The Surprisingly Swift Decline of U.S. Manufacturing Employment Justin R. Pierce Board of Governors of the Federal Reserve System Peter K. Schott Yale School of Management & NBER

Disclaimer Any opinions and conclusions expressed herein are those of the authors and do not necessarily represent the views of the U.S. Census Bureau, the Board of Governors or its research staff. All results have been reviewed to ensure that no confidential information is disclosed. 2

Post-War U.S. Manufacturing Employment 3

4

mill over 3 years

Post-War U.S. Manufacturing Employment mill over 3 years

Introduction The sharp decline in US manufacturing employment since 2001 is closely linked to a change in US trade policy: –Chinas receipt of Permanent Normal Trade Relations in late 2000 PNTR did not change actual tariff rates: Chinese imports were already eligible for low NTR rates typically reserved for WTO members –But for China, NTR required contentious annual renewals –Failure would increase tariffs to Smoot-Hawley levels The potential for large tariff increases likely discouraged: –US firms from openings plants in China –Chinese firms from making investments to export to US With PNTR, the possibility of future tariff hikes was eliminated 7

Before Continuing… 8 …note that manufacturing real value added continues to rise at the historical pace

Outline US-China Trade Policy Data Baseline results Alternate explanations Additional results Conclusion 9

US NTR and Non-NTR Tariffs NTR = Normal Trade Relations –Synonym for Most Favored Nation (MFN) The US has two basic tariff schedules –NTR tariffs: for WTO members; generally low –Non-NTR tariffs : for non-market economies; generally high; set by Smoot-Hawley (1930) So how does China fit into these categories? 10

US-China Trade Policy, (February) China was granted temporary NTR status by the US Congress Temporary NTR requires annual re-approval by Congress 2000 (October) U.S. Congress grants China PNTR, eliminating the risk that a failed vote might lead to a jump in tariffs 2001 (December) China enters WTO Annual renewals of MFN status were uncertain

Measuring the Policy Change Measure the effect of the policy as: –NTR Gap = Non-NTR Tariff – NTR Tariff Measures extent to which tariffs could increase prior to PNTR Varies across industries We can preview the results in two simple figures that use public data 12

Preview of Findings – Employment Public NBER-CES Data 13 High- and low-gap industries follow roughly parallel trends in two decades prior to PNTR

Preview of Findings – Employment Public NBER-CES Data 14 High- and low-gap industries follow roughly parallel trends in two decades prior to PNTR After PNTR the series diverge with employment falling most sharply in the high-gap industries most affected by PNTR (Note: we use the gap as a continuous variable in our regression analysis.)

Preview of Findings – Trade Public Census Trade Data 15 Divergence is also evident in trade data

Preview of Findings – Trade Public Census Trade Data 16 Divergence is also evident in trade data Imports from China in the more- exposed products jump after PNTR

17 Divergence is also evident in trade data Imports from China in the more- exposed products jump after PNTR This trend is not present in imports from rest-of-world (ROW) Preview of Findings – Trade Public Census Trade Data

18 Divergence is also evident in trade data Imports from China in the more- exposed products jump after PNTR This trend is not present in imports from rest-of-world (ROW) Find similar results for number of U.S. importers, Chinese exporters and importer-exporter pairs Preview of Findings – Trade Public Census Trade Data

Related Research Employment and trade liberalization –Lots of papers –Autor et al. (2012); Bloom et al. (2012) Investment under uncertainty –Lots of papers –Trade: Handley (2012); Handley and Limao (2012, 2013) Jobless recoveries –Manufacturing: Faberman (2012) –Overall: Jaimovich and Siu (2012) Supply-chain linkages –US manufacturing: Ellison, Glaeser and Kerr (2010) –Trade: Baldwin and Venables (2012) 19

Outline US-China Trade Policy Data Baseline results Alternate explanations Additional results Conclusion 20

NTR Gaps NTR Gap = Non-NTR Tariff – NTR Tariff Compute for each HS8 product using the ad valorem equivalent NTR and non-NTR rates from Feenstra, Romalis and Schott (2003) available for The NTR Gap for industry i is the mean over the gaps of the HS8s in that industry 21

Distribution of 1999 NTR Gap 22 The gap is large in economic terms Varies substantially across industries, allowing for identification of effect of PNTR 89 percent of the variation in the NTR gap across industries arises from variation in non-NTR rates, set in 1930 Mean: 0.32 Std: 0.15

Census Data LBDxxxxxxxx xxxxxxxxxxxx CMxxxxxxxxx xxxxxxxxxxxx LFTTD Annual employment of all U.S. establishments, Employment + other attributes for all manufacturing establishments every five years, 1977(5)2007 Transaction-level US import data: value, importer ID, foreign exporter ID 23

Outline US-China Trade Policy Data Baseline results Alternate explanations Additional results Conclusion 24

Empirical Strategy We use a difference-in-differences strategy to examine the link between PNTR/WTO and U.S. manufacturing employment outcomes – 1 st difference: industries with higher vs lower NTR Gaps – 2 nd difference: outcomes after 2001 vs before PNTR coincides with the 2001 peak, so compare employment d years after 2001 with employment d years after the 1990 peak 25

Industry-Level OLS Diff-in-Diff Using the LBD (i=industry; t=NBER peak {1990,2001}; d=1:6 years after peak) 26 Cumulative percent change in industry i employment d years after NBER peak t={1981,1990,2001} DID Term Interaction of indicator variable for 2001 peak and continuous, time-invariant own-industry NTR Gap Industry and peak-year fixed effects (control for cyclicality) Industry attributes Separate regression for d=1:6 years after each peak

Basic Industry-Level Regressions Bold=statistically significant at 10% level 27 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR

Basic Industry-Level Regressions Bold=statistically significant at 10% level 28 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR Use LBD to examine outcomes 1:6 years after peak, e.g., compare , , and

Basic Industry-Level Regressions Bold=statistically significant at 10% level 29 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR Industries with higher NTR gaps experience larger employment declines following PNTR, as expected

Basic Industry-Level Regressions Bold=statistically significant at 10% level 30 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR Absolute magnitude of DID coefficient rises over time; i.e., relative losses are persistent

Basic Industry-Level Regressions Bold=statistically significant at 10% level 31 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR Effect attenuated in high K/L industries, magnified in high S/L industries But these controls generally are not statistically significant

Basic Industry-Level Regressions Bold=statistically significant at 10% level 32 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i ln(K/L it ) ln(S/L it ) Observations 652 R Fixed Effects i,t Employment Weighted Yes Implied Impact of PNTR Multiply DID coefficient by average NTR gap to assess implied impact of PNTR Post-2001 growth is 3.4 to 15.6 percentage points lower than post growth

Outline PNTR Data Baseline results Alternate explanations Additional results Conclusion 33

Alternate Explanations Alternate explanations must explain: –Timing: employment declines and Chinese imports rise with PNTR in 2001 –Variation across industries: outcomes are larger for industries most affected by the policy change We consider a wide range of stories… 34

Alternate Explanations Changes in Chinese Policy –Lower import tariffs –Elimination of export licensing requirements –Elimination of production subsides –Reduced barriers to foreign investment Union Resistance in the US Popped US tech bubble Rising Chinese competitiveness End of Textile and Clothing Quotas 35

Alternate Explanations Changes in Chinese Policy –Lower import tariffs – Brandt et al. (2013) –Elimination of export licensing requirements – Bai et al. (2007) –Elimination of production subsides – Girma et al. (2007) –Reduced barriers to foreign investment – Nunn (2007) Union Resistance in the US – unionstats.org Popped US tech bubble – IT dummy; control for prior growth Rising Chinese competitiveness – capital and skill intensity End of Textile and Clothing Quotas – Khandelwal et al. (2013) 36

Correlation of 1999 Gap with Other Industry Attributes Bold=statistically significant at 10% level NTR Gap is: Negatively correlated with K/L, S/L, change in Chinese subsidies, and US union membership Positively correlated with contractibility, share of Chinese firms eligible for export licenses, MFA dummy, and advanced technology indicator Dependent variable: 1999 NTR Gap ln(K/L) ln(S/L) Nunn s Contract Intensity Chinese Import Tariffs ( ) Chinese Subsidy ( ) Share of Chinese Firms Eligible to Export (1999) {MFA Apparel} Union Membership {Advanced Technology Products} {Anti-Dumping Filings, } Prior Growth NTR Industries326 R-squared Covariate Mean Covariate Std Dev Notes: Table reports the results of industry-level OLS regressions summarizing the relationship between the 1999 NTR gap and noted industry attributes. Coefficient for constant is suppressed. See text for a discussion of these attributes and their sources. 37

Full Specification Where possible, we include all these covariates and their interactions with a post-PNTR dummy to allow for potential changes in relationships after 2001 These interactions yield a very flexible specification 38 Now Includes 1999 NTR Gap (as before) & All other industry attributes

Alternate Explanations Changes in Chinese Policy –Lower import tariffs – Brandt et al. (2013) - o –Elimination of export licensing requirements – Bai et al. (2007) - x –Elimination of production subsides – Girma et al. (2007) - x –Reduced barriers to foreign investment – Nunn (2007) - x Union Resistance in the US – unionstats.org - x Popped US tech bubble – IT dummy; control for prior growth - o Rising Chinese competitiveness – capital and skill intensity - o End of Textile and Clothing Quotas – Khandelwal et al. (2013) - x 39

40 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i {post-PNTR} x Contract Intensity i {post-PNTR} x China Import Tariffs i {post-PNTR} x China Licensing i {post-PNTR} x China Subsidies i Anti-Dumping Filings i {post-PNTR} x Anti-Dumping Filings i {post-PNTR} x 1{Advanced Tech i } {post-PNTR} x 1{MFA Apparel} i ln(K/L it ) {post-PNTR} x ln(K/L it ) ln(S/L it ) {post-PNTR} x ln(S/L it ) Union Membership it {post-PNTR} x Union Membership it Prior Growth it {post-PNTR} x Prior Growth it NTR i Observations 652 R Fixed Effects i,t Employment Weighted Yes

41 Percent Change in Industry Employment Years After NBER Peak (LBD) {post-PNTR} x NTR Gap i {post-PNTR} x Contract Intensity i {post-PNTR} x China Import Tariffs i {post-PNTR} x China Licensing i {post-PNTR} x China Subsidies i Anti-Dumping Filings i {post-PNTR} x Anti-Dumping Filings i {post-PNTR} x 1{Advanced Tech i } {post-PNTR} x 1{MFA Apparel} i ln(K/L it ) {post-PNTR} x ln(K/L it ) ln(S/L it ) {post-PNTR} x ln(S/L it ) Union Membership it {post-PNTR} x Union Membership it Prior Growth it {post-PNTR} x Prior Growth it NTR i Observations 652 R Fixed Effects i,t Employment Weighted Yes DID term remains negative and significant

Implied Impact of PNTR 42 Estimated impact of PNTR is reduced when controlling for alternate explanations but remains substantial

Placebos & Alternate DID Specification Estimate relationship between NTR Gap and employment growth separately in pre-PNTR period and post-PNTR period 43

44 Percent Change in Industry Employment Years After 1990 PeakYears After 2001 Peak NTR Gap i Contract Intensity i China Import Tariffs i China Licensing i China Subsidies i Anti-Dumping Filings i {Advanced Tech i } {MFA Apparel} i ln(K/L it ) ln(S/L it ) Union Membership it Prior Growth it NTR i Observations 326 R Fixed Effects i,t Employment Weighted Yes Notes: Each column displays the results of an OLS regression of the cumulative percent change in industry (i) employment after 1990 on the noted covariates (see text). Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. Estimates for the constant and fixed effects are suppressed.

45 Percent Change in Industry Employment Years After 1990 PeakYears After 2001 Peak NTR Gap i Contract Intensity i China Import Tariffs i China Licensing i China Subsidies i Anti-Dumping Filings i {Advanced Tech i } {MFA Apparel} i ln(K/L it ) ln(S/L it ) Union Membership it Prior Growth it NTR i Observations 326 R Fixed Effects i,t Employment Weighted Yes Notes: Each column displays the results of an OLS regression of the cumulative percent change in industry (i) employment after 1990 on the noted covariates (see text). Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. Estimates for the constant and fixed effects are suppressed. No relationship between NTR Gap and employment growth in pre-PNTR period Negative and statistically significant relationship in post-PNTR period

Placebos & Alternate DID Specification Estimate relationship between NTR Gap and employment growth separately in pre-PNTR period and post-PNTR period –No relationship pre-PNTR; negative relationship post-PNTR Estimate a DID for the two peaks prior to 2001: 1990 vs

47 Years After NBER Peak (LBD) {t=1990} x NTR Gap i {t=1990} x Contract Intensity i {t=1990} x China Import Tariffs i {t=1990} x China Licensing i {t=1990} x China Subsidies i Anti-Dumping Filings i {t=1990} x 1{Advanced Tech i } {t=1990} x 1{MFA Apparel} i ln(K/L it ) {t=1990} x ln(K/L it ) ln(S/L it ) {t=1990} x ln(S/L it ) {t=1990} x Union Membership it Union Membership it Prior Growth it {t=1990} x Prior Growth it {t=1990} x NTR i Observations 652 R Fixed Effects i,t Employment Weighted Yes Notes: Each column displays the results of an OLS regression of the cumulative percent change in industry (i) employment on a difference-in-differences term and various industry attributes. There are two observations for each industry corresponding to growth up to six years after the 1981 and 1990 peaks, indexed by t. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. Estimates for the constant and fixed effects are suppressed. No relationship between DID estimator and employment growth in placebo period

Placebos & Alternate DID Specification Estimate relationship between NTR Gap and employment growth separately in pre-PNTR period and post-PNTR period –No relationship pre-PNTR; negative relationship post-PNTR Estimate a DID for the two peaks prior to 2001: 1990 vs 1981 –No relationship Alternate DID specification with annual data: NTR Gap + Other industry attributes

49 ln(Emp it ) 1{year=1991} x NTR Gap i {year=1992} x NTR Gap i {year=1993} x NTR Gap i {year=1994} x NTR Gap i {year=1995} x NTR Gap i {year=1996} x NTR Gap i {year=1997} x NTR Gap i {year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i {year=2006} x NTR Gap i {year=2007} x NTR Gap i Observations 5868 R Employment WeightedYes Fixed Effects i,t Additonal CovariatesNoK/L, S/LAll

50 ln(Emp it ) 1{year=1991} x NTR Gap i {year=1992} x NTR Gap i {year=1993} x NTR Gap i {year=1994} x NTR Gap i {year=1995} x NTR Gap i {year=1996} x NTR Gap i {year=1997} x NTR Gap i {year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i {year=2006} x NTR Gap i {year=2007} x NTR Gap i Observations 5868 R Employment WeightedYes Fixed Effects i,t Additonal CovariatesNoK/L, S/LAll Interactions of year dummies with NTR gap are not significant prior to PNTR

51 ln(Emp it ) 1{year=1991} x NTR Gap i {year=1992} x NTR Gap i {year=1993} x NTR Gap i {year=1994} x NTR Gap i {year=1995} x NTR Gap i {year=1996} x NTR Gap i {year=1997} x NTR Gap i {year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i {year=2006} x NTR Gap i {year=2007} x NTR Gap i Observations 5868 R Employment WeightedYes Fixed Effects i,t Additonal CovariatesNoK/L, S/LAll Interactions of year dummies with NTR gap are not significant prior to PNTR Interactions are negative and significant after PNTR

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain Exposure –Trade Conclusion 52

Other Countries Our paper focuses on effects of a U.S. trade policy Now we compare employment outcomes in the U.S. to those in EU Useful test case because EU did not have the policy change that took place in U.S. –EU granted permanent NTR status to China in 1980, did not have annual renewals We estimate relationship between NTR gap in EU and again in US using an alternative data source (UNIDO) and an alternate specification 53

54 Dependent Variable: ln(Emp it ) EUUS 1{year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i Observations R Employment Weighted Yes Fixed Effects i,t Notes: Each column displays the results of an ISIC-industry (i) level OLS regression of the log of manufacturing employment on year (t) fixed effects, industry fixed effects and interactions of year fixed effects with the 1999 U.S. NTR gap. Coefficient estimates for all but the latter are suppressed. Data are from the UNIDO INDSTAT4 database (see text) for 1997 to U.S. data for 1993 are missing from this dataset. Industries for which data are not available in all years for a particular country are dropped. Regressions are weighted by 1997 employment. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. US versus EU Employment Data from UNIDO; Bold indicates statistical significance

55 Dependent Variable: ln(Emp it ) EUUS 1{year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i Observations R Employment Weighted Yes Fixed Effects i,t Notes: Each column displays the results of an ISIC-industry (i) level OLS regression of the log of manufacturing employment on year (t) fixed effects, industry fixed effects and interactions of year fixed effects with the 1999 U.S. NTR gap. Coefficient estimates for all but the latter are suppressed. Data are from the UNIDO INDSTAT4 database (see text) for 1997 to U.S. data for 1993 are missing from this dataset. Industries for which data are not available in all years for a particular country are dropped. Regressions are weighted by 1997 employment. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. US versus EU Employment Data from UNIDO; Bold indicates statistical significance U.S. No effect of NTR gap prior to PNTR

56 Dependent Variable: ln(Emp it ) EUUS 1{year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i Observations R Employment Weighted Yes Fixed Effects i,t Notes: Each column displays the results of an ISIC-industry (i) level OLS regression of the log of manufacturing employment on year (t) fixed effects, industry fixed effects and interactions of year fixed effects with the 1999 U.S. NTR gap. Coefficient estimates for all but the latter are suppressed. Data are from the UNIDO INDSTAT4 database (see text) for 1997 to U.S. data for 1993 are missing from this dataset. Industries for which data are not available in all years for a particular country are dropped. Regressions are weighted by 1997 employment. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. US versus EU Employment Data from UNIDO; Bold indicates statistical significance U.S. No effect of NTR gap prior to PNTR Negative and significant after PNTR

57 Dependent Variable: ln(Emp it ) EUUS 1{year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i Observations R Employment Weighted Yes Fixed Effects i,t Notes: Each column displays the results of an ISIC-industry (i) level OLS regression of the log of manufacturing employment on year (t) fixed effects, industry fixed effects and interactions of year fixed effects with the 1999 U.S. NTR gap. Coefficient estimates for all but the latter are suppressed. Data are from the UNIDO INDSTAT4 database (see text) for 1997 to U.S. data for 1993 are missing from this dataset. Industries for which data are not available in all years for a particular country are dropped. Regressions are weighted by 1997 employment. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. US versus EU Employment Data from UNIDO; Bold indicates statistical significance U.S. No effect of NTR gap prior to PNTR Negative and significant after PNTR E.U. No effect of NTR gap on employment

58 Dependent Variable: ln(Emp it ) EUUS 1{year=1998} x NTR Gap i {year=1999} x NTR Gap i {year=2000} x NTR Gap i {year=2001} x NTR Gap i {year=2002} x NTR Gap i {year=2003} x NTR Gap i {year=2004} x NTR Gap i {year=2005} x NTR Gap i Observations R Employment Weighted Yes Fixed Effects i,t Notes: Each column displays the results of an ISIC-industry (i) level OLS regression of the log of manufacturing employment on year (t) fixed effects, industry fixed effects and interactions of year fixed effects with the 1999 U.S. NTR gap. Coefficient estimates for all but the latter are suppressed. Data are from the UNIDO INDSTAT4 database (see text) for 1997 to U.S. data for 1993 are missing from this dataset. Industries for which data are not available in all years for a particular country are dropped. Regressions are weighted by 1997 employment. Robust standard errors are displayed below each coefficient. Coefficients in bold are statistically significant at the 10 percent level. US versus EU Employment Data from UNIDO; Bold indicates statistical significance U.S. No effect of NTR gap prior to PNTR Negative and significant after PNTR E.U. No effect of NTR gap on employment The disparity in outcomes provides further evidence against competing explanations: technological change aggregate shocks in China

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain Exposure –Trade Conclusion 59

Examine Other Industry Outcomes Using CM 60 Now turn to CM Only available in years ending in 2, 7 but a rich set of characteristics is available Post-PNTR period is ; pre-PNTR period is

Examine Other Industry Outcomes Using CM 61 Effect on employment is similar to that in primary specification

Examine Other Industry Outcomes Using CM 62 Effect on employment is similar to that in primary specification Effect on production workers nearly twice as large as non-production workers

Examine Other Industry Outcomes Using CM 63 Effect on employment is similar to that in primary specification Effect on production workers nearly twice as large as non-production workers Capital intensity increases, but effect not statistically significant Skill intensity (share of non-production workers) does increase Trade-induced skill-biased technical change?

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain Exposure –Trade Conclusion 64

Margins of Adjustment Job Destruction (JD) – PC: plant contraction at continuing firms – PD: plant death at continuing firms – FD: firm death Job Creation (JC) – PE: plant expansion at continuing firms – PB: plant birth at continuing firms – FB: firm birth 65

Implied Impact of PNTR 66 Contribution of exaggerated job destruction Contribution of anemic job creation JC contributes 17 to 41 percent across Relates to research by Faberman (2012), who notes changes in manufacturing JC and JD rates after 2001

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain Exposure –Trade Conclusion 67

Plant-Level OLS Diff-in-Diff Using the CM 68 Examine effect of PNTR at plant-level –See if industry results are driven by death –Control for heterogeneity within industries Similar specification, but now include plant attributes as controls –TFP, age, capital and skill intensity as well as their interaction with post-PNTR indicator Use plant-level NTR gap –Weighted average gap across the products a plant produces Focus on continuing plants

Plant-Level OLS Diff-in-Diff Using the CM 69 Effect on total employment is similar to the effect found in the industry specification

Plant-Level OLS Diff-in-Diff Using the CM 70 Effect on production workers is again about twice as large as that for non- production workers Production hours decline at a similar magnitude to production employment I.e., the decline in employment is not due to making remaining workers work more

Plant-Level OLS Diff-in-Diff Using the CM 71 Plants with the average NTR gap are 9.2 percent more likely to exit

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain exposure –Trade Conclusion 72

Supply chain exposure For each plant p we compute upstream and downstream NTR gaps using data from the 1997 US input-output table 73 Plant p Upstream Weighted average NTR gap of industries j that supply the industries produced by p, using the values from the IO total requirements table as weights Downstream Weighted average NTR gap of industries k that make use of industries produced by p, using values from the IO total requirements table as weights

Plant-Level OLS Diff-in-Diff Using the CM 74 Upstream and downstream effects are positive and significant for plant death Could reflect supply chain co-location (Baldwin and Venables 2013) In total, own-, up- and downstream NTR gaps associated with a relative increase in the probability of death of 15.4 percent

Outline US-China Trade Policy Data Baseline results & Alternate explanations Additional results –Other Countries –Other Outcomes –Margins of Adjustment –Plant-level –Supply-chain Exposure –Trade Conclusion 75

Effect of PNTR on U.S. Imports from China (c=country; h=HS product) Trade data not available till 1990s So, amend DID specification and estimate a triple difference: – 1 st difference: products with higher vs lower NTR Gaps – 2 nd difference: imports from China versus other trading partners – 3 rd difference: vs Examine import value, number of US importers, number of Chinese exporters, and number of importer-exporter pairs 76 Triple difference estimator Country and product fixed effects Growth rate of value or number of importer, exporters, or pairs from country c Interactions of difference variables

Effect of PNTR on U.S. Imports from China (LFTTD; Bold=statistically significant at 10% level) 77 Post-PNTR US imports from China increase in same goods where domestic employment is lost Growth in number of trading firms is consistent with greater incentives to invest in trading relationships as uncertainty declines

Outline PNTR Data Baseline results Alternate explanations Additional results Conclusion 78

Conclusions Strong link between US manufacturing job loss and PNTR U.S. imports from China increase in same industries where employment declines occur, along with number of U.S. importers, Chinese exporters and importer-exporter pairs Results are robust to inclusion of proxies for wide array of alternate explanations, as well as alternate specifications Our measure of the effect of the policy change (NTR Gap) has no relationship with manufacturing employment in the EU, which was not subject to policy change Effects of PNTR experienced through both elevated job destruction and suppressed job creation, and effect is somewhat magnified through supply chain linkages 79

Thanks 80