ARE THE UTILIZATION RATES OF

Page ǀ 1 ARE THE UTILIZATION RATES OF

U.S.

?

A

US

University of Groningen Faculty of Economics and Business

Master Thesis International Economics and Business

Student: Guillaume Durand Student ID number: s3095282

Student email: g.durand@student.rug.nl Date May 31st_{, 2017}

Supervisor: Dr. Tristan Kohl

Page ǀ 2

A

Since 1985, the U.S. have signed 14 trade agreements with 20 partner countries. Investigat-ing the efficiency of such agreements is paramount as negotiations keep beInvestigat-ing set up in the hope of implementing more trade agreements. The U.S. is no exception and its agreements will be stud-ied after a review of the literature on utilization of trade agreement. It appears that U.S. agreements follow the global trend of small utilization rates. Their efficiency is confirmed as trade is overall increasing thanks to the agreements but their utilization is dependent on many factors. Stemming from this literature on utilization rates and data restrictions, the factors deemed most influential are the preferential margin, exchange rate variations and eligibility to competing tariff saving schemes. After analysing each of these factors empirically it will emerge that eligibility to competing agree-ments and the preferential margin have a positive influence on TA utilization rates when measured as a discount, while the impact of exchange rate variations is null. After robustness check it will however appear that eligibility to other tariff saving schemes has a negative impact on utilization rates as is generally put forward in the literature. U.S. trade agreements are therefore similarly impacted by specific factors as other agreements studied in the literature.

Page ǀ 3

T

C

ABSTRACT ... 2

LEXICON ... 4

SECTION 1:INTRODUCTION ... 5

SECTION 2:LITERATURE REVIEW ... 6

1. DEFINING UTILIZATION RATES ... 6

2. PREFERENTIAL TARIFF MARGIN ... 7

3. RULES OF ORIGIN (ROOS) ... 8

4. FIRM SIZE AND TRADE VOLUME ... 10

5. EXCHANGE RATES AND THEIR IMPACT ON ROOS ... 11

6. DIAGONAL AND BILATERAL CUMULATION ... 12

7. ELIGIBILITY TO MORE THAN ONE TARIFF SAVING SCHEME ... 12

8. LEVEL OF INFORMATION AND TIME ... 13

9. PRESENCE OF LOW OR ZERO MFN TARIFF AND ACTIVE INVESTMENT SCHEMES ... 14

SECTION 3:METHODOLOGY ... 15

1. DATA ... 15

2. UTILIZATION RATES OF U.S. TRADE AGREEMENTS ... 19

3. INFORMATION AND LEARNING EFFECT ... 20

4. BILATERAL AND DIAGONAL CUMULATION ... 21

5. STRICT RULES OF ORIGIN ... 22

6. COMPLIANCE COSTS ... 23

7. LEVEL OF PREFERENTIAL MARGIN... 24

8. EXCHANGE RATES VARIATIONS ... 25

9. ELIGIBILITY TO OTHER TARIFF SAVING SCHEMES ... 26

10. ROBUSTNESS CHECKS WITH DIFFERENT SETS OF INDEPENDENT VARIABLES ... 28

SECTION 4:RESULTS ... 29

1. UTILIZATION RATES AND PREFERENTIAL MARGIN ... 30

2. UTILIZATION RATES AND EXCHANGE RATES VARIATIONS ... 33

3. UTILIZATION RATES AND GSP ELIGIBILITY ... 35

4. ROBUSTNESS CHECKS ... 36

SECTION 5:CONCLUSION ... 37

REFERENCES ... 39

APPENDIX 1.SUMMARY OF PAPERS CITED IN THE LITERATURE REVIEW ... 44

APPENDIX 2.SUMMARY OF THE U.S. TRADE AGREEMENTS AND THE DATE OF FIRST USE ... 49

APPENDIX 3.UTILIZATION RATES OF U.S. TRADE AGREEMENTS FOR U.S. IMPORTS ... 50

APPENDIX 4.CUSTOMS CLAIMS FOR U.S. IMPORTS BY TRADE AGREEMENT ... 51

APPENDIX 5.DESCRIPTION TABLES FOR FACTORS METHODOLOGY ... 54

APPENDIX 6.VARIABLES SPECIFICATION ... 55

APPENDIX 7.TESTS ON MAIN VARIABLES BEFORE THE REGRESSIONS ... 57

APPENDIX 8.U.S. IMPORTS CLAIMS ACROSS IMPORT PROGRAMS ... 58

APPENDIX 9.ELIGIBILITY TO THE U.S.GSP OVER 1989-2017 BY TA PARTNER ... 59

APPENDIX 10.COUNTRIES CLAIMING BENEFITS FROM CIVIL AIRCRAFT AND GSP SCHEMES ... 60

APPENDIX 11.RULES OF ORIGIN REQUIREMENT FOR IMPORTS BY TA ... 62

Page ǀ 4

L

Terms specifically related to the topic of Free Trade Agreements and their utilization rates will be defined in this lexicon to provide clear knowledge to the reader. This will also allow the writing to be unencumbered by constant definitions of the concepts used. Other concepts will be explained later in the research as their definition pertain to a referred study or a continuous stream of thought.

Free Trade Agreement (FTA): treaty (such as the North America Free Trade Agreement (NAFTA)) between two or more countries to establish a free trade area where commerce in goods and services can be conducted across their common borders without tariffs or hindrances but in which capital and/or labor may not move freely. Member countries usually impose a uniform tariff (called common external tariff) on trade with non-member countries. Such agreements will only be referred to as trade agreement or TA throughout this research.

Regional Trade Agreement (RTA): defined by the World Trade Organization as reciprocal trade agreements between two or more partners, they include trade agreements and custom unions. Most Favoured Nation (MFN) tariff rate: in current usage, MFN tariffs as defined by the World Bank, are what countries promise to impose on imports from other members of the WTO, unless the country is part of a preferential trade agreement (such as a free trade area or customs union). This means that in practice, MFN rates are the highest (most restrictive) tariffs WTO members charge one another.

Preferential tariff rate: as defined by the World Bank, show that virtually all countries in the world joined at least one preferential trade agreement, under which they promise to give another country's products lower tariffs than their MFN rate. In a customs union (e.g. the Southern Africa Customs Union or the European Community) or a free trade area (e.g. NAFTA), the preferential tariff rate is zero on essentially all products. These agreements are reciprocal as all parties agree to give each other the benefits of lower tariffs. Some agreements specify that members will receive a percentage reduction from the MFN tariff, but not necessarily zero tariffs. Preferences thus differ between partners and agreements.

Generalized System of Preferences (GSP): many countries, particularly the wealthier ones, give developing countries unilateral preferential treatment rather than reciprocal. The largest of these programs is the Generalized System of Preferences (GSP) which was initiated in the 1960s. The European Union, Japan, and the United States offer multiple unilateral preference programs, the EU's Everything But Arms (EBA) program is one example. Exporting countries may have access to several preference programs for a given trade partner and product.

Page ǀ 5

S

1:

I

According to the World Trade Organization records for 2016, there were 647 regional trade agreements notified to the organization across all countries. This increasing trend is coupled with booming trade and a rise in preferential trade regimes. Such a tendency is also found in the US, with the signing of 14 different trade agreements with 20 countries all over the globe between 1985 and 2012. The trade agreements (TAs) make up for reduced and/or eliminated tariffs on specific sectors and/or products. To this end, trade between the partner countries and the U.S. has globally been increasing over the course of the last decades. In 2016, the United States total trade volume was composed of USD 1.455 trillion for exports and USD 2.189 trillion for imports, creating a large trade deficit. Among these USD 2.189 trillion for U.S. imports in 2016, USD 749 billion were imported from trade agreement partners, that is to say a little above a third of all imports. Following these 14 different TAs, 47% of the U.S. goods for exports went to a TA partner country in 2015 for a total value of $710 billion.

Because tariffs are one of the main barriers to trade between countries, it could be expected that such trade agreements are highly utilized to reap as much benefit from a duty-saving scheme as possible. This aspect is studied through utilization rates of trade agreements which allow further investigation in the efficiency of said TAs. The literature related to these utilization rates of trade agreements however, shows that trade agreements have rather low utilization rates. This literature is globally focusing on large trading areas such as Asian economies, North America through NAFTA or the EU. This type of analysis could therefore be applied to the U.S. and its trade agree-ments in order to prove whether the U.S. trade agreeagree-ments’ utilization rates are indeed low and if that is the case, what affects their utilization the most.

As far as the extent of the literature review and the search for publications on the topic of TAs utilization rates, there is no research concerning the U.S. outside of NAFTA and the US-Chile agreement. Consequently, this research will be the first one to study all 14 TAs signed by the U.S. and therefore bring a clearer picture of their TA utilization rates. These trade agreements relate to 20 different countries: Australia, Bahrain, Chile, Colombia, CAFTA-DR (Costa Rica, Dominican Republic, El Salvador, Guatemala, Honduras and Nicaragua), Israel, Jordan, Korea, Morocco, NAFTA (Canada and Mexico), Oman, Panama, Peru and Singapore.

It will appear that utilization rates of U.S. trade agreements are comparable to other TAs across the globe even though certain rates differ from the tendency. The influence of the preferen-tial margin will prove to be positive and significant when measured as a share of the MFN tariff rate but not necessarily when measured as an absolute gap. However, the impact of exchange rates variations will remain unconfirmed while the eligibility to GSP will be shown to be positively related to utilization rates.

Page ǀ 6

S

2:

L

R

Utilization rates of preferential trade agreements are a fairly recent field of study as most of the trade agreements being used today were implemented during the past 20 years. The literature on the topic seem to be concentrated on utilization rates for Asian economies, the U.S. with NAFTA and the EU. However, such literature shows that utilization rates of TAs are remarkably low considering the potential benefits for firms involved in international trade. This literature ex-plains the low utilization rates with a number of factors: differences in the definition of the utiliza-tion rate, preferential tariff margin, rules of origin, firm size, trade volume, exchange rates and their impact on Rules Of Origins (ROOs), diagonal & bilateral cumulation, eligibility to more than one TA, level of information, time, presence of low or zero MFN tariff and active investment schemes. All these factors will be discussed individually in the following paragraphs to give a clear picture of their impact on utilization rates.

1. Defining Utilization Rates

There are several issues that the literature covers in order to explain these low utilization rates, beginning by the very meaning of the utilization rate concept. This definition has been an issue in the literature especially since variations of the concept itself could have the most important impact on the results presented. The general definition of this utilization rate, as given by Hama-naka (2013) and Hayakawa et al (2017) is the share of dutiable imports with a margin of preference that uses a TA as a part of total imports. However, this rate is neither very realistic nor very effec-tive to assess the usefulness of a TA as trade under a zero MFN tariff for example, is excluded entirely. This is why the usage rate is the most appropriate measure of the concept as it is the ratio of goods actually imported under the TA over the goods eligible for such TA preference. Such a definition is shared by Hayakawa et al (2014) when measuring utilization of TA by firms, (Nilsson, 2010; Nilsson, 2011; and Nilsson, 2016). In order to avoid confusion, this research will only use this concept under the definition of the usage rate.

Blind & Ziltener (2014) in their study of the Japan-Switzerland TA make a clear difference between several definitions regarding the utilization rate. They distinguish the General Utilization Rate (GUR) from the Adjusted Utilization Rate (AUR). The former only encompasses the share of total trade that benefits from reduced or eliminated tariffs from a trade agreement while the latter is defined as the value of trade benefitting from reduced or eliminated tariffs from the TA over the value of trade in goods included in the TA. This differentiation is made to show that the utilization rate of trade agreement could easily be underestimated or overestimated depending on the values involved in the denominator of the ratio.

Page ǀ 7 define as the ratio of total imports value eligible for any type of TA over the total value of imports eligible for TAs and MFN tariffs. They also define the potential rate of utilization of a country as the ratio of exports from a country to another, under a preference scheme over the total exports of the country to any destination. With these different definitions they find that TA utilization rates can vary between 50% and 86% proving that a reliable definition of the concept is crucial in stud-ying this concept.

Candau et al (2004), on their part, choose to define TA utilization rate as the ratio of the share of imports eligible for the regime over the share of imports that requested such a preference. But this definition is overshadowed by the fact that not all products for which a preferential treat-ment was requested actually go through with it as showed by Nilsson (2010). Kim & Cho (2010) also define the utility ratio of TAs as the ratio of exports which actually receive preferential tariff treatment in an importing country over the total exports from a certain country in a certain period. All these definitions pertain to the same literature but make results from different studies somewhat incomparable as their framework differ. Although there are different definitions of uti-lization rates, most papers seem to agree on the very definition that will be used in this research. Furthermore, the literature seems to agree on one factor: the preferential tariff margin is positively related with TA utilization rate.

2. Preferential Tariff Margin

Most of the literature on TA utilization rates indeed agree that the preferential margin has a positive impact on utilization rates. The main benefit of using a preference is the reduction in duties, in most cases to zero. Therefore, the higher the preferential margin, the higher the tariff saving opportunity and therefore the higher should be the probability that a preferential tariff rate is utilized. The preferential margin refers, as indicated in the Lexicon, to the difference between the MFN and the preferential tariff rates. Many scholars show that the TA utilization rate is closely linked with the preferential margin for different groups of countries and agreements (Candau et al, 2007; Hakobyan, 2015; Hamanaka, 2013; Hayakawa, 2014; Hayakawa et al, 2009; Hayakawa et al, 2014; Hayakawa et al, 2017; Keck & Lendle, 2012; Kim & Cho, 2010; Nilsson et al, 2009 and Nilsson, 2016).

As a consequence, TAs are expected to have higher utilization rates when their preferential margin is bigger. Reciprocally, TAs are expected not to be utilized at all when the margin is null considering the costs implied by the application for a preferential tariff rate. Incidentally, the coef-ficients for the tariff margin are expected to be significantly positive on utilization rates, which gives the first hypothesis of this research:

H1: Larger preferential margins and therefore potentially higher tariff saving opportunities are expected to have a significantly positive impact on utilization rates.1

1 _{Even though hypotheses are formulated for most of the factors stemming from the literature, it is}

Page ǀ 8 Subsequently, for the case of reduced preferential margin especially, the costs linked with the utilization of a TA are becoming very important for firms to decide whether they will use a TA or not. Such costs are mostly represented by rules of origin and incidentally certificates of origin.

3. Rules of Origin (ROOs)

Rules of origins are the requirements a firm must fulfill in order to benefit from the prefer-ential tariff rate of a TA. Ratananarumitsorn et al (2008) show that in most instances there are three categories of Rules of Origin (ROOs): the Wholly Obtained (WO) rule, the Value Content (VC) rule and the Change in Tariff Classification (CTC) rule. The WO rule is defined as a product that must be wholly obtained in the country to benefit from the preferential regime whereas the VC rule corresponds to a certain percentage of the product that must be local content, based on a value-added ratio of origin. On the other hand the CTC rule is defined as an originating product that must be transformed from a product in one tariff classification to a product in another classification according to criteria specified in the TA. Hayakawa et al (2017) use a similar definition of regional Value Content but also show that most of these content requirements are between 30% and 40% for the value added ratio. For the case of NAFTA, Carrère & De Melo (2004) include the ROO of TECH instead of the VC rule which is defined as the presence or not of a technical (TECH) re-quirement. In their econometrics analysis they show that an approximate 10% extra preference margin would be needed to compensate for the costs linked with a Regional Value Content (RVC) requirement. Moreover, Hayakawa et al (2016) show that in the case of RVC-related rules, indus-tries requiring higher levels of administrative costs through documentation exhibit lower utilization rates of TAs. Similar results are shown by Hiratsuka et al (2009) for Japanese firms and Nilsson (2016) for EU exports.

As shown by Candau et al (2004) and François et al (2006), ROOs were constructed in order to avoid trade diversion between countries. This possibility meant that firms could make their products go through different countries to benefit from tariff reductions or elimination without having any activity in the country other than transit. Kim & Cho (2010) give the example of Sin-gapore which is mainly used as a transit hub for products in Asia and due to these rules of origin, TAs with this country have, comparatively, very low utilization rates. However, rules of origin also show that such a case makes for less than optimal market situations, as a more efficient producer could exist but not be employed because he does not fulfill the rules of origin required to benefit from a preferential tariff scheme.

Hakobyan (2015) finds that products with higher local content share, therefore products that will fulfill ROOs more easily are linked with higher utilization rate of preference scheme in the case of the U.S. GSP (defined in the Lexicon). Hayakawa et al (2014) also show that ROOs restrictiveness is one of the main factors in TA utilization for the Korea-ASEAN2 TA. Because

2 _{Association of South East Asian Nations (ASEAN) is a free trade area of ten economies: Brunei,}

Page ǀ 9 such costs are important for firms, Demidova & Krishna (2008) demonstrate that the most produc-tive firms use TAs in their exports whereas the least producproduc-tive ones use general tariffs (MFN rate) because they cannot afford to pay the fixed costs linked with the request for the preferential tariff rate. Estevadeordal (2000) constructed a restrictiveness index for ROOs in a quantitative analysis to show the importance of potential cost-raising effects of restrictive ROOS. This index shows that products with the lowest levels of restrictiveness are expected to be more exported under the TA. Similar results are also shown in Kim & Cho (2010) for Korean TAs but they go further by proving that TAs would be more utilized if ROOs restrictiveness were lowered. Moreover, scholars studied rules of origin to another extent by showing that such rules make it harder for firms to make use of economies of scale as input requirements to benefit from the scheme may vary between destina-tions and therefore between agreements as shown in Candau et al, 2004.

To fulfill the requirements set by rules of origin, firms need to provide certificates of origin in order to justify that they are allowed to benefit from the preferential tariff regime. In their paper, Bureau et al (2007) show that the main reason for low utilization rates of trade agreements are the requirements of rules of origins. In order to benefit from the reduced or eliminated tariff, firms need to comply with certain types of requirements depending on the TA and the goods involved. This process involves certification and traceability of the goods, incurring administrative costs for the firm in question and a delay for the processing of the request. They later state that uncertainty regarding preferences eligibility of products and the risk of financial penalties if violations of com-plex rules are found, are also reasons for low utilization rates.

Anson et al (2005) and Bureau et al (2007) show that compliance costs to utilize a TA, including administrative costs, undo most of the preferential tariff access to the partner(s). More precisely, Anson et al (2005) show that margin of compliance costs for firms using TAs ranges between 3% and 5% of the trade value for the case of NAFTA. Furthermore, they conclude that for the case of North-South trade with preferential tariff access, it may actually offer very limited market access to the participating countries of the South due to these costs. Nilsson (2011) find similar results as high fixed compliance costs are one of the main deterrent to TA utilization. It was also shown that for the US, the higher the degree of processing in the production of goods, the more likely are firms not to use the TA because of the rising compliance costs (Bureau et al, 2007; Hakobyan, 2015).

Page ǀ 10 after deliberating on the application filed by a firm. Hiratsuka et al (2009) explain that the self-certificate system is advantageous because the administrative costs in terms of issuance fee are reduced even if the costs for verifications of such documents cannot be reduced. The hybrid system, used broadly by the EU in their TAs, shows approved exporters that are able to use the self-certif-icate system. These approved exporters are under a criterion of governance by an authority while the rest of exporters use the third party certificate system, it is hybrid in the sense that both systems are used simultaneously.

As a consequence, TAs are expected to have lower utilization rates when rules of origin are stricter. Such rules can encompass higher value-added ratio of origin requirement and thus make it more difficult for firms to be eligible for preferential tariff rates. Consequently, coefficients for rules of origin are expected to be significantly negative on utilization rates, giving the hypothesis:

H2: Higher value-added ratio requirement of ROOs are expected to have a negative impact on TA utilization rates.

Furthermore, TAs are expected to have lower utilization rates when their overall compli-ance costs are higher. As these costs cover all aspects incurred by a firm when requesting a prefer-ential tariff treatment, higher costs are reducing the benefit of the tariff saving opportunity. This factor is also expected to be most important in the case of small firms and/or small trade volumes as the tariff saving potential is minor in these cases. Consequently, coefficients for compliance costs are expected to be significantly negative on TAs utilization rates, giving the third hypothesis:

H3: Higher levels of compliance costs when requesting a preferential tariff rate are expected to negatively impact utilization rates.

Rules of origin and incidentally certificates of origin with their fixed costs are therefore recognized as a main cause for low utilization rates of TAs. However, the size of the firm involved in trade and its trade volume are also major determinant in TA use.

4. Firm Size and Trade Volume

Blind & Ziltener (2014), following the major part of the literature on the topic, find that large firms are the main users of trade agreements. This is mainly due to the fixed costs linked with using a trade agreement, that large firms are expected to be more able to cover than Small and Medium Enterprises (SMEs). This similar point was studied by Demidova & Krishna (2008) as stated earlier, with the most productive firms using TAs and the least productive ones using the MFN tariff rate, as this approach was first theorized by Melitz (2003). Hayakawa et al (2009) also show that the larger the affiliate or the more diversified the origins of products procurement, the more likely the firms are to use a TA. Hiratsuka et al (2009) also show similar results for Japanese firms.

Page ǀ 11 exports not using the preference regime falls drastically. Similar results were obtained by Hama-naka (2013) and Hayakawa et al (2014) in which a main reason for Asian firms not to use TAs is the small size of their shipments and therefore potentially low tariff saving benefits. Hayakawa & Laksanapanyakul (2017), Keck & Lendle (2012) and Nilsson (2016) find analogous results of larger exports trade values being linked with higher utilization of TAs or low utilization because of low duty-savings. Hiratsuka et al (2009) go further by showing that Japanese firms would only consider using a TA if the benefits (understood as the preference margin) are higher than 10% when compared to the MFN rate (this measure is the discount and will be used later in this re-search), considering the volumes and fixed administrative costs involved in the transactions.

The size of the firm involved or their trade volume is, as shown by these studies an im-portant factor of TA utilization. Nonetheless, as these firms trade internationally, exchange rates play a significant part in the value and value-added ratio of their transactions. Exchange rates are therefore also considered as a reason for low TA utilization rate.

5. Exchange Rates and their impact on ROOs

Hayakawa et al (2017) expose the impact of exchange rates changes on the utilization rates of TAs considering the importance of ROOs for the exports of ASEAN countries to Korea between 2007 and 2011. In such rules of origin, the value-added ratio measure is defined as one minus a ratio of non-originating input price to export product price, and is dependent on the exchange rate. A depreciation of the exporter’s currency against the importers’ enhances TA utilization because the value-added ratio requirement of rules of origin is improving. The authors prove these effects to be even stronger for higher levels of demand elasticity. As a consequence, variations in exchange rates bring uncertainty to the decision of using TAs as ROOs fulfillment is varying along with such exchange rates.

TAs are thus expected to have lower utilization rates when the country originating the im-ports have high exchange rate variations. Eligibility to the preferential tariff rate rests upon the value-added ratio of origin expressed in local currency therefore if this currency is highly volatile, firms are expected not to utilize a TA as much as a TA linked to a stable currency. Exchange rate variations create unpredictability in eligibility to the preferential tariff rate and probably overall lower utilization of the TA. Such variations are likely to reflect the economic situation of the coun-try, therefore GDP levels and growth will be controlled for in assessing this factor. As a conse-quence, controlling for GDP levels and growth, the coefficients of exchange rate variations are expected to be significantly negative on TAs utilization rates, giving the next hypothesis:

H4: More volatile exchange rates measured by their levels of yearly variations are expected to have a negative impact on TA utilization rates through harder fulfillment of ROOs.

Page ǀ 12 6. Diagonal and Bilateral Cumulation

In the case of rules of origin, Hayakawa et al (2013) and Hayakawa (2014) show that the difficulty to meet the requirements can be somewhat relaxed when there is a possibility for either diagonal or bilateral cumulation in the TA. Bilateral cumulation is defined as materials originating from one country that could be considered as originating from the partner country and vice versa. The value added ratio requirement for ROOs can therefore be reached more easily by firms trying to benefit from a bilateral TA. On the other hand, diagonal cumulation is defined as products orig-inating from one country that could be considered as origorig-inating from all the TA partner countries in the case of a multilateral TA. These cumulation possibilities understandably make local input requirements of ROOs easier to fulfill, for firms to benefit from a preferential tariff rate. For the case of Thai exports to Japan in 2010 studied by Hayakawa (2014), such schemes were linked with higher levels of trade, meant as trade under the specific TA, and that cumulation is positively re-lated with TA utilization. Moreover, Hayakawa shows that in the case of a TA between more than two countries without diagonal cumulation, there is little to no difference between this multilateral TA and multiple bilateral TAs.

Hence TAs with possibilities for bilateral or bilateral cumulation are expected to have higher utilization rates than TAs without these options. As these cumulation rules allow a product to be originating from either of the partner countries, it is therefore easier to reach the value-added ratio requirement to be eligible for the preferential tariff rate. This effect is expected to be even more powerful for multilateral agreements as products can be originating from a larger number of countries and amount for this value-added ratio. As a consequence, the coefficient for cumulation rules are expected to be significantly positive on utilization rates of TAs, hence the next hypothesis:

H5: bilateral or diagonal cumulation are expected to have a positive impact on TA utilization rates

The case of multiple TAs and rules of origin is also expressed in the possibility for products to be eligible for multiple TAs at the same time. Considering trade can only be conducted under one TA, this factor could lead to major underestimation of TA utilization rates and be of overall great importance for this study.

7. Eligibility to more than one tariff saving scheme

Hakobyan (2015) for the case of the U.S. GSP, shows that goods eligible for multiple trade preferences schemes are linked with lower utilization rates because goods can only be traded under one scheme. Similar results are found in Wignaraja (2014) and in Keck & Lendle (2012) and prove this point by using the examples of the ATPA3 and the GSP program for the US. They show that overall, 87% of U.S. imports from the ATPA countries that are eligible to at least one scheme enter

3 _{The Andean Trade Preference Act (ATPA) was enacted in 1991 between the U.S. and Andean countries}

Page ǀ 13 under ATPA, but only 3% under GSP. The remaining 10% enter under MFN. Taking into account only products that are covered by GSP, the utilization rate of the program is 25%, but 72% of GSP eligible imports actually enter under the ATPA, which brings the utilization rate of TAs (no matter which one) to 97%. Multiple TAs are also studied in Pomfret et al (2010) for the case of Australia with 6 different trade agreements. However the increasing number of zero MFN tariff rates con-cerning Australia is making the utilization rate of TAs decrease as the preferential margin is dras-tically reduced if not eliminated entirely.

As explained by Kawai & Wignaraja (2011), Asia suffers from the ‘spaghetti bowl prob-lem’ or ‘noodle bowl probprob-lem’. There has been in recent years a proliferation of TAs in Asia lead-ing to many agreements overlapplead-ing each other and offerlead-ing different tariff rates for the same products from the same country. This issue therefore created an unnecessarily complicated system of preferences between Asian economies. Stemming from the fact that products can only be ex-ported/imported under one TA, utilization rates for such agreement can be expected to be artifi-cially low when studied individually.

TAs are expected to have lower utilization rates when products are eligible for another preferential treatment scheme. As trade can only be conducted under one agreement or none at all if using the MFN tariff, individual utilization rates are expected to be lower for all agreements than in the case of a single TA available. Although the impact of eligibility can be ambiguous as more than one TA could raise awareness to any preferential treatment and thus increase overall utiliza-tion rates, individual rates are still expected to be lower. Therefore coefficients for the eligibility to several TAs are expected to be significantly negative on utilization rates, hence the next hypoth-esis:

H6: Eligibility to more than one preferential tariff rate is expected to have a negative impact on TAs utilization rates.

With eligibility to multiple TAs, firms could often lack the information required to make the best choice and benefit from the greatest preference margin between all agreements. Conse-quently, information and time are also found to be of importance in TAs utilization rates.

8. Level of information and time

Page ǀ 14 was not given immediately to all sectors after the implementation of the Chile-U.S. treaty showing that firms’ benefits are varying over time.

Hamanaka (2013) also shows that a major issue for the utilization rate of TAs is that recog-nition of the agreement may take some time. As a consequence, this information and understanding lag after the implementation may affect utilization rates for the first years of available data. Another aspect of time is, as shown by Hamanaka (2013) for the case of Asia, that firms will often not go to the trouble of requesting a preferential tariff treatment for their products because they are time sensitive. Due to the rules of origins and incidentally the certificates of origins, a certain amount of time is needed to process the request, which can be too long for certain firms and certain prod-ucts. Delays from the administration are therefore also given as a main reason for low utilization rates of TA in Wignaraja (2014). To measure the effect of compliance costs for firms, Carrère & De Melo (2004) measured the administrative costs of trading within NAFTA between the U.S. and Mexico. They found that such administrative costs account for 45% and 42% of total compliance costs in 2000 and 2001 respectively. From this study the authors conclude that a significant learning effect could be the basis for low utilization rates in the first years of the preferential access.

Utilization rates of TAs are expected to increase more significantly during the five first years after their implementation as firms become increasingly aware of the scheme. The coeffi-cients for the impact of awareness of the TA in the first five years are therefore expected to be more important than in the following years. The next hypothesis follows as:

H7: Utilization rates of TA are expected to rise more significantly during the five years after its implementation than in their later years.

For all these reasons explained in the literature but not only, Asian economies and espe-cially countries in ASEAN are found to have very low TA utilization rates. Due to multiple TAs and good data, this region is studied in particular in the literature of this topic.

9. Presence of low or zero MFN tariff and active investment schemes

Hamanaka (2013) shows that TA utilization rates in Asia are mostly very low because a large part of the trade within the region is covered by zero MFN tariffs making it pointless to incur the compliance costs linked with the TA application to benefit from the same zero tariff. Hiratsuka et al (2009) give the example of the Information Technology Agreement (ITA)4 under which duties have been completely eliminated for information technologies. Nilsson (2010) uses the example of the EU GSP program for which only 25% of EU imports from GSP countries are subject to positive MFN tariffs because of generalized low MFN tariffs. Hayakawa et al (2009) show that these low utilization rates are also due to the active investment schemes in the Asian region. Because of these schemes, firms do not need TAs to benefit from reduced or eliminated tariffs, which have a clear

4 _{Although only 82 countries are members of the ITA, it covers nearly 97% of world trade in IT products.}

Page ǀ 15 negative impact on TA utilization rates in the region. More specifically, the investment schemes provide tariff reducing opportunities to firms investing in a specific country, which incidentally replaces trade. Firms making Foreign Direct Investments (FDI) can thus benefit from advantageous opportunities making, in some situations, TAs as a tariff saving scheme obsolete. Promoting in-vestment consequently comes at the expense of trade which is a negative influence on TAs utiliza-tion rates.

Furthermore, they show that tariffs are generally already very low in Asia, making the ben-efits very low for firms when there are fixed costs involved with TAs and certificates of origin. Similar results are also showed in Hiratsuka et al (2009) that proves Japanese firms have low TA utilization rates because there are no additional benefit to be incurred by using a TA or because there is no TA with the country in question.

Such an extensive analysis for the Asian economic region calls for a study of the U.S. trade agreements not only to show whether U.S. TAs utilization rates are low as well, as explained in the literature, but also to provide insights on whether the low TA utilization rates in Asia are mostly due to regional conditions or factors applicable elsewhere. Studying utilization rates of trade agree-ments can be considered a good tool to evaluate the efficiency of said trade agreeagree-ments. As a con-sequence, studying the different TA signed by the U.S. and their utilization rates can help explain whether they are successful and why.

As have been showed, TA utilization rates are subject to many factors that influence whether firms use such agreements or not but also influence the data itself. From the literature, factors acting on firm decision are: rules of origin and incidentally certificates of origin, firm size, trade volume, preferential margin, level of information, eligibility to other preference schemes, level of MFN tariff rate, possibility for TA cumulation rules of origin and exchange rates. Factors influencing the data used in the studies are: the definition of the concept of utilization rate, time lag, and the way data is collected at the firm level. Appendix 1 provides a concise summary of the papers used in this literature review and the factors that will be studied in the subsequent analysis of TAs utilization rates.

S

3:

M

Page ǀ 16 This research will uncover whether factors studied in previous papers are relevant in the case of U.S. TAs.

This thesis will investigate factors influencing utilization of the trade agreements signed by the U.S. with 20 partner countries. Although this analysis is as thorough as possible, some factors and therefore some potential parts of this analysis will be left out because of the lack of data. Any kind of firm level analysis will not be conducted in this research as the quality and consistency of potential databases cannot be trusted to bring reliable results. A considerable part of the data on TA firm utilization is collected at the firm level through different types of questionnaires or inter-views. Hamanaka (2013) shows that such data should include how much of the firm’s trade value is concerned by the utilization of TAs instead of only counting firms using TAs or not. In such a situation, no matter whether the firm is large or small, firms are treated equally as a ‘0’ or a ‘1’ depending on whether their use the TA or not. Consequently, TA utilization rates could be overes-timated or underesoveres-timated at the firm level, bringing a severe bias in the results of such studies. Using other types of firm level data such as certificates of origin needed for rules of origin also brings an overestimation of utilization rates. This overestimation is expressed in Hayakawa et al (2013) and Hamanaka (2013) where exporters are not necessarily exporting their products under the program they requested, and trade that is not conducted even though the firm obtained the certificate of origin to benefit from the TA. These possibilities both lead to an overestimation of the utilization rate. For these reasons, any kind of firm level data will not be used in this research and factors such as firm size will not be investigated.

Furthermore, in their paper, Blind & Ziltener (2014) demonstrate that there could be sea-sonal variations in trade data stemming from the fact that imports and exports can vary depending on the time of the year. As a consequence, data involved in utilization rates computations will only be dealt with as yearly data to solve this issue. For the matter of data comparability, Hamanaka (2013) shows that TA utilization rates should be compared across various trade agreements and not across countries. A comparison made across countries could bias the results and therefore the con-clusions of the analysis. As all the agreements that will be discussed in this research are partnered with the U.S., comparisons between agreements and countries are equivalent in the case of bilateral TAs.

Page ǀ 17 these datasets were assembled in one to join all U.S. imports for consumption of the period 1989-20165.

As defined extensively by the U.S. International Trade Administration, imports for con-sumption measure the total value of merchandise that physically clears customs, or goods with-drawn from customs bonded warehouses or U.S. Foreign Trade Zones which immediately enter consumption channels. This measure does not include merchandise being held in bonded ware-houses of U.S. Foreign Trade Zones (USFTZ) as it is specifically withdrawn from consumption. These bonded warehouses and USFTZs are intermediary places used for storage and handling goods for which payment of duties is delayed until they reach customs territory. This import for consumption data was extracted at the HTS-10 digit level to be as precise as possible. The HTS digit codes were developed by the World Customs Organization (WCO) to make a clear classifi-cation of internationally traded products. Different codes correspond to different levels of preci-sion, six digits stand for universal standards while the seven to ten digits are for individual country of import. Choosing HTS-10 digits codes allows the data to contain the HS code but also the duty as well as further definition and makeup as higher numbers of digits mean a higher level of preci-sion. Each product code contained in the USITC database gives the customs value of import per programme, allowing the computation of utilization rates and the investigation of the influence of several factors on them. This research will therefore study U.S. imports at the country-product-year level for each of the 14 trade agreements.

Data extracted from the USITC historical and current database gives the necessary obser-vations to compute utilization rates and preferential margins. Supplemental data is required to as-sess the impact of other factors, as a consequence, data on exchange rate variations, GDP levels and GDP growth data was extracted from the World Bank Databank. Datasets were transformed to match the specifications of the trade data and allow the econometric analysis within one single dataset. GDP level and growth are used as control variables in the analysis to solve for the potential problem of correlation between variables.

The analysis will be performed using a regression technique developed by Sergio Correia (2017) as a Stata extension named ‘REGHDFE’. This alternative technique, unlike the Ordinary Least Squares (OLS) method, accounts for high-dimensional fixed effects. Hill, Griffiths and Lim (2012) showed that fixed effect models were preferred to OLS as they recognise variations in the data as non-random. The data is therefore treated as containing points correlated to one another within a certain group. This argument is of the utmost relevance in the case of TAs utilization rates as there can be co-dependence between variations within countries, years, or products. Conse-quently, a bias might be created when not accounting for these fixed effects.

Using panel data, the regressions using this technique can absorb several fixed effects at once. Considering the research and dataset at hand, fixed effects will be used to control for time and country-product. These fixed effects are meant to absorb variations in the variables that might

5 _{The raw trade data downloaded from the USITC database was handled by dr. Tristan Kohl with the coding}

Page ǀ 18 bias the results. For example, an economic crisis happening during a specific period would be controlled for by the time fixed effect as the variations are specific to this period. Moreover, vari-ations at the country-product level happening because of a sanction or a restriction will be absorbed by the related fixed effect as well. Consequently, regressions will absorb either for year and coun-try-product fixed effects or councoun-try-product fixed effect, making some of the potential control var-iables obsolete.

As the literature showed, the trade agreements signed by the U.S. are generally thought to have small utilization rates for a number of reasons. Globally, trade has been growing very fast during the past three decades, however since the implementation of these trade agreements, trade between the U.S. and these partners has grown particularly quickly. Appendix 3 provides a concise summary of each of the trade agreements with their name, type and implementation date. These 20 countries have seen their trade levels with the U.S. increase significantly considering imports of goods and services data values in constant 2010 U.S. dollars from the World Bank. For example, for the Australian-U.S. TA, U.S. imports from Australia are comparatively volatile with vast changes between years ranging between -5.70% (1991) and 24.97% (1989). After the implemen-tation of the TA in 2005, U.S. imports became more stabilized, around 10% increase per year ex-cept for 2009 which is likely to be a direct consequence of the 2007 economic crisis. The TA thus allowed a less volatile evolution in U.S. imports from Australia linked with an overall higher growth in imports levels since its implementation. Moreover, in the case of the U.S.-Oman TA implemented in 2009, U.S. imports from Oman were already increasing fast before the TA (20% increase in 2008) but started increasing even faster after its implementation (22.51% in 2010 and 48.18% in 2011). Furthermore, when taking into account the dataset of U.S. imports in goods and services between 2007 and 2015 (in constant 2010 U.S. dollars) for all countries, the growth rate is 20.25% between the years 2007 and 2015. However, when only the TA partner countries are taken into account, the growth rate of U.S. imports is 28.43%. This period was chosen to avoid the potential consequences of the economic crisis while having as many TAs implemented as possible. Consequently, trade between the U.S. and its TA partners tend to grow faster than its total trade even though TAs with the U.S. may not be said to be highly utilized in the literature.

Page ǀ 19 2. Utilization rates of U.S. trade agreements

As explained in the literature review, utilization rates are defined as the share of custom values imported under FTA schemes in the total customs value for FTA eligible products. The data extracted from the USITC database gives all U.S. imports for consumption from the 20 TA partner countries during the period 1989-2016 in terms of customs value. These values at the country-product-level were computed to give utilization rates at this level of detail and then averaged at the country-year level to give comparable TA utilization rates as shown in Appendix 3. The utilization rates at the country-year level were not computed directly as the highest level of detail will be used in the analysis of the influence of factors.

In the literature, utilization rates of the U.S. trade agreements were considered to be low as most papers show results around 50-60% for NAFTA and much lower for TAs in Asia. This trend is confirmed by the data in Appendix 3 and similar rates are found for other agreements than NAFTA although there are exceptions on the high and low ends of the spectrum. Only in a few instances such as the US-Jordan or the US-Oman agreements do utilization rates go above 55-60%. In most cases, only half or less than half of the total customs value eligible for preferential tariff treatment actually claimed the TA tariff rate. Such utilization rates justify the need for research on factors influencing utilization of TAs as there are reasons behind what are commonly considered low rates. The literature on the topic usually place a threshold around 70%, above which TAs are considered to be highly utilized. Following this threshold, several U.S. TAs come fairly close but only the U.S.-Jordan and U.S.-Oman agreements can be considered highly utilized as shown in the Appendix.

Some exceptionally low numbers can be explained by specific conditions. For example, the TA between the U.S. and Singapore is very little utilized as any of its utilization rates are below 10%. These rates mean that since the implementation of the agreement, every year, less than 10% of U.S. imports from Singapore in terms of total customs value, claimed the US-Singapore TA to benefit from the preferential tariff rate. These low rates can most likely be understood as a conse-quence of strict rules of origin. In order to benefit from a TA and therefore obtain the reduction or elimination of tariffs, firms need to comply with certain requirements of origin. Considering Sin-gapore is a relatively small country that is mainly used an entry hub for other Asian economies, fulfilling requirements of products origin is made much harder than in larger economies such as Australia or Canada. It is however, impossible to test for the impact of country size as it would require firm level data of extreme precision for the reasons behind the use or non-use of the TA. Although this explanation is economically sound, it is to be taken carefully as the data shows that most MFN tariff rates for Singapore are already zero or close to zero. As a consequence, there is little incentive for firms to request the preferential tariff treatment as the tariff saving opportunity is limited. Both these factors are certainly contributing to overall low utilization rates for the U.S.-Singapore TA.

Page ǀ 20 Oman but its utilization rates are much smaller (and decreasing) compared to the two other coun-tries. However, Israel is much more developed than the two other countries, with a GDP of nearly 300 billion dollars in 2015 (World Bank Data) while Jordan and Oman have GDP levels of 38 and 70 billion dollars respectively for 2015. With a much more developed economy, it could be ex-pected that rules of origins are easier to fulfill for Israel than Jordan or Oman and therefore, that its utilization rate be higher. A larger economy could be expected to be positively related to production capacity and thus higher probability of fulfilling the origin requirement of TAs. As this is not the case, other factors are clearly influencing this outcome. In such a situation, it is likely that a com-paratively larger difference between MFN and preferential tariff rates play a particularly important role in these high utilization rates. This higher difference would mean a larger tariff saving oppor-tunity for firms importing from Jordan or Oman compared to other countries. As was the case for the low utilization rates as well, the importance of this factor cannot be proven unless further econ-ometric analysis is conducted. Such analysis might prove the opposite even though this specific type of influence is economically intuitive.

Utilization rates for each of these 14 agreements give heterogeneous results even if they belong to a generalized range of utilization rates as shown in the literature. While controlling for different environments and economic conditions, this research will try to provide clear results as to how are influenced these utilization rates.

3. Information and learning effect

The learning effect influencing the utilization rate of TAs is expressed by a more or less steep increase in TAs claims during the first years after its implementation. Such an increase can be explained by firms becoming progressively more aware of the existence of a new preferential tariff scheme. The graphs in Appendix 4 showing the evolution of preferential programs claims between 1989 and 2016 help studying this effect when considering whether all or no programs are claimed. Appendix 3 shows the evolution of TAs utilization rates between 1989 and 2016 but the case of Israel cannot be studied for this factor as the data available only starts in 1989 and this agreement was implemented in 1985 therefore leaving the crucial first years of the TAs action missing.

Page ǀ 21 These results are confirmed by the utilization rates in Appendix 3, although there seem to be an increase in the first years after a TA implementation, no general conclusion can be drawn as some agreements do not observe this trend. For example, Australia and Singapore show more or less constant utilization rates over the years, even in their first ones. The remaining countries, namely Chile, Costa Rica, Morocco, Canada and Mexico also incur a rather steady trend in their numbers of TAs claims. The lack of clear rise in TAs claims would suggest an absence of learning effect or importers stopping their use of the TA while others start using it following the increasing awareness of the agreement.

This learning effect can also be ambiguous as TAs require several years of negotiations and then a more or less publicized implementation. Firms that are going to benefit the most from this type of agreements are likely to be concerned about TAs with specific countries and therefore would know about these negotiations very soon. However, as shown in the literature, many firms lack this type of information and thus can be expected to learn about a TA only later. These firms and especially SMEs are making the bulk of the potential learning effect as company and shipment sizes make the duty savings small in comparison to multinationals.

Such a learning effect could be seen intuitively on the progression of TAs claims but it is impossible to disentangle this influence from other possible causes. What is understood as a time lag for information to reach more firms that would utilize the preferential schemes if they knew about it could be encompassed in a larger cause. In order to test for this learning effect, we would need firm level data of great precision as the specific reasons for using a TA or not are essential. As a consequence, we cannot test for this factor and cannot make any clear conclusion on it. Ap-pendix 4 also shows that the utilization rates of the US-Israel TA become smaller over time. When considering a lag in awareness effect, as firms learn about potential benefits from using this agree-ment there should be an increasing trend. As a consequence, there are other factors influencing utilization rates of trade agreements. Such factors could include bilateral or diagonal cumulation.

4. Bilateral and diagonal cumulation

As defined earlier, bilateral cumulation in a TA means that materials originating from one country can be considered as originating from the partner country and vice versa. Considerations of product origins are crucial in rules or origin because there is a value added ratio requirement to be able to benefit from the preferential tariff rate when the product is not wholly obtained from the partner country. A bilateral cumulation rule means that in the case of the US-Bahrain TA for ex-ample, products from the U.S. can be considered from Bahrain and vice versa. It is therefore easier to reach the value added ratio of origin for the partner country and benefit from the tariff saving opportunity.

Page ǀ 22 The table shows that nearly all the TAs have the cumulation provision embedded in their texts. The presence or not of a cumulation rule is not stated in the Harmonized Tariff Schedule of the United States (2017) for the TAs with Jordan and Israel. As all other U.S. TAs have this provi-sion and similar value added requirement ratios, it is assumed that this proviprovi-sion is also valid in the case of Jordan and Israel. As a consequence, it is impossible to test for the impact of the cumulation rule on TA utilization rate compared to its absence as the presence of this rule in all agreements is in direct conflict with the scientific method. The method requires a group of subjects/observations under the influence of the factor to be tested and a control group that is not influenced by this factor at all. Within the limitations of the sample, there is no possibility to have a control group for this factor.

However, this factor analysis could be performed with the comparison of countries having the bilateral cumulation rule against the countries having the diagonal cumulation. Unfortunately, even if this analysis showed results of higher utilization rates with a diagonal cumulation compared to a bilateral cumulation rule, no conclusion on the effect of cumulation itself could be reached. The cumulation effect would not be disentangled from the exterior economic situation as the anal-ysis would only compare between agreements with the rule. Consequently, concluding that diago-nal cumulation has a more positive or negative effect on utilization rates compared to the bilateral cumulation gives no indication on the overall impact of cumulation (no matter its type) on utiliza-tion rates.

For these reasons, there will be no further econometric analysis for this factor considering our sample. The impact of this learning effect is highly suspected to be a positive influence on TA utilization rates but as all TAs in the sample benefit from this rule, no econometric analysis will be performed.

5. Strict rules of origin

As defined in the literature and given by the United States Harmonized Tariff Schedule (2017 Basic Edition), rules of origin correspond to products that must be wholly obtained from the TA partner country or have a minimum value-added ratio of origin from this country to be eligible to the preferential tariff rate. The rules of origin are detailed for each U.S. trade agreement in Ap-pendix 11. Although the percentage required for the value-added ratio is missing for most TA in the HTSUS, the ones detailed are very similar. To be eligible to the preferential tariff rate, products must be wholly obtained or have at least 35% of value-added originating from the TA partner country. This percentage includes value added originating from the U.S. or other TA partners as well following the bilateral or diagonal cumulation as explained earlier. Furthermore, the US-Israel agreement for example, puts a 15% cap for the value-added originating in the U.S. within the global ratio. This point further confirms the assumption of bilateral cumulation in the case of the US-Israel TA.

Page ǀ 23 very precise data on rules of origin is needed for all the TAs as well as very precise firm level data. Knowing whether a firm is using a TA or not because of the rules of origin is essential in this analysis but this level of detail cannot be reached with the available data.

The percentages given above refer to the agreement-specific part of the Harmonized Tariff Schedule, with a certain number of missing values. The document then describes the customs treat-ment per product at the HTS 8-digit level and gives substantial details on tariffs rates. Although value-added ratio requirement is not detailed at this level of precision and therefore cannot be stud-ied as utilization rates, studying rules of origin is of extreme complexity. Even when assuming a singular value-added ratio for all agreements, special treatments are granted for specific products within most trade agreements. Consequently, if we assume different value-added ratios and special treatments, rules of origin become an intrinsically complex issue that require extensive data. The influence of rules of origin are thus economically very probable but will be neither proven nor refuted in this research considering the data available and the sample limitations.

6. Compliance costs

Compliance costs refer to the sum of different costs a firm incurs when requesting the pref-erential tariff treatment for import. This total encompasses the fixed administrative cost of applying for the TA tariff rate, the variable costs of verifying the claims but also any cost stemming from the delays in carrying the transaction such as warehousing. Because of the complex nature of the compliance costs, it is especially hard to test for this factor. As was the case for the learning effect and rules of origin, very detailed data is required to pursue an econometric analysis of the impact of compliance costs on utilization rates of trade agreements. Given that such firm level data is either unavailable or unreliable, it is impossible to test this factor with certainty within the bound-aries of this research.

As a consequence, we expect that utilization rates of trade agreements are decreasing when compliance costs are increasing but will remain unconfirmed as we cannot verify this claim. This hypothesis is economically intuitive as a firm is expected to decrease its utilization of a TA when the overall costs linked with this utilization are increasing. The rise of compliance costs could not only be expressed by an increased monetary cost but also through longer administrative procedures linked with longer delays, a complicated verification procedure or heavier controls after trade. The cost in itself is not necessarily negatively impacting the utilization rates, it is rather the decrease in the tariff saving benefit compared to other potential claiming schemes. Utilization rates are thus expected to decrease because the rise in compliance costs makes other options such as not claiming a TA at all, less costly.

Page ǀ 24 by using a TA as compared to when not using it should uncover a factor of large influence on such utilization rates. Trade agreements and their preferential margins are therefore studied next.

7. Level of preferential margin

As detailed in the literature review, the preferential margin refers to the difference between the MFN tariff rate and the preferential tariff rate of the TA. This margin is expressed as an absolute gap between the two rates, the larger the gap, the larger the benefit for firms. For each country-product-year trio, the MFN and preferential tariff rates were downloaded from the USITC Interac-tive Tariff and Trade database version 3.1.0. The margin gap was computed as the absolute differ-ence between the MFN and the preferential tariff rates for the three components of the tariff rates: ad valorem, specific and additional rates. Ad valorem rates are the most common and correspond to a customs duty of a specified percentage of the products value as defined by the World Bank. For non-ad valorem tariffs, as defined by the World Bank as well, specific rates are a fixed customs charge per unit of quantity. Thus, the ad valorem is expressed as a percentage of the value for the duty while the specific rate is directly a monetary duty value per quantity. The additional part of the tariff, on the other hand, corresponds to the part of the tariff that is neither ad valorem nor specific. In other words, the additional part of the tariff assembles all other measures expressed in tariffs and can be measured in different ways.

These margins are given by the variables ‘pref_ad_val_gap’, ‘pref_specific_gap’ and ‘pref_other_gap’ as shown in Appendix 6. Then these gaps were computed as a percentage share of the MFN tariff rate given by the variables ‘pref_ad_val_discount’, ‘pref_specific_discount’ and ‘pref_other_discount’. Compared to the gap variables, the discount variables give the preferential margin as a percentage of MFN tariff rate instead of an absolute difference. In creating these vari-ables it is assumed that the preferential tariff rate is lower than the MFN rate. Preferential tariff rates are supposed to be more advantageous than MFN rates in terms of duty, lower MFN rates than preferential rates would lead to an underestimation of utilization rates as these products rep-resent special cases. However in the case of 20,861 out of 8,454,320 observations the MFN tariff rate is lower than the preferential rate. As these specific occurrences only amount to 0.25% of the total observations, they will be omitted of all econometric analysis as they represent a potential bias to the results.

Page ǀ 25 This preferential margin is expected to have a particularly strong influence on TAs utiliza-tion rate as it directly defines the tariff saving potential of the TA. Both these measures are expected to be of major influence of utilization rates of TAs, as stated in Hypothesis 1. The preferential margin factor will be investigated at the country-year-product level of U.S. imports for consump-tion from each of the 20 countries of the sample, between 1989 until 2016. The specificaconsump-tion of the regression for the preferential margin measured as the absolute gap between MFN and preferential tariff rate is as follows:

(1) 𝑈𝑅𝑖𝑗𝑡 = 𝛽0+ 𝛽1𝐺𝐴𝑃1𝑖𝑗𝑡+ 𝛽2𝐺𝐴𝑃2𝑖𝑗𝑡 + 𝛽3𝐺𝐴𝑃3𝑖𝑗𝑡+ 𝛽4𝐺𝐷𝑃_𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝑡 + 𝑒𝑖𝑗𝑡

Where URijt gives the TA utilization for country i and product j at year t and is regressed on the different part of the preferential margin and the GDP growth. GAP1ijt, GAP2ijt and GAP3ijt respec-tively correspond to the absolute difference between the MFN tariff rate and the preferential rate for the ad valorem, specific and additional parts of the total duty for country i and product j at year t. GDP_growthit is the GDP growth of country i at year t to control for potential correlation between dependent and independent variables. Finally, eijt is the error term of this regression.

Furthermore, the specification of the regression measuring the impact of the preferential margin measured as the gap share of the MFN tariff rate (discounts) is as follows:

(2) 𝑈𝑅_𝑖𝑗𝑡 = 𝛽₀+ 𝛽₁𝐷𝐼𝑆𝐶1_𝑖𝑗𝑡+ 𝛽₂𝐷𝐼𝑆𝐶2_𝑖𝑗𝑡+ 𝛽₃𝐷𝐼𝑆𝐶3_𝑖𝑗𝑡+ 𝛽₄𝐺𝐷𝑃_𝑔𝑟𝑜𝑤𝑡ℎ_𝑖𝑡 + 𝑒_𝑖𝑗𝑡 Where, as above, URijt stands for the TA utilization rate for each country-product-year trio but is regressed on the preferential margin measured as a gap share. Thus, DISC1ijt, DISC2ijt and DISC3ijt respectively correspond to the preferential margin of the ad valorem, specific and additional parts of the tariff rates computed as a discount. The discount itself corresponds to the absolute gap (MFN minus preferential tariff rate) share in the MFN ad valorem, specific and additional parts of the total duty. As above again, GDP_growthit is the GDP growth for each country-year pair and eijt is the error term of this regression.

8. Exchange rates variations

Page ǀ 26 2010 USD to correct for inflation was extracted from the World Bank Databank as well as GDP growth as a yearly percentage change for each of the 20 partner countries. GDP level is further computed to be used as a percentage share of the U.S. GDP level in order to avoid scale differences between variables in the analysis. Data extracted from the World Bank Databank was transformed to match the specifications of the trade dataset previously constructed to allow the analysis within a single dataset.

To this end, the impact of exchange rate variations will be tested for each of the local cur-rencies on their respective utilization rates. All the local curcur-rencies and their country are shown in Appendix 5, Table 4. El Salvador is a unique case in this instance as the legal tender was the Sal-vadoran colón until December 31, 2000 but the economy was then dollarized and the U.S. dollar became the local currency in January 1, 2001. All subsequent analyses for exchange rate variations will be made at the country level to account for this fact.

As exchange rates are highly dependent on the local economic situation, the regression will be controlled for GDP level and GDP growth in order to eliminate the effect of differences in economic conditions. As a consequence, the specification of the impact of exchange rate variations on TA utilization rates is as follows:

(3) 𝑈𝑅_𝑖𝑗𝑡 = 𝛽₀+ 𝛽₁𝐸𝑅𝑉_𝑖𝑡+ 𝛽₂𝐺𝐷𝑃_𝑔𝑟𝑜𝑤𝑡ℎ_𝑖𝑡+ 𝛽₃𝐺𝐷𝑃_𝑙𝑒𝑣𝑒𝑙_𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒_𝑖𝑡+ 𝑒_𝑖𝑗𝑡 Where URijt gives the TA utilization for country i and product j at year t and is regressed on the exchange rate variations and controlled for GDP growth and level. ERVit stands for the annual variation in exchange rate between the U.S. dollar and the local currency for each country i year t pair. GDP_growthit is the GDP growth of country i at year t and GDP_level_relativeit corresponds to the GDP level as a share of the U.S. GDP for each country i year t pair. Lastly, eijt is the error term of this regression

9. Eligibility to other tariff saving schemes

Some of the 20 countries that have a TA with the U.S. are eligible to other tariff saving schemes besides their bilateral or multilateral agreements with the US as shown in Appendix 5, Table 5. Under these different schemes, trade can be conducted with reduced or eliminated tariffs and therefore these agreements are directly competing with the TAs signed with the US. The avail-ability of different schemes can thus be expected to reduce the utilization rates of such TAs as trade can only be conducted under one program or none at all, if no import program is claimed.

As presented in Appendix 8, there are 11 other schemes allowing a country to benefit from these reduced tariffs. In terms of claims number, the ATPA, GSP, CBERA, CBTPA, APTA and Civil Aircraft concentrate most of the non-FTA claims.