Robots, Tasks, and Trade - World Trade Organization · Data I Cross-country industry-level panel...
53
Robots, Tasks, and Trade Erhan Artuc Paulo Bastos Bob Rijkers World Bank June 2019
Transcript of Robots, Tasks, and Trade - World Trade Organization · Data I Cross-country industry-level panel...
Robots, Tasks, and Trade
Erhan Artuc
Paulo Bastos
Bob Rijkers
World Bank
June 2019
Introduction
I Advances in automation, robotics and artificial intelligence
I Concerns about potential disruptive impacts
I Evidence that adoption of industrial robots had significant
economic impacts in high-income countries
I Positive effects on industry productivity and avg. wages, but
reduced employment share of low-skilled workers in the OECD
(Graetz and Michaels, 2018)
I Sizable negative effects on employment and wages across
commuting zones in the US (Acemoglu and Restrepo, 2017)
Introduction (cont.)
I So far, robot adoption has been largely confined to a small
number of high-income countries
I But lower-income countries might be indirectly affected
(Rodrik, 2018)
I In an integrated global economy, robot adoption in the North
may impact:
1. relative costs and international specialization
2. wages and welfare, even among non-adopters
I We examine implications of industrial robot use for
North-South trade, wages and welfare
Introduction (cont.)
I What is an industrial robot?
Automatically controlled, reprogrammable, multipurpose
manipulator programmable in three or more axes, which
may be either fixed in place or mobile for use in industrial
automation applications.
I Perform variety of repetitive tasks with consistent precision
I Common applications include:
I Assembling
I Dispensing
I Handling
I Processing
I Welding
Introduction (cont.)
Figure: Industrial robot in auto industry
Introduction (cont.)
Figure: Industrial robot in the electronics industry
Introduction (cont.)
Figure: Robot prices relative to wages
Theory
I Develop a task-based Ricardian trade model combining several
ideas from the literature:
I Productivity differences across countries and sectors (Eaton
and Kortum, 2002)
I Two-stage production, and trade in intermediates and
final goods (Yi, 2003; Caliendo and Parro, 2015)
I Robots can take over some tasks previously performed by
humans (Acemoglu and Restrepo, 2017)
Key implications of the model
I Higher wages lead to robotization of tasks
I Lower robot prices induce robotization and reduce production
costs
I Change in production costs varies by stage
I Robotization impacts comparative advantage and trade
patterns
I If robotization in the North increases:
I Exports to South in robotized industry increase
I Imports from South in robotized industry can rise or fall
I Impact on wages is non-linear
Data
I Cross-country industry-level panel data for 1995-2015,
linking several data sources through consistent definition of
industries:I Industrial robots by country-industry-year from International
Federation of Robotics
I Based on yearly survey of robot suppliers
I Measures deliveries of “multipurpose manipulating industrial
robots”
I Covers about 90 percent of industrial robots market
I Industry-country-year data on labor hours, material inputs, IT
capital and non-IT capital from EUKLEMS
I Industry-level measure of replaceability (Graetz and Michaels,
2018)
I Bilateral trade and tariff data from BACI and UNCTAD
TRAINS
Robotization and initial GDP per capita
Belgium
BulgariaCroatia
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Latvia
Netherlands
Poland
Portugal
Romania
Slovak Republic
Slovenia
Spain
Sweden
United Kingdom
United States
0.5
11.
5lo
g(1+
robo
ts/h
ours
)
0 10000 20000 30000 40000 50000Initial GDP per capita (constant 2010 USD)
Robotization and replaceability
AgricultureMining
Food products
Textiles
Wood, paper, printing
Chemicals
Rubber and plastics
Metal
ElectronicsMachinery
Automotive
Other manufacturing
Utilities
ConstructionEducation
Other non-manufacturing0.5
11.
52
2.5
log(
1+ro
bots
/hou
rs)
0 20 40 60% of replaceable jobs
Empirical Strategy
I Main specification:
Tradenmit = βRobotsnit + τnmt + θit + ε (1)
I n indexes developed country, m indexes developing country, i
indexes sector, and t year
I τnmt is fixed effect by exporter-importer-year
I θit denotes an industry-year fixed effect
I standard errors are clustered by developed country
I Various forms of endogeneity possible:
I import competition could impact robot adoption
I increased exports to South might also impact adoption
incentives
I Attenuation bias caused by measurement error in robot stocks.
Empirical Strategy: Theory-Consistent IV Strategy
I Instrument robot stock by country-industry-year with
interaction between:
1. initial GDP per capita (proxy for labor costs)
2. pre-determined industry replaceability index (proxy for
robotization frontier)
3. global robot stock (proxy for exogenous robot price)
I Assess robustness using alternative specifications and IV
strategies
Impact of robotization on North-South trade
OLS IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
log(1 + robotshours
) 0.1534** 0.4117*** -0.2583*** 0.6144*** 1.1839*** -0.5695***
(0.0690) (0.1238) (0.0684) (0.2156) (0.3146) (0.1529)
First stage:
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027*** 0.0027***
(0.0003) (0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813 888,813 888,813
R2 0.6351 0.7738 0.3912 0.6314 0.7655 0.3887
Kleibergen-Paap rk Wald F-stat 111.795 111.795 111.795
Importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using baseline estimation sample. Columns (1)-(3) report the OLS results, while
columns (4)-(6) report the IV and corresponding first stage estimates. Robust standard errors clustered by developed country in parenthesis.
***1% level, **5% level, *10% level.
Additional controls
OLS IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
log(1 + robotshours
) 0.1694** 0.4309*** -0.2615*** 0.6370*** 1.2130*** -0.5760***
(0.0635) (0.1216) (0.0712) (0.2145) (0.3175) (0.1561)
log (1 + material inputs) 0.0029 0.0939 -0.0910 -0.1710 -0.1970 0.0259
(0.1013) (0.1662) (0.1140) (0.1227) (0.2142) (0.1225)
log (1 + IT capital) 4.8065 7.8370 -3.0306 12.0230*** 19.9068** -7.8838
(3.9050) (7.3223) (6.2491) (3.4734) (8.7868) (7.4821)
log (1 + non-IT capital) -19.3898** -31.3726* 11.9828 -31.2832*** -51.2645*** 19.9813*
(7.2922) (16.5164) (11.0901) (7.2316) (15.6678) (11.3884)
log (1 + tariffs) 0.5462*** 0.2371 0.3091*** 0.5589*** 0.2584* 0.3005***
(0.0725) (0.1650) (0.0989) (0.0643) (0.1500) (0.0925)
First stage:
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027*** 0.0027***
(0.0003) (0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813 888,813 888,813
R2 0.6354 0.7742 0.3913 0.6318 0.7659 0.3889
Kleibergen-Paap rk Wald-F-stat 105.254 105.254 105.254
Importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using baseline estimation sample. Columns (1)-(3) report the OLS results, while
columns (4)-(6) report the IV and corresponding first stage estimates. Robust standard errors clustered by developed country in parenthesis.
***1% level, **5% level, *10% level.
Alternative instruments: robotization in other countries
IV IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
log(1 + robotshours
) 0.5825** 1.4069*** -0.8244*** 0.5762** 1.2756** -0.6995**
(0.2333) (0.3403) (0.2576) (0.2745) (0.4565) (0.2756)
First stage:
log(1 + robotshours
) in two countries 0.2743** 0.2743** 0.2743**
with most similar GDP per capita (0.1000) (0.1000) (0.1000)
log(1 + robotshours
) in four countries 0.3247** 0.3247** 0.3247**
with most similar GDP per capita (0.1182) (0.1182) (0.1182)
Observations 786,797 786,797 786,797 854,589 854,589 854,589
R2 0.6282 0.7608 0.3855 0.6255 0.7580 0.3874
Kleibergen-Paap rk Wald-F-stat 7.523 7.523 7.523 7.546 7.546 7.546
Importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using the baseline estimation sample and alternative instruments. Columns
(1)-(3) report the OLS results, while columns (4)-(6) report the IV and corresponding first stage estimates. Robust standard errors clustered by
developed country in parenthesis. ***1% level, **5% level, *10% level.
Imports of intermediates versus other goods
BEC classification Schott (2004) classification
OLS IV OLS IV
Dependent variable: log(1 + imp) log(1 + imp) log(1 + imp) log(imp)
(1) (2) (3) (4)
A. Intermediate goods
log(1 + robotshours
) 0.1094 0.6815** 0.2132*** 0.8566**
(0.0788) (0.3240) (0.0573) (0.3124)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027***
(0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813
R2 0.5137 0.5027 0.5278 0.5101
Kleibergen-Paap rk Wald-F-stat 111.795 111.795
B. Other goods
log(1 + robotshours
) 0.1507*** 0.5627*** 0.1136* 0.4929**
(0.0511) (0.1680) (0.0619) (0.1906)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027***
(0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813
R2 0.6259 0.6226 0.6246 0.6220
Kleibergen-Paap rk Wald-F-stat 111.795 111.800
Importer-exporter-year effects Y Y Y Y
Industry-year effects Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using the sub-samples for imports of intermediate and other goods. Columns
(1)-(3) report the OLS results, while columns (2)-(4) report the IV and corresponding first stage estimates. Robust standard errors clustered by
developed country in parenthesis. ***1% level, **5% level, *10% level.
Counterfactual simulations
I World Input-Output Database 2005 for calibration
I Use calibrated model to perform counterfactual simulations
I 3 countries: representative North, representative South, Other
I Examine effects of reduction in robot prices on trade patterns,
labor allocation, wages and welfare
Effects of robot price reductions on labor allocation
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-40
-35
-30
-25
-20
-15
-10
-5
0
% C
hang
e in
Num
ber
of W
orke
rs
Labor Allocation in Robotized Industries
SouthNorthOther
Effects of robot price reductions on wages
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-3
-2
-1
0
1
2
3
4
% C
hang
e in
Rea
l Wag
e
Change in Wages
SouthNorthOther
Effects of robot price reductions on real gdp
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
1
2
3
4
5
6
% C
hang
e in
Rea
l GD
P
Change in GDP
SouthNorthOther
Conclusion
I We examined theoretically, empirically and quantitatively the
impacts of robotization on trade, wages and welfare
I Main takeaways:
1. Robot adoption in the North promotes trade with the South
(both in intermediates and final goods)
2. Input-output linkages and trade in intermediates are important
in modulating effects of robotization on trade
3. Robots impact wages and welfare, even among non-adopters
4. Robotization initially depresses wages in the North
5. Northern robotization leads to higher wages and welfare in the
South
6. Positive effects in the South more likely under lower trade costs
Additional slides
Wage-productivity decoupling
Figure: Labor productivity and wages in the US
Robot prices, wages, and labor demand
0 10 20 30 40 50 60 70 80 90
% reduction in wR
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Coe
ffici
ents
a
nd
m,i
m,i
Theory (cont.)
I Demand for ω ∈ Ss is given by
qm,i (ω) =
(pm,i (ω)
Pm,is
)−σi
Qm,is ,
where:
Pm,is =
1
µ(S is)
[∫Ss
(pm,i (ω)
)1−σi
dω
] 1
1−σi,
I µ(S is) denoting the share of varieties associated with stage s
in industry i .
Theory (cont.)
Prices and international trade
I The price of the composite good from industry i and stage s
can be expressed as
Pm,is = ψ
∑n
(τm,n,i
(rn,is
)αn,is,F
(Pn,is−1
)αn,is,M (
Ωn,iwnL
)αn,is,T
)−θi− 1
θi
I The probability that country n charges the lowest price in
country m for a stage s variety is given by
πm,n,is =
ψn,is,4τ
m,n,i(rm,is,F
)αm,is,F(Pn,is−1
)αn,is,M (
Ωn,iwnL
)αn,is,T
Pm,is /ψi
2
−θi
.
Theory (cont.)
Nominal factor prices
rm,is,F =αm,is,FY
m,is
Fm,is
,
wmL =
(ψm,is,1
ψ4
) 1
αm,iT ,s α
m,iT ,sY
m,is
Lm,is
Ξm,i
Ωm,i,
where Ξm,i is a measure of labor demand
Ξm,i =(1− K i )(Lm,iA + Lm,iN )
Lm,iN
,
Related literature
1. Literature on impacts of industrial robotsI Graetz and Michaels (2018), Acemoglu and Restrepo (2017)
2. Literature using variants of Eaton and Kortum (2002) to
assess implications of different aspects of globalization for
welfare and income distribution
I Yi (2003), Dekle et al. (2008), Chor (2010), Waugh (2010),
Fieler (2011), Arkolakis et al. (2012), Parro (2012), Burstein
et al. (2013), Caliendo and Parro (2015), Caliendo et al.
(2015) and Donaldson (2018)
3. Broader literature on links between trade and technologyI Freund and Weinhold (2004), Verhoogen (2008), Lileeva and
Trefler (2010), Bustos (2011), Burstein et al. (2016), Bloom et
al. (2016), Burstein and Vogel (2017), Atkin et al. (2017),
Fort (2017), Bastos et al. (2018), and Autor et al. (2017).
Robotization by sector
01
23
01
23
01
23
01
23
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
Agriculture Mining Food products Textiles
Wood, paper, printing Chemicals Rubber and plastics Metal
Electronics Machinery Automotive Other manufacturing
Utilities Construction Education Other non-manufacturing
log(
1+ro
bots
/hou
rs)
Robotization by country
0.5
11.
52
0.5
11.
52
0.5
11.
52
0.5
11.
52
0.5
11.
52
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015
Belgium Bulgaria Croatia Czech Republic Denmark
Estonia Finland France Germany Greece
Hungary Ireland Italy Latvia Netherlands
Poland Portugal Romania Slovak Republic Slovenia
Spain Sweden United Kingdom United States
log(
1+ro
bots
/hou
rs)
Robotization by sector-country
01
23
4
1995 2000 2005 2010 2015
Agriculture
01
23
4
1995 2000 2005 2010 2015
Mining
01
23
4
1995 2000 2005 2010 2015
Food products
01
23
4
1995 2000 2005 2010 2015
Textiles
01
23
4
1995 2000 2005 2010 2015
Wood paper printing
01
23
4
1995 2000 2005 2010 2015
Chemicals
01
23
4
1995 2000 2005 2010 2015
Rubber and plastics
01
23
4
1995 2000 2005 2010 2015
Metal
01
23
4
1995 2000 2005 2010 2015
Electronics
01
23
4
1995 2000 2005 2010 2015
Machinery
01
23
4
1995 2000 2005 2010 2015
Automotive
01
23
4
1995 2000 2005 2010 2015
Other manufacturing
01
23
4
1995 2000 2005 2010 2015
Utilities
01
23
4
1995 2000 2005 2010 2015
Construction0
12
34
1995 2000 2005 2010 2015
Education
01
23
4
1995 2000 2005 2010 2015
Other non-manufacturing
log(
1+ro
bots
/hou
rs)
Robotization by country-sector
01
23
4
1995 2000 2005 2010 2015
Belgium
01
23
4
1995 2000 2005 2010 2015
Bulgaria
01
23
4
1995 2000 2005 2010 2015
Croatia
01
23
4
1995 2000 2005 2010 2015
Czech Republic
01
23
4
1995 2000 2005 2010 2015
Denmark
01
23
4
1995 2000 2005 2010 2015
Estonia
01
23
4
1995 2000 2005 2010 2015
Finland
01
23
4
1995 2000 2005 2010 2015
France
01
23
4
1995 2000 2005 2010 2015
Germany
01
23
4
1995 2000 2005 2010 2015
Greece
01
23
4
1995 2000 2005 2010 2015
Hungary
01
23
4
1995 2000 2005 2010 2015
Ireland
01
23
4
1995 2000 2005 2010 2015
Italy
01
23
4
1995 2000 2005 2010 2015
Latvia
01
23
4
1995 2000 2005 2010 2015
Netherlands
01
23
4
1995 2000 2005 2010 2015
Poland
01
23
4
1995 2000 2005 2010 2015
Portugal0
12
34
1995 2000 2005 2010 2015
Romania
01
23
4
1995 2000 2005 2010 2015
Slovak Republic
01
23
4
1995 2000 2005 2010 2015
Slovenia
01
23
4
1995 2000 2005 2010 2015
Spain
01
23
4
1995 2000 2005 2010 2015
Sweden
01
23
4
1995 2000 2005 2010 2015
United Kingdom
01
23
4
1995 2000 2005 2010 2015
United States
log(
1+ro
bots
/hou
rs)
Changes in robotization and initial GDP per capita
Belgium
Czech Republic
Denmark
Finland
France
Germany
Hungary
Italy Netherlands
Poland
Portugal
Romania
Slovak Republic
Spain Sweden
United KingdomUnited States
0.5
11.
5C
hang
e in
log(
1+ro
bots
/hou
rs)
0 10000 20000 30000 40000 50000Initial GDP per capita (constant 2010 USD)
Changes in robotization and replaceability
Agriculture
Mining
Food products
TextilesWood, paper, printing
Chemicals
Rubber and plastics
Metal
ElectronicsMachinery
Automotive
Other manufacturingUtilities Construction
EducationOther non-manufacturing
0.5
11.
5C
hang
e in
log(
1+ro
bots
/hou
rs)
0 .2 .4 .6proportion of jobs that are replaceable
Alternative robotization measure
OLS IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
log(1 + obs.robotshours
) 0.1387** 0.3854*** -0.2467*** 0.5679*** 1.0943*** -0.5264***
(0.0669) (0.1180) (0.0646) (0.1960) (0.2843) (0.1392)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0030*** 0.0030*** 0.0030***
(0.0003) (0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813 888,813 888,813
R2 0.6351 0.7739 0.3913 0.6313 0.7656 0.3890
Kleibergen-Paap rk Wald-F-stat 115.912 115.912 115.912
Importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using the baseline estimation sample and an alternative robotization measure
(the observed stock of robots, ignoring depreciation). Columns (1)-(3) report the OLS results, while columns (4)-(6) report the IV and
corresponding first stage estimates. Robust standard errors clustered by developed country in parenthesis. ***1% level, **5% level, *10% level.
Excluding outliers: robotization and dependent variable
censored
OLS IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
log(1 + robotshours
) 0.1771** 0.4384*** -0.2725*** 0.5830** 1.1704*** -0.5942***
(0.0738) (0.1327) (0.0720) (0.2173) (0.3223) (0.1370)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027*** 0.0027***
(0.0002) (0.0002) (0.0002)
Observations 870,826 871,295 870,942 870,826 871,295 870,942
R2 0.6205 0.7624 0.4095 0.6179 0.7559 0.4070
Kleibergen-Paap rk Wald-F-stat 129.528 128.008 128.850
importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, excluding outliers from the baseline estimation sample (robotization and
dependent variables censored). Columns (1)-(3) report the OLS results, while columns (4)-(6) report the IV and corresponding first stage
estimates. Robust standard errors clustered by developed country in parenthesis. ***1% level, **5% level, *10% level.
Using inverse hyperbolic sine to account for zeros
OLS IV
Dependent variable: sinh−1(imp) sinh−1(exp) sinh−1(imp)− sinh−1(imp) sinh−1(exp) sinh−1(imp)−sinh−1(exp) sinh−1(exp)
(1) (2) (3) (4) (5) (6)
sinh−1( robotshours
) 0.1431** 0.3479*** -0.2048*** 0.5236*** 0.9625*** -0.4389***
(0.0621) (0.1044) (0.0560) (0.1827) (0.2513) (0.1236)
First stage :
Replaceability * init. GDPpc * global robot stock 0.0035*** 0.0035*** 0.0035***
(0.0003) (0.0003) (0.0003)
Observations 888,813 888,813 888,813 888,813 888,813 888,813
R2 0.6382 0.7753 0.3868 0.6349 0.7683 0.3850
Kleibergen-Paap rk Wald-F-stat 109.622 109.622 109.622
importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using the inverse hyperbolic sine to account for zeros in the regression variables.
Columns (1)-(3) report the OLS results, while columns (4)-(6) report the IV and corresponding first stage estimates. Robust standard errors
clustered by developed country in parenthesis. ***1% level, **5% level, *10% level.
Heterogeneity by income level of non-OECD countries
OLS IV
Dependent variable: log(1 + imp) log(1 + exp) log( 1+imp1+exp
) log(1 + imp) log(1 + exp) log( 1+imp1+exp
)
(1) (2) (3) (4) (5) (6)
A. High-income non-OECD countries
log(1 + robotshours
) 0.3187*** 0.4851*** -0.1664** 1.0488*** 1.1087*** -0.0599
(0.0800) (0.1202) (0.0626) (0.2927) (0.2944) (0.1987)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027*** 0.0027***
(0.0003) (0.0003) (0.0003)
Observations 80,080 80,080 80,080 80,080 80,080 80,080
R2 0.6674 0.8266 0.3889 0.6593 0.8217 0.3886
Kleibergen-Paap rk Wald-F-stat 110.960 110.960 110.960
B. Low and middle-income non-OECD countries
log(1 + robotshours
) 0.1370* 0.4044*** -0.2674*** 0.5716** 1.1913*** -0.6197***
(0.0688) (0.1247) (0.0702) (0.2094) (0.3193) (0.1593)
First stage :
Replaceability * initial GDPpc * global robot stock 0.0027*** 0.0027*** 0.0027***
(0.0003) (0.0003) (0.0003)
Observations 808,733 808,733 808,733 808,733 808,733 808,733
R2 0.6366 0.7677 0.3939 0.6333 0.7587 0.3908
Kleibergen-Paap rk Wald-F-stat 111.830 111.830 111.830
Importer-exporter-year effects Y Y Y Y Y Y
Industry-year effects Y Y Y Y Y Y
Notes: Table reports OLS and IV results of equation (1) in text, using the sub-samples for non-OECD countries depending on their level of
income per capita. Columns (1)-(3) report the OLS results, while columns (4)-(6) report the IV and corresponding first stage estimates. Robust
standard errors clustered by developed country in parenthesis. ***1% level, **5% level, *10% level.
Effects of robot price reductions on cost reduction
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
: U
nit C
ost o
f Tas
ks /
Wag
e
: Change in Unit Cost with Robotization
SouthNorthOther
Effects of Northern robotization on North-South trade in
automated sector
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
1
2
3
4
5
6
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Nor
th to
Sou
th
Relative Robotized Ind. Exports of North to South
TotalIntermediate
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Sou
th to
Nor
th
Relative Robotized Ind. Exports of South to North
TotalIntermediate
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
1
2
3
4
5
6
% C
hang
e in
Exp
orts
of N
orth
to S
outh
Robotized Ind. Exports of North to South
TotalIntermediate
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
0.5
1
1.5
2
2.5
3
3.5
% C
hang
e in
Exp
orts
of S
outh
to N
orth
Robotized Ind. Exports of South to North
TotalIntermediate
Effects of robot price reductions on robot use, cost
reduction, labor allocation and wages
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1R
obot
Use
per
Wor
ker
Robot Use per Worker in Robotized Industries
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
: U
nit C
ost o
f Tas
ks /
Wag
e
: Change in Unit Cost with Robotization
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-40
-35
-30
-25
-20
-15
-10
-5
0
% C
hang
e in
Num
ber
of W
orke
rs
Labor Allocation in Robotized Industries
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-3
-2
-1
0
1
2
3
4
% C
hang
e in
Rea
l Wag
e
Change in Wages
SouthNorthOther
Effects of robot price reductions on exports
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-25
-20
-15
-10
-5
0
5
10
15
% C
hang
e in
Exp
orts
(R
obot
ized
Fin
al)
Change in Exports - Robotized Final
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-15
-10
-5
0
5
10
% C
hang
e in
Exp
orts
(R
obot
ized
Inte
rmed
iate
)
Change in Exports - Robotized Intermediate
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
10
20
30
40
50
60
70
% C
hang
e in
Exp
orts
(O
ther
Fin
al)
Change in Exports - Other Final
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
10
20
30
40
50
60
70
% C
hang
e in
Exp
orts
(O
ther
Inte
rmed
iate
)
Change in Exports - Other Intermediate
SouthNorthOther
Effects of robot price reductions on exports, real GDP and
consumer prices
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
2
4
6
8
10
12
14
16%
Cha
nge
in E
xpor
ts (
All
Sec
tors
)
Change in Total Exports (All Sectors)
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-2
0
2
4
6
8
10
12
14
% C
hang
e in
Exp
orts
/GD
P (
All
Sec
tors
)
Change in Total Exports /GDP (All Sectors)
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
1
2
3
4
5
6
% C
hang
e in
Rea
l GD
P
Change in GDP
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0.9
0.92
0.94
0.96
0.98
1
1.02
% C
hang
e in
CP
I
CPI
SouthNorthOther
Effects of Northern robotization on North-South trade in
automated sector: frictionless trade
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
1
2
3
4
5
6
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Nor
th to
Sou
th
Relative Robotized Ind. Exports of North to South
TotalIntermediate
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Change in Robots/Workers
0
2
4
6
8
10
12
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Nor
th to
Sou
th
Relative Robotized Ind. Exports of North to South
TotalIntermediate
Effects of Northern robotization on South-North trade in
automated sector: frictionless trades
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Change in Robots/Workers
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Sou
th to
Nor
th
Relative Robotized Ind. Exports of South to North
TotalIntermediate
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Change in Robots/Workers
0
2
4
6
8
10
12
% C
hang
e in
Rel
ativ
e E
xpor
ts o
f Sou
th to
Nor
th
Relative Robotized Ind. Exports of South to North
TotalIntermediate
Effects of robot price reductions on robot use: frictionless
trade
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Rob
ot U
se p
er W
orke
r
Robot Use per Worker in Robotized Industries
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Rob
ot U
se p
er W
orke
r
Robot Use per Worker in Robotized Industries
SouthNorthOther
Effects of robot price reductions on real wages: frictionless
trade
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-3
-2
-1
0
1
2
3
4
% C
hang
e in
Rea
l Wag
e
Change in Wages
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
-2
-1
0
1
2
3
4
5
% C
hang
e in
Rea
l Wag
e
Change in Wages
SouthNorthOther
Effects of robot price reductions on real GDP: frictionless
trade
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
1
2
3
4
5
6
% C
hang
e in
Rea
l GD
P
Change in GDP
SouthNorthOther
0 10 20 30 40 50 60 70 80 90
% Reduction in Robot Price
0
1
2
3
4
5
6
% C
hang
e in
Rea
l GD
P
Change in GDP
SouthNorthOther
References I
Acemoglu, Daron and Pascual Restrepo, “Robots and Jobs: Evidence from US Labor Markets,” 2017. NBER
Working Paper 23285.
Arkolakis, Costas, Arnaud Costinot, and Andres Rodriguez-Clare, “New Theories, Same Old Gains?,” American
Economic Review, 2012, 102 (1), 94–130.
Atkin, David, Amit Khandelwal, and Adam Osman, “Exporting and Firm Performance: Evidence from a
Randomized Experiment,” Quarterly Journal of Economics, 2017, 132 (2), 551–615.
Autor, David, David Dorn, Gordon Hanson, Gary Pisano, and Pian Shu, “Foreign Competition and Domestic
Innovation: Evidence from U.S. Patents,” 2017. Unpublished manuscript.
Bastos, Paulo, Joana Silva, and Eric Verhoogen, “Export Destinations and Input Prices,” American Economic
Review, 2018, 108 (2), 353–392.
Bloom, Nicholas, Mirko Draca, and John Van Reenen, “Trade Induced Technical Change? The Impact of Chinese
Imports on Innovation, IT and Productivity,” Review of Economic Studies, 2016, 83 (1), 87–117.
Burstein, Ariel and Jonathan Vogel, “International Trade, Technology, and the Skill Premium,” Journal of Political
Economy, 2017, 125 (5), 1356–1412.
, Eduardo Morales, and Jonathan Vogel, “Changes in Between-group Inequality: Computers, Occupations and
International trade,” 2016. Unpublished Manuscript.
, Javier Cravino, and Jonathan Vogel, “Importing Skill-Biased Technology,” American Economic Journal:
Macroeconomics, 2013, 5 (2), 32–71.
Bustos, Paula, “Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of MERCOSUR
on Argentinian Firms,” American Economic Review, 2011, 101 (1), 304–340.
References II
Caliendo, Lorenzo and Fernando Parro, “Estimates of the Trade and Welfare Effects of NAFTA,” Review of
Economic Studies, 2015, 82 (1), 1–44.
, Maximiliano Dvorkin, and Fernando Parro, “Trade and Labor Market Dynamics: General Equilibrium Analysis
of the China Trade Shock,” 2015. NBER Working Paper No. 21149.
Chor, Davin, “Unpacking Sources of Comparative Advantage: A Quantitative Approach,” Journal of International
Economics, 2010, 82 (2), 152–167.
Dekle, Robert, Jonathan Eaton, and Sam Kortum, “Global Rebalancing with Gravity: Measuring the Burden of
Adjustment,” IMF Staff Papers, 2008, 55 (3), 511–540.
Donaldson, Dave, “Railroads of the Raj: Estimating the Impact of Transportation Infrastructure,” American
Economic Review, 2018, 108 (4-5), 899–934.
Eaton, Jonathan and Samuel Kortum, “Technology, Geography and Trade,” Econometrica, 2002, 70 (5), 1741–79.
Fieler, Ana Cecilia, “Non-Homotheticity and Bilateral Trade: Evidence and a Quantitative Explanation,”
Econometrica, 2011, 79 (4), 1069–1101.
Fort, Teresa C., “Technology and Production Fragmentation: Domestic versus Foreign Sourcing,” Review of
Economic Studies, 2017, 84 (2), 650–687.
Freund, Caroline and Diana Weinhold, “The Effect of the Internet on International Trade,” Journal of International
Economics, 2004, 62 (1), 1069–1101.
Graetz, Georg and Guy Michaels, “Robots at work,” Review of Economics and Statistics, 2018.
Lileeva, Alla and Daniel Trefler, “Improved Access to Foreign Markets Raises Plant-level Productivity...For Some
Plants,” Quarterly Journal of Economics, 2010, 125 (3), 1051–1099.
References III
Parro, Fernando, “Capital-Skill Complementarity and the Skill Premium in a Quantitative Model of Trade,”
American Economic Journal: Macroeconomics, 2012, 5 (2), 72–117.
Verhoogen, Eric, “Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing Sector,” Quarterly
Journal of Economics, 2008, 123 (2), 489–530.
Waugh, Michael, “International Trade and Income Differences,” American Economic Review, 2010, 100 (5),
2093–2124.
Yi, Kei-Mu, “Can Vertical Specialization Explain the Growth of World Trade?,” Journal of Political Economy,
2003, 111 (1), 52–102.
