Learning the Right Lessons from the ‘China Shock’

An influential 2013 paper by economists David Autor, David Dorn, and Gordon Hanson finds an average reduction in manufacturing employment of 90,000 jobs per year from 1990 to 2007 because of U.S. competition with imports from China. Those economists published another important paper, along with economists Daron Acemoglu and Brendan Price, that found import growth from China led to 200,000 fewer jobs per year in the U.S. from 1999 to 2011.

In my new paper for the Aspen Economic Strategy Group, I try to place those numbers in perspective:

In a typical month, five million workers separate from their employers. In the manufacturing sector, typical monthly separations are 350,000.

Moreover, these papers focus on only half the story of increased trade openness. Trade liberalization is associated with increased imports and exports. And economic theory suggests that employment reductions in sectors exposed to import competition should be roughly balanced by employment increases in export-intensive sectors and in other sectors. (Note that I am not criticizing the papers; I am pointing out that the way the papers have been discussed in the media and the policy community is missing an important understanding of the actual content of the studies themselves.)

The economists Robert C. Feenstra, Hong Ma, and Yuan Xu attempt to estimate the effects of trade openness on both sides of the ledger. In a 2019 paper, they expand the Acemoglu et al. framework to incorporate not only U.S. imports from China but also imports from the rest of the world, exports to China, and exports to the rest of the world.

From my AESG paper:

When they study the effects of trade more broadly, Feenstra, Ma, and Xu find that the job losses identified by Acemoglu et al. and Autor, Dorn, and Hanson are fully offset by job gains due to US exports.

Specifically, in their preferred specification they find job losses of 533,000 due to import competition between 1999 and 2011. These losses were offset by job gains of 411,000 due to exports. Since export-driven gains were greater than import-driven losses during the 1990s, over the entire 1991–2011 period they find a net gain of 379,000 jobs. As robustness checks, Feenstra, Ma, and Xu examine specifications using only Autor, Dorn, and Hanson-style instruments and ending their time period in 2007. They find job losses from Chinese import competition totaling 671,000 alongside export-driven job gains of 1.2 million.

I argue that much of the policy debate reflects the wrong lessons from the “China shock,” including that trade reduces employment. Indeed, I argue that trade is not about jobs at all — it is about consumption, productivity growth, and wage growth.

What, then, are the right lessons?

There are two important lessons for economists and policymakers from the “China shock” literature. First, the labor market is less fluid than many economists had thought, and it is harder for workers specialized in one sector with declining opportunities—in the case of trade, in import-competing sectors — to reallocate to other sectors with expanding opportunities. The second lesson is that workers may be less willing to relocate from regions with declining opportunities to regions with expanding opportunities than many economists had thought.

These are generalizable lessons that apply to labor market disruptions broadly, regardless of the source of the disruption. For example, the development of generative artificial intelligence raced forward in 2023 and portends substantial labor market disruption (Strain 2024b). The energy transition away from fossil fuels could create a situation similar in kind to the China shock, given the geographic concentration of that industry. These lessons from the China shock will apply to AI and the energy transition.

Other lessons from the China shock are important for understanding that episode but may be of limited generalizability. China’s export growth was explosive, with its share of world manufacturing exports rising from 3 percent in 1995 to 18 percent in 2014, to 21 percent in 2020. And the reallocation of workers across sectors was likely severely adversely affected by the 2008 global financial crisis and Great Recession, in which the US unemployment rate peaked at 10 percent and there were as many as six unemployed workers for every one job opening in the labor market. To the extent that adverse effects on import-competing workers created Keynesian aggregate-demand reductions, post-2007 economic slack was a major contributor.

The paper is well titled: “Protectionism is Failing and Wrongheaded: An Evaluation of the Post-2017 Shift toward Trade Wars and Industrial Policy.” Check it out here.