What is so special about Trade in Services? 1

Transcription

1 What is so special about Trae in Services? By Daniel Mirza (School o Economics, Nottingham, U) aniel.mirza@nottingham.ac.uk an Giuseppe Nicoletti (OECD, Paris) giuseppe.nicoletti@oec.org Abstact: This article argues that trae in services has a speciic eature that oes not apply to trae in goos. As the trae service is prouce where it is consume (i.e. in the host country), we propose that it must use interactively inputs rom both the export an host countries. We consier the O-ring theory to show this proposition. We apply this original ramework to a new ataset on bilateral trae in services rom the OECD an in that wages rom both sies o the transaction an other policies aecting them or the prouct markets are inee aecting trae in services in the same way. This ining coul also explain why we in lower trae in bilateral services than goos. The title o the paper is not as original as some woul ten to think. It has been borrowe rom section titles in two other papers on trae in services (i.e. Philippa Dee, 200 an Mattoo, 999), which stress some speciic eatures to services but that are ierent rom that shown here. We thank participants at the Miwest International Meetings 2003 in Pittsburgh an especially Alex Skiba or his excellent comments an suggestions on an earlier rat o the paper. Financial support rom the Leverhulme Trust uner the programme grant F4/BF is grateully acknowlege. The opinions engage are personal an o not engage the OECD neither its member countries.

2 . Introuction We know much more about trae in goos than we o about trae in services. The most common reason is the lack o internationally comparable an well-structure ata. This urge 6 main international organisations 2 to publish a Manual on Statistics o International Trae in Services (MSITS) that makes recommenations or governments an institutions to provie coherent conceptual ramework within which countries can structure the collecte statistics relate to trae in services 3. However, our limite knowlege in that area has another explanation. Services, apart rom tourism, were use to be consiere non-traable in many trae textbooks an ew theories have been evelope so ar to explain the speciic eatures that services coul have over goos. Following Dearor (985), Melvin (988) o seminal contributions by suggesting that the existence o trae in input services (or proucer services) together with output commoities in an Heckscher-Ohlin type moel leas to a reconsieration o the law o comparative avantage. One important orce altering that law is the act that the proucer service is elivere abroa while its owners continue to consume in their home country. Ethier an Horn (992) propose another characteristic relate to consumer type services: the prouction unction o the consumer service is ierent rom that o commoities as it implies a service-prouct that is more customise to the taste o the consumer 4. Ater assuming this istinction, the authors show how reeing trae in services jointly or not with trae in gooscoul result in higher welare. Other theories moel implicitly consumer service trae through moelling oreign commercial presence. Markusen s joint work on FDI, with ew other researchers, summarise recently in his book Multinational Firms an the Theory o International Trae (2002) might be applie to goos as well as to commercial presence o oreign owne irms proviing services to consumers 2 The organisations are UN, EC, IMF, OECD an the WTO an the Unite Nations Conerence on Trae an Development. 3 Quote rom page o the Manual on Statistics o International trae in Services, jointly publishe by the above organisations. 4 See the survey o Philippa Dee (200) or more insights on Ethier an Horn (992) s work.

3 abroa. Besies, Markusen, Rutheror an Tarr (2000) explicit the role o proucer services (i.e managerial an engineering consulting) that are transerre through oreign irect investment. While they appear to be very insightul, all these theories can harly explain the large majority o international trae lows in services observable rom available atasets. Although the GATS istinguishes among our moes o supply in service trae: / Cross-borer supply (e.g. transport, inancial services, consulting, etc ), 2/ Consumption abroa (e.g. tourism), 3/ Commercial presence (e.g. the activity o oreign ailiates) an 4/ movement o iniviuals (e.g. temporary movement o workers), National Balance o Payments Statistics ientiy only service transactions between resients an non-resients. Thus, only moes, 2 an part o moe 4 are actually reporte, which rules out all attempts to explain observe trae in services igures with commercial presence type theories relate to consumer services. Also, igure reports that Transport an Tourism are by ar the most important trae services, ollowe by Financial Intermeiation an inally, Consulting, Real Estate activities an Telecommunications. Again, this leaves out little power to alternative existing theories like those o Melvin, Ethier an Horn or Markusen, Rutheror an Tarr. In this article, we propose an original set up base -rom the supply sie- on the O-ring theory evelope by Michael remer (993). A service that is exporte rom one country to another is the result o a process using ierent tasks in both countries, which interact with each other. In act, rom table 2 one can see that in most observe inustries, the prouction process o the trae service is complete in the host country (i.e. where it is consume). Thus, some tasks are perorme by omestic inputs an others are perorme by oreign inputs. I one o these tasks in either country were imperectly perorme it woul aect the prouctivity o the whole chain o tasks with which it interacts an thus the prouction/trae o the elivere service. Hence, the quality o the unertaken task matters in both countries. At the extreme, i one o these tasks in either country happen not to be unertaken, the whole prouction/consumption o the service sol abroa vanishes. A irect implication is that any given group o tasks unertaken by prouction actors in the host (importing) country cannot be substitute to services o the same type o actors in the exporting country to prouce the eventually trae service. Many examples ten to support the argument: In Transport or example, route inrastructure or airports in the two countries are neee in

4 orer to supply internationally the service. In orer to set up a telecommunication, the two countries shoul be equippe with computers, reliable telephone cables, etc. In tourism, Tour Operators in one country sell holiays to be consume in a oreign country, which price not only epens on say, marketing or avertisements provie in the ormer country but also on the costs o accommoation, catering an other leisure activities perorme in the latter. Labour is also neee in both countries to perorm all o these tasks. Besies, the intensity o trae epens on the perormance quality o both omestic an oreign tasks. This aspect, we argue, is a speciic eature to trae in services. Although prouction o inal goos coul be ragmente over ierent proucing sites in several countries, those tasks unertaken abroa by oreign actors o prouction might not interact with those at home. In act, omestic actors can still be able to prouce the inal goo with an alreay prouce stock o intermeiary inputs, whether this stock is locally provie or prouce in a oreign country. This characteristic or the internationally trae service has some implications or policy. As tasks rom both countries interact, all the moes o regulations rom both countries coul be aecting the quality o the tasks, which coul result in higher input prices, higher price o the trae services an hence lower bilateral trae. This is clearly not the case or trae in manuacturing where regulations in one country might increase the relative competitiveness o its partners. We use a new OECD ataset on bilateral trae in services to test or the interaction o tasks in both countries. Also, our work stans as one o the irst attempts to assess the eterminants o bilateral trae in services together with a very recent stuy by Grünel an Moxnes (2003). Our inings are consistent with the O-ring theory. Especially, labour wages, inrastructure in transport an that in telecommunications in both countries are aecting interactively bilateral trae in services. We also prouce evience on host country s regulations (tax wege, prouct market regulations) having similar negative eects on bilateral exports in services than the source country s regulations. 2. Some stylise acts So ar, the OECD has mae available only aggregate values o bilateral trae in services, which oes not inorm about the type o service trae within country pairs. Thus, in orer to unerstan what kin o

5 services are hien behin this aggregation, we look at another OECD ata that provies sectoral inormation o total service trae by country (i.e. not broken up into country-partner pairs) 5. Figure shows the average inustry structure o OECD in exports an imports services. Most services trae, whether exporte or importe, relate to tourism (aroun 30 per cent) an transport (aroun 25 per cent), ollowe by the business services (2 per cent) an inancial services (6 per cent). The composition o services trae is similar at the iniviual country level 6. In other wors, a large share o trae in services is relate to international movements o manuacture goos an people. Yet, the most striking eature o services trae is that trae intensity appears to be much lower than or manuacture goos, with both exports an imports lows in manuacturing several times higher than the corresponing lows in services in all OECD countries or which ata are available. Inee we have compare two ratios in each OECD country: the proportion o trae in goos to total GDP in goos an the proportion o trae in services to total GDP in business services. It appears that the trae intensity in goos ratio is 2 to 7 higher in most countries, but in some extreme cases, like Mexico it is even 0 times higher. See table or more etails 7. In international trae textbooks, lower services trae intensities are oten relate to the cost o transport, which is, or some services, much higher than the cost o shipping manuacture goos. 8 Also, the inluence o geographical an structural actors, such as location an size o the economy, on trae in services an manuacture goos may ier. Cross-country patterns o export intensities an import penetration ratios suggest that these actors play partly the same role as or trae in goos (Figure 2). Trae is strong in relatively small an well-locate countries -- such as Austria, Belgium, the Netherlans an Irelan -- an weak in relatively large or remote countries -- such as the Unite States, Japan an Australia. 5 Data on trae in services by sector (bilateral inormation is not available) are base on the OECD Statistics on International Trae in Services Results available upon request. 7 Such gaps between trae in goos an services coul be narrowe once cross-borer supply through commercial presence (i.e. FDI) is taken into account. However, it can also be argue that the Commercial Presence (C.P) type o supply has to be compare with Greeniel Investments (G.I) instea o pure trae in goos, as both CP an GI are consiere to be Foreign Direct Investment. 8 For instance, services provie by a barber are harly traeable between cities or regions within a country, not to mention across borers. However, the provision o many other services, incluing some o the most ynamic ones

6 However, these patterns are less clear than or manuacturing. We perorm a Spearman correlation test an in that inee, the cross-country correlation between export intensities in manuacturing an services is relatively low (aroun 0.2), though the correlation o import penetration ratios is higher (aroun 0.6). This suggests that other orces are impinging on the openness o OECD economies to services trae. 3. Some Theory 3. Deman sie As or trae in goos, one can obtain a bilateral import eman unction or services rom a Sub-utility CES unction where consumers etermine the quantities o each service variety to consume uner their buget constraint. To see this, assume there are F 2 traing countries, each o them being associate with a representative consumer as well as a representative service sector proucing a ierentiate goo (inustry inex k implicit in the whole section). The representative consumer o country, {,..., F}, maximises the Spence-Dixit-Stiglitz sub-utility unction D n U = α x = v= U subject to his or her buget constraint, where: σ v σ x v stans or total eman o variety v aresse to its proucer in country an n or the number o varieties prouce in country an available in country. The parameter o preerence σ σ α can be viewe as proxies o home bias or also revealing taste or perceive quality o goos (See Feenstra (994), Hea an Mayer (2000) or also Erkel Rousse an Mirza (2002)). Finally, σ is the elasticity o substitution between the ierent available varieties ( σ > ). The irst orer conitions lea to eman equations per variety. Summing over the number o varieties supplie by country i, one can obtain the bilateral export expression o country i to country j: over the past two ecaes (such as communication, inancial intermeiation an business services) involves lower transportation costs, which are urther ecreasing as inormation an communication technologies (ICT) sprea out.

7 with σ P E σ EX = N. xv = N. α.. () IP IP σ σ vi D N σ IP = α P the price inex o the composite prouct an E the total expeniture = v= evote to consuming the ierentiate goo in country. While it is very convenient to assume the existence o varieties in Tourism (e.g. several sites to visit, ierent hotels an restaurants where to go, etc ), this hypothesis might not be very suitable or transport services where they coul be assume to be relatively homogenous. We can show however that a particular case o the eman unction () woul arise i we consier CES unctions where goos are only ierentiate by their country o origin (Armington type hypothesis) but still perectly homogenous within each country (see or instance Dearor, 995). Equation () woul be retrieve, but as only one variety is prouce by country ( N = ), the number o varieties woul not enter the equation anymore. 3.2 Supply sie As note earlier, on the supply sie an O-ring prouction unction (remer, 993) is consiere. Assume a prouction o a variety o a trae service that consists o ierent tasks, which interact with each other. For simplicity, all varieties are treate equally which allows to remove the subscript v rom the ollowing relations. A group o tasks are unertaken in the omestic country, an another group in the oreign country. Each task k in a country h ( h {, }) employs a number L k, an L k, o omestic an oreign inputs respectively. The perormance q o these inputs ranges rom 0 to 9. I the task is perectly perorme (q=) then the quality o the task an the contribution to the prouction process are maximal. I the task is not perorme (q=0), then the chain o prouction stops an the output is 0. Put ierently, ierent tasks are complementary within an between countries 0. The supply o a service that is trae abroa can then be expresse by: Y = qk Lk, k= k= +, q L. (2) k, k, 9 The variable q is also thought to measure the probability to perorm perectly the task. 0 See remer an Maskin (996) or more etails about complementarity o tasks.

8 All o these tasks are associate with a quality, h {, } o the perormance that might ier q k, h across countries. I oreign inputs o not participate in the prouction process, then q = 0, an the transaction between the two countries is not realise (Y=0). I the same service ha to be sol to the omestic market an assuming a quality o the perormance iering across countries because o structural ierences, then the technology between a trae service an its peer sol at home, woul be ierent. Inee, as equation (3) below suggests, Y = q k=, L. (3) k the same elivere service Y woul imply only omestic inputs. This simple result is very speciic to the nature o the transaction as the prouction process o a service en where it is to be consume (i.e. in the importing country). k, In remer s original ramework (993), the proucers ha to choose the quality an price o inputs that maximise their proit unction. Here instea, proucers choose the quantity o labour or each task, given a certain quality o perormance an price o inputs. This eviation rom remer s ramework leas to a relation that is very close to a Cobb-Douglas unction where irms have to choose optimal quantities o ierent inputs given actor prices. However, the speciic eature that trae in services oes not occur i oreign inputs are not accounte or is more relevant to the spirit o the O-ring theory. We assume that equilibrium wages are provie rom competitive labour markets an o not epen on the quality o the perormance. Quality can be thought to be an externality, unction o other actors outsie the irm but speciic to the service inustry. This externality coul be relate to structural policies in each country like inrastructure o routes, telecommunications, market regulations in the service sector, etc. Given wages an perormance quality, a ual cost expression o the service prouction unction can be easily erive. Total cost can then be expresse as: TC = Y / w q w q k = k= k + Noting µ the mark up o the irm realise on the oreign market, CM the marginal cost o the service an τ the transaction cost rom traing, the bilateral price expression obtaine at equilibrium is: This is one by minimising total costs uner the constraint o a given prouction (equation 2).

9 k k k k k k k T q w q w Y P..,,,, + = = = µ (4) Again, in the Armington case o perect competition but with proucts ierentiate by country o origin, the price woul equal the marginal cost o exports which is a particular case o equation 4 above, where the mark up i µ woul equal. Note rom (4) that an increase in the quality o the tasks, or in increase in wages in either country reuces export prices. Replacing (4) in (), an noting = k h x h x k, the geometric mean o a variable x in country or, the expression o trae in services becomes: σ σ σ σ σ σ σ σ σ µ α = IP q q w w T Y N E EX... (5) 4. Data Data on trae in services with partner isaggregation (bilateral trae) is taken rom the OECD Statistics on International Trae in Services Inormation or trae in services is available or two years only, where the ata is covering 20 countries an 27 partners. Activity type ata in services was taken rom STAN-OECD. In act, ata relative to Wholesale, Hotel an Retaurants (sector in ISIC nomenclature), Construction (ISIC 45), Transport an Communication (ISIC 60-64), Financial Intermeiation (ISIC 65-67) an Real Estate/ Business activities (ISIC70-74), were aggregate to one big sector o traable services. We coul then extract some country ata relative to value ae, an average wages in traable services.

10 / The expeniture o the importing country ( E ) is proxie by value ae ( VA ) in services 2 an the composite variable N. Y relative to the source country is represente by services value ae o the exporting irms ( VA ). 2/ Geometric Average o Wages o the exporting an importing countries w ) an ( w ), are proxie by ( the ratio o total compensation to total employment in traable services (resp. calle hereater wage ). One can show that this ratio is very close to the weighte average o wages. wage an 3/ Transaction costs T are represente by Geographical istance (ist). Aroun 30 percent o trae in services is in the transport sector which price shoul be, by nature, highly aecte by istance. Also, istance shoul inorm on the extent o networks in trae in services in general as it has been shown to be the case or goos (see Rauch, 996). Transactions are also expecte to be reuce in Free Tae Areas ( FTA ). 4/ The geometric averages o perormance quality in the omestic an oreign countries q an q are represente by a series o variables that we propose to introuce progressively in the teste equation. Average Human capital in the source an importing countries (resp. hc an hc ) are the irst o these variables. Human capital is provie by the OECD an is measure as the average number o years o eucation in the population. We also expect in both countries the inrastructure in transport an telecommunications as to increase the quality o the perorme task in relate services. Inee, even i actors o prouction have aequate skills in running a particular task, the absence o such inrastructures relate to services woul cancel out that task an thereby the whole cross-borer transaction. Inrastructure 2 A better proxy woul have been apparent consumption in services. However, as we lack o prouction in services ata or some countries that is necessary to compute apparent consumption, we preerre taking value ae. In any case, value ae an apparent consumption were very correlate (more than 0.95) or countries where ata on prouction was available.

11 inicators are new variables provie by the OECD an base on measures o quality an quantity o telecommunications an transport 3. We also introuce two Prouct market regulation variables in the service sector that enter the equation to test in an alternative manner. Regulations might aect an eicient unctioning o the market an thus might reuce the average quality o unertaken tasks represente by q an q. Also, market regulation in the importing country shoul aect mark-ups µ realise on that market which, in our setting, reuces exports. The two alternative inicators representing prouct market regulations in both countries are: Barriers to entry in services an Prouct market Regulations. The latter combines barriers to entry with other observe regulations in services in 998. These ata are also provie by the OECD. By reerring to the proxies above, the empirical counterpart o the service equation (4) when transorme into log can be written as: log EX t = a.logva t + a.logva 2 t a 3 log ist + a 4 ta (6) a 5 log wage t a 6 log wage t + a 7 log q t + a 8 log q t + λ + λ + λ + u t t As alreay note, the variables representing the quality o perormance are progressively inserte to the equations we estimate. Let us also assume in a irst stage that the price inex variable ( IP ), the bran image ( α ) an mark-ups µ ) that evolve with the partner country an possibly over time t are to be ( capture by the ixe eects λ an λ t. t The 3-imension-panel structure o the ata implies to a country ( λ ), partner ( λ ) an time speciic eects ( λ t ). However, their introuction woul create multi-collinearity, as some o the RHS variables use are mainly varying across countries (country speciic variables) or partners (partner speciic) whereas 3 The telecoms inex is a weighte average o several components: Mainlines per capita inex, Mobiles per capita, the percentage o Digital lines, Answer Seizure Ratio an Fault Clearance Rate. The transport inrastructure inex is also a weighte average o Aircrat Departures per capita an Length o Motorways per capita.

12 the variance in time is very small (only 2 years available). The ist attempt is then to run OLS type regressions where the three types o ixe eects are constraine to equal the constant: λ = λ + λ + λ. t However, we propose hereater an alternative metho that eals with multicolinearity problems associate with inclusion o ixe eects. As in Erkel Rousse an Mirza (2002) an Nicoletti et al (2003), we express the variables in eviations rom: / the mean exporter ( log EX. t ) an 2/ the mean importer ( log EX. t ). In aition however, we propose in this paper to run the two new obtaine relations simultaneously by applying the Seemingly Unrelate Regression (SUR) metho. This TLS-SUR metho woul then account implicitly or ixe eects but also or correlation among the resiuals o the two equations. Namely, the variables o equation (6) can be expresse in eviations rom the means / across countries an 2/ across partners, which leas to two equations o trae in services TLS an TLS respectively: log EX a 5 t = a.logva log wage t t + a a 7 3 log q log ist t + a 9 + a + v t 4 ta (TLS ) an log EX t = a.logva 2 t a 3 log ist + a 4 ta (TLS ) a 6 log wage t + a 8 log q t + a 0 + e t We have assume + λ an a 0 + e ij, t = λ + u, t with a9 an a0 two a 9 v i, t = + u, t intercepts, while v, an t e j, t are two transorme resiuals. By expressing the variables in eviations rom the means we reuce the number o parameters to be estimate in each TLS-type equation. In act, in the TLS equation, export country speciic eects an variables (inexe by or the couple t ) are accounte or in a non-parametric way. In the same ashion, import country speciic eects variables (inexe by or the couple t ) are implicitly accounte or in the TLS equation. Notice that in that equation, the price inex variable that appears in the theoretical relation but that is non observable rom the ata is alreay ( IP ) accounte or in a non-parametric way.

13 The transorme resiuals are correlate by construction. The SUR metho that we propose by running simultaneously the two equations accounts or correlation among these two resiuals an hence shoul lea to more eicient parameter estimates. Also, we constrain the coeicients o variables reappearing in both equations (like Distance an FTA) to be the same across the two equations. 5. Results Table 2 shows the irst results base on OLS equations an TLS separate equations (run nonsimultaneously). In these benchmark regressions the quality o perormance is represente only by human capital. In eqwuation, we run a typical gravity equation on services where no variables speciic to the importing country that are relate to inputs an perormance quality are inserte. The eects o value ae relate to both countries an that o the ormation o a Free Trae area are positive an signiicant while istance an high wages in the omestic countries reuce competitiveness. These results are very comparable to those we coul obtain in a regression o trae in goos. However, when we introuce the wage an human capital variables relative to the importing country in equation 2, the R2 increases o 2 percentage points. In aition, the eect o wages in the importing country are negative an statistically signiicant with a magnitue that is very close to that o the exporting country wage eect. The same observation can be mae on the quite similar eects in magnitue an signiicance that are observe on the variable o human capital relative to both countries. We then run the two transorme least squares regressions TLS an TLS separately to account or ixe an time eects (equations 3 an 4 respectively in table 2). For most variables, the results appear to be similar, at least in their statistical signiicance, to the uniie OLS regressions. Nevertheless, the FTA variable eect is positive an signiicant in the signiicantly smaller in the TLS regression only, while the istance negative eect is TLS regression. However, as the parameters on istance on the one han, an FTA on the other han, shoul be equal across the two TLS equations in theory, we apply hereater this constraint to our ollowing speciications. Equation 5 in table 3 presents the SUR-TLS regression where the two TLS are run simultaneously. We obtain the same results than those o column 2 in table 2, all consistent with theory. Thus, even when accounting implicitly or ixe eects an running simultaneously the two TLS regressions we retrieve the original result o wages an human capital in the importing country are aecting its import o services in the way suggeste by our theoretical ramework.

14 Our observe wage variable is only an imperect proxy o costs as it might inorm on the prouctivity o labour as well, which eect might be positive on trae. In that respect, the parameter on wages is a mixe eect o pure cost an prouctivity an shoul then be biase ownwar compare to a pure cost eect. Now, the introuction o human capital shoul partly capture that prouctivity eect. However, one coul still cast oubt on wages as being ully capturing costs even when human capital is consiere in the regression. Especially, i the wages in the importing country are more responsive to prouctivity than costs, then the negative eect obtaine on these wages is more the relection o higher prouctivity in the importing country. Then in that particular coniguration, what makes an exporting country less competitive is the higher prouctivity o its partner. This leas us to a more traitional reasoning that we use to make or trae in goos, at os with our applie O-ring theory. The only way to show then that the O-ring theory is still the best to it the ata is to emonstrate that the inormation containe in the wage variable is more a cost than a prouctivity inormation. There are two alternative ways to o so. First, instea o consiering wages one might consier a component o it that inorms on pure cost. Alternatively, one coul run an instrumental variable regression where wages woul be instrumente by pure cost type variables. Inee, equation 6 o table 3 shows the results o a regression where wages have been replace by wege taxes on labour, that are suppose to inorm purely on costs 4. Inee, the wege tax on labour in both the exporting an importing countries is aecting negatively bilateral exports. Moreover, the parameters absolute values are higher than those relate to wages in equation 5, which is consistent with the assumption that the latter hol mixe inormation on costs an prouctivity. The signiicance o the rest o the parameters remains the same as expecte but because wages have been replace by one o their component, the values o the parameters are now lower. In act, wages might be correlate to other variables in the regression like value ae or human capital. When wages are omitte rom the equation to test, an as the wege variable cannot ully capture wage variation, some inormation previously hel by wages coul then be retrieve in the resiuals. This then might prouce a correlation between the explanatory variables an the resiuals, which ens up biasing ownwar their parameters. In orer to avoi this scenario to be repeate we then re-introuce wages to the TLS equations to test, but now we propose to instrument wages. 4 Recall that average wages are the ratio o total compensation to employment where compensation inclues necessarily labour taxes on wages.

15 Hence, hereater we set up a 3SLS type regression that is basically a SUR metho with some enogenous explanatory variables to which we apply instrumental variables. We thus instrument wages by the wege tax in 998, employment an most other exogenous variables at han (Value Ae, istance, FTA) 5. The important point to mention here is that we nee instruments that inorm on costs. It is obvious rom above that wege taxes is inee inorming on one o the cost parts o wages. Besies, the relation between employment an wages shoul also be o a cost nature either when it is to consier the supply sie or the eman sie o the labour market. Notice however that we o not consier human capital, as an aitional instrument, as we know that human capital inorms about prouctivity, not on costs. Equation 7 rom table 3, shows the results. The same results than those o equation 5 are retrieve with higher parameter estimates (in absolute values) on import an export country wages. Hereater, we maintain the same set up to run alternative regressions. The mean quality o the perormance in either country has been proxie so ar by average human capital. We propose hereater to exten the number o variables that coul represent perormance. Equation 8 as to our speciication telecom an transport inrastructure variables relate to both countries. Their eect appears to be positive an statistically signiicant. Also, either relate to telecommunication or transport inrastructure, the eect o inrastructure on trae in services in the import country is the same than that o its peer rom the export country. Table 3 shows an aitional Chi2 test that oes not reject the equality o the coeicients. Hence, we constrain these coeicients to be the same or the ollowing regressions. This result suggests that the quality o perorming a task in one country, is not only relate to the unerlying inrastructure o that country but to the partner s country inrastructure as well. Goo inrastructure in one country is a necessary but not a suicient conition to be able to trae in services. Both countries shoul be well equippe in orer or the service to be trae. It is important to notice however, that FTA is not signiicant when we introuce these inrastructure variables. Equations 9 an 0 introuce two alternative variables o market regulation. In equation 9, the barriers to entry variable is consiere. Only, the importing country s barriers to entry seem to be reucing imports to that country. The exporting country s barrier to entry oes not appear to have a signiicant eect. This result seem to be consistent with our theory as long as market regulations are only aecting mark-ups obtaine rom exporting in the host market (see theoretical equation 5 above). The quality o perorming a task woul not be then relate to barriers to entry. However, the secon inicator o market regulation that 5 Both import an export country wages were instrumente in the same manner. The instruments are all expresse as eviations rom the means as to match the transormation alreay mae or the instrumente variables. We have introuce some exogenous variables as instruments to increase the number o instruments in orer to obtain more

16 is more general than the Barriers to entry one, appears to be aecting trae in services rom both ens o the transaction as coeicients relate to both countries variables are negative an statistically signiicant. Then, the quality o perorming a task might be aecte by stringent measures to regulate the markets in both countries. 6. Conclusion an Discussion 7. Bibliography Dearor, A. (985): Comparative Avantage an International Trae an Investments in Services, in Stern, R. (eitor), Trae an Investment in Services: Canaa/US Perspectives, Ontario Economic Council, pp Dearor, A. (995): Determinants o Bilateral Trae: Does Gravity Work in a neo-classical Worl?, NBER Working Paper 5377 Dee, P. (200): Trae in Services, mimeo o the Prouctivity Commission, Australia Erkel Rousse, H. an D. Mirza (2002): Import Prices Elasticities: reconsiering the Evience, Canaian Journal o Economics, 35(2), pp Ethier, W. an H. Horn (99): Services in International trae, in E. Helpman an A. Razin (eitors), International Trae an Trae Policy, MIT Press, Cambrige Massachusetts, pp Feenstra R. (994): "New Prouct Varieties an the measurement o International Prices", American Economic Review, 84(), pp Grünel, L.A an A. Moxnes (2003): The Intangible Globalization: Explaining the Patterns o International Trae an Foreign Direct Investment in Services, Norwegian Institute o International Aairs mimeo. precise estimations. Note however, that we have conucte other regressions where only wege an employment were consiere an oun very similar results. Figures are provie upon request.

17 Hea. & T. Mayer (2000): Non-Europe: The Magnitue an Causes o Market Fragmentation in the EU, Weltwirtschatliches Archiv 36, pp remer, M. (993): The O-Ring Theory o Economic Development, Quarterly Journal o Economics, pp remer, M. an E. Maskin (996): Wage Inequality an Segregation by Skill, NBER working paper 578. Markusen, J. (2002): Multinational Firms an the Theory o International Trae, the MIT Press, Cambrige Massachusetts. Markusen, J., T. Rutheror an D. Tarr (2000): Foreign Market in Services an the Domestic Market or Expertise, NBER Working Paper Mattoo, A. (2000): Trae in Services Econmics an Law, Worl bank mimeo Melvin, J. (989): Trae in Proucer services: A Heckscher-Ohlin approach, Journal o Political Economy, 97(5), pp Rauch J. (996): Networks versus market in international trae, NBER Working Paper 567. UN, EC, IMF, OECD, UNCTAD, an WTO (2002): Manual on Statistics o International trae in Services

18 Table :Comparison o trae intensity in services an trae intensity in goos Goos Services country exp. Intensity () imp. Intensity (2) exp. Intensity (3) imp. Intensity (4) Ratio (/3) Ratio (2/4) AUT CAN CZE DEU DN FIN FRA GBR HUN ITA JPN OR MEX NLD NOR POL PRT SWE USA Note: / exports (import) intensity in goos equals the ratio o exports (imports) to GDP in goos (manuacturing+agriculture+mining) 2/ exports (import) intensity in services equals the ratio o exports (imports) to GDP in internationally traable business services

19 Figure. Composition o services trae in the OECD area, 999 OECD average 2 Panel A: Exports 3 Wholesale, Retail Construction Financial intermeiation Transport o goos Tourism Real estate an business activities Post an telecommunications Panel B: Imports 3 Wholesale, Retail Construction Financial intermeiation Transport o goos Tourism Real estate an business activities Post an telecommunications. Service trae reporte by balance o payments statistics inclues only cross-borer supply an consumption abroa. Service supplie through commercial presence or movement o iniviuals are exclue. 2. Simple average. 3. Ratio o exports or imports in each inustry to total exports or imports. Source: OECD.

20 Figure 2. Intensity o global trae in services, 980s an 990s Panel A. Export Intensity Japan Mexico Unite States Germany Australia Canaa orea Italy Finlan France Polan Unite ingom Sween Spain New Zealan Panel B: Import penetration 2 Portugal Turkey Irelan Denmark Switzerlan Norway Icelan Hungary Greece Netherlans Czech Republic Belgium Austria Unite States Japan Mexico. Export intensity is eine as the ratio o exports to GDP. 2. Import penetration is eine as the ratio o imports to omestic absorption. Source: OECD. Turkey Spain Polan Australia Greece France Switzerlan Italy Germany Unite ingom orea Canaa Portugal Finlan Sween Hungary New Zealan Denmark Norway Icelan Czech Republic Netherlans Austria Belgium Irelan

21

22 Table 2: Determinants o trae in services- Basic Equations OLS OLS2 TLS TLS log VA exporter 0.829*** 0.89*** 0.856*** [0.046] [0.037] [0.033] log VA importer 0.509*** 0.888*** 0.787*** [0.042] [0.036] [0.036] log istance -0.95*** *** *** *** [0.065] [0.049] [0.077] [0.057] ta.02***.560*** *** [0.45] [0.2] [0.92] [0.2] log wage exporter *** *** *** [0.069] [0.064] [0.053] log wage importer *** *** [0.048] [0.050] log hc exporter 2.73*** 2.859*** 2.082*** [0.464] [0.36] [0.298] log hc importer 3.650*** 3.590*** [0.357] [0.39] Constant -4.50*** *** *** [.379] [.563] [0.042] [0.042] Observations R-square White-Robust stanar errors in brackets * signiicant at 0%; ** signiicant at 5%; *** signiicant at %

23 Table 3: Determinants o trae in services- complete speciications SUR-TLS SUR-TLS2 3SLS-TLS VI 3SLS-TLS inratelecoms 3SLS-TLS +agbe 3SLS-TLS +apmr log VA exporter TLS 0.928*** 0.566*** 0.957***.053***.05***.08*** [0.027] [0.028] [0.027] [0.03] [0.024] [0.025] log VA importer TLS 0.823*** 0.508*** 0.80*** 0.872*** 0.844*** 0.842*** [0.03] [0.026] [0.03] [0.026] [0.026] [0.026] l_istance TLS=TLS *** *** *** -0.76*** *** *** [0.045] [0.067] [0.049] [0.054] [0.054] [0.053] ta TLS=TLS.207*** 0.75***.89*** [0.3] [0.60] [0.24] [0.44] [0.4] [0.35] log wage exporter TLS *** *** -.06*** *** *** [0.037] [0.038] [0.045] [0.035] [0.035] log wage importer TLS *** *** -0.80*** *** *** [0.043] [0.042] [0.038] [0.037] [0.037] log hc exporter TLS 2.549*** 0.990** 2.555*** 2.234*** 2.04*** 2.039*** [0.30] [0.467] [0.3] [0.278] [0.275] [0.286] log hc importer TLS 2.88*** 2.205*** 3.46*** 2.597*** 2.329*** 2.422*** [0.349] [0.462] [0.343] [0.287] [0.283] [0.278] log wege exporter TLS -.44*** [0.203] log wege importer TLS -0.84*** [0.28] (continues next page)

24 Table 3 (continue): Determinants o trae in services- complete speciications SLS-TLS 3SLS-TLS inratelecoms 3SLS-TLS 3SLS-TLS SUR-TLS SUR-TLS2 VI +agbe +apmr log inra-transport TLS 0.335*** (exporter) [0.085] log inra-transport TLS 0.340*** (importer) [0.06] log inra-transport TLS=TLS 0.262*** 0.275*** [0.049] [0.050] log inra-telecoms TLS 2.58*** (exporter) [0.535] log inra-telecoms TLS 2.43*** (importer) [0.39] log inra-telecoms TLS=TLS 2.09*** 2.308*** [0.37] [0.306] log Av. Entry Barr TLS (exporter) [0.090] log Av. Entry Barr TLS *** (importer) [0.08] log Av. PMR TLS *** (exporter) [0.28] log Av. PMR TLS *** (importer) [0.7] Constant TLS 0.088* ** ** [0.046] [0.057] [0.046] [0.043] [0.043] [0.042] Constant TLS *** *** *** *** *** *** [0.042] [0.058] [0.043] [0.056] [0.044] [0.043] Observations R-square TLS R-square TLS Chi2-test: / transport_exp=transport_imp / telecoms_exp=telecoms_imp 0.02 Robust stanar errors in brackets * signiicant at 0%; ** signiicant at 5%; *** signiicant at %