Exchange Rate and Inflation Differentials With Imported Intermediate Inputs.

Transcription

1 Preliminary and incomlete. Comments and suggestions most welcome. Exchange Rate and Inflation Differentials ith Imorted Intermediate Inuts. A Factor-ecific odel with International Fragmentation of Production Antonella Liberatore* Fabio dogati * Paer to be resented at the Italian Trade tudy Grou meeting Roma, November 2008 Abstract and ummary of Conclusions JEL Ref numbers: Inflation in Oen Economies International Fragmentation of Production Exchange Rates and Inflation Politecnico di ilano, Deartment of anagement, Economics and Industrial Engineering. <antonella.liberatore@it.ibm.com>, <fabio.sdogati@olimi.it>

2 Introduction Over the ast ten years the Euroean Central Bank succeeded in reaching and maintaining medium-run rice stability in the Euro Area 1. Nevertheless, though inflation rates converged (and decreased) during the stage II of EU, since 1999 inflation differentials among the member countries have been increasing (Busetti et al. 2006). To be sure, inflation differentials among countries belonging to a monetary union are a natural adjustment mechanism to local shocks; nevertheless, inflation differentials among EU countries, esecially because of their olicy imlications, have been raising a growing concern among Euroean economists. Before the Euro began circulating, the well-known Balassa-amuelson effect was a common exlanation for this henomenon: different growth rates between tradable and non-tradable good sectors roduce a real areciation of the exchange rate which has to be balanced with higher inflation rates. howing a relevant growth in roductivity, Ireland was regarded to be the best examle of the Balassa-amuelson effect. ince the introduction of the Euro, as soon as inflation differentials began to increase, a variety of emirical studies have been addressing the issue, considering cyclical reasons, the role of national olicies (fiscal olicy above all), and demand behaviour among the ossible causes of this henomenon (Rogers et al 2001; Blanchard 2001; ECB 2003). Another branch of literature considers instead the effects of globalisation on inflation rate: aearance of emerging countries on the world trade scene and growing trade integration contributed, according to some economists, to maintain a low and stable inflation rate in most industrialised countries. Nevertheless, not everybody shares such oint of view: low-cost imorts from emerging countries do not affect the inflation rate automatically, as international trade only affects relative rices, not the aggregate rice level 2. The main result of most studies, however, is that low-cost imorts could have an effect on inflation dynamics only in the short run (IF 2006). Though extremely imortant toward our understanding of the issue of what determines inflation differentials, the two strands of literature referred to above heavily discount two issues: the role of the exchange rate as an indeendent originator of inflation, and the role of structural variables in the emergence of country-secific differences in the reaction of the aggregate rice level to exchange rate shocks. Indeed, a debate focusing on the structural reasons of the henomenon, namely the structural differences among member countries and 1 This statement does not aly to the whole eriod beginning July 2007, which has been characterized as one of credit crunch or, even, credit anic. 2 For two oosite visions, see Rogoff (2005) and Ball (2006). 2

3 their different degree of exosure to extra-eu trade, is only recently emerging in the literature, and then with attention being aid to differential effects of exchange rate movements on national inflation rates (ECB, 2003; Honohan and Lane, 2003). This aer aims at analyzing the differential imacts of exchange rate changes on national inflation rates both theoretically and emirically. It differs from reviously ublished work in several resects. First, we are interested on the effects of a change in the rice of intermediate imorts, and not of the imorted finished roducts, on the aggregate rice level. This channel is imortant because intermediate inuts affect domestic roduction costs directly, but to our knowledge such channel has not been studied to the extent we believe it deserves. Consequently, we are interested in analysing the effect of a change in the rice of intermediate imorts both at country and at industry level: we exect different rice dynamics for different degrees of integration of industries (and countries) in the global value chain. econdly, we model the link between exchange rate and inflation by considering the role of internationally integrated roduction rocesses, beside oenness to trade. The imortance of this channel is self-evident: imorted intermediate inuts affect domestic roduction costs directly. e do not know of any aer addressing the issue this way. Finally, we look for the net effect of an exchange rate shock on the inter-country inflation differential, that is, for the effect of the shock on both the imort and the exort side: the imort-side effect of the shock will deend in our model uon the degree of integration of the national roduction structure on the world s, whereas the exort-side effect will deend uon the degree of exosure of the home country to the world s change in demand for domestic roducts due the exchange rate shock. The aer is organised as follows. The first aragrah lays the theoretical foundations of the model as a whole drawing on both the candinavian model of inflation for a small oen economy and the recent literature on international fragmentation of roduction rocesses. These two strands of literature constitute the basic framework for our model and allow us to identify recisely the core of our interest, namely the link between rices of intermediate roducts and rices of final ones. In the second aragrah we build a model in which the aggregate rice level of a small oen economy deends both on its degree of integration in the world roduction rocess and on its exosure to trade on the exort side: the first link is ensured by an elastic domestic demand for imorted intermediate inuts, the second by an elastic foreign demand for domestic exorts of final manufacture roducts. Emirical testing of the theoretically derived relations follows in aragrah three: we test the redictions of the model using a samle of eleven countries belonging to the Euroean Economic and onetary 3

4 Union and we use anel data models (ooled OL and Fixed Effects models) to estimate the imact of integration ad exosure on inflation rate. Emirical testing is carried out at the industry level as well. Paragrah four summarizes the main results. 1. Theoretical and Emirical Foundations Aiming at analysing the henomenon of inflation differentials both theoretically and emirically, our work needs to be founded in both theoretical and emirical literature. From the theoretical oint of view, our work has imortant affinities with the candinavian model of inflation, which relates the inflation rate of a small oen economy to different roductivity growth rates between tradable and non-tradable good sectors (Aukrust, 1977). The core feature of this model, borrowed in our framework, is the distinction between an exosed sector, which suffers international cometition, and a sheltered sector, not cometing in the international market but linked to the exosed sector by the erfect mobility of labourers. This structural, cost ush aroach assumes great imortance in exlaining inflation differentials if monetary causes are left out by definition. The other main branch of theoretical literature on which our work is founded is related to international fragmentation of roduction rocesses. Jones and anyal (1982) ublished the first main contribution in this field, where intermediate roducts are finally let enter a general equilibrium model. During the nineties various contributions analysed the international fragmentation of roduction within the traditional international trade models: the international division of labour enhances the comarative advantage rincile and amlifies the benefits of roductive secialization. (Arndt, 1996, 1997, 1998; Deardorff, 1998, 2001; Jones, Kierzkowski, 2001). Due to historical reasons, most contributions focused on the imlications of international fragmentation on factors rewards and emloyment; nevertheless, this literature allowed us to oint out the link between rices of intermediate roducts and rices of final goods 3. The animated debate about causes and imlications of the observed inflation differentials among Euro Area countries, instead, is mostly emirical. Alberola (2000) ublished the first emirical work on this issue, identifying two ossible causes of the differentials: on one hand, the ingle arket drives rice level convergence, imlying temorary differences in inflation rates; on the other hand, these differentials may be 3 From a strictly theoretical oint of view, we borrowed the basic framework of the well-known secific factor model (Jones, 1971), in turn derived from the general equilibrium framework set by Jones (1965). 4

5 consequences of country-secific shocks. In the following years a number of emirical studies followed the one ublished by Alberola, trying to exlain the ossible causes of the henomenon. The first ossible cause is linked to the role of domestic demand, i.e. the resence of asymmetric consumtion atterns (or different elasticities of substitution) in different countries. Government olicies can also affect HICP inflation through changes in administered rices and indirect taxes and thus such national measures may add to inflation disersion within the euro area. Neither domestic demand nor institutional reasons, however, seem to have had a major imact on inflation differentials (ECB, 2003). Cyclical reasons, instead, certainly contributed to exlain, at least artly, the resence of inflation differentials among member countries, esecially in the case of Ireland 4 (Rogers et al, 2001; Blanchard, 2003). The most traditional exlanation for the henomenon is the so-called Balassa- amuelson effect, which relates high inflation rates to large roductivity differentials between tradable and non tradable sector. A aer by De Grauwe and kudenly (2000) shows that this exlanation is quite convincing, even if it is worth considering the difficulty to be found when attemting to isolate the Balassa-amuelson effect from other factors 5 (ECB, 2003). The last grou of ossible causes considered in literature has an international dimension, and refers to external factors like oil rice or exchange rate: if Euro Area countries show different degrees of deendence from oil imorts, different degrees of extra-eu trade or different ass-through coefficient, for instance, the imact of similar shocks (the same variation in the rice of oil or the same exchange rate movement) may have differential effect on national inflation rates. uch external factors seem to have contributed to the increase of inflation differentials, esecially for countries like Netherlands, Greece and Portugal, whose inflation rate, during the first years after the introduction of the Euro, seemed to have been affected by the dereciation of the currency (ECB, 2003). The external factors exlanation aears to us to be the most suited to exlain the henomenon, and we would certainly subscribe to a statement to the effect that the exchange rate is a major factor in exlaining inflation divergence, as exchange rate movements are the most imortant source of the variation of imort rices (Honohan and Lane 2003). The authors conclude that countries whose nominal effective exchange rate dereciates more than the Euroean average will have higher inflation rate. This result was confirmed a few years later: if the Euro s dereciation contributed the rise of inflation in the most exosed (to extra- EU trade) countries, the areciation that followed contributed to reduce inflation 4 But also Portugal, Greece and, to a lesser extent, ain and Netherlands. 5 onetary factors, for instance. 5

6 differentials, driving a shar fall in inflation rate in the more externally oriented member countries (Honohan and Lane 2005). The above mentioned literature, even though emhasizing the imortance of the exchange rate in exlaining inflation differentials among Euro Area countries, neglects the role that intermediate imorts may lay, since it regards the different degrees of exosure to external trade in finished goods as the unique channel through which the exchange rate may influence rices in an oen economy. In an internationally integrated roduction rocess, instead, any exchange rate change causes also a variation in the rices of intermediate roducts, thus affecting national roduction costs directly. 2. A ecific-factor odel with International Fragmentation of Production 2.1 Basic tructure of the odel Consider a small, erfectly cometitive economy roducing two commodities, manufacturing () and services (). The manufactured good is traded in the world market at a given rice ( ). ervices are, instead, non tradable. The economy emloys three factors of roduction: each sector makes use of a secific factor -caital, K, in the service sector, and intermediate roduct I in manufacturing, along with labour (L), which is free to move across industries to equalise the value of its marginal roduct. Neither labour nor caital are internationally mobile. Endowment of caital and labour are given and treated as arameters. ecific factor I, instead, is imorted from the rest of the world at a given rice, I, assumed constant in world currency. Technology, with variable coefficients, is described by the columns of the A matrix 6 : a A = a L I a a L K where a ij indicates the number of units of inut i required to roduce one unit of commodity j. Each inut coefficient a ij is a function of the relative cost of factors (ex. w a = I f ), I 6 The formulation of the model is borrowed from Jones (1965). 6

7 and it is chosen among all the ossible combinations of factors (reresented by the unit isoquants) in order to minimise unit roduction costs. The following relationshis show the full emloyment conditions for the three factors of roduction: a + a L [1] L L = a K = K [2] a I = I [3] Equation [1] and [2] reresent the full emloyment of the two factors the economy is endowed with; equation [3], instead, is the economy s demand for the intermediate roduct I, imorted from the rest of the world and emloyed to roduce an amount of manufactured roduct, given its rice in world currency, in fact, we have that services, The manufactured roduct s rice, I in domestic currency. Being I = I e 7. I the intermediate roduct s rice, is determined in the world market; the rice of, instead, is endogenous. Equations [4] and [5] reresent the maximum rofit conditions in a cometitive equilibrium: a + a w = [4] I K I L L a r + a w = [5] where w reresents the nominal wage, r is the return of caital, I, and reresent the rices of intermediate inut, manufactured roduct and services, all exressed in domestic currency. Trade balance is defined by equation [6]: the value of the country s exorts (a share of roduction, EXP ) equals the value of imorts of intermediate inut I: = I [6] EXP I Combining eq. [6] and [3] we obtain the ercentage of total roduction of the country exorts: 7 As usual, we indicate with the letter e the domestic currency rice of foreign exchange. 7

8 EXP I ai = [7] As exected, the share of manufactured roduct the country sells in the world market deends on the terms of trade, er unit of outut, I, and on the intensity with which intermediate inut is emloyed I I a I. Note that ϑi a =, i. e. the distributive share of factor I on the rice of the manufactured roduct; consequently, trade balance imlies the equality between the ercentage of exorted and the distributive share of the intermediate inut on the rice of the manufactured roduct: EXP = ϑ [7 ] I e finally define the elasticity of substitution between the two commodities, manufactured roduct and services: ˆ ˆ C σ D = [8] ˆ ˆ where ˆ C indicates the ercentage of consumed within the economy. 2.2 Effects of a Change in the Terms of Trade Consider an exogenous fall in the rice of the imorted intermediate inut due to an areciation of the domestic currency: if the rice of the manufactured good is assumed, as for now, constant in domestic currency, the shock imroves the country s terms of trade. The maximum rofit condition for the exosed sector becomes: a e + a w = [4 ] I I L Any movement of the exchange rate (consider, in this instance, an areciation of the domestic currency) generates a change in the factor rice ratio w in the manufacturing I e 8

9 sector. Consequently, the coefficients a L and I a will change, together with the allocation of labour in the two sectors. The variation of the inut coefficients deends on the elasticity of substitution between the factors of roduction in the two sectors: firms, indeed, will substitute labour with intermediate inut, whose rice decreased. Being the wage equal, the reduction of the rice of the imorted factor will generate extra rofit in the manufacturing sector, if the rice of is held constant. This is shown in relation [9], which reresents the equation of change of the maximum rofit condition for the manufacturing sector: ϑ eˆ + ϑ wˆ = 0 [9] I L Let the symbol ^ indicate the relative change of the variable below and ϑ ij the distributive share of factor i on the rice of commodity j 8 : ai I e ϑ I = ; a e + a w I I L al w ϑ L =. a e + a w I I L The reduction of I (in domestic currency) allows the country to imort a greater amount of intermediate inut and the manufacturing sector to exand its outut, which is consistent with the increase in the endowment of the secific factor of that industry. The exansion of the outut attracts labourers from the service sector, and the ercentage increase in salary which restores the equilibrium is given by eq. [9 ]: ϑi wˆ = eˆ [9 ] ϑ L The bigger the distributive share of the intermediate inut on the rice of, the stronger is the growth in nominal wage following an areciation of the domestic currency. If ϑ I is greater than ϑ L wages increase, in ercentage terms, more than the exchange rate itself. The erfect mobility of the labour force ermits the transmission of the exchange rate shock from the manufacturing sector to the services and drives the equalisation of nominal 8 The distributive shares are considered constant in both sectors. 9

10 wage in the two industries. The adjustment of the marginal roductivities takes lace through the reallocation of workers and the consequent variation of the amounts of and roduced. If the total labour force is given, equation [10] must hold: λ Lˆ λ Lˆ = 0 [10] + The revious exression is obtained differentiating equation [1]; Lˆ and Lˆ reresent the relative change in the workers allocated in the manufacturing and service sectors, resectively, while the coefficient λ j is the fraction of labourers allocated in the j sector (for al L examle, λ = = ). a + a L L L e can exress eq. [10] in function of the ercentage variation of the amount roduced of and and of the â ij, namely the variations of the inut coefficients: λ ( aˆ + ˆ ) + λ ( aˆ + ˆ) = Lˆ [10 ] L L The coefficients â ij vary in function of the changes in the factors rice ratio; given the elasticities of substitution in both sectors, σ and σ, we can write 9 : aˆ aˆ L I = ϑ σ (ˆ e wˆ ) ; I = ϑ σ (ˆ e wˆ ) ; L aˆ aˆ L K = ϑ σ ( rˆ wˆ ) ; K = ϑ σ ( rˆ wˆ). L From equation [8], known that ˆ = ˆ EXP = ˆ C 10, we obtain the ercentage variation in the amount of roduced, if ˆ = 0 : ˆ = ˆ + σ ˆ [8 ] D 9 The equation is derived from the definition of elasticity of substitution between factors of roduction and from the maximum rofit condition. 10 ince = C + EXP, we have ˆ = ϑ I ˆ EXP + ϑl ˆ C. olving for C and substituting into [8], we obtain [8 ]. 10

11 Differentiating eq. [3.2] we obtain the ercentage variation in the amount of services roduced: ˆ = ˆ [11] a K ubstituting in equation [10 ] the exressions found for ˆ and Ŝ, together with the ercentage variation in salary, we obtain the ercentage variation in the return of caital deending on the exchange rate variation: ( σ ) D σ ( σ σ ) + σ ϑ I λ rˆ = eˆ 1 [12] ϑl λϑk D The effect of an areciation of the domestic currency on the return of caital, then, deends on the value of the ratio λ λ ϑ K ( σ D σ ) ( σ D σ ) + σ, i.e. on substitution ossibilities between the factors of roduction and on the elasticities of substitution on demand s side. Call A the ratio λ λ ϑ K ( σ D σ ) ( σ D σ ) + σ this term. Four main scenarios might occur: : the effect of an areciation on the return of caital will deend on 1. If the elasticity of substitution in the manufacturing sector is bigger than σ D, A is negative. This can haen, for instance, if σ = 0 : in this case consumtion (and roduction) D of the two commodities varies in the same roortion; if domestic currency areciates the roduction of both and increases, and the return of caital will grow more than the wage itself. 2. The case A = 0 occurs if we assume that both consumer references and roduction functions in the two sectors are Cobb-Douglas. If that, the roduction of services will not change, while the ercentage increase in manufacturing roduction will equal the rise in salary, and the return of caital in the service sector will increase by the same amount, too. 3. If the value of A falls between 0 and 1 the ercentage variation in the return of caital will be smaller than the variation in salary. This occurs, for instance, if substitution ossibilities in the manufacturing sector are small (or even zero): the increase in roduction 11

12 attracts labourers in the manufacturing sector, while the reduction in the amount of services roduced damens the increase in the return of caital. 4. Finally, the ercentage variation of r is negative ( A > 1) if consumer demand is very elastic to the relative rice of the two commodities: the reduction of outut in the service sector is such as to reduce the marginal roduct of the secific factor in that sector. The effect of an exchange rate shock on the final rice of services is described by eq. [13], equation of change of the maximum rofit condition in the service sector: ϑ rˆ +ϑ wˆ = ˆ [13] K L ubstituting the exressions found for rˆ and ŵ in the revious exression, we can write the ercentage variation in the rice of services deending on the variation of the nominal wage: ( ϑ A) ˆ = wˆ 1 [13 ] K or deending on the exchange rate variation: ϑi ˆ = eˆ ( 1 ϑk A) [13 ] ϑ L Once again, the effect of an areciation of the national currency on the final rice of services will deend, both in direction and in size, on the value of the A coefficient. As ˆ is a weighted average of the variations of wage and return of caital, the rice of services decreases only if the reduction of r more than comensates the increase in w 11. The following table summarises the results obtained: Table 1. Effects of an areciation of the national currency on relative return of factors, roduction and rice of services as a function of A rˆ wˆ rˆ A < 0 > 1 wˆ ˆ Ŝ ˆ ˆ > 0 ˆ > 0 ˆ > 0 11 Actually, a high elasticity of substitution in demand would be quite unconvincing, the two commodities being manufactured roducts and services. 12

13 A = 0 rˆ = 1 wˆ 0 < A < 1 rˆ < 1 wˆ A > 1 rˆ < 0 wˆ ˆ > 0 ˆ = 0 ˆ > 0 ˆ > 0 ˆ < 0 ˆ > 0 ˆ > 0 ˆ < 0 ˆ > 0, A < 1 ϑ K ˆ < 0, 1 A > ϑ K An areciation of the national currency determines an increase in the roduction of for any value of the A coefficient, i.e. indeendently from both demand and technology. This is consistent with the fact that, if the country imorts more intermediate inut, the total endowment of the country for that factor increases, and so does the roduction of the sector which emloys that factor as secific. The ercentage increase in the amount of roduced is greater the bigger is A, that is the stronger are the substitution ossibilities between factors in the manufacturing sector and the more elastic is consumer demand. On the contrary, the ercentage variation in the roduction of services decreases as A increases, and the same occurs to the rice of services. 2.3 Effects of an Areciation of the Domestic Currency Consider now an areciation of the national currency that has an imact on both the imort and the exort side. That is, the areciation makes the imorted intermediate inut cheaer but, at the same time, causes a loss in international cometitiveness, as a art of the manufacturing roduction, EXP, is traded in the world market. The economy is still rice taker in the international market for manufactured roducts, but the rice of, (and constant) in world currency. is given The full emloyment conditions for the factors of roduction do not change; the maximum rofit condition for the manufacturing sector, instead, becomes: ai I e + al w = e [4 ] while in the service sector is the same as before: 13

14 a r + a w = [5] K L Trade balance imlies that the imorts of intermediate factor equals, in value, the amount of manufactured roduct sold in the world market: that is: EXP e = I I e [6] I EXP = [6 ] I where the share of manufacturing sector s roduction the country exorts equals, as before, the distributive share of the intermediate factor in the rice (or cost) of : EXP = ϑ [7 ] I The internal demand elasticity is now defined as follows: ˆ C ˆ σ D = [8] ˆ eˆ remembering that ˆ = ˆ + eˆ, and the world rice of is assumed constant. The effect of an areciation of the national currency in the sector is obtained differentiating eq. [4 ]: ϑ eˆ + ϑl wˆ = eˆ [17] I from which we derive the ercentage variation in nominal wage that maintains the equilibrium in the manufacturing sector: eˆ = wˆ [9 ] 14

15 In this case an areciation of the national currency reduces nominal salary in the same ercentage: if the manufactured roduct becomes more exensive for foreign consumers, the entrereneur is forced to cut the costs of roduction to remain active in the world market. The variation in salary is transmitted from the manufacturing to the service sector and generates once again a reallocation of labourers between the two industries. The full emloyment condition (in differential terms) ermits to determine the variation in the return of caital following the shock in the exchange rate: ( aˆ + ˆ ) + λ ( aˆ + ˆ ) = 0 λ [10 ] L L The ercentage variation in the amount of services roduced is directly derived from eq. [2] (the total endowment of caital is fixed): ˆ = aˆ = ϑ σ ( rˆ wˆ ) [11] K L hile roduction in the manufacturing sector varies according to eq. [8], remembering that ˆ = ˆ C = ˆ EXP. ubstituting into [10 ] the revious exressions we obtain the ercentage variation in the return of caital, which equals the variation in salary and exchange rate: rˆ = eˆ = wˆ [17] An areciation of the domestic currency generates a reduction of nominal wage and of the return of caital identical, in ercentage terms, to the exchange rate variation. Consequently, relative rices of factors in the two sectors do not change: both technological coefficient and the amount roduced of and are not altered by the shock. The reduction of the nominal return of factors generates the fall in the rice of services: ˆ = eˆ [18] Thus, the areciation of the domestic currency has a deflationary effect. 15

16 2.4 An Intuitive Exlanation of Our Analytical Results: the Role of Integration and Exosure Before moving on to emirically test the redictions of the model, we want to briefly discuss the roerties of this theoretical framework. In our version of the secific-factor model, imorted intermediate inuts are not an inut as energy might be: rather, they are manufactured goods and can substitute for labour in the roduction of the home manufactured good. This roerty allows for a secific role of the relationshi between degree of international fragmentation of roduction and degree to which domestic rices react to exchange rate shocks. In the baseline scenario in which the home country roduces two goods, only one of which requires the imorted intermediate inut in roduction and is at same time the only exorted one, a shock to the domestic currency rice of the intermediate inut yields a direct imact on the structure of costs of the industry roducing the manufactured goods. Yet, inter-industry labour mobility ensures that the shock will indirectly affect the other industry as well, even though the imorted good is not an inut in its roduction rocess. It follows that the final, net effect on the good s rice deends on the degree to which the changing imorted inut rice generates a terms of trade change. Indeed, a change in imorted inut rices leading to a terms of trade imrovement requires that nominal labour reward increases, which induces in turn an increase in the non-tradable good s rice: the aggregate inflationary effect will be the higher, the larger the share of the imorted inut on the manufactured, tradable good s rice. Consider now the case of a Euro area country exorting the manufactured good to a non Euro Area country. In this instance, an areciation of the home currency will reduce the home currency rice of the imorted inut without affecting the terms of trade: manufacturing firms will be comelled to cut domestic roduction costs, which will induce a fall in both wages and the rice of the non-traded good. The final aggregate effect on the home economy will be deflationary, and the more so the more the manufacturing industry is integrated in the world s roduction network. To ut it more straightforwardly: the effect of the home currency areciation will be inflationary if firms are rice takers in the home currency, whereas it will be deflationary if they are rice takers in foreign currency rices. The degree of integration of domestic manufactures in the global value chain does not generate, er se, inflation or deflation: a variation of the rice of the imorted intermediate inut (or an exchange rate movement) triggers a transmission mechanism that, through the labour market, roduces a change in the rice of final goods. The direction of the variation, that is, whether there will be an inflationary or deflationary bout, deends uon the degree to 16

17 which national roducers are able to set the rice of the manufactured good indeendently of any exchange rate variation. Intuitively, an entrereneur who exorts a large share of outut to the world market will suffer exchange rate variations more than an entrereneur who only sells in the home market. This is what we mean by exosure, and roose to measure with the country s exort erformance. The extent of the variation, instead, deends on the degree of integration of the economy in the world roduction rocess. Now, it is aarent that as far as EU countries are concerned exorts toward another country of the area is tantamout to selling to a different area of the same country, since the single currency does not leave room for the effects of exchange rates on rice cometitiveness. This is the reason why we treat external exosure, that is, extra-ue exort erformance, and internal exosure, or infra-ue exosure, searately both in the model and in our emirical work. If a country s exorts are shied within the Euro area, then an areciation of the Euro will induce a terms of trade imrovement insofar as imorted (from the Euro area) intermediate inuts will be available at a lower domestic currency rice but will have no detectable effect on the rice of exorts. On the other hand, exorts shied to a country not belonging to the area will be less rice-cometitive after the areciation of the domestic currency: the home country gains from the areciation on the imort (cost) side, but loses on the exort (revenue) side. Under such conditions we exect that areciation will lead to lower wages and, accordingly, deflation. 3. Testing the odel s Predictions This aragrah aims at emirically testing the relationshi between an economy s degree of integration in the global value chain and its inflation rate. In our theoretical model a change in imorted intermediate rice (or an exchange rate movement) triggers a transmission mechanism which changes the overall rice level, the direction and the extent of this variation deending on the degree of exosure to external trade and on the degree of deendence on imorted intermediate inuts. The latter variable is measured using the ratio of intermediate imorts coming form extra-eu countries on total roduction; as for the former, we consider the ratio of exorts on total roduction, distinguishing between intra-eu and extra-eu exosure. e tested emirically the results of the model using a samle of eleven countries belonging to the Euroean Economic and onetary Union (Austria, Belgium, ain, Germany, France, Italy, Greece, Ireland, Finland, Netherlands and Portugal). Using data 17

18 covering the eriod , we investigated the role of international division of labour in exlaining, al least artly, inflation differentials among EU countries. e use anel data models (ooled OL and Fixed Effects models) to assess the imact of integration 12 and exosure on the domestic inflation rate. The analysis was also carried out individually for the following industries: Textiles, Chemicals, etal Products and achinery, Electrical and Otical Equiment, Transort Equiment, ervices. Formally, we submit to emirical testing two searate statements: 1. An areciation of the national currency generates a variation in the general rice level; the greater is the distributive share of the intermediate inut on the total rice of the traded finished roduct (i.e., the more the economy is integrated in the world roduction rocess), the bigger is the variation of the country s aggregate inflation rate. 2. The direction of the variation in the inflation rate deends on the degree of external exosure of the economy: we exect that an areciation of the domestic currency is deflationary for countries (and sectors) whose exosure to extra-eu trade is higher. The econometrical model we used to verify the above hyotheses is the following: ijt = β + β β β β ( int_exosure ) + ε 0 1( ro _ grijt ) + 2 ( ex _ integrationijt ) + 3 ( ex _ exosureijt ) + 4 ijt ijt where i refers to countries, j to sectors and t to time eriod (year). Variables are defined and measured as follows: ijt : growth rate of urchaser s rices in sector j for country i (source: Eu Klems). ro _ gr ijt : labour roductivity growth rate for sector j in country i; i. e. value added for hour worked 13 (index, source Eu Klems). ex _ integration = ex _ rate integration ijt t coefficient ijt : this term measures the joint effect of a movement of the exchange rate and of a variation in the degree of integration of the economy in the world roduction rocess. e need to measure the combined effect because an high (or low) value of the variable integration has no effect, er se, on the inflation rate: in our model the change in the rice of the imorted intermediate inut (which is catured, in 12 The variable integration used for estimation includes the exchange rate s contribution: integratio n = imorted intermediate inut ratio exchange rate variation. 13 The indicator is borrowed from IF (2006). 18

19 our case, by the exchange rate movement) causes a variation of the overall rice level; the integration variable, instead, determines the magnitude of the effect. ex _ rate t is the ercentage variation in the nominal effective exchange rate of the Euro 14, (the samle contains countries belonging to the Euro Area); the index is calculated considering the 44 most imortant trade artners (source: Eurostat). integratio n coefficient ijt : the sector level integration coefficient has been estimated from country-level data 15 : int _ coefficient ijt int _ coefficientit intermediate consumtion = value of outut ijt jt where the integration coefficient at country level is given by the total intermediate imorts divided by the total intermediate consumtion. Estimating the integration coefficient this way, the imorted intermediate share is the same for all sectors belonging to the same country. As in the theoretical model, the indicator reresents the cost share of the intermediate factor on the total rice of the finished good (source: Eurostat, Eu Klems). ex _ exosure ijt : measure of the degree of exosure to extra-eu trade: ( extra EU _ exorts) ( value of outut) ijt ijt, where the total exorts of sector j are divided by the total outut (in value) of the same sector. (source: Eurostat, OECDstat 16 ). int_exosu re ijt : measure of the degree of exosure to intra-eu ( intra EU_exorts) trade: ( value of outut) ijt ijt (source: Eurostat, OECDstat). ε ijt : error term. The equation is estimated on a anel of data of 66 cross section units (sector j in country i) observed for 11 eriods ( ). The countries we considered are Austria, Belgium, France, Finland, Germany, Greece, Ireland, Italy, Portugal, Netherlands, ain; the sectors considered are Textiles, Chemicals, achinery, Electrical and Otical Equiment, Transort Equiment, ervices. e estimated three different anel data models: 1. ooled OL : baseline model. 14 Data before 1st of January 1 are roxies calculated by ECB. ource: Eurostat. 15 Data on sector-level imorts of intermediate roducts are not available. 16 Data base OECDstat was used only for services. 19

20 2. LDV1: fixed-effects model (Least quare Dummy Variable), with sector and country dummies. 3. LDV2: fixed-effects model with joint dummies (sector-country). e estimated the equation using the full samle first, than on eleven country sub-samles (with sector dummies) and finally on six sector sub-samles (with country dummies). The results of the estimates are shown in table 2, 3 and 4. Table 2. Estimates from the full samle rod_growth ext_integration ext_exosition int_exosition OL LDV1 LDV *** *** -0,1028 *** (0.0051) (0.0414) (0,0372) *** *** -3,8386 *** (0.7745) (0.8210) 0, ,0503 (0.0081) (0.0043) (0,0345) *** 0,0104 (0.0029) (0.0019) (0,0114) ector dummies NO YE Country dummies NO YE JOINT N obs. Adj. R-squared ,2191 F *** *** 3,68541 *** Table 3. Estimates from six sector sub-samles VARIABLE COEFFICIENT P-VALUE 20

21 TEXTILE rod_growth -0,5594 0,0505 * ext_integration -2,2439 0,0029 *** ext_exosure -0,0884 0,6634 int_exosure -0,0690 0,6738 CHEICAL rod_growth -0,1047 0,2032 ext_integration -11,8987 <0,00001 *** ext_exosure -0,1122 0,0001 *** int_exosure 0,1409 0,3163 ACHINERY rod_growth -0,1589 0,1100 ext_integration -3,3764 <0,00001 *** ext_exosure 0,3116 0,2002 int_exosure -0,0052 0,1018 ELECT. rod_growth -0,0657 <0,00001 *** OPTICAL ext_integration -1,6850 0,0064 *** EQUIPENT ext_exosure 0,0465 0,5168 int_exosure 0,0340 0,0557 * TRANP. rod_growth -0,0601 0,0000 *** EQUIPENT ext_integration -0,6503 0,1962 ext_exosure -0,1639 0,2222 int_exosure 0,0999 0,2571 ERVICE 17 rod_growth 0,0336 0,1357 ext_integration 0,0674 0,8382 ext_exosure 1,9988 <0,00001 *** 17 Data on internal exosure are not available for the service sector. 21

22 Table 4. Estimates from eleven country sub-samles VARIABLE COEFFICIENT P-VALUE AUTRIA rod_growth -0,0286 0,4760 ext_integration -1,7992 0,1087 ext_exosure 0,1320 0,0742 * rod_growth 0,0349 0,0179 ** BELGIU rod_growth -0,1998 0,0366 ** ext_integration -3,2719 0,0720 * ext_exosure 0,0533 0,6703 rod_growth 0,0287 0,1573 PAIN rod_growth -0,0563 0,2999 ext_integration -3,9447 0,0162 ** ext_exosure 0,4250 0,0879 * rod_growth 0,2160 <0,00001 *** FINLAND rod_growth -0,1555 0,0982 * ext_integration -7,4658 0,0041 *** ext_exosure -0,0802 0,1839 rod_growth -0,1189 0,5282 FRANCE rod_growth -0,0504 0,0582 * ext_integration -5,9771 0,0335 ** ext_exosure -0,0786 0,4723 rod_growth 0,0288 0,6907 GERANY rod_growth 0,0849 0,0995 * ext_integration -2,4284 0,0150 ** ext_exosure -0,0303 0,9011 rod_growth 0,2079 0,6067 GREECE rod_growth -0,3549 0,0185 ** ext_integration -5,1348 0,1476 ext_exosure -0,1717 0,6849 rod_growth -0,1371 0,5681 IRELAND rod_growth -0,0683 0,2514 ext_integration -3,8422 0,0014 *** ext_exosure -0,0126 0,9565 rod_growth -0,0003 0,9441 ITALY rod_growth -0,5623 0,0204 ** ext_integration -9,7747 0,0058 *** ext_exosure -0,1192 0,7901 rod_growth 0,0573 0,6177 NETHERLAND rod_growth -0,0372 0,3909 ext_integration -2,9546 0,1833 ext_exosure -0,0795 0,

23 rod_growth 0,0589 0,1613 PORTUGAL rod_growth -0,0418 <0,00001 *** ext_integration -10,5044 0,0325 ** ext_exosure -0,0888 <0,00001 *** rod_growth 0,1366 0,0705 * The integration coefficient is statistically significant in the first model: the affect of an areciation of the national currency is deflationary. In the LDV1 model the internal exosure is significant, too, and shows the exected ositive sign. The last model controls for the joint time-invariant effects of sectors and countries: the coefficient of integration is again negative and significant, while the exosure coefficient is statistically zero. Table 3 shows the results of the estimates for the sector sub-samles, calculated with country dummies. The table shows interesting results: the integration coefficient is significant for Textiles, achinery and Chemicals. For this last industry, in articular, the estimated arameter value is External exosure is never significant (excet for the service sector); the exosure to intra-eu trade, instead, seems to affect rices in the exected ositive sign. The results of the estimates on the country sub-samles (with sector dummies) are shown in table 4. The integration coefficient is significant (at least 10% of confidence) for eight over the eleven countries considered. In all cases, including the non-significant ones, the coefficient has negative sign and aears relevant: for Portugal and Italy, for instance, the estimated beta is aroximately -10. The variables aear significant and show the exected sign only for Portugal. For ain and Austria, instead, both intra and extra-eu exosure seems to ositively affect inflation. Overall, the results obtained artly validate the model s redictions: the integration coefficient, as exected, is quite always negative and significant, thus confirming that an areciation of the national currency has deflationary effect, the more the economy is integrated in the international value chain. If we exclude Portugal, however, the country-level analysis does not reveal a significant imact of all the variables on the inflation rate at the same time. e thus erformed a further analysis on four sub-samles, slitting the full samle according to the values of the variables of external integration and extra-eu exosure. e estimated an equation in which the deendent variable is the inflation rate, including only the exchange rate variation among the regressors (roductivity growth serves as control). 23

24 Table 5. Estimates from four exosure-integration sub-samles Low Exosure High Exosure Low Integration var. coeff. -value var. coeff. -value Prod_growth -0, <0,00001*** Prod_growth -0, ,06488* Ex_rate_var -0, ,06988*** Ex_rate_var -0, ,05744* High Integration var. coeff. -value var. coeff. -value Prod_growth -0, ,00667*** Prod_growth -0, ,00101*** Ex_rate_var -0, ,00355*** Ex_rate_var -0, ,00007*** In the four sub-samles an areciation of the domestic currency is deflationary and the estimated coefficient associated to the exchange rate variation is always statistically significant. As redicted by our theoretical model, a high degree of integration in the world s roduction rocess magnifies the deflationary effect of the currency s areciation: the estimated beta of the exchange rate is always bigger (in modulus) in the second row of the matrix. The weakest effect, as exected, is associated with low values of both integration and external exosure. oreover, for high values of integration the effect of an areciation on inflation rate is almost the same for high and low values of external exosure 18, thus stressing the relevance of integration as a determinant factor in exlaining inflation rate in an oen economy. 4. Concluding remarks Literature on the effects of exchange rate fluctuations on national inflation rates is not articularly abundant. Furthermore, much of it deals with the issue within a framework where traded goods are final, i. e., goods entirely roduced in one country. In this framework inflation can be observed either because of an increase in the domestic currency rice of imorted consumtion goods or because of an increase of the domestic currency rice of energy and raw materials. Finally, domestic inflation is rarely modelled as the effect of a change in the degree of cometitiveness of domestic outut due to exchange rate changes. 18 The distance between the two coefficients is about one standard deviation. 24

25 In this aer we address the question of why domestic currency rices react as much as they do to a given exchange rate change. Clearly, an answer to this question ought to highlight the reason behind different reactions of domestic rices in different countries, such as those we insist calling members of the Euroean Economic and onetary Union. Large ortions of contemorary literature on inflation differential within the EU consists of econometric exercises conducted at a very high level of methodological sohistication, but makes no reference to the role of common external shocks such as it might be one coming from the external valuation of the Euro. The model we adot in this aer is very straightforward. Building mostly on research on secific-factor models ioneered by Jones, we osit that the home country is roducing and consuming two goods, a manufactured one and a service. The manufactured good is roduced through domestic labour and an imorted intermediate inut, which the world sulies at a given foreign currency rice for any amount demanded; labour and imorted inut are substitutes in roduction. The service good is roduced through domestic labour and domestic caital; neither labour nor caital are traded internationally. Finally, the manufactured good is roduced for exorts as well as for domestic consumtion, whereas all services roduced are shied to the home market only. There are two features of this model we have exloited in this aer. First, by letting the imorted intermediate inut enter as such in just one industry, we are able to derive the differential imact that a change of its domestic currency rice has on the two industries, the one roducing the traded good and the one roducing the non traded one. This oens the way to emirical testing of the differential imact of exchange rate changes at the industry level within any given country. econdly, we can study the very same issue under two different conditions: first, by assuming that the exchange rate change has no effect on a country s cometitiveness on the exort market via the domestic currency rice of the exorted good, we study the within-eu trade effects of the shock. But in a second scenario we allow for extra-eu trade, so that the exchange rate shock hits the domestic economy by altering the foreign currency rice of its manufactured exorts. It follows that an exchange rate shock will affect a given industry (in any country) the more the higher the degree of integration of that industry in the world roduction structure; and that it will affect a given industry differently in countries where it is differently integrated in the world roduction network. Just the same is to be exected about the effects of exchange rate changes on the exort side of the economy: different degrees of exosure of an industry 25

26 to cometition on the exort market will generate different reallocation of resources on the domestic economy. The model yields the results we reorted in Tables 1. e then moved to econometrically test the analytical results. e obtain a number of results which not in the least contradict the a riori according to which integration and exosure, as we mean these exression in this aer, seem to lay a relevant role in the determination of the inflation rate of an oen country. In articular, the coefficient of the variable caturing integration is negative and statistically significant in most cases, both for the whole samle and for the sub-samles (countries and sectors). Intra-EU and extra-eu exosure aear significant in some cases, but the sign of the coefficient is not always consistent with our exectations. In order to understand the role of this two variables we use a model of the inflation rate including only the exchange rate and the roductivity growth rate as indeendent variables. e estimate this model on four searate subsamles, classified according to the values ( high and low, the median being the threshold) assumed by the variables of integration and external exosure. In this way we are able to distinguish the contribution of exchange rate from the one of integration. In all subsamles the exchange rate coefficient is negative and significant: an areciation is always deflationary. For low values of integration and exosure the effect is weaker, while for higher values of integration the effect of the areciation is relevant, both for high and low values of external exosure. Overall, emirical evidence is coherent with the results of the theoretical model, highlighting the imortance of the intermediate imorts channel to exlain inflation differentials among EU countries. 26

27 References ALBEROLA, E., Interreting inflation differentials in the Euro Area, Economic Bulletin, , Aril ARNDT, ven, Globalization and the oen economy, North American Journal of Economics and Finance, Vol. 8, No. 1, , ARNDT, ven, uer-secialization and the gains from trade, Contemorary Economic Policy 16 (4), October AUKRUT, Odd, Inflation in the Oen Economy, a Norwegian odel, in L. B. Krause,.. alant, eds, orldwide Inflation Theory and Recent Exerience, the Brookings Institution, BALL, Lawrence., Has Globalization Changed Inflation?, NBER orking Paer No , BLANCHARD, Olivier, Country Adjustments within Euroland: Lessons after two years, BUETTI et al., Inflation Convergence and Divergence within the Euroean onetary Union, P No. 574, January DEARDORFF, Alan V., Fragmentation in simle trade models, North American Journal of Economics and Finance, DEARDORFF, Alan V., Fragmentation across cones, in. Arndt and H. Kierzkowski (eds.), Fragmentation, New Production Patterns in the orld Economy, Oxford University Press, DE GRAUE, Paul, and Frauke KUDELNY,, Inflation Differentials in EU, Center for Economic tudies Discussions Paer eries (DP) 00.15, ECB, Inflation Differentials in the Euro Area: Potential Causes and Policy Imlications, etember HONOHAN, Patrick and Phili R. LANE, Divergent inflation rates in EU, Economic Policy 18 (37), HONOHAN, Patrick and Phili R LANE, Exchange Rates and Inflation under EU: An Udate, III Discussion Paer No. 31, IF, How has Globalisation Affected Inflation?, in IF, orld Economic Outlook, JONE, Ronald., imle General Equilibrium odels, The Journal of Political Economy 73, No. 6, December JONE, Ronald., A Three-Factor odel in Theory, Trade and History, in Trade, Balance of Payments and Growth, edited by Jagdish Bhagwati, Ronald. Jones, Robert undell, and Jaroslav Vanek. Amsterdam: North-Holland, 1971.