Product Quality and the Human Capital Content of Swedish Trade in the 1990s

Transcription

1 Produt Quality and the Human Capital Content of Swedish Trade in the 1990s By Lars M Widell * Dep. of Eonomis (ESI) Örebro University Abstrat In an earlier study, Widell (2005), we alulated the average human apital ontent of Swedish trade in exports relative to imports in manufaturing setors and found that the human apital ontent were higher in imports over the period However, one of the shortomings when alulating the fator ontent of a ountry s net trade is the use of a ommon tehnology matrix for all ountries, ompared to using eah ountry s own fator input requirements. This paper propose a method to irumvent this problem by ombining a strutural measure developed by Lundberg & Wiker (1997), whih is based on the Heksher-Ohlin-Vanek equation and used in Widell (2005), with the onept of vertial intraindustry trade. The proposed method makes a quality adjustment in the Lundberg & Wiker measure, refleting the problem of olleting data on eah of a ountry s trading partner s fator input requirements, when alulating the fator ontent in imports. The method is then used to (re)alulate the average human apital ontent of Swedish trade in exports relative to imports in 1997 and 2000 and the new results reveal that the average human apital ontent of Swedish exports were higher relative to imports for those years, ontraditing the previous empirial results. Keywords: Fator ontent of trade, intra-industry trade, produt quality JEL lassifiation odes: F11, F19 Preliminary version prepared for the ETSG onferene in Dublin, September Date: * SE , Sweden, Telephone: ; Lars.Widell@esi.oru.se. This paper has benefited greatly from disussions and omments from Joakim Gullstrand, Lars Lundberg and seminar partiipants at Örebro University. 1

2 1. Introdution Widell (2005) reently found that the average human apital ontent in Swedish imports was higher relative to exports in manufaturing setors during the period In the study he used a strutural measure derived from the Heksher-Ohlin-Vanek (HOV) equation and he inter alia alulated the human apital ontent of Swedish net trade and evaluated different measurement issues evolving from those alulations. The results showed that the alulations were dependent on measurement pratie in the evaluation of a single year, but less dependent in the evaluations over time. The result that the average human apital ontent in Swedish imports was higher relative to exports were not in aordane to our a priori expetations due to the fat that Sweden is highly endowed with skilled labor. 1 One explanation of the outome may be that Swedish fator input requirements were used in alulating the skill ontent in both exports and imports. 2 The use of a ommon tehnology matrix follows both from the theory behind the Heksher-Ohlin-Vanek model and from the onventional fator ontent of trade literature. However, there is some evidene that the adoption of a ommon tehnology matrix for all ountries will lead to attenuation bias. 3 One reason may be that a ommon tehnology matrix exludes the possibility of fator ontent in intraindustry trade. It would, however, be better to use the fator input requirements of eah of Sweden s trading partners when alulating the skill ontent of imports, sine if fator prie equalization fails, fator ontents should be measured by the prodution tehnology of the exporting ountry. However, data on fator input requirements for all, or even only the most important trading partners, are hard to obtain. Other possible explanations for the results in Widell (2005) are i) that the servie setors are exluded from the alulations, due to lak of redible data and ii) the exlusion of the publi setor from the alulation, whih is very skill-intensive and relatively big in Sweden in omparison with other ountries, the supply of skilled workers available to the manufaturing setors beome relatively smaller ompared to other ountries. 1 Partiularly sine Widell (2005) measures skilled labor as labor with at least a post-seondary eduational attainment level. 2 See equation 3.7 below or equation 11 in Widell (2005). 3 See Trefler (1993 and 1995), hapter 3 in Wood (1994), Davis & Weinstein (2000) and Widell (2005). 2

3 Another possible way to advane in the alulations, taken in this paper, is to use the idea of vertial intra-industry trade (VIIT). With perfet information, a speifi produt variety sold at a higher prie must be of higher quality than the variety sold at a lower prie. 4 Aording to Sutton (1986), given two varieties of the same good offered at the same prie, the one good with the highest quality will be preferred by all onsumers. Hene, if produts that are both exported and imported are vertially differentiated, it is reasonable to think that the differene in produt quality is assoiated with differenes in skill ontent, i.e. high quality goods requires higher ontent of skilled labor than low quality goods. 5 The remainder of this paper is strutured as follows. Setion 2 surveys some of the related literature, with speial emphasis on Sweden and on the onnetion between human apital and produt quality. Setion 3, whih are divided into three subsetions, address theoretial, methodologial and data questions. Subsetion 3.1 outline the theoretial bakgrounds on whih the proposed model (method) is based, whih is further desribed in subsetion 3.2 and in subsetion 3.3 a brief explanation of the data used in this study is provided. Setion 4 is divided into three subsetions, where in 4.1 the empirial results are shown, disussed and ompared with those in Widell (2005); in 4.2 a sensitivity analysis of the results is performed and disussed and in 4.3 some poliy impliations are provided. Finally, onluding remarks are provided in the losing setion Related literature In a survey artile by Leamer (1992) he noted that there are only two empirial findings that have altered the way eonomists think of the auses of international trade. The first is the Leontief paradox, whih gave birth to the extensive fator ontent of trade literature; the seond is the great degree of intra-industry trade atalogued by Grubel & Lloyd (1975). Sine 1975 there have been several other studies onfirming the big share of intra-industry trade, e.g. Abd-el-Rahman (1991) on Frenh data, Crespo & Fontoura (2001) on Portuguese data, Greenaway et al. (1994) on UK data and Lundberg (1982) and Greenaway & Torstensson (1997) on Swedish data. 4 Stiglitz (1987) gives several examples also in ases with imperfet ompetition. 5 Celi (1999) also uses the assumption that VIIT is driven by differenes in skill ontent. Greenaway et al. (1994) and (1995) impliitly assumes the same, sine they use the ratio of manual workers to total employment of eah industry as a proxy variable for quality differentiation. 3

4 Until the beginning of the 1990s, most of the literature on intra-industry trade has thought of produt differentiation as a horizontal phenomenon, meaning that different varieties of a speifi produt are of a similar quality. Reently, there has been a distintion between horizontal intraindustry trade (HIIT) and VIIT that has grown in importane, where VIIT arises when different varieties of the same produt are of different qualities. 6 Sine international speialization does not only emerge in ountries speializing in produing produts in different industries, but of different varieties of produts within the same industry both vertially and horizontally, we have to separate the different types of international trade. One reason for doing this is that the results from a trade expansion have different impliations on fator markets, sine the determinants behind VIIT and HIIT differ (see e.g. Aturupane et al. (1999) and Torstensson (1992, 1996a)). Empirially Greenaway et al. (1995) and Greenaway et al. (1999) show that VIIT is markedly more important than HIIT for the UK. Fontagné & Freudenberg (1997) find a similar result for the EU. Hansson (1994) finds that VIIT in Swedish manufaturing in 1983 is an important determinant of intra-industry trade and that it is most ommon in trade with other developed ountries. One problem that arises here is how to disentangle HIIT and VIIT from the trade data. Two of the more prominent studies in the field are Greenaway et al. (1994) and Fontagné & Freudenberg (1997). Both studies use unit value ratios together with a dispersion fator to find the two forms of intraindustry trade, but they do it in slightly different ways. They both alulates unit value ratios for a produt and Greenaway et al (1994), on the one hand, heks if the alulated value lies within an interval [(1-α), (1+α)], whih indiates HIIT, or not, whih indiates VIIT. Fontagné & Freudenberg (1997), on the other hand, use the interval [1/(1-α), (1+α)], with a similar interpretation. In a reent study, Azhar & Elliot (2004) use a different method based on the share of HIIT or VIIT in total intra-industry trade. They use an index that has symmetrial limits and is equally distributed between both lower and upper bounds. In this study we will, however, use a logarithmized version of the Greenaway et al. (1994) dispersion fator. 7 6 The distintion between HIIT and VIIT draws here on Greenaway et al. (1994). Lloyd & Grubel (2003), however, note that there was a different distintion earlier in the literature, dated bak to the mid 1970s, where HIIT meant the exhange of ompeting or substitute produts and where VIIT meant the exhange of produts at different stages in the prodution proess of the produts. 7 This is further disussed in setion

5 There is not, to our knowledge, a straightforward theoretial link between produt quality, prie and human apital. 8 Greenaway & Milner (1986) has suggested, though, that it is human apital and not physial apital that is, to a great extent, positively related to quality. Torstensson (1996b) draws on this suggestion and investigates whether physial or human apital is most important in influening the quality of vertially differentiated produts. He finds from a study of Swedish imports of vertially differentiated produts at SITC 3-digit level in the 1980s, that human apital is the main determinant behind the quality of prodution. Redding (1996) argues that a ountry s sarity in eduation and other types of training is highly related to firms produt quality investments. Hansson (1994), however, studies manufaturing setors in 1983 and finds that the relationship between quality (measured by prie per tonne) and apital intensity (physial and/or human) has a weak empirial support. Greenaway & Torstensson (1997) finds that abundant human apital endowment inreases the quality of the OECD ountries manufaturing exports, using Swedish data from 1969, 1981 and They also find that human apital beomes a more important determinant of quality over time. 3. Theory, methodology and data 3.1 Theoretial bases of the proposed model The standard multifator, multiommodity, and multiountry setting model for prediting fator servies trade is the HOV model. 9 The model show that ountries, if trade is balaned, who have an abundant relative endowment of a fator will have an embodied net export of that fator and a net import of the fator in whih they have a sare relative endowment. 10 Intra-industry trade, i.e. simultaneous exports and imports of the same statistial produt group, sometimes denoted two-way trade, is an area of researh where publiations began in the 1960s. The term intra-industry was first used by Balassa (1966) and has now beome one of the tools in eah trade eonomist s toolbox. In 1971 Grubel & Lloyd presented a measure of intra-industry trade, later known as the Grubel-Lloyd index, whih have been highly used ever sine. The evolution sine the 1970s ha 8 Stokey (1990), however, develops a growth model with heterogeneous labor, differentiated by human apital, as the only fator of prodution and where she assumes that higher-skilled labor produes higherquality produts. 9 See Vanek (1968) and Leamer (1980). 10 The HOV-model is further desribed in Widell (2005). 5

6 been haraterized by Greenaway & Milner (1986, p. 2) as; a phenomenon in searh of a theory. One ommonly used link between the Heksher-Ohlin trade model and intra-industry trade was presented by Falvey (1981) and further developed by Falvey & Kierzkowski (1987). 11 Their model, ommonly named the neo-heksher-ohlin model of intra-industry trade, onsists of two ountries and two fators, where one fator is industry speifi. This fator an freely move between firms within a given setor but not between different setors. They also assume that at least one setor produes vertially differentiated goods using labor and the industry speifi fator as inputs. 12 Falvey (1981) defines a setor by the range of produts the industry speifi apital equipments an produe. The tehnology exhibits onstant returns to sale and markets are perfetly ompetitive. 13 The outome of the model shows that relatively apital-abundant ountries speialize in and export high quality produts and the relatively laborabundant ountries speialize in and export low-quality produts. 14 Balboni (2004), however, builds a model that explains trade in produts of different quality that have idential fator intensity at any relative prie. This is onsistent with the empirial findings that VIIT is observed even at a muh disaggregate level of statistial lassifiation. Two other ways to measure quality has been proposed in the literature by Cooper et al. (1993) (hedoni priing) and Brenton & Winters (1992) (prie elastiities). A problem with both those methods are that they are too data intensive and more appropriate to use in analyses of individual produt markets ompared to the multiprodut analysis arried out here. 3.2 The Method The most popular way to separate different quality levels in produt datasets is to alulate a unit value by dividing the monetary value of trade 11 Further studies have used their model e.g. Celi (1999), Ferragina & Pastore (2004) & Torstensson (1996b). 12 The authors assume that it is apital that is industry speifi in the model. Further desriptions of the neo-heksher-ohlin model an be found in hapter 2 in Greenaway and Milner (1986). 13 Some ompeting models, e.g. Brander (1981), Krugman (1981) and Helpman & Krugman (1985), rely on imperfet ompetition together with horizontal produt differentiation and eonomies of sale when explaining intra-industry trade. Davis (1995), however, provides a model based on omparative advantage that does not involve inreasing returns and Flam & Helpman (1987) provides a model that fous on trade between North and South with a fous on Riardian supply effets. 14 The model assumes that the prodution of higher-quality produts requires a higher quantity of apital per unit of labor. 6

7 by the quantity, whih gives a prie per tonne. 15 The rationale for using suh unit values is that a produt of a higher quality should demand a higher prie, whih means that it an be onsidered as an, although not perfet, indiator of quality. 16 When measuring the human apital ontent of net trade in Sweden, it is reasonable to suppose that differenes in produt quality are assoiated with differenes in skill ontent. This implies that a high quality produt should inorporate a high ontent of skilled labor. The method proposed in this study builds on two assumptions that are not tested here. The first assumption is that the prie, measured by unit-value ratios, of a produt p in a given setor i is proportional to the input requirements of skilled labor in that setor, i.e. ai, p, f = b UVRi, p, where b is a onstant and subindex f indiate a fator or fators; seond, that the onstant b is idential among all setors. The alulation of VIIT adjusted average human-apital ontent of trade in Swedish exports relative to imports follows a step-by-step proedure aording to the following. First, we alulate the unit-value ratio between Swedish exports and imports for eah single bilateral trade flow, using data from COMTRADE at SITC rev. 3 (4/5-digit). 17 Following Greenaway et al. (1995) we distinguish between horizontal- and vertial IIT, using, UVR i, p, UV UV x i, p, ( 1 α ) ( 1+ α ), (3.1) m i, p, where i denote industry, p produt, ountry and α dispersion fator. This dispersion fator is in various studies set to 0.15 or 0.25; and we will use both values in this study as a omparison. 18 If the UVR i, p, is less then (1-α), the lower bound, it is thought of as low quality VIIT and if the UVRi,p, is greater than (1+α), the upper bound, it is thought of as high quality VIIT. All other ratios between (1-α) and (1+α) are defined as HIIT. The UVR i, p, will also be logarithmized in order to ensure symmetry between the lower and upper bounds in terms of their relative distane from unity when 15 Torstensson (1991), on the ontrary, uses unit values per item as an alternative to unit values per tonne. 16 See Stiglitz (1987). However, Lüthje & Nielsen (2002) disuss that unit-value ratios are poor indiators of quality, but we haven t found a better alternative measure of quality, so we will stik to using unit values. 17 Here we also follow Aturupane et al. (1999), Crespo & Fontoura (2001) and Gullstrand (2002) and let small trade volumes be removed from the sample together with trade flows without any information on quantity. The limit value we use is 20,000 USD. 18 We will also use α=0 to see how sensitive the results are from the use of different dispersion fators. 7

8 distinguishing between VIIT and HIIT. 19 following equation, This exerise gives us the x UV i, p, log UVRi, p, log( 1 α ) log log( 1 α m + ), (3.2) UV i, p, whih is the equation used for distinguishing between VIIT and HIIT. The unit value ratios that are revealed to be of the horizontal ategory will be set to 1 while those that are revealed to be vertial will keep the alulated values from equation (3.1). 20 Seondly, we aggregate the unit value ratios for eah produt separately using trade weights, i.e. by SITC rev. 3 produts, aording to, ( Xi, p, + Mi, p, ) ProdUVR i, p= UVR i, p, * ( Xi, p, + Mi, p, ) UVR Ω p p, (3.3) where Ω p is the set of unit value ratios (UVR i,p, ) that belong to a speifi produt p. Thirdly, we aggregate the produts that belong to eah industry group aording to the 4-digit level of ISIC rev.3. The alloation of produts to industries was based on a onordane table between SITC rev. 3 (4/5 digit) and ISIC rev. 3 (4-digit) attained from the Eurostat lassifiation server Ramon. 21 This gives us the following equation, IndUVR = ProdUVR * ( Xi, p+ Mi, p) ( X, + M, ) i i, p p i p i p p Ω i i, (3.4) where Ω i is the set of produts that belongs to a speifi industry i. This gives a set of industry weights that will, in the next step, either sale up or sale down the skill intensity in the separate industries respetively. These 19 See Fontagné & Freudenberg (1997) and the disussion about the proportionality effet in Azhar & Elliot (2004) for a more thorough treatment of this issue. 20 The separation between HIIT and VIIT is made by equation (3.2), but equation (3.1) is used (after setting the horizontal ategory to 1) hereafter, sine unit value ratios less than one will have a negative value in its logged form. 21 See 8

9 industry weights are used as a weighting sheme in alulating the human apital ontent of trade in Sweden in 1997 and The equation we will use to measure the human apital ontent of trade is derived from the Heksher-Ohlin-Vanek equation, w F AT = V s V, (3.5) where F is fator ontent of trade in ountry, A is a vetor or matrix of fator input requirements, T is a vetor of net trade in ountry, V is the endowment vetor of a speifi fator/fators in ountry, s is a vetor of ountry s share of world onsumption, and V w is the world endowment vetor of a speifi fator/fators. We an write the net trade in embodied servies of prodution fators f for ountry, i.e. an element in AT, and summing over industries i, as I I I I I I I f = i if = i if i if = i i if i i if i= 1 i= 1 i= 1 i= 1 i= 1 i= 1 i= 1, (3.6) F Ta Xa Ma X xa M ma where T i, is net exports for industry i in ountry, X i and M i are exports (imports) from (to) industry i in ountry, x i and m i the share of the i:th industry in the total exports (imports) from (to) ountry, and aif the use of fator f per unit of prodution from the i:th industry. The omparison of the fator ontent of trade an be made in differene form, as in equation (3.6) above, or as a ratio, 23 z ft I xita i= 1 = I ma i= 1 it ift ift. (3.7) If a superior (high quality) VIIT is revealed within an industry in equation (3.4), the skilled fator input requirements in imports, i.e. the denominator in equation (3.7), will be weighted down by the IndUVR i -number and the opposite for inferior (low quality) VIIT. 24 These adjustments give us the following equation, 22 The hoie of years is further disussed in setion This is a strutural measure developed by Lundberg & Wiker (1997). 24 The IndUVR s are not used to atually indiate the skill-intensity of a partiular produt or industry. Instead they are used to sale relative fator intensity among partner ountries given a partiular fator intensity that is industry speifi. 9

10 z ft = I i= 1 I xa i= 1 it ift ( ma it ift IndUVRi ). (3.8) Thes z-measure has a simple interpretation as the average requirements of a fator f, weighted by trade shares, per unit of exhange of exports, in thousands of Swedish kronor, ompared to the average requirements of the imports, adjusted for quality differenes. This gives us information about the differene in export and import struture with respet to a partiular fator's intensity in produts and servies, regardless of the trade balane. 25 This is the final equation in the step-by-step proedure. Finally, a note on the fator input requirements, i.e. variable a ift in equation (3.8). Theoretially we should use total fator input requirements in the alulations of the fator ontent of trade (see e.g. Hamilton & Svensson (1983) or Deardorff (1984)). However, sine Statistis Sweden only has published symmetrial input-output tables on a highly aggregated level 26, we will use the diret fator input requirements when alulating the z- value in equation (3.8). Widell (2005) have shown that, using equation (3.7), both the hoie of total- over diret fator input requirements on a 2- digit level (ISIC rev. 3) and the hoie of aggregation level of the data do matter in a ross-setional approah (as is done here). Feenstra & Hanson (2000) also show that aggregating different industries together gives a substantial aggregation bias. 3.3 Data The data used in this study are taken both from the COMTRADE database maintained by the United Nations Statistis Division and from various databases maintained by Statistis Sweden. The COMTRADE-data onsists of Swedish ommodity trade, both imports and exports, with all its partner ountries, i.e. on a multilateral basis, aording to SITC rev.3 at 4/5-digit level for the years 1997 and The data are reported both in values (in USD) and in quantity units. 27 The data from Statistis Sweden were olleted from the databases register based labor market statistis (RAMS), industrial statistis/finanial statistis (IS/FS) and the foreign trade statistis. From RAMS we have olleted both total wages and wages by eduational groups at 4-digit ISIC rev.3; and from IS/FS we have 25 See Lundberg & Wiker (1997) for further details. 26 Statistis Sweden has published IO-tables for 1995 and 2000 on a 2-digit level ISIC rev The COMTRADE data will only be used in alulating the IndUVR s. 10

11 olleted total wages and gross prodution at the same 4-digit ISIC rev. 3 level. Wage taken from RAMS is annual earnings and wage taken from IS/FS is labor osts inlusive of soial seurity osts. Exports and imports at the same ISIC level were olleted from the foreign trade statistis. The years available are Further explanation of the data from Statistis Sweden an be found in Widell (2005). The onversion key between SITC rev.3 and ISIC rev.3 were available from the Eurostat server RAMON. Here we limit ourselves to investigate manufaturing industries only, i.e. setors (ISIC rev.3), due to data availability. The hoie of years in the study is restrited by the available data. In this study we alulate the human apital ontent of net trade in 1997, with all data originating from 1997, and in 2000, with produt trade data from 2001 and all other data from In 1997 the COMTRADE data ontains around observations on import values and around observations on export values. In the alulations of unit value ratios, equation (3.1), we use approximately bilateral observations. After alulating equation (3.3) we end up with around observations (produts) and when alulating equation (3.4) we end up with around 110 (industry setors). In the dataset from Statistis Sweden we have observations representing around 200 industry setors. When mathing those datasets we get a perfet math for around 25 per ent of the industry setors representing around 50 per ent of total trade value for all observations within the manufaturing setors. 4. Results and analysis When presenting the empirial results in this setion we alulate equation (3.8) using one dataset inluding all of Sweden s trade partners, the olumn named z-value (all ountries), and one dataset using only Sweden s trade with the OECD ountries, the olumn named z-value (OECD). 28 We also inlude alulations using different values of the dispersion fator α, olumn Dispersion fator, evaluating the importane of this fator in this type of studies. 28 The OECD dataset is a subset of the all ountries dataset. 11

12 4.1 Empirial results The results from alulating the average human apital 29 ontent of exports relative to imports in Swedish manufaturing industries in 1997, using equation (3.8), are reported in table 4.1 below. [Table 4.1 about here] The first row show the result from the alulation performed in Widell (2005) for 1997, 30 the seond to fourth rows show the alulations of equation (3.8) using different dispersion fators respetively. Both the alulations using data from all ountries and those using data from only OECD ountries, give an upward adjustment of the z-value ompared to the results in Widell (2005). The z-value of the former is slightly higher than that of the latter, whih is in aordane to our expetations. Due to the fat that the dataset ontaining OECD ountries only onsists of highly developed ountries (North), we would expet the skill ontent in produts produed in those ountries to be of a similar quality to those produts produed in Sweden, whih in turn will give a small adjustment of the z- value in either diretion, depending on the skill ontent in the traded produts. Here we notie that the z-value in the OECD-alulations is adjusted upwards mirroring that the skill ontent in Swedish exports is higher then the OECD average. In the dataset ontaining all ountries, both developed/high skilled ountries (North) and less developed/low skilled ountries (South) are lumped together. Sine South has a relatively large supply of unskilled labor ompared to North that has a relatively large supply of skilled labor, we would expet the quality of a produt produed in South to be lower than the same statistial produt produed in North, resulting in a bigger effet on the z-value ompared to that of the OECD ountries. 31 We would also expet this z-value to be higher than both that of the OECD ountries and that in Widell (2005). When omparing the different dispersion fators, it doesn t seem to be important whih fator to use in the alulations of the z-value. In the ase of using data from the whole dataset we find that the dispersion fator has 29 A broad definition of human apital inludes all harateristis of the labor fore that gives in return a higher produtivity of the worker, e.g. eduation, on-the-job and of-the-job training. We use eduational attainment as our measure of human apital, sine the other two are hard to measure. 30 See figure 4.3 in Widell (2005), whih is based on the 4-digit level of ISIC rev See Wood (1994) for a similar argument about quality differenes between North and South. 12

13 an insignifiant effet on the resulting z-value, and in the ase of using data from only the OECD ountries, we find that the dispersion fator have a very small effet on the z-value. This low, or no, impat of the dispersion fator on the alulations may be a onsequene of the weighted aggregations performed in equations 3.3 and 3.4. Table 4.2 below reports the same alulations as table 4.1 using trade data from 2001 and fator input requirements data from [Table 4.2 about here] The results from the alulations are similar to those in 1997, both when using the dataset overing all ountries and in the use of the dataset overing the OECD ountries. We see an upward adjustment of the z- values that are slightly higher in magnitude in 2001 than in The hoie of dispersion fator seems to be unimportant even in this year. Comparing the results from tables 4.1 and 4.2 with those in Widell (2005) we find that the z-values in 1997 and 2000 are lower than those reported here. Sine the z-values in 1997 and 2000 in Widell (2005) also are below 1 and that the z-values alulated in this study for those years are above 1, the interpretation of the z-value in this study is the opposite. In this study we find that the average human apital ontent in Swedish exports is higher ompared to the imports both in 1997 and in Those results follow our a priori expetations stated in the introdution. The results drawn from tables 4.1 and 4.2 above also orroborates with those of Torstensson (1991), who found that the quality of Swedish imports were generally lower than that of Swedish exports in the mid 1980s. 4.2 Sensitivity analysis A possible link between produt quality and human apital ontent of trade on one hand and the traditional fator ontent of trade literature on the other hand is tehnologial differenes. In the traditional fator ontent of trade literature, the poor fit of the Heksher-Ohlin trade model have by many authors been asribed the assumption of equal tehnologies aross ountries. 32 Trefler (1993) and (1995) develops extended versions of the HOV-model allowing for tehnologial differenes between ountries. In the 1993 study he allows all fators in every ountry to differ in their 32 See Widell (2005) and referenes therein. 13

14 produtivities, while in the study from 1995 he allows the fator input requirements matrix to differ aross ountries. The differene between the two ways of allowing for tehnologial differenes is that in the first ase, Trefler assumes that a fator in a ountry is, for example, 5 perent more produtive than in another ountry, while in the seond ase, he assumes that a ountry needs 5 perent more of a prodution fator to produe the same amount of a produt than another ountry. It is the latter ase that form the basis of the model used in the sensitivity analysis. In order to test how sensitive the results are from using unit value ratios as a measure of produt quality, we have realulated equation (3.8), without using quality weights, but instead using a Trefler (1995) inspired adjustment of the fator input requirements. This is alulated as follows, z ft I VH a ift xa it ift tot i= 1 a ift =, I VL a ift ma it ift tot i= 1 a ift (4.1) VH where a ift is the average fator input requirements of those setors that are revealed to produe high quality produts (VIIT high), a VL ift is the average fator input requirements of those setors that are revealed to produe low quality produts (VIIT low); and a is the average fator input requirements for all setors. The motivation for using average values is that a high quality good are assumed to be produed with higher skill ontent and vie versa with a low quality good. tot ift [Table 4.3 about here] As is seen in table 4.3 above, there is a large differene in the results when using average fator input requirements as indiator of produt quality ompared to unit value ratios. The results are also sensitive to whih year the separation between high and low quality trade takes plae. When using produt trade data from 1997 we end up with a value below both the z_eg value and the z_eg (VIIT-orr) value, but when using produt trade data from 2001 we end up above the two. 14

15 4.3 Poliy impliation This study has thus far followed a positive approah. However, in this subsetion we use the attained results and make some poliy impliations. If we think of vertial produt differentiation as an intra-industry produt yle, a ountry with an abundant relative endowment in high skilled labor will produe and export produts that intensively use this fator. As the quality of the produt dereases, i.e. the more standardized the produt gets; the ountry will abandon the prodution of this partiular produt quality and introdue higher qualities or new produts. The abandoned qualities will instead be produed in ountries with an abundant relative endowment of low skilled labor. 33 This hange in prodution fators has an effet on the omparative advantages of a ountry. The alulation of adjusted z-values is, in the light of this, an important indiator of what type of produts (quality) that are produed in a ountry relative to its trading partners. This, in turn, indiates that eduational poliy is needed to strengthen a ountry s omparative advantage. Thus, this is important sine ountries endowments of human apital will be inreasingly important in the determination of industrial loation, international speialization and trade. The results presented in this study show that the average quality of Swedish imports is generally of a lower quality than the quality of Swedish exports suggesting that an eduational poliy aiming at inreasing the skill intensity of the Swedish labor fore is preferable. The results from this study an also be interpreted as suggesting that the government should enourage those employees losing their jobs in low skilled setors (or ompanies) either to searh for work in high skilled setors or to enhane their human apital through eduation. 5. Conluding remarks In this paper we remodeled/adjusted a strutural measure developed by Lundberg & Wiker (1997), whih builds on the Heksher-Ohlin-Vanek equation, by using the onept of VIIT. Due to the fat that a great part of a ountry s exports and imports takes plae within the same statistial produt groups, and that those produts are often vertially differentiated, 33 This senario an also be explained by the Flam & Helpman (1987) North-South model with quality upgrading. 15

16 the strutural measure, equation 3.7 above, needs to be adjusted in order to take this fat into aount. This is done by adding an extra term in the denominator (see equation 3.8) measuring the unit value ratio between exports and imports for eah industry. If the adjusted z-equation (equation 3.8) reveals the true average human apital ontent of Sweden s export relative to imports, there is higher skill ontent in Swedish exports ompared to imports in both 1997 and This result alters that of our earlier study (Widell 2005) using the standard z-equation (equation 3.7), whih revealed that Swedish imports had a higher skill ontent, ompared to exports for the whole time period However, we want to stress that these alulations hinges on the idea that a speifi produt variety sold at a higher prie must be of higher quality than the variety sold at a lower prie. The empirial results in this study also highlight the importane of an eduational poliy aimed at strengthen a ountry s omparative advantage. Sine Sweden is a high skilled ountry with a positive net export, an inrease in the skill level of the Swedish labor fore through human apital formation (eduation et.) is important for future welfare. 16

17 Referenes Abd-el-Rahman, K, [1991], Firms Competitive and National Comparative Advantages as Joint Determinants of Trade Composition, Weltwirtshaftlihes Arhiv, 127, pp Aturupane, C, Djankov, S, & Hoekman, B, [1999], Horizontal and Vertial Intra-Industry Trade between Eastern Europe and the European Union, Review of World Eonomis, 135, No. 1, pp Azhar, A & Elliot, R, [2004], On the Measurement of Vertial and Horizontal Intra-Industry Trade: A Geometri Exposition, Graduate Shool of Management Researh Paper Series, GRPS 03/04. (Paper presented at the European Trade Study Group 6 th Annual Conferene, ETSG 2004, 9-11 September, Nottingham, England). ( Balassa, B, [1966], Tariff Redutions and Trade in Manufatures Among the Industrial Countries, Amerian Eonomi Review, 56, No.3, pp Balboni, A, [2004], A General Equilibrium Model of Vertial Intra- Industry Trade and Inter-Industry Trade with Complete Speialization of Countries, mimeo, University of Paris Dauphine. (Paper presented at the European Trade Study Group 6 th Annual Conferene, ETSG 2004, 9-11 September, Nottingham, England). ( Brander, J, [1981], Intra-Industry Trade in Idential Commodities, Journal of International Eonomis, 11, pp Brenton, P & Winters, A, [1992], Estimating the Trade Effets of 1992, Journal of Common Market Studies, 30, pp Celi, G, [1999], Vertial and Horizontal Intra-industry Trade: What is the Empirial Evidene for the UK?, Disussion Paper 49, Università degli Studi di Salerno, Italy. Cooper, D, Greenaway, D & Raynor, A, [1993], Intra-Industry Trade and Limited Produer Horizons: An Empirial Investigation, Review of World Eonomis, 129, pp Crespo, N & Fontoura, M, [2001], Determinants of the Pattern of Horizontal and Vertial intra-industry Trade: What Can We Learn From Portuguese Data?, mimeo, ISEG, Tehnial University of Lisbon, Portugal. Davis, D, [1995], Intra-Industry Trade: A Heksher-Ohlin-Riardo Approah, Journal of International Eonomis, 39, pp Davis, D & Weinstein, D, [2000], International Trade as an Integrated Equilibrium : New Perspetives, Amerian Eonomi Review: Papers and Proeedings, 90, No. 2, pp

18 Deardorff, A, [1984], Testing Trade Theories and Prediting Trade Flows, h. 10 in Jones, R & Kenen, P (ed.), Handbook of International Eonomis, Vol. I, Elsevier Siene Publishers B.V., Netherlands. Falvey, R, [1981], Commerial Poliy and Intra-Industry Trade, Journal of International Eonomis, 11, pp Falvey, R & Kierzkowski, H, [1987], Produt Quality, Intra-Industry Trade and (Im)perfet Competition, hapter 11 in Kierzkowski, H (ed.), Protetion and Competition in International Trade: Essays in Honor of W. M. Cordon, Blakwell Publishers, Oxford, UK. Feenstra, R & Hansen, G, [2000], Aggregation Bias in the Fator Content of Trade: Evidene from U.S. Manufaturing, Amerian Eonomi Review: Papers and Proeedings, 90, No. 2, pp Ferragina, A & Pastore, F, [2004], Fator Endowment and Market Size in EU-CEE Trade. Would Human Capital Change the Atual Quality Trade Patterns, Disussion Paper No. 1139, Institute for the Study of Labor (IZA), Germany. Flam, H & Helpman, E, [1987], Vertial Produt Differentiation and North-South Trade, Amerian Eonomi Review, 77, No. 5, pp Fontagné, L & Freudenberg, M, [1997], Intra-Industry Trade: Methodologial Issues Reonsidered, CEPII Working Paper, No. 97/02. Fukao, K, Ishido, H & Ito, K, [2003], Vertial Intra-Industry Trade and Foreign Diret Investment in East Asia, Journal of the Japanese and International Eonomies, 17, pp Greenaway, D, Hine, R & Milner, C, [1994], Country-Speifi Fators and the Pattern of Horizontal and Vertial Intra-Industry Trade in the UK, Review of World Eonomis, 130, pp Greenaway, D, Hine, R & Milner, C, [1995], Vertial and Horizontal Intra-Industry Trade: A Cross Industry Analysis for the United Kingdom, Eonomi Journal, 105, pp Greenaway, D & Milner, C, [1986], The Eonomis of Intra-Industry Trade, Blakwell, Oxford, UK. Greenaway, D, Milner, C & Elliot, R, [1999], UK Intra-Industry Trade with EU North and South, Oxford Bulletin of Eonomis and Statistis, 61, No. 3, pp Greenaway, D & Torstensson, J, [1997], Eonomi Geography, Comparative Advantage and Trade Within Industries: Evidene from the OECD, The Trade Union Institute for Eonomi Researh (FIEF) Working Paper, No. 144, Stokholm, Sweden. 18

19 Grubel, H & Lloyd, H, [1975], Intra-industry trade: the theory and measurement of international trade in differentiated produts, MaMillan, London, UK. Gullstrand, J, [2002], Splitting and Measuring Intra-Industry Trade, Dissertation No. 98, Lund Eonomi Studies, Lund University, Sweden. Hamilton, C & Svensson, L, [1983], Should Diret or Total Fator Intensities be Used in Tests of the Fator Proportions Hypothesis?, Weltwirtshaftlihes Arhiv, 119, pp Hansson, P, [1994], Produt Quality and Vertial Produt Differentiation as Determinants of Intra-Industry Trade in Swedish Manufaturing, FIEF Working Paper No. 122, Stokholm, Sweden. Helpman, E & Krugman, P, [1985], Market Struture and Foreign Trade: Inreasing Returns, Imperfet Competition, and the International Eonomy, MIT Press, Cambridge, Massahusetts, USA. Krugman, P, [1981], Intra-Industry Speialization and the Gains from Trade, Journal of Politial Eonomy, 89, No.5, pp Leamer, E, [1980], The Leontief Paradox, Reonsidered, Journal of Politial Eonomy, 88, No. 3, pp Leamer, E, [1992], Testing Trade Theory, NBER Working Paper, No. 3957, Cambridge, MA, USA. Lloyd, P & Grubel, H, [2003], Intra-Industry Trade, Edgar Elgar Publishing, Cheltenham, UK. Lundberg, L, [1982], Intra-Industry Trade: The Case of Sweden, Review of World Eonomis, 118, No. 2, pp Lundberg, L & Wiker, P, [1997], "Skilled Labor and International Speialization in OECD Countries", International Review of Applied Eonomis, 11, No. 3, pp Lüthje, T & Nielsen, J, [2002], Tests of the Empirial Classifiation of Horizontal and Vertial Intra-Industry trade, Eonomis Disussion Papers, Syddansk University. Redding, S, [1996], The Low-Skill, Low-Quality Trap: Strategi Complementarities Between Human Capital and R&D, Eonomi Journal, 106, No. 435, pp Stiglitz, J, [1987], The Causes and Consequenes of the Dependene of Quality on Prie, Journal of Eonomi Literature, 25, pp Stokey, N, [1990], Human Capital, Produt Quality, and Growth, Quarterly Journal of Eonomis, 106, No. 2, pp Sutton, J, [1986], Vertial Produt Differentiation: Some Basi Themes, Amerian Eonomi Review: Papers and Proeedings, 76, No. 2, pp

20 Torstensson, J, [1991], Quality Differentiation and Fator Proportions in International Trade: An Empirial Test of the Swedish Case, Review of World Eonomis, 127, pp Torstensson, J, [1992], Fator Endowments, Produt Differentiation and International Trade, Dissertation No. 47, Lund Eonomi Studies, Lund University, Sweden. Torstensson, J, [1996a], Determinants of Intra-Industry Trade: a Sensitivity Analysis, Oxford Bulletin of Eonomis and Statistis, 58, No. 3, pp Torstensson, J, [1996b], Can Fator Proportions Explain Vertial Intra- Industry Trade?, Applied Eonomis Letters, 3, pp Trefler, D, [1993], International Fator Prie Differenes: Leontief was Right!, Journal of Politial Eonomy, 101, No. 6, pp Trefler, D, [1995], The Case of the Missing Trade and other Mysteries, The Amerian Eonomi Review, 85, No. 5, pp Vanek, J, [1968], The Fator Proportions Theory: The N-Fator Case, Kyklos, 4, pp Widell, L, [2005], "On Measurements of the Fator Content of Trade: The Case of Sweden", ESI Working Paper Series, No. 2005:7, Örebro University. Wood, A, [1994], North-South Trade, Employment and Inequality: Changing Fortunes in a Skill-Driven World, Clarendon Press, Oxford, UK. 20

21 Appendix Variable definitions and onstrution + data availability Table A.1 Variable definitions and data desription Variable definition Data soure - Export (produt) - Import (produt) Trade data (exports and imports in values and quantities) from the UN statistis database COMTRADE, lassified aording to SITC rev.3 and reorded at the 4/5-digit level. Values are reorded in - Export (industry) - Import (industry) Fator input requirements (a ift in equation 3.6, 3.7 and 3.8) Variables: - Total wages by industry - Total wages by eduational group and industry Fator input requirements (a ift in equation 3.6, 3.7 and 3.8) Variables: - Total wages by industry - Gross prodution by industry US dollars. Trade data (exports and imports in values) from the foreign trade statistis at Statistis Sweden, lassified aording to ISIC rev. 3 (SNI-92) and reorded at the 4-digit level. Values are reorded in Swedish kronor. Data olleted from Statistis Sweden s database RAMS (register based labor market statistis). All data is lassified aording to ISIC rev.3 and reorded at the 4-digit level. Data olleted from Statistis Sweden s database IS/FS (industrial statistis/ finanial statistis). All data is lassified aording to ISIC rev.3 and reorded at the 4-digit level. Note! The COMTRADE database an be found at: and the various databases from Statistis Sweden an be found at: 21

22 Tables Table 4.1 The average human apital ontent of Swedish exports ompared to imports in Model Dispersion fator z-value (all ountries) z-value (OECD) Widell (2005) Eq Eq Eq Note! The all ountries z-value is based on a dataset inluding all of Sweden s trade partners; and the OECD z-value is based on a dataset inluding Sweden s trade with the OECD ountries only. All alulations are based on equation 3.8, exept the Widell (2005), whih is the alulated z-value for 1997 in that study. Table 4.2 The average human apital ontent of Swedish exports ompared to imports in Model Dispersion fator z-value (all ountries) z-value (OECD) Widell (2005) Eq Eq Eq Note! The all ountries z-value is based on a dataset inluding all of Sweden s trade partners; and the OECD z-value is based on a dataset inluding Sweden s trade with the OECD ountries only. All alulations are based on equation 3.8, exept the Widell (2005), whih is the alulated z-value for 2000 in that study. Table 4.3 Calulation of equation 4.1 Year z-value Note! All alulations are performed on 4-digit ISIC rev.3 level. 22