Trade and Worker Deskilling Rui Costa 1 , Swati Dhingra 1 , 2 and - - PowerPoint PPT Presentation

Trade and Worker Deskilling

Rui Costa1, Swati Dhingra1,2 and Stephen Machin1,2

1Centre for Economic Performance, London School of Economics 2Department of Economics, London School of Economics

April 2019

Motivation

• Globalisation has been pinpointed as one of the possible culprits of low

productivity growth and real wage stagnation in many developed economies.

• This has motivated work looking at the impact of international trade on the

labour market, but many fundamental questions are far from resolved.

• This paper offers complementary new evidence on connections between trade

and the labour market, taking an empirical approach that differs from existing work.

Approach and Contribution

• The focus is on a big world event that sent shock waves globally – the EU

referendum vote of June 23, 2016 in the UK.

• The unexpected result of the vote induced an unprecedented overnight fall in

the value of sterling.

• Biggest drop ever among the world’s four major currencies since the

collapse of Bretton Woods.

• Unexpected vote-induced sterling depreciation → Trade→ Workers

Trade and Worker Outcomes

• Exports are generally thought of as improving worker outcomes, but the effect
• f imports on workers is less clear.
• Early studies report a negative impact of imports on workers (Grossman 1987,

Revenga 1992), but there is mixed evidence as the nature of imports has changed.

• There has been a huge rise in trade in intermediate goods and services,

and they now comprise two thirds of international trade. (Johnson and Noguera 2012)

• Workers may be hurt through easier offshoring of tasks, but they can

benefit through lower production costs from offshoring. (Grossman and Rossi-Hansberg 2006)

• Imports hurt workers if they substitute for domestic labour, but can benefit

them if they are complementary: empirically, endogeneity and anticipation are “first-order” concerns for studying causal effects of trade

• n workers. (Goldberg and Pavcnik 2016)

Summary

• Based on the pre-referendum trade structure and the referendum-induced

sterling depreciation across currencies, workers in industries more exposed to intermediate imports-weighted sterling depreciation → Saw a bigger rise in their industry’s intermediate import price → Experienced bigger cuts in wages and job-related education training.

• A 1% higher increase in the price of intermediate imports of an industry

lowered wages by 0.35-0.55%, implying complementarity among intermediate imports and workers.

Real World (Yesterday)

Matthew Naylor, managing director of Naylor’s Flowers, said: the drop in the value

• f the pound had reduced wages for workers. "On the ground, as a British employer,

we don’t have the reputation we used to have as a place to better yourself," added Mr

• Naylor. (Sky News 2019)

Advances

• This paper advances trade and labour market research by:
• Studying the impact of trade on all workers in the economy, not just in
• manufacturing. (Liu and Trefler 2011, Ebenstein et al. 2014)
• Utilising the large currency depreciation to lever plausible exogenous

variation in imports, including in services. (Hummels et al. 2018)

• Providing the first evidence of actual worker-level outcomes being

adversely affected by the current surge in nationalist politics (Auer et al. 2018, Fajgelbaum et al. 2019)

• The content of this work relates to two strands of literature
• Impacts of trade, intermediate inputs/offshoring and exchange rates on

labour market outcomes (Feenstra and Hanson 1999, Campa and Goldberg 2001, Pierce and Schott 2016)

• Actual impacts of nationalist politics (Breinlich et al. 2017, Amiti et al.

2019)

Structure of the Talk

• Context of Events of June 23/24, 2016
• Research Design and Data
• Results and Discussion
• Conclusions

EU Referendum

• On June 23 2016, UK voters were asked: “Should the United Kingdom remain

a member of the European Union or leave the European Union?”

• By 10pm, polling stations across the country closed and YouGov polls

suggested a 52% Remain win.

• By 10.15pm, UKIP’s Nigel Farage conceded Remain "will edge it,”

sending sterling to its 2016 high of 1.5 against the dollar.

• At 12.20am on June 24, Sunderland voted to Leave by a significant
• margin. Sterling dropped more than on Black Wednesday.
• By 1.30am, Brexit showed up in Google searches and bookmakers

changed their odds in favour of Leave winning the referendum.

• Other Leave wins followed and by 2.17am, Nigel Farage tweeted he is "so

happy with the results in North East England".

• 72% of eligible voters cast a ballot and 52% of these chose “Leave”.
• Country ranking of depreciation is consistent with flight for safe haven assets
• Fell more against Japanese Yen, US Dollar.
• Forex investors moved to Gold, Government Bonds.

Sterling depreciation from the Brexit news shock Figure 1 Table 1

EU Referendum - Exchange Rates

EU Referendum Sterling Depreciation Across Countries

Country Currency % Country Currency % Japan Japanese Yen

• 11.1

Israel New Israeli Sheqel

• 6.8

United States US Dollar

• 8.0

Switzerland Swiss Franc

• 6.6

Saudi Arabia Saudi Riyal

• 8.0

Turkey Turkish Lira

• 6.5

Hong Kong Hong Kong Dollar

• 7.9

Malaysia Malaysian Ringgit

• 6.3

Thailand Thai Baht

• 7.6

Denmark Danish Krone

• 6.1

China Chinese Yuan

• 7.5

Euro Zone Euro

• 6.0

Singapore Singapore Dollar

• 7.4

Czech Republic Czech Koruna

• 5.9

Taiwan Taiwan Dollar

• 7.2

South Korea Korean Won

• 5.7

Russia Russian Ruble

• 7.2

South Africa South African Rand

• 5.3

India Indian Rupee

• 7.1

Hungary Hungarian Forint

• 5.2

New Zealand New Zealand Dollar

• 7.1

Norway Norwegian Krone

• 5.2

Australia Australian Dollar

• 6.9

Sweden Swedish Krona

• 5.1

Canada Canadian Dollar

• 6.9

Poland Polish Zloty

• 4.3

Research Design

• Initial trade-weighted depreciation for each industry→Trade prices → Wages

and job-related education & training.

• The reduced form approach has some similarities to recent papers on trade and

labour markets, which typically examine worker outcomes with initial trade or trade policy exposure as explanatory variable across manufacturing industries (e.g. Schott and Pierce 2016).

Data

• The paper uses individual and industry-level data for the UK, combined with

trade data from a variety of sources

• Worker outcomes - wages and training - come from the quarterly Labour Force

Survey (LFS)

• The time-frame of the analysis covers four years (sixteen quarters)

pre-referendum (2012Q3-2016Q2) and eight quarters post-referendum (2016Q3-2018Q2)

• Data on trade comes from the Office of National Statistics (ONS), including

both publicly available data and customized data obtained through freedom of information requests.

• The import and export price indices are from ONS data (current price and

chained volume measures).

• In order to construct the relevant industry-specific trade measures (85

industries), we use data from UN COMTRADE and International Trade in Services (ITIS) for the year 2015 combined with the 2014 Input-Output table from ONS.

• Observe trade shares Sdxo, Ssio for services trade; SsiSio for goods trade.

Difference-in-difference Estimates Table 2 Intermediate imports weighted depreciation Table 3 Exports weighted appreciation Figures 1 to 4 Trade prices and Worker outcomes

Results - Intermediate Imports Weighted Depreciation

Post - Pre Referendum Changes Intermediate Import Prices Export Prices Wages Training (1) (2) (3) (4) Highest Quintile 0.081 0.105 0.060

• 0.008

(0.008) (0.010) (0.007) (0.003) 2nd Highest Quintile 0.043 0.118 0.075 0.003 (0.009) (0.009) (0.010) (0.002) Middle Quintile 0.074 0.108 0.077

• 0.002

(0.004) (0.012) (0.008) (0.003) 2nd Lowest Quintile 0.052 0.101 0.087

• 0.003

(0.015) (0.013) (0.007) (0.003) Lowest Quintile 0.037 0.102 0.087

• 0.001

(0.007) (0.009) (0.008) (0.001) Difference-in-Differences 0.044

• 0.006
• 0.027
• 0.007

(Highest - Lowest Quintile) (0.010) (0.015) (0.010) (0.003)

Results - Exports Weighted Appreciation

Post - Pre Referendum Changes Intermediate Import Prices Export Prices Wages Training (1) (2) (3) (4) Highest Quintile 0.052 0.097 0.077

• 0.004

(0.017) (0.004) (0.009) (0.003) 2nd Highest Quintile 0.062 0.125 0.081

• 0.001

(0.013) (0.015) (0.009) (0.002) Middle Quintile 0.047 0.108 0.064

• 0.004

(0.009) (0.006) (0.011) (0.002) 2nd Lowest Quintile 0.070 0.107 0.073 0.002 (0.007) (0.013) (0.008) (0.002) Lowest Quintile 0.051 0.096 0.089

• 0.002

(0.010) (0.007) (0.009) (0.002) Difference-in-Differences 0.001 0.001

• 0.011
• 0.003

(Highest - Lowest Quintile) (0.020) (0.008) (0.013) (0.003)

Results - Intermediate Import Prices

Results - Export Prices

Results - Wages

Results - Training

Reduced Form Evidence Table 4 Trade Prices Figure 5 Intermediate Import Price Trends Figure 6 Wage Trends

Reduced Form/First Stage - Trade Prices and Exchange Rates

Full Sample Restricted Samples Excluding Excluding Announcement QE (1) (2) (3)

• A. Intermediate Import Prices

Log(Intermediate Imports Weighted 0.375 0.411 0.526 Depreciation) X Post-Referendum (0.100) (0.107) (0.137)

• B. Export Prices

Log(Intermediate Imports Weighted 0.164 0.171 0.233 Depreciation) X Post-Referendum (0.111) (0.114) (0.136) Log(Exports Weighted 0.006 0.007 0.009 Appreciation) X Post-Referendum (0.016) (0.016) (0.019) Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 1870 1360

Pre/Post Trends in Intermediate Import Prices

Pre/Post Trends in Wages

Reduced Form and IV Estimates Tables 5 and 6 Wages Tables 7 and 8 Training

Wages and Trade, A

• A. Full Sample

Reduced Form IV Reduced Form IV (1) (2) (3) (4) Log(Intermediate Imports Weighted

• 0.201
• 0.204

Depreciation) X Post-Referendum (0.075) (0.074) Log(Exports Weighted

• 0.013

Appreciation) X Post-Referendum (0.014) Log(Intermediate Import Prices)

• 0.515
• 0.545

(0.251) (0.226) Log (Export Prices) 0.075 (0.101) First-Stage F Statistic 14.1 4.6 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 123,489 123,489 123,489 123,489

Wages and Trade, B

• B. Restricted Samples

Exclude Announcement Exclude QE Reduced Form IV Reduced Form IV (5) (6) (7) (8) Log(Intermediate Imports Weighted

• 0.206
• 0.190

Depreciation) X Post-Referendum (0.074) (0.087) Log(Intermediate Import Prices)

• 0.483
• 0.343

(0.221) (0.175) First-Stage F Statistic 14.1 14.0 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 117,380 117,380 92,492 92,492

Training and Trade, A

• A. Full Sample

Reduced Form IV Reduced Form IV (1) (2) (3) (4) Log(Intermediate Imports Weighted

• 0.069
• 0.071

Depreciation) X Post-Referendum (0.021) (0.022) Log(Exports Weighted

• 0.001

Appreciation) X Post-Referendum (0.007) Log(Intermediate Import Prices)

• 0.184
• 0.196

(0.073) (0.076) Log (Export Prices) 0.024 (0.044) First-Stage F Statistic 14.1 4.9 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 578,282 578,282 578,282 578,282

Training and Trade, B

• B. Restricted Samples

Exclude Announcement Exclude QE Reduced Form IV Reduced Form IV (5) (6) (7) (8) Log(Intermediate Imports Weighted

• 0.073
• 0.048

Depreciation) X Post-Referendum (0.022) (0.022) Log(Intermediate Import Prices)

• 0.177
• 0.090

(0.073) (0.053) First-Stage F Statistic 14.8 14.7 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 548,079 548,079 434,431 434,431

Skill Variations: Wages

Full Sample Restricted Samples Excluding Excluding Announcement QE IV IV IV (1) (2) (3) Log(Price of Intermediate Imports)× Graduates

• 0.616
• 0.607
• 0.601

(0.216) (0.191) (0.176) Log(Price of Intermediate Imports)× Non-Graduates

• 0.216
• 0.184
• 0.042

(0.252) (0.225) (0.196) First-Stage F-Statistic 6.8 7.3 12.8 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 123,489 117,380 92,492

Skill Variations: Training

Full Sample Restricted Samples Excluding Excluding Announcement QE IV IV IV (1) (2) (3) Log(Price of Intermediate Imports)× Graduates

• 0.205
• 0.199
• 0.148

(0.067) (0.067) (0.061) Log(Price of Intermediate Imports)× Non-Graduates

• 0.118
• 0.124
• 0.023

(0.085) (0.087) (0.074) First-Stage F-Statistic 6.9 7.4 11.3 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 578,282 548,079 434,431

Comparison - Offshoring Studies

• dlogW/dlogPM = ˆ

θMW = [−0.35, −0.55]

• Offhsoring studies typically use trade values, instead of prices. Import demand

elasticity for intermediates in Alfaro et al (forthcoming) is -8.6.

• Compare with Hummels et al. (2014) study of offshoring rise on Danish

matched employer-employee manufacturing data. Estimated Wage Elasticity ˆ θMW /(−7.6) Hummels et al estimate (1) (2) Graduate 0.08 0.03 to 0.07 Non-Graduate 0 to -0.03

• 0.06 to -0.11

Graduate/Non-Graduate 0.08 to 0.11 0.09 to 0.18

Discussion

• The reduction in wages and training after the Brexit vote is the first piece of

evidence on the worker impacts of the recent surge in economic nationalism.

• The rationale that protectionist barriers will protect domestic production and

undo the economic damage suffered by those who lost out earlier needs to account for the sharp rise in the fragmentation of production into global value chains.

• In the long-term this disinvestment can jeopardize potential worker earnings

and human capital development, particularly in the context of an economy with chronic skill shortages like the UK.

Conclusions

• The depreciation that resulted from the Brexit vote has significantly increased

the costs for industries relying more on global value chains in their production.

• The response to this cost shock was a significant price pass through and the

resultant higher costs of intermediate imports caused a relative fall in the training and wages of workers in the most affected industries.

• The decision to leave the EU (not even the actual exit) has already detrimentally

affected workers’ welfare in a way contrary to the beliefs expressed by some about the likely repercussions from the leave vote.

Origin Country Structure - Intermediate Import Shares

Destination Country Structure - Export Shares

Robustness - Inclusion of Pre-Referendum Import Shares, Wages Reduced Form

Intermediate Wages Import Prices First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.383

• 0.205

Depreciation) X Post-Referendum (0.107) (0.072) Share of Imports in

• 0.035
• 0.026

Intermediates X Post-Referendum (0.037) (0.025) Log(Intermediate Import Prices)

• 0.430

(0.210) First-Stage F Statistic 8.3 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 123,489 123,489

Robustness - Inclusion of Pre-Referendum Import Shares, Training Reduced Form

Intermediate Training Import Prices First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.371

• 0.068

Depreciation) X Post-Referendum (0.101) (0.021) Share of Imports in

• 0.035

0.003 Intermediates X Post-Referendum (0.035) (0.009) Log(Intermediate Import Prices)

• 0.177

(0.068) First-Stage F Statistic 8.5 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 578,282 578,282

Robustness - Inclusion of Pre-Referendum Import Shares, Wages Reduced Form

Intermediate Wages Import Prices First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.383

• 0.205

Depreciation) X Post-Referendum (0.107) (0.072) Share of Imports in

• 0.035
• 0.026

Intermediates X Post-Referendum (0.037) (0.025) Log(Intermediate Import Prices)

• 0.430

(0.210) First-Stage F Statistic 8.3 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 123,489 123,489

Robustness - Inclusion of Pre-Referendum Import Shares, Training Reduced Form

Intermediate Training Import Prices First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.371

• 0.068

Depreciation) X Post-Referendum (0.101) (0.021) Share of Imports in

• 0.035

0.003 Intermediates X Post-Referendum (0.035) (0.009) Log(Intermediate Import Prices)

• 0.177

(0.068) First-Stage F Statistic 8.5 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 578,282 578,282

Robustness - Intermediate Import Prices Reduced Form with Import Competition Channel

Full Sample Restricted Samples Excluding Excluding Announcement QE (1) (2) (3) (4) (5) Log(Intermediate Imports Weighted 0.375 0.350 0.383 0.490 Depreciation) X Post-Referendum (0.100) (0.091) (0.098) (0.128) Log(Total Imports Weighted 0.169 0.147 0.165 0.214 Depreciation) X Post-Referendum (0.050) (0.046) (0.048) (0.059) Sample Size 2,040 2,040 2040 1870 1360

Robustness - Wages Reduced Form with Import Competition Channel

Full Sample Restricted Samples Excluding Excluding Announcement QE (1) (2) (3) (4) (5) Log(Intermediate Imports Weighted

• 0.201
• 0.187
• 0.192
• 0.177

Depreciation) X Post-Referendum (0.075) (0.076) (0.075) (0.089) Log(Total Imports Weighted

• 0.081
• 0.068
• 0.074
• 0.066

Depreciation) X Post-Referendum (0.054) (0.052) (0.051) (0.055) Controls Yes Yes Yes Yes Yes Time Dummies, Industry Dummies Yes Yes Yes Yes Yes Sample Size 123,489 123,489 123,489 117,380 92,492

Robustness - Training Reduced Form with Import Competition Channel

Full Sample Restricted Samples Excluding Excluding Announcement QE (1) (2) (3) (4) (5) Log(Intermediate Imports Weighted

• 0.069
• 0.060
• 0.065
• 0.041

Depreciation) X Post-Referendum (0.021) (0.020) (0.022) (0.022) Log(Total Imports Weighted

• 0.055
• 0.051
• 0.046
• 0.036

Depreciation) X Post-Referendum (0.017) (0.015) (0.017) (0.017) Controls Yes Yes Yes Yes Yes Time Dummies, Industry Dummies Yes Yes Yes Yes Yes Sample Size 578,282 578,282 578,282 548,079 434,431

Robustness - Pre-Post Referendum Changes in Trade Quantities

Full Sample Restricted Samples Excluding Excluding Announcement QE (1) (2) (3)

• A. Intermediate Imports

Log(Intermediate Imports Weighted 0.072 0.088 0.115 Depreciation) X Post-Referendum (0.079) (0.085) (0.102)

• B. Exports

Log(Intermediate Imports Weighted

• 0.461
• 0.493
• 0.639

Depreciation) X Post-Referendum (0.289) (0.305) (0.366) Log(Exports Weighted 0.010 0.009 0.004 Appreciation) X Post-Referendum (0.053) (0.057) (0.066) Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 1870 1360

EU Referendum - Exchange Rates - 7-Day and 15-Day

(a) 7-Day (b) 15-Day

Robustness - Wages with 7-Day Exchange Rate Window

Intermediate Import Prices Wages First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.383

• 0.189

Depreciation) X Post-Referendum (0.116) (0.074) Log(Intermediate Import Prices)

• 0.495

(0.270) First-Stage F Statistic 10.9 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 123,489 123,489

Robustness - Training with 7-Day Exchange Rate Window

Intermediate Import Prices Training First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.370

• 0.073

Depreciation) X Post-Referendum (0.113) (0.024) Log(Intermediate Import Prices)

• 0.197

(0.084) First-Stage F Statistic 10.8 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 578,282 578,282

Robustness - Wages with 15-Day Exchange Rate Window

Intermediate Import Prices Wages First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.402

• 0.102

Depreciation) X Post-Referendum (0.129) (0.054) Log(Intermediate Import Prices)

• 0.252

(0.169) First-Stage F Statistic 9.2 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 123,489 123,489

Robustness - Training with 15-Day Exchange Rate Window

Intermediate Import Prices Training First Stage Reduced Form IV (1) (2) (3) Log(Intermediate Imports Weighted 0.386

• 0.052

Depreciation) X Post-Referendum (0.127) (0.024) Log(Intermediate Import Prices)

• 0.136

(0.076) First-Stage F Statistic 9.2 Controls, Time Dummies, Industry Dummies Yes Yes Yes Sample Size 2040 578,282 578,282

Hours and Trade

• A. Full Sample

Reduced Form IV Reduced Form IV (1) (2) (3) (4) Log(Intermediate Imports Weighted

• 0.052
• 0.051

Depreciation) X Post-Referendum (0.032) (0.032) Log(Exports Weighted 0.009 Appreciation) X Post-Referendum (0.007) Log(Intermediate Import Prices)

• 0.137
• 0.103

(0.105) (0.104) Log (Export Prices)

• 0.069

(0.039) First-Stage F Statistic 14.3 5.0 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 571,335 571,335 571,335 571,335

Hours and Trade

• B. Restricted Samples

Exclude Announcement Exclude QE Reduced Form IV Reduced Form IV (5) (6) (7) (8) Log(Intermediate Imports Weighted

• 0.050
• 0.061

Depreciation) X Post-Referendum (0.033) (0.039) Log(Intermediate Import Prices)

• 0.122
• 0.116

(0.097) (0.088) First-Stage F Statistic 15.0 14.9 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 541,567 541,567 429,214 429,214

Overtime Hours and Trade

• A. Full Sample

Reduced Form IV Reduced Form IV (1) (2) (3) (4) Log(Intermediate Imports Weighted

• 0.147
• 0.140

Depreciation) X Post-Referendum (0.049) (0.051) Log(Exports Weighted

• 0.017

Appreciation) X Post-Referendum (0.001) Log(Intermediate Import Prices)

• 0.392
• 0.314

(0.176) (0.168) Log (Export Prices)

• 0.122

(0.049) First-Stage F Statistic 14.1 4.9 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 578,282 578,282 578,282 578,282

Overtime Hours and Trade

• B. Restricted Samples

Exclude Announcement Exclude QE Reduced Form IV Reduced Form IV (5) (6) (7) (8) Log(Intermediate Imports Weighted

• 0.141
• 0.150

Depreciation) X Post-Referendum (0.052) (0.051) Log(Intermediate Import Prices)

• 0.342
• 0.285

(0.163) (0.120) First-Stage F Statistic 14.8 14.7 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 548,079 548,079 434,431 434,431

Robustness - Wages Of All Workers

• A. Full Sample

Reduced Form IV Reduced Form IV (1) (2) (3) (4) Log(Intermediate Imports Weighted

• 0.247
• 0.251

Depreciation) X Post-Referendum (0.082) (0.082) Log(Exports Weighted

• 0.009

Appreciation) X Post-Referendum (0.012) Log(Intermediate Import Prices)

• 0.652
• 0.661

(0.312) (0.287) Log (Export Prices) 0.029 (0.142) First-Stage F Statistic 14.4 4.9 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 159,957 159,957 159,957 159,957

Robustness - Wages Of All Workers

• B. Restricted Samples

Exclude Announcement Exclude QE Reduced Form IV Reduced Form IV (5) (6) (7) (8) Log(Intermediate Imports Weighted

• 0.251
• 0.253

Depreciation) X Post-Referendum (0.080) (0.103) Log(Intermediate Import Prices)

• 0.607
• 0.478

(0.271) (0.233) First-Stage F Statistic 14.9 14.8 Controls, Time Dummies, Industry Dummies Yes Yes Yes Yes Sample Size 152,065 152,065 119,953 119,953

Robustness - Time Window and Alternative Training Measure

Excluding Observations Received training in Close to Referendum Last Three Months Wages Training Training - 3 months IV IV IV (1) (2) (3) Log(Intermediate Import

• 0.489
• 0.176
• 0.319

Prices) X Post-Referendum (0.252) (0.071) (0.139) First-Stage F Statistic 13.5 13.9 14.1 Controls Yes Yes Yes Time and Industry Dummies Yes Yes Yes Sample Size 121,954 571,049 578,282