Using Utilization Rates to Identify Rules of Origin Reforms: The Case of EU Free Trade Area Agreements

Governments are increasingly entering FTAs and mega-regionals to secure market access for their firms. Utilization rates are used to monitor whether firms are using these FTAs. This paper is part of a recent stream of studies to dash out enduring myths that preferences are not used when preferential MFN rates are low or for unknown or vague reasons. Contrary to this sort of conventional wisdom this study advocates that low utilization rate is a valuable and unequivocal sign that reform of rules of origin and related administrative procedures is needed to make the FTA attractive and meaningful to the private sector. By using a “repeated offender” methodology this paper identifies a series of product specific rules of origin (PSROs) causing low utilization rates. Such PSROs are the candidate for reforms towards more lenient requirements that are commercially viable for firms.

1 Introduction * Utilization rates are increasingly used to monitor the effective use of preferential trading arrangements of reciprocal or autonomous nature. The methodology of calculating utilization rates has been first conceived in UNCTAD back to 1975 and most recently reaffirmed in WTO in the context of the debate over preferential rules of origin for LDCs.
The National Board of Trade Sweden/UNCTAD report found that utilization rates are 90 percent for partner country exporters and 67 percent for EU exporters.
This means that the EU partner country exporters use the EU's free trade agreements to a higher degree-by 23 percentage points with respect to EU exporters. Why is that? What is the reason for such asymmetrical utilization rates? In a FTA, rules of origin and related administrative procedures are the same for both partners.
Moreover, while average figures are relatively high, there are pocket of underutilization linked to specific sectors and product specific rules of origin.
The present study deepens the analysis of the previous report to identify the reasons for 1) asymmetry of utilization rates and 2) the critical products responsible for such pockets of underutilization linking it to product specific rules of origin (PSRO) and related administrative procedures The study develops a new methodology to identify PSRO defined as repeated offenders at the 4-digit level of the harmonized system (HS) as a reason for systematic pockets of underutilization in EU FTAs.
This study brings new evidence that the reason for such low utilization of trade preferences is mainly due to stringent rules of origin and their related administrative procedures. This paper is part of a recent stream of studies to dash out enduring myths that preferences are not used when preferential MFN rates are low or for unknown or vague reasons. Contrary to this sort of conventional wisdom this study advocates that a low utilization rate is a valuable and unequivocal sign that reform of rules of origin and related administrative procedures is needed to make the preferential trading arrangement attractive and meaningful to the private sector.

Linking utilization rates to rules of origin
Governments are increasingly negotiating FTAs and mega regionals to create market access for their companies. Only recently, Governments are realizing the importance of the implementation and effective use of FTAs by their companies. Hence, it is not surprising that the utilization of trade preferences has been object of a revival also thanks to the work done by LDCs in getting to the WTO Nairobi Decision on preferential rules of origin. Such Decision contains an obligation of preference granting countries to LDCs to notify to the WTO secretariat on the utilization rates of such trade preferences. Robert Schuman Centre for Advanced Studies Working Papers This in turn meant that the WTO had to agree on a definition of utilization rate 1 and that the performance of the various trade preferences schemes granted to LDC is object of analysis by the WTO secretariat 2 and members, especially LDCs 3 .
There have been a series of interesting findings based on the first part of this study 4 where asymmetrical utilization rates were identified in the case of the EU-Switzerland FTA and the EU-South Korea FTA. Such findings were brought to the attention of WTO members and have been further developed in the Committee on Rules of Origin 5 Utilization rates may reveal a merciless picture of a trade preference either of unilateral or reciprocal nature. This is the main reason why Governments gradually became wary or made a variety of objections to the use of utilization rates and began to question: a) The use of utilization rates to measure the effectiveness of an FTA or a given scheme of trade preferences, and; b) The link among low utilization and rules of origin.
The objections raised under (a) are that some trade preferences may not be used simply because there is an overlapping of trade preferences or other competing duty rebates schemes, such as the possibility of duty drawback 6 . However, such argument in reality reinforces the usefulness of utilization rates to measure the effectiveness of various competing and overlapping trade preferences as utilization rates clearly show that one preferential scheme may be better than the other one allowing trade policy choices and an improved exploitation of the more advantageous trade preferences.
The link of utilization rates with rules of origin under (b) has been demonstrated in variety of studies and corroborated by factual evidence and testimonials from companies. Yet it stirs debates in negotiating circles and in the literature.
The utilization rates provide figures for the value of trade that has not received preferences at the time of customs clearance while it was eligible for preferential rates of duty. However, the utilization rates does not provide a clear and definitive reason on why the preferential rate has not been requested or granted at the time of customs clearance. From a point of view of customs law, the denial of the preferential duty rate means that at the time of importation the preferential rate of duty has not been granted and MFN rate of duty has been instead being imposed.
The reason for such denial of preferential treatment from the point of view of customs law may be due to one of the following factors: See for instance the submission of the LDC "Further evidence from utilization rates" about the utilization of the preferences granted by Switzerland to LDC contained in WTO document G/RO/W/186 of 6 May 2019. This document is of particular importance since it marked the start of a dialogue between the WTO LDC, assisted by UNCTAD group and Switzerland that permitted to identify the reasons for low utilization of trade preferences linked to transhipment requirements. 4 See The Use of the EU's Free Trade Agreements Exporter and Importer Utilization of Preferential Tariffs available at https://unctad.org/webflyer/use-eus-free-trade-agreements-exporter-and-importer-utilization-preferentialtariffs 5 See the Direct consignment rules and low utilization of trade preferences submission by the LDC WTO group WTO document G/RO/W/191 of 9 October 2019 and the Note from WTO secretariat »Impact of direct consignment requirements on preference utilization by Least Developed countries, WTO document G/RO/W/187/Rev.1 of 24 February 2020 6 For instance the GSP/AGOA in the case of unilateral trade preferences might be overlapping. In the case of FTAs, a valuable example is the Vietnam -Japan preferential trade relations where there is 1) a Japan-ASEAN FTA ,2) a Vietnam-Japan bilateral FTA, 3) The Comprehensive and Progressing Trans Pacific Partnership CP-TPP and when implemented the Regional Comprehensive Partnership Agreement (RCEP). On top Vietnam still beneficiary of the Japan GSP

Using Utilization Rates to Identify Rules of Origin Reforms: The Case of EU Free Trade Area Agreements
European University Institute 3 a) Lack of presentation of documentary evidence related to the originating status of a product presented to customs such as a certificate of origin (CO) or an exporter or importer declaration (ED/ID) due to failure to comply with product specific rules of origin requirements 7 ; b) Lack of presentation of documentary evidence related to ancillary administrative requirements related to origin, such as direct consignment or related procedures (i.e. lack of bill of lading, lack of certificate of non manipulation, mismatch of the certificate of origin/exporter declaration with bill of lading, third country invoicing 8 and other related documentary evidence); c) Low preferential margin meaning that there has been a deliberate choice by a firm not to use of the trade preference because of the low preferential margin that made compliance with rules of origin not profitable, d) Any other reason that are usually attributed in literature to different factors such as existences of competitive duty rebate schemes, drawback, etc.
There is hardly any evidence tracking or identifying what is the most common reason for the failure to provide a certificate of origin or an exporter or importer declaration (ED/ID). There are a number of empirical studies based on econometrics 9 . Evidence from the business sector in beneficiary countries and from business surveys, literature and field activities indicate that in the majority of cases, failure to provide a CO/ED/ID is mainly due to excessive stringency of product-specific rules of origin (PSROs) or related administrative documentary evidence 10 .
The most recent findings and studies are converging that non-utilization of trade preferences is due to rules of origin requirements under (a) and (b) that are mainly related to non compliance with the rules of origin, such as an ad valorem percentage requirement, a change of tariff classification or a specific working of processing requirement or a failure to comply with ancillary administrative requirements related to direct shipment, third country invoicing or related administrative requirements. The inability of a firm to comply with (a) and (b) may be closely related. The lack of documentary evidence related to origin under (b) has been one of most underestimated reason of low utilization of rules of origin. Recent research 11 has shed some lights on this area that has not yet attracted the attention that deserves especially in terms of potential for trade facilitation. 7 The failure to comply with PSROs requirement is mainly due to lack excessive stringency of a given PSRO requiring according a) an ad valorem percentage that was not reachable given the sourcing policy of a firm b) a Change of tariff classification requirement that impeded the use of a non originating input, (c) a working or processing requirement that was not possible to perform given the insufficient manufacturing capacity of a firm 8 For a definition of third country invoicing, one of the most common reason for non utilization of trade preferences not related to origin see Guidelines on certification of origin, 2014 and 2018, World Customs Organization.WCO 9 See Inama, Rules of origin in international trade, 2021, forthcoming Cambridge University Press. and Jiso Ji Jisoo Yi Rules of origin and the use of free trade agreements: a literature review World Customs Journal, 2016, 10 See see various presentation from private sectors representatives discussing the link between RoO and ability to use trade preferences presentation of Jon Edward at UNCTAD available at https://unctad.org/system/files/non-officialdocument/aldc2014_06_edwards_en.pdf and and related presentation of Bianchi as importer available at https://unctad.org/system/files/non-official-document/aldc2015_Florence_p11_Bianchi.pdf. See, Roberto Soprano, The Challenge of Designing 'New' Rules of Origin in International, Trade paper made for European University Institute (2016). See also UNCTAD studies, "Improving market access for least Developed countries "UNCTAD/DITC/TNCD/4, may 2001 available at https://unctad.org/system/files/officialdocument/poditctncd4.en.pdf; Trade preferences for LDCs: an early assessment and possible improvement, UNCTAD/ITCD/TSB/2003/8, available at https://unctad.org/system/files/official-document/itcdtsb20038_en.pdf.; See WTO document Modalities for the calculation of preferences utilization G/RO/W/161 25 August 2016 and subsequent notes prepared by the WTO secretariat on preference utilization; 11 See Stefano Inama and Pier Paolo Ghetti « The real cost of rules of origin: A business perspective to discipline rules of origin in a post COVID-19 scenario in Global Trade and Customs Journal, Global Trade and Customs Journal, Volume 15, Issue 10, 2020 and submission from the LDC WTO group and notes from the WTO secretariat listed in footnote 6 above Robert Schuman Centre for Advanced Studies Working Papers Thanks to recent debates in WTO Committee on Rules of Origin fuelled by the research conducted by the authors 12 , it has become evident that the administrative requirements to comply with rules of origin under (b) are a decisive factor for the utilization or non-utilization of trade preferences. The requirement of documentary evidence of direct consignment is the most glaring example of administrative requirements that have a bearing on utilization rates. 13 Low preferential margins under (c) and other reasons under (d) are most frequent anecdotal arguments to dismiss the existing link between utilization rates and rules of origin. A recent report from the National Board of Trade Sweden (Hereinafter the National Board of Trade Sweden report) 14 based on transaction and firm level data analysis contains valid findings against such general assumptions:  It is generally assumed that companies in general fail to use free trade agreements due to complicated rules and administrative burdens. This analysis, however, indicates that companies tend to utilize the tariff preferences when large import transaction values are involved and the potential duty savings are high.  It is generally assumed that free trade agreements are underutilized by small companies. This analysis, however, indicates that small companies might be even more efficient users of free trade agreements than large companies.  It is generally assumed that the level of tariff reductions is the main driver of companies' use of free trade agreements. This analysis, however, indicates that the size of the tariff reduction is less significant than the import transaction value.
In fact, the results of this report should set the end for the enduring myth that low preferential margins are a reason for low utilization. In spite of the numerous econometric studies to quantify the cost of rules of origin there is no evidence for such assumption.
The most intuitive argument against the notion of low preferential margin as justification for low utilization stems from the fact that the MFN rate for cars in the US is as low as 2,5%. Yet successive US presidents under NAFTA and most recently the Trump administration have spent their time to devise stringent rules of origin in NAFTA and USMCA to oblige more US content in cars using as leverage the preferential margin of 2.5%. In order to save the preferential margin of 2.5% car manufacturers have to comply with stringent rules of origin under USMCA going up to a 75% of regional value content with workers earning not less than 16 USD per hour. According to USTR office this provision "ensure that United States producers and workers are able to compete on an even playing field, and incentivize new vehicle and parts investments in the United States" If a low preferential margin, such as 2.5%, is not commercially meaningful, such restrictive rules of origin would be ineffective to incentivize cars manufacturers to comply with USMCA since they always have the option to pay the 2.5% and not comply with such stringent rules. However, the large majority of manufacturers are expected to comply with such new requirements as shown by the high utilization rates for cars under NAFTA.
It should be obvious that even a preferential margin as low as 2.5% could be commercially meaningful when related to a sector like automotive where the trade value is significant and competition is fierce.
Low utilization rates are also justified by the existence of other duty rebates and/or duty drawbacks. This argument is an anecdotal variant of (d) as it is rather obvious that duty rebates and drawbacks are different customs procedures that are not competing or should not be confused with preferential duty rates. Normally data entry by customs separates under different columns duty drawbacks, rebates and utilization rates of trade preferences.
This study conducted on the basis of the findings of the previous report on the use of EU FTAs 15 starts from the final consideration of the precedent report that even though the use of the possibilities of tariff reductions in the EU's free trade agreements is on average about 75 percent, it is important to identify possible pockets of low utilization at a more detailed level.
These pockets of low utilization have to be identified in the individual free trade agreements and at an industry and/or product-specific level. The present study aims at showing a methodology to use utilization rates to identify "repeated offenders", i.e. product specific rules of origin that are causing low utilization of trade preferences.
Utilization rates as argument for a reform of rules of origin has been raised in literature 16 and recently in WTO by the LDCs on the basis of a study prepared for the WTO LDC by UNCTAD. 17 In this study the spectacular raise of utilization rates and volume of exports following the reform of Canadian and EU rules of origin were contrasted with the stagnant performance of the US GSP/AGOA and Japan GSP rules of origin that have not undergone any significant reform since their inception in early 1970s 18 . Yet questions were raised about the casual link of such reforms of rules of origin with the corresponding rise in utilization rates and volume of exports. Testimonials made in the WTO Committee on Rules of Origin by the firms benefitting from such reforms were noted but did not totally remove perplexities.
It is obvious that utilization rates of a given preferential agreements represent a check-up of the status of well being of a preferential trade arrangement. Low utilization rates are an unequivocal sign that the preferential trade arrangement is not working properly. Conversely, high utilization rates should not be blindly considered as a positive sign since such high utilization rate may hide low trade volume and/or a low coverage of the preferential trade agreement. This is the case of the US GSP in the case of LDC countries for instance 19 . The reality is that like a medical check-up, the results of the analysis should be read by trade experts that are able to draw the necessary conclusions confronting them with other datasets and evidence from firms.
This study shows how the application of a methodology based on filtering criteria joined with statistical analysis permits to identify a "malaise" in FTAs, such as the overly restrictive productspecific rules of origin and/or related administrative provisions that have been found to cause low utilization rates in various FTA, across FTAs under different direction of trade.
In certain cases, the result of this analysis has later been confirmed by the findings of other surveys, like the report of the implementation of EU-South Korea FTA and the reforms that have been undertaken in recent FTAs negotiations, such as the case of the EU-Mexico FTA and the EU-Japan FTA. In other cases, further research and counterfactual with the private sector should be carried out using the results of this analysis. Robert Schuman Centre for Advanced Studies Working Papers This study, like the previous National Board of Trade Sweden/UNCTAD report, is based on data from 17 EU free trade agreements (including a customs union in the case of Turkey), i.e. the bilateral trade relations, between the years 2009 and 2013, where data on preference utilization are available on both parties, in order to make a comparative analysis (see Figure 1). Not all free trade agreements covered in this analysis have been in force for all these years and the data on the EU's preference utilization, i.e. EU exports and/or partner country exports, are complete for all years in the period. The development of the preference utilization before and after the years 2009 and 2013 is not considered in the analysis since the data from partner countries are not available.
The EU's total number of free trade agreements (and customs unions) within different continents as of 2013, i.e. 33 free trade agreements, or rather bilateral free trade relations, as well as the free trade agreements covered by this analysis (in cursive), were the following: The data on preference utilization in this report is based on import data that are collected and processed by the different parties concerned. The existence of different sources for the import data might limit the comparability between the partner countries and the EU due to possible differences in the quality of the data and the different methods possibly used. Furthermore, with the only exception of Switzerland, specific duties have not been converted to ad valorem equivalents, preferences within tariff rate quotas have not been considered in the calculations, and data on the EU's unilateral scheme of preference utilization, the 'generalised scheme of preferences', are not excluded in the calculations on preference utilization for the partner countries that might benefit from both preferential schedules albeit this occurrence has been limited by the entry into force of the GSP regulation of 2014 progressively excluding from GSP benefits those countries that have entered.

EU FTAs 20 -Identifying the reasons of asymmetrical utilization
As it has been pointed out in the National Board of Trade Sweden/UNCTAD report, the "average total preference utilization rate (2009)(2010)(2011)(2012)(2013) of the EU's free trade agreements is 90 percent for partner country exporters and 67 percent for EU exporters. The average 'preference utilization rate' for both parties is 75 percent. This implies that the EU's free trade agreements are used to a generally high degree." 20 The import data used in this analysis are originally collected by the partner countries and have thereafter been processed by the European Commission. In addition, data from United Nations Comtrade International Statistics Database and the UNCTAD TRAINS Database have been used. The existence of different data stemming from a series of partner countries might limit the comparability between the partner countries and the EU. In addition, the calculations of the EU's export and duty values with regard to the utilized tariff preferences in the EU's free trade agreements are approximations based on the real tariff preference utilization rates of the EU exports. The data presented in the report is, however, the closest approximation of the reality that is currently available. As shown in Figure 1 below, both the simple and weighted average utilization rates 21 of EU imports from the selected partners 22 are relatively high ranging from 87% to almost 91 % over the 2009 and 2013 period. 23 These high figures are matching a preliminary analysis made by Inama (2013) finding that the utilization rates of NAFTA and other US trade FTAs have been high, in some case above the 90% figure. 24

Figure 1 -Aggregated utilization Rates of EU Imports from selected FTA Partners Simple and Weighted Average
The overall high utilization rate of EU FTAs imports depicted in the above figure 1 above in reality captures only a fraction of the total picture. When the analysis is carried out at more disaggregated level, by direction of trade, partner and products, the utilization rates vary substantially indicating areas where PSROs and other administrative issues related to origin are problematic.
In fact, a closer analysis depicts a more fragmented scenario. Figure 2 below shows the utilization rates of EU exports to the same selected partners over the same period. The utilization rate is significantly lower, ranging around 70%, with lower utilization rates in selected chapters, such garments of HS chapter 61, knitted and crocheted garments, 62, not knitted and crocheted garments and 64, shoes. 21 The simple average is the arithmetic mean of utilization rates calculated at the HS 4-digit product level. The weighted average is calculated as the sum of imports receiving preferential treatment divided by the sum of imports eligible for the preferential treatment over all HS-4 digit level products and FTA partners. 22   The National Board of Trade Sweden/UNCTAD report acknowledged such a disparity shown in figure and 2 above: "The partner country exporters, however, use the EU's free trade agreements to a higher degree at a relative levelby 23 percentage points.
Admittedly, aggregated data of EU imports from partners countries contained in Table 1 below, showed a significant disparity on the utilization rates according to the direction of trade where the EU exports to the partner countries were showing significantly lower utilization rates than the corresponding exports from the partner countries. Table 1suggests an immediate reflection and a question: In an FTA, the rules of origin and related administrative procedures are the same. How is it possible to explain such an asymmetry in utilization rates depending on the direction of trade?
In addition, such asymmetry is somewhat counterintuitive since the EU partners, especially developing countries, should face more difficulties in complying with rules of origin requirements given their lower industrial base and possibilities of sourcing inputs is less developed and diversified than the EU counterpart. At more disaggregated level, it was shown that utilization rates vary substantially by direction of trade and partner, with substantial pockets of low utilization. The EU-South Korea FTA, the EU-Mexico FTA and the EU-Switzerland FTA, were objects of further analysis 26 given their relatively high asymmetry of utilization rates, their high values of trade and because they are representatives of different phases of evolution of EU rules of origin in the network of EU FTAs.
The EU-South Korea FTA is one of the first FTAs of the so-called 'new generation'; the EU-Mexico FTA was first signed at the time of the Pan Euro Med rules of origin in mid 1990s and the EU-Switzerland FTA is one of the first FTA entered by the EU back in the 1970s.   Figure 3(a), between 2011 and 2013, EU covered imports from South Korea increased from 8.6 billion USD to 15.5 million USD and utilization rates rose from 59% to 82.4%. Over the same period, the corresponding utilization rates of South Korean imports from the EU followed a similar pattern with increasing utilization rates even though starting from a lower level of utilization and remaining significantly lower throughout the period, reaching 67% in 2013 (see Figure 3(b)).
The asymmetrical evolution of the EU-South Korea FTA has been recognized during a comprehensive independent evaluation 27 carried out by a consulting company and a Commission staff working document 28 presenting its own evaluation as regards the findings and conclusions of the consultant company who carried out the evaluation. As shown below, in Figure 4, the 2019 report of the Commission 29 acknowledges that even if the utilization rates of the EU exports have increased from 50% in 2012 to 71% in 2016, there is still a rather wide gap compared to the utilization rate of 86% of South Korean exports to the EU.

Figure 4 -Preference utilization rates in Korea and the EU (%)
Source: European Commission staff working document (2019) 30 The interim report of the consulting company 31 indicated some reasons for such asymmetric utilization: a) low preferential margins associated to a cost/benefit analysis of using preferences b) South Korean Government promotion about using the FTA.
Besides these generic factors, the Mid Term review identified the following substantive reasons for such asymmetrical utilization: c) delays by EU customs authorities of delivering approved exporters status d) the direct consignment rules provisions contained in the EU-Korea FTAs e) non-compliance with product specific rules of origin.
In his working paper, the Commission, drawing from the consultant report, acknowledged that asymmetrical utilization rates was mainly due to burdensome customs procedures, namely: -"That the administrative burden related to the approved exporter status (AES) vary across Member States and can be time-and resource-consuming. Also, the exporters needed time to adapt". First, it should be noted that adaptation may take one year (i.e. 2012) but it is difficult to argue that exporters and customs still need adaptation after 4 years of practice. Moreover the AES has been in use for decades in the EU. -"the direct transport rule. It is considered to be burdensome for some industries who use regional hubs in Asia such as Singapore". This issue is linked with the drafting of the provision on direct transport provision in the EU-South Korea FTA.
First it is not clear why the Commission deviated from its established practice when negotiating the text of the EU-South Korea FTA protocol as further explained below. Second it is also not clear why it has not been possible to come to an understanding with South Korea under the Joint Committee on Rules of origin to find a solution in the course of the years. Apparently South Korea requires large trade offs for changing this provision.
The Commission mentions in the working paper that a number of actions has been taken to improve the asymmetric situation, namely: a) developing relevant web tools to assist exporters b) raising the issue of direct transport on numerous occasions with the South Korean authorities to reach an FTA amendments package.
Yet it may be wondered how these measures have been timely and effective in addressing the stumbling blocks that have been clearly identified using utilization rates as a monitoring tool.
Finally, the assertion in the Commission working paper that the low utilization rates in the machinery, electronics and chemical sectors "can be attributed to the complexity of the process to obtain proofs of origin for all the multiple product components as a result of the global value chain production" reinforces the sentiment that the Commission may not have wholeheartedly accepted the use of utilization rates and related instruments as monitoring tool to identify and address area of reforms on rules of origin and related administrative procedures.
As pointed out earlier in a FTA such the EU-South Korea the rules of origin and customs procedures are the same. South Korea is also exporting goods of the machinery, electronics and chemical sector to the EU why then the utilization rates should be lower for the EU, why the complexity of proof of origin and sourcing should affect more the EU than South Korea? There are not definite answers to these questions except to double-check these assumption with EU companies.
Once again the most recent National board of Trade Sweden report dismisses such enduring myths by showing that SMEs and large companies tends to use more trade preferences according to the value of transactions rather than by sector specific constraints or sourcing requirement Another reason raised by the National Board of Trade Sweden is that the analysis of preference utilization should focus more on importers, since the importers are the ones that benefit directly from the duty savings or might face the risk of having to pay the duties if there is any error in the request. The importers are likely the drivers of preferences utilization 32 .
As outlined in the preceding section, a finding of low or asymmetrical utilization should be accompanied by a text base analysis of the FTA protocol on rules of origin and fact-finding meetings and discussion with private sector.
In the case of the EU-South Korea protocol on rules of origin there is a rather striking difference in the texts on direct transport provisions from other agreements that lies at the heart of the question. Like many other FTAs, the EU-South Korea Protocol on rules of origin states that products must be transported directly between the EU to South Korea, and vice versa, in order to benefit from the tariff preferences of the FTA. According to such provision, exporters must provide customs authorities in the destination country with evidence verifying that the direct transport provision has been satisfied. The 32 If seen from this perspective the preference utilization by EU importers is high (above 90 percent) and the preference utilization by partner country importers is low (about 67 percent). In order to increase the preference utilization of EU exporters, the focus should be to facilitate the trade for partner country importers and their contact with customs authorities and other relevant stakeholders in these countries. usual article related to direct transport 33 contained in the EU FTAs provides that; " Evidence that the conditions set out in paragraph 1 have been fulfilled shall be supplied to the customs authority, in accordance with the procedures applicable in the importing Party, by the production of: (a) evidence of the circumstances connected with transshipment or the storage of the originating products in third countries; (b) a single transport document covering the passage from the exporting Party through the country of transit; or (c) a certificate issued by the customs authorities of the country of transit: (i) giving an exact description of the products; (ii) stating the dates of unloading and reloading of the products and, where applicable, the names of the ships, or the other means of transport used; and (iii) certifying the conditions under which the products remained in the country of transit.

(d) any substantiating documents to the satisfaction of the customs authorities of importing country
However paragraph 2 of article 13 titled direct transport of the EU-South Korea FTA does not contain the provision under (d), thus limiting the alternatives to provide documentary evidence for direct transport to a trough bill of lading as laid down in paragraph (b) above or what is normally defined a non manipulation certificate under (c).
These requirements of evidence of direct shipment are far from the best practices as recommended by WCO and presently discussed during the revision of Annex K on rules of origin of the Kyoto Convention. As resulted from interviews with private sectors and reported in various studies 34 and notes, firms are using hubs to split consignments 35 , mainly in Singapore.
In fact, EU exporters make use of logistical hubs (mainly in Singapore) for storage and related operations, such as repackaging and labelling, prior to distributing their products to various Asian markets, including South Korea 36 . Such repackaging and labelling in the country of transit (Singapore) disqualifies goods according to the rigid requirements of article 13 in the EU-South Korea protocol of origin mentioned above. Moreover, it has turned out, at interviews with the private sector, that the Korean Customs adopted a rather overzealous application of the protocol on rules of origin, including such provision. 37 In order to benefit from the preferential tariffs of the FTA, some companies have chosen to ship goods directly from the EU to South Korea. However, in these cases, companies cannot react swiftly to demand fluctuations, as shipping from the EU to South Korea can take well over a month 38 . Discussions have taken place to address this challenge but no solution has been found.
Most recently, during the negotiations of the FTA with Japan, the Commission seems to have learned the lesson by 1) inserting the non alteration clause that is replacing the concept of direct transport requirement and 2) the introduction of the exporter/importer self certification without prior approval. The introduction of self certification by the importer is an innovative feature in EU FTAs that till recently used "approved exporters" as self certification mechanisms. Under the "approved exporters" system only exporters that were approved by EU national customs administration or partner country were allowed to self certification. The EU-Japan FTAs introduced novelties since according to article 3.16 of protocol on rules of origin of the EU-Japan FTA "The importer shall be responsible for the correctness of the claim for preferential tariff treatment and compliance with the requirements provided for in this Chapter. origin that the product is originating made out by the exporter; or (b) the importer's knowledge that the product is originating.
Such innovation seems to have encountered a series of initial implementation obstacles that generated a number of guidance notes from both parties: EU 39 and Japan 40 .
Asymmetrical and low utilization have also been recorded in the case of the EU-Mexico FTA. In this case, the utilization rates are not only asymmetrical but also relatively low in both directions of trade, ranging between from 66% per cent to 70% per cent in the case of imports from Mexico into the EU, as shown in Figure 5, and from 48% per cent to 36% per cent in the case of EU imports into Mexico.

Figure 5 -European imports from Mexico and utilization rates
These low utilization rates are further analysed in section 3.3 below, where the application of the "repeated offenders" methodology detected numerous pockets of low utilization in the EU-Mexico FTA. In particular, the methodology identified the PSRO in the automotive sector in the EU-Mexico FTA as one the major culprits for low utilization. In fact, the original EU-Mexico FTA provided for stringent rules of origin in the automotive sector that have been liberalized in the course of the negotiations 41 for the updating of EU-Mexico FTA. The PSRO for cars and other vehicles (HS headings 8701-8707) is now equal to 45% max NOM (Non-Originating Material). Therefore, for the entire automotive sector, the permanent product-specific rules of origin are the same as in the EU-Japan or EU-Korea FTA. Figure 6 below, depicts a similar asymmetrical patter in the EU Switzerland FTA, raising a number of questions since such FTA is in existence since the 1970s and by any means should be a well-tested and widely known FTA for traders and the business community. Still, while the utilization rate is as high as 93 per cent for goods imported in the EU the corresponding utilization rates of Swiss imports from the EU is ranging from 73 per cent in 2011 with a lowering trend to 66 per cent in 2013. 39

Figure 6 -European imports from Switzerland and utilization rates
As there has not been a similar study on the implementation of the EU-Switzerland FTA, it is not possible at the moment to identify the reasons for such asymmetrical low utilization. However, a submission by the LDC WTO group 42 in the context of utilization rates of the Swiss GSP by LDCs found that the Swiss utilization rates for garments and clothing imports of HS Chapters 61 and 62 are ranging between zero and 49%, much lower than those observed in the EU that amount to 95% on average.
Given that EU and Swiss rules of origin for garments of chapters 61 and 62 are identical, the LDCs requested to clarify the reasons for such lower utilization rates in the Swiss market. In addition, it was also found that pearls and precious stones of HS Chapter 71 also showed extremely low utilization rates. Since these are primary products, normally considered as originating, they are mostly wholly obtained in the LDCs, it was difficult to find a reasonable explanation for low utilization. During bilateral meetings between Switzerland and the LDCs trade data was made available by the Swiss delegation to identify the reasons for such low utilization. The data provided by the Swiss delegation showed the utilization rates for direct imports, i.e. those that were imported directly into Switzerland, recorded very high utilization rates while indirect imports (mainly through the EU) showed significantly lower utilization rates. These findings points to the direct shipment in the EU-FTA 43 and present in the Swiss GSP as possible culprit for such low utilization, taking into account that Switzerland is a landlocked country where most of goods are transiting third countries before reaching the final destination. This preliminary finding is an useful indicator of a "malaise" or improper functioning of these preferential trading arrangements that would need to be tested in counterfactuals with firms and traders.

A methodology to identify the link between product specific rules of origin and low utilization of trade preferences
The previous section has shown evidence that a reading of the trends of utilization rates in conjunction with counterfactuals from the private sector can track the reasons for low utilization rates and lead to a solution whenever trade policymakers recognize the problem and take swift action. 42  The repeated offender methodology is a tool aiming at identifying those product-specific rules of origin and/or administrative related procedures that are responsible for pockets of low utilization in FTAs, and therefore where reform of PSRO may have significant trade effects.
This section conveys a strong policy message that, similarly to the EBA and Canadian GSP case 44 , reforms of rules of origin in EU FTAs could generate strong market responses for the EU exporters and importers, as well as FTA partners.
To do so, specific methodologies based on utilization rates and input/output matrices can be applied in order to (1) identify the PSRO where a reform would be the most beneficial and (2) which type of PSRO should be adopted to replace those product-specific rules of origin that are found to be causing low utilization

The "Repeated Offenders" Methodology
The repeated offender methodology is a tool aiming at identifying the product-specific rules of origin or administrative related procedures that are responsible for pockets of low utilization in FTAs, and therefore where reform of PSRO may have significant trade effects.
While the preceding sections have documented and discussed the magnitude of the trade flows affected by low utilization, the repeated offenders methodology applied in this section to the existing trade data and utilization rates of EU-FTAs, makes it possible to systematically identify the sectors at HS 4-digit level in which PSRO may be causing low utilization of trade preferences. The methodology comprises five steps divided into two parts, as reported below.
Part A: Identifying low utilization at product specific level (HS 4 digits 45 )

Filtering the data to identify "critical products"
a. The data are filtered to keep observations referred to as "critical products" showing a low utilization rate (UR) and high preference margin (PM) at the HS 4 digit-level, on average 46 . The thresholds applied in the present paper are UR<70% and PM>2pp.
b. After the filtering described in (a) above, observations are ranked in descending order of covered imports for each specific agreement and separately for each directions of trade (EU exports and EU imports). Rank 1, 2, 3, etc. respectively refer to the first, second, third, etc. most exported or imported critical product between the European Union and a specific FTA partner.
c. The analysis focuses on the R products at the top of the critical products ranking (Rank<=R) 47 . These products considered after the filtering of (a) and the selection made under (b) are referred to as "Critical Products at rank R".
The reasoning behind the filtering under (a) is twofold: 44 See Getting to better rules of origin for Least Developed Countries, forthcoming, UNCTAD, 2020 45 Such level of disaggregation is explained by the fact that most of PSRO of the EU-FTAs are set at HS 4 digit level. 46 The panel dataset is aggregated to a cross-session by averaging imports and preference margin over all available years for a given product and partner. See Technical Appendix available on request from the authors. 47 Alternatively, the authors considered the application of trade shares thresholds. Given the various level of export/import diversification across EU FTA partners and the high degree of specialization of some economies, such criterion is difficult to apply in practice. As an illustration, while 900 critical products are identified at rank 50 in EU imports (50 for each 18 FTA), restricting the cumulative import shares up to a maximum of 2% results in 1189 critical products, distributed highly unevenly over 8 countries: Egypt (270), Island (375), Korea (1), Morocco (3), Peru (240), Serbia (4), Tunisia (203) and Turkey (93). For all other FTAs, the most exported products to the EU represent a share in EU imports from that partner higher than the threshold of 2%. A criterion independent from export/import diversification is therefore more appropriate to account for all economies and avoid excluding crucial critical products from the analysis.
(i) to capture a significant size of non-optimal utilization with trade values that are not too low to be of a sporadic nature, and (ii) a margin of 2pp could be considered significant taking into account that preferential margin of 2,5pp on motor vehicles in the US market has generated around 100 pages of rules of origin text in NAFTA and even more in USMCA, still commanding headline in the press at this very days. 48

Identification of "repeated offenders"
Broadly speaking, the concept of the repeated offenders are products that have been identified as reporting a consistently low utilization rates across various FTAs or within an FTA in both directions of trade and with significant trade values in terms of covered imports. Following the terminology defined in step 1, two types of repeated offenders have been singled out: a. Bilateral repeated offenders (BROf): critical products at rank RB in both directions of trade (EU exports and EU imports) within a given agreement.
b. Repeated offenders across agreements (AROf): critical products at rank RA in one direction of trade, across a number NA of FTA partners.
The advantage of this methodology is to avoid considering isolated cases, for which the low utilization rate might result from various exogenous factors. Instead, with Bilateral Repeated Offenders, the products are critical, not only for the exports of an FTA partner to the EU or for the EU exports to that given partner, but for both parties at the same time.
Across-Agreements Repeated Offenders refers either to the critical products among the top RAI most imported products to the EU from at least NAI (>1) FTA partners, or to the top RAX most exported products from the EU to at least NAE (>1) destination markets. The value of the parameters applied in this study are given in section 3.2.
Part B: Identify PSRO for reform: Linking "critical products" and "repeated offenders" to PSRO of FTAs The second part of the methodology matches the critical products and repeated offenders that have been identified to their respective PSRO

Matching critical products with PSRO
From step (1) above, critical products can be matched with PSRO. This step only constitutes a preliminary analysis as a transition to the application of the repeated offenders' stricter criteria.

Matching bilateral repeated offenders with PSRO
Within a given FTA, PSRO are identical in both directions of trade. Therefore, a product that is critical for EU exports to the partner country AND for the partner's exports to the EU may be the sign of inadequate PSRO within the agreement. Therefore, it is crucial to match the bilateral repeated offenders with their respective PSRO. This makes it possible to examine these rules and understand what are the underlying reasons for low utilization in order to make recommendations for reform. Robert Schuman Centre for Advanced Studies Working Papers

Matching bilateral & cross-agreements repeated offenders with PSRO as priority candidates for reform
This step consists in identifying the critical products that are included in both bilateral and crossagreements repeated offenders (for EU imports, EU exports and for both), and then to match them to the PSRO of each concerned FTAs.
Although PSRO are not identical across agreements, some similarities can be observed. Therefore, critical products that are bilateral repeated offenders in several agreements might be good candidate for reform, requiring to match and analyze the PSRO of these two dimensional (bilateral and crossagreements) repeated offenders.

Box 1 -"Repeated offenders" methodology Bilateral vs. across agreements at EU exports and imports
Looking jointly at AROf in both directions of trade, EU exports and imports makes it possible to identify the sectors that are critical for EU exports to several FTA partners and at the same time for EU imports from several FTA partners. In contrast to bilateral repeated offenders, this approach does not require the sectors to be critical at the bilateral level. In other words, the FTA partners to which EU export the critical product and those from where sourcing takes place are not necessarily the same. This is particularly important when looking at sectors that are not exported and imported at the same time in large amounts to/from a given FTA partner, for structural or cyclical reasons.
In the present paper, only HS2710 is excluded from the bilateral analysis. However, when significant differences are observed between products identified with or without bilateral restrictions, an additional step could be added to the methodology before jointly merging AROf and BROf to PSRO. This would allow to analyze and compare the PSRO of the "repeated offender" sector in each AROf agreement, and attempt to identify possible specific common feature leading to low utilization, such as stringent restrictiveness or unnecessary administrative burden.
Critical products identified as repeated offenders across agreements on EU exports and/or EU imports could be matched with their respective PSRO without the bilateral restriction. As PSROs may vary across agreements, this would require a comparison across all the agreements where the product is considered as critical. Alternatively, a preliminary analysis can be conducted based on the agreement(s) showing the lowest rank (i.e. for which the critical product is among the top most imported/exported critical product).
Annexes 1 and 2 contain diagrams showings for each FTA examined the results of such filtering in both directions of trade.

Data
The analysis is using the National Board of Trade Sweden/UNCTAD dataset composed of 17 FTAs (see section 0), with the following exceptions:

EU imports
o Exclusion of Nicaragua as EU sourcing, given the unavailability of FTA duty rate to compute the preference margin.
As a result, 14 reporting FTA (EU export destinations) and 18 EU sourcing countries are considered in the analysis. The panel dataset is first averaged over the available years. The analysis is therefore applied to the resulting cross section of countries and tariff headings. The detailed dataset is described in the technical appendix.
In this analysis, the parameters are set as follows: RB = 30: Bilateral repeated offenders are defined as critical products that are critical and lie among the top 30 critical products in terms of covered imports in both direction of trade within a given agreement. RAI = RAX = 10; NAI = NAX =2: Across agreements repeated offenders are defined as the products that lie among the top 10 most exported (resp. imported) critical products to (from) at least two FTA destination (sourcing) countries.
Threshold for critical products are applied as previously defined, utilization rates below 70% and preference margin above 2pp.  Table 2 provides an overview of EU critical trade. Column (2) indicates the value imports of tariff headings covered by the FTA exhibiting a preference margin above 2 percentage points and a weighted utilization rate lower than 70% while column (3) reports the corresponding share out of the total covered imports independently of the utilization rates or preference margin rates.

Identification of "critical products"
The most striking evidence is that the share of critical products is systematically higher for EU exports than for EU imports. Out of the total Korean imports from the European Union that are covered by the FTA, 73.9% has been found critical, showing low utilization, 50% on average (see column 5) and a preference margin above 2pp. In contrast, with a similar utilization rate of 51%, the share of critical trade amounts only to 25.5% when considering EU imports from South Korea. Therefore, to correctly characterise this finding utilization rates are not asymmetrical but the amounts traded on the critical tariff headings are. Such findings can be interpreted in two ways, both of them pointing towards a fixed cost in utilizing the trade preferences, in line with earlier discussion on the administrative procedures related to rules of origin discussed in section 0: (i) Asymmetry in utilization rates is observed only for relatively high values of utilization rates.
Therefore, conditional on reaching high utilization above 70%, EU exporters face more difficulties than foreign exporters to raise utilization to higher values. This "residual non-utilization" of the preference, or "utilization rate ceiling" lower than 100% could be due to fixed costs and administrative procedures preventing exporters to use the preference despite the massive potential gain resulting from the massive amount trade on these critical tariff lines. The answer may also be found on the importer side, since importers seem to be the drivers for preference utilization.
(ii) The utilization rates are asymmetric only when the preference margin is relatively low. Therefore, the similar efforts are placed by exporters in both direction of trade to utilize the preference when the preference margin is commercially meaningful, independently of the traded amount. However, considering potential economies of scale, higher trade values are expected to translate into higher utilization rate as the marginal cost related to the use of the trade preference decreases. Since we do not observe such pattern, but in contrast, utilization rates that are independent of trade values, the findings point towards fixed costs, possibly resulting from administrative requirements related to rules of origin.
In the case of the EU-Switzerland agreement, the asymmetric utilization rates discussed in section 0 and reported in Table 1 amount to 72% on EU exports and 91% on EU imports. Because figures in Table 2 capture exclusively critical trade with utilization rates below 70%, only 196 tariff headings are considered for EU exports and 55 for EU imports (see column (4)), leading to lower shares of critical trade of respectively 11% and 10% for EU covered exports/imports to/from Switzerland, and to utilization rates of 45% (EU exports) and 48% (EU imports). It is worth noting that this similar magnitude of critical exports/imports shares and of utilization rates does not necessarily rule out asymmetries at the HS heading level.
Indeed, if Table 2 provides an interesting preliminary overview of the potential asymmetries within FTAs and of the variability of critical trade shares between FTAs, a more disaggregated analysis is needed to study the major source of heterogeneity related to the product characteristics. Table 3 therefore reports the EU's most exported critical products to FTA partners, exhibiting utilization rate below 70% and a preference margin above 2pp. Columns (1) and (2) report the product HS 4-digit code and its corresponding description while column (3) indicates the name (ISO 3 code) of the considered FTA partner. The utilization rate in column (4), preference margin in (5) and the value of covered imports in (6) all refer to imports from the European Union to the FTA partner reported in column (3). Filters to identify the critical products are applied in columns (4) for utilization rates and in column (5) for preference margin. Results are reported in descending order of the FTA partners' covered imports of column (6) 49 . For completeness and comparison purpose, column (7) and (8), report the corresponding non-filtered utilization rates and preference margin in the other direction of trade, from the FTA partner reported under (3) to the EU, with the amount of covered imports in column (9).
Preliminary observations can be drawn from Table 3. First, most of the critical products are petroleum oils, cars and parts thereof, and machinery of chapter 84. Second, beyond this sectoral concentration, FTA partners importing significant amounts of such products from the EU are mostly Mexico, Korea, Egypt, Switzerland and Morocco.
It has to be noted that the preponderant volume of trade not fully utilizing trade preferences under HS Chapter 27, petroleum oils and derivatives, are often imposed MFN duties between 3 and 4pp. Given the high values of trade, this represents a tremendous amount of potential duty savings. As an illustration, with a utilization rate of 62% and an average preference margin of 3.4pp, the amount of MFN duty paid on the 4.2 billion exports of products of HS 2710 to Turkey can be evaluated at 55 million USD 50 is often imposed an MFN duty of 2.5 % in the EU. The second most exported critical products are cars and parts of cars that are attracting a high average of MFN rates in Mexico.  (6); nes. stands for "not elsewhere specified"; Only critical products with FTA covered imports above USD 450 million are reported. Table 4 provides a similar snapshot in the reverse side of trade i.e. EU imports from partner countries with data sorted in descending order of covered imports (6). The results mirrors to a certain extent the findings of the previous table in the sense that there is a relatively high concentration of products and FTA partners that are recording low utilization. Chemicals and Switzerland appears to be countryproduct pairs showing substantial volumes of trade and suffering from low utilization. Cars and parts of motor vehicles also feature low utilization in this direction of trade from the same partners, mostly Mexico and Switzerland. Besides these specific products, it has to be noted a great variety of products and partners recording a low utilization rates.

Repeated offenders
Bilateral repeated offenders Table 5 reports the bilateral repeated offenders for each FTA, namely the critical products with UR<70%, PM>2pp lying among the top 30 products in terms of covered imports for both EU imports and partner country imports. Observations for each FTA are sorted in descending order of total covered trade (EU imports and partner country imports) reported in column (4). The latter corresponds to the sum of column (8) and (11).
The table further strengthens the previous findings pointing to a number of products and FTAs that have appeared in both Table 3 and Table 4, with high trade values, such as Switzerland for motorcycles, bicycles and selected garments, South Korea for a variety of HS headings, Mexico and Turkey, etc.
Looking only at products with covered imports above 80 million USD, which represent twenty-four products of the fifty-seven reported in Table 5, there is a rather concentrated pattern of FTAs that are routinely encountering problems with different products like Mexico (9 products out of twenty-four) or South Korea (seven products out of twenty-four). With trade above 200 000 million, sixteen products remain in the list of bilateral repeated offenders, seven for each EU-Mexico FTA and South Korea FTA, one for EU-Egypt FTA and one for EU-Switzerland FTA (see rows highlighted in green in Table 5 below). Robert Schuman Centre for Advanced Studies Working Papers  Table 6 and Table 7 below report the subsequent step of the methodology, that is, the identification of those products at HS 4-digit level that are found to report a low utilization rates across a number of FTAs for EU exports or imports. It is obvious that the more a product is found to report a low utilization rate in more than one FTA for a significant amount of trade, the more it is likely that the PSRO associated to that product is causing such low utilization.

Across FTAs repeated offenders
Despite the fact that PSROs can vary across agreements, the repeated offenders highlights that some PSROs are consistently inadequate across various agreements and need to be given a serious consideration for potential reform. Table 6 reports the observations that are among the top ten products exported to the EU (RAI = RAX = 10) with a utilization rate below 70% and a preference margin above 2pp for at least two different FTAs (NAI = NAX =2). The first two columns refers to the product identified as repeated offender. Columns (3) reports the FTAs for which the product is critical together with the rank of such product in the EU critical imports within this FTA. For example, HS 8703 is the first critical products in terms of EU covered imports from Montenegro (MNE1), the third for Egypt (EGY3), the fourth for Albania (ALB4) and Norway (NOR4), the fifth for Switzerland (CHE5), etc. All the countries reported in column 3 therefore face difficulties to use the preference when exporting to the EU. Note that the case of Switzerland for 8703 already appeared in Table 4. This analysis therefore strengthen the previous findings by showing that the product is not only critical for Switzerland exports to the EU but also for a number of other FTAs whose lower export values did not allow them to appear in Table 4, but for which HS 8703 still corresponds to a significant amount of trade since it lies among the top 10 critical products in terms of covered exports to the EU. Table 6 displays the critical product rank until a maximum of 30. However, the selection criteria for the product to be considered as repeated offender and therefore be reported in the table is that at least two FTAs report for the same product a rank of being critical below or equal to 10. In the previous example for HS 8703, not two but eight FTAs are satisfying the condition (in bold). 51 Column (4) reports the EU covered imports from all partners listed in column (3). This therefore represent the total value of "critical" imports at rank 30. Colum (5) shows the preference margin while column (6) and (7) are the simple and weighted average utilization rates of EU imports from the countries listed in (3). A simple average is calculated as the arithmetic average of the utilization rates of each country reported in (3) for the specific product while the weighted utilization rates is the sum of imports from the group of partners receiving the preference divided by the covered imports from the same partners.
Overall, the table finds again some of the products that appear regularly during this analysis as recording low utilization: petroleum oils, cars and parts of car thereof, a chemical product. In addition and differently from previous figures there are a rather conspicuous number of electrical machinery of chapter 85.
Indeed, the utilization rates, both arithmetic and weighted, reported in column (6) and (7) as explained above are showing extremely low utilization rates as low as 22 to 27% for boards and panels of HS 8537, or 17% to 8% for television of HS 8528. Utilization rates below 35% are reported in red. 51 Another example is HS 8708 that reads as follows: MEX1; LBN2; ISL3; CHL4; DZA11; PER14. The selection criteria (the reason why this product appears in the table) is because four countries report this product as critical with a rank below 10 (Mexico, Lebanon, Island, and Chile), which is above the selection criteria minimum threshold of two countries (4 >NAI =2).

Using Utilization Rates to Identify Rules of Origin Reforms: The Case of EU Free Trade Area Agreements
European University Institute 27 Some FTAs appear more regularly than others. For EU covered import values above 50 million, South Korea has nine critical products with rank below or equal to 10 and is followed by Algeria with seven products and Mexico with six products. 52  (3) and associated imports in (4) are restricted to critical products at rank 30. Table 7 reports the results of the analysis in the opposite direction of trade when EU exports to the FTA partners. The reading of the table is similar as for EU imports with the product HS code and description in columns (1) and (2) and the partner's name and rank in column (3). Colum (4) shows the covered imports from the EU aggregated for all FTA partners listed under (3) and is therefore an indication of the overall exports of EU to the various FTAs where the utilization rate is below 70% and the preference margin above 2pp.

Table 5 -Across agreements repeated offenders -EU imports Critical products with product rank<=10 for at least two FTA sourcing partners (in bold) (sorted in descending order of EU covered imports (4))
The most striking finding is the magnitude of trade flows that are significantly higher than on the EU import side. Therefore, despite that the utilization rates are not systematically lower when considering EU exports as compared to EU imports 53 , results of Table 7 highlight a potentially important trade policy indication. They clearly show that despite important amount of trade, EU exporters are facing difficulties in fully utilizing trade preferences and/or partner countries are facing difficulties in administering RoO in the FTA, leading to potential massive losses in terms of duty savings.
In addition in most of the cases, the utilization rates is significantly lower than the ceiling of 70% for both arithmetic and weighted utilization rates reported in column 6 and 7. For the majority of products showing a significant amount of trade in terms of US billions i.e. the first 9 products at HS-4 digit, the utilization is around 50% or lower and this is equivalent to billions of trade flows from the EU to the trading partners that have not utilized trade preferences.
The repeated offenders are petroleum products, cars and parts thereof, machinery of chapter 84 and chemicals. As in the previous figure, a number of FTA partners seems to appear regularly as importers of EU goods that have not benefitted from trade preferences. For trade values above 100 000 million, Egypt leads the pack with 9 products with rank below or equal to 10, followed by Switzerland and Korea reporting 8 products, and 6 for Croatia, Mexico and Montenegro 54 .
The 10 products appearing in both Table 6 and Table 7 are highlighted in each table. They correspond to the HS headings that are critical at rank 10 for a minimum of 2 FTAs, for EU imports and EU exports. It can already be noted that except HS 2710, all of these products have also been identified as bilateral repeated offenders in Table 5.
These nine remaining tariff headings will therefore be proposed as priority candidate for reformed PSRO in the last step of the methodology combining AROf and BROf (section 3.3.5). 53 When applying the repeated offenders methodology, the asymmetry found on non-filtered aggregated data in the first report does not systematically hold but is still valid when considering the magnitude of critical trade that is much higher for EU exports than EU imports. See discussion in section 0. 54 Further frequencies are as follows: five for Albania and Serbia, four for Morocco, three for Algeria and Turkey, 2 for Island and Nicaragua, and 1 for Macedonia. Robert Schuman Centre for Advanced Studies Working Papers Note: Utilization rate lower than 35% reported in red. ARO at exports AND imports, highlighted in red. FTA partners reported in (3) and associated imports in (4) are restricted to critical products at rank 30.

Matching critical products with their respective PSROs
Using the most exported and imported critical products reported in Table 3 and Table 4 above, a preliminary identification and analysis of the corresponding PSRO can be carried out. Table 8 and Table 9 below report the various products previously identified as being critical and their corresponding PSRO (column (5)). For each product, a reformed rule is proposed in column (6). The proposals for alternative PSROs are tentative examples of alternative PSRO drawn mostly from the recent EU-Japan FTA and subject to further analysis and checks with the private sector of the specific industry to verify the findings and listen to suggestions. A forthcoming study on convergence of PSROs will provide more options of alternative PSROs based on best practices. 55 Robert Schuman Centre for Advanced Studies Working Papers

Table 7 -EU most exported critical products at rank 30 and corresponding rules of origin (FTA partners critical imports from EU in descending order of covered imports, USD million (4))
HS Product FTA Cov. Imp.

(1) (2) (3) (4) (5) 2710
Petroleum oils, etc, TUR 4'234; Operations of refining and/or one or more specific process(es) (1) or Other operations in which all the materials used are classified within a heading other than that of the product. However, materials of the same heading as the product may be used, provided that their total value does not exceed 50% of the ex-works price of the product For purposes of heading 2710, the following processes confer origin: (a) Atmospheric distillation: A separation process in which petroleum oils are converted, in a distillation tower, into fractions according to boiling point and the vapor then condensed into different liquefied fractions. (b) Vacuum distillation: Distillation at a pressure below atmospheric but not so low that it would be classed as molecular distillation.  58 The origin of petroleum products has not been the object of sufficient studies even if some complex origin questions may arise. See "The Petroleum Industry and Free Trade Agreements: How Oil Companies can Benefit" available at https://www.slideshare.net/PeterMachielse/the-petroleum-industry-and-free-trade-agreements and the NAFTA origin problem for Canadian producers, https://business.financialpost.com/commodities/energy/canadas-oilpatch-pays-america-60-million-a-year-to-export-crude-and-usmca-may-not-help 59 In reality the original product specific rules of origin under the EU-Mexico FTA of 2000 listed a number of subdivisions for the heading 8708 that are not reported in this table for the sake of brevity,

HS Product
FTA Cov. Imp.  However, the value of all the materials of headings 2932 and 2933 used shall not exceed 20 % of the ex-works price of the product or Manufacture in which the value of all the materials used does not exceed 40 % of the ex-works price of the product  CTSH;  A chemical reaction, purification, a change in particle size, production of standard materials, isomer separation or biotechnological processing is undergone;  MaxNOM 50 % (EXW) KOR 11  2924 Carboxyamidefunction CHE 983 Manufacture from materials of any heading, except that of the product. However, materials of the same heading as the product may be used, provided that their total value does not exceed 20 % of the exworks price of the product or Manufacture in which the value of all the materials used does not exceed 40 % of the ex-works price of the product  CTSH;  A chemical reaction, purification, a change in particle size, production of standard materials, isomer separation or biotechnological processing is undergone;  MaxNOM 50 % (EXW); 2934 Other heterocyclic compounds CHE 573 Manufacture from materials of any heading.

Rules of Origin
However, the value of all the materials of headings 2932, 2933 and 2934 used shall not exceed 20% of the ex-works price of the product or Manufacture in which the value of all the materials used does not exceed 40 % of the ex-works price of the product  CTSH;  A chemical reaction, purification, a change in particle size, production of standard materials, isomer separation or biotechnological processing is undergone;  MaxNOM 50 % (EXW); 0802 Other nuts TUR 519 Manufacture in which: -all the fruit and nuts used are wholly obtained, and -the value of all the materials of Chapter 17 used does not exceed 30 % of the value of the ex-works price of the product All products of chapter are wholly obtained 2710 Petroleum oils DZA 502 Operations of refining and/or one or more specific process(es) (2) or other operations in which all the materials used are classified within a heading other than that of the product. However, materials of the same heading as the product may be used, provided that their total value does not exceed 50 % of the ex-works price of the product CTH except from biodiesel of subheadings 3824.99 and 3826.00; or Distillation or a chemical reaction is undergone, provided that biodiesel (including hydrotreated vegetable oil) of heading 27.10 and subheadings 3824.99 and 3826.00 used is obtained by esterification, transesterification or hydrotreatment. 2922 Oxygenfunction aminocomp.
CHE 228 Manufacture from materials of any heading, except that of the product. However, materials of the same heading as the product may be used, provided that their total value does not exceed 20 % of the exworks price of the product or Manufacture in which the value of all the materials used does not exceed 40 % of the ex-works price of the product  CTSH;  A chemical reaction, purification, a change in particle size, production of standard materials, isomer separation or biotechnological processing is undergone;  MaxNOM 50 % (EXW); 0808 Apples, pears, quinces CHL 218 Manufacture in which: all the fruit and nuts used are wholly obtained, and the value of all the materials of Chapter 17 used does not exceed 30% of the ex-works price of the product Eliminate or lessen the requirements for use of sugar of chapter 17

Matching bilateral repeated offenders with PSROselected examples
In Table 10 below, a few examples of bilateral repeated offenders that have been matched with their corresponding rules of origin are provided. Similarly to the previous section, for each case, a reformed product-specific rule of origin is proposed in the last column.

Matching bilateral & cross-agreements repeated offenders with PSRO as priority candidates for reform
The last step of this analysis has been to combine the two types of Repeated Offenders and to keep only the products that are included in both type of repeated offenders, the bilateral and cross-agreement repeated offenders.
There are four options to combine the bilateral repeated offenders with the cross agreement repeated offenders as follows: a. EU exports: 19 repeated offenders identified corresponding to 12 HS headings.
b. EU imports: 35 repeated offenders identified corresponding to 22 HS headings.
c. Both directions of trade taken together as a cumulative and therefore stricter condition ((a) and (b)): 16 repeated offenders identified corresponding to 9 HS headings.
d. EU exports and imports as alternative conditions ((a) or (b)): 38 repeated offenders identified corresponding to 25 HS headings. Table 11 reports the results of methodology (d) and allows to identify the repeated offenders of the more restrictive methodologies (a)-(c). Notes: Data sorted in descending order of (i) total critical trade at rank 30 (sum of (8) and (9)), ie. EU imports and exports of critical products at rank 30 from/to partners reported in (7) and (11), and (ii) covered bilateral trade within a given tariff heading (EU imports and exports, (3)); Highlights reading grid: Red -9 repeated offenders at the bilateral level and across agreements for both EU imports and EU exports, corresponds to methodology case (c). Blue -3 bilateral repeated offenders repeating across agreements on the EU export side only. Red and Blue RPOf together corresponds to methodology case (a) White -26 bilateral repeated offenders repeating across agreements on the EU import side only. Red and white RPOf together corresponds to methodology case (b); Interestingly, the 16 repeated offenders identified in this section have already been detected in Table 6 and Table 7. The bilateral restriction imposing that the good is exported imported in large quantities from/to the same country applied in this section only results in the exclusion of HS 2710 (see discussion in section 3.1, Box 1). A brief comparison of the PSROs under the respective FTAs identified as bilateral and cross agreement repeated offender and the proposed reformed PSROs clearly shows that the proposed PSROs excerpted from most recent EU FTAs are remarkably more liberal than the previous ones.
This final finding strengthens once again the link among stringency of PSROs and low utilization of trade preferences. This being said the proposed PSROs should be tested and validated as a viable reformed PSROs with the firms operating in the specific industrial sector. Note: CR R30 refers to trade of critical products at rank 30, the total of EU exports and imports of critical products with rank below or equal to 30 to/from FTA partners.

Conclusion and way forward
Utilization rates can offer a merciless picture of the effective use of a preferential trade arrangement and this is the main reason why Governments and trade officials are wary of utilization rates especially when linked to rules of origin. This is not surprising since low utilization rates are showing to Governments and trade officials that certain rules of origin and related procedures in a PTA are not being negotiated effectively, forcing them back to the negotiating table and to their parliament. Robert Schuman Centre for Advanced Studies Working Papers In some cases low utilization rates in specific sectors may also be the deliberate result of targeted negotiation aiming at insulating specific industries from the competition deriving from FTA 62 The analysis in this study clearly shows that certain rules of origin provisions, for instance article 13 on direct transport of the EU-South Korea FTA could have been better formulated, that the product specific rules of origin in the EU-Mexico FTA needed revisions that were recently introduced and that further analysis is needed to understand the reasons for asymmetrical low utilization of the EU-Switzerland. On this latter case the recent debates in the WTO Committee on Rules of Origin over low utilization rates of the Swiss GSP by LDCs points to administrative requirements related to direct transport and territoriality requirement as being possible causes for such low utilization.
The methodology adopted in this paper clearly identifies critical products and product-specific rules of origin (PSROs) that need reform as a result of bilateral FTAs and cross-FTAs analysis carried out in this study. These repeated offender product-specific rules of origin are the obvious candidate for reform. Some findings of this report have already found their way and have been incorporated in the EU-Japan FTA and the modernized EU-Mexico FTA especially in automotive. Much more need to be done. The results of this study identify other product-specific rules of origin in the petroleum, chemical, garments and machinery sectors as repeated offenders and responsible for pockets of low utilization rates. Further consultations with the private sector to discuss these findings and counterfactuals would make it possible to better understand the reasons for such low utilization and formulate or validate the alternative rules of origin based on best practices suggested in this study.
Finally, it is extremely important to make clear that this study advocates a methodology to identify the reasons for low utilization due to rules of origin and related administrative procedures that should by used by Government and trade officials. It does not suggest magical solutions. It suggests that Government and trade officials recognize the use of utilization rates and a rigorous analysis as a valid tool to identify and diagnose clearly and unequivocally the reasons for pockets of low utilization of a FTA, mostly due to rules of origin and related administrative procedures.
Such rigorous analysis and its results, as unpleasant as it may be, should be recognized, seriously considered, and acted upon by Governments and trade officials rather than resort to denial attitudes and anecdotal evidence based on enduring myths. 62  Utilization Rates (lines, right axis) and value of covered imports (bars, left axis) for the five top products (HS code on the right) in descending order of average covered imports with average preference margin > 2 and average UR<70% Utilization Rates (lines, right axis) and value of covered imports (bars, left axis) for the five top products (HS code on the right) in descending order of average covered imports with average preference margin > 2 and average UR<70% Utilization Rates (lines, right axis) and value of covered imports (bars, left axis) for the five top products (HS code on the right) in descending order of average covered imports with average preference margin > 2 and average UR<70%