Tariffs and non-tariff measures: substitutes or complements. A cross-country analysis

Transcription

1 Bank i Kredyt 48(1), 2017, Tariffs and non-tariff measures: substitutes or omplements. A ross-ountry analysis Eyal Ronen* Submitted: 29 April Aepted: 3 November Abstrat Alongside the global tariff liberalizatio a growing body of evidene demonstrates the rise in the use of non-tariff measures (NTMs), whih suggests a substitution effet between these two import poliy instruments. Yet, detailed eonomi data reveals that in ountries with lower tariff rates (developed ountries), the use of NTMs is signifiantly lower ompared to developing ountries, whih implies a possible omplementary effet between tariffs and NTMs aross nations. Using a dataset of Kee, Niita and Olarreaga (2009) on ad valorem tariff equivalents of NTMs, at a very disaggregated produt level, this paper explores the determinants of NTMs and their substitutability/omplementarity relations with tariff barriers. While exploiting the ountry variatio it demonstrates the dereasing trend of substitutability between the two import poliy instruments with the rise in eonomi development. In partiular, a signifiant omplementarity orrelation exists between the two trade measures among the wealthiest nations, implying a stronger ommitment to freer trade. Keywords: non-tariff measures, tehnial barriers to trade, WTO JEL: F13, F14, F53 * University of Warsaw, Faulty of Eonomi Sienes; eronen@wne.uw.edu.pl.

2 46 E. Ronen 1 Introdution Over the past two deades, international trade has inreased rapidly, largely due to a signifiant gradual elimination of tariff protetion. Tariff redution has been ahieved either by suessive rounds of multilateral trade negotiations, by unilateral liberalizatio or by the reation of preferential trade agreements (PTAs). Sine a notable share of PTAs was among developing ountries, whih originally ommened with higher tariff levels, it is no surprise that these ountries in partiular have pursued more far-reahing tariff elimination. Yet, the average tariff levels in low-inome ountries are still signifiantly higher than developed ountries (WTO 2016). Alongside the redution of tariff rates, aumulated evidene shows a growing propensity in the use of non-tariff measures to trade (NTMs 1 ) by many ountries, whih partially offsets the advanements ahieved by lowering tariffs. Contrary to tariff measures, whih were originally introdued in order to realize eonomi and trade objetives, the purported intention behind the imposition of NTMs was to design publi objetives whih are non-protetionist by nature. These poliy measures often serve as the first-best instrument to advane various soial, politial or environmental protetion objetives, as well as health and onsumer protetion. Nevertheless, these instruments have beome popular in ahieving eonomi goals, mainly laiming to orret market ineffiienies whih arise from information asymmetries or imperfet ompetition. However, as the imposition of suh proedures reates a benefiial advantage for players who partiipate in the trade arena, it is no wonder that the use of NTMs has expanded. The politial eonomy literature demonstrates how poliymakers, who fae pressures to protet domesti produers, may hoose to use NTMs more extensively. At the same time, it laims that NTMs are merely alternative hannels of protetionism in disguise. These ations may even be intensified when the redution of tariffs adversely affets the loal produers performanes or in times of eonomi downturn. The use of NTMs varies onsiderably aross ountries, differing aording to the speifi type hose affeting diverse produts, and flutuating over time. Nevertheless, a ommon motive in various approahes whih study their inidene validates the aelerated expansion of NTMs over time. In partiular, some praties suh as tehnial barriers to trade (TBTs) and sanitary and phytosanitary (SPS) measures, whih aount for a large share of all NTMs, have been on the rise sine the mid-90s, as shown in the inreasing number of notifiations of SPS and TBT measures reported to the WTO (Figure 1). Moreover, this upward diretion is well demonstrated by the substantial inrease in the use of anti-dumping measures over the last two deades, mostly by developing ountries. As the ontinuous redution of tariffs and the expanding use of NTMs move in opposite diretions, inreasing attention has been drawn to the possible substitution effet between these two import poliy measures. The ommon onsensus shared among sholars and supported by several empirial analyses laims that NTMs often enable ountries to enhane restritiveness, manipulate the terms-of-trade and relaim possible eonomi losses due to tariff liberalization. These studies, whih fous mostly on speifi ountries or partiular NTMs, demonstrate the substitutability relation between tariffs and NTMs. This negative orrelation emphasizes that new NTM restritions simply replae the traditional ones (namely tariffs), in order to ahieve similar objetives. 1 Despite minor differenes, the terms non-tariff measures and non-tariff barriers are often used interhangeably. In this paper the term non-tariff measures will be used.

3 Tariffs and non-tariff measures: substitutes or omplements This paper tests the proposition that although from the broad perspetive a negative orrelation may be found, a deeper analysis would reflet heterogeneity among ountries. In partiular, it will verify the diminishing trajetory of the substitutability between the two import poliy instruments with the rise in GDP per apita. It will provide evidene that the more developed a ountry is, the less likely that NTM praties our. In low-inome ountries, the substitutability effet is dramatially high; however, the rationalization is ounter-intuitive, sine high levels of applied tariffs are found alongside a modest imposition of NTMs. Low levels of NTMs are the outome of the omplexity and high osts assoiated with operating administrative and regulatory systems. Seondly, low-inome ountries are more revenue oriented, thus depending mostly on tariff inome. Thirdly, their dependeny on world trade as well as their ommitment to freer trade is less robust than the rihest ountries. Lastly, low-inome ountries enjoy greater flexibility in terms of binding overhang gaps, whih signifiantly dereases their motivation to impose NTMs. The binding overhang gaps are essentially the differenes between the bound levels and applied tariff rates on eah produt granted to ountries under WTO rules. On the other hand, high-inome ountries that generally levy lower levels of tariffs are omparatively more ommitted to transpareny and non-disrimination praties that result in lower levels of NTMs. Those developed ountries are generally at the frontline of WTO disussions, hene they are more involved in the proess of eliminating TBT and SPS measures. This means taking upon themselves the removal of trade barriers more extensively than elsewhere. Moreover, sine high-inome ountries tend to rely more profoundly on international trade, it leads them to seek better onditions, both for domesti importers, as well as for their loal exporters in foreign markets. Suh interests o-exist with the neessity to attrat imports at the lowest osts possible for the benefit of domesti onsumers as well as for importers of intermediate goods. These reasons are expeted to lead to less substitutability or even a greater omplementarity orrelation between the two import poliy measures. The proposition mentioned above is further supported by evidene based on a reent dataset of Kee, Niita and Olarreaga (2009) aompanied by the authors alulations. Figure 2 portrays how ountries with lower levels of tariffs also impose lower NTMs. This ross-ountry analysis demonstrates that omplementary orrelation exists between the two import measures, alongside negative orrelation with the level of eonomi development. Moreover, data on NTMs overage ratio draw similar onlusions regarding the positive orrelation of tariff levels and the number of produts affeted by NTMs. Additionally, a strong orrelation is found between the average number of NTMs at the hapter level and the tariffs (UNCTAD 2013). Taken together, these results demonstrate how a low tariff regime may be paired with a less stringent NTM poliy. Suh findings reinfore the omplementary relation between the two import poliy instruments, hene reating a paradox with the shared onsensus on substitutability. Despite their signifiant impat on trade and the reporting requirements to notify NTMs to the WTO with lear information regarding the produts affeted, systematially olleted data regarding their impliations is still hardly available. Eonomi studies address NTMs in a narrow setting, mainly in a qualitative approah, whereas in pratie NTMs are broadly employed and hange over time alongside the onstant appearane of new types of NTMs that may require other analyti methodologies. Throughout the years, several attempts were made in order to ollet information on NTMs (i.e. MAST, TRAINS). Nevertheless, to some extent, these quantifiations usually address speifi types of measures, ountries, and produts, and it is still not suffiiently omprehensive or laks the broader perspetive.

4 48 E. Ronen Reently, this gap has been addressed by Kee, Niita and Olarreaga (2009) who tried to overome various hallenges and reated a unique dataset of ad valorem equivalents (AVE) of NTMs at a disaggregated level (i.e. the 6-digit level of the HS lassifiation). Their work provides estimations for tariff equivalents of NTMs of 104 ountries (developing and developed), while omprising information regarding more than 30 types of NTMs. Among these measures are prie ontrol, quantity restritions, monopolisti measures and tehnial regulations, agriultural domesti support and others. The extensive data set onsists of observations on ad valorem equivalents of NTMs expressed as a perentage of the value of the produt, making them diretly omparable with tariffs. The possible diret omparison provides the starting point for this paper, as it allows an eonometri-based analysis of the substitutability/omplementarity orrelation of NTMs and tariff barriers to be undertaken. Moreover, the urrent analysis will outline the heterogeneity aross ountries and emphasize the diminishing negative orrelation between the two import measures with the rise in ountries level of eonomi development. By performing an in-depth and omprehensive analysis whih differentiates between ountry groups, this paper aims to add an important dimension to the basi orrelation analysis offered by Kee, Niita and Olarreaga (2009), whih fouses on the broad perspetive. Seondly, it will ontribute to the existing knowledge by investigating some additional determinants of NTMs, providing a more preise haraterization and motivation behind their pervasiveness. One imperative explanatory variable for that matter is the binding overhang ratio, whih represents the flexibility between bound and applied tariffs, and in turn the inentive to alternate to NTMs. Other ontrol fators whih are taken into onsideration and influene the restritiveness of NTMs are trade openness, import revenues, and others. The paper is omprised of five setions. Following an introdutio the seond setion portrays the haraterization of the use of NTMs aross ountries. The third setion outlines the literature review, starting with the relations between tariffs, NTMs and trade, followed by a desription of the approahes used to quantify the restritiveness of NTMs, sueeded by the partiular methodology of Kee, Niita and Olarreaga (2009) whih served for obtaining the data on AVE of NTMs in the analysis. The fourth setion presents the methodology whih was hosen to ondut the eonometri analysis in the paper, aompanied by a disussion of the results of the estimations, along with several robustness heks. The last setion of the paper underlines the main onlusions whih an be drawn from the researh. 2 The harateristis of the use of NTMs aross ountries The sope of use of non-tariff measures varies signifiantly aording to their type, nature or the objetives to be ahieved. Evidene indiates that, in general, the average ountry imposes tehnial barriers to trade (TBT) on about 30% of produts and trade. Sanitary and phytosanitary (SPS) measures, whih are exlusively related to agriulture and food produts, are imposed on more than 60% of agriultural produts, whih in fat represents slightly less than 15% of overall trade. Among non-tehnial measures, pre-shipment inspetions, for example, affet on average approximately 20% of trade and produts (UNCTAD 2013). NTMs also differ onsiderably aross ountries, depending muh on eah ountry s omparative advantage and politial eonomy preferenes. NTMs imposed on agriultural produts are likely to be greater and more restritive in ountries with a stronger omparative advantage in produing

5 Tariffs and non-tariff measures: substitutes or omplements agriultural produts. These fators are demonstrated in the use of SPS measures and quantity and prie ontrol measures, whih tend to be more predominant in developing ountries. Countries, whih rely heavily on domesti prodution of traditional setors suh as agriulture will use these instruments more extensively ompared to developed ountries. At the same time, riher ountries, whih are often onerned about shielding import-ompeting industries, or preserving the interests of infant industries, are found to impose TBTs more extensively than elsewhere. The desriptive statistis presented in Table 2 point out the differenes between the impositions of NTMs aross ountry groups. They provide a snapshot of the average statistis on various variables suh as tariffs, NTMs, and others. These figures are grouped aording to the level of eonomi development of the ountries (based on GDP per apita ategories). The figures indiate that exept for the lowest inome group, the riher a ountry is in terms of GDP per apita, the lower its use of NTMs. While the AVE of NTMs in lower-middle inome ountries is roughly 15%, in upper-middle inome ountries and in high-inome ountries the average is 10% and less than 9%, respetively. The exeption to this rule is the ase of low-inome ountries, where although the average AVE of NTMs is relatively low (4%), as demonstrated in Figure 3, there is a large variane aross these ountries. For example, in the Afrian ontinent, Rwanda (0.3%) and Kenya (1.3%) are at the lowest end, while Sudan and Nigeria are on the highest end with an average AVE of NTMs whih reahes 40%. Complementary studies show that although on average NTMs are utilized for slightly less than half of the list of 5,000 produts, in the Afrian ontinent Tanzania and Senegal use NTMs substantially less than Egypt or Uganda. In Latin Ameria, the imposition of NTMs by Argentina is double that of Chile or Paraguay. More examples ould be found in Asia, where Bangladesh, Syria and the Philippines utilize NTMs muh more than Cambodia or Indonesia. Furthermore, as seen in the desriptive statistis, riher ountries are more ommitted to the redution of all types of barriers to trade, as internationally agreed under the general priniples of the WTO. Therefore, the average use of NTMs among the rihest ountries is the lowest found aross all ountries. These ountries are typially more open and dependent on international trade, and onsequently are less likely to employ NTMs ompared to ountries, whih are more self-reliant. On the other hand, with some exeptions, the less developed a ountry is, the less likely it will be open to ompeting for import flows. The low-inome ountries, whih rely profoundly on revenue generating tax measures suh as import tariffs, will prefer not to operate a ostly and omplex administration. Furthermore, low-inome ountries also enjoy a bigger overhang gap (the differene between bound and applied tariffs), whih allows them the flexibility to inrease their level of atual tariffs legally. While exploring the use of NTMs aross ountries, it is essential to takle speifi harateristis of ountries suh as export performanes as well as those who are the natural trading partners. In general, high-inome ountries tend to export more than lower inome ountries (by share of GDP), and their exports are mostly direted to other rih nations. The low and middle-inome ountries still trade mostly among themselves, though in the last deade these ountries inreasingly expand their exports to higher inome ountries at the expense of their traditional markets. Trade data shows that the developing ountries share in world merhandise exports have expanded dramatially from 17% to 43% during the last 25 years (WTO 2016). Moreover, although the low-inome ountries may enjoy better market aess onditions relative to the past, they still fae larger finanial and manufaturing onstraints. These apaity onstraints make it even harder for them to overome the barriers posed by NTMs. Given the fat that low-inome

6 50 E. Ronen ountries speialize in traditional exports suh as agriulture, textile, and apparel goods, and suffer from higher apaity onstraints, it is even more worrisome that these ountries fae substantial obstales in the form of high substitutability of tariffs with NTMs in their main exporting markets. This problem is further amplified by evidene of reent ITC business surveys on NTMs, whih indiate that exporters of agriultural produts report more problems related to TBT/SPS measures than exporters of manufatured goods 59% ompared to 34% respetively (WTO 2012). 3 Literature review 3.1 NTMs, tariffs and trade in the literature In reent years, a growing interest has emerged around the harateristis of NTMs and partiularly, over their impat on trade restritiveness and on the welfare of nations. This paper builds both on existing politial eonomy theories, as well as on the empirial literature on the determinants of governments imposition of NTMs in response to dereasing tariffs. It primarily attempts to ontribute to the strand of literature analysing the politial eonomy environment of trade poliy, whih is based on Grossman and Helpman (1994), and has been tested empirially in several studies. While the authors laid down the foundations of the role of domesti interest groups, others, suh as Mansfield and Bush (1995), have foused on the domesti politial influene of institutions and the impat of deteriorating maroeonomi onditions. The later found that these fators explain the variane between ountries in the demand for NTMs protetion by pressure groups. Lee and Swagel (1997) establish that ountries pair these two trade measures in order to protet vulnerable industries whih are politially important or those threatened by import ompetition. Yu (2000) highlights the importane of transpareny and the presene of informed onsumers on the government s deision to substitute voluntary export restraints (VERs) with tariffs. Furthermore, Yu laims that an inrease in foreign ompetition will not ause the government to substitute NTMs for tariffs; however, a rise in the government s valuation of politial ontribution might do. The vast majority of the literature suggests that substitutability between NTMs and tariffs exists alongside the implementation of preferential trade agreements (PTAs). These agreements lower the rate of protetio but often do not redue the domesti pressure for protetionism. The law of onstant protetion phenomena suggests that produers who are well-proteted by tariffs may are less for NTMs relative to industries adversely affeted by the eonomi impat of dereasing tariffs, whih may often reeive NTMs protetion as a substitute (Bhagwati 1988). By employing data on Turkey s tariffs and NTMs, Limao and Tovar (2011) exploit the variation in tariff onstraints generated by multilateral agreements and PTAs. They establish a ausal impat of the resulting tariff onstraints on the likelihood and restritiveness of NTMs. By onsidering the differenes in the size of EU member states in a PTA, they show that if the ommon EU tariff had onstrained Turkey in its tariff setting, this ould have had a ausal impat on protetion via NTMs on non-eu exporters. They find evidene of poliy substitution between tariff ommitments imposed via the WTO and the PTA with the EU and the inreasing probability of Turkish NTMs. An important role in the ountry s deision regarding the extent of the use of NTMs is aredited to the market power it has over its trading partners. Broda, Limao and Weinstein (2008) demonstrate

7 Tariffs and non-tariff measures: substitutes or omplements that signifiantly higher NTMs are used in import-ompeting setors, where there is a greater ability to affet foreign exporters pries. Moreover, ountries with suffiient market power, or even small ountries (mainly non-wto members), who in ertain produts fae lower export supply elastiities (inelasti supply), will harge higher tariffs and also be driven to fore a more protetionist approah by imposing NTMs. Aisbett and Pearson (2012) establish the substitutability orrelation between SPS measures and tariffs, by suggesting that ountries manipulate their environmental and health standards for protetionist purposes. The authors laim that there is a rae to the bottom, meaning that tariff liberalization puts downward pressure on standards in ountries whih already have low standards (namely developing ountries) and upward pressure on ountries with high-standards (developed ountries). Bagwell and Staiger (2014) introdue the globalization fatigue hypothesis and laim that developed ountries suffer from insuffiient bargaining power in the multilateral and bilateral negotiations arena, relative to developing ountries. That, in tur may trigger them to tighten the imposition of NTMs in order to form trade poliy spae (or make room) for future negotiations with developing ountries. Using data on speifi trade onerns (STCs), Beverelli, Boffa and Kek (2014) find lear substitutability between past redutions in applied tariffs and SPS measures in both developed and developing eonomies, and same negative orrelation between tariffs and TBTs only in developed ountries. In line with these results, however, from the exporting ountry perspetive, Orefie (2015) shows that lowering tariff by 10% orresponds to a 0.18% and 0.36% higher probability of observing an STC on SPS and TBT respetively. This figure is even further magnified when the raising and the imposing ountry belong to the same inome group. While interest in studying TBT and SPS has been growing in reent years, muh of the relationship between tariff liberalization and the detailed level of speifi NTMs fous mainly on anti-dumping (AD) measures. Moore and Zanardi (2011), who study how past trade liberalization impats the deision to adopt antidumping laws, demonstrate that exept for heavy users of AD among the developing ountries there is no statistially signifiant substitution effet of trade liberalization on AD initiatives. By ontrast, Feinberg and Reynolds (2007), who based their analysis on 24 ountries for the period from 1996 to 2003, identify that tariff redutions inreased both the likelihood and number of AD petitions, espeially for developing ountries. Their onept of quid pro quo implies a quiet agreement whih exists between nations on swithing traditional tariff poliies by NTMs. They onlude that multilateral trade redutions are the ause of the reent growth in new users of anti-dumping poliies. Bown and Tovar (2011) reaffirm the substitution effet by analysing India, whih reently beame a heavy user of antidumping measures. They demonstrate how India s liberalization reforms in the early 1990s have resulted in a higher probability of AD filings and inreased safeguard restritions. In the reviewed literature it is not always evident that substitutability between tariffs and NTMs prevails. At times, the two measures follow the same diretio i.e. when ountries harge high tariffs, they also employ high NTMs and vie versa. Support for the existene of omplementarity orrelation is predominantly aredited to the neessity to protet domesti prodution of sensitive onsumer goods, textiles, apparel, and agriulture (Ray 1981). Moreover, Trefler (1993) and Lee and Swagel (1997) provide evidene for the positive orrelatio showing that the measures are often used together to inrease the protetion granted to import-ompeting setors. Both argue that NTMs are less likely to be imposed on export-oriented industries, at least partly beause of fear of foreign retaliation. Trefler uses a two-equation strutural model of the determinants of NTMs and imports aross US industries,

8 52 E. Ronen taking into aount variables suh as import penetration and fators like apital and labour. Lee and Swagel (1997) use disaggregated ross-ountry, ross-industry data for wages, produtio trade barriers and trade flows of manufatured goods aross 41 ountries. After aounting for industry and ountry- -speifi fators, ountries tend to protet espeially the weak industries, delining setors, politially important setors or those threatened by import ompetition. Lee and Swagel onlude that the ausality between the motives for using the trade measures might be reversed, or that trade barriers ould influene industry onditions rather than poliymakers responding to industry-speifi alls for protetion. Both papers argue that import penetratio or its growth, is positively orrelated with the willingness of poliymakers to impose NTMs. Dean et al. (2009), in their ross-ountry analysis, find that in the ase of fruits and vegetables, as well as for apparel produts, the joint use of tariffs and NTMs signifiantly redues the impat of NTM on prie. They use ity level retail prie data to diretly estimate the average impat of ore NTMs on the pries of 47 onsumer produts grouped into four separate setors for more than 60 ountries in Their model attempts to explain the observed prie gaps due to NTMs, given observed differenes in loal markups, transport osts and differenes in tariffs, in addition to some random unexplained fators. Moreover, they suggest that in some setors, the restritiveness of NTMs is highly orrelated with ountry inome; however, they do not provide interpretations as to this relationship. Essaji (2010) proves that the motivation to inrease the use of TBTs reflet the growing awareness of onsumption externalities. His onlusion puts a question mark over the aspiration of governments to protet domesti firms profits in a tariff onstrained environment. Essaji uses a two-ountry Cournot duopoly model in order to demonstrate how governments will hoose to inrease tehnial regulations, on ondition that the net marginal benefit of the regulation inreases with falling tariffs. It further reinfores the intuition that tariffs and regulations may be omplements, in ases where tariffs fall and ause a substantial inrease in the onsumption externalities. As already see the literature review provides various viewpoints regarding the orrelation between partiular NTMs and individual ountries or ountry groups. Yet it laks an overall perspetive along with an in-depth analysis of the determinants of NTMs, and the orrelation between all NTMs with their orresponding tariffs, while differentiating ountry groups. The urrent paper aims to address this gap by adding to the existing literature an analysis of the orrelation between the use of both trade poliy measures, while unovering the heterogeneity aross ountries. In partiular, it will provide evidene for the diminishing negative trajetory of the orrelation between the two import measures, with the rise in the ountry s level of eonomi development. Furthermore, it will supplement ommon knowledge by providing additional determinants for the use of NTMs, suh as the overhang gap, whih an further explain the motivation behind their ourrene. 3.2 Quantifiation of NTMs The quantifiation of NTMs is required in order to measure their impat on trade restritiveness and on welfare as well as their relation with additional maroeonomi variables. Unlike tariffs, for whih the available quantitative databases enable the evaluation of their levels and hanges, NTMs are muh more hallenging to quantify. The main explanation for this diffiulty is that NTM restritions may take many different forms and often the information regarding these measurements are not publily

9 Tariffs and non-tariff measures: substitutes or omplements available or satisfatorily transparent. In order to better investigate aspets regarding the impat of NTMs, one must use reliable methods, whih allow the transformation of qualitative praties into measurable quantities. Generally, these methods allow the alulation of the ad valorem equivalents of NTMs, i.e. the ad valorem tariff rate that would indue the same level of imports. Approahes for quantifiation of NTMs Along the years, several analytial approahes were developed in order to takle the hallenging task of quantifiation of NTMs. The approahes use various methodologies to overome the heterogeneous nature of NTMs and partiularly the lak of available detailed information regarding their implementation aross produts. The approahes differ in the various assumptions adopted, as well as by the eonometri tools whih were used for estimating their value, onlusiveness, and hanges over time. The most ommon methodology is the inventory approah presented by the frequeny index and the overage ratio. This approah allows the quantifiation of the inidene of NTMs and aptures the perentage of produts that are subjet to one or more NTMs in the ase of the frequeny index. The overage ratio basially measures the perentage of imports that are subjet to one or more NTMs. An additional approah, alled the prie (or wedge) gap, approximates the degree to whih a speifi regulatory measure or poliy intervention raises domesti pries above international pries. These approximations are built on omparing pries of goods affeted by an NTM with goods unaffeted by the NTM. The main disadvantage of the prie gap method is that it is often diffiult to reate two prie measurements for the same good and establish that one fully reflets the effets of an NTM, whereas the other is unaffeted. Several prie-based eonometri tehniques attempt to build on the foundations behind the prie gap method and expand it to several ountries and produts simultaneously. An example for suh an exerise on various OECD ountries is found in Deardorff and Stern (1997). Alternatively, the literature proposes the quantity-based eonometri approahes, whih allow the estimation of the impat of trade poliies, suh as NTMs, on trade flows. These methods usually employ analysis of trade data using the gravity model, fator ontent model or ombinations of features from both models. The trade data may be omprised of import values or quantities, or similarly, export measurements. Helpma Melitz, and Rubinstein (2008) used the famous gravity equation in their empirial framework, in whih given estimations of import demand elastiities serve to drive prie effets or ad valorem equivalents of NTMs. Notwithstanding the many advantages of these approahes in the quantifiation of NTMs, they also attrat ertain ritiism. The most ommon ritique is that estimations of NTMs should be rafted with detailed knowledge of produts and markets. Ferrantino (2006) adds that estimations should be done while analysing one produt and ountry at a time; however, the olletion of data requires exessive resoures and often disallows the apturing of many produts and ountries at one. This leads to a trade-off between handiraft and mass-produed estimates of NTM effets. Moreover, in reent years, thanks to tehnial improvements some omputable modelling simulation methods were developed in order to measure the effets of hanges in NTMs on a wide variety of eonomi parameters. These simulations were originally developed in order to estimate the effets of poliy hanges suh as tariffs on pries, produtio or on maroeonomi indiators suh as GDP or welfare. The most familiar example in that respet is the omputable general equilibrium

10 54 E. Ronen (CGE) simulation employed by the global trade analysis projet (GTAP). Lastly, several external soures of informatio suh as business surveys, may omplement the approahes mentioned above. These surveys aim to address the diffiulties that exporting firms fae, mainly in developing ountries. An example is the International Trade Centre (ITC), whih assembles responses from firms to the most burdensome NTMs and ways in whih they are affeted. Additionally, the CoRe NTMs database assembles information from various soures, among them, the US Trade Representatives National Trade Estimate Reports on Foreign Trade Barriers and the EU s Market Aess Trade Barriers database. Ad valorem equivalents of NTMs by Kee, Niita, and Olarreaga The model that will be employed in the urrent paper uses observations from a dataset of ad valorem tariff equivalent of non-tariff measures (AVE of NTMs) developed by Kee, Niita, and Olarreaga (2009). The authors estimated the AVE of NTMs using a quantity-impat approah ombined with approximations of import demand elastiity of nearly 5,000 produts in 104 ountries. Their non- -linear least square estimatio whih is based on the gravity model, allows the apture of information regarding the impat of various NTMs on eah ountry s imports. The NTMs inlude, among others, prie ontrol measures, quantitative restritions, monopolisti measures, anti-dumping and ountervailing measures, tehnial regulations and agriultural domesti support. The dataset onsists of ad valorem equivalents of NTMs, speified at the tariff line level (6 digit harmonized system of lassifying goods). At the outset, the authors use the following equation: ln m Σ = α α β ε k Core DS n + kc + βn, Core + ln DS + 1 ln( + tn ) + μn,, (1) m n, is the import value of good n in ountry evaluated at exogenous world pries, whih are all normalized to unity so that imported quantities equal m n,, t is the tariff and ε is the import k demand elastiity, whih was obtained extraneously, C is a set of variables that ontrol for k fator endowments (agriultural land, apital, labour fore, GDP, et.). The effet of ore NTMs at the ountry level is estimated by the interation term between the NTM dummy (for the presene of an NTM) and the vetor of fator endowments of the ountry C, while DS denotes the agriulture domesti support given to a produt. α are tariff line dummies that apture any good-speifi effet, while β is the parameter that aptures the impat that the NTM imposed on good k in ountry i has on the orresponding imports. k At the seond stage, the estimators are transformed into prie equivalents, using the elastiities of import demand. ln m ln Core = ln m ln p d ln p d ln Core = ε ave Core (2) ave Core = ε Core β 1 e ln m Core = 1 ε (3)

11 Tariffs and non-tariff measures: substitutes or omplements The main advantage of the model is that it goes beyond the traditional approah relying on overage and frequeny indies. It is far more informative than other gravity-based approahes, whih have dominated the evaluation of the effets of speifi NTMs. Furthermore, the methodology allows a diret omparison of the relative effets on imports of removing tariffs versus removing NTMs, and providing a partiular tariff equivalent of NTMs affeting eah produt, at any ountry. 4 Eonometri methodology and data In this researh paper, the determinants of the government s use of NTMs are examined as a ompensative reation to the existing low levels of MFN applied tariffs. Moreover, the Government s deision regarding the magnitude of the imposition of NTMs is studied with respet to the extent of the binding overhang, as well as to several supplementary ontrol variables, whih are desribed below. The size of the sample whih was developed for this purpose is omprised of approximately 200,000 observations, enompassing data on 61 ountries and between 3,500 to 4,500 tariff lines per ountry. The reason for the exlusion of ertain ountries from the original dataset is the lak of information on ertain independent variables. Moreover, it allows some of the explanatory variables to interat with eah other, and improve the goodness of fit of the model. The sample used in the estimation inludes a non-negligible part of the world eonomy and therefore, it is representative of the world as a whole in some dimensions. The sample omprises 61 ountries from all ontinents, while the average per apita GDP in the sample is USD 7,300, whih is slightly higher than the world average of USD 6,400. The list of ountries omprises 75% of the world s population and lose to 80% of its GDP (in PPP). This is due to the fat that it inludes 14 out of the world s 20 largest eonomies, among them the USA, Japa Frane, China, Italy, Canada and others. The eonometri methodology used in this analysis is the following: ln( 1 + AVE _ NTM ) = α + α ln( 1+ Tariff ) + α ln( 1+ OverHang ) + + α + α 4 i, n 7 i, n 5 6 ln( Openess ) + α ln( Cus _ Rev ) + α ln( 1+ Tariff ) ( D _ NonAgri) + i, n i, n i, n i, n i, n i, n ln( 1+ Tariff ) ( D_ Eon _ Dev) + εi, n i, n i, n i, n i, n (4) For the purpose of this study, a log-linear transformation of the ordinary least squares (OLS) model has been employed. The dependent variable used in all the speifiations is ln(1 + AVE_NTM i,n ), whih is the natural logarithm transformation of the ad valorem equivalents of NTMs. It is important to note that the number 1 has been added in order to ontrol the zero values. As mentioned previously, the dataset on the AVE of NTMs was obtained from the eonometri estimations of Kee, Niita and Olarreaga (2009). Although the authors use information on NTMs from the period 1992 to 2002, the data on the AVE of NTMs employed in the urrent analysis is for one partiular year during 2001 and 2003, sine the original estimation builds on the average trade data between the latter years. Eah observation in the urrent analysis represents a tariff equivalent of NTMs in a speifi ountry i on tariff line n. Regrettably, sine the original dataset on the AVE of NTMs is speified exlusively for one year per ountry, it laks the time-series dimension needed for reating a panel data analysis. Consequently, the eonometri approah used in this analysis uses a ross setional dataset based on the ross-ountry produt-level.

12 56 E. Ronen The eonometri analysis is omprised of a vetor of variables, whih may aount for ontrol variables explaining the use of NTMs. The main ontrol variable in the analysis is denoted as ln(1 + Tariff i,n ), whih is a vetor of the most favoured nations (MFN) applied tariff rates. Similar to the alulation of the dependent variable, the number 1 has been added to the tariff values in order to ontrol the zero values. Data for the variable is provided for eah of the ountries and expressed at the 6-digit HS level. The data soure is the UNCTAD trade analysis and information system (TRAINS) database aessible via the World Bank, world integrated trade solution (WITS) software. In order to maintain the required onsisteny with the methodology used for the alulation of the AVE of NTMs, the tariffs used are for the most reent year for whih data is available between 2001 and As the theory predits, the orrelation between Tariffs and NTMs is expeted to be negative for the whole sample of ountries. However, the heterogeneity among ountries when grouped aording to their level of eonomi development is preditably signifiant. Here, it is antiipated that the less developed group of ountries will tend to substitute their tariffs with higher levels of NTMs. As the level of eonomi development inreases, the substitutability dereases and the orrelation beomes positive for the rihest ountries. The seond major ontrol variable used in the analysis is tariff binding overhang (denoted as ln(1 + OverHang i,n ), whih represents the differene between the MFN applied tariff and the bound tariff. Data on the binding overhang was olleted from the world integrated trade solution (WITS), whih provides data from two different soures of WTO and TRAINS database (maintained by UNCTAD). Similar to the alulation of the dependent variable, the number 1 has been added to the overhang values in order to ontrol the zero values. In the analysis, a disaggregated data on binding overhang was used at a 6-digit level of HS to be ompatible with the data on the AVE of NTMs and Tariffs. It measures the degree of flexibility available in eah ountry within its WTO obligations and often-alled tariff water. The exess binding overhang is generally low in developed ountries and in manufaturing setors; however, it may reah very high levels in developing eonomies or in agriultural produts. These stylized fats are well demonstrated in the desriptive statistis, where the more advaned the eonomy, the signifiantly higher the MFN bound rates are than the MFN applied rates (i.e. higher binding overhang). Moreover, it is expeted that in ases of high binding overhang rates, ountries may raise tariffs legally without breaking their WTO ommitments, hene they will not alternate them by using NTMs. A third ontrol variable is Openness, whih represents the share of trade in goods and servies as a perentage of GDP. The data is olleted from the Global Development Network Growth Database, for the similar speifi year of the previous variables. This ratio is frequently used to measure the importane of international transations relative to domesti transations. Although this ratio is referred to as trade openness, the term openness may be slightly misleading, sine a low ratio does not neessarily imply high barriers to foreign trade, but may refer to fators suh as the size of the eonomy and geographi remoteness from its urrent and potential trading partners. It is likely that ountries that are dependent on international trade (mostly the more developed ountries) will employ fewer NTMs ompared to ountries that are more self-reliant. The next variable Cus_Rev represents the share of revenues of ountries from imports taxes, as a perent of their total tax revenues. The data soure is the World Bank and refers to eah of the ountries aording to the year speified. In general, NTMs differ from tariffs by the fat that these measures do not generate revenues to ountries. Nevertheless, this analysis allows

13 Tariffs and non-tariff measures: substitutes or omplements an examination of whether losses of tariff revenues to ountries ould explain the motivation for inreasing their AVE of NTMs. It is expeted that the low-inome ountries that rely on import revenues are less likely to use NTMs sine those are signifiantly less of a soure of inome. Several dummy variables are inluded in the eonometri analysis. These dummies are mostly used in the interation terms in order to validate the hypothesis that substitutability dereases with the rise in GDP per apita. Moreover, these dummies allow an examination of whether there is a substantial differene between the orrelation in tariffs and the AVE of NTMs for the agriulture produts ompared to non-agriulture produts. The first dummy variable is D_Eon_Dev, whih refers to the level of eonomi development of ountries (i.e. GDP per apita, aording to the lassifiation of the World Bank). It takes the following values: 0 = low-inome ountries; 1 = lower-middle inome ountries; 2 = upper-middle inome ountries; 3 = high-inome ountries. The dummy D_Eon_Dev is used in the analysis as part of an interation term, together with ln(1 + Tariff i,n ). The reason the interated variables were also not inluded separately is that the effets of these variables are aptured by ountry dummies. By reating the interation term, it allows the apture of the partiular orrelation between tariffs and NTMs of eah ountry group. By estimating the oeffiients of the 4 groups, differentiated orrelations reinfore the hypothesis that the less developed the ountries are, the more likely that NTMs are served in order to substitute for tariff protetion. Additional dummy variables are D_Non_Agri, whih is a binary dummy whih takes a value of 1 if the tariff line is assoiated with a non-agriulture produt, and equals 0 if the tariff line refers to agriulture produts. D_EU is also a binary dummy that takes a value of 1 if the ountry is a member state of the European Unio and equals 0 otherwise. Although at the time that the researh was onduted there were 28 member states, sine the data refer to earlier years (i.e is the latest), and some data on NTMs were missing for some member states, only 10 member states are inluded in the sample. D_OECD is a binary dummy that takes a value of 1 when the ountry is a member of the Organization for Eonomi Cooperation and Development (OECD), and 0 otherwise. The sample used in the analysis is omprised of 15 OECD members, and the soure of the information is the OECD website. Both EU member states and ountries belong to the OECD are the more advaned eonomies, and as suh it is likely that they are strongly ommitted to the WTO s objetives to failitate trade, hene they impose relatively lower NTMs. 5 Estimation results 5.1 General regressions results The results of the regression analysis for the entire sample of ountries are presented in Table 3. The first two olumns report the findings of the basi speifiatio whih inlude simply the main explanatory variables Tariffs and OverHang, while both ountry and produt fixed effets are ontrolled simultaneously. Country fixed effets allow to ontrol for the fat that some ountries may have higher levels of both tariffs and NTMs than others due to their stronger partiipation in multilateral and bilateral trade agreements. Controlling for tariff line fixed effets indiates that some produts may differ by their levels of both tariffs and NTMs due to domesti politial eonomy strengths. From the third olumn onwards, additional plausible ontrol variables whih potentially determine

14 58 E. Ronen the prevalene of NTMs, are inluded. This was designed in order to test for the sensitivity of the results to the inlusion of additional explanatory variables, suh as trade openness and ustoms revenues. As a aveat, while these orrelations are relatively robust, eonometris alone annot exlude the possibility of reverse ausality between partiular variables (i.e. trade openness) and the dependent variable. Therefore, an additional analysis of the partial orrelation has been performed in order to prove the strength and diretion of the orrelation between the variables employed in the analysis. It shows that whilst ontrolling for the effet of other variables, the orrelation is suffiiently signifiant and robust (the results are shown in tables 5 and 6). In olumn 4, the evidene is provided for the oeffiients of tariffs and binding overhang for non-agriultural produts relative to agriulture produts. Column 5 presents an interation term of tariffs and D_Eon_Dev, whih shows the distintion between different ountry groups aording to the level of eonomi development (i.e. GDP per apita). The goodness of fit of the model is satisfatory as approximately % of the variations in the NTMs are explained by the regressors speified. Sine the primary fous of this paper is to unveil evidene of a substitution effet between the applied tariffs and restritiveness of NTMs, attention should first be drawn to the two ontrol variables Tariff and OverHang. The findings are onsistent with the expetations by most of the literature of trade protetion theory. It shows that from an overall perspetive whih overs all ountries in the sample the predetermined tariffs have a statistially signifiant and negative orrelation with the import poliy to impose NTMs. The result reinfores the idea that although ountries agreed to lower tariffs during the Uruguay Round negotiations, they subsequently replaed some of this liberalization with a stringent NTM regime. The oeffiient for the model estimating this orrelation to the whole sample of ountries implies that a 1% tariff derease, leads to a 4.7% higher AVE of NTMs. In terms of eonomi magnitudes, it suggests that an inrease in tariffs from 1% to their mean level of 9.3% (an 830% inrease) dereases the AVE of NTMs by 39%, whih is a onsiderable impat. The degree of flexibility provided by the tariff binding shedules is onsiderably different aross ountries. It is generally lower in developed ountries and in the manufaturing setor but reahes high levels in developing eonomies or in speifi agriultural produts. In olumn 2 the ontrol variable OverHang is introdued and while the oeffiient of tariffs (first row) does not hange, the estimator for the binding overhang is found to be high and inversely assoiated with the use of NTMs. This learly reinfores the notion that a bigger overhang gap allows ountries to legally raise tariffs without breaking their WTO ommitments. Without suh a neessity, ountries with a bigger binding overhang are less likely to substitute tariffs with NTMs. From olumn 3 onwards, an additional set of variables is revealed in order to aount for other determinants of the use of NTMs. These ontrol variables inlude ln(openness), whih denotes the level of partiipation in world trade; ln(cus_rev), whih is the share of revenues from imports, expressed as a perent of total tax revenues. Moreover, the speifiation of these models inludes an interation term between the ountry s GDP and a dummy for eah setor (HS setion lassifiation). This interation term allows to ontrol for similar politial eonomy influenes of same size eonomies at the setoral level. As politial eonomy theory suggests, speifi interest groups may push governments to impose higher NTMs ompared to other industries whih are less organized. Adding this term inreases the goodness of the whole model, from R-squared of 35.6% to 37.5%. Notie that the tariff oeffiient remains relatively unhanged and signifiant at the one perent level, while it slightly inreases the negativeness of the seond ontrol variable OverHang.

15 Tariffs and non-tariff measures: substitutes or omplements Countries differ by their degree of reliane on international trade, and more speifially developed ountries, whih depend more on trade, generally impose on average fewer NTMs. By reduing their overall barriers to trade in their own borders, these ountries gain better market aess and enhane their ability to penetrate easily to their ounterpart markets. Moreover, the fear of retaliatory measures on the foreign demand for their exports similarly results in fewer NTMs. Other ountries, typially the less developed ones, demonstrate a more self-reliant trade poliy based on domesti produtio and therefore these ountries prefer to inrease their barriers to protet sensitive industries from foreign ompetition. Consequently, as predited, the estimators of the ontrol variable Openness are found to be negatively orrelated and statistially signifiant with AVE of NTMs, at the 1% level. The oeffiient is found to be 0.3%, meaning that an inrease of 1% in the share of trade in goods and servies (as a perentage of GDP) will result in an % derease in the use of distortive NTM restritions. The results are in line with Mansfield and Bush (1995), who use a different ontrol variable, i.e. the ratio of a ountry s imports to world imports, for measuring ountries partiipation in world trade. The findings are also onsistent with Mihalopoulos (1999), who notes that frequeny ratios of quantity and prie ontrol measures tend to be higher in ountries with lower levels of per apita inome and lower degrees of openness. When explaining the restritiveness of NTM restritions, the oeffiient on the share of revenues from imports (perent of total tax revenues) is found to be statistially signifiant, negative and relatively small. The negative sign of the oeffiient indiates that an improvement in the ountry s revenue from tariffs indues a fall in NTM protetion. The oeffiient shows that the AVE of NTMs are inversely assoiated with hanges in the share of revenues from imports and that a 1% rise in this variable indues a 1.1% fall in NTM protetion. The higher the import revenue as a share of total tax revenues, the less likely a ountry is to apply NTMs (and vie versa: a lower share of revenues from ustoms is assoiated with a higher prevalene of NTMs). Sine NTMs differ from tariffs by the fat that these measures are not designed to generate revenues to ountries, it seems worrisome that losses of revenues are in fat a soure of onern for ountries and a motivation for an inrease in the imposition of NTMs. Naturally, a onsiderable diversity among ountries onsists in the sensitivity towards loal produers interests. Some may favour proteting import-ompeting industries at the expense of the exporting setors. This is partiularly predominant when protetion of agriultural produts is involved. Evidene shows that tariffs on agriultural produts are on average muh higher than those on non- -agriultural produts, although there is onsiderable divergene between ountries. Countries with a speial interest in agriultural produts would most probably pair their high level of applied tariffs with a more regulated trade regime. This way, ountries may keep all options on the table, meaning that when global pressure rises to remove one form of protetio they ould still keep the other measure ative. An empirial validation is given in olumn 4 displaying a omparison between the orrelation of tariffs and the AVE of NTMs for non-agriulture produts relative to agriulture produts. This omparison yields the result that a 1% redution of tariffs on non-agri-produts generates a rise of 5.7% in the restritiveness of NTMs, and this oeffiient is statistially signifiant. The onlusion is that while the two import instruments are found to be paired in the two ases, the substitutability is stronger for non-agriulture produts ompared to agriulture produts.

16 60 E. Ronen 5.2 Correlation aross ountry groups The most important results of the model are presented in olumn 5 of Table 3. The speifiation of this estimation introdues a supplementary interation term between ln(1 + Tariff i,n ) and the dummy variable D_Eon_Dev. The objetive of this exerise is to allow to differentiate between ountry groups while ontrolling for the heterogeneity driven by the extent of eonomi development. The data is disaggregated into four groups; based on the World Bank lassifiation eah group represents a relatively homogenous luster of ountries, grounded on their level of GDP per apita. The results of the estimation suggest that ontrolling for the applied tariffs jointly with the level of eonomi development plays a substantial role in determining the orrelation between tariffs and the AVE of NTMs aross ountry groups. From a broad perspetive, the model provides evidene whih reinfores the substitutability between NTMs and tariffs for the majority of ountries. However, a deeper examination renders some additional onlusions about the trajetory of this orrelation in the light of the ountry variation. The results of the estimation validate the main proposition of the paper. They onfirm that although the orrelation between tariffs and NTMs is negative, the sale of substitutability diminishes gradually the more developed a ountry is. This trend ontinues until the negative orrelation hanges to positive for the group of high-inome ountries. Despite the fat that part of the estimators may be affeted by the sample size, the trend seems to be evident, and the estimations are statistially signifiant in all groups. The estimators indiate that a 1% derease in the tariff rate leads to approximately a 13.4% inrease in the AVE of NTMs in low-inome ountries. In lower-middle inome and in upper-middle inome ountries, a 1% derease in tariff rate is assoiated with an 11.4% and 9.4% inrease in the restritiveness of NTMs respetively. While in all ountry groups mentioned above, the estimators are found to be negatively orrelated and statistially signifiant, the orrelation between the two import measures is found to be signifiantly positive in the high-inome ountries. A similar derease in tariffs is assoiated with a dramati fall of 23.8% in the restritiveness of NTMs. These last results onfirm that in high-inome ountries, NTMs are more likely to be restritive, the higher the tariff on a produt is, and vie versa. The applied tariffs are relatively high in the least developed ountries; thus, the estimations imply that the restritiveness of NTMs in these ountries is typially low. Although the orrelation is found to be negative, the interpretation is fairly ounter-intuitive. The ost of operating an administration system whih deals with NTMs is usually high and it is naturally omplex. Moreover, these ountries rely on revenues generated from import taxes. Consequently, low-inome ountries use NTM restritions less extensively, and therefore the orrelation is negative. Nevertheless, the orrelation also implies that for speifi produts where tariffs are low, the imposition of some restritions on imports still exists, mainly in order to preserve domesti protetion. Sine most of the middle-inome ountries have undergone the deepest proess of tariff liberalizatio there is no wonder that they pratie NTMs more exessively. As supported by the politial eonomy theory, this negative orrelation is mainly a subsequent reation to tariff liberalization. Governments that are subjet to pressures from domesti interests, adversely affeted by the dismantling of tariff barriers, are often impelled to inrease alternative hannels of protetionism. That, in tur is demonstrated in the strong substitution orrelation between the two import poliy instruments. Claims against these ountries for using NTMs exessively are being rejeted by

17 Tariffs and non-tariff measures: substitutes or omplements the inreasing awareness of health and environmental auses. However, the negative and signifiant orrelation demonstrates the ompensative effet between the two import measures. As for the ase of the rihest ountries, it is evident that the orrelation is strongly positive, meaning that if a ountry belongs to the high-inome ountry group, it will most likely pair the two trade measures. Sine the average applied tariffs in high-inome ountries are relatively low, it implies a low restritiveness of NTMs as well. Several reasons ould support these findings. Firstly, these ountries are at the frontline of the WTO negotiations on TBT and SPS measures, and are generally more involved in designing large parts the world trade regime. This multilateral arena aims to ahieve the goal of failitating trade obstales, as well as simplifying market aess to the developing world. Therefore, the most developed ountries take upon themselves these objetives genuinely, whih naturally means removing trade barriers more extensively than elsewhere. Moreover, sine high- -inome ountries tend to rely more profoundly on international trade, it leads them to seek better onditions, both for domesti importers, as well as for their loal exporters in foreign markets. Lastly, these interests oexist with the neessity to attrat imports at the lowest osts possible for the benefit of domesti onsumers as well as for importers of intermediate produts. 5.3 Robustness heks and additional speifiation Several robustness heks were performed in this last part of the paper. These heks were designed to test the validity of the results obtained in the general model when estimating different ountry groups aording to geographial or organizational lusters. The results of these heks are robust and they reinfore the proposition of the paper. The developing ountries demonstrate a strong substitution orrelation in omparison to the rest of the world, while the rihest ountries suh as the EU member states omplement tariffs with an NTM regime. The results of the regression analysis are presented in Table 4. MENA. A speial geographial group that was examined is the Middle East and North Afrian group (MENA ountries). Despite the geographial and regulatory proximity, alongside inreasing trade relations with the EU, most of these ountries are still onsidered developing by nature. These ountries GDP per apita is relatively low and the level of tariffs is rather high. Therefore, it is interesting to hek whether the level of eonomi development influenes more of the MENA ountries NTM poliy, ompared with the mentioned proximities to the EU. The model estimate that tariff is inversely orrelated with the AVE of NTMs, suggesting that a MENA import line whose tariff is 1% higher than another MENA member s import line tariff has 6.7% lower AVEs for its imposed NTM. The estimator is found to be statistially signifiant; however, it is relatively smaller than that found for the low-inome ountry ase. The negative oeffiient implies that the orrelation between tariff and AVE of NTMs is substitutable and despite the low level of GDP per apita, the influene of the EU has a positive impat on the restritiveness of NTMs. Meanwhile, a MENA ountry tariff line whose binding overhang is 1% higher than other MENA ountry tariff lines exhibits 2.5% higher AVEs of NTMs relative to the other tariff line. OECD. The OECD organization is omprised of a heterogeneous group of ountries. However, these developed ountries typially have open eonomies and a relatively low level of applied tariffs. Despite some minor exeptions, this rule applies also to the low level of binding overhang.

18 62 E. Ronen The estimation for the tariff orrelation with the AVE of NTMs suggests that an OECD import line whose tariff is 1% higher than other OECD members import line tariffs has 15% higher AVEs for its imposed NTM. The estimator is found to be statistially signifiant, and the positive high oeffiient implies that the orrelation between tariff and AVE of NTMs is omplementary, for the reasons mentioned in the previous sub-setion. EU. The most remarkable positive orrelation between the two import poliy measures is found in the group of ountries that belong to the European Union. These ountries are onstrained to a single ustoms union regime, meaning that the MFN applied tariffs for all these ountries are similar for eah produt. The similarity, however, does not apply to all NTMs, whih vary to some extent aording to eah ountry s national regulations and domesti administration. The oeffiient in the ase of the EU demonstrates how an EU import line whose tariff is 1% higher than another EU member state s import line tariff has 25.3% higher AVEs for its imposed NTM. The estimator is found to be statistially signifiant, and the high positive oeffiient implies a strong omplementarity between the tariff and the restritiveness of NTMs. The fat that the EU does not substitute between the two import measures is fairly reasonable. The EU is part of an extensive network of PTAs, and the elimination proess of the restritiveness of NTMs started as early as in 1973, the year in whih all internal tariffs were abolished. Furthermore, the EU has been one of the promoters of trade failitation issues within the WTO and has taken upon itself the role of eliminating distortive measures to trade mostly, in favour of the developing world. With respet to the binding overhang, it is found to be inversely orrelated with the AVE of NTMs at the one-perent level. Sine the binding overhang of the EU is relatively small, this suggests that a 1% hange of the gap is assoiated with a 5.5% derease in the restritiveness of NTMs. 6 Conlusions The main objetive of this researh is to rejet the ommon hypothesis that non-tariff measures (NTMs) serve as protetionist instruments in light of the global obsolete use of tariffs in a similar manner aross all ountry groups. Moreover, it seeks to establish that the ommitment to lower all forms of trade barriers inreases with the ountry s level of eonomi development. In order to realize these objetives, an eonometri analysis has been undertake while exploiting a unique dataset of ad valorem tariff equivalents of NTMs at the 6-digit HS lassifiation. The empirial results reaffirm that from a global perspetive the orrelation between tariffs and NTMs is negative and statistially signifiant. Yet, the study asserts that while taking into aount the ountry variatio the substitutability effet diminishes with the rise in the ountry s level of eonomi development. This proves that the higher a ountry s GDP per apita, the less likely that it ompensates for the adverse effets of tariff elimination with a growing restritiveness of NTMs, while alternative onsiderations beome relevant. Furthermore, the relationship between the two import poliy measures hanges its sign from negative to strongly positive in high-inome ountries at the lowest levels. The paper argues that the most advaned eonomies have taken upon themselves stronger international ommitments than less developed ountries in order to ahieve protetionist relief in a genuine manner. Suh omplementarity is also attributed to the fat that these ountries are highly dependent on international trade, as well as to the response of poliy makers to onsumer s interests to freer trade. This outome is shown in the regression estimatio whih learly demonstrates

19 Tariffs and non-tariff measures: substitutes or omplements the signifiant extent to whih high-inome ountries tend to pair their low levels of tariffs with similarly less restritive NTMs. The empirial results of the analysis are further onfirmed by the robustness heks of the orrelation between the two import poliy instruments when different ountry lusters suh as the MENA ountries, OECD and EU member states are examined. Moreover, the analysis allows for additional ontrol variables to serve as determinants for the extent of the restritiveness of NTMs. Among these variables, an imperative role is given to the binding overhang gap and its effet on the AVE of NTMs. The binding overhang, whih represents the differene between ountries binding ommitments and their atual applied tariffs, is a signifiant poliy measure pratied by WTO member states. The results of the estimations validate that the less developed a ountry and the bigger the overhang gap, the less likely that the ountry imposes restritive NTMs. The negative estimators reinfore the notion that ountries that enjoy a larger binding overhang gap may raise their tariff legally without breaking their WTO ommitments, hene they do not need to resort to ompensatory measures suh as NTMs. The empirial framework employed in this study was based on a broad dataset, although it was limited in years. Therefore, in future studies and based on the availability of the AVE of NTMs it would be useful to investigate whether the findings reported in this paper held over reent years. Referenes Aisbett E., Pearson L.M. (2012), Environmental and health protetions, or new protetionism? Determinants of SPS notifiations by WTO members, Crawford Shool Researh Paper, 12 13, Crawford Shool of Publi Poliy. Bagwell K., Staiger R.W. (2014), Can the Doha Round be a Development Round? Setting a Plae at the Table, Globalization in an Age of Crisis: Multilateral Eonomi Cooperation in the Twenty-First Century, University of Chiago Press. Beverelli C., Boffa M., Kek A. (2014), Trade Poliy Substitution: Theory and Evidene from Speifi Trade Conerns, World Trade Organization. Bhagwati J. (1988), Protetionism, MIT Press. Bown C.P., Tovar P. (2011), Trade liberalizatio antidumping, and safeguards: evidene from India s tariff reform, Journal of Development Eonomis, 96(1), Broda C., Limao N., Weinstein D.E. (2008), Optimal tariffs and market power: the evidene, Amerian Eonomi Review, 98(5), Dean J.M., Feinberg R., Signoret J., Ludema R., Ferrantino M. (2009), Estimating the prie effets of non-tariff barriers, B.E. Journal of Eonomi Analysis and Poliy, 9(1), Deardorff A.V., Stern R.M. (1997), Measurement of non-tariff barriers, OECD Eonomis Department Working Papers, 179, OECD Publishing. Essaji A. (2010), Trade liberalizatio standards and protetio B.E. Journal of Eonomi Analysis and Poliy, 10(1), Feinberg R.M., Reynolds K.M. (2007), Tariff liberalization and inreased administrative protetion: Is there a quid pro quo?, The World Eonomy, 30, Ferrantino M. (2006), Quantifying the Trade and eonomi effets of non-tariff measures, OECD Trade Poliy Working Papers, 28, OECD Publishing.

20 64 E. Ronen Grossman G., Helpman E. (1994), Protetion for sale, Amerian Eonomi Review, 84, Helpman E., Melitz M., Rubinstein Y. (2008), Estimating trade flows: trading partners and trading volumes, Quarterly Journal of Eonomis, 123(2), Kee H.L., Niita A., Olarreaga M. (2009), Estimating trade restritiveness indies, The Eonomi Journal, 119(534), Lee J.W., Swagel P. (1997), Trade barriers and trade flows aross ountries and industries, Review of Eonomis and Statistis, 79(3), Limao N., Tovar P. (2011), Poliy hoie: theory and evidene from ommitment via international trade agreements, Journal of International Eonomis, 85(2), Mansfield D., Bush M.L. (1995), The politial eonomy of nontariff barriers: a ross-national analysis, International Organizatio 49(4), Mihalopoulos C. (1999), Trade Poliy and Market Aess Issues for Developing Countries, the World Bank. Moore M.O., Zanardi M. (2011), Trade liberalization and antidumping: Is there a substitution effet?, Review of Development Eonomis, 15(4), Orefie G. (2015), Non-tariff measures, speifi trade onerns and tariff redutio Working Papers, , CEPII Researh Center. Ray E. (1981), Tariff and non-tariff barriers to trade in the United States and abroad, The Review of Eonomis and Statistis, 63(2), Trefler D. (1993), Trade liberalization and the theory of endogenous protetion: an eonometri study of the U.S. import poliy, Journal of Politial Eonomy, 101, UNCTAD (2013), Non-tariff measure to trade: eonomi and poliy issues for developing ountries, UNCTAD/DITC/TAB/2012/1, United Nations Publiations. WTO (2012), Trade and publi poliies: a loser look at non-tariff measures in the 21st entury, World Trade Organization. WTO (2016), World Trade Statistial Review, World Trade Organization. Yu Z. (2000), A model of substitution of non-tariff barriers for tariffs, Canadian Journal of Eonomis, 33(4), Aknowledgements First and foremost, the author wishes to express his gratitude to Jan Jakub Mihałek, as well as to Jan Hagemejer, both of the Faulty of Eonomi Sienes, University of Warsaw, for the inspiring disussions and produtive feedbaks, ongoing interest and onstant enouragement. Furthermore, the author thanks Paola Cononi, Université libre de Bruxelles and Maurizio Zanardi, Lanaster University, for their onstrutive and helpful suggestions, whih signifiantly ontributed to this paper. The researh for this paper has been supported by the framework of the Dissettle projet. Dissettle: Dispute Settlement in Trade: Training in Law and Eonomis, Marie Curie Initial Training Network (ITN), finaned by the European Commission; IP: FP7-PEOPLE-2010-ITN.

21 Tariffs and non-tariff measures: substitutes or omplements Appendix Table 1 Variables and soures Variable Definition Soure AVE_NTM Tariff OverHang Openness Cus_Rev GDP GDPp D_Eon_Dev (dummy) D_Non_Agri (dummy) Tariff. Non_Agri D_OECD (dummy) D_EU (dummy) D_MENA (dummy) Country FE Produt FE Ad valorem equivalents of non-tariff measures Average applied tariffs (6-digit HS lassifiation) Gap between MFN applied tariff rate and bound rate (6-digit) Share of trade in goods and servies, as % of GDP Customs and other import duties, as % of tax revenue Gross domesti produt, urrent pries (USD billion) Gross domesti produt per apita, in urrent pries (USD) 0 = low-inome ountries 1 = lower-middle inome ountries 2 = upper-middle inome ountries 3 = high-inome ountries 0 = agriulture produts 1 = non-agriulture produts Interation term of tariff and D_Non_Agri 0 = non OECD ountry 1 = OECD member ountry 0 = non EU member states 1 = EU member states 0 = non MENA ountry 1 = MENA ountry Fixed effets of ountry variables Fixed effets of produts lines (6-digit HS lassifiation) the World Bank (Kee, Niita, Olarreaga 2009) Tariff Download Faility (WTO) Tariff Download Faility (WTO) Global Development Network Growth Database the World Bank International Monetary Fund International Monetary Fund the World Bank the World Bank OECD website European Commission

22 66 E. Ronen Table 2 Statistial desription Country group Variables Mean Standard deviation Min Max Observations Tariff ,742 AVE_NTMs ,742 OverHang ,202 Low-inome ountries GDP ,221 GDPp ,742 Openness ,742 Tr_Non_OECD ,827 Cus_Rev ,174 Lower-middle inome ountries Upper-middle inome ountries High-inome ountries Tariff ,474 AVE_NTMs ,474 OverHang ,768 GDP ,474 GDPp ,474 Openness ,474 Tr_Non_OECD ,080 Cus_Rev ,041 Tariff ,068 AVE_NTMs ,068 OverHang ,172 GDP ,522 GDPp ,068 Openness ,522 Tr_Non_OECD ,068 Cus_Rev ,984 Tariff ,135 AVE_NTMs ,135 OverHang ,852 GDP , , ,135 GDPp ,135 Openness ,135 Tr_Non_OECD ,589 Cus_Rev ,815

23 Tariffs and non-tariff measures: substitutes or omplements Table 3 Regression results; dependent variable: ln(1 + AVE_NTMs) ln(1 + Tariff) ln(1 + OverHang) ln(openness) ln(cus Rev) ln(1 + Tariff) D_Non_Agri ln(1 + Tariff) low inome ln(1 + Tariff) lower-middle inome ln(1 + Tariff) upper-middle inome ln(1 + Tariff) high inome Constant *** (-11.06) *** (-3.64) *** (-11.06) *** (-7.00) 0.147*** (8.04) *** (-11.56) *** (-17.45) *** (-3.60) ** (-3.83) 0.851*** (13.20) *** (-17.42) *** (-3.66) ** (-3.94) *** (-13.68) 0.854*** (13.24) *** (-14.72) * (-0.42) * (-1.36) *** (-6.18) *** (-22.24) *** (-8.80) 0.238*** (8.03) 0.724*** (11.10) Country FE Yes Yes Yes Yes Yes Produt FE Yes Yes Yes Yes Yes R Adjusted R F statistis Observations 206, , , , ,554 Notes: t statistis in parentheses; p-value: * p < 0.05, ** p < 0.01, *** p <

24 68 E. Ronen Table 4 Regression results, robustness heks; dependent variable: ln(1 + AVE_NTMs) MENA OECD EU ln_(1 + Tariff) *** (-6.56)*** 0.150*** (4.90)*** 0.253*** (2.84)*** ln_(1 + OverHang) 0.025*** (3.37)*** 0.006*** (1.73)*** *** (-7.49)*** ln_openness *** (-31.51)*** *** (-2.79)*** *** (-1.22)*** ln_cus_rev *** (-18.79)*** *** (-3.56)*** *** (-10.17)*** Country FE Yes Yes Yes Produt FE Yes Yes Yes R *** 0.338*** 0.457*** Adjusted R *** 0.338*** 0.456*** F statistis *** *** *** Observations 21,443 71,657 47,289 Notes: t statistis in parentheses; p-value: * p < 0.05, ** p < 0.01, *** p < Table 5 Correlations between variables (observations = 206,554) Variable ln(1 + AVE_NTMs) ln(1 + AVE_NTMs) 1 ln(1 + Tariff) ln(1 + Over Hang) ln(openness) ln(gdp) ln(gdpp) ln(cus_rev) ln (1 + Tariff) ln(1 + OverHang) ln(openness) ln(gdp) ln(gdpp) ln(cus_rev)

25 Average MFN applied tariff rates (unweighted) SPS & TBT notifiations Tariffs and non-tariff measures: substitutes or omplements Table 6 Partial and semipartial orrelations of ln(1 + AVE_NTMs) with observations = 206,554 Variable Partial orrelation Semipartial orrelation Partial orrelation ˆ 2 Semipartial orrelation ˆ 2 Signifiane value ln(1 + Tariff) ln(1 + OverHang) ln(openness) ln(gdp) ln(gdpp) ln(cus_rev) Figure 1 World average MFN applied tariff rates vs. SPS and TBT notifiations 16 % SPS notifiations TBT notifiations World average MFN applied tariff rates 0 Soure: alulations based on data of the World Bank and WTO I-TIP.

26 70 E. Ronen Figure 2 World average MFN applied tariff rates vs. SPS & TBT notifiations AVE of ore NTMs all lines % 50 TZA SDN 40 PHL SEN NGA DZA CIV EGY MYS 30 JOR 20 BRA MEX COL PRY RUS EU27 NZL JPN VEN 10 USA AUS NOR TUR CHN CAN SAU ISL ZAF 0 % Average MFN applied tariff rates Soure: alulations based on Kee, Niita, Olarreaga (2009).

27 Tariffs and non-tariff measures: substitutes or omplements Figure 3 Mean of ad valorem equivalent of NTMs aross ountries Mean ad valorem equivalent of NTMs Less than No data Soure: alulations based on Kee, Niita, Olarreaga (2009).