Jones.
The real question is whether American workers have suffered ills due to trade — more specifically, due to trade with China. Autor, Dorn, and Hanson show empirically that where imports from China increased the most, the pain to workers was most acute. But this does not prove their point. That would require them to defend their preferred cause against alternative explanations.
In the case of whether trade caused harm, there are two prominent alternatives:
Workers were hurt by technological change. Whereas factories used to need low-skilled workers to pull levers, they now need higher-skilled workers to reprogram computer-guided machine tools. This is the sort of argument put forward by Claudia Goldin and Lawrence Katz.
Workers were hurt by domestic competition. Textiles that used to be made in the Northeast moved to the Southeast. Auto production that used to take place in Michigan and Ohio moved to Tennessee and Alabama.
The possibility to test against these alternatives depends on the use of models that allows for them to be taken into account, and on data that allows one to disentangle the effects. To see the potential problem, suppose that China mostly exported goods produced by unskilled labor. And suppose that these good were produced under conditions conducive to machines replacing workers (both plausible assumptions).
The effects would be difficult to disentangle if China emerged as a trading power at the same time that factory automation progressed — which is exactly what we saw. So if we thought of the cheapest way to make a certain good, we might have come up with a ranking:
- Cheapest: import from China
- Next cheapest: automate the factory
- Most expensive: the traditional labor-intensive approach, made in the United States
In this case, even though we would see a correlation between job loss and Chinese imports, the argument that China was the cause of the labor woes would be inaccurate. Absent China, those woes would still have occurred as the factories automated. (One piece of evidence that supports the relative importance of automation is that the trend has been appearing globally, including in China).
Setting aside the question of trade’s importance relative to other explanations, what about China’s role? Here, Autor, Dorn, and Hanson adopt a particularly unfortunate, blinkered approach that has been central to a number of advocacy studies in the China realm — the research focuses on China but ignores supply from the rest of the world.
In the earlier example, the alternative to importation from China was automation. What if the alternative to importation from China was to import from Vietnam, rather than to produce in the United States? In that case, there would be no marginal impact of China trade — it would be a question of sourcing.
This is more than just a hypothetical. In Thursday’s Washington presentation, Autor suggested that a desirable policy would be one in which workers could be temporarily shielded from trade shocks and given time to adjust. As it happens, we have had just such a program all along. When China joined the WTO in 2001, the United States gained access to a special “safeguard” program known as Section 421. This set a very low hurdle for reimposing protection against China. A complainant had only to show “market disruption.”
So why didn’t we see many instances of tariffs being slapped on Chinese goods in the 2000s as China emerged as a great trading power? Critics had an easy explanation — the Bush administration was soft on China. I served on a couple of the interagency panels that reviewed these cases, though, and would offer an alternative explanation: There are more than two countries in the world.
In each of the two Section 421 cases I heard, the importers made credible presentations that, were tariffs to be imposed, they would switch their sourcing from China to Vietnam, or to India, or Brazil. In one case, the factory move was estimated to take three weeks. In another, contingent contracts were already in place. Producing in those places cost a bit more than in China, which is why they weren’t the original sourcing countries, but they were cheaper than the United States. So what benefit would U.S. workers have seen in blocking China trade? None. That’s why we recommended against imposing tariffs.
The Obama administration came in promising to remedy this. They did apply tariffs on Chinese tire imports under Section 421, in 2009. The benefit to U.S. workers? As shown by Gary Hufbauer, essentially none (with costs to the rest of the economy).
This does more than call into question one of Autor, Dorn, and Hanson’s potential remedies. It calls into question the premise of their analysis. If the alternative to imports from China was imports from other developing nations, then the impact of China on U.S. workers was negligible. While China’s emergence clearly had some impact, the blinkered approach does little to help us pin down what that impact was.
Although Autor, Dorn, and Hanson abjure protection as a remedy, authors often have little say about how their analyses are used, even when those analyses are correct. In a charged election year, with China suspicion running high, it becomes even more vital that analyses be well-conceived.
By Phil Levy