The impact of free trade agreements on Middle East and North Africa exports of intermediate and final goods

Cardozo, Adriana; Martínez‐Zarzoso, Inmaculada; Vogler, Paula L. Article — Published Version The impact of free trade agreements on Middle East and North Africa exports of intermediate and final goods The World Economy Provided in Cooperation with: John Wiley & Sons Suggested Citation: Cardozo, Adriana; Martínez‐Zarzoso, Inmaculada; Vogler, Paula L. (2021) : The impact of free trade agreements on Middle East and North Africa exports of intermediate and final goods, The World Economy, ISSN 1467-9701, Wiley Periodicals, Inc., Hoboken, USA, Vol. 45, Iss. 5, pp. 1501-1527, https://doi.org/10.1111/twec.13214 This Version is available at: https://hdl.handle.net/10419/264571 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc/4.0/ World Econ. 2022;45:1501–1527. wileyonlinelibrary.com/journal/twec | 1501 Received: 11 November 2020 | Revised: 6 October 2021 | Accepted: 14 October 2021 DOI: 10.1111/twec.13214 ORIGINAL ARTICLE The impact of free trade agreements on Middle East and North Africa exports of intermediate and final goods AdrianaCardozo1 | InmaculadaMartínezZarzoso1,2 | Paula L.Vogler1 1Department of Economics, University of Goettingen, Göttingen, Germany 2Department of Economics, University Jaume I, Castellón, Spain Correspondence Inmaculada MartínezZarzoso. University of Goettingen and University Jaume I, Department of Economics, Platz der Göttingen Sieben 3, 37073, Germany. Email: [email protected] Abstract This paper is the first to analyse the impact of free trade agreements (FTAs) and the harmonisation of rules of origin (RoO) on Middle East and North African (MENA) countries’ exports differentiating between final and intermediate goods for a global sample of trade partners. Data on exports from four MENA countries (Egypt, Jordan, Morocco and Tunisia) to 61 destinations over the period 1995– 2016 are used to estimate a structural gravity model applying a Poisson Pseudo Maximum Likelihood (PPML) estimator. Moreover, the paper estimates the effect of the progressive adoption of the PanEuropeanMediterranean RoO. Results show that FTAs have been overall successful in increasing MENA exports. This is particularly true for FTAs that eliminate protection on agricultural products. In contrast to the existing literature, we find that the agreements concluded with European countries raise MENA exports, whereas no significant impact is found for the application of the PanEuropean RoO. KEYWORDS free trade agreement, gravity model, Middle East and North Africa, panel data, Poisson Pseudo Maximum Likelihood This is an open access article under the terms of the Creative Commons AttributionNonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2021 The Authors. The World Economy published by John Wiley & Sons Ltd. 1502 | CARDOZO et al. 1 | INTRODUCTION The Middle East and North Africa (MENA) region constitutes a heterogeneous group of countries mostly characterised by below average economic performance and a generally low level of integration into the global economy, combined with a high degree of export concentration in primary commodities (Rouis and Tabor,41). Despite exhibiting a great potential for trade due to its strategic location among European, African and Asian markets and its increasing availability of human capital, the MENA region is among the areas of the world that has not yet been able to profit from trade integration as an engine for economic growth (Wood & Yang, 33). The question of whether Free Trade Agreements (FTAs) have increased MENA trade integration remains controversial, considering the positive impact they could have on economic growth, overall development and political stability. In particular, countries expanded their economic cooperation with the European Union (EU), which is the most important trading partner for the Southern Mediterranean area, due to its geographical proximity and colonial ties. In the course of the Barcelona Process, this historical relationship has been deepened through the conclusion of association agreements between the EU and each Mediterranean country1, covering the removal of tariffs for industrial goods. Moreover, the application of the PanEuropean rules of origin (RoO), which were introduced progressively from 2010 on, established full cumulation and the same RoO across participating countries. Likewise, the European Free Trade Area (EFTA) states, Turkey and the Mediterranean countries have also increased their cooperation in the same timeframe. Furthermore, Jordan and Morocco expanded their economic cooperation with the USA by signing trade agreements in 2001 and 2006 respectively. Finally, Jordan concluded agreements with Singapore and Canada in 2005 and 2012, which in contrast to the agreements with the EU and Turkey, cover not only trade in industrial goods but also liberalisation in agriculture. It is an open question whether these FTAs have increased trade in final or in intermediate goods or both. The differentiation between intermediate and final goods is crucial when analysing the impact of FTAs on trade, given the growing importance that intermediate goods’ trade has gained over time in the development of global value chains. For this reason, and as pointed out by de MelloSampayo (17), not only trade in final goods but also trade in intermediate goods has to be considered in order to measure countries’ competitiveness. This paper investigates the impact of the FTAs on the exports of four MENA countries, namely Egypt, Jordan2, Morocco and Tunisia (from now on referred to as Med4) to 61 destinations over the period from 1995 to 2016. These countries are the most open in terms of concluded FTAs and the most diversified economies in the region, since oil and related products do not account for the bulk of their export earnings, as is the case for Libya and Algeria (The Economic Complexity Observatory, 31)3. It is reasonable to expect that the FTA effect will differ across sectors (Anderson & Van Wincoop, 3). This paper distinguishes between final and intermediate goods, since due to the growing internationalisation of production processes, these types of goods display different 1As of March 2018, the EU has concluded bilateral agreements with Tunisia, Israel, Morocco, Jordan, Egypt, Algeria and Lebanon. See Table A2 for more details on these agreements. 2Although strictly speaking Jordan does not belong to the group of Mediterranean countries, it is for convenience considered as being one of them. 3The private sector remains relatively small in the oilexporting economies of the MENA region (International Monetary Fund, 37) and consequently, the expected impact of trade agreements on such countries’ exports is fairly limited. | 1503 CARDOZO et al. dynamics (Athukorala & Yamashita, 4) and the impact of trade policy is heterogeneous across sectors (Anderson & van Wincoop, 3). Results of existing studies4 investigating the impact of FTAs ratified by Med4 are mixed at best. To our knowledge only two studies use disaggregated data, but both of those focus on a restricted sample of products and countries. Bensassi et al. (11) find a positive effect of FTAs on MENA exports, while Parra et al. (25) and MárquezRamos and MartínezZarzoso (24) do not find a positive and significant impact. Similarly, Cieślik and Hagemejer (15) and Freund and PortugalPérez (36), who base their analysis on aggregate data, fail to find a positive impact on MENA exports but do report an increase in imports from the EU. The main contribution of this paper is threefold. First, diverging from the existing literature for these countries, we estimate the gravity model of trade using the Poisson pseudo maximum likelihood (PPML) estimator in order to account for the presence of a significant share of zero trade flows, which is particularly relevant when using disaggregated data. Second, we use sectoral data classified according to the Broad Economic Categories (BEC) and estimate the model for 3and 5year time intervals to account for adjustments to trade policy. The main results show that the agreements have generally been successful in increasing MENA exports, but the effects differ across types of goods. For instance, the agreements that include actual liberalisation in agriculture have a greater impact on Med4 exports than the ones covering only industrial goods. Contrary to other studies, we find a positive impact of the EuroMed agreement on exports of final goods across different specifications. The remainder of this paper is organised as follows. In Section 2, we review the relevant literature, paying special attention to studies that consider to some extent the effect of integration on sectoral exports. The empirical approach is discussed in Section 3, which also includes a description of the data and variables used, as well as some stylised facts concerning the evolution of trade in final and intermediated goods. Section 4 presents and discusses the results. Finally, Section 5 concludes with a summary of the main ideas put forward in the paper and some policy recommendations. 2 | LITERATURE REVIEW ON MENA TRADE INTEGRATION We summarise in this section the findings of closely related papers that focus on the MENA region or cover a global sample, including NorthSouth as well as SouthSouth agreements, and use sectoral or aggregated data without focusing on a single sector5. We start by presenting the outcome of research using sectoral data (MárquezRamos & MartínezZarzoso, 24; Bensassi et al., 11) and follow with research that uses more aggregated trade flows. In both cases, the gravity model of trade is the methodological framework (Cieślik & Hagemejer, 15; Freund and PortugalPérez, 36; Parra et al., 25). Finally, we refer to papers that focus on a global sample and show some specific results for the MENA region. 4Parra et al. (26), MárquezRamos and MartínezZarzoso (24), Bergstrand et al. (12) and Cieślik and Hagemejer (14). 5For studies investigating North– South agreements, see Péridy (28), Bergstrand et al. (2011) and Bensassi et al. (10). For studies also including South– South agreements, see Parra et al. (26), MárquezRamos and MartínezZarzoso (24) and Cieślik and Hagemejer (14). Augier et al. (5) investigate the impact of the PanEuropean RoO on trade in textiles between MENA countries and the EU. To date very few studies focus exclusively on regional agreements like GAFTA and the Agadir agreement (Abedini and Péridy (1) and Péridy (27)). 1504 | CARDOZO et al. MárquezRamos and MartínezZarzoso (24) is the only paper that distinguishes between intermediate and final goods when investigating trade effects of FTAs in the MENA region. However, its main focus is on the participation of MENA countries in EuroMed production networks, that is, the link between imports of intermediate products and exports of final products. As an additional contribution, the paper estimates the impact of the EuroMed agreements, TurkeyMed FTAs and the USAMorocco FTA on intermediate and final goods exports of Algeria, Egypt, Morocco and Tunisia to OECD countries over the period 1995– 2008. A gravity model of exports of final goods is estimated, which includes as a regressor lagged imports of intermediate goods from the EU and the rest of the world (RoW) in order to capture the effect of the participation in regional production networks. Dummy variables to proxy for the existing FTAs and for the adoption of the PanEuropean RoO are also included to fully account for the effect of the EuroMed process. The estimation results show that imports of intermediate goods from the RoW and the EU positively impact exports of final goods. The authors conclude that the MENA countries have indeed become more integrated into regional production networks, as intermediate goods imports from the EU have a positive effect on these countries’ final goods imports. This positive effect is particularly observed for capital goods and transport equipment. Finally, it is shown that the effect of the changes in RoO is stronger than the impact of the tariff elimination since no significant effect is found for any of the FTAs (MárquezRamos & MartínezZarzoso, 24). Adopting a slightly different perspective, Bensassi et al. (11) use highly disaggregated trade data for a number of sectors to investigate the impact of the Barcelona Process on the exports of the same four North African countries over the same period (1995– 2008). The authors analyse the extent to which the extensive and intensive margins of their exports to France, Italy, Germany and Spain6 have been impacted by the EuroMed FTAs, also analysing the effect of RoO. They use as proxies for the intensive and extensive margins of trade the average value per shipment and the number of products exported respectively. These are the dependent variables in a gravity model of trade estimated using logged dependent variables. In contrast to MárquezRamos and MartínezZarzoso (24), Bensassi et al. (11) find that the EuroMed agreements have been successful in increasing MENA country exports to their European partners and that this is mainly due to an increase in the intensive margin of trade, that is, these countries export more of the varieties they were already exporting. This finding is in accordance with the fact that North African countries are mainly exporters of goods with a low technological content that can be easily substituted by other countries on the international market7. The authors find, however, that effects differ by sector and that the increase in exports due to changes in RoO is higher than the effect of the EuroMed agreements. Concerning the RoO, diagonal cumulation allowed the MENA countries to use cheaper or better quality inputs, increasing demand for their final products in Europe. Finally, no significant effect is found for imports from the EU, whereas a significant and negative effect on total exports is 6Intensive margin of trade refers to the mean value of an individual shipment or the quantity of every variety exported, whereas the extensive margin refers to the number of exporting firms or the number of varieties exported (Bensassi et al., 10, pp. 228– 229). 7These results are in line with the findings of Chaney (13) regarding the elasticity of substitution and its effect on the two margins of trade. According to this author, a high elasticity of substitution makes the intensive margin more sensitive to changed trade barriers but makes the extensive margin less sensitive, so that in the case of the MENA countries, the decrease in trade costs is expected to impact the intensive margin. | 1505 CARDOZO et al. found for imports from the RoW, which also holds for the number and quantity of goods exported. One possible interpretation is that imports from the RoW are replaced by EU inputs, which reduces the variety of goods that are exported to the RoW. This result might be due to other FTAs in force. In this context, the FTA between the USA and Morocco might have diverted exports to the USA (Bensassi et al., 11). Among the studies that focus on total trade, Cieślik and Hagemejer (15) investigate the impact of a range of trade agreements on the imports of seven MENA countries over the period 1980– 2005. These authors use a large sample of partner countries8. In contrast to the other papers summarised above, they find that the EuroMed agreements only increased MENA imports from the EU but had no significant effect on their exports. This is in line with the findings of Bergstrand et al. (12) regarding their results for EuroMed FTAs with Tunisia and Morocco. Similar results are obtained for the EFTAMed agreements. More recently, Freund and PortugalPérez (36) and Parra et al. (25) extended the analysis of Cieślik and Hagemejer (15), distinguishing between broad categories of goods. Freund and PortugalPérez (36) distinguish between nonoil and nonnatural resources imports, whereas Parra et al. (25) differentiate between industrial and agricultural goods in order to account for the different liberalisation schedules. The period covered in the latter paper is 1994– 2010, and hence accounts for almost all existing trade agreements of interest with the exception of the TurkeyJordan and the CanadaJordan, agreements which entered into force in 2011 and 2012 respectively. Regarding the effect of the EuroMed agreements on trade in manufactured goods, Parra et al. (25) also find a positive and significant impact of FTAs on MENA imports and a negative and significant effect on MENA exports. They argue that firms that have been mostly selling to the domestic market could not survive the newly induced competition from EU imports given the overall low productivity (Parra et al., 25). With respect to the other agreements of interest, a positive and significant coefficient is obtained for the TurkeyMed agreements on MENA imports. Regarding the impact on trade in agricultural goods, no significant effect of the EuroMed FTAs can be found, which is not surprising considering the fact that these goods have not been subject to significant changes in the liberalisation schedule. From these results it follows that the agreements that include liberalisation in agriculture are more beneficial for MENA trade integration as these countries clearly possess a comparative advantage in these goods. Interestingly, a positive effect on MENA agricultural exports is found for the TurkeyMed agreements, although these agreements include only a limited number of concessions for trade in agriculture. However, in contrast to the EuroMed agreements, these concessions are included in all TurkeyMed agreements, which might be the reason for this result. Finally, with respect to the distinction between NorthSouth and SouthSouth agreements, both types of agreements appear to positively influence the integration of the MENA countries. The results of previous studies that have been presented in this section show that the effect of the FTAs of interest in the MENA region depends not only on the agreement but also on the approach and specification chosen. In particular, the effect of the EuroMed agreements is ambiguous, as two of the three studies that use disaggregated trade data – MárquezRamos and MartínezZarzoso (24) and Parra et al. (25) – find no positive effect of the EuroMed agreements on MENA exports, whereas Bensassi et al. (11) find that these agreements were indeed successful 8In addition to the EUMed agreements, they account for the EFTA agreements, bilateral agreements between the MENA countries themselves, agreements with Canada, Mexico and the USA and FTAs with countries in Central and Eastern Europe. Additionally, to Algeria, Egypt, Morocco and Tunisia, their dataset includes Israel, Jordan and Turkey. 1506 | CARDOZO et al. in increasing MENA exports. However, it might be difficult to directly compare these results as there are a number of differences regarding the sample of countries, the analytical approach, the trade partners, the level of disaggregation and the years covered. It has been shown that the effect depends on the type of goods, whether differentiating between specific sectors or between agricultural and manufactured goods in general. Most recent papers have adopted a more advanced methodological approach, in that they include zero trade flows and estimate the gravity model of trade using a PPML approach, as recommended by Yotov et al. (43). Of the papers analysing the gravity model in this framework, a few consider a global sample of countries and also present estimates of different FTAs (Baier et al., 8,9). Using total exports for a global sample of countries over almost five decades, Baier et al. (8) find that the effect of Customs Unions (CU) for total trade is more than twice the effect of FTAs and that, whereas CU membership mainly affects the intensive margin of trade9, FTA membership affects the intensive and extensive margins equally. Using a smaller sample of 70 countries over the period 1986 to 2006, Baier et al. (9) account and test for bilateral heterogeneity in the effect of economic integration agreements on trade flows, finding considerable evidence that different pairs of countries are affected differently by the same agreement. Also the direction of trade matters, with the effects being mostly asymmetric. They present estimates of the partial FTA effect for each agreement in their sample, some of which correspond to the MENA region, including JordanUSA, EFTAMorocco, EFTATurkey, Agadir, EUEgypt, EUMorocco and MoroccoUSA. The results show nonsignificant RTA effects for all of them, whereas the effects for most of the EUEastern European countries’ FTAs are positive and significant. However, the sample period ends in 2006, and given that the effects take time to materialise and most of these FTAs were ratified in the late 1990s and early 2000s, this finding is not surprising. The authors acknowledge that FTAs can have very different effects across industries, and suggest this as a new avenue for further research. None of the above works distinguish between the effect on intermediate and final goods, which is the main novelty of this paper together with the fact the sample has been extended to include more recent years. 3 | EMPIRICAL STRATEGY AND DATA To analyse the impact of trade agreements in the MENA region, we use the gravity model of international trade, which has become increasingly popular in the trade literature. Indeed, it is considered the workhorse of international trade research due to its empirical robustness and great explanatory power (Feenstra, 19; Head and Mayer, 20; Kepaptsoglou et al., 21). Derived from Newton's law of gravitation, the gravity model of trade predicts bilateral trade flows between two countries as a function of their economic mass and distance from one another (Pöyhönen, 28; Tinbergen, 32). Trade is expected to be positively related to the respective GDP of the trade partners and negatively related to the geographical distance between them, which is considered a proxy for all sorts of trade costs. In its most widely accepted specification, which has a theoretical basis, the gravity model accounts for relative trade costs by incorporating the socalled multilateral resistance terms (MRT), as shown by Anderson and Van Wincoop (2): 9The intensive margin measures the average exports per product of the goods already exported and the extensive margin measures exports in products that had not previously been exported. | 1507 CARDOZO et al. where Xijt denotes bilateral trade between country i and j at time t; Yit, Yjt and Yt denote the GDP of countries i, j and the world at time t, respectively; tijt denotes trade costs (typically proxied by distance) and σ (>1) is the elasticity of substitution. Pit and Pjt are the countryspecific MRT, which decrease if a country is remote from the main world markets. Moreover, relative trade costs are not only determined by physical factors such as distance but also by trade policy factors such as high tariff barriers or other trade costs such as nontariff barriers (Bacchetta et al.,6). Therefore, in addition to bilateral distance, a number of variables are typically incorporated to account for trade costs between two countries, such as dummies for a common language, colonial ties, a common border and the existence of trade agreements (Parra et al., 25). In this context, it is expected that countries with similar cultural features such as a common language or colonial ties will trade more as they are likely to better understand each other's business practices than firms operating in less similar environments (Bacchetta et al., 6). Finally, trade agreements will reduce the price of the traded goods in the partner's market and are thus expected to have a positive impact on trade. With respect to the estimation of the gravity model, two major challenges arise that can be addressed using panel data. The first challenge is related to the presence of unobserved heterogeneity. It refers to the fact that trade between country pairs is determined by a number of observable and unobservable characteristics. While the observable factors can be measured and incorporated into the gravity model, unobserved factors are not measurable. In the context of the gravity model, these unobservable factors, the socalled MRTs, are of major importance and should be incorporated in the estimation. The simplest and widely used method consists of adding several sets of fixed effects, as, for example, exporter and importer fixed effects (Bacchetta et al., 6, p. 107), which capture all unobservable countryspecific characteristics that are time invariant. However, according to Baldwin and Taglioni (10, p. 800), these importer and exporter fixed effects should be time variant since otherwise the results would be biased.10 In this paper, given that we have sectoral trade data, we use exportersectortime and importersectortime fixed effects (Head & Mayer, 20). While accounting for the MRTs, these timevarying fixed effects absorb the effect of variables such as the sectoral value added of the importer and exporter countries, as well as of other observable countryspecific factors that vary by country, sector and time, namely industrial policies and real effective exchange rates (Yotov et al., 43). We diverge from the standard approach of estimating the model in its loglinear form as this approach leads to the loss of information in the case of zero trade flows. This method is adequate when the zeros are believed to be arbitrarily missing data or random rounding errors and thus carry no information (Bacchetta et al., 6). But if these zeros actually reflect zero trade or regular rounding errors associated with small trade flows, then dropping these observations will lead to a loss of information and will thus produce inconsistent estimates (Bacchetta et al., 6). Moreover, (1) X ijt = YitYjt Ywt ( tijt PitPjt )1 − 𝜎 10De MelloSampayo (16) proposes the incorporation of a competitiveness factor into the gravity model, which captures the effect of competition between countries to sell intermediate products to a specific country. In the author's words, the competition factor allows treating foreign trade, and specifically trade in intermediate goods, directed to a specific country as interdependent on decisions to trade with alternative countries. This approach is particularly relevant if a country is faced with multiple competing alternatives when taking a decision on from where to import an intermediate good. 1508 | CARDOZO et al. estimating the gravity model in its loglinear form can lead to misleading conclusions in the presence of heteroscedasticity as the log transformation affects the disturbances in the sense that the errors will have generally heterogeneous variances. The PPML estimator overcomes this challenge as it does not assume homoscedasticity and is thus valid with general forms of heteroscedasticity (Silva & Tenreyro, 30). For these reasons, to analyse the impact of trade agreements in the MENA region, we estimate a gravity model of international trade in its multiplicative form using PPML. While zero trade is less of a problem when using aggregate trade flows, it is especially relevant for this empirical application as the unit of analysis is sectoral trade flows, which contain more zeros than aggregated data11. We assume that zero trade flows in our dataset are explained by the fact that some goods are not traded between country pairs because of high trade costs or the fact that export items in a given commodity cannot compete in foreign markets. In this case, dropping the observations is not appropriate as they carry information. While the data extracted from UN Comtrade did not contain any zeros, balancing the data to obtain all possible importer, exporter, sector and year combinations has led to a large number of observations with missing trade values, either because they were not reported or because they were actually zero12. While there are a number of suitable ways to overcome the problem of zero trade flows13, PPML is preferred here as it is straightforward in its application and avoids the theoretically inconsistent method of replacing zero trade flows with an arbitrary value or the application of one of the more complex alternative methods.14 In addition to the properties stated above, PPML has one key advantage in that it allows the researcher to control for endogeneity and heterogeneity issues by including a rich set of fixed effects at the country and sectoral level15. The sectoral fixed effects should also be allowed to vary by origin, destination and time as countries might have a comparative advantage in a certain commodity that explains a high level of exports. Furthermore, sectoral trade flows are subject to fluctuations over time, which affect individual countries differently depending on their economic structures. Clearly, however, the inclusion of pair fixed effects does not allow the estimation of the standard gravity covariates such as distance, contiguity, language and religion, as they get absorbed by these effects. Some authors argue that pair fixed effects account for additional trade costs that are not captured by the standard gravity covariates and are thus better suited to proxy trade costs (Yotov et al. 43) when the main aim is the identification of the effect of a timevariant bilateral variable, such as FTA. Another reason for including pair fixed effects is the fact that for trade with Med4 countries, it seems plausible to expect that unobservable factors like stability in bilateral 11Note that the data used here are only disaggregated to a relatively low degree (two digits) and that this problem is accentuated as the degree of disaggregation rises (Yotov et al., 43, p. 19). 12Balancing the dataset led to an additional 19,582 observations, which corresponds to approximately onethirds of all observations. These zero flows are nearly equally distributed between the OECD and nonzero trade flows. 13Yotov et al. (43) (p. 19) present five possible solutions to this problem. 14According to MartínezZarzoso (38), inferences drawn from simulation studies like the one conducted in Silva and Tenreyro (31) should be handled with caution, as minor changes in the simulation setting can lead to different outcomes. As such, finding the best estimator for a given dataset requires a large variety of tests. Hence, the PPML estimator cannot be seen as a generally superior estimation method for gravity models. 15For the implementation of this estimation method, the newly available stata command ppml_panel_sg (Zylkin, 44) was employed. | 1515 CARDOZO et al. its loglinear form, meaning a large share of observations with zero trade flows drop out and they also cover a shorter time period. The finding that the EuroMed integration process had a nonsignificant impact on MENA exports of intermediate goods is in line with results of other studies and can be associated with increased integration in regional production networks in the course of the Barcelona Process, as suggested in MárquezRamos and MartínezZarzoso (24). A robust effect only for final goods is plausible in this context, if companies transferred the assembly of final products to MENA countries. Alternatively, the finding of a significant effect only for final goods could be explained by the fact that the elimination of import tariffs in MENA countries enabled MENA exporters to source cheaper and/or better quality inputs from European trade partners. This efficiency gain in sourcing is likely to have positively impacted the competitiveness of MENA exporters and thus the exports of final goods. Such a result is in line with MárquezRamos and MartínezZarzoso TABLE 3 The Impact of FTAs on MENA Exports of Final and Intermediate Goods Dependent Variable: MENA4 exports to the reporting importers Explanatory variables Final Goods Intermediate Goods (1) (2) (3) (4) Eumed 0.379*** 1.272*** 0.022 −0.093 (0.121) (0.129) (0.116) (0.086) Turmed 0.351 0.229 0.664*** 1.483*** (0.216) (0.536) (0.115) (0.233) Usamed 0.142 0.599*** 0.788*** −0.161 (0.222) (0.220) (0.286) (0.117) Jorfta 0.836*** 1.290*** 0.635 −1.296*** (0.126) (0.419) (0.643) (0.446) RoO_EU 0.012 1.317** 0.175 −0.389 (0.073) (0.519) (0.114) (0.237) Ln dist −1.698*** −1.127*** (0.047) (0.043) Contig −1.031*** 0.220** (0.126) (0.110) Colony 0.341*** 1.114*** (0.101) (0.092) Lang 1.863*** −0.112 (0.126) (0.110) Observations 28,745 29,217 23,228 23,356 R20.997 0.846 0.973 0.776 ImporterSectorTime FE Yes Yes Yes Yes ExporterSectorTime FE Yes Yes Yes Yes Pairsector FE Yes Yes Notes: Robust standard errors in parentheses. ***p<.01, **p<.05, *p<.1. Errors clustered by id: exporterimportersector. ImporterSectorTime FE capture the inward multilateral resistance. ExporterSectorTime FE capture the outward multilateral resistance. Pairsector FE are exporterimportersector FE. 1516 | CARDOZO et al. (24), who find a positive and robust effect of lagged imports of intermediate goods from the EU on MENA exports of final goods. Contrary to this, the effect of the TurkeyMed agreement on trade is statistically significant and positive for intermediate goods exports across specifications. As with the EuroMed agreement, the coefficient for intermediate goods is at least twice as large when using standard gravity variables compared to pair fixed effects (column (4) in Table 3). The agreement led to an increase of 94% in exports of intermediate goods considering model results using pair fixed effects.24 24In contrast, MárquezRamos and MartínezZarzoso (24) find no significant effect of the TurkeyMed agreement on the two types of goods. Parra et al. (26) investigate the impact of the TurkeyMed agreement on trade in manufactured goods and find a positive but insignificant effect of this agreement on Turkey's imports from MENA countries. For agricultural trade, however, the authors find a significant effect amounting to an 89% increase in MENA exports to Turkey. TABLE 4 The Impact of FTAs on MENA Exports of Final and Intermediate Goods to OECD countries Dependent Variable: MENA4 exports from the reporting importers Explanatory Variables Final Goods Intermediate Goods (1) (2) (3) (4) Eumed 0.384*** 1.499*** −0.004 −0.490*** (0.131) (0.156) (0.176) (0.113) Turmed 0.354* 0.214 0.523*** 1.342*** (0.203) (0.585) (0.085) (0.320) Usamed 0.053 0.623*** 0.853*** 0.214** (0.171) (0.204) (0.302) (0.100) Jorfta 0.777*** 1.133*** 0.475 −0.618 (0.126) (0.385) (0.737) (0.444) RoO_EU −0.031 1.609*** 0.386** −0.129 (0.092) (0.525) (0.160) (0.327) Ln dist −2.100*** −1.160*** (0.070) (0.053) Contig −2.572*** 2.835*** (0.340) (0.318) Colony 0.336*** 0.933*** (0.104) (0.098) Lang 1.831*** 0.251*** (0.131) (0.090) Observations 17,283 17,525 13,740 13,820 R20.997 0.874 0.986 0.912 ImporterSectorTime FE Yes Yes Yes Yes ExporterSectorTime FE Yes Yes Yes Pairsector FE Yes Yes Notes: Robust standard errors in parentheses. ***p<.01, ** p<.05, *p<.1. Errors clustered by id: exporterimportersector. ImporterSectorTime FE capture the inward multilateral resistance. ExporterSectorTime FE capture the outward multilateral resistance. Pairsector FE are exporterimportersector FE. | 1517 CARDOZO et al. This finding might be explained by the fact that MENA exports to Turkey had not been subject to tariff reductions before the Barcelona Process, unlike exports to Europe. Thus, the TurkeyMed agreements might have prompted Turkish importers to source goods from MENA countries that had not been competitive in terms of prices prior to the agreement. Another reason might be the SouthSouth nature of this trade agreement. Even though Turkey clearly exhibits a higher level of development than the Med4, it would seem reasonable to expect that these countries are more similar in terms of their economic structure than most European countries are to the Med4. Similarities among countries could make them rather unnatural trading partners due to a lack of complementary of trading schemes, as suggested by Magee (23). However, similarities could also help to ensure favourable conditions when negotiating trade agreements with countries of a similar development level (United Nations Conference on Trade and Development, 42). Moreover, the fact that the effect is larger and more robust for trade in intermediate goods might be explained by the fact that these goods may have a higher elasticity of substitution than final goods. According to Chaney (14), the demand for a given good is relatively insensitive to changes in trade costs if this good is highly differentiated and thus has a low elasticity of substitution. If one assumes that the intermediate goods imported by Turkish firms are relatively undifferentiated and can be easily substituted, this could explain the large effect of the trade increase. Considering the impact of USA agreements with Jordan and Morocco, the coefficient is not significant in the case of final goods when controlling for all possible unobserved pair characteristics (column (1)), but it is significant and positive in the case of intermediate goods (column (4)). However, in contrast to the agreement with Jordan, tariffs on agricultural products are phased out over a far longer period in the USAMorocco agreement (see Section 2.4). Furthermore, the existence of the qualified industrial zones (QiZs) in Jordan definitely contributes to this large effect as these zones significantly increased trade between the two countries, which was probably reinforced by the existence of the FTA. Busse et al. (13) also find that the agreement between Jordan and the USA significantly boosted the exports of the former. As the effect for both trade agreements was estimated jointly, conclusive statements on the different effects cannot be made. MárquezRamos and MartínezZarzoso (24) investigate the impact of the USAMorocco FTA and find no significant effect of this agreement on final or on intermediate goods. An explanation for this result might be the fact that the agreement entered into force in 2006 and their data cover only the years until 2008. Parra et al. (25) also analyse the effect of this agreement using years up to 2010 and actually find a positive and significant effect on Egyptian exports of industrial and agricultural goods occurring 2years after the agreement entered into force. Parra Robles et al. (40), who examine the effect of the agreements the USA have concluded with Israel, Jordan and Morocco, find a positive and significant effect of these agreements on exports of industrial and agricultural goods with the effect being larger for manufactured products. In the case of the agreements that Jordan concluded with Singapore and Canada in 2005 and 2012, respectively, there is a positive and large impact on final exports (column (1), Table 3). Both agreements include liberalisation in agricultural and industrial goods, which might explain this large effect. For intermediate goods, no significant effect is detected for the specifications using pair fixed effects, while for the specification using the standard gravity variables a negative and highly significant effect is detected. In both cases, the coefficients from the standard gravity specification are larger than when using pair fixed effects. Our results diverge from those of Busse et al. (13), who estimate the effect of the SingaporeJordan FTA also using a PPML estimator and find a negative and significant effect of this agreement on Jordanian exports; however, they did not use sectoral MRTs. MárquezRamos and 1518 | CARDOZO et al. MartínezZarzoso (24) on the other hand, find that this FTA has a positive and significant effect on exports of manufactured goods but find a negative effect on agricultural exports. However, results are not strictly comparable to the ones shown in this paper, as the impact of FTAs between Canada and Jordan is not included in the abovementioned studies. This is indeed not surprising given that the agreement is quite recent. According to the World Trade Organization (35), there is, however, a high degree of trade complementarity between Canada and Jordan, such that it is expected that the agreement will lead to trade creation rather than causing trade diversion. This has also been confirmed in simulations predicting that Jordan's exports of apparel and accessories, vegetables as well as a number of chemical and mineral products will increase due to this agreement. Regarding the effect of the more flexible RoO that apply for trade among the Med4, Turkey, the EU and EFTA states, no significant effect is found for either type of good that is robust across specifications. Only for final goods do we find a positive and significant effect for the specifications without pair fixed effects that are of similar magnitude for both samples and exceed the effect of the EuroMed trade agreement per se. The finding that the changed RoO had a higher effect than the tariff reduction is in line with MárquezRamos and MartínezZarzoso (24), who find a positive and significant effect of the changed RoO on MENA exports of final goods. Bensassi et al. (11) also find that the effect of the more flexible RoO is greater than the effect of the simple tariff reduction. Unfortunately, the latter study does not distinguish between final and intermediate goods. The fact that the effect of the RoO on final goods is significant only for the specification without fixed effects might be explained by the lack of sufficient variation in this variable. In summary, as suggested by the theory, timeinvariant pair fixed effects seem to be a better proxy for trade costs than the standard gravity variables. Regarding the effect of the trade agreements on the imports of MENA trade partners, it is evident that each agreement significantly affects either final or intermediate goods, if the most accurate specifications in columns (1) and (4) are considered. This indicates the importance of distinguishing between these types of goods in economic analyses as they display different dynamics. Finally, with respect to trade in agriculture, results show that the agreements including liberalisation in agriculture lead, as expected, to a larger increase in imports when the percentage change is considered. This underlines the often stated importance of increasing market access for MENA agriculture goods in European markets. In general, however, it can be said that all the agreements considered are found to significantly increase imports of the MENA trade partners. In this context, the positive effect of the EuroMed agreements is surprising, as similar analyses have found that these agreements do not affect MENA exports. Clearly, the use of a different estimation method is one factor explaining this result. To investigate whether this promising result holds, the following section performs a robustness check using intervals rather than the full set of years, as a number of authors state that it is more appropriate to use intervals. 4.1 | Robustness: Impact for OECD countries, for different time intervals and by sector As additional robustness checks, the model for final and intermediate goods is first estimated for a subsample of OECD countries, then using intervals of 3 and 5years, and finally for specific sectors. These results are available upon request. The results for OECD countries are presented in Table 4. Even though using a slightly reduced sample of countries, which in any case includes | 1519 CARDOZO et al. the partners of all FTAs considered, should not influence the estimates much, it matters for the TurkeyMed agreements and the changes in RoO. For the other agreements, however, there is no significant difference in the coefficients. The approach for the second robustness check is chosen as trade agreements are typically phased in over a period of 5– 10years and it is thus not possible to capture their full economic effect in singleyear observations (Baier & Bergstrand, 7). Furthermore, according to these authors, as the economic effects of an FTA include a change in the terms of trade that tend to have lagged effects on the volume of trade, it is reasonable to expect that trade agreements might still have an effect on trade 10years after the agreement entered into force. To account for this fact, it is common in the trade literature to estimate the gravity model using data for intervals rather than for data pooled over consecutive years. Following the recommendation in Yotov et al. (43), we chose to estimate the model using 3and 5year intervals. The results in Table A2 (in the Appendix) show that the effect of the relevant trade agreements is robust across the different intervals for the EuroMed agreements and Jordan's FTAs. This could indicate that the tariff eliminations that occurred directly when these agreements entered into force matter the most. Furthermore, the results indicate that the effect of the tariff eliminations is persistent (at least for time horizon considered here). However, as every third and fifth year was kept in the dataset regardless of when the single trade agreements entered into force, the explanations provided are more like educated guesses as it could be the case that the trade agreement dummy switches to one at the start of an interval, when the agreement has actually already been in force for 1 or 2years. If one were interested in specifically investigating the effect after a certain number of years following the entry into force, lags would have to be included for the single dummies. However, as the main interest of this paper does not lie in investigating the possible delay in the impact of trade agreements, it was deemed sufficient to investigate the robustness of the estimates as depicted. As before, the coefficients for the USAMed agreements are significant for trade in intermediate goods, but the coefficient obtained for the 3year intervals is of much larger magnitude. This finding might be explained by the fact that for some goods that highly benefited from the trade agreement, tariffs were not eliminated directly when the agreement entered into force. For the TurkeyMed agreements and the changes in RoO, the time interval considered matters, as the coefficients display a different pattern of significance. For the TurkeyMed agreement and trade in intermediate goods, the coefficient loses significance for the 3year intervals but is significant in the case of 5year intervals. This could indicate that the effect of the initial tariff elimination is not persistent, whereas additional tariff eliminations that occur 5years after the agreement entered into force significantly impact trade. For the changes in RoO, the estimation results using intervals are more promising than the results presented in the preceding section. The results for final goods clearly show the relevance of accounting for adjustments over time as this policy change is found to greatly increase imports when 5year intervals are considered. This result seems reasonable, especially if one considers the emergence of production networks, as companies cannot adjust their production processes in a short timeframe. However, as the coefficient is compared to the other coefficients of much higher magnitude, it might be the case that the 5year intervals overestimate the effect of this policy instrument. On the other hand, since other studies find that the changes in the cumulation regime have a much greater effect than the trade agreements per se, this strong effect might actually confirm this finding. For intermediate goods, on the other hand, a significant effect is found only for the 3year period. As in the case of final goods, it is plausible to expect that it takes more than 1year for MENA exporters to change their source of inputs. The finding that the positive effect did not persist 1520 | CARDOZO et al. might be explained by the fact that the new input sources were less efficient in terms of quality, which in the longer run negatively affected exports. Although not significant, the negative coefficient found for the 5year intervals could support this explanation. Summarising, it can generally be said that the effect of the trade agreements is robust, although there are differences for the TurkeyMed agreements and the RoO. Finally, the effect of the different trade agreements has been estimated for single subsectors to account for the expected heterogeneity between them (See Tables A4 and A5 in the Appendix, for subsectors belonging to final and intermediate goods respectively). It seems that sector 6, consumer goods, is mainly driving the results for final goods, whereas sectors 1 and 2, food and beverages and industrial suppliers, are driving the results for intermediate goods. 5 | CONCLUSION This paper investigates the impact of a number of trade agreements on the exports of final and intermediate products of four MENA countries – Egypt, Jordan, Tunisia and Morocco – during the period from 1995 to 2016. Unlike the other MENA countries, the analysed countries are relatively well integrated into the global economy considering the number of agreements concluded. Nevertheless, they still lag behind in terms of economic development and would thus benefit from the growth potential generated by increasing exports. To examine the extent to which these agreements have been successful and to consistently identify the impact of the relevant agreements, the analysis employed a gravity model specification that includes a rich set of fixed effects to control for endogeneity and heterogeneity at country and sector level. This specification was then estimated using the PPML estimator to account for the large share of zero trade flows in disaggregated data. The estimation results show that the concluded agreements have been overall successful in increasing exports of the Med4. In particular, the positive impact of the EuroMed agreements is promising as the majority of studies that analyse the impact of these agreements also using disaggregated data find no positive impact on MENA exports. Furthermore, the results indicate that it is important to distinguish between intermediate and final goods, as it was found that the agreements had an impact either on final or on intermediate goods. In this context, the results show that the EuroMed agreements as well as Jordan's FTAs with Canada and Singapore have been successful in boosting exports of final goods. The respective effects amount to an increase in exports of 46% and 130% respectively. For intermediate goods, the TurkeyMed and USAMed agreements were found to have increased exports of the Med4 by 94% and 119% respectively. These effects show that the agreements including liberalisation in agricultural goods do indeed have a greater effect than those including only industrial goods. This is in line with expectations, as MENA countries clearly have a comparative advantage in these products. With respect to the effect of the PanEuropean RoO, it was found that it takes several years until they actually affect exports. This result is surprising as other studies have found that the effect of this change was instantaneous and bigger than the impact of the trade agreements per se. Regarding the impact on the different sectors, it was found that mostly undifferentiated goods profited from the tariff liberalisation. These findings are in line with the theories indicating that goods that are easily replaced by others – in other words, that are highly substitutable – are also more sensitive to tariff changes. Additionally and most interestingly, a positive effect of the EuroMed agreements was found on exports of food and | 1521 CARDOZO et al. beverages. Thus, the agricultural concessions granted by the EU and EFTA seem to have been successful in improving MENA market access to Europe. From these findings, a number of implications arise for policy making and further research. The results presented in this paper clearly support the need to expand the liberalisation of trade in agriculture with European partners as it was found that the existing concessions in the EuroMed and EFTAMed agreements benefit MENA exports of food and beverages. This implication is underlined by the fact that the agreements including trade in agriculture are, in line with other studies, found to have a greater impact on MENA exports than those covering only industrial goods. With respect to future research, there is a need for more evidence on the effects of MENA trade integration for specific countries using PPML estimation, as the results for the EuroMed agreements and the changes in RoO differ with respect to those obtained previously. Furthermore, as the chosen classification method distinguishes between goods only very broadly, it could be interesting to take a closer look at the goods included in the sectors to identify a more specific direction for export promotion at the sectoral level. Most importantly, it would be interesting to investigate whether improved export performance actually leads to economic growth. This is especially true for the EuroMediterranean Association Agreements that aim to generate stability through openness and the related economic success. ACKNOWLEDGMENTS We would like to thank the anonymous referees and the editors for their helpful comments and suggestions. Inmaculada MartínezZarzoso is also grateful to the financial support received from Grant PID2020114646RBC42 funded by MCINAEI/ 10.13039 / 501100011033 and from projects project UJIB202057 (University Jaume I) and PROMETEO2018/108. ORCID Adriana Cardozo https://orcid.org/0000-0003-4635-0287 Inmaculada MartínezZarzoso https://orcid.org/0000-0002-3247-8557 REFERENCES Abedini, J., & Péridy, N. (2008). The Greater Arab Free Trade Area (GAFTA): An estimation of its trade effects. Journal of Economic Integration, 23(4), 848– 872. Anderson, J. E., & Van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle. American Economic Review, 93(1), 170– 192. Anderson, J. E., & Van Wincoop, E. (2004). Trade costs. Journal of Economic Literature, 42(3), 691– 751. https://doi. org/10.1257/00220 51042 177649 Athukorala, P.- C., & Yamashita, N. (2006). Production fragmentation and trade integration: East Asia in a global context. The North American Journal of Economics and Finance, 17(3), 233– 256. Augier, P., Gasiorek, M., & LaiTong, C. (2004). Rules of origin and the EUMed partnership: The case of textiles. World Economy, 27(9), 1449– 1473. Bacchetta, M., Beverelli, C., Cadot, O., Fugazza, M., Grether, J.- M., Helble, M., Nicita, A., & Piermartini, R. (2012). A practical guide to trade policy analysis. World Trade Organization. Baier, S. L., & Bergstrand, J. H. (2007). Do free trade agreements actually increase members’ international trade? Journal of International Economics, 71(1), 72– 95. https://doi.org/10.1016/j.jinte co.2006.02.005 Baier, S., Bergstrand, J. H., & Clance, M. W. (2018). Heterogeneous effects of economics integration agreements. Journal of Development Economics, 135, 587– 608. Baier, S. L., Yotov, Y. V., & Zylkin, T. (2019). On the widely differing effects of free trade agreements. Journal of International Economics, 116, 206– 226. https://doi.org/10.1016/j.jinte co.2018.11.002 1522 | CARDOZO et al. Baldwin, R., & Taglioni, D. (2007). Trade effects of the euro: A comparison of estimators. Journal of Economic Integration, 22(4), 780– 818. Bensassi, S., MárquezRamos, L., & MartínezZarzoso, I. (2012). Economic integration and the two margins of trade: The impact of the Barcelona process on North African countries’ exports. Journal of African Economies, 21(2), 228– 265. Bergstrand, S., Sunesen, E., & Thelle, M. (2011). Ex post Assessment of six EU Free Trade Agreements, An econometric assessment of their impact in trade. DG Trade. European Commission. Busse, M., Gröning, S., & Groening, S. (2012). Assessing the impact of trade liberalization: The case of Jordan. Journal of Economic Integration, 27(3), 466– 486. Chaney, T. (2008). Distorted Gravity: The intensive and extensive margins of international trade. American Economic Review, 98(4), 1707– 1721. Cieślik, A., & Hagemejer, J. (2009). Assessing the impact of the EUsponsored Trade Liberalization in the MENA Countries. Journal of Economic Integration, 343– 368. https://doi.org/10.11130/ jei.2009.24.2.343 De MelloSampayo, F. (2009). Competingdestinations Gravity Model: An Application to the Geographic Distribution of FDI. Applied Economics, 41(17), 2237– 2253. De MelloSampayo, F. (2017). Competingdestinations Gravity Model Applied to Trade in Intermediate Goods. Applied Economics Letters, 24(19), 1378– 1384. https://doi.org/10.1080/13504 851.2017.1282109 Egger, P. H., & Lassmann, A. (2012). The language effect in international trade: A metaanalysis. Economics Letters, 116(2), 221– 224. Feenstra, R. C. (2004). Advanced International Trade. Theory and Evidence. Princeton University Press. Freund, C., & PortugalPerez, A. (2013). Assessing MENA’s trade agreements. In E. M. Gasiorek (Ed.), The Arab spring implications for economic integration (pp 137– 157). A VoxEU.org Book. Retrieved from https://voxeu. org/conte nt/arabsprin gimpli catio nsecono micinteg ration Head, K., & Mayer, T. (2014). Gravity Equations: Workhorse, Toolkit, and Cookbook. In Handbook of International Economics, Vol. 4 (pp. 131– 195). Elsevier. International Monetary Fund (2016). Economic diversification in oilexporting Arab countries. In Annual meeting of Arab ministers of finance. Retrieved from https://www.imf.org/en/Publi catio ns/Polic yPaper s/Issue s/2016/12/31/Econo micDiver sific ation - inOilExpor tingArabCount riesPP5038 Kepaptsoglou, K., Karlaftis, M. G., & Tsamboulas, D. (2010). The gravity model specification for modeling international trade flows and free trade agreement effects: A 10year review of empirical studies. The Open Economics Journal, 3(1), 1– 13. Larch, M., Wanner, J., Yotov, Y. V., & Zylkin, T. (2017). The Currency Union Effect: A PPML Reassessment with HighDimensional Fixed Effects. Cesifo Working Papers. https://www.cesifo.org/en/publi katio nen/2017/ worki ngpaper/ curre ncyunion - effec tppmlreasses sment - highdimen siona lfixed Magee, C. S. (2003). Endogenous preferential trade agreements: An empirical analysis. Contributions in Economic Analysis & Policy, 2(1), 1– 19. MartínezZarzoso, I. (2013). The log of gravity revisited. Applied Economics, 45(3), 311– 327. MárquezRamos, L., & MartínezZarzoso, I. (2014). Trade in intermediate goods and EuroMed production networks. Middle East Development Journal, 6(2), 215– 231. https://doi.org/10.1080/17938 120.2014.961327 Parra Robles, M. D., MartínezZarzoso, I., & Suárez Burguet, C. (2016). The impact of FTAs on MENA trade in agricultural and industrial products. Applied Economics, 48(3), 2341– 2353. Parra, M., MartinezZarzoso, I., & SuàrezBurguet, C. (2016). The impact of FTAs on MENA trade in agricultural and industrial products. Applied Economics, 48(25), 2341– 2353. https://doi.org/10.1080/00036 846.2015.1119792 Péridy, N. (2005a). Toward a PanArab Free Trade Area: Assessing trade potential effects of the Agadir Agreement. The Developing Economies, 43(3), 329– 345. Péridy, N. (2005b). The trade effects of the EuroMediterranean partnership: What are the lessons for ASEAN countries? Journal of Asian Economics, 16(1), 125– 139. https://doi.org/10.1016/j.asieco.2004.12.001 Pöyhönen, P. (1963). A tentative model for the volume of trade between countries. Weltwirtschaftliches Archiv, 90(1), 93– 99. Rouis, M., & Tabor, S. (2013). Regional economic Integration in the Middle East and North Africa: Beyond Trade Reform. Directions in development: Trade. World Bank. © World Bank. Retrieved from http://hdl.handle. net/10986/ 12220 | 1523 CARDOZO et al. Shepherd, B. (2012). Alternative Gravity Model Estimators. In The Gravity Model of International Trade: A User Guide. United Nations Economic and Social Commission for Asia and the Pacific (ESCAP). Silva, J. S., & Tenreyro, S. (2006). The log of gravity. The Review of Economics and Statistics, 88(4), 641– 658. https:// doi.org/10.1162/rest.88.4.641 The Economic Complexity Observatory (2017). The Economic Complexity Observatory: An Analytical Tool for Understanding the Dynamics of Economic Development. Workshops at the TwentyFifth AAAI Conference on Artificial Intelligence. http://atlas.media.mit.edu/en/ Tinbergen, J. (1962). An analysis of world trade flows. Shaping the World Economy, 5(1), 1– 117. https://doi. org/10.1002/tie.50600 50113 United Nations Conference on Trade and Development (2007). Regional cooperation for development. Trade and development report. United Nations. Wood, C. A., & Yang, J. S. (2016). MENA export performance and specialization– the role of financial sector development and governance. World Bank Policy Research Working Paper. https://openk nowle dge.world bank.org/ handl e/10986/ 24157 World Bank (2010). Mirror Data with UN COMTRADE. https://wits.world bank.org/wits/wits/witsh elp/Conte nt/ Data_Retri eval/T/Intro/ B2.Impor ts_Expor World Trade Organization (2012). The Potential Economic Effects of FTA between Jordan and Canada. Technical report. World Trade Organization. Yotov, Y. V., Piermartini, R., Monteiro, J.- A., & Larch, M. (2016). An advanced guide to trade policy analysis: The structural gravity model. World Trade Organization. Zylkin, T. (2017). PPML_PANEL_SG: Stata module to estimate "structural gravity" models via Poisson PML. How to cite this article: Cardozo, A., MartínezZarzoso, I., & Vogler, P. L. (2022). The impact of free trade agreements on Middle East and North Africa exports of intermediate and final goods. The World Economy, 45, 1501– 1527. https://doi.org/10.1111/twec.13214 1524 | CARDOZO et al. APPENDIX TABLE A1 Classification by Broad Economic Categories (BEC) 1 – Food and beverages 11 – Primary (I) (111: Mainly for industry, 112: Mainly for household consumption) 12 – Processed (F) (121: Mainly for industry, 122: Mainly for household consumption) 2 – Industrial supplies not elsewhere specified 21 – Primary (I) 22 – Processed (I) 3 – Fuels and lubricants 31 – Primary (I) 32 – Processed (I) 4 – Capital goods (except transport equipment), and parts and accessories thereof 41 – Capital goods (except transport equipment) (C) 42 – Parts and accessories (I) 5 – Transport equipment and parts and accessories thereof 51 – Passenger motorcars (NC) 52 – Other (F) (521: Industrial, 522: Nonindustrial) 53 – Parts and accessories (I) 6 – Consumer goods not elsewhere specified 61 – Durable (F) 62 – Semidurable (F) 63 – Nondurable (F) 7 – Goods not elsewhere specified (NC) Source: United Nations Statistics Division (2017), Manual of the Fifth Revision of the BEC, page 8. https://unsta ts.un.org/unsd/ trade/ class ifica tions/. (I) denotes intermediate, (F) final, (C) capital goods and (NC) not classified goods. TABLE A2 Overview of the FTAs and RoO FTA Country Partners Entry Force Full Liberal RoO EUMED Tunisia EU countries 1998 2009 PanEuropean diagonal and Full cum. Morocco 2000 2010 Jordan 2002 2014 Tun, Mor, Egypt 2004 2016 USAMED Jordan USA 2001 2010 Full cum Tun, Mor Morocco 2006 2020/2026 CJFTA Jordan Canada 2012 2017 Bil cum TURMED Tunisia Turkey 2005 2014 PanEuropean diagonal cum. Morocco 2006 2015 Egypt 2007 2020 Jordan 2011 2018 JSGPFTA Jordan Singapore 2005 2015 Bil cum Note: CJFTA and JSGP are considered together in the empirical analysis under the name Jorfta.