HOV with technology and consumption dissimilarity

Transcription

1 bilaterally HOV with technology and consumption dissimilarity Neil Foster and Robert Stehrer The Vienna Institute for International Economic Studies (wiiw) Version: Study carried out within the 7 th EU framework program project WIOD Grant agreement number April 24-26, WIOD conference, Groningen, The Netherlands

2 bilaterally Heckscher-Ohlin-Vanek and Ricardo with technology differences Technology differences Accounting for traded intermediates Role of consumption patterns Introducing consumption dissimilarity Results preliminary /further steps

3 bilaterally Heckscher-Ohlin model Ricardo: Technology differences determine patterns of trade 2 countries, 2 sectors, 1 factor (labour) Countries specialize in goods with relative productivity advantage England: Textiles; Portugal: Wine Heckscher-Ohlin: Factor endowments determine patterns of trade 2 countries (L, C), 2 sectors (textiles, computers), 2 factors Same technologies across countries Identical and homothetic preferences Factors can move between sectors but not between countries Note: Heckscher and Ohlin have been well aware of the strictness of assumptions and limitations (see Baldwin, 2010)

4 bilaterally Country L: Labor abundant; country C: capital abundant Autarky patterns Relative price of labor lower in country L Country L produces relatively more of the labor intensive good Textiles in country L relatively cheaper Heckscher-Ohlin: Factor endowments determine patterns of trade Free trade Country L specializes in textiles Relative price of textiles increases in country L Relative price of textiles decreases in country C Factor price convergence (Stolper-Samuelson) and equalization County L exports textiles; Country C exports computers

5 bilaterally Issues Notorious problems with generalizations Strong assumptions (relaxations might make predictions ambiguous) Nonetheless: model became workhorse for decades Looking at trade in factors helps Vanek, 1968; Travis, 1964; Melvin, 1968 Allows to increase dimensionality of model in any direction Example above: County L is net exporter of labor

6 bilaterally However Still little empirical support of model in factor trade Little empirical support (e.g. Leontief paradoxon, 1953 and 1956) US was net exporter of labor (though capital abundant country) Leontief, 1953 and 1956; many others Results on testing predictions:... like a coin toss... (Trefler, 1993) Allowing for technology differences helps Trefler, 1993 and 1995; others Supporting original explanation given by Leontief ( Leontief was right! ) As US workers are more productive, US is - in productivity adjusted terms - labor abundant Davis and Weinstein (2002); Reimer (2006); Trefler and Zhu (2010) Including traded intermediates Trefler and Zhu (2010); Deardorff (1982)

7 bilaterally with traded intermediates t r f d f b f tf r = d f (I A) 1 t = b f t = b }{{} f f r sf r b f f = Vf r sf r V f }{{} MeasuredFCT PredictedFCT... Measured net trade in factor f... NC 1 vector direct factor inputs... NC 1 vector direct and indirect factor inputs I... Identity matrix A... NC NC coefficients matrix t r... Country r trade vector (exports positive, imports negative) V r f... Country r s endowment with factor f s r... Share of country r in world consumption V f... World endowment with factor f

8 bilaterally Interpretation Predicted FCT: Vf r sf r V f Measured FCT: d f (I A) 1 t r Underlying assumptions Identical and homothetic preferences Factor-price equalization (in productivity adjusted terms) Measuring FCT allows for productivity differences and trade in intermediates Strong consumption similarity (SCS): f pr = s r f Country consumes proportion s r of final goods produced by every country Necessary and sufficient condition that Vanek prediction holds Weak consumption similarity (WCS): f r = s r f Country r s consumption of good is proportional to what world consumes (regardless of where good is produced) Standard HOV: Condition if technologies are equal

9 bilaterally Empirical applications t r f = d f (I A) 1 t r = V r f s r f V f Assuming same technologies: HOV prediction is like a coin toss Trefler, 1993 and 1995; Davis and Weinstein, 2001; Trefler and Zhu, 2010 Allowing for technology differences: HOV prediction is met quite will Various tests: sign test, rank correlation test, regression However, large amount of missing trade i.e. predicted trade is larger than measured trade in factors Home bias Non-homothetic preferences Quality of products traded bilaterally Solutions: Trefler and Zhu (2010): Imposing structures on selected sectors Cassing and Nishioka (2010): Introducing a vector of consumption differences)

10 bilaterally Accounting for consumption dissimilarity t r f = d f (I A) 1 t r = V r f s r f V f = b f f r b f s r f f = V r f = V r f b f s r f If b f S r f

11 bilaterally Assumptions on consumption structures: S r Case Preferences Sourcing structures SCS Homothetic Proportional (1) Non-homothetic Proportional (2) Homothetic Empirical sourcing at country level (3) Homothetic Empirical sourcing at country-industry level (4) Non-homothetic Empirical sourcing at country level HOV Non-homothetic Empirical sourcing at country-industry level

12 bilaterally Tests Sign test (90% and more) Rank correlations are high and significant (ρ = 0.9) Already with assumption of SCS Only slight improvements when introducing modifications Regression tests: FCT measured,r = α + β FCT predicted,r + ε r Various specifications (OLS, FE, RE,...)

13 bilaterally Strong consumption similarity: Total employment _1, EMP 50 Measured FCT 0-50 DEU FRA GBR ITA ESP CAN AUT CZE CYP BGR BRA AUS BEL SWETUR TWN MEX DNK GRC FIN PRT IRL POL HUN LUX SVN SVK MLT EST LTU LVA ROU NLD KOR RUS Predicted FCT

14 bilaterally Non-homothetic and country-level sourcing (selected sample): Total employment _4, EMP 10 BRA Measured FCT 0-10 FRA CAN AUS ESP KOR NLD ITA CZE BGR LUX CYP HUN LTU EST LVAMEX ROU SVK SVN MLT POL DNK FIN PRT IRL TWN SWE GRC AUT TUR BEL RUS -20 GBR -30 DEU Predicted FCT

15 bilaterally Non-homothetic and country-level sourcing: Total employment 200 _4, EMP 100 CHN Measured FCT 0 JPN OTH IDN IND -100 USA EU Predicted FCT

16 bilaterally Results (RE) Table 12 - Slope coefficients - Random effects Total sample Reduced sample Total High Medium Low Capital Total High Medium Low Capital Strong consumption similarity FCTp s.e R Non-homothetic preferences, proportionality FCTp s.e R Homothetic, Non-proportionality at country level FCTp s.e R Homothetic, Non-proportionality at country-industry level FCTp s.e R Non-homothetic, Non-proportionality at country level FCTp s.e R

17 bilaterally A bilateral test of HOV Derived from world general equilibrium condition (Hakura, 2001) Note: Not to be confused with bilateral factor content of trade (Helpman, 1984; Staiger, 1986; Choi and Krishna, 2004; Lai and Zhu, 2007; Foster and Stehrer, 2012) ( ) b(t r t c ) = (V r V c ) + b S c (S r ) 1 I f r ( ) = (V r V c ) + b S c (S r ) 1 I S r f ( ) = (V r V c ) + b S c S r f

18 bilaterally Results (RE) Table 17 - Slope coefficients - RE Total sample Reduced sample Emp. High Medium Low Capital Emp. High Medium Low Capital Strong consumption similarity FCTp s.e R Non-homothetic preferences, proportionality FCTp s.e R Homothetic, Non-proportionality at country level FCTp s.e R Homothetic, Non-proportionality at country-industry level FCTp s.e R Non-homothetic, Non-proportionality at country level FCTp s.e R

19 bilaterally HOV was - empirically - never alive HOV + Ricardo + Home bias + Gravity + Linder are doing the job Give credit to other aspects of H&O (see Baldwin, 2010): Scale effects Specialization dynamics Role of factor endowment changes and technology upgrading