1 The Causal Question

Does increased import competition from China cause adverse labour market outcomes for US workers? This question is deceptively simple. Rising Chinese imports coincided with declining manufacturing employment in the United States from the 1990s onward, but correlation does not imply causation. US demand patterns were changing at the same time—workers might have been leaving manufacturing voluntarily—and Chinese exports are endogenous to US industry conditions. Autor et al. [2013] developed a causal identification strategy that moved this debate decisively forward and reshaped the empirical trade literature.

2 Identification Strategy: The Bartik Shift-Share Instrument

Autor et al.'s identification relies on a shift-share (Bartik) instrument. The core idea is to isolate variation in Chinese import exposure that is driven by China's rising global competitiveness rather than by local US labour market conditions.

Define the exposure of commuting zone i to Chinese imports as:

where L_ijt is employment in industry j in commuting zone i at time t, L_ujt is national employment in industry j, ΔM_ucjt is the change in US imports from China in industry j, and L_it is total employment in commuting zone i. The exposure measure is a weighted average of national-level import growth, where the weights reflect the local economy's initial industry mix.

The endogeneity concern is that commuting zones specialising in industries facing more import competition may have been in secular decline for reasons unrelated to China (e.g., automation, domestic demand shifts). To address this, Autor et al. instrument ΔIPW_uit with import growth in those same industries in other high-income countries (Japan, Australia, Canada, New Zealand, and several European nations):

where ΔM_ocjt is Chinese import growth in other high-income countries. The logic: China's cost competitiveness in specific industries caused import surges across all high-income markets; this common supply-side push is unrelated to demand-side shocks specific to US local labour markets. Autor et al. [2013] provide three validity arguments: (1) the instrument is highly correlated with US import growth, satisfying relevance; (2) other high-income countries' import demand did not respond to US-specific demand shocks, satisfying independence; (3) the instrument does not directly affect US labour market outcomes except through US import exposure, satisfying the exclusion restriction.

3 Data and Setting

‍The paper uses a long panel from 1990-2007, divided into two sub-periods (1990-2000 and 2000-2007). The unit of observation is the commuting zone (CZ), a geographic unit developed by Tolbert and Sizer [1996] that aggregates counties into functional labour markets based on commuting patterns. The United States has approximately 722 commuting zones covering the entire continental US.

The main outcomes are changes in manufacturing employment as a share of working-age population, changes in wages, and changes in non-employment (capturing transitions to unemployment and out of the labour force). Industry employment data come from the Census of Population (1990, 2000) and the American Community Survey (2000-2007). Import data come from the UN Comtrade database.

4 Key Findings

Employment. The 2SLS estimates show that a 1,000 dollars increase in per-worker import exposure reduces manufacturing employment as a share of working-age population by approximately 0.596 percentage points. Given that the average increase in Chinese import exposure was about 1,800 dollars per worker over 1990-2007, this implies a decline of roughly 1 percentage point in the manufacturing employment share. This is economically large: it accounts for roughly 25% of the actual decline in manufacturing employment over this period.

Wages and non-employment. Import exposure reduces wages in manufacturing and also in non-manufacturing sectors, suggesting negative spillovers through local labour demand. It substantially raises non-employment transitions to unemployment and labour force exit—rather than generating smooth reallocation to other sectors. This finding challenged the standard Heckscher-Ohlin prediction that trade liberalisation generates efficiency-improving labour reallocation.

Social insurance. Autor et al. [2013] also show that higher import exposure leads to significantly higher government transfers per capita—disability insurance, unemployment insurance, retirement benefits, and income support—suggesting that displaced workers relied on the social safety net rather than finding new employment.

5 Limitations

Exclusion restriction. Several authors have questioned whether the instrument satisfies the exclusion restriction. Goldsmith-Pinkham et al. [2020] show that Bartik instruments identify a weighted average of industry-level estimates, and the instrument is valid only if the industry-level shocks (here, industry shares) are exogenous. Regions specialising in industries that face Chinese competition may have been on declining trajectories for reasons correlated with other high-income countries' import patterns.

Spillovers across commuting zones. The analysis treats commuting zones as independent. Yet labour mobility and inter-regional supply chains imply that a trade shock in one region spills over to others, violating the assumption of cross-unit independence that underpins the local labour market design.

General equilibrium. The estimates are partial-equilibrium. A general-equilibrium model would account for price adjustments, factor mobility, and changes in other industries that occur as a result of the China shock. Caliendo et al. [2019] build a quantitative GE model and find that the overall welfare effect of Chinese imports for the US was positive in aggregate despite large distributional costs for manufacturing workers.

Reanalysis debates. Acemoglu et al. [2016] extended the ADH framework to study broader automation and trade effects, while Bloom et al. [2016] replicated and extended the analysis to the UK and European markets, finding similar patterns. The robustness of the main findings across many replications has bolstered confidence in the original result.

6 What We Learn

The Autor et al. [2013] paper demonstrates several methodological lessons beyond its specific findings:

1. Bartik instruments can be powerful when the shares (initial industry mix) and shifts (national import changes) are plausibly exogenous to local shocks. Formal assessment of each component as developed by Borusyak et al. [2022] and Goldsmith-Pinkham et al. [2020] is now standard.
2. Local labour markets matter. The use of commuting zones as the unit of analysis revealed persistent geographic concentration of trade losses that aggregate analysis obscured.
3. Labour reallocation is slow and costly. The findings challenge simple models of frictionless labour markets and provide motivation for active labour market policy in trade-exposed regions.
4. Identification requires theory. The exclusion restriction is not self-evident; it requires a model of why Chinese export supply shocks affect other high-income countries similarly and for common reasons.

References

1. Acemoglu, D., Autor, D., Dorn, D., Hanson, G. H., and Price, B. (2016). Import competition and the great US employment sag of the 2000s. Journal of Labor Economics, 34(S1):S141-198.
2. Autor, D. H., Dorn, D., and Hanson, G. H. (2013). The China syndrome: local labor market effects of import competition in the United States. American Economic Review, 103(6):2121-2168.
3. Bloom, N., Draca, M., and Van Reenen, J. (2016). Trade induced technical change? The impact of Chinese imports on innovation, IT and productivity. Review of Economic Studies, 83(1):87-117.
4. Borusyak, K., Hull, P., and Jaravel, X. (2022). Quasi-experimental shift-share research designs. Review of Economic Studies, 89(1):181-213.
5. Caliendo, L., Dvorkin, M., and Parro, F. (2019). Trade and labor market dynamics: general equilibrium analysis of the China trade rise. Econometrica, 87(3):741-835.
6. Goldsmith-Pinkham, P., Sorkin, I., and Swift, H. (2020). Bartik instruments: what, when, why, and how. American Economic Review, 110(8):2586-2624[cite: 4].
7. Tolbert, C. M. and Sizer, M. (1996). US commuting zones and labor market areas: a 1990 update. Economic Research Service Staff Paper, 9614[cite: 4].

‍