Introduction Theorical Framework Empirical implementations Estimation Conclusion

Theory and evidence Daniel M. Bernhofen Journal of International Economics

Accepted 14 March 1997

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Table of contents

1

Introduction

2

Theorical Framework

3

Empirical Implementation

4

Estimation

5

Conclusion

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Literature Review New Trade Theory Originality Motivations

Literature Review

Traditional theories of comparative advantage fail in explaining IIT Emergence of an Intra-Industry Trade literature : Balassa (1966), Grubel (1967) : empirical works Grubel and Lloyd (1975) : Grubel and Lloyd index of IIT Brander (1981) : provides a theorical framework

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Literature Review New Trade Theory Originality Motivations

New Trade Theory :

1

Large numbers models : Krugman, Helpman 1 2

monopolistic competition IIT in differenciated products

IIT stems from preferences for variety and scale economics. 2

Small numbers model : Brander (1981) 1 2 3

duopolistic, segmented markets homogeneous products Cournot-Nash game

IIT stems from incentives to capture foreign monopoly rent.

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Literature Review New Trade Theory Originality Motivations

Originality

positive approach : explaining the trade pattern

allows for country specific differences

focused on a single industry : petrochemicals

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Literature Review New Trade Theory Originality Motivations

Motivations

1

fill the lack of theory-based empirical studies

2

IIT : « Not a figment of the product grouping » (Finger, 1975)

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Hypotheses Profit functions Grubel and Lloyd Index Theorical results

Hypotheses

Two countries Single industry (nh , nf firms) Oligopolistic model Homogeneous product Constant marginal costs (ch , c f ) Same demand behaviors

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Hypotheses Profit functions Grubel and Lloyd Index Theorical results

Profit functions

Solving a Cournot-Nash game : πh =xhh p(Qh/sh )+xhf p(Qf /sf )-ch (xhh +xhf )

πf =xff p(Qf /sf )+xfh p(Qh/sh )-cf (xff +xfh )

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Hypotheses Profit functions Grubel and Lloyd Index Theorical results

Grubel and Lloyd Index (1975)

with 0 ≤ ρ ≤ 1

1

hf −Xfh | ρ = 1 − |X Xhf +Xfh

2

hf −nf xhf | ρ = 1 − |nnhh xxhf +nf xhf

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Hypotheses Profit functions Grubel and Lloyd Index Theorical results

Therorical results

Benchmark : same cost (c h = cf ), same market size (s h = sf ), same market concentration (nh = nf ). 1

Different market concentrations : (nh 6=nf )

2

Different costs : (c h 6=cf )

3

Different market sizes : (s h 6=sf )

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Hypotheses Profit functions Grubel and Lloyd Index Theorical results

Theorical results

Proposition : Asymetry in any parameter decreases the IIT level. Exemple with different degrees of market concentration (nh 6=nf ) : −nf | ρ = 1 − |nnhh +n V decreasing level of IIT as nh and nf diverge. f

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Aim Data

Aim

Test empirically the proposition in a specific domain

Grubel-Lloyd index ρ = ρ

∂ρ

Show ∂

|ch −cf | ch +cf

∂ρ

< 0; ∂

|sh −sf | sh +sf



|ch −cf | |sh −sf | |nh −nf | ch +cf ; sh −sf ; nh −nf

∂ρ

< 0;

Daniel M. Bernhofen

∂

|nh −nf | nh +nf



<0

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Aim Data

Data

Petrochemicals Industry analysis of bilateral intra industry trade : From 1988 to 1992 , USA & Germany 38 traded petrochemicals : homogeneous, standardized Varying levels of IIT for each products. Oligopolistic competition, capital intensive, fixed costs. Profit determined by the number of producers and production costs. Cournot-Nash model, a reasonable framework.

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Aim Data

Data

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Aim Data

Data Dependant variable : Grubel Lloyd Index : it −Mit | IITit = 1 − |X Xit +Mit

Independant variables : COSTDIF : raw materical costs, utility costs. Standford Research Institute. SIZEDIF : we use proxy for each country’s market size : Annual product sales. FIRMDIF : Number of firms in USA (SOC), in Germany (WPDP).

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Aim Data

Potential bias tied to petrochemicals industry

Intrafirm trade

Vertical Integration

Symmetric transport cost

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Check correlation Functionnal form Robustness and Regressions Results

Potential bias Analysis of a potential bias : Source of error : ∂ρ

we supposed ∂

|ch −cf | ch +cf

∂ρ

< 0; ∂

|sh −sf | sh +sf

∂ρ

< 0; ∂

market structure configurations.

|nh −nf | nh +nf

< 0 hold in all

A high correlation between variables implies an underestimation of IIT Correlation test between pairs of explanatory variables. Result : No positive correlation Each independant variables explain "distinct influences" on IIT . Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Check correlation Functionnal form Robustness and Regressions Results

The Functionnal form choice 1

OLS : IITit =α + β COSTDIFit + γSIZEDIFit + δ FIRMDIFit + εit risk that the predicted value of IIT index lies outside [0,1]

2

Logit :   transformation log

3

IITit 1−IITit

= α + β COSTDIFit + γSIZEDIFit + δ FIRMDIFit + εit

Non linear least square : 1 IITit = 1+exp−(α+β COSTDIFit +γSIZEDIF + εit it +δ FIRMDIFit )

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Check correlation Functionnal form Robustness and Regressions Results

Robustness and Regressions Robustness Few data, missing observations per year. Large amount of "zero IIT" observations. Check robustness of estimations with regards to "zero IIT" outliers. Regressions 1

Regression with annual dummy variables.

2

Annual Regression. 1 2

With full observations. With restricted observations : excluding zero IIT observations Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Check correlation Functionnal form Robustness and Regressions Results

Results

Figure: Estimates for pooled data

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Check correlation Functionnal form Robustness and Regressions Results

Results for annual estimation

Daniel M. Bernhofen

IIT and Strategic Intereaction

Introduction Theorical Framework Empirical implementations Estimation Conclusion

Conclusion

IIT « Not a figment of the product grouping » Through an extention of the Brander Model, Find determinants of IIT. Test theory on petrochemicals industry. Empirically relevant. Negative impact of country specific differences on intensity of bilateral IIT. Pave the way for "positive" studies on IIT under oligopolistic competition.

Daniel M. Bernhofen

IIT and Strategic Intereaction