Trade Induced Technical Change?The Impact of Chinese Imports on Innovation,

Diffusion and Productivity

Nick Bloom (Stanford)

Mirko Draca (Warwick)

John Van Reenen (LSE & CEP)

Chicago

April 2015

What is the impact of Southern trade on Northern technical change?

• Limited empirical evidence, partly from lack of micro data and partly due to a lack of North-South trade natural experiments

• Theoretical literature also inconclusive (e.g. ambiguous effects of competition on innovation & adoption)

• But this is a major economic and political issue because of the rapid growth of Chinese & low-wage country imports

05

10

15

19

80

19

81

198

21

98

31

98

41

98

51

986

198

71

98

81

98

91

99

01

991

199

21

99

31

99

41

99

51

996

19

97

19

98

19

99

20

00

200

12

00

22

003

20

04

20

05

200

62

00

7

Low wage countries list from Bernard, Jensen and Schott (2006). Countries <5% GDP/capita of US 1972-2001.

Low-wage % of imports in Europe and the US

China

All low wage

Clear political importance – for example tires trade war in 2010

Major election issue in 2012

Summary: we study the impact of Chinese imports on technology in Europe (1/2)Use panel data on ~90,000 firms & establishments in 1990s & 2000s in 12 EU countries. We find that higher threat of Chinese imports leads to:

A) Within-firm increase in volume of innovation (patenting and R&D) as well as greater IT, productivity & management quality

B) A reallocation of jobs towards higher tech firms (as measured by patents, TFP, IT, etc.)

So there is aggregate technological upgrading in the North from liberalization with a lower wage country like China

Results robust to using IV strategy of quota relaxation post China’s entry into the WTO

Overall magnitudes moderate but rising: China “accounts” for:• ≈ 15% of increase in IT, patents & productivity 2000-2007

Suggests the impact on innovation is another positive outcome (alongside static gains) from low wage country trade

Summary: we study the impact of Chinese imports on technology in Europe (2/2)

Some econometric ‘case-studies’ illustrate our results

Freeman and Kleiner (2005) look at a largeUS shoe maker’s response to increasinglow wage country competition

Bartel, Ichniowski & Shaw (2007) on US valve manufacturers’ response to cheaper imports

Bugamelli, Schivardi & Zizza (2008) Italian manufacturers reactions to Chinese import threats

All find very similar changes:• Increased innovation to develop new product ranges• Investment in IT, worker skills and management practices

Caveat: Main empirical analysis is partial equilibrium

Build Endogenous Growth GE model with “Trapped factors”– Some factors of production with firm-specific investment

(skilled labor, management time, specific capital, etc.)– China shock reduces profits from old products– Lower opportunity cost to shifting trapped factors to

innovate in new products Chinese are not making– Amplifies the standard positive effect on innovation from

larger market size

Consistent with case-studies & results from the paper:

1. Import competition increases innovation

2. Mainly within the narrowly impacted industry

3. Main response from imports from low skilled countries

Seems to fit a ‘Trapped-Factor’ innovation model (Bloom, Romer, Terry & Van Reenen, 2014)

Related Literature• Trade & Productivity. De Locker & Goldberg (2014) survey; Empirics: Pavcnik (2002), Trefler (2004), Dunne et al (2008) Theory: Heterogeneous firms Melitz (2003), Arkollakis et al (2008, 2010); Redding & Melitz (2015); product switching: Bernard et al (2007); offshoring Grossman & Rossi-Hansberg (2008)

• Trade & Innovation. Empirics: Bustos (2011), Acemoglu et al (2015); Theory: market size Krugman (1980), Grossman & Helpman (1992), Eaton & Kortum (2002), Atkeson & Burstein (2010); competition (Raith, 2002; Schmidt, 1997); Aghion et al, 2005); learning (Coe & Helpman, 1995; Acharya & Keller, 2008); Inputs (Goldberg et al, 2010a,b),

• Impact of China. Bernard Jenson & Schott (2006); Ossa & Hsieh (2010); Levchenko & Zhang (201)

• Low wage countries & labour markets: Machin & Van Reenen (1998), Autor, Dorn & Hanson (2013, 2014)

Data

Within plant/firm effects

Reallocation effects between plants/firms

Extensions & Robustness

Innovation and Productivity data: firm panel

• European Patent Office data (patents and citations) matched to population of public and private firms (living & dead) from BVD AMADEUS for 12 European countries 1996-2005

• France, Italy, Spain and Sweden have good materials data so estimate production functions to get TFP (separately by industry). Use de Loecker (2011) version of Olley-Pakes

• Harte Hanks plant survey on IT. Focus on computers per worker as consistent across time & countries, but also use other software-based (e.g. ERP)

• Smaller samples:– Subset 459 quoted firms reporting R&D for 5+ years– 1,576 firms with management data

Trade data: UN Comtrade• Trade data at 6-digit level product matched to 4-digit SIC

using Feenstra, Romalis, & Schott (2006) concordance

• Our main measure is IMPCH = (Chinese Imports/All Imports):• Well measured annually at 4-digit SIC level

• Also use import penetration measures– Chinese imports/apparent consumption– Chinese imports/production

Chinese Imports into EU by SIC-2, 2000-2005

0 .02 .04 .06 .08 .1mean of down5

2523313934363222263538272433293037282120

5-year change in export share, 2000 to 2005, for our sample

LeatherApparel

Furniture

Toys

FoodTobaccoChemicals

Transport

Fabricated metals

Textiles

Data

Within plant/firm effects

Reallocation effects between plants/firms

Extensions & Robustness

Basic Technology Equation

ln CHijkt jkt kt i ijktY IMP f

Chinese import share

patents, IT, R&D, TFP, management

Fixed Effects

fkt : country*time dummies

Cluster at industry-by-country (jk)

Estimate in (5-year) long differences

Example: For IT i = plants (22,957)j = industries (366)k = countries (12)jk = 2,816 cellst = 2000,…,2007

(1) (2) (3)

ΔlnPATENT Δln(IT/N) ΔTFP Δln(R&D) ΔManagement

Method 5 year diffs 5 year diffs 5 year diffs 5 year diffs 3 year diffs

Change in 0.321*** 0.361** 0.257*** 1.213** 0.891***Chinese Imports

(0.102) (0.076) (0.072) (0.549) (0.337)

Sample period

2005-1996 2007-2000 2005-1996 2007-1996 2010-2004

# units 8,480 22,957 89,369 459 1,576

# ind*cty clusters

1,578 2,816 1,210 196 579

Obs 30,277 37,500 292,167 1,626 3,607

Table 1A: Within Firm OLS Results

Notes: SE clustered by industry-country, Country-year dummies included. Estimate TFP separately by industry (on 1.4m obs). Use Olley-Pakes (1996)/de Loecker (2011). Management data from Bloom and Van Reenen (2010), management mean (std dev.) is 3.09 (0.59). Because of short-panel run regressions in 3-year diffs.

Endogeneity - use MFA policy experiment• The Multi Fiber Agreement (1974) restricted apparel and

textile exports from developing countries

• When China entered the WTO in 2001 it gained access to this phased abolition, occurring between 2002 and 2005 (Brambilla, Khandelwal & Schott, 2010)

• As Chinese textile products came off quota there was huge surge of imports into EU (led to “bra-wars”)

• 31,052 patents in the textile & apparel sub-sample

• Use (value weighted) proportion of HS6 products covered by a quota in each 4 digit SIC in 2000

IV with share of SIC4 on quota under MFA – almost random - making this a great instrument

| % industry covered by ussic | quotas (all stages) 4-digit | Mean Freq.------------+---------------------------------------------------- 2211 | .77447796 210 (BROADWOWEN FABRIC, COTTON) 2221 | .23278008 63 (BROADWOWEN FABRIC, SILK) 2231 | .02347782 134 (BROADWOWEN FABRIC, WOOL) 2321 | .86472106 32 (MEN'S AND BOYS' SHIRTS) 2322 | 1 22 (MEN'S AND BOYS' UNDERWEAR) 2323 | .78554922 26 (MEN'S AND BOYS' NECKWEAR) 2325 | .05432023 10 (MEN'S AND BOYS' TROUSERS) 2329 | .74802500 12 (MEN'S AND BOYS' CLOTHING NEC) 2337 | .48232245 137 (WOMEN’S AND GIRLS SKIRTS) 2339 | 0 22 (WOMEN’S AND GIRLS CLOTHING NEC)

| % industry covered by ussic | quotas (all stages) 4-digit | Mean Freq.------------+---------------------------------------------------- 2211 | .77447796 210 (BROADWOWEN FABRIC, COTTON) 2221 | .23278008 63 (BROADWOWEN FABRIC, SILK) 2231 | .02347782 134 (BROADWOWEN FABRIC, WOOL) 2321 | .86472106 32 (MEN'S AND BOYS' SHIRTS) 2322 | 1 22 (MEN'S AND BOYS' UNDERWEAR) 2323 | .78554922 26 (MEN'S AND BOYS' NECKWEAR) 2325 | .05432023 10 (MEN'S AND BOYS' TROUSERS) 2329 | .74802500 12 (MEN'S AND BOYS' CLOTHING NEC) 2337 | .48232245 137 (WOMEN’S AND GIRLS SKIRTS) 2339 | 0 22 (WOMEN’S AND GIRLS CLOTHING NEC)

IV with share of SIC4 on quota under MFA – almost random - making this a great instrument

Δln(IT/N) ΔChinese

Imports Δln(IT/N)

Method OLS First Stage IV

ΔChinese Imports 1.284*** 1.851***(0.172) (0.400)

Quotas removal 0.088***(0.019)

       Sample period 2005-2000 2005-2000 2005-2000

Number of units 2,891 2,891 2,891

industry clusters 83 83 83

Observations 2,891 2,891 2,891

Table 2A: IV estimates using changes in EU textile & clothing quotas - IT

SE clustered by 4 digit industries, Country-year and site type dummies included.All columns using just textiles and apparel sample

ΔlnPATENTS ΔlnPATENTS Δln(TFP) Δln(TFP)

Method OLS IV OLS IV

ΔChineseImports

0.902***(0.087)

1.629**(0.326)

ΔChinese 1.160*** 1.864*Imports (0.377) (1.001) 1st Stage F-stat   24.1  11.5

Sample period 2005-1999 2005-1999 2005-1999 1999-2005

Units 1,866 1,866 12,247 12,247

Industry clusters 149 149 177 177

Observations 3,443 3,443 20,625 20,625

Table 2A- Cont: IV estimates using changes in EU textile & clothing quotas – Patents and TFP

SE clustered by 4 digit industries, Country-year dummies included

Maybe quota sectors had pre-existing trends? Table 3: Control for these & results robust

  (1) (2) (5) (6)

Dependent Variable: Δln(PATENTS) Δln(PATENTS) ΔTFP ΔTFP         

Quotas removal 0.129** 0.216** 0.143*** 0.178***(SIC4*year>=2001) (0.063) (0.105) (0.018) (0.037)

Firm-specific trends? No Yes No Yes

Sample period 2005-1992 2005-1992 2005-1992 2005-1992

Number of units 2,435 2,435 16,495 16,495Number industry clusters

159 159 187 187

Observations 14,768 14,768 55,791 55,791

SE clustered by 4 digit industries, Country-year and dummies included.All columns using just textiles and apparel sample

*No significant correlation between quota IV and pre-WTO change in observables like TECH, industry size, wages, etc. (see Table A3)

Other ways of controlling for unobserved technology shocks• Include three digit industry trends (Table 1B)

• Use level of Chinese imports share in 1999 in all EU & US four digit industry as IV for subsequent growth (Table 8)– Trade analogue of Card’s “ethnic enclave” instrument

• Include lagged dependent variable (Table A4)

• Find supportive results to the quota IV approach

Data

Within plant/ firm effects

Reallocation effects between plants/firms

Extensions & Robustness

5

5

ln [ * ]

ijkt

n CHijkt ijkt jkt

n CH n n njkt ijkt kt

N TECH IMP

IMP TECH f

C) Employment Equation

expect αn < 0

Employmentgrowth

If high TECH plants “protected” from Chinese imports then γn > 0

expect δn > 0

-.2

-.1

0.1

1 2 3 4 5 1 2 3 4 5

Quintiles of initial IT Intensity

EMPLOYMENT GROWTH BY INITIAL IT INTENSITY

Bottom Quintile China Import Growth

IT HIGH IT HIGHIT LOW IT LOW

Top Quintile China Import Growth

Quintiles of initial IT Intensity

Em

plo

yme

nt

Gro

wth

Dependent Variable: Δln(N) Δln(N) Δln(N)TECH Measure: Patents IT TFPChinese Import Growth -0.434*** -0.379*** -0.377***

(0.136) (0.105) (0.096)Ln(pat stock/worker) at t-5 0.348***

(0.049)Ln(pat stock/worker) at t-5* 1.434**Chinese imports growth (0.649)IT intensity (t-5) 0.230***

(0.010)(IT/N) (t-5)*Chinese Imp Growth 0.385**

(0.157)Ln(TFP) at t-5 0.136***

(0.012)Ln(TFP) at t-5* 0.795***Chinese import growth (0.307)Clusters 1,375 2,816 1,210Observations 19,844 37,500 292,167

Table 4A: High Tech firms shed less jobs when faced with rising Chinese imports

SE clustered by country- industry, all standard additional controls included

5

5

[ * ]

ijkt

s CHijkt ijkt jkt

s CH s s sjkt ijkt kt

SURVIVAL TECH IMP

IMP TECH f

C) Survival Equation

expect αs < 0

SurvivalIf high TECH plants partially “protected” from effect of Chinese imports then γs > 0

expect δs > 0

Dependent Variable Survival Survival SurvivalTECH measure: patents IT TFPChange in Chinese Imports -0.079 -0.182** -0.208***

(0.051) (0.072) (0.050)

Ln(patent stock/workert-5) *Change

in Chinese Imports0.288**(0.138)

(IT/N) t-5*Change in Chinese Imports 0.137(0.112)

ln(TFPt-5) *Change in Chinese Imports 0.110*(0.059)

IT Intensity (IT/N)t-5 -0.002(0.006)

Ln(patent stock/workert-5) -0.008(0.009)

Ln(TFPt-5) -0.003

(0.003)Observations 7,985 28,624 60,883

Table 4B: High tech firms more likely to survive Chinese imports

SE clustered by up to 2,863 country- industry pairs, lagged size & all standard additional controls

Magnitude calculation

Table 6: Aggregate effect of trade on technology, 2000-2007, % of Technology Measure that Chinese trade ‘accounts for’

Measure Within (%) Between (%) Exit (%) Total (%)

Patents 5.1 6.7 2.1 13.9

IT Intensity 9.8 3.1 1.2 14.1

TFP 9.9 2.4 0.2 12.5

Notes: calculated for the regression sample using OLS coefficients

Dependent Variable: Δln(PATENTS) Δln(IT/N) Δln(TFP)Change in Chinese Imports 0.368* 0.399*** 0.326***

(0.200) (0.120) (0.072)

Sample period 2005-1996 2007-2000 2005-1996

Observations 6,888 7,409 5,660

Dependent Variable: Δln(PATENTS) Δln(IT/N) Δln(TFP)Change in Chinese Imports 0.171** 0.361** 0.164***

(0.082) (0.076) (0.051)Years 2005-1996 2007-2000 2005-1996Observations 30,277 37,500 241,810

Table A5: Comparison of Industry level results to firm results confirms simulated magnitudes

Panel A: Industry-level regressions

Panel B: Firm-level (firms allocated to single SIC4)

Data

Within plant/firm effects

Reallocation effects between plants/firms

Extensions & Robustness

Extensions & Robustness

• Dynamic Selection (we bound the bias)• General Equilibrium• Heterogeneity of China effects

– Within firm innovation effects strongest in industries where there are more “trapped factors”

• Other low wage countries (yes, similar to China)• Lawyer effects on patents (find no evidence for this)• Offshoring (find some effect on IT and TFP)• “I-Pod” story (firms innovate here to produce in China – but

imports reduce profits)• Exports (find small & insignificant effects)• Product Switching

Dependent Variable: Δln(PAT) Δln(PAT)

BaselineWorst case

Lower Bound

Change in Chinese 0.321*** 0.271**Imports (0.102) (0.098)

Number of units 8,480 8.732

Number of clusters 1,578 1,662

Observations 30,277 31,271

Table 7: Dynamic Selection - Worst Case Lower Bounds

Use Manski & Pepper (2000) idea of imputing lower bound of zero patents to all the firms who exited to calculate worst case lower bound. Similar results for Negative Binomial model & using quota IV

What about general equilibrium – does that reduce China’s estimated impact?Bloom, Romer, Terry & Van Reenen (2014) detail model of impact of falling trade costs with China on global Growth in a GE model of endogenous growth with “trapped factors”

Calibrating this model to 1996-2007 we find that China raises:

- Long-run global growth in OECD by ≈ 0.2% (market size)

- Short-run (10 year) growth by further ≈ 0.2% (trapped factors)

Quantitatively important (equivalent to 16% ↑ in consumption)

GE effects magnify the impacts we find in micro data

Any evidence for trapped factor mechanism? Looks at heterogeneity of Chinese import effect

• We want to capture the firm specific ‘trapped’ capacity to innovate

• Two measures of ‘trapped factors’– Industry wage premia (estimated via UK LFS)– Measured firm-level TFP (can be show to incorporate

trapped factors)

• We allow China effect to be heterogeneous across firms/industries depending on these proxies

Change Chinese Imports 0.321*** 0.202*** 0.284* -2.466***(0.102) (0.092) (0.157) (0.848)

Industry wage premia   -0.411***        (0.069)    Change Chinese Imports   2.467***    * Industry Wage premia   (1.171)    TFPt-5       -0.287***

      (0.050)Change Chinese Imports       1.464****TFPt-5       (0.462)

       Number of units 8,480 8,480 5,014 5,014Number of clusters 1,578 1,578 1,148 1,148

Number of Observations30,277 30,277

14,500 14,500

The heterogeneity of the China treatment effect consistent with trapped factor mechanism

Dependent variable: Δln(Patents)

Change in Chinese

0.182*** 0.063 0.182*** Imports (0.074) (0.125) (0.073)

Change in Non-China 0.152 Low Wage Imports (0.128)

Change in All Low 0.106*** Wage country Imports (0.040)

Change in High Wage 0.003 0.004 Country Imports (0.019) (0.019)

Change in World Imports

Observations 29,062 29,062 29,062 29,062

Low wage countries list taken from Bernard et al (2006). Defined as countries <5% GDP/capita relative to the US 1972-2001.Chinese imports normalized by domestic production

Is there something specific about China? No: similar to all low-wage countries

The lawyer effect?

Maybe firms just patenting more to protect IP?

So investigate this in three ways:

• R&D: seem to spend more on innovation

• Cites/patents: should drop if more marginal ideas patented. We find the opposite

• Timing of patents: if this is simply a legal response should happen immediately (or in advance), while it is an innovation response more likely to be lagged (which is what we find, Table A14)

What about offshoring instead?

Is effect all driven by firms offshoring low value inputs to China?

Investigate this by generating a Chinese offshoring proxy

(based on Feenstra-Hansen, 1999)• Weight Chinese imports/apparent consumption by SIC 4-

digit input-output tables (US 2002 tables)• Proxies how much Chinese imports are increasing for each

industry averaged across its sourcing industries

Dependent Variable ΔPATENTS Δln(IT/N) ΔTFPChange in Chinese Imports 0.313*** 0.279*** 0.189***

(0.100) (0.080) (0.082)Offshoring (ΔChinese Imports 0.173 1.685*** 1.396*** in source industries (0.822) (0.517) (0.504)Observations 30,277 37,500 30,608

Table 1D: Offshoring matters for more compositional measures (IT and TFP) but not innovation

Note: We also find bigger jobs shakeout for firms who have branches in China

Dependent Variable: Δln(Employment) Δln(EU Producer Prices)

Δln(Profits/Sales)

Change in Chinese -0.411*** -0.453** -0.112**

Imports (0.133) (0.217) (0.052)

Observations 8,788 259 5,372

Industry-country pairs 1,913 131 2,295

Aggregation SIC4 SIC2 SIC4

Years 2005-1996 2006-1996 2007-2000

The IPod effect: innovate because of low costs? No, imports reduce industry profits

Notes: Estimation by OLS in 5 year long differences, SE clustered by industry-country pair, country-year dummies included,

Dependent Variable ΔPATENTS Δln(IT/N) Δln(TFP)Change in Chinese Imports 0.322*** 0.361*** 0.254***

(0.100) (0.076) (0.072)Change in Exports to China -0.243 0.028 -0.125

(0.200) (0.118) (0.126)Number of Observations 30,277 37,500 292,167

Is this just exporting to China? No exporting per se has little effect

Conclusions

Empirics• Find trade-induced increases in innovation, IT and TFP• Occurs within and between plants and firms• Relatively large and growing: China “accounts” for ~15% of

increase in aggregate IT, patents and TFP and rising• Other low-wage countries trade similar effect, but high-wage

countries trade appears to have no effect

Model• Trapped-factors model seems to be best fit

– firms have trapped factors that keep them in the industry– so innovate new products to escape competition

Back Up

Trapped Factor Model

• Set Up– Endogenous Growth with North (R&D performing high

tech) and South – Consider that in short run some factors are trapped in a

firm (specific due to sunk costs irreversibilities)– Parameter with Northern barriers to Southern trade

import changes over time. Consider a shock to this • Results

– Standard result that lowering trade barriers increases innovation/growth through market size effect

– Trapped factors re-deployed from production to innovation sector in firms affected by shock. This gives an extra short-run kick to innovation

Table 1B: Include SIC-3 trends

Dependent variable: Δln(PATENTS) Δln(IT/N) ΔTFP

Change in Chinese 0.191* 0.170** 0.128**Imports (0.102) (0.082) (0.053)Number of Units 8,480 22,957 89,369

Number of country by industry clusters 1,578 2,816 1,210

Observations 30,277 37,500 292,167

Table A3: Regressing 2000-1995 changes on 2000 Quota IV, Industry Level

Δln(PATENTS) Δln(TFP)

Δln(Output/Labor)

Δln(Capital/Labor) Δln(Labor) Δln(Wages)

QUOTA -0.263 0.010 0.006 0.059 -0.053 0.032

(0.195) (0.023) (0.041) (0.071) (0.035) (0.023)Observations 203 115 115 114 113 116

Table 8: Initial Conditions IV

  (1) (2) (3) (4) (5) (6)  

Dependent VariableΔln

(PATENTS)Δln

(PATENTS)Δln(IT/N)

Δln(IT/N) ΔTFP ΔTFP  

Method: OLS IV OLS IV OLS IV  

Change in Chinese 0.321*** 0.495** 0.361*** 0.593*** 0.257*** 0.507*  

 Imports (0.117) (0.224) (0.106) (0.252) (0.087) (0.283)  

Initial Condition IV              

First-stage F-Statistic   95.1   38.7   14.5

Sample period 2005-1996 2005-1996 2007-2000 2007-20002005-1996

2005-1996

Number of Units 8,480 8,480 22,957 22,957 89,369 89,369  

#of industry clusters 304 304 371 371 354 354  

Observations 30,277 30,277 37,500 37,500 292,167 292,167  

Computer Intensity

Employment Number of Sites

Austria 0.50 352 1067

Denmark 0.69 148 510

Finland 0.59 173 677

France 0.55 243 2911

Germany 0.52 435 3679

Ireland 0.63 196 350

Italy 0.55 222 2630

Norway 0.72 131 362

Spain 0.49 175 1018

Sweden 0.60 161 1168

Switzerland 0.60 179 1346

United Kingdom 0.64 270 3567

Summary Statistics: 12 Country Panel

Output Quotas rather than Input Quotas matter most

Dependent Var. Δln(IT/N) Δln(IT/N) Means

Method Reduced Form Reduced Form (standard dev)Output Quota 1.284*** 0.133*** 0.094Removal (0.172) (0.045) (0.232)

Input Quota 0.311 0.031Removal (0.342) (0.041)

Observations 2,891 2,891

Notes: Input quotas are calculated using the Feenstra-Hansen method but using quotas instead of import flows. 489 SIC4 clusters.

List of low wage countriesAlbaniaAngolaBangladeshBeninBoliviaBurkina FasoBurundiCambodiaCameroonCentral African RepChadChinaComorosCongoDjibouti

EgyptEquatorial GuineaEthiopiaGambiaGhanaGuineaGuinea-BissauGuyanaHaitiHondurasIndiaIndonesiaIvory CoastLao People's Dem. Rep.

MadagascarMalawiMaliMauritaniaMongoliaMoroccoMozambiqueNepalNicaraguaNigerNigeriaPakistanPapua New GuineaPhilippines

RwandaSenegalSierra LeoneSri LankaSudanSurinameSyriaTanzaniaTogoUgandaViet NamYemenZambiaZimbabwe

Low wage countries list taken from Bernard, Jensen and Schott (2006). They defined these as countries with less than 5% average per capita GDP relative to the Unites States in the period between 1972-2001.

Sample All All AllTextiles

& Apparel

Textiles &

Apparel

Method OLS OLS OLS OLS IV

Change in Chinese 0.144** 0.099** 0.166** 0.277***

Imports (0.035) (0.043) (0.030) (0.053)

Change in IT intensity 0.081** 0.050*

(0.024) (0.026)F-test of excluded IVs

9.21

Observations 204 204 204 48 48

Tab A10 China associated with skill upgrading (wage bill share of college educated in UK)

SE are clustered by 74 SIC3 industries; 2006-1999, UK LFS data, IV is height of quota pre-WTO; all columns control for year dummies, regressions weighted by industry employment in 1999

Dependent variable

Plant Switches Industry

Plant Switches Industry

Δln(IT/N) Δln(IT/N)

ΔChinese 0.138*** 0.132*** 0.466***Imports (0.050) (0.050) (0.083)IT intensity (t-5) -0.018**

(0.008)Switched Industry 0.025**

(0.012)0.023*(0.012)

Employment growth

-0.002(0.006)

Observations 32,917 32,917 32,917 32,917

Table A11: INDUSTRY SWITCHING

“Switched Industry” is a dummy if a plant switched its main four digit industry over a five Year period. SE clustered by country*industry pair. 2000-2007.

Tab A14 Dynamics - Patent effect largest at long lags

PANEL A: PATENTS, Δln(PATENTS) (1) (2) (3) (4) (5) (6) (7)

5-year lag of Chinese Imports Change 0.328*** 0.013

5CHtIMP (0.110) (0.163)

4-year lag of Chinese Imports Change 0.394*** 0.280*

4CHtIMP (0.110) (0.149)

3-year lag of Chinese Imports Change 0.402*** -0.005

3CHtIMP (0.120) (0.178)

2-year lag of Chinese Imports Change 0.333*** 0.074

2CHtIMP (0.113) (0.136)

1-year lag of Chinese Imports Change 0.314*** -0.069

1CHtIMP (0.102) (0.145)

Contemporaneous Chinese Imports Change 0.321*** 0.203 CHtIMP (0.102) (0.163)

Number of country-industry pairs 1,578 1,578 1,578 1,578 1,578 1,578 1,578 Number of Firms 8,480 8,480 8,480 8,480 8,480 8,480 8,480 Observations 30,277 30,277 30,277 30,277 30,277 30,277 30,277

Table A13 Results on IT appear broadly robust to using other ICT diffusion measures

Databases ERP Groupware

Growth in Chinese Import Share

0.002(0.070)

0.040(0.034)

0.249***(0.083)

Highest Quintile Growth in Chinese Import Share

0.020**(0.010)

0.013***(0.005) 0.034**

(0.014)

Quintile 4 0.030**(0.010)

0.006(0.005)

0.021(0.013)

Quintile 3 0.043***(0.010)

0.014***(0.005)

-0.008(0.013)

Quintile 2 0.024***(0.011)

0.010**(0.005)

-0.018(0.013)

Obs. 24,741 24,741 24,741 24,741 24,741 24,741

Note: All changes in long (5-year) differences. Includes country, year and site-type controls. All standard-errors clustered by country and SIC-4 cell

(1) (2) (3) (4) Dep. variable: Δln(PATENTS) Δln(PATENTS) ΔTFP ΔTFP Quotas removal 0.207** 0.490*** 0.201*** 0.204*** *I(year>2000) (0.098) (0.157) (0.038) (0.047) Include lagged dependent variable(t-5)? No Yes No Yes IV lagged dependent variable? No Yes No Yes Years 2005-1995 2005-1995 2005-1995 2005-1995 Number of units 675 675 675 675 Number of industry clusters 104 104 104 104 Observations 6,075 6,075 3,107 3,107

Table A4 Controlling for lagged technology

Find Cites/Patents constant as Chinese imports rise – so no evidence patent quality falling

Dependent variableCites/Patent

Cites/Patent

Cites/Patent

Log(1+Cites/Patent)

OLS OLS OLS OLS

Growth in Chinese Imports

0.090(0.242)

0.027(0.843)

0.082(0.242)

0.090(0.242)

Log (Patents) 0.020***(0.007)

4 digit industry controls Yes n/a n/a n/a

Firm fixed effects No Yes Yes Yes

Country-year fixed effects Yes Yes Yes Yes

Observations 21,273 21,273 21,273 21,273

SE are clustered by firm