International trade and local labor markets: Do foreign and domestic shocks affect regions differently?

Abstract

Despite the attention given to international trade in discussion of the economic struggles of many U.S. regions, it is unclear whether international trade shocks impact local economies more, or differently than shocks originating within the domestic economy. A challenge in making this discernment is separating trade shocks from common or domestic shocks. Therefore, using U.S. county-level data for 1990-2010, this study carefully constructs shocks to local economies, isolating those arising from international imports and exports to assess whether trade shocks have different effects from domestic shocks. In confirmatory analysis, we also employ a novel combination of IV and matching strategies. We examine a variety of indicators including employment growth, population growth, employment rates, wage rates and poverty rates. The results suggest that international trade shocks have some different effects than overall domestic shocks, though likely less than commonly perceived. We also find that domestic shocks dominate international trade shocks in explaining variation in regional labor market outcomes.

Heterogeneity in the local employment multipliers in the United States

“…The 2SLS estimated coefficient of 1.072 is relatively larger than the reduced form of estimate of 0.829. The averaged reduced form of multiplier ranges between 1.600 and 1.829, which is within the range of typically estimated local employment multipliers (see, e.g., Bartik & Sotherland, 2019;Jones & Yang, 2018;Partridge et al, 2017;Tsvetkova & Partridge, 2016;van Dijk, 2017van Dijk, , 2018. The 2SLS estimates may appear to be an improvement over the OLS estimates but as noted by Bartik and Sotherland (2019), the 2SLS approach omits some of the agglomeration and congestion effects following a demand shock.…”

“…In constructing their shift-share instruments, Partridge et al (2017) and Tsvetkova and Partridge (2016) used all industries, both tradables and nontradables. In constructing their shift-share instruments, Partridge et al (2017) and Tsvetkova and Partridge (2016) used all industries, both tradables and nontradables.…”

“…Starting from the seminal contribution of Moretti (2010), a growing body of empirical studies has estimated local employment multipliers (Adams, 2016;Bartik & Sotherland, 2019;van Dijk, 2017van Dijk, , 2018Fuller & Chapman, 2018;Gerolimetto & Magrini, 2016;Jones & Yang, 2018;Partridge, Rickman, Rose Olfert, & Tan, 2017;Tsvetkova & Partridge, 2016). Moretti (2010) used the shift-share instrumental variable's strategy to estimate the local multiplier effect in the US between 1980 and 2000.…”

Two tales of two U.S. states: Regional fiscal austerity and economic performance

“…In the following formulation, the national and industry mix effects are combined (im): The sum of the industry mix effects across industries (i), including overall national growth, is the predicted change in regional employment from period 0 to t that is attributable to its employment composition of industries in time 0, which then is converted to a rate of change. The industry mix effect reflects employment effects of international trade shocks, national productivity shocks and national industry restructuring (Partridge et al, 2017) and is often used as an exogenous instrument for employment growth (e.g., Bartik, 1991;Moretti, 2010).…”

The response of U.S. regional demographics to import shocks

“…The 2011 5-year sample combines the 1-year ACS samples for each year from 2007 to 2011; we use this and the 2000 decennial Census sample to construct the first differences for the 2000-2007 period. Covariates include black proportion, manufacturing employment share, expected employment growth based on industry mix (following Partridge et al, 2017), and educational attainment group shares unconditional on population. This is done to make the sample size comparable to the decennial Census samples (which represent 5% of the population, whereas the 1-year ACS samples capture 1%).…”