The Heckscher-Ohlin model: Empirics I (factor content studies)

Transcription

1 The Heckscher-Ohlin model: Empirics I (factor content studies) Robert Stehrer Version: May 1, 2013

2 Types of empirical studies: Factor contents of trade studies Cross-section studies Commodity composition methods: regress measure of trade performance on determinants of comparative advantage Baldwin (1971); Branson and Junz (1971); Harkness and Kyle (1971); Branson and Monoyios (1977); Harkness (1978); Stern and Maskus (1981); Stern (1976); etc. Cross-country method: regress measure of commodity characteristic (e.g. capital intensity) on country characteristic Balassa (1979); Baldwin (1979); Bowen (1983); Balassa (1986); etc. Test assumptions of model (Non)reversibility of factor intensities Homothetic preferences Assumption of identical production functions

3 The Leontief paradox (Leontief, 1953) Input-output accounts for US allows to compute labor L and capital K used in each industry Production of commodity requires inputs from own and other industries Can be described by matrix capturing inter-industry transactions (per unit of gross output) B L = I + B + B 2 + = (I B) 1 US trade data to compute amounts of K and L directly and indirectly used for Result: production of US exports and imports E.g. Capital embodied in exports = k Le Hypothesis: US is capital abundant, therefore should be net exporter of capital (embodied in exported goods) using (HO) assumption that foreign countries use same technology as US Exports Imports Capital stock ($ million) $ 2.5 $ 3.1 Labor (person-years) Capital/Labor $ $ Remarkable result: K/L is higher in imports than in exports which contradicts the HO result (under the assumption that the US was capital abundant in 1947)

4 Further study, Leontief (1956) on 1951 data (and more detailed industry level): Paradox still appears but less strong When removing 19 industries (as non-competitive imports) the ratio becomes less than one Various explanations (see e.g. Deardorff, 1984) US and foreign technologies are not the same Ignoring other factors (e.g. land, skilled labor) Unusual year (1947) US was not engaged in free trade Taking account of this: Paradox continued to occur (in some/most cases) Debate continued: Eg. Leamer (1980) argued that Leontief test was wrong and suggested alternative test based on (factor contents of trade)

5 However note: Leontief did not interpret his result as refutation of HOS theorem Argued that labor is more productive in US; thus, in efficiency terms (i.e. taking productivity into account) US was labor abundant country Other Leontief type studies: Tatemoto and Ichimura (1959) on Japan; Bharadwaj (1962) on India-US trade; Roskamp (1963) on Germany. On US: Baldwin (1971); Stern and Maskus (1981). Leamer (1984); Bowen, Leamer and Sveikauskas (1987) using more sophisticated method Trefler (1993, 1995) Davis and Weinstein (2001)

6 Heckscher-Ohlin-Vanek (HOV) model (Vanek, 1968) Idea Calculate factor content of exports and imports E.g. how much labour is (directly and indirectly) embodied in a country s exports Application: Calculate CO2-Emission of demand Trade in value added terms Applicable in many dimensions Assumptions and notations Countries i = 1,..., C Industries j = 1,..., N Factors k = 1,..., M Assume (for the moment) that Technologies are identical and homothetic Factor price equalization prevails (under free trade)

7 Total factor requirements (per unit of output) a a 1M A M N = a N1... a NM Matrix A includes both direct and indirect factor inputs A BA = à A(I B) = à A = ÃM N(I B N N ) 1 where à is the matrix of direct factor requirements and B the input-output matrix.

8 YN 1 i... vector of outputs (in country i) DN 1 i... vector of (domestic) demands (in country i) Vector of net exports (Output minus domestic demand) TN 1 i = Y i D i from which the factor content of trade is FM 1 i = AT i Similarly, demand for factors in country i is AY i which - under full employment assumption - equals endowments AY i = V i

9 Under assumptions where identical and homothetic tastes across countries AD i = s i AD w D w N 1... vector of world demand s i... country i s share in world consumption, i.e. s i = p D i /p D w Under balanced trade - expenditure = income in each country - p D i = p Y i s i = p Y i /p Y w = GDP i /GDP w As world consumption equals world production AD i = s i AD w = s i AY w 0 = s i V w where last condition follows from full-employment assumption at world level.

10 The factor content of trade can therefore be written as F i = AT i = AY i AD i = V i s i V w If country i s endowment of factor k, Vk i, relative to world endowment exceeds country i s share of world GDP, we say country i is abundant in that factor. From this it follows that Fk i > 0 i.e. country is expected to be net exporter of this factor (e.g. Leontief expected US to be net exporter of capital) Conversely, if country i is relatively scarce in factor j.

11 Accounting for technological differences (Trefler, 1993 and 1995) Two possibilities (though related) Differences in productivity of factors Differences in factor requirement matrix A Sketch of argument Let πk i denote productivity of factor k in country i relative to US πk i V k i denotes effective factor endowment A denotes amount of effective factors needed per unit of output Factor price equalization still holds (in terms of effective factor prices) Implies - under assumption of identical technologies - that A is same across countries

12 Factor content of trade is still F i = AT i and HOV equation (in terms of effective factor endowments) Fk i = πi k V k i C si π j k V j k j=1 MC equations and M(C 1) productivity parameters As world exports must equal world imports (in factor services) one can drop one country, e.g. US This implies (Trefler, 1993) Equations will be independent ( almost everywhere ) Productivity parameters enter linearly Exists a solution for πk i such that equation above holds HOV is no longer testable but becomes identity by choosing productivity parameters How used: Test whether πk i > 0 Compare πk i to other measures like wages

13 Further results Trefler (1998) Davis and Weinstein (2001) taking other than US technology matrix into account by estimating a c ij = α c + β jk + γ k ( K c L c ) + ε c ij Using estimated coefficients ˆα c, ˆβ jk and ˆγ k they construct matrices  c and calculate factor content of exports as ˆF cr =  c X cr Factor content of trade when intermediates are traded (Zhu and Trefler, 2010; Deardorff, 1982) Global input-output matrix FCT r = f k Ltr

14 Analysis based on Trefler and Zhu (2010) Measuring FCT when intermediates are traded Predicted FCT: Vf r s r V f Measured FCT: d f Ltr (following Trefler and Zhu (2010) and Deardorff (1982)) But requires stronger assumption on consumption patterns Results (based on WIOD data) comparing predicted versus measured FCT: Sign test: 90% and more Rank correlation: ρ = 0.9 Regression based tests: FCT measured,r = α + βfct predicted,r + ε r reveal correct signs but lot of missing trade (i.e. predicted FCT is larger than measured FCT)

15 HOV test for employment following Trefler and Zhu (2010) 50 _1, EMP 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 Source: Foster and Stehrer, Predicted FCT

16 HOV test for employment following Trefler and Zhu (2010) but allowing for home-bias in consumption _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 -30 Source: Foster and Stehrer, Predicted FCT

17 Related topic: More and more production stages are offshored (i.e. activities abroad) Intermediates trade FDI Contracts Policy debate Lost jobs? Increasing competitiveness (cost efficiency)

18 Example: Production of ipod Linden and Kraemer, 2010 Overview

19 Example: Car Baldwin, 2011 Overview

20 Many other case studies... T-shirts Barbie dolls NOKIA N95 Salmon... Various approaches... Various approaches (Global supply chains, offshoring/outsourcing, fragmentation of production, slicing up the value chain, trade-in-tasks,...)

21 Measures of international fragmentation of production Narrow and broad offshoring measure (Feenstra and Hanson) Vertical specialization (Hummels et al., 2001) Foreign and domestic value added content of exports Value added exports (VAX; Johnson and Noguera, 2012) Decomposition of value added exports (Koopman et al., 2010; Stehrer, 2012b) Trade balances in VA terms (Stehrer, 2012a) Factor content of trade (Trefler and Zhu, 2010) Global value chain approach (Timmer et al., 2012) International upstreamness/downstreamness and Distance to consumer (Fally, 2012)

22 Multi-national models (3 countries, 2 industries) z11 11 z12 11 z11 12 z12 12 z11 13 z z Z 11 Z 12 Z z22 11 z21 12 z22 12 z21 13 z22 13 Z = Z 21 Z 22 Z 23 z = z12 21 z11 22 z12 22 z11 23 z Z 31 Z 32 Z 33 z z22 21 z21 22 z22 22 z21 23 z22 23 z11 31 z12 31 z11 32 z12 32 z11 33 z12 33 z21 31 z22 31 z21 32 z22 32 z21 33 z22 33 and corresponding for B = Zˆx 1 and L = (I B) 1 Value added (factor income) created in each country per unit of gross output Vector of (bilateral) exports or final demand structures

23 Example: Offshoring measures Austria - Transport Equipment (NACE 34t35) Share of intermediate inputs sourced from abroad Narrow in % of GO Narrow in % of II Broad in % of GO Broad in % of II

24 Example: Vertical specialization Austria Overview

25 Example: Foreign value added content in exports Austria

26 Example: VAX ratio compared to domestic VA content of exports Austria, in % VAXR DVAiX