Deep Agreements and the Internationalization of Production Alberto Osnago, Nadia Rocha WTO Michele Ruta IMF

University of Exeter, 30 May 2013

Deep agreements and int. production

• Preferential trade agreements (PTA) are usually thought as tariff reductions (i.e. shallow agreements) • Deep agreements cover in addition many other provisions: – Technical barriers to trade (TBT) measures – Sanitary and phytosanitary (SPS) measures – Investment – Intellectual property rights (IPR) protection – Anti-corruption – Competition policy – … Deep PTAs and vertical FDI 2

Deep agreements and int. production

• Depth of PTAs and the international fragmentation of production have changed over time Source World Trade Report (2011) Deep PTAs and vertical FDI 3

Deep agreements and int. production

• This paper: • Digs further into the relationship between deep trade agreements and the internationalization of production • • Specific question: How are deep agreements and vertical FDI related?

• • • • General idea: Deep provisions improve the contractual environment However, different provisions affect the contractibility of various inputs differently FDI respond to (and, possibly, determine) these institutional changes Deep PTAs and vertical FDI 4

Related literature

•

International organization of production

– Antras (2003); Antras and Helpman (2004); Nunn and Trefler (2012); … •

Domestic institutions and intra-firm trade

– Grossman and Helpman (2005); Antras and Helpman (2008); Bernard et al. (2010); … •

Trade agreements and internationalization of production

– Lawrence (1996); Baldwin (2010); Orefice and Rocha (2012); Blanchard and Matschke (2012); Baltagi et al. (2008); … Deep PTAs and vertical FDI 5

Outline 1.

–

Theory: Vertical FDI and deep PTAs

Antras and Helpman (2008)

2. Data description and methodology

– Depth and composition of deep agreements – Vertical FDI

3. Empirical findings

– Depth of PTAs and vertical FDI – Composition of PTAs and vertical FDI

4. Summary and work ahead

Deep PTAs and vertical FDI 6

Vertical FDI and deep PTAs

• Based on Antras and Helpman (2008) – Property rights approach of the international organization of production under asymmetric contractual frictions • Two countries: North and South – North is high-cost and has complete contracting – South is low-cost and has weaker contracting institutions relative to North • Final good producers located in North – They have to produce headquarter (HQ) services, but can source components domestically or from South Deep PTAs and vertical FDI 7

Vertical FDI and deep PTAs

• Final good production in an industry combines HQ service (h) and components (m) with technology 𝑞 θ = θ ℎ 𝜔 η η 𝑚 𝜔 1−η 1−η • Sectors vary by intensity with which they use HQ services (η) • Firms within sectors vary by productivity ( θ ) Deep PTAs and vertical FDI 8

Vertical FDI and deep PTAs

• Each input is produced with a continuum of activities in the interval [0,1] 𝑙 = exp 0 1 log 𝑥 ℎ 𝑖 𝑑𝑖 with 𝑙 = ℎ, 𝑚 • Imperfect contractibility in South – Only activities in range contractible 0, 𝜇 𝑙 with 0 ≤ 𝜇 𝑙 ≤ 1 are • Contractibility depends on domestic institutions in South (λ) and on PTA provisions (γ) 𝜇 ℎ = ℎ(λ, 𝛾 1 , … , 𝛾 𝑇 ) and 𝜇 𝑚 = 𝑚(λ, 𝛾 1 , … , 𝛾 𝑇 ) Deep PTAs and vertical FDI 9

Vertical FDI and deep PTAs

• Differences in contractibility across production processes and countries reflect technological and institutional variation • Deep provisions in PTAs are a determinant of institutional variation and affect contractibility of inputs • Examples: – – Anti-corruption provisions reduce contractual insecurity (↑ 𝜇 ℎ and ↑ 𝜇 𝑚 ) IPR provisions improve contractibility of HQ services (namely R&D) by protecting patents (↑ 𝜇 ℎ ) – TBT/SPS provisions improve contractibility of components by promoting standardization / mutual recognition (↑ 𝜇 𝑚 ) Deep PTAs and vertical FDI 10

Vertical FDI and deep PTAs

• Final good producer has three alternatives to obtain components (m) – Domestic sourcing (D) – Foreign outsourcing (O) – FDI, i.e. foreign integration (V) • Profits depend on control and location choice and can be expressed with the standard form 𝜋 𝑖 = 𝑍 𝑖 𝜗 − 𝑤 𝑁 𝑓 𝑖 with 𝑖 = 𝐷, 𝑂, 𝑉 where 𝜗 is a function of productivity θ and profitability 𝑍 𝑖 depends on HQ intensity η (Antras, 2003) Deep PTAs and vertical FDI 11

Vertical FDI and deep PTAs

• Domestic production, foreign outsourcing and FDI coexist π V π O π D -w N f D -w N f O -w N f V ϑ D ϑ O ϑ V Domestic prod Outsource FDI Deep PTAs and vertical FDI ϑ 12

-w N f D -w N f O -w N f V

Vertical FDI and deep PTAs

π V ’ π V π O ’ π O π D ϑ D ϑ O ’ ϑ O ϑ V ’ ϑ V Outsourcing Deep PTAs and vertical FDI FDI ϑ 13

Vertical FDI and deep PTAs

• PTA provisions that improve contractibility of HQ services ( ↑𝜇 ℎ ) decrease FDI π V ’ π V π O ’ π O π D ϑ D ϑ O ’ ϑ O ϑ V ϑ V ’ -w N f D -w N f O -w N f V Outsourcing Deep PTAs and vertical FDI FDI ϑ 14

Vertical FDI and deep PTAs

• Summary of model’s prediction: – Depth of agreements is associated to more offshoring, • but relationship with FDI is ambiguous Association is stronger for HQ intensive industries – Discipline improving contractibility of components (HQ services) are associated with increasing (decreasing) FDI – • • Property rights versus transaction cost (TC) model Better contractual institutions are always associated to more outsourcing and less FDI in the TC model Empirical analysis provides an indirect test of the two theories Deep PTAs and vertical FDI 15

•

Depth and composition of PTAs

We use WTO data on the content of PTAs • Comprehensive mapping and coding of 100 PTAs signed between 1958-2011 • We restrict the sample to PTAs signed by Germany (EU), Japan, and USA • We are left with 59 agreements (40 signed by EU, 12 by JAP, and 11 by USA) • PTA dataset has been constructed following the methodology of Horn et al. (2010) – Horn et al. (2010) identify a set of 52 policy areas – Legal enforceability of PTA obligations is established according to the language of the agreements Deep PTAs and vertical FDI 16

Depth and composition of PTAs

Country

Germany Japan USA All

N. Of Agreements N. Of Partners

37 78 11 14 11 17 59 109

Avg n. Of Provisions

21.93

18.51

18.06

20.99

Min n. of Provisions

5 13 9 5

Max n. of Provisions

35 23 21 35 Deep PTAs and vertical FDI 17

Depth and composition of PTAs

• To analyze the depth of PTAs, we construct three variables (as in Orefice and Rocha, 2013): – Total count of enforceable provisions (# Provisions) – Top 5 and top 10 provisions with the highest degree of commonality across the agreements • To analyze the composition of PTAs, we classify provisions into two groups: – HQ-provisions: GATS, TRIPS, IPR, investment, and movement of capital – M-provisions: SPS, TBT, consumer protection, customs, and export taxes Deep PTAs and vertical FDI 18

Depth and composition of PTAs

Germany Japan USA HQ-provisions

GATS TRIPS IPR Investment Movement of capital

M-provisions

SPS TBT Consumer protection Customs Export taxes 11 22 22 13 23 6 8 4 35 27 7 8 11 11 5 11 10 7 10 10 10 11 10 8 8 9 8 11 10 10 Deep PTAs and vertical FDI 19

Number of agreements with specific provisions by country Deep PTAs and vertical FDI 20

Measuring vertical FDI

• FDI data has been constructed using the ORBIS database assembled by Bureau van Dijk – ORBIS includes location, ownership, detailed sector level, and operational data (e.g. revenues) for more than 100 million firms in Europe, Americas, and Asia-Pacific region • We restrict our analysis to subsidiaries in any country owned by parent firms located in Germany, Japan, and USA in 2003, 2007, and 2011 – We have the revenues of 125,212 subsidiaries – We can identify 42,984 ultimate owner parents Deep PTAs and vertical FDI 21

•

Measuring vertical FDI

We follow the methodology in Alfaro and Charlton (2009) to identify • the ownership structure: • We consider a firm to be a parent if it owns a minimum of 25.01% of another firm • and the type of link between a subsidiary and its parent firm (i.e. vertical versus other forms of FDI): • Let S be the set of 6-digits NAICS codes of the subsidiary and P be the set of 6-digits NAICS code of the parent • An element x of S is an input of an element z of P if the total requirements coefficient of the US IO table > 0.03

Deep PTAs and vertical FDI 22

•

Measuring vertical FDI

We can identify 4 types of connections: • Horizontal FDI: if S and P share any element • Vertical FDI: if any element of S is an input of any element of P • Complex FDI: if S and P share any element and any element of S is an input of any element of P • Non-identified: if none of the above is satisfied  The share of vertical and horizontal links we obtain is in line with the findings of Alfaro and Charlton (2009) and Lanz and Miroudot (2011) Horizontal Complex Vertical Deep PTAs and vertical FDI 23

Measuring vertical FDI

Deep PTAs and vertical FDI 24

Depth of PTAs and vertical FDI

• Ideally, we would want information on intra-firm trade, but data are not available • Quantification of vertical FDI –

FDI ijkt

is the aggregate value of the revenues of subsidiaries in country (destination) j owned by parents operating in sector k, country i (US, Japan, or Germany) at time t • Example: – Vertical FDI of the car sector is the sum of revenues of all the foreign-owned subsidiaries that produce car inputs, such as plastic, seat-belts, glass. Deep PTAs and vertical FDI 25

Depth of PTAs and vertical FDI

•

Other explanatory variables:

– HQ intensity (η) is constructed as the log of capital expenditures divided by total worker wages using data from the American Manufacturing Survey (AMS) in 2007 – Rule of law is taken from the Worldwide Governance Indicators database – GDP and GDP per capita come from the World Bank – Distance, contiguity, colony relationship, common language come from Mayer and Zignago (2011) Deep PTAs and vertical FDI 26

Depth of PTAs and vertical FDI

•

FDI ijkt =α+β 1 DEPTH ijt + β 2 INSTITUTIONS jt + γ 1 X jt +γ 2 X ij + δ t + δ k + δ i + δ it

• • • where k is sector, t is time, i and j are country indexes (i for the "origin" country and j for the "destination" country) DEPTH ijt is # Provisions and Top 5 and top 10 provisions

X jt

are controls for characteristics of the destination country that vary over time (GDP and GDP per capita)

X ij

are country-pair variables (distance, contiguity, common language, colonial relationship)

δ t , δ k , δ i, δ it

fixed effect (1) are time, sector, country (origin), and country-time Deep PTAs and vertical FDI 27

Depth of PTAs and vertical FDI

FDI and Deep Integration

VARIABLES PTA (1) 0.527** (0.208) (2) (3) FDI (log of revenues in 1000$) (4) N. of Provisions 0.0169** (0.00736) log(Top 5) 0.543* (0.297) log(Top 10) Dummy=1 if Rule of Law η>average 0.802*** (0.284) 0.306** (0.126) 0.806*** (0.285) 0.282** (0.123) 0.804*** (0.284) 0.282** (0.124) 0.478** (0.218) 0.803*** (0.284) 0.290** (0.125) Observations R-squared Year FE 4,951 0.249

Yes 4,914 0.246

Yes 4,914 0.245

Yes 4,914 0.246

Yes Industry-4dig FE Country FE Yes Yes Yes Yes Yes Yes Yes Yes Country-Year FE Yes Yes Yes Yes Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

All regressions control for distance, contiguity, colony relationship, common language, BIT, a dummy for China, GDP, GDP per capita, and remoteness of the country of the subsidiary.

Deep PTAs and vertical FDI 28

Composition of PTAs and vertical FDI

• FDI ijkt =α+β 1 μ M ijt + β 2 μ H ijt + β 3 DEPTH ijt + β 4 INSTITUTIONS jt + γ 1 X jt +γ 2 X ij + δ t + δ k + δ i + δ it • • • where k is sector, t is time, i and j are country indexes (i for the "origin" country and j for the "destination" country) μ M ijt , μ H ijt described below and DEPTH ijt is as described above

X jt

are controls for characteristics of the destination country that vary over time (GDP and GDP per capita)

X ij

are country-pair variables (distance, contiguity, common language, colonial relationship)

δ t , δ k , δ i, δ it

fixed effect (2) are time, sector, country (origin), and country-time Deep PTAs and vertical FDI 29

Composition of PTAs and vertical FDI

•

We construct two variables

μ ℎ

and

μ 𝑚

:

–

Dummy

μ 𝑙 = 1, if there is at least one provision of the l-type in the PTA 2 if all provisions of l-type in PTA –

Discrete

μ 𝑙 = 1 if at least one provision of l-type, 0 otherwise where 𝑙 = ℎ, 𝑚 Deep PTAs and vertical FDI 30

Composition of PTAs and vertical FDI

FDI, different provisions, and depth

(1) VARIABLES Dummy μ H 0.678* (0.389) Dummy μ M (2) 0.968*** (0.288) (3) (4) FDI (log of revenues in 1000$) -0.665

(0.551) 1.448*** (0.393) Discrete μ H 0.269

(0.221) Discrete μ M (5) 0.569*** N. of Provisions -0.00476

(0.0155) -0.0114

(0.0102) -0.00283

(0.0155) 0.000308

(0.0183) (0.172) 0.000703

(0.00747) Rule of Law Observations R-squared Year FE Industry FE Country FE Country-Year FE 0.243** (0.108) 7,108 0.337

Yes Yes Yes Yes 0.267** (0.110) 7,108 0.339

Yes Yes Yes Yes 0.271** (0.109) 7,108 0.339

Yes Yes Yes Yes 0.237** (0.108) 7,108 0.337

Yes Yes Yes Yes 0.286** (0.111) 7,108 0.338

Yes Yes Yes Yes Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(6) 0.133

(0.233) 0.534*** (0.182) -0.00775

(0.0179) 0.287** (0.111) 7,108 0.338

Yes Yes Yes Yes All regressions control for distance, contiguity, colony relationship, common language, BIT, a dummy for China, GDP, GDP per capita, and remoteness of the country of the subsidiary.

Deep PTAs and vertical FDI 31

Conclusion –future work

• • • So far we have said little about direction of causality – Control decisions of firms are expected to respond to depth and composition of PTAs, but firms lobby on content of trade agreements – These aspects are not captured by the model and empirical strategy On the theory side: – Need to embed FDI decisions in a model of endogenous PTA formation On the empirical side: – Need to instrument PTA depth and composition – For PTA depth between country i and country j , we can use instruments from poli-science literature (e.g. affinity of political systems) or average depth signed with other countries – Not obvious how to instrument for composition of PTA Deep PTAs and vertical FDI 32

Conclusion -summary

• We use the AH model to guide our analysis of the relationship between deep PTAs and the internationalization of production • We exploit two new datasets on depth and composition of PTAs and on vertical FDI • Consistently with the theory, we find that: 1. Depth of PTA is associated to an increase in FDI (this finding is not robust) 2. PTA provisions that improve the contractibility of components relative to HQ activities are associated to more FDI (this supports PR over TC approach) Deep PTAs and vertical FDI 33

-w N f D -w N f O -w N f V

Transaction costs model

π V π O ’ π O π D ϑ D ϑ O ’ ϑ O ϑ V ϑ V ’ Outsourcing Deep PTAs and vertical FDI FDI ϑ 34

• •

Intuition PR model

Higher component contractibility (left) increases the optimal revenue share of final good producer Higher HQ contractibility (right) lowers the optimal revenue share of final good producer Β(η) Β(η) η Deep PTAs and vertical FDI η 35

Sectors with low HQ intensity

• Domestic production and foreign outsourcing coexist, but no FDI π O π D π V ϑ D ϑ O -w N f D -w N f O -w N f V Domestic prod Outsourcing ϑ Deep PTAs and vertical FDI 36

-w N f D -w N f O -w N f V

Sectors with low HQ intensity

ϑ D ϑ O ’ ϑ O Outsourcing Deep PTAs and vertical FDI π O ’ π O π D π V ’ π V ϑ 37

Location choice

Deep PTA

Control choice Final-Good Producer Domestic sourcing Intra-Firm Trade inputs Offshoring Foreign sourcing Arm’s Length Trade Vertical Integration (FDI) Foreign Outsourcing

Location choice

• Location choice in a simple model of global sourcing with heterogeneous firms π O π D -w N f D -w N f O ϑ D ϑ O Domestic production Offshoring Deep PTAs and vertical FDI ϑ 39

-w N f D -w N f O

Location choice

ϑ D ϑ O ’ ϑ O Offshoring Deep PTAs and vertical FDI π O ’ π O π D ϑ 40

Ownership: ORBIS definition

• The definition of ownership “concern the minimum percentage that must characterize the path from a subject company up to its Ultimate Owner” Firm 1 54% Firm 2 35% Firm 3 100% Firm 4 • • Considering a path of minimum 25.01%: the ultimate owner of firm 4 is firm 1 Considering a path of minimum 50.01% : the ultimate owner of firm 4 is firm 3 Deep PTAs and vertical FDI 41

Distribution of Vertical, Horizontal, and Complex FDI

Type

Vertical Horizontal Total

Number of Subsidiaries

25,230

Share

26,904 52,134 48.39

51.61

100 • The share of vertical and horizontal links is in line with the findings of Alfaro and Charlton (2009) and Lanz and Miroudot (2011) Deep PTAs and vertical FDI 42