The Trade Impact of EU Tariff Margins: an Empirical Assessment

social sciences $€ £ ¥

Article The Trade Impact of EU Tariﬀ Margins: An Empirical Assessment

Maria Cipollina 1,* and Luca Salvatici 2 1 Department of Economics, University of Molise, 86100 Campobasso, Italy 2 Department of Economics, University of Roma Tre, 00145 Rome, Italy; [email protected] * Correspondence: [email protected]; Tel.: +39-0874-404338

Received: 31 May 2019; Accepted: 9 September 2019; Published: 12 September 2019

Abstract: This article provides an assessment of how the EU trade policies affect EU imports. The main contribution is that we compute a theoretically consistent measure of the EU tariff margin and estimate the elasticities of substitution at the sectoral level, using a structural gravity model that includes domestic trade flows. Our analysis is related to the most recent gravity literature and the identification strategy is based on the existence of a sufficient variation of the tariffs applied by the EU to different markets of origin. We use cross-section data (more than 5000 tariff lines and 188 exporters, including the EU28 Member States, in the year 2017), to obtain structural gravity estimates of trade substitution elasticities. Since tariffs greatly differ by product, an in-depth analysis should take place at the tariff line. Moreover, we use the information provided by the Eurostat Comext database on the tariff regime of imports, so we distinguish the Most Favored Nation (MFN) from the preferential trade flows. The estimated elasticities can be used to calculate the counterfactual change in total EU imports that would follow either from the removal of trade preferences or from the removal of trade policies.

Keywords: gravity model; European trade policy; bilateral tariﬀ margin; preferential trade agreements

1. Introduction The main goal of this paper is to assess the impact of EU trade policies on imports to the EU. We consider a sample of 188 countries, including intra-EU trade flows, and discuss the most relevant issues, from a methodological point of view, that are important for analyzing the trade-creating impacts of trade policies. The analysis is based on the gravity model, starting from the theoretical model (see Anderson 1979; Anderson and Wincoop 2003), and we compute an explicit measure of the bilateral tariff margin at a disaggregated level. We focus on EU trade for several reasons. It is the biggest player in the world trade even if tariffs are still significant in specific sectors, e.g., agriculture, though over time a large number of preferential trade agreements have been concluded between EU and many developing countries. Our paper refers to recent studies using the gravity model that argue that regressions should be estimated with data that include also intra-national sales (Heid et al. 2017; Feenstra et al. 2018). Our paper also refers to the recent literature that use updated econometric methods to estimate the elasticity of substitution using disaggregated data (Baier and Bergstrand 2001; Caliendo and Parro 2015; Romalis 2007; Imbs and Isabelle 2009; Corbo and Osbat 2013). This analysis contributes to the strand of the related literature in two ways. The first contribution is the estimation of the impact of tariff margins in a theoretically grounded gravity model that includes domestic trade flows. In particular, we measure the bilateral tariff margin as the advantage or disadvantage of each exporter with respect to other competitors, for disaggregated data producers (as defined by Low et al. 2009; Cipollina et al. 2017) and argue that the intra-national trade is indeed necessary to properly assess the trade policies impact. The second contribution of our analysis is to define an empirical strategy that considers both

Soc. Sci. 2019, 8, 261; doi:10.3390/socsci8090261 www.mdpi.com/journal/socsci Soc. Sci. 2019, 8, 261 2 of 15 international and intra-national trade data, and this implies that the tariff margin computation takes into account the domestic competitors and the fact that intra-EU trade flows are duty-free. In particular, we argue that there is not such a thing as a non-discriminatory trade policy: Trade policies introduce additional costs but these costs only matter in relative terms, i.e., with respect to the costs paid by each exporter to reach the importing markets. The main reason for including both domestic and foreign goods is represented by the possibility to assess the effects of trade policies taking into account trade creation as well as trade diversion impact. Therefore, we attempt to assess both the liberalization and protectionist nature of EU trade policies by measuring the effects of multilateral, bilateral, and unilateral agreements on international trade relative to intra-EU trade. The empirical strategy developed in this article is based on the variation of margins across 5388 tariff lines and 188 exporters (including the EU28 Member States) to the EU market in the year 2017. Working at a disaggregated level involves a very large number of observations that raises the problem of zeros in trade flows and leads to a computational problem. Therefore, we opt to take advantage of the cross-section dimension, rather than panel since trade policies are likely to vary across products and exporters. Specifically, using data cross-section we obtain structural gravity estimates of trade substitution elasticities taking into account all available information available regarding (1) the preference utilization, since we distinguish Most Favored Nation (MFN) and preferential trade flow; (2) domestic and international trade flows. Finally, the estimated elasticities are used to calculate the counterfactual change in total EU imports that would follow either from the removal of trade preferences or from the removal of trade policies. The article is structured as follows. Section2 provides a brief literature review; Section3 introduces the theoretical gravity model and discusses the empirical methodology; Section4 includes data and descriptive analyses; Section5 presents and discusses the results; finally, Section6 concludes.

2. A Brief Literature Review Since the early 1960s, the European Union (EU) is actively engaged in a wide range of negotiations to strengthen existing measures within the strategic partnership, especially with developing countries, in order to enhance their integration in the world trading system and to promote a process of economic development and industrialization. The EU’s trade policy is part of Strategic Plan 2016–2020 with the aim of boosting employment and creating a more sustainable economy (European Commission 2016). Trade preferences are the common instrument of trade policy. They include reduction or, in many cases, elimination of tariff barriers on imports from beneficiary countries. Nowadays, most developing countries can export to the EU with preferential market access under various preferential schemes. This paper is most closely related to the more recent literature testing the impact of EU preferential agreements on trade volume using highly disaggregated data (Cipollina et al. 2017; Scoppola et al. 2018). Even if the expectation of the positive impact of preferences on trade is by far and large confirmed, such impact is affected by the presence of complex rules that often accompany preferential schemes. The higher the preferential margin, the higher the probability should be that preference is used, but for various reasons not all imports of products that are nominally eligible for preferential treatment enter the granting country at the preferential rate. Costs related to fulfilling rules of origin requirements, and other formalities that can be specific to each shipment are often attached to using a preference, so that preferences may not be used unless volumes are important enough to result in substantial duty savings. Furthermore, the complexities of rules of origin are part and parcel of all preferential agreements. As a result, available preferences are not always fully utilized. Although preferences might be considered rather generous, other complex rules (including non-compliance with the relevant rule of origin) are an important obstacle for exporters of goods (De Melo and Nicita 2018). Since we do not know the utilization rates of different schemes, we used the available information on tariff regime (MFN or preferential flow) and applied tariff to each trade flow accounting that the importing country will apply the MFN tariff if the product fails to meet the country’s rules that determine the product’s country of origin. Soc. Sci. 2019, 8, 261 3 of 15

With respect to the recent literature, our research attempts to examine the trade impacts of EU policies accounting for intra-national sales. Recent works (Heid et al. 2017; Feenstra et al. 2018) using the gravity model, argue that regressions should be estimated with data that include not only international trade but also intra-national sales. There are three tiers of aggregation in the CES framework (Feenstra et al. 2018): (i) The disaggregation is across goods in the upper-tier; (ii) across home and foreign products in the middle-tier; (iii) and across foreign sources in the lower-tier. Feenstra et al. (2018) call their middle-tier elasticity (across home and foreign products) the “macro” elasticity, while they call their lower-tier elasticity (across foreign source countries) the “micro” elasticity. Evaluating the diﬀerence between the elasticity of substitution between home and foreign goods and between varieties of foreign goods, requires two ingredients: A model that allows for such a nested CES structure and a data set that has both home and foreign supplies at exactly the same level of disaggregation. Feenstra et al.(2018) provide both ingredients. Cipollina et al.(2017) focus on elasticity among foreign suppliers. In this article, we estimate the macro elasticity across home and foreign products, and therefore the few results provided in the literature are not readily comparable with ours.

3. Structural Gravity Model

3.1. The Gravity Equation Gravity models utilize the gravitational force concept as an analogy to explain the volume of trade (Tinbergen 1962). The ability to correctly approximate bilateral trade ﬂows makes the gravity equation one of the most successful “empirical fact” in the international trade literature. Following Anderson(1979) and Anderson and Wincoop(2003) theoretically grounded gravity model, we start from the well-known gravity equation:

σ k − Pi imk = αkMk , (1) i i (1 σ) Πk −

k k k k where αi is the consumer preference parameter, M is the expenditure on import, Π and Pi are, respectively, the product k import price index computed across all exporters i and the domestic price of imported good k from the country i, while σ represents the elasticity of substitution between varieties and it is greater than 1. The domestic price is given by: k k k k Pi = pi ci 1 + ti , (2) k k with pi representing the fixed free-on-board (FOB) export price of a physical unit, ci > 1 capturing k the transport costs that differ by destination and product, and ti that is the bilateral applied ad valorem tariff. In the spirit of Cipollina et al.(2017), the assumption of the separation of tari ffs from other trade cost components allows the computation of the price index Πk as a weighted average tariff factor 1 + Tk applied to product k computed as a CES aggregator:

1   1 σ X ( ) − k  k k k k 1 σ  k Π =  α p c 1 + t −  = 1 + T . (3)  i i i i  i

In our model, the Equation (3) is crucial because for each product it measures the overall incidence of the EU’s trade policies. Furthermore, being aggregated across all exporters, it also represents an explicit measure of the multilateral trade resistance. Substituting Equation (1) becomes: Soc. Sci. 2019, 8, 261 4 of 15

σ k + k − pi 1 ti imk = αkMk , (4) i i (1 σ) (1 + Tk) − k k and defining the bilateral tariff margin (btmi ) as the ratio between the reference tariff factor (1+T ) and the applied tariff factors faced by each exporter: 1 + Tk k btm = , (5) i + k 1 ti we can estimate the following gravity equation:

  σ αkMk pkβ γk − k i  i i  im =   . (6) i (1 + Tk)  k  btmi

The bilateral tariff margin depends both on the higher rates and on the share of exporters paying those rates, then incorporates the “multilateral nature” of trade policies. The main idea is that the bilateral tariff margin is not confined to the country-specific structure of tariffs but it depends on the whole structure of applied tariffs. Therefore the “reference tariff” used to compute it takes into account the competitive advantage, or disadvantage, that each exporter faces with respect to other competitors. In this article, we compute the actual tariff margin in relative terms and on a bilateral basis taking into account the multilateral nature of trade policies. When the reference tariff is lower than the applied, the margin is between “0” and “1” and signals the existence of negative margins, depending on the disadvantage of the country with respect to other competing exporters.

3.2. Econometric Approach Working at a disaggregated level, the presence of many zero trade flows create obvious problems in the log-linear form of the gravitational equation. These zeros may be the result of rounding errors, that are more likely to occur for small or distant countries and, therefore, the probability of rounding down will depend on the value of the covariates, leading to the inconsistency of the estimators. Many “zeros” may also be due to missing observations which are wrongly recorded as zero. This problem is more likely to occur when small countries are considered and, again, measurement error will depend on the covariates. There has been a long debate concerning what is the best econometric approach in order to avoid the bias that would be implied by eliminating the observations with zero flows. Tobit models rely on rather restrictive assumptions that are not likely to hold since the censoring at zero is not a ‘simple’ consequence of the fact that trade cannot be negative: Zero flows, as a matter of fact, do not reflect unobservable trade values but they are the result of economic decision making based on the potential profitability of engaging in bilateral trade at all. Zero trade flows could be the result of economic decisions based on the potential profitability of engaging in bilateral trade at all, so they should be treated properly in order to overcome a sample selection problem. Several authors consider the Heckman two-step estimator (Heckman 1979) as the best procedure (Helpman et al. 2008; Martin and Pham 2016). Nevertheless, the bilateral trade flows are collected from multiple countries and heteroskedasticity may be a challenge especially in the common practice of logarithmic transformation. Recent empirical analyses argue that because of the presence of heteroskedasticity, gravity type models should be estimated in multiplicative form and recommend maximum likelihood estimation techniques based on the Poisson specification of the model suggested by Silva and Tenreyro(2006). They showed that the estimation of the gravity model by the Pseudo Poisson Maximum Likelihood (PPML) specification are consistent in the presence of heteroskedasticity and are reasonably efficient, especially in large samples. Soc. Sci. 2019, 8, 261 5 of 15

As recent empirical analyses (Fally 2015; Yotov et al. 2016) we estimate the gravity model in multiplicative form, using the PPML estimator, for the following reasons:

- We could not accept the null hypothesis of homoskedasticity, and the PPML estimator is generally well behaved even when the conditional variance is far from being proportional to the conditional mean; - The fact that the dependent variable has a large proportion of zeros does not aﬀect the performance of the estimator; - Estimation of gravity with PPML including exporter and product (HS6 digit), that enables to control for any other observable and unobservable characteristics that vary over exporters and sectors, respectively, is consistent with the ‘structural approach’ to gravity analyses.

Accordingly, we estimate the following regression n o k k γ k imi, = exp σln btmi + EXPi + δPRODHS6 + εi . (7)

Note that the fixed effects absorb all country-specific and time-invariant controls (such as GDP, distance, colonial status, and common language) usually included in gravity estimations. Since they are not of particular interest here, it is preferable to use the large set of fixed effects described above (as argued by Ornelas and Ritel 2018) that are also controls for all unobservable trade costs. We estimate regression (7) for 21 sectors defined according to the EU Sections of the Harmonized Commodity Description and Coding System (the full list of the commodity classification is available from the authors upon request, or for more detail see: https://trade.ec.europa.eu/tradehelp/eu-product-classification-system). For each sector, we obtain an estimated coefficient of the elasticity of substitution, σˆ, allowing elasticities to differ across HS sections. In a first step, we estimate our gravity Equation (7) using a relative bilateral tariff margin, computed taking the applied MFN duty as reference tariff. Once estimating each parameter σˆ, we compute the CES tariff margins. Then, we iterate the process (in the spirit of Head and Mayer 2014; and Cipollina et al. 2017) to obtain a new set of elasticities with the CES margins, we stop until the estimated coefficient does not change at the second-decimal digit. Finally, following Lai and Zhu(2004) we consider two possible scenarios computing the percentage change due to the hypothetical elimination of trade policies:

1. We remove the preferential policies setting all tariffs equal to “zero” and estimate the counterfactual EU imports. The difference between the counterfactual (i.e., free-trade) and predicted flows represents the trade decrease resulting from the protectionist impact of EU tariffs. When all tariffs are removed, the numerator (i.e., the reference tariff T) decreases faster than the denominator k (i.e., the bilateral tariff ti). This increases to 1 the pre fi values of the extra-EU countries facing k negative margins (i.e., pre fi < 1). Accordingly, their export to the EU will increase at the expense of intra-EU trade flows and exports originating from extra-EU countries facing positive margins k (i.e., pre fi > 1). The reduction in exports suffered by the preferred countries corresponds to the value of preference erosion in the case of unilateral liberalization by the EU. However, it is worth emphasizing that preference erosion is ubiquitous: Even the change of bilateral tariff will affect the CES aggregator modifying the margins of each exporter; 2. We remove the preferential policies setting all bilateral tariffs equal to the MFN rate and estimate the counterfactual EU imports. In this case, the difference between the counterfactual (i.e., non-discriminatory protection) and predicted flows represents trade flows that either would or would not take place without preferences. The former represents the trade diversion (Viner 1950). As far as the latter is concerned, unlike Cipollina et al.(2017), the CES reference tariff takes into account extra- as well as intra-EU trade. Hence, the trade flows generated by the EU trade preferences include both trade creation and trade diversion. Soc. Sci. 2019, 8, 261 6 of 15

Soc. Sci. 2019, 8, x FOR PEER REVIEW 6 of 15 4. Data Sources exporters*5388To perform products*1 the empirical year) analysis, since afor data each set exporter was built all including goods that intra-national are never EUexported trade flowsand (astherefore proxy fornot domesticproduced sales)were eliminated and international from the EU samp imports.le. Data We on trade estimate at the a panelHS6 level model, of detail covering are importstaken from of 5388 the commoditiesEurostat Comext based database on the (see WTO http definition,://fd.comext.eurostat.cec. from 188 countrieseu.int/xtweb/) (including theand EU) data to EU28on tariffs in the are year from 2017. TRAINS The number database of observations(see http://r0. isunctad.org/trains/) 1,006,180 rather than which 1,012,944 is integrated (1 importer into the188 × exporters*5388WITS software products*1 (see http://wits.worldbank.org/witsweb/default.aspx). year) since for each exporter all goods that are We never use exported the information and therefore on notpreferential produced and were non-preferential eliminated from (MFN) the sample. trade flow Datas, on provided trade at by the the HS6 Comext level of database, detail are and taken we from set thethe Eurostat level of duty Comext used database for the (seecomputation http://fd.comext.eurostat.cec.eu.int of the bilateral tariff margins/xtweb equal/) and to the data applied on tari MFNffs are tariff when the flow is registered as MFN trade and to the preferential applied tariff otherwise. In this from TRAINS database (see http://r0.unctad.org/trains/) which is integrated into the WITS software way, we associate to each flow the duty effectively paid and focus on the flows that effectively benefit (see http://wits.worldbank.org/witsweb/default.aspx). We use the information on preferential and from the preferences. Figure 1 shows the percentage of imports associated with the tariff regime and non-preferential (MFN) trade flows, provided by the Comext database, and we set the level of duty intra-EU members for each sector. Most imports (64%) take place among EU members, whereas the used for the computation of the bilateral tariff margins equal to the applied MFN tariff when the flow rest is divided between duty-free imports (17%) and imports paying positive MFN duties (19%). The is registered as MFN trade and to the preferential applied tariff otherwise. In this way, we associate to section that accounts for the lowest share of intra-EU trade, Section XIV (natural and precious metals), each flow the duty effectively paid and focus on the flows that effectively benefit from the preferences. also shows the highest percentage of duty-free imports (64%). At the other extreme in Section I Figure1 shows the percentage of imports associated with the tari ff regime and intra-EU members for (animal products) the intra-EU trade share reaches 75% and only 2% benefits from duty-free each sector. Most imports (64%) take place among EU members, whereas the rest is divided between treatment. Sections as X (paper and paperboard and articles thereof) and XXI (works of art) enter in duty-free imports (17%) and imports paying positive MFN duties (19%). The section that accounts the EU market under an MFN duty-free regime. for the lowest share of intra-EU trade, Section XIV (natural and precious metals), also shows the In our analysis we exclude sections where trade policies depend on political rather than highest percentage of duty-free imports (64%). At the other extreme in Section I (animal products) the economic motivations, this is the case of excluding Section XIX (arms and ammunition), and all intra-EUsections tradewhere share there reaches is no room 75% for and preferences only 2% benefits because from all tariffs duty-free are equal treatment. to “zero”, Sections or sections as X (paper that andpresent paperboard trivial percentages and articles of thereof) preferential and XXItrade (works flows, of e.g., art) Sections enter in V the (mineral EU market products) under and an MFNXIV duty-free(natural and regime. precious metals).

100% 1 0 0 2 7 3 5 4 4 4 6 2 6 3 5 13 8 12 5 14 17 9 14 11 17 19 13 6 25 11 16 14 6 15 18 15 21 20 80% 10 10 16 14 12 0 2 2 4 17 20 27 29 22 19 27 3 14 49 23 60% 64 0 2 0 85 1

40% 83 79 75 76 72 77 73 74 73 70 64 61 58 61 58 63 56 56 51 20% 39 28 15 0%

INTRA-EU trade MFN duty-free flows Positive MFN tariff flows Preferential trade

FigureFigure 1.1. StructureStructure of ofEU28 EU28 imports imports by section. by section. Source: Source: Elaboration Elaboration on data by on COMEXT data by COMEXTand TRAINS; and TRAINS;2017. 2017.

Figure 2 shows the MFN and applied tariffs for the sections included in our analysis. We present both the simple and trade-weighted averages: Instances when the latter is higher than the former

Soc. Sci. 2019, 8, 261 7 of 15

In our analysis we exclude sections where trade policies depend on political rather than economic motivations, this is the case of excluding Section XIX (arms and ammunition), and all sections where there is no room for preferences because all tariffs are equal to “zero”, or sections that present trivial percentages of preferential trade flows, e.g., Sections V (mineral products) and XIV (natural and precious metals). Figure2 shows the MFN and applied tari ffs for the sections included in our analysis. We present Soc. Sci. 2019, 8, x FOR PEER REVIEW 7 of 15 both the simple and trade-weighted averages: Instances when the latter is higher than the former indicateindicate that that higher higher tari tariffsffsare areassociated associated withwith inelasticinelastic import demand curves. Only Only 15% 15% of of extra-EU extra-EU importsimports are are preferential preferential trade.trade. UsingUsing thethe MFNMFN ratesrates as a benchmark, the the most most protected protected sectors sectors are are animalanimal products, products, foodstu foodstuffsffs and and beverages, beverages, textiles textiles and and footwear footwear (Sections (Sections I, IV, XI,I, IV, and XI, XII). and Animals, XII). foodstuAnimals,ff, andfoodstuff, beverages and productsbeverages present products high present shares hi ofgh intra-EU shares of trade, intra-EU while trade, in the while case of in the the textile case sectorof the alltextile EU importssector all are EU subject imports to are positive subject duty. to positive duty.

MFN tariffs Bilateral applied tariffs

XX: Misc. Manufactured Articles 1.7 1.5 XIX:Arms 2.4 1.9 XVIII: Instruments 0.8 0.7 XVII:Transport 5.4 3.0 XVI: Machineries 1.2 1.0 XV: Metals 2.2 1.5 0.3 XIV: Pearls and Precious stones 0.2 XIII: Stone/Glass 4.2 3.2 11.2 XII: Footwear / Headgear 8.4 10.5 XI: Textiles 5.4 1.9 IX: Wood & Articles of Wood 1.5 4.3 VIII: Raw Hides, Skins, Leather, & Furs 3.1 5.0 VII: Plastics/Rubbers 3.5 2.4 VI: Chemicals & Allied Industries 1.9 0.4 V: Mineral Products 0.4 8.4 IV: Foodstuffs, Beverages and Tobacco 3.7 5.6 III: Fats & Oils 3.8 4.0 II: Vegetable Products 1.4 9.3 I: Animal & Animal Products 4.4 4.3 Overall 2.5

FigureFigure 2. 2.Tari Tariffsffs (%, (%, trade-weighted trade-weighted mean). mean). Source: Source: Elaboration Elaboration on on data data by COMEXTby COMEXT and and TRAINS; TRAINS; 2017. 2017. 5. The Results 5. The Results 5.1. Elasticities of Substitutions by Sections 5.1. TableElasticities1 reports of Substitutions the econometric by Sections results of Equation (7) for the 16 sections under investigation. ResultsTable 1 show reports very the di econometricfferent elasticities results across of Equation sectors, (7) almost for the all sections16 sections show under significant investigation. estimates, with the exception of Sections VIII (raw hides and skins, leather, fur skins, and articles thereof) and XII (footwear, headgear, umbrellas,Table 1. and Estimated other textileelasticities articles), of substitution XIII (ceramic by section. and glassware), XV (metals), XVIII (instruments and apparatus), and XX (miscellaneous manufactured articles). Sections VIII Sections Sigma (𝛔) N. of Obs. and XII seem to be characterized by an inelastic demand as confirmed by the observation that the I: Animal and Animal Products 5.2 *** 4.5798 trade-weighted average tariffII:s areVegetable higher Products than simple averages 19.5 (Sections*** 45.545 VIII and XII) and by a low share of preferential trade (SectionsIII: Fats XVIII and Oils and XX). The most 20.2 of imports*** of7.024 Sections XIII and XV are among EU countries andIV: Foodstuffs, preferences Beverages, seem to and be notTobacco relevant for9.4 *** the international 28.676 trade. VI: Chemicals and Allied Industries 3.3 ** 152.831 VII: Plastics/Rubbers 14.3 *** 39.659 VIII: Raw Hides, Skins, Leather, and Furs −0.9 12.471 IX: Wood and Articles of Wood 7.3 ** 19.865 XI: Textiles 12.8 *** 160.859 XII: Footwear/Headgear 1.4 9.001 XIII: Stone/Glass −0.8 24.829 XV: Metals 2.5 105.729 XVI: Machineries 14.1 *** 151.063

Soc. Sci. 2019, 8, 261 8 of 15

Table 1. Estimated elasticities of substitution by section.

Sections Sigma (σˆ ) N. of Obs. I: Animal and Animal Products 5.2 *** 4.5798 II: Vegetable Products 19.5 *** 45.545 III: Fats and Oils 20.2 *** 7.024 IV: Foodstuffs, Beverages, and Tobacco 9.4 *** 28.676 VI: Chemicals and Allied Industries 3.3 ** 152.831 VII: Plastics/Rubbers 14.3 *** 39.659 VIII: Raw Hides, Skins, Leather, and Furs 0.9 12.471 − IX: Wood and Articles of Wood 7.3 ** 19.865 XI: Textiles 12.8 *** 160.859 XII: Footwear/Headgear 1.4 9.001 XIII: Stone/Glass 0.8 24.829 − XV: Metals 2.5 105.729 XVI: Machineries 14.1 *** 151.063 XVII: Transport 14.8 *** 26.398 XVIII: Instruments 2.6 36.007 XX: Misc. Manufactured Articles 5.54 23.304 1 Robust standard errors, clustered by country-product, are shown in parentheses. Data refer to 2017. All specifications include exporter and product fixed effects. ** significant at 5 per cent level; *** significant at 1 per cent level. Source: Elaboration on data by COMEXT and TRAINS; Software Stata/SE 13.1.

In Sections II (vegetable products) and III (oils and fats), the estimates for the elasticity of substitutions reach quite a high value (respectively, 19 and 20) showing high substitutability among country varieties. This is also true for Sections VII (plastics), XVI (machinery), and XVII (transport equipment), with an estimated coefficient equal to 14. Indeed, from this perspective, the most differentiated sectors seem to be those where estimates are not significant. The few results provided in the literature are not readily comparable with ours since they only consider substitutability among foreign suppliers, because the observations on intra-EU trade are not included in the sample. However, as expected, our macro elasticities between home and import goods are smaller than the micro elasticity between foreign sources of imports. Table2 shows the structure of EU imports based on the CES preference margins and preferential status. More than 18 percent of preferential imports, corresponding to €189,929 million, are actually associated with a positive margin. Whereas, for the two percent of preferential imports, the exporters face tariffs that, despite being lower than MFN duties, are not lower than those faced by their competitors. This is especially relevant in the case of agricultural products (Sections I–IV). However, this does not imply that the preferences are necessarily irrelevant since export flows may be lower without preferences (due to trade creation or diversion) or under free trade (due to preference erosion). In some cases, when particularly high bilateral unit values lower the value of the bilateral ad valorem equivalent, the bilateral ad valorem equivalent MFN tariff could be lower than those faced by competitors (the trade flow corresponds to €11,108 million). This is especially relevant in the case of Sections VI (chemicals) and XVI (machinery and mechanical appliances).

5.2. Trade Effect The focus of our analysis is the impact of the EU tariff schedule. Differences among bilateral tariffs lead both to a protectionist and a preferential impact, and it is not possible to distinguish them unless the whole tariff structure is considered. The inclusion of intra-EU trade allows us to compute two counterfactual scenarios. In both cases, preferences are removed but in one case this is the consequence of free trade (all duties are removed), and in the other, it is the consequence of increased protection since all imports are subject to MFN duty. Soc. Sci. 2019, 8, 261 9 of 15

Table 2. Trade ﬂows based on the CES tariﬀ margins, preferential status, and shares with respect to total EU imports.

Preferential Trade Flows MFN Flows Sections Margin > 1 Margin = 1 Margin < 1 Margin > 1 Margin = 1 Margin < 1 189,929 468 17,926 11,108 340,979 477,131 Overall (18) (0) (2) (1) (33) (46) 6423 7 2725 44 1451 6515 I: Animal and Animal Products (37) (0) (16) (0) (8) (38) 12,118 402 4659 603 24,783 8099 II: Vegetable Products (24) (1) (9) (1) (49) (16) 3712 0 261 0 75 6281 III: Fats and Oils (36) (0) (3) (1) (61) IV: Foodstuﬀs, Beverages, and 9458 59 1301 560 13,217 7927 Tobacco (29) (0) (4) (2) (41) (24) VI: Chemicals and Allied 13,091 0 568 4482 82,313 61,502 Industries (8) (0) (3) (51) (38) 16,390 0 737 342 4313 37,748 VII: Plastics/Rubbers (28) (1) (1) (7) (63) 954 0 177 195 5765 4878 IX: Wood and Articles of Wood (8) (1) (2) (48) (41) 49,836 0 7345 38 1374 52,759 XI: Textiles (45) (7) (0) (1) (47) 42,094 0 21 4842 205,692 195,021 XVI: Machineries (9) (0) (1) (46) (44) 35,853 0 132 2 1994 96,400 XVII: Transport (27) (0) (0) (1) (72) Millions of €; shares of total EU imports (in parenthesis). Source: Elaboration on data by COMEXT and TRAINS; 2017.

When we compute trade by setting all tariffs equal to the MFN duties, extra-EU exporters will export less due to the higher protection faced. Taking into account the sectors for which we got statistically significant estimates, Figure3 shows that EU preferences generate additional imports, that is, trade that would not take place without preferences, for €94,142 million (representing 9% of predicted trade). Since we consider intra-EU trade flows, this includes both trade creation and trade diversion impact. On the other hand, imports that would take place if preferences were removed are equal to €47,612 million (around 5% of predicted trade): This represents the trade diversion impact of EU trade policies. The net effect, corresponding to the trade creation effect, is equal to €46,530 million (that is 4% of predicted trade). The trade effect due to preferences differs greatly across sectors. If we focus on the agricultural sector (Sections I–IV), we can say that even if agricultural goods are particularly protected by non-tariff measures (animal products are likely covered by Sanitary and Phytosanitary rules, while beverages are covered by several regulations and internal taxes that restrict trade) the presence of preferences may be relevant and have a positive effect on the level of trade. Indeed, the large trade effect for vegetable products (Section II) and oils and fats (Section III) is not surprising, considering that they have by very elastic import demand (see Table1) and, in general, the agricultural sector also shows a large share of preferential trade (see Figure1).Other sections that present a high trade creation e ffect are plastic (VII) and transport (XVII), these are sections where there might be room for further liberalization on a preferential basis (see Figure1). Soc. Sci. 2019, 8, x FOR PEER REVIEW 9 of 15

16,390 0 737 342 4313 37,748 VII: Plastics/Rubbers (28) (1) (1) (7) (63) 954 0 177 195 5765 4878 IX: Wood and Articles of Wood (8) (1) (2) (48) (41) 49,836 0 7345 38 1374 52,759 XI: Textiles (45) (7) (0) (1) (47) 42,094 0 21 4842 205,692 195,021 XVI: Machineries (9) (0) (1) (46) (44) 35,853 0 132 2 1994 96,400 XVII: Transport (27) (0) (0) (1) (72) Millions of €; shares of total EU imports (in parenthesis). Source: Elaboration on data by COMEXT and TRAINS; 2017.

5.2. Trade Effect The focus of our analysis is the impact of the EU tariff schedule. Differences among bilateral tariffs lead both to a protectionist and a preferential impact, and it is not possible to distinguish them unless the whole tariff structure is considered. The inclusion of intra-EU trade allows us to compute two counterfactual scenarios. In both cases, preferences are removed but in one case this is the consequence of free trade (all duties are removed), and in the other, it is the consequence of increased protection since all imports are subject to MFN duty. When we compute trade by setting all tariffs equal to the MFN duties, extra-EU exporters will export less due to the higher protection faced. Taking into account the sectors for which we got statistically significant estimates, Figure 3 shows that EU preferences generate additional imports, that is, trade that would not take place without preferences, for €94,142 million (representing 9% of predicted trade). Since we consider intra- EU trade flows, this includes both trade creation and trade diversion impact. On the other hand, imports that would take place if preferences were removed are equal to €47,612 million (around 5% of predicted trade): This represents the trade diversion impact of EU trade Soc.policies. Sci. 2019 The, 8, 261net effect, corresponding to the trade creation effect, is equal to €46,530 million (that10 of 15is 4% of predicted trade).

Missing trade Additional trade

5,940 XVII:Transport 23,254 3,910 XVI: Machineries 11,608 27,142 XI: Textiles 120 31,053 IX: Wood & Articles of Wood 277 2,816 VII: Plastics/Rubbers 8,285 499 VI: Chemicals & Allied Industries 1,994 2,601 IV: Foodstuffs, Beverages and Tobacco 5,691 Soc. Sci. 2019, 8, x FOR PEER REVIEW 10 of 15 998 III: Fats & Oils 1,980 tariff measures (animal products are likely covered2,623 by Sanitary and Phytosanitary rules, while II: Vegetable Products 7,797 beverages are covered by several regulations and internal taxes that restrict trade) the presence of I: Animal & Animal Products 963 preferences may be relevant and have a positive effect2,202 on the level of trade. Indeed, the large trade effect for vegetable products (Section II) and oils and fats (Section III) is not surprising, considering that theyFigure have 3. Trade Tradeby very effect eff elasticect due due to import topreferences: preferences: demand Results Results(see forTable se forctors 1) sectors and, with in with significant general, significant thepreference agricultural preference impact impact sector (CES also shows(CESreference a large reference tariff). share tari At ffof worl). preferential Atd worldprices; prices;millions trade millions of (see €. Source Figu of €.: reElaborationSource 1).Other: Elaboration on sections data by on COMEXTthat data present by COMEXT and TRAINS;a high and trade creationTRAINS;2017. effect 2017. are plastic (VII) and transport (XVII), these are sections where there might be room for further liberalization on a preferential basis (see Figure 1). FiguresThe trade4 4 and andeffect5 5present presentdue to thepreferences the impact impact of ofdiffers trade trade preferencesgreatly preferences across for for sectors. country country If groups wegroups focus of of countrieson countries the agricultural defined defined accordingsector (Sections toto income I–IV), levels. we can Low say and that lower even middle if agricultural income countries, goods are defined definedparticularly according protected to the by World non- Bank classification, are the ones with the largest trade increase in relative terms (especially Asian Bank classification, are the ones with the largest trade increase in relative terms (especially Asian countries, such as Bangladesh,Bangladesh, Pakistan,Pakistan, Cambodia,Cambodia, Philippines,Philippines, Indonesia,Indonesia, and Sub-SaharanSub-Saharan Africa countries, such as CCôteôte d’Ivoire, Ghana, Madagascar, Ethiopia,Ethiopia, Uganda).Uganda). However, itit is worth noting thatthat thethe highesthighest tradetrade flowsflows fromfrom high-incomehigh-income OECDOECD thatthat wouldwould notnot taketake placeplace without preferences concerns importsimports fromfrom Korea, Korea, Switzerland, Switzerland, Israel, Israel and, and Chile, Chile, under under specific specific preferential preferential treatment treatment and freeand tradefree trade agreements. agreements.

Missing trade Additional trade

30,661 Upper middle income 38,678 6,809 Lower middle income 30,166

135 Low income 3,322

7,803 High income: OECD 21,521

2,204 High income: nonOECD 455

Figure 4.4. Trade eeffectﬀect duedue toto preference:preference: ResultsResults by countriescountries basedbased on incomeincome levels (CES reference taritariff).ﬀ). At world prices;prices; millionsmillions ofof €€.. Source: Elaboration Elaboration on data data by COMEXT and TRAINS; 2017.

The upper middle-income countries benefit significantly from EU preferences both in relative and absolute terms. This result is due to the existing trade agreements with countries in Latin America, in particular to the EU-Mercosur (Argentina, Brazil, Paraguay, and Uruguay) free trade agreement and the agreement with Mexico, though some trade barriers still remain; and to the Stabilization and Association Agreement between the EU and the Western Balkan partners from which the EU mainly imports machinery and appliances, metals, and chemicals. On the other hand, the middle-income group of countries in Asia and the Pacific (especially, China) are those most negatively affected by the existing preferential schemes. Actually, these are countries with higher trade flows than would have been the case if preferences were removed.

Soc. Sci. 2019, 8, 261 11 of 15 Soc. Sci. 2019, 8, x FOR PEER REVIEW 11 of 15

Missing trade Additional trade

SubSaharan Africa 134 3,118

South Asia 3,510 15,501

North America 3,786 100

Middle East & North Africa 831 10,589

Latin America & Caribbean 1,821 8,864 40,126 Europe & Central Asia 1,970

East Asia & Pacific 35,558 15,843

Figure 5.5. TradeTrade eﬀeffectect due due to to preference: preference: Results Results by countriesby countries according according to regions to regions (CES reference(CES reference tariﬀ). Attariff). world At prices;world prices; millions millions of €. Source of €.: Source Elaboration: Elaboration on data on by data COMEXT by COMEXT and TRAINS; and TRAINS; 2017. 2017.

TheFigure upper 6 shows middle-income the trade effect countries of the benefit actual significantly EU trade policies from EU through preferences the computation both in relative of andthe absolutecounterfactual terms. trade This resultflows is that due towould the existing be record tradeed agreements if all tariffs with were countries removed. in Latin The America, overall inprotectionist particular toimpact the EU-Mercosur of EU trade policies (Argentina, is quite Brazil, large Paraguay, and amounts and to Uruguay) €310,210 freemillion, trade around agreement 30% andof predicted the agreement trade. These with Mexico, are imports though that some would trade take barriers place under still remain; free trade. and Larger to the Stabilizationdecrease of trade and Associationdue to protection Agreement is registered between for the textile EU andproducts the Western (XI), so Balkanthat the partners major exporters from which of textiles the EU (such mainly as importsChina, India, machinery and Vietnam) and appliances, should metals,be particularly and chemicals. affected by a liberalization policy. Other sections that mightOn the suffer other from hand, trade the middle-incomeliberalization are group the agriculture of countries (Sections in Asia I–IV) and theand Pacific the plastic (especially, sector China)(VII). Comparing are those mostresults negatively so far, notwithstanding affected by the existingthe existence preferential of several schemes. preferential Actually, schemes, these arethe countriesoverall stance with of higher the EU trade trade flows policy than seems would to haveremain been quite the protectionist. case if preferences were removed. FigureResults6 showshows that the €37,282 trade e millionffect of the(around actual 4% EU of trade predicted policies trade) through of imports the computation would disappear of the counterfactualunder free trade. trade These flows are that exports would belost recorded by countr if allies tari actuallyffs were enjoying removed. an The advantage overall protectionist due to the impactpreferential of EU treatment trade policies and isrepresents quite large what and amountsdeveloping to €countries310,210 million, may lose around as a30% consequence of predicted of trade.multilateral These liberalization are imports that(prefere wouldnce takeerosion). place If under we look free at trade. the results Larger for decrease different of products, trade due the to protectionmost protected is registered sections for appear textile to products be textiles (XI), (XI), so thatvegetables the major (Section exporters II), and of textiles oils and (such fats as (Section China, India,IV), whereas and Vietnam) the lowest should protection be particularly impact is found affected for by cereals a liberalization (Section V). policy. Other sections that might suffer from trade liberalization are the agriculture (Sections I–IV) and the plastic sector (VII). Comparing results so far, notwithstanding the existence of several preferential schemes, the overall Missing trade Additional trade stance of the EU trade policy seems to remain quite protectionist. Results show that €37,282 million (around 4% of predicted trade) of imports would83,293 disappear under free trade. These are exportsXVII:Transport lost by countries5,223 actually enjoying an advantage due to the preferential treatment and representsXVI: Machineries what developing countries may50,787 lose as a consequence of 3,507 90,017 multilateral liberalization (preference erosion). If we look at the results for different products, the most XI: Textiles 19,774 protected sections appear to be textiles (XI), vegetables (Section II), and oils and fats (Section IV), IX: Wood & Articles of Wood 1,049 whereas the lowest protection impact is found for71 cereals (Section V). 33,406 VII: Plastics/Rubbers 2,319 7,727 VI: Chemicals & Allied Industries 502 17,724 IV: Foodstuffs, Beverages and Tobacco 1,929 6,978 III: Fats & Oils 883 16,392 II: Vegetable Products 2,236 2,836 I: Animal & Animal Products 837

Soc. Sci. 2019, 8, x FOR PEER REVIEW 11 of 15

Missing trade Additional trade

SubSaharan Africa 134 3,118

South Asia 3,510 15,501

North America 3,786 100

Middle East & North Africa 831 10,589

Latin America & Caribbean 1,821 8,864 40,126 Europe & Central Asia 1,970

East Asia & Pacific 35,558 15,843

Figure 5. Trade effect due to preference: Results by countries according to regions (CES reference tariff). At world prices; millions of €. Source: Elaboration on data by COMEXT and TRAINS; 2017.

Figure 6 shows the trade effect of the actual EU trade policies through the computation of the counterfactual trade flows that would be recorded if all tariffs were removed. The overall protectionist impact of EU trade policies is quite large and amounts to €310,210 million, around 30% of predicted trade. These are imports that would take place under free trade. Larger decrease of trade due to protection is registered for textile products (XI), so that the major exporters of textiles (such as China, India, and Vietnam) should be particularly affected by a liberalization policy. Other sections that might suffer from trade liberalization are the agriculture (Sections I–IV) and the plastic sector (VII). Comparing results so far, notwithstanding the existence of several preferential schemes, the overall stance of the EU trade policy seems to remain quite protectionist. Results show that €37,282 million (around 4% of predicted trade) of imports would disappear under free trade. These are exports lost by countries actually enjoying an advantage due to the preferential treatment and represents what developing countries may lose as a consequence of multilateral liberalization (preference erosion). If we look at the results for different products, the Soc.most Sci. protected2019, 8, 261 sections appear to be textiles (XI), vegetables (Section II), and oils and fats (Section12 of 15 IV), whereas the lowest protection impact is found for cereals (Section V).

Missing trade Additional trade

83,293 XVII:Transport 5,223 XVI: Machineries 50,787 3,507 90,017 XI: Textiles 19,774 1,049 IX: Wood & Articles of Wood 71 33,406 Soc. Sci. 2019, 8, x FOR PEER REVIEWVII: Plastics/Rubbers 2,319 12 of 15 7,727 VI: Chemicals & Allied Industries 502 Figure 6. Trade effect due to protection: Results for sectors with significant preference impacts (CES IV: Foodstuffs, Beverages and Tobacco 17,724 reference tariff). At world prices; millions of €. Source:1,929 Elaboration on data by COMEXT and TRAINS; 6,978 2017. III: Fats & Oils 883 16,392 II: Vegetable Products 2,236 Figures 7 and 8 show the protectionist impact of EU tariffs for different country groups. Exports I: Animal & Animal Products 2,836 from the high-income OECD countries group are837 the most negatively affected (first of all, North America, i.e., US and Canada), followed by those from upper middle-income countries. In terms of preferenceFigure erosion, 6. Trade low-income effect due to and protection: lower middle-income Results for sectors countries with significant are the groups preference that impacts risk more. This (CESis especially reference the tari ffcase). At for world Asian prices; countries millions such of €. Source:as India, Elaboration Indonesia, on dataVietnam, by COMEXT Sri Lanka, and and Bangladesh.TRAINS; 2017.

Missing trade Additional trade

621 Low income 2,133

19,511 High income: nonOECD 198

29,815 Lower middle income 15,139

102,239 High income: OECD 6,235 158,024 Upper middle income 13,577

Figure 7.7. TradeTrade e ffeffectect due due to protection:to protection: Results Results by countries by countries according according to income to levelsincome (CES levels reference (CES tarireferenceff). At tariff). world prices;At worl millionsd prices; of millions€. Source: of €. Elaboration Source: Elaboration on data byon COMEXTdata by COMEXT and TRAINS; and TRAINS; 2017. 2017. Figures7 and8 show the protectionist impact of EU tari ffs for different country groups. Exports from the high-income OECD countries group are the most negatively affected (first of Missing trade Additional trade all, North America, i.e., US and Canada), followed by those from upper middle-income countries. In terms of preference erosion, low-income and lower middle-income countries are the groups that risk Sub-Saharan Africa 488 more. This is especially the case for Asian1,388 countries such as India, Indonesia, Vietnam, Sri Lanka, and Bangladesh. South Asia 15,314 9,220

North America 55,656 44

Middle East & North Africa 7,405 4,047

Latin America & Caribbean 15,424 3,172

Europe & Central Asia 19,265 12,809 196,658 East Asia & Pacific 6,601

Figure 8. Trade effect due to protection: Results by countries according to income levels (CES reference tariff). At world prices; millions of €. Source: Elaboration on data by COMEXT and TRAINS; 2017.

Soc. Sci. 2019, 8, x FOR PEER REVIEW 12 of 15

Figure 6. Trade effect due to protection: Results for sectors with significant preference impacts (CES reference tariff). At world prices; millions of €. Source: Elaboration on data by COMEXT and TRAINS; 2017.

Figures 7 and 8 show the protectionist impact of EU tariffs for different country groups. Exports from the high-income OECD countries group are the most negatively affected (first of all, North America, i.e., US and Canada), followed by those from upper middle-income countries. In terms of preference erosion, low-income and lower middle-income countries are the groups that risk more. This is especially the case for Asian countries such as India, Indonesia, Vietnam, Sri Lanka, and Bangladesh.

Missing trade Additional trade

621 Low income 2,133

19,511 High income: nonOECD 198

29,815 Lower middle income 15,139

102,239 High income: OECD 6,235 158,024 Upper middle income 13,577

Figure 7. Trade effect due to protection: Results by countries according to income levels (CES Soc. Sci.reference2019, 8, 261tariff). At world prices; millions of €. Source: Elaboration on data by COMEXT and TRAINS;13 of 15 2017.

Missing trade Additional trade

Sub-Saharan Africa 488 1,388

South Asia 15,314 9,220

North America 55,656 44

Middle East & North Africa 7,405 4,047

Latin America & Caribbean 15,424 3,172

Europe & Central Asia 19,265 12,809 196,658 East Asia & Pacific 6,601

Figure 8.8. TradeTrade e ffeffectect due due to protection:to protection: Results Results by countries by countries according according to income to levelsincome (CES levels reference (CES tarireferenceff). At tariff). world At prices; worl millionsd prices; ofmillions€. Source: of €. Elaboration Source: Elaboration on data byon COMEXTdata by COMEXT and TRAINS; and TRAINS; 2017. 2017. 6. Conclusions The aim of this article was to provide a thorough empirical analysis of the EU’s tariffs, estimating their effects on bilateral trade flows while controlling for their multilateral impact and including a full set of fixed effects. Estimating the impact of trade policies involves complex issues due to the difficulty in correctly specifying the gravity equation. Here, we carried out a simple but theoretically consistent gravity analysis of EU tariffs using a complete, well-documented dataset. We ran estimations at the disaggregated level of individual tariff lines and identified relationships on the basis of the extensive variation available in bilateral trade data. That is to say, a variation in trade costs across exporters provides the price variation needed to trace the slope of the EU import demand curves. This characteristic should be acknowledged, as the literature of international trade has recognized years ago the existence of a multilateral component of resistance, catching the fact that exports from country A to country B depend on the commercial costs of all possible suppliers. In other words, focusing on absolute costs means that other important general equilibrium effects that operate through the price index are often ignored. A crucial feature is the measurement of trade policy treatment. In particular, the policy variable must take into account the relative changes in bilateral tariff in relation to duties paid by other exporters. This allows accounting for the multilateral resistance component capturing the fact that exports from country i to country j depend on trade costs across all possible suppliers. In this article we presented a general framework in which the magnitude of trade policies’ impacts on trade was structurally estimated, using very detailed data including intra-national flows. We computed the bilateral tariff margin as the ratio between a reference tariff factor and the applied tariff factor faced by each exporter. Such a choice is consistent with the observation that bilateral trade depends also on the market conditions applied to other countries and not only on direct market conditions. The result of the trade policy can be an advantage or a disadvantage with respect to other countries, and the greater is the relative advantage (disadvantage) the higher (lower) is the expected trade flows. Whereas Cipollina et al.(2017) assess the impact of EU preferences considering only extra-EU trade flows, we developed a model with both extra and intra-EU trade flows at the most disaggregate tariff line level. The use of intra-national trade flow data is consistent with gravity theory (Yotov 2012; Yotov et al. 2016) and the most recent literature has pointed out that it allows a consistent identification of multilateral (Heid et al. 2017) as well as bilateral trade policies (Dai et al. 2014; Bergstrand et al. Soc. Sci. 2019, 8, 261 14 of 15

2015). Moreover, since intra-EU trade is substantial, it is important to allow foreign and domestic producers to be active in the same sector producing similar goods. Cipollina et al.(2017) show that preference margins can be positive or negative when they are computed with regard to the tariffs paid by other exporters. These points to a renewed interpretation of the trade creation and trade diversion concepts, through multilateral resistance terms, of changes in k k tariffs included in T . A decrease in ti is directly increasing bilateral imports from i, while also changing the relative trade costs through its impact on the reference tariff. EU consumers, therefore, reallocate demand according to new relative prices, diverting trade from non-preferred countries and reducing demand for domestic producers. Since we took into account the intra-EU duty-free trade, the reference tariffs turned out to be much lower. This led to a significant reduction, or disappearance, of the preference margins, though they are likely to have a larger impact since they affect total consumption rather than imports only. An important byproduct of our approach is that it can be used to obtain estimates of the elasticity of substitution which is the single most important parameter in the international trade literature. Since tariffs are a direct price-shifter, gravity theory can be used to recover the elasticity of substitution directly from the estimate of the coefficient on bilateral margins. Although bilateral measures of bilateral tariffs have previously been used to identify the trade elasticity in structural gravity frameworks, e.g., Cipollina et al.(2017), we were able to estimate a more comprehensive elasticity by also taking into account the impact on domestic goods. Our model allows differences in trade elasticities between the various sectors, which means that consumers can react differently to price changes in different sectors. This is quite significant as tariffs vary substantially across the various sectors, and this means that the failure to take into account this heterogeneity between the various sectors can lead to distorted results. We find that the impact of the community policy, in terms of protection, is much stronger than the increase in extra-EU trade due to the preferential policy. In the debate on how effective the EU preferences are, our analysis shows that, even if the EU seems to lean towards trade liberalization through a proliferation of preferential agreements, the preferences do not seem to be very effective in incrementing trade. Nonetheless, preferences play a significant role in specific products or exporters.

Author Contributions: Conceptualization, M.C. and L.S.; Data curation, M.C.; Formal analysis, M.C. and L.S.; Investigation, M.C. and L.S.; Methodology, M.C. and L.S.; Supervision, L.S.; Writing—original draft, M.C. and L.S.; Writing—review & editing, M.C. and L.S. Funding: This research received no external funding. Acknowledgments: We would like to thank participants to the session “Trade policy relevance of CGE parameters” at the 22nd Annual Conference on Global Economic Analysis “Challenges to Global, Social, and Economic Growth”, GTAP Conference, 19–21 June 2019, Warsaw (Poland). Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

Anderson, James E. 1979. A Theoretical Foundation for the Gravity Equation. The American Economic Review 69: 106–16. Available online: http://www.jstor.org/stable/1802501 (accessed on 10 September 2018). Anderson, James E., and Eric Van Wincoop. 2003. Gravity with Gravitas: A Solution to the Border Puzzle. The American Economic Review 93: 170–92. [CrossRef] Baier, Scott L., and Jeffrey H. Bergstrand. 2001. The growth of world trade: tariffs, transport costs, and income similarity. Journal of International Economics 53: 1–27. [CrossRef] Bergstrand, Jeffrey H., Mario Larch, and Yoto V. Yotov. 2015. Economic Integration Agreements, Border Effects, and Distance Elasticities in the Gravity Equation. European Economic Review 78: 307–27. [CrossRef] Caliendo, Lorenzo, and Fernando Parro. 2015. Estimates of the trade and welfare effects of Nafta. Review of Economic Studies 82: 1–44. [CrossRef] Cipollina, Maria, David Laborde Debucquet, and Luca Salvatici. 2017. The tide that does not raise all boats: An assessment of EU preferential trade policies. Review of World Economics 153: 199–31. [CrossRef] Soc. Sci. 2019, 8, 261 15 of 15

Corbo, Vesna, and Chiara Osbat. 2013. Trade Adjustment in the European Union. A Structural Estimation Approach, European Central Bank Working Paper Series, no. 1535; Frankfurt: ECB Working Paper. Dai, Mian, Yoto V. Yotov, and Thomas Zylkin. 2014. On the Trade-Diversion Effects of Free Trade Agreements. Economic Letters 122: 321–25. [CrossRef] De Melo, Jaime, and Alessandro Nicita, eds. 2018. Non-Tariff Measures: Economic Assessment and Policy Options for Development. Geneva: United Nations Conference on Trade and Development. European Commission. 2016. Strategic Plan 2016-2020, DG Trade. Available online: https://trade.ec.europa.eu/ doclib/docs/2016/august/tradoc_154919.pdf (accessed on 1 June 2019). Fally, Thibault. 2015. Structural Gravity and Fixed Effects. Journal of International Economics 97: 76–85. [CrossRef] Feenstra, Robert C., Philip Luck, Maurice Obstfeld, and Katheryn N. Russ. 2018. In Search of the Armington Elasticity. The Review of Economics and Statistics 100: 135–50. [CrossRef] Head, Keith, and Thierry Mayer. 2014. Gravity equations: Workhorse, toolkit, and cookbook. In Handbook of International Economics. Amsterdam: Elsevier, vol. 4, pp. 131–95. Heckman, James J. 1979. Sample selection bias as a specification error. Econometrica 47: 153–61. [CrossRef] Heid, Benedikt, Mario Larch, and Yoto Yotov. 2017. Estimating the Effects of Non-discriminatory Trade Policies within Structural Gravity Models. CESifo Working Paper no. 6735. München: CESifo Working Paper Series. Helpman, Elhanan, Marc Melitz, and Yona Rubinstein. 2008. Estimating trade flows: Trading partners and trading volumes. Quarterly Journal of Economics 123: 441–87. [CrossRef] Imbs, Jean, and Méjean Isabelle. 2009. Elasticity Optimism. International Monetary Fund Working Paper Series, 09/279; Washington: IMF. Lai, Huiwen, and Susan Chun Zhu. 2004. The Determinants of Bilateral Trade. Canadian Journal of Economics 37: 459–83. [CrossRef] Low, Patrick, Roberta Piermartini, and Jürgen Richtering. 2009. Multilateral solutions to the erosion of nonreciprocal preferences in nonagricultural market access. In Trade Preference Erosion: Measurement and Policy Response. Edited by Bernard Hoekman, Will Martin and Carlos A. Primo Braga. Washington: World Bank, pp. 219–67. Ornelas, Emanuel, and Marcos Ritel. 2018. The Not-So-Generalized Effects of the Generalized System of Preferences. CEP Discussion Paper No 1578. London: Centre for Economic Policy Research. Martin, Will, and Cong S. Pham. 2016. Estimating the Gravity Model When Zero Trade Flows are Frequent and Economically Determined. World Bank Policy Research Working Paper No. 7308. Washington: World Bank. Romalis, John. 2007. Nafta’s and cusfta’s impact on international trade. Review of Economics and Statistics 89: 416–35. [CrossRef] Silva, JMC Santos, and Silvana Tenreyro. 2006. The log of gravity. The Review of Economics and Statistics 88: 641–58. [CrossRef] Scoppola, Margherita, Valentina Raimondi, and Alessandro Olper. 2018. The impact of EU trade preferences on the extensive and intensive margins of agricultural and food products. In Agricultural Economics. Milwaukee: International Association of Agricultural Economists, vol. 49, pp. 251–63. Tinbergen, Jan J. 1962. Shaping the World Economy: Suggestions for an International Economic Policy. The Twentieth New York: Century Fund. Viner, Jacob. 1950. The Customs Union Issue. New York: Carnegie Endowment for International Peace. Yotov, Yoto V. 2012. A Simple Solution to the Distance Puzzle in International Trade. Economics Letters 117: 794–98. [CrossRef] Yotov, Yoto V., Roberta Piermartini, José-Antonio Monteiro, and Mario Larch. 2016. An Advanced Guide to Trade Policy Analysis: The Structural Gravity Model. Geneva: World Trade Organization.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).