The impact of EU preferential trade agreements on Foreign Direct Investments 1

Transcription

1 The impact of EU preferential trade agreements on Foreign Direct Investments 1 Paola Cardamone Margherita Scoppola DRAFT January 2010 Abstract: The aim of this paper is to assess the impact of the EU preferential trade agreements (PTAs) on the outward stocks of foreign direct investments (FDI) of the EU. We estimate a model based on the knowledge-capital theory of the multinational enterprise over the period by using a large sample of 173 host countries. Explanatory variables include measures of the level of bilateral protection and an indicator of deep integration provisions of PTAs. A dynamic panel model with fixed effects has been used, in order to take into account the dynamic behaviour of FDI, that is the fact that past FDI could affect current FDI, and the heterogeneity bias. Overall results show that EU FDI are a mix of horizontal and vertical FDI. The level of the EU protection negatively affects FDI while the tariffs applied by host countries exert a positive impact on FDI. Deep integration provisions do not seem to affect FDI. The results suggest that unilateral trade agreements could be more effective in enhancing EU FDI than reciprocal trade agreements. Results are also confirmed when splitting the sample in groups of host countries on the basis of their membership to PTAs with the EU and of their level of income. Key words: regional integration, FDI, knowledge-capital model, dynamic panel data JEL code: F15, F21, F23, C33 Corresponding author: Dipartimento di studi sullo sviluppo economico University of Macerata, Italy (scoppola@unimc.it). University of Calabria, Department of Economics and Statistics I Arcavacata di Rende (CS), Italy (p.cardamone@unical.it). 1 Financial support received from the Italian Ministry of Education, University and Research (Scientific Research Program of National Relevance 2007 on European Union policies, economic and trade integration processes and WTO negotiations ) is gratefully acknowledged. 1

2 The impact of EU preferential trade agreements on FDI 1. Introduction Over the past two decades the number of preferential trade agreements (hereinafter PTAs) has dramatically increased; more than one third of world trade takes now place within preferential agreements (World Bank, 2004). The new wave of PTAs mainly involves the US and the European Union (EU), which very actively promote both North-North and North- South agreements, even though a number of South-South PTAs has also emerged. The recent PTAs - especially the North-South agreements - differ from the past also because of the inclusion in the agreements of a number of non-trade provisions in areas such as investments, services, competition policy, intellectual property rights and standards. Although trade liberalization remains very important in many agreements, there are expectations that these deep integration provisions, by improving the overall economic climate and by locking-in domestic reforms, actively help developing countries to fully exploit the preferential access opportunities. There is a widespread literature assessing the impacts of the EU PTAs on trade, with contrasting results. As for the EU integration, early contributions found evidence of trade creation (e.g., Aitken, 1973; Bergstrand, 1985) while later papers have found the effect on trade amongst member countries either negative or insignificant (e.g. Frankel, 1997, Soloaga, Winters, 1999). Even less clear-cut is the evidence of the trade impact of the EU PTAs with developing countries (e.g. Martinez-Zarzoso, Nowak-Lehman, 2003; Persson, Wilhelmsson 2007, Agostino et al, 2008). In more recent years interest is growing for the possible dynamic effects of the EU PTAs and, among them, for their potential positive impact on investments. The perspective of a preferential access to the EU market together with the deep integration commitments should promote foreign direct investments (FDI); this is expected to generate technological spillovers, to create new industries and, overall, to contribute to reinforce the ability of third countries to fully exploit the benefits of the trade preferences. Although these issues are considered very important in the general political debate on the EU trade and development policies, to date they have not been adequately considered by the literature. A number of studies have focused on the impact of the EU and of its enlargements on FDI (e.g. Adams et al. 2003; Baltagi et al, 2008) but, to the best of our 2

3 knowledge, there are no papers addressing the issue of what is the impact of the overall EU PTAs on the pattern of FDI: is there FDI creation or diversion and which, among them, is likely to prevail? Does the nature of the PTA matters as for its impact on FDI? And what is the effect of the deep integration provisions, if any? The aim of this paper is to provide tentative answers to these questions by assessing the impact of the EU trade agreements on the investments of the EU firms in third countries. For this purpose an empirical model based on the knowledge-capital model of the multinational enterprise (Markusen, 2002; Bergstrand, Egger, 2007) is used to estimate the impact of both trade and deep integration provisions of PTAs on the outward stocks of FDI of the EU. The study covers all third countries and all PTAs under negotiation or already signed by the EU during the examined period ( ). This paper differs in a number of respects from other studies assessing the impact of the PTAs on FDI (e.g. Adams, 2003; Yeyati et al 2004; Medevedev, 2006; Baltagi et al 2008; Tekin-Koru, Waldkirch, 2009). First, while all studies use a dummy to take into account the PTA, in this paper we use an explicit measure of the bilateral average tariffs; this provides us directly a measure of the elasticity of FDI to tariffs with a straightforward interpretation of the impact of a reduction of tariffs. Further, by using bilateral tariffs it is possible to assess the impact of asymmetric liberalization, that is, the impact of unilateral versus reciprocal liberalization, which would not be possible with the dummy. Second, all studies (with the only exceptions of Adams et al, 2003 and Lesher, Miroudot, 2006) do not explicitly consider the deep integration provisions of the PTAs; in this paper the non-trade subjects of all PTAs of the EU have been examined and an index proxies the existence of credible commitments in the various non-trade areas. Third, a dynamic specification is used to take into account the very likely impact that past bilateral stocks of FDI have on current bilateral FDI (Egger, 2001) and that the impact of deep integration provisions may be more consistent in the long-run than in the short-run. The results offer a number of interesting insights. Overall, the estimations indicate that the trade provisions of the EU PTAs have both an investment creation and an investment diversion effect, while deep integration does not seem to affect the outward stocks of the FDI of the EU. These results are, by and large, confirmed when we split the sample of host countries on the basis of their membership to a PTA with the EU and of their level of income. 3

4 The paper is organised as follows. The next section provides the theoretical background to the empirical exercise. The third section illustrates the empirical model and the data used and deals with the estimations issues; the fourth presents and discusses the results obtained. Finally, the last section offers a number of concluding remarks. 2. Foreign direct investments and preferential trade agreements: theoretical background Trade costs are one of the key variables determining multinational activity. Early papers usually explained either horizontal or vertical FDI within general equilibrium trade models in a two-country framework. 2 Horizontal FDI occur when a multi-plant firm locates the same activity in different countries. Markusen (1984) and Markusen and Venables (2000) have shown that horizontal FDI are likely to be important when plant economies of scale are low with respect to firm economies of scale and countries are similar; further, high trade barriers represent a key factor in explaining the choice of producing abroad (tariff-jumping). With vertical FDI different stages of the production process are geographically disperse. Early models of vertical FDI emphasize that they are likely to be more important when countries differ in factor endowment (Helpman, 1984). More recent papers have integrated these two kinds of FDI within a unique framework which explain both the choice of replicating the same activities in many locations and that of fragmenting geographically the production into stages. In the knowledge-capital model the firm could either build multiple plants (horizontal FDI) or geographically separate headquarters activities, which are skilled-labour intensive with respect to the intermediate and final production, from the production plants (vertical FDI) (Markusen, 2002; Bergstrand, Egger, 2007). Horizontal FDI are likely to prevail if countries are similar in size and in relative endowments and trade costs are high, while vertical FDI could exist when countries differ in factor endowments, especially if the skilledlabour abundant country is small, and trade costs are low. The literature on multinational enterprises, thus, suggests that the impact of trade liberalization depends upon the prevalence of one or the other kind of FDI. To investigate the impact of discriminatory trade liberalization, a three country framework is needed. A number of theoretical papers has specifically addressed the issue of the impact of trade agreements on FDI by means of partial equilibrium models in a three- 2 Recent papers have expanded this traditional distinction by emphasising the more complex integration strategies of multinational enterprises (e.g. Baltagi et al, 2007; Ekholm et al, 2007). 4

5 country framework (e.g. Motta, Norman 1996; Montout, Zitouna, 2005; Ekholm et al 2007). Although using different settings, these papers share the view that the formation of a free trade area incentives export-platform FDI from both inside and outside firms, and that the overall impact on FDI depends upon the initial situation. If before the agreement (inside and outside) firms do not invest abroad, then the reduction/elimination of internal tariffs may stimulate both inside and outside firms to invest in the low cost member country and to export from that plant to the other member countries; thus, the regional agreement has an investment creation effect. However, if initially firms are already horizontally integrated in the partner countries, then the trade agreement may have an investment diversion effect: firms concentrate production in one partner country the one with relative cost advantages in production - from which they export to other member countries. For the purpose of this paper, we are particularly interested in the impact of a PTA on the pattern of the outward FDI of one partner country. Table 1 summarises the expected impacts of a number of variables and of different types of PTA on the basis of the above mentioned literature. The variables included in the table are those which have been found to be the most important determinants of FDI also by previous empirical studies (e.g. Carr et al, 2001; Markusen, Maskus, 2002; Baltagi et al, 2008). The first variable is the size of the market of host and parent countries. Market size is crucial in determining the possibility to exploit plant economies of scale; the larger the size of the markets, the easier the covering of the plant costs. Hence, we expect horizontal FDI to be positively correlated with the market size. Horizontal FDI are also expected to be positively influenced by market similarity, as predicted by general equilibrium trade model, while differences in factor endowments explain vertical FDI. Trade costs may have different impacts depending upon the nature of FDI. Host country barriers to trade positively affect horizontal FDI, while they are expected to have no impacts on vertical FDI or a negative impact if subsidiaries in the host country use inputs imported from the home country. Conversely, home country barriers to trade are expected to have a negative impact on vertical FDI, especially on export platform FDI when intermediate goods are produced in the low cost partner country and then are shipped back to the home country. Finally, other costs that firms face when they invest abroad are also important FDI determinants; these costs generally depend on the overall economic, political and social climate for (foreign) investments in the host country. Obviously, high investment costs negatively affect both vertical and horizontal FDI. 5

6 The expected impacts of a PTA are straightforward. Table 1 considers two key features which distinguish the PTAs of developed countries and, more specifically, those of the EU. First, trade concessions may be unilateral - such as the preferential treatment granted to the ACP countries before the Cotonou agreement and the Everything but arms initiative - or reciprocal (e.g. the free trade agreement between the EU and the EFTA countries). In the first case the PTA implies the reduction/elimination only of the home country trade barriers; we should expect a positive impact on vertical FDI, because firms face lower cost when shipping the product back to the home country, while the impact on horizontal FDI should be negligible. A reciprocal liberalization implies the reduction/elimination of both home and host countries barriers to trade. While this may always have positive effects on vertical FDI, we expect a negative impact on FDI if before the agreement horizontal FDI are prevailing. In the latter case, regional integration could imply the dismantling of plants in host countries. More recent regional agreements cover a number of other subjects beyond trade in goods (deep integration) (World Bank, 2004). Often agreements include investment liberalization provisions, such as the elimination of local content requirements, the removal of barriers to international capital flows and limitations of the foreign investors participation in the domestic economic activities. These provisions are expected to have a positive effect on both horizontal and vertical FDI, as they reduce investment costs. Other provisions include commitments to liberalize services, rules to protect the intellectual property rights (IPR), mutual recognition of standards, enforcement of competition policy, and dispute settlement. A number of these provisions are likely to have a positive effect on FDI. An effective protection of IPR appears to be a key factor influencing the choice to invest abroad, as it reduces the risk of a dissemination of the knowledge-based intangible assets of multinational firms; nevertheless, the empirical evidence of a positive relation between IPR protection rules and FDI is weak (Maskus, 2000). Clear procedures for dispute settlements also may favour FDI, while it is less clear the likely impact of mutual recognition of standards, as standards are non-tariff barriers; hence, their elimination may also have an investment diversion effect. Also the impact of the enforcement of competition policy is not straightforward; the literature on multinational firms have shown that their presence may be positively or negatively correlated with the degree of market power (Barba Navaretti, Venables, 2004). One argument which is frequently used to support the view that deep integration has an overall positive effect on FDI is that, by locking-in existing domestic reforms, it reduces the political risk for foreign investor. Empirical studies tend also to support this view, as they found a positive relation between deep integration and FDI, even though results appears to be not robust to 6

7 changes in the empirical setting (e.g. Adams, 2003; Medvedev 2006; Lesher, Miroudot, 2006). Overall, the impact on FDI of deep integration provisions of PTAs is difficult to predict, as there are good reasons to believe that the overall effect could be either positive or negative. 3. The empirical model The empirical model here used is more parsimonious compared to the one proposed by Carr et al. (2001) and used by Markusen and Maskus (2002) to test the knowledge-capital model. In particular, it includes only four control explanatory variables and the variables of interest, that is, trade costs and PTA specific variables. The main reason for this reduced specification is that the original knowledge-capital specification seems to perform better with cross-sectional data than with panel data (Egger, Merlo, 2007). The basic cross-section specification for the knowledge-capital model typically includes interaction terms between skilled labour relative endowments and other explanatory variables, such as the differences in GDP and trade costs. However, including these variables leads to multicollinearity among regressors in the time dimension of panel data. Moreover, the effect of the interacted terms on the dependent variable should be partially captured in our estimations by the fixed effects and the lagged dependent variable. 3 Unlike previous works, we assume the EU as the home country, instead of considering members countries as single home countries. The main reason is that the focus of this paper is the impact of PTAs on the FDI in third countries and not within the EU; the variables of interest are specific to the EU and do not change across EU members. 4 Hence, we adopt the following specification: ln ( FDI ) = β + β ln( sumgdp) + β ln( relgdp) + β ln( relskill) + β ln 4 EUjt ( dist) + β 5 ln( host _ tax) + β 6 ln( eu _ tax) + β 7deepEUjt + δ 0trend t + u EUjt EUj 0 1 EUjt EUjt 2 EUjt EUjt 3 EUjt + [1] 3 Actually, we also estimated the model with interaction terms, but coefficients of all interaction terms were not significant. Furthermore, the inclusion in the model of interaction terms reduced also the significance of the other coefficients. These results could be a direct consequence of multicollinearity. It is worth noting that also investment costs have been excluded; this is mainly because of the difficulties in finding data for the whole period and all countries covered by the study. Nevertheless, the impact of the investment cost should be captured by the fixed effects and by the lagged dependent variable in the dynamic model. 4 Intra-EU FDI are likely to be influenced by the specific features of the single member countries; nevertheless this paper examines only the extra-eu FDI. The assumption made here is that the FDI towards third countries are not affected by the features of the home member country of the multinational firms, like the size of the market or the factor endowment; rather, EU firms when deciding about exports and FDI towards third countries consider in fact the EU as a single (home) market. 7

8 where subscript j refers to the host country (j=1,...173), t indicates the year (t=1995,...,2005); u ijt is the error term and trend indicates a trend variable which captures all specific time factors which affect the dependent variable. FDI indicates the EU outward stocks of FDI, sumgdp is the sum of GDPs of host and EU countries, relgdp is the parent-to-host relative GDP and relskill is the parent-to-host relative skilled-labour endowments. The latter two variables are determined by the ratio between GDPs or skilled-labour endowments of the EU and each host country. 5 Dist is the distance between each host country and the EU, host_tax indicates trade tariffs applied to the EU by host countries, while eu_tax indicates trade tariffs applied by the EU to host countries. Deep is an index indicating if a PTA contains deep integration provisions; the index takes also into account the stage at which the PTA between the EU and the host country is; it is 0 if there is no agreement or the agreement does not contain deep integration provisions, while its value ranges from 1 to 3 - depending upon the proceeding status of the PTA - when the PTA includes deep integration provisions. 6 Outward stocks of FDI come from the Eurostat database, which reports the data on FDI coming from the balance of payments of member countries. Data on GDP are from the World Development Indicators (WDI) The skilled labour endowment of each country is measured by the secondary school enrolment provided by the WDI The distances are those provided by CEPII and are given by the simple distance in kilometres between the capitals of each host country and Brussels. A distinguished feature of this paper is the way in which trade costs and the existence of a PTA are considered. Previous studies have all used a dummy variable to proxy the existence of a PTA (Yeyati et al., 2004; Medvededv, 2006; Ekholm et al., 2007; Baltagi et al., 2008; Tekin-Koru,Waldkirch, 2009). 8 However, dummies may capture a range of other country specific effects which are contemporaneous to the PTA implementation. Furthermore, the dummy variable does not capture the impact of different level of trade preferences, as it imposes that tariffs granted under different preferential schemes are the same, nor the different impact that host and home tariffs may have on FDI. For these reasons, we use the 5 The specification proposed by Carr et al. (2001) considers the difference between skills or GDP of host and partner countries, rather than the ratio. However, Carr et al. (2001) adopt a specification in levels. Studies which consider a specification expressed in logarithm generally use the logarithm of the ratio (that is the difference of the logarithms) (Egger, Winner, 2006; Baltagi et al., 2007; Egger, Merlo, 2007; Egger, 2001 and 2008). 6 More specifically, deep is 1 if negotiations have started, 2 if the agreement has been signed and 3 if the agreement is in force. 7 Because of the huge amount of missing values, secondary school enrolment data, instead of tertiary school enrolment or labour force with tertiary/secondary education, have been used to measure the skilled labour endowment. 8 Only Baltagi et al. (2008) and Tekin-Koru and Waldkirch (2009) consider the knowledge-capital model specification. 8

9 weighted sum of bilateral applied tariffs provided by WITS as a proxy of the trade costs for each country. WITS provides data on tariffs disaggregated at six digit level for each pairs of countries. In aggregating tariffs the use of import values as weights leads to an endogeneity bias due to the fact that if tariffs are very high, imports should be very low or nil. A weighted sum has been thus computed following the MacMap procedure (Bouet et al., 2004). All countries have been split in five groups on the basis of their level of development. Then, the weighted sum of tariffs has been obtained by using as weights the imports of each country from the group of countries the exporter belongs to. By this way, the endogeneity bias due to the use of bilateral imports in the weighting procedure is reduced (Cipollina and Salvatici, 2008). Information about deep integration provisions of PTAs is provided by the relevant EU regulations. The deep integration variable is positive if the PTA includes specific commitments in investment liberalization - which is considered a key feature to promote foreign investments (World Bank, 2004) and in at least two of the other subjects, i.e. services, protection of IPR, standards, enforcement of competition policy, and dispute settlement. Table 2 reports descriptive information about the variable included in the model. The Table shows that on average the EU FDI directed to higher income countries are more than twenty times higher than those directed toward developing countries, most of which enjoy trade preferences. As expected, differences in GDP and skill endowments are very high for lower income countries and developing countries with a PTA with the EU. Countries which have not a PTA with the EU apply, on average, lower tariffs to the EU exports with respect to PTA countries, while the EU tariffs are lower for the trade preferred and lower income countries. Estimating equation [1] by OLS could raise a problem of heterogeneity bias which may be due to observable and non-observable factors specific to each country-pair. From an econometric perspective, the omission of such factors leads to a mis-specification of the equation [1], and produces biased and inconsistent estimates. To take into account country individual effects, country specific dummies are included in equation [1], that is the error term of equation [1] is decomposed as: u EUjt = α j + ε EUjt, where j country fixed effects andε EUjt is the idiosyncratic error term. α indicates time-invariant Furthermore, when dealing with the stocks of FDI it is very plausible that past bilateral FDI affect current bilateral FDI (Egger, 2001). Thus, a dynamic specification could 9

10 be more appropriate. Since OLS and fixed effect estimators yields biased and inconsistent estimates with a dynamic panel specification, the Arellano and Bond (1991) estimator - which is based on a generalised method of moments (GMM) applied on the first-differenced equation - is here employed. Blundell and Bond (1998) proposed a system-gmm approach which combines first-differenced model (with lagged levels of FDI as instruments) and level model (with lagged differences of FDI as instruments). As already argued by Egger and Merlo (2007) this approach is not appropriate in our context. Indeed, system-gmm requires that levels of the series of the bilateral FDIs among countries should not deviate systematically from their long-run value, but this is not plausible in the case of the stocks of FDI. Thus, the Arellano and Bond (1991) estimator is here employed. 4. Results Table 3 reports the results obtained by estimating equation [1] by fixed effects and dynamic GMM, as discussed in the previous section. We also report estimates obtained through OLS for comparison. 9 As for the OLS estimates, the coefficients have the expected sign, except for the differences in the skilled-labour endowments variable (skillrel), which shows a negative and significant coefficient. The joint size of the EU and host country markets (sumgdp) has a considerable positive and significant effect on the outward FDI of the EU, as shown by the values of the coefficient, in line with most empirical studies on the determinants of FDI. Conversely, relative GDP, as expected, has a negative impact. Among variables concerning PTAs, only the deep integration (deep) variable seems to have a significant effect on the FDI of the EU with, as expected, a positive sign. Results partly change when we take into account heterogeneity bias by considering a fixed effect model. 10 While the fixed effect estimates confirm the relevance of the market size and the negative impact of market differences, unlike the OLS estimates differences in skilled labour endowments exert here a positive impact on EU FDI. These findings suggest that the pattern of the EU FDI is a mix of vertical and horizontal FDI, in line with the results of previous empirical studies dealing with the FDI of the US (e.g. Carr et al, 2001; Baltagi et al 2007). Confirmation to this hypothesis is given also by the coefficients of the EU tariffs 9 It is worth noting that we should have 1903 (11*173) observations. However, there are many missing values for GDP, secondary school enrolment and host tariffs for less developed countries. Moreover, for more than four hundreds observations, the EU FDI to developing and less developed countries are equal to zero. Taking the logarithm, these observations are dropped. 10 The distance variable has been dropped here, as it is absorbed by country fixed effects, which capture the effects of all specific country variables. 10

11 (EU tax): we observe a significant and negative effect of EU duties on FDI. This means that high EU tariffs prevent the location of production processes abroad, as firms have to face high trading costs to re-enter the final products into the EU market. Thus, the results suggest that the PTA implies an investment creation effect. The presence of deep integration provisions (deep) in the agreement has a positive but not significant impact on the FDI. Table 3 reports also the results obtained for the dynamic model using the Arellano and Bond (1991) estimator. 11 The dynamic specification seems to work well, as indicated by the serial correlation tests and the Hansen test. Serial correlation tests show, as expected, a firstorder autocorrelation because we are considering a first differenced model, but the absence of second-order autocorrelation. Furthermore, the Hansen test does not reject the null hypothesis that instruments are exogenous. As results show, past FDI significantly and positively affect current FDI. Thus, the dynamic model estimated through the Arellano and Bond (1991) procedure should be considered as the most appropriate. Results for the dynamic model also confirm the considerable positive impact of market size, and that FDI are larger the more similar is the countries market size and the greater the differences in skilled-labour endowments. The coefficient of the skilled-labour differences is considerably higher than in the fixed effect estimation and significant at the 1% level. Thus, the estimations seem to provide a rather strong and robust evidence of the coexistence of factor endowment based FDI, together with market-oriented FDI. The policy variables confirm, by and large, the results already obtained with the fixed effect estimation. Deep integration commitments do not affect investments by the EU firms in foreign countries. We have also estimated their long-run impact in order to verify if deep provisions could affect EU FDI in a long-run perspective. The long-run coefficient and the relative standard error are reported at the bottom of the Table 12 and show that also in the long-run deep integration provisions do no significantly affect the EU FDI. The EU trade protection has a negative impact on FDI, supporting the hypothesis of the existence also of vertical FDI. Unlike the fixed effect estimation, in the estimation of the dynamic model the coefficient of the host country tariffs (host tax) is significant at the 10% level and, as expected, is positive, indicating that a high level of trade protection in the host country incentives (tariff-jumping) FDI. This result reinforces the hypothesis of a mixed pattern of FDI. Further, it is worth noting that the coefficient of the EU tariff is about the 5% 11 In estimating the first-differenced equation through the GMM estimator we take into account the fixed effects as well; hence, also in this estimation the distance variable has been dropped. β. 1 t 1 12 The long-run coefficient is given by ( ) deep β FDI 11

12 of the value of the host country tariff coefficient; this means that a percentage reduction of the host average tariff would yield a change of the FDI twenty times greater than an identical percentage reduction of the EU tariff. These findings overall suggest that the PTAs of the EU may have both an investment creation and an investment diversion effect. Agreements which unilaterally reduce the EU tariffs should incentive FDI in the partner countries; any percentage reduction of the EU tariff has the effect to increase by 0.014% the outward FDI of the EU in partner countries. Conversely, for those agreements reducing tariffs on a symmetric bilateral basis, we should expect that investment diversion more than offset investment creation; our empirical evidence for the EU suggests that the same percentage reduction of the average tariff would yield a reduction of horizontal FDI in the host country twenty times higher than the increase in vertical FDI. The main implications is that if the aim of the EU and of host countries signing a new agreement is also to create new EU investments in the host countries, then a EU unilateral liberalization, rather than a reciprocal one, is much more appropriate. A further extension of the analysis is the estimation of equation (1) by splitting the sample in two groups: developing countries with a PTA under negotiation, signed or in force during the examined period - with the EU, and the other countries. The first group, hence, includes countries eligible for GSP, the ACPs, the European Neighbourhood Policy (ENP) countries, the Mercosur countries, Chile, Mexico and South Africa. Table 4 reports the obtained estimates. The significance of the lagged dependent variable and tests on serial correlation show that the estimates through the Arellano and Bond (1991) estimator is the more appropriate. In the latter estimation, results confirm for both groups of countries the general findings as for the size of joint market and the difference in labour skills, which both positively affect the EU FDI; on the contrary, relative GDP has a (negative) significant impact only for the second group of countries. Moreover, as for the first group of countries, host tariffs enhance FDI, while EU tariffs reduce FDI. Similarly to the estimates for the total sample of countries, also in the case of the developing countries with a PTA with the EU the absolute coefficient of host tariffs is much higher than that observed for the EU tariffs. This confirms our general finding that unilateral trade agreements may be more effective in enhancing EU FDI towards developing countries rather than reciprocal liberalization, as in the latter case investment diversion may more than offset the investment creation effect of a PTA. Conversely, EU FDI towards the second group of countries - which include the developed countries but also the few developing countries which do not have any PTA with the EU - are not affected by tariffs of host countries, while the coefficient of the EU tariffs is significant 12

13 although the sign is not as predicted by the theory (Table 4). Finally, deep integration provisions have a positive but not significant effect on the EU FDI in countries which do not have a PTA with the EU, both in the short and in the long-run. On the contrary, the coefficient is significant but negative for host countries which have a PTA with the EU. This result contrasts with previous studies which, as mentioned before, have found (weak) evidence either of a positive effect or no impact (e.g. Adams, 2003; Medvedev 2006; Lesher, Miroudot, 2006). This negative impact of deep provisions on FDI in PTA countries is even greater in the long-run. Estimations have been also run by splitting the sample in high and upper-middle income countries and low and lower-middle income countries, according to the classification of the World Bank (Table 5). Also for these estimates, the Arellano and Bond (1991) method should be considered as the more appropriate. For both groups of countries, EU FDI are positively correlated with the size of joint market and negatively affected by the relative market size. The coefficient of relative labour endowment is positive and significant for high and upper-middle countries, while it shows a negative and significant sign for the low and lower-middle income countries. 13 Moreover, host trade protection incentives the EU FDI in low and lower-middle income countries, confirming the presence of tariff-jumping FDI in this group of countries, while results do not provide such an evidence for the high and uppermiddle income countries. EU duties have a positive and significant effect on FDI towards both groups of countries; as mentioned before, this result is not in line with the predictions of the theory. Finally, Table 5 shows that deep integration provisions have a significantly negative effect only on the EU FDI towards low and lower-middle income countries, both in the short-run and in the long-run, confirming the evidence found for PTA countries. In other words, estimations made with different aggregates of countries confirm, by ad large, that deep integration provisions negatively influence the EU FDI toward developing countries. To further investigate this issue, equation (1) has been estimated by including in the model, instead of deep, a dummy variable representing the single deep provision (investment liberalization, service liberalization, standard recognition, protection of IPR, enforcement of competition or dispute settlement). The main objective is to verify if, despite the aggregate negative impact of deep integration on FDI in developing countries, there is some single provision especially IPR and investment liberalization which, on the basis of previous 13 This result is not supported by the theory of FDI and is not in line with most previous studies (e.g., Carr et al, 2001; Baltagi et al., 2007 e 2008) which have all found a positive relationship. It should be noted, however, that all previous papers cover a limited number of countries and do not include, as in this paper, lower income countries. 13

14 literature, are expected to positively influence investment showing a positive coefficient. This dummy is equal to one if the PTA includes that single provision and zero otherwise. Since these dummies are highly correlated, and cannot be included in the model at the same time, six different estimations have been run. Results, which are reported in the Appendix, confirm those obtained by estimating the model with the aggregate variable deep. 14 The coefficients of the single deep provision dummies are all negative and significant for developing countries while they are not significant for the other groups of countries. Hence, also these results confirm that deep provisions disincentive EU FDI in lower income countries and countries with a PTA with the EU. 5. Concluding remarks The aim of this paper is to assess the impact of the PTAs involving the EU on the pattern of the outward stocks of EU FDI in third countries. Using a sample of 173 host countries and the EU as the home country, we estimate an empirical specification of the knowledge-capital model over the period Unlike previous studies, which have limited the analysis to the impact of the EU and its enlargements on FDI, in this paper all host countries, including developing and less developed ones, and all PTAs have been considered. Furthermore, bilateral aggregated tariffs are used as proxies of each country trade protection, instead of dummy variables. Finally, we include an indicator of deep integration provisions in PTAs. From an econometric point of view, heterogeneity bias, which could be due to the likely correlation between country specific effects and regressors, is taken into account by including in the model country fixed effects. Furthermore, we consider the fact that past stocks of FDI are likely to affect the current ones and estimate a dynamic panel model by using the Arellano and Bond (1991) estimator. Overall results show that the pattern of the EU FDI is a mix of vertical and horizontal FDI. Markets size has a considerable positive impact on the outward stocks of FDI; further, FDI are larger the more similar is the countries market size and the greater the differences in skilled-labour endowments. These findings are, by and large, in line with previous empirical studies. Estimations provide a rather robust evidence that the EU tariffs have a negative impact on the outward FDI; thus, PTAs reducing EU tariffs are expected to create EU FDI in third countries. There is also evidence that duties applied by host countries to the EU have a positive effect on the outward FDI; this suggests that PTAs may have also an investment 14 As some deep integration dummies are very similar, coefficients are identical. In all these cases we report the results once, that is, in one column only. 14

15 diversion effect. Because the size of the coefficient of the host country tariffs is much greater than the one of the EU country tariff, we should expect that the same reduction of the host and EU level of tariffs would result in an investment diversion effect much greater than the investment creation effect. Thus, one of the main policy implications of the paper is that unilateral trade agreements could be more effective in enhancing EU FDI than reciprocal trade agreements. A further result is that deep integration has not a significant effect on the outward stocks of the EU FDI; this is in line with the findings of the few other papers which have tested the impact of deep integration provisions. These results are also confirmed, by and large, when splitting the sample of host countries in countries with a PTA with the EU versus other countries and in higher versus lower income countries. Unilateral liberalization is confirmed to be more effective in enhancing EU FDI towards countries which have a PTA with the EU. An interesting and quite robust result is that all deep integration provisions have a negative effect on the EU FDI in countries with a PTA with the EU and, more generally, in low and lower-middle income countries. Finally, it is worth mentioning that this paper has considered the effects of PTAs only on the FDI coming from the EU; however, the EU PTAs may well attract also multinational firms from third countries. Hence, further research should also include the assessment of the investment creation and diversion effect of the EU PTAs as for non EU firms. References Adams, R., Dee, P., Gali J., McGuire G The trade and investment effects of preferential trading arrangements. Old and new evidence, Working paper Productivity Commission, Canberra. Agostino M.R., Aiello F., Cardamone P., Evaluating the impact of Non- Reciprocal Trade Preferences Using Gravity Models. Applied Economics, forthcoming. Aitken, N., The Effect of the EEC and EFTA on European Trade: a Temporal Cross-section Analysis, American Economic Review 63 (1973): Arellano M., Bond S.R., Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58, Baltagi, B.H., Egger, P. Pfaffermayr, M., Estimating models of complex FDI. Are there third-country effects? Journal of Econometrics, 140, Baltagi, B.H., Egger, P. Pfaffermayr, M., Estimating regional trade agreement effect on FDI in an interdependent world, Journal of Econometrics, 145,

16 Barba Navaretti, G., Venables A.J. (2004) Multinational Firms in the World Economy, Princeton University Press. Bergstrand, J. H., The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence, Review of Economics and Statistics 67 : Bergstrand, J.H., Egger, P., A knowledge-and-physical-capital model of international trade flows. Foreign direct investment and multinational enterprises, Journal of International Economics, 73, Blundell, R.W., S.R. Bond, Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics, 87, Bouët A., Decreux Y., Fontagné L., Jean S., Laborde D., A consistent, advalorem equivalent measure of applied protection across the world: The MacMap-HS6 database. CEPII WP Carr, D., Markusen, J.R., Maskus, K.E., Estimating the knowledge-capital model of the multinational enterprise. American Economic Review 91, Cipollina M. and Salvatici L., Measuring Protection: Mission Impossibile?. Journal of Economic Surveys, 22, Egger P., European exports and outward foreign direct investment: A dynamic panel data approach. Review of World Economics (Weltwirtschaftliches Archiv), 137, Egger P., On the Role of Distance for Bilateral Trade. The World Economy. 31, Egger P., Merlo V., The Impact of Bilateral Investment Treaties on FDI Dynamics. The World Economy, 30, Egger P., Winner H., How Corruption Influences Foreign Direct Investment: A Panel Data Study. Economic Development and Cultural Change, 54, Ekholm, K., Forslid, R., Markusen, J.R., Export-Platform Foreign Direct Investment. Journal of the European Economic Association, 5(4), Frankel, J Regional Trading Blocs in the World Trading System, Washington, DC: Institute for International Economics. Helpman, E., A Simple Theory of Trade with Multinational Corporations. Journal of Political Economy 92, Lesher, M., Miroudot, S., Analysis of the economic impact of investment provisions in regional trade agreements, OECD Trade Policy Working paper.36. Markusen, J.R., Multinationals, multi-plant economies, and the gains from trade. Journal of International Economics 16, Markusen, J.R., Multinational Firms and the Theory of International Trade. MIT Press, Cambridge, MA. Markusen, J.R., Maskus, K.E., Discriminating among alternative theories of the multinational enterprise. Review of International Economics 10, Markusen, J.R., Venables, A.J., The theory of endowment, intraindustry and multinational trade. Journal of International Economics 52,

17 Martinez-Zarsoso I., Nowak-Lehman F., Augmented Gravity Model: An Empirical Application to Mercosur-European Union Trade Flows. Journal of Applied Economics, 6, Maskus, K Intellectual Property Rights in the Global Economy. Washington, DC: Institute for International Economics. Medvedev, D Beyond Trade. The Impact of Preferential Trade Agreements on Foreign Direct Investment Inflows, World Bank Policy Research Working Paper 4065, November Montout, S., Zitouna, H Does North-South Integration Affect Multinational Firms Strategies? Review of International Economics, 13, 3: Motta, M., Norman, G., Does Economic Integration Cause Foreign Direct Investment? International Economic Review 37, Persson, M.,Wilhelmsson, F., 2007 Assessing the Effects of EU Trade Preferences for Developing Countries, in Y. Bourdet, J. Gullstrand and K. Olofsdotter (eds), The European Union and Developing Countries: Trade, Aid and Growth in an Integrating World, Edward Elgar. Soloaga, I. and A. L.Winters, How has Regionalism in the 1990s Affected Trade Policy? World Bank Policy Research working paper Tekin-Koru A., Waldkirch A., North American Integration and Canadian Foreign Direct Investment. MPRA Paper 12968, University Library of Munich, Germany. World Bank Global Economic Prospects 2005: Trade, Regionalism, and Development. Washington, DC: World Bank. Yeyati E.L., Stein E., Daude C., The FTAA And The Location Of FDI. Working Papers Central Bank of Chile 281, Central Bank of Chile. 17

18 Table 1: Determinants of FDI and the expected impacts of a PTA Horizontal FDI Vertical FDI Markets size + 0 Market size differences - 0 / - Difference in factor endowments 0 + Host country tariffs + 0 / - Home country tariffs 0 - Investment costs - - PTA reciprocal trade liberalization - + unilateral trade liberalization 0 + deep integration?? Table 2. Descriptive Analysis, All host countries Host countries with a PTA with the EU Host countries with no PTA with the EU High and uppermiddle income host countries Low and lowermiddle income host countries Mean (CV) Mean (CV) Mean (CV) Mean (CV) Mean (CV) FDI stock (in thousand Euro) (.13) (.26) (.26) (.19) (.28) GDP of the Host countries (in billions Euro) 158, (.) 49, (.) 676, (.) 317, (.) 35, (.) GDPrel (in thousand Euro) 1,273, (.38) 1,516, (.42) 117, (.77) 721, (.38) 1,700, (.42) SKILLrel 2.59 (.97) 2.92 (1.01) 1.28 (3.17) 1.35 (3.95) 3.73 (1.13) HOST TAX 8.35 (1.32) 9.46 (1.48) 4.02 (1.22) 7.11 (1.34) 9.48 (1.38) EU TAX 1.10 (.67) 0.88 (.58) 2.02 (1.07) 1.24 (.79) 0.99 (.58) Total observations Note: CV is the coefficient of variation. Source: own computations. 18

19 Table 3. Estimation results. Dependent variable: FDI stocks (in logarithm) ( ) OLS with cluster Fixed Effects Arellano-Bond (1991) FDI(t-1) (.09) ** sumgdp (5.82) *** (3.77) *** (2.98) *** GDPrel (.06) *** (.68) *** (1.32) *** SKILLrel (.32) ** (.55) ** (.61) *** HOST TAX (.17) (.12) (.14) * EU TAX (.04) (.02) *** (.01) * DEEP (.12) *** (.11) (.54) DIST (.18) trend (.13) *** (.09) *** (.07) *** costant (201.99) *** (128.64) *** Observations R-squared Wald-Chi Square Hansen test (p-value) (.23) AR(1) test -1.7 (p-value) (.09) AR(2) test 0.77 (p-value) (.44) Long-run coefficient DEEP (.73) Notes: standard errors in parenthesis. ***; **, * indicate significance at 1%, 5% and 10% level, respectively. : standard errors are adjusted by clustering observations at the country level. 19

20 Table 4. Estimation results for developing countries with a PTA with the EU. Dependent variable: FDI stocks (in logarithm) ( ) Developing countries with a PTA Others Fixed Effects Arellano-Bond (1991) Fixed Effects Arellano-Bond (1991) FDI(t-1) (.08) * (.06) *** sumgdp (4.63) *** (4.42) *** (6.1) *** (1.19) *** GDPrel (.77) *** (1.24) (1.47) (1.16) ** SKILLrel (.64) * (.74) *** (1.27) (.65) *** HOST TAX (.22) ** (.26) ** (.14) * (.07) EU TAX (.02) *** (.02) *** (.03) (.) *** DEEP (.14) (.48) *** (.17) *** (1.49) trend (.12) *** (.11) *** (.15) (.03) *** costant (157.93) *** (208.31) *** Observations R-squared Wald-Chi Square Hansen test (p-value) (.39) (.98) AR(1) test (p-value) (.08) (.1) AR(2) test (p-value) (.42) (.26) Long-run coefficient DEEP (.69) *** (1.77) Notes: standard errors in parenthesis. ***; **, * indicate significance at 1%, 5% and 10% level, respectively. 20