THE CANADIAN PREFERENTIAL TARIFF REGIME AND POTENTIAL ECONOMIC IMPACTS OF ITS EROSION. by Przemyslaw Kowalski 1.

THE CANADIAN PREFERENTIAL TARIFF REGIME AND POTENTIAL ECONOMIC IMPACTS OF ITS EROSION by Przemyslaw Kowalski 1 Przemyslaw.Kowalski@oecd.org Preliminary draft not for citation, comments welcome Version 29 July 2005 Prepared for International Symposium on "Preference Erosion: Impacts and Policy Responses" Geneva, June 13-14, 2005 1 Przemek Kowalski is an economist at the Trade Directorate of the OECD Secretariat. The views expressed in this paper do not necessarily represent those of the OECD or any of its members. The author would like to thank Raed Safadi and Simon Evenett for their helpful comments, Karinne Logez for her statistical assistance and Caroline Mirkovic for research assistance. The assistance of Diane Kelloway of the Finance Canada in providing the underlying trade data and associated technical explanations was greatly appreciated and made this paper possible. All remaining errors are those of the author. 1

Executive summary In 2003, Canada s imports from NAFTA partners accounted for approximately 65% of its total imports. This agreement offered significant margins of preferences over MFN duties and, to a lesser extent, over GPT and LDCT treatments. Canada s imports form developing countries accounted for 18.6% of its total imports. 74% of imports from developing countries entered Canada under MFN treatment at an average tariff rate of 5.8%, 15% under GPT treatment at an average tariff rate of 2.2% and 0.7% under LDCT treatment at an average tariff rate of 0%. The extensive reliance of several developing countries on non-preferential access to Canada s market suggests their strong interest in further liberalisation of the MFN regime. Cases of over-reliance on preferential trade with Canada are limited to few LDCs where trade under preferential schemes accounts for the bulk of total exports to Canada. As a result, the latter group of countries appears as potential opponents to further cuts to MFN tariffs. One of the possible solutions to the preference erosion problem that may be considered in the DDA negotiations is a simultaneous lowering of MFN and corresponding preferential rates so that preferential margins are preserved. This paper finds that out of 19,147 records where positive imports from developing countries were recoded in 2003 simultaneous reductions of MFN and preferential rates would be in principle feasible on up to 7,582 lines covering 66% of total imports from developing countries entering under non-reciprocal preferential schemes. 7,382 lines covering 24% of preferential trade would not be affected by MFN cuts since on these lines both the MFN and preferential rates are already zero. Declines of preferential margins would be inevitable on 4,183 lines covering 20% of total preferential imports from developing countries including all preferential trade with LDCs. As far as the value of preferences is concerned, the main beneficiaries include several more advanced developing countries such as Mexico, South Korea, China, India, Brazil, Israel, Malaysia Hong Kong-China as well as some LDCs such as Bangladesh, Cambodia or Haiti. Expressed as percentage of total beneficiary s exports to world the value of preferences does not typically exceed 1%. Nevertheless, several LDC countries exhibit higher-than-average ratios of value of preferences to total exports (0.52% in Bangladesh, 0.43% in Cambodia, 0.28% in Haiti, 0.23% in People s Democratic Republic of Laos, 0.22% in Maldives, 0.20% in Lesotho). Remarkably, this seems to be a consequence of the 2003 extension of the LDC scheme to textiles and clothing. Imports under the LDC scheme exhibit a heavy concentration in textile and clothing products Five HS textile and clothing chapters (61-65) accounted for 98% of imports entering under the LDC scheme. High tariff-line concentration of value of preferences in the case of LDCT and, to a lesser extent, GPT schemes suggests that exclusion of relatively few lines from MFN liberalisation could alleviate a substantial part of potential negative effects of preference erosion in the Canadian market. However, the benefits of such an option should be compared with benefits (to Canada and its other trading partners) that could be obtained by reducing MFN tariffs on these lines. Bearing in mind the qualifications associated with the use of the particular model and data, a conclusion that could be drawn from the simplified model assessment implemented in the paper, is that the welfare impact of Canadian preferences is very small for most developing 2

countries analysed here. Notwithstanding the significant variation in the utilisation and value of preferences across beneficiaries, this is consistent with the generally small shares of preferential trade with Canada in these countries total trade. Therefore, erosion of preferential schemes does not appear as a major hurdle to further reduction of Canadian MFN rates. 3

Introduction This paper seeks to identify Canada s trade partners and products that may be potentially vulnerable to the problems of preference erosion following MFN liberalisation under the Doha Development Agenda. This is achieved through a two-track approach. First, a detailed statistical analysis of the structure of Canadian preferences is undertaken drawing on the trade preferences database collected by the OECD Secretariat. 2 The database contains actual trade flows under preferential arrangements for Canada in years 1998, 2002 and 2003 (for description of database coverage and methodology see Lippoldt and Kowalski, 2005). The statistical analysis includes overviews of: structure of preferential trade flows; associated tariff levels and preferential margins; utilisation, utility and coverage rates; the value of preferences by scheme and beneficiary as well as information on key tariff lines where the value of preferences is concentrated. The second complementary approach in this paper employs the standard GTAP model and the GTAP 6 database to examine the trade and welfare impacts of removing Canada s preferential duties. An overview of Canadian tariff preferences The extent and potential economic impacts of Canada s preferential schemes on beneficiary countries need to be considered in relation to market access conditions that Canada s other trading partners face in the Canadian market. In particular the trade impact of GPT and the LDC schemes appear to have been limited by the relatively low protection levels afforded to other trading partners be it on the MFN basis or through reciprocal trade agreements (free trade areas). Indeed, judging by the simple average tariffs, Canada s tariffs are relatively moderate. Simple average tariff on all MFN tariff lines was estimated at around 6.8% in 2002 (WTO, 2003) and simple average tariff on tariff lines where imports from developing were positive was estimated at 5.8% in 2003 (see Table 3). However, noticeable differences are reported in the levels of protection of agricultural and non-agricultural products (WTO, 2003). For example, the average MFN tariff calculated for Agriculture 3 in 2002 was 21.7% as compared to an average of 4.2% for Non-agricultural products (WTO, 2003). Within agricultural products category, the highest tariffs are placed against imports of Dairy products (237%), Live animals and products thereof (52.7%), Beverages and spirits (8.3%) and Fruits and vegetables (4.8%). Within non-agricultural products, average MFN tariff rate in Textiles and clothing was calculated at 9.9%. It is also reported that tariff escalation continues to inhibit exports of downstream products to Canada by countries exporting under MFN treatment (WTO, 2003). This is the case particularly for imports of food and beverages, textiles and clothing, wood products, chemicals and nonmetallic mineral products. Currently, Canada has five reciprocal preferential agreements: Canada-Chile Free Trade Agreement, Canada-Costa Rica Free Trade Agreement, Canada-Israel Free Trade Agreement, Canada-United Sates Free Trade Agreement, and North-American Free Trade Agreement. These reciprocal agreements offer significant margins of preferences over MFN duties and, albeit to a lesser extent, over GPT and LDCT treatments (see Table III.2 in WTO, 2003). For example, under NAFTA 98.8% and 93.8% of tariff lines of imports originating, respectively, from the US and Mexico were duty free in 2002 (WTO, 2003). The average import duties for United States and for Mexico under NAFTA were 2.6% and 2.7% while those for the GPT and LDCT were respectively 5.4% and 4.1%. The significance of these advantages and their potential impact on market access of other trading partners is further strengthened by the sheer shares of the NAFTA partners in the Canadian market in 2004 imports from the US and Mexico accounted together for around 63% of total Canadian imports (see Table 1). The remaining free trade agreements also 2 Tariff-line data on Canada s imports from developing countries were kindly provided by Finance Canada. 3 According to the WTO definition. 4

grant substantial advantages to respective trading partners but the volumes of trade are significantly smaller as compared to NAFTA. Non-reciprocal preferential schemes Table 2 presents basic information on Canada s non-reciprocal preferential tariff schemes with respect to developing countries. Currently Canada provides non-reciprocal tariff preferences to developing countries under the Generalized Preferential Tariff (GPT), the Least Developed Country Tariff (LDCT) and under the CARIBCAN. However, certain Canadian imports from developing countries are classified under other preferential (reciprocal and non-reciprocal) schemes (see Table 3 for the composition of preferential trade flows). The Generalized Preferential Tariff (GPT) scheme took effect in 1974 and was subsequently renewed and expanded in 1984, 1995 and 2004. It is worth mentioning that the 1995 revision aimed to take into account the effects of erosion of preferential margins following the Uruguay Round negotiations, mainly through expansion of product coverage and lower GPT duty rates (UNCTAD, 2001). At present GPT rates granted for selected agricultural and industrial products 4 range from duty-free to MFN rates. However, the evolution of Canada's tariff preferences in favour of developing countries over the past two decades reflects mainly a number of special measures introduced for LDCs (see e.g. Weston, 2003). In 1983 LDCs were granted a zero rate on GPT-covered products with exception of clothing, footwear, certain labour-intensive industrial products as well as some agricultural products. As of 1999, and reflecting consultations held since 1994, the GPT (and LDC) product coverage was extended by some 220 product lines and the GPT tariffs were lowered to two thirds of the corresponding MFN rates. As pointed out by Weston (2003), in addition to improving market access under GPT, this also resulted in effective cut of margins enjoyed by the LDC countries. In 2000 further 570 tariff lines were added to the duty free list for LDCs bringing the share of duty-free lines for LDCs to 90%. Nevertheless, a number of sensitive products were still excluded. Weston (2003), citing UNCTAD (2001) and UNCTAD and Commonwealth Secretariat (2001), indicates that at the beginning of 2000s, despite the apparently high tariff line coverage, the LDC programme of Canada was not granting real market access opportunities for LDCs: only 15 products were exported to Canada by LDCs; only around 30% of non-oil, non-arms imports from LDCs were duty-free in 2000; of the Quad countries, Canada had the highest proportion of imports form LDCs facing tariffs above 5%. Indeed, the extension of product coverage in 2000 led to very limited changes as textiles and clothing products remained excluded. These accounted for 38% of total LDC exports to Canada in 2000 (Kowalski and Lippoldt, 2005). As discussed below, however, an unambiguous increase in the utilization rates of LDCT scheme by some LDC beneficiaries points to beneficial effects of more flexible rules of origin (ROO) (see Rules of origin below). From 1 January 2003 all remaining tariff and quota restrictions on imports from LDCs (with the exception of supply-managed agricultural products 5 and Myanmar) were removed. The initiative included textiles and clothing and a modification of ROOs. Before the extension to textiles and clothing, excluded products were 93% of total dutiable LDC imports (Lippodlt and Kowalski, 2005). In 2003, the share of excluded products was expected to go down to almost zero which, as data indicates, is what actually happened (see Utilisation, utility and coverage of preferences). As argued by Weston (2003) the new LDC initiative was controversial not only because of the potential impact on Canadian producers (mainly in the 4 Excluded products encompass: textiles, footwear, products of the chemical, plastic and allied industries, specialty steels and electron tubes (UNCTAD, 2001). 5 Supply-side managed products referred to here are dairy, poultry and eggs products the supply of which is regulated in Canada through a system of quotas. 5

clothing industry) but also because of the short-term nature of benefits it would afford to LDCs in the context of inevitable phase out of MFA quotas in 2004. It was expected that benefits to LDC exporters that would profit from the new initiative would disappear with expansion of quotas on non-ldc, in particular Chinese, clothing products. Weston (2003) also points out that uncertainty associated with non-binding nature of preferential margins, associated conditionalities and supply constraints in the exporting developing countries have all contributed to the scope of benefits associated with these preferential schemes. Rules of origin Rules of origin (ROO) are employed under preferential as well as non-preferential tariff schemes in order to require a minimum level of local content in products imported from eligible suppliers. They can help to ensure that the products imported under the preferences are not merely transhipped from noneligible countries via the eligible suppliers with little or no local value added. While under certain circumstances the ROO can play a role in ensuring the intended beneficiary countries actually reap the benefits from preferential programmes they are also perceived as the main reason for underutilisation of preferences (e.g. Inama (2003)). In the 1980s Canada introduced more generous rules of origin for LDCs than for other developing countries with a minimum requirement of 40% of local value added compared to 60% required for other GPT countries. The 2000 reform of the system further relaxed ROO allowing up to half of 40% minimum value added originate from another developing countries (Weston, 2003). The 2003 LDC initiative included a modification of ROO for textiles and clothing products. To be eligible under the new ROOs the cloth has to be cut and sewn or fabric woven from yarn produced in the eligible country. The new cumulation system allows inputs from all beneficiary countries provided that a minimum of 25% of value added originates from the exporting LDC country. For goods other than textiles and apparel the existing pre-2003 rules of origin for the LDCT remained in force. Other forms of compliance verification To be eligible for the LDCT or GPT, in addition to ROO requirements, goods must satisfy the requirement of certification and direct shipment. Direct shipment requires in essence that goods are either shipped directly from an eligible country or transhipped through an intermediate country provided that the goods remained under customs transit control in the intermediate country and did not undergo additional processing, trade, consumption or storage exceeding 6 months in the intermediate country (CCRA, 2003; UNCTAD, 2001). The required documentation consists of a thorough bill of landing (TBL) and related shipping documents if a TBL does not specify all points of transhipping (CCRA, 2003). The system distinguishes between two types of certificates of origin, one for non-textile and apparel goods and one for textile and apparel goods, that should be used to claim entitlement to the benefits of duty-free market access under the LDCT. The verification procedure involves an origin questionnaire or letter which is sent to the exporter for completion and return to Canada Customs and Revenue Agency (CCRA). This document is the basis on which an auditor determines whether goods are eligible or not. If additional information is required the auditor may request a visit to the premises of the exporter. The value of preferences A number of approaches to measuring the benefits from preferential arrangements, or alternatively the losses from their erosion, have been employed in the existing literature. At the theoretical level, trade preferences and regional agreements can be analysed in the first instance within the same analytical setting. This is because both kinds of arrangements share discriminatory properties, in that they involve 6

geographically selective trade liberalisation. The static trade and welfare effects of regional agreements and trade preferences have to be examined in a second-best context, in which the final judgement as to any economic benefits that might accrue is an empirical question. Among the most popular methods for measuring the benefits of preferential schemes are simple calculations of the value of benefits based on fixed trade values, estimations of trade creation/trade diversion impacts, or general equilibrium evaluations. In the first of these approaches, the benefit to the preference-receiving country is usually defined as the difference between the MFN rate and the preferential rate multiplied by the value of imports under the given preferential scheme evaluated at world prices. The effect of preference erosion is then calculated as the difference between the value of the preference before and after a multilateral liberalization. An example of such an approach in the literature is the assessment of preference erosion in agricultural products by Yamazaki (1996). A limitation of this methodology is that changes in MFN tariffs are likely to induce changes in the volumes traded under both preferential and nonpreferential schemes resulting in lower benefit to preference-receiving countries after MFN liberalisation. Disregarding these changes to trade volumes may result in underestimation of the impacts of preference erosion. A number of studies improve upon this approach by modelling the demand and supply schedules in partial equilibrium models (see e.g. Subramanian, 2003, and Alexandarki and Lankes, 2004). This methodology differentiates products by country of origin and controls for trade creation/diversion effects in response to changes in trade protection measures. Under this approach, MFN liberalisation typically results in an increased demand for products imported under MFN treatment and decreased demand for imports entering under preferential rates. The advantage of this approach is its relative ease of interpretation and the possibility of its application at the very detailed level of product classification an advantage over the data-intensive general equilibrium assessments that have to be conducted at a high level of aggregation. Its shortcoming is that it cannot relate changes in tariffs and trade on one good to those on other goods - i.e. it is partial equilibrium. Where a far-reaching reform is under consideration such as currently is the case in the DDA, this can be a handicap and result in unrealistic estimates of the economic value of preferential trading arrangements. Computable general equilibrium evaluations of MFN liberalisation capture the effects of substitution between imports and domestic production, imports from preferential to non-preferential sources in the preference-giving country, changes in demand for intermediate inputs, reallocation of productive resources across industrial sectors, terms of trade and balance of payments effects. It can therefore better capture some of the costs inherent in preferential trading arrangements such as, for example, preferences-driven concentration of resources in relatively uncompetitive activities. In such a framework, in addition to the potentially negative impact in a particular preference-receiving sector (e.g. in terms of output reduction), economy-wide implications of reallocation of productive resources towards other activities are evaluated. This important advantage enables accounting for the package nature of multilateral trade agreements where the potential negative effects associated with a particular sector or preferential scheme can be analysed in conjunction with other effects. One drawback of CGE modelling is its high data requirements and the resulting high-level of aggregation of different types of economic activity a feature that may be particularly problematic in a study of preference erosion where the focus is on selected product categories and in a limited group of small countries. No matter what approach one takes, caution in interpreting the results is called for. Most analyses of the liberalisation benefits arising from preference schemes rely on trade flow data, estimates of changes in preferential margins, coverage or exclusion ratios, changes in trade shares, and the like. One of the problems with this kind of analysis is that even where a relationship seems to exist between changes in a preferential arrangement and changes in trade flows or shares, it is often necessary simply to assume a causal link, despite the possibility that a range of other factors may have influenced an observed outcome. 7

Similarly, the absence of a positive trade response in the presence of preferences does not necessarily mean that the preferences are deficient -- the constraint could lie elsewhere. Indeed, as we will see later in the paper, benefits from the Canadian preferential schemes may have been limited on account of relatively low to modest MFN duties, and/or the number of FTAs that Canada has concluded with its major trading partners. Structure and utilization of preferences Data The analysis presented below draws on tariff preference database collected by the OECD Secretariat. The Canadian data covering tariff lines for which there were imports from developing countries in 1998, 2002 and 2003 were provided by Finance Canada. The database consists of bilateral import flows at the HS-8 digit level differentiated by treatment claimed at the time of import. In addition to LDCT and GPT treatments, imports from developing countries have been registered under Commonwealth Caribbean Countries Tariff, British Preferential Tariff, Chile tariff, General Tariff, Mexican Tariff, Mexican-United States tariff, United States Tariff, Canada-Israel Agreement Tariff and Costa Rica Tariff. The database makes a distinction between total and dutiable imports for each specific tariff line and treatment. Dutiable imports refer to the portion of imports that was used for calculating duty paid or payable. 6 An overview of 2002 and 2003 data underlying calculations in this part of the paper is presented in Table 3, including an indication of the number of tariff lines on which at least one positive import flow between Canada and a developing country occurred under the specific tariff treatment. The corresponding minimum, maximum and simple average tariffs are also provided. For the sake of comparison, an average MFN rate calculated on corresponding tariff lines is also presented. The last column indicates the number of lines on which ad valorem duties were specified. The source data on preferential trade of Canada provided by Finance Canada include descriptions of specific duties. Non-ad valorem tariffs on records concerning Canada s trade with developing countries take mostly the form of mixed rates. 7 The share of non ad valorem tariffs on MFN lines on which trade with developing countries was recorded in 2003 was 2.6%. For GPT this figure was even lower (1.4% corresponding to 0.26% of value of GPT imports) and for LDCT it was nil. Additionally, only for 0.40% of value of imports entering Canada under the GPT scheme in 2003 was the corresponding MFN duty rate specified in non-ad valorem terms. For LDCT the corresponding ratio was 0.14%. In view of the relatively low incidence of non ad valorem items and given the non-availability of data on imported quantities or collected tariff revenue, the ad valorem components of mixed and compound rates are used in calculations referring to these lines. This is arguably an ad hoc and crude way of dealing with specific rates but, compared with excluding these items from calculations, it delivers a more realistic 6 The data set contains some problematic entries where for example positive values of dutiable imports are recorded for duty free records. It is important to note that Canada s simplified tariff regime that came into effect on January 1, 1998 was reported to have caused misunderstandings with respect to coding of documents, duty rates, classifications etc. This could explain these anomalies for 1998. However, they do also occur in other years. 7 Mixed rates take the form of a conditional expression determining either an ad valorem or a specific tariff (for example X /kg but not less than Y % or X /kg but not to exceed Y %). Other types of specific duties used in the Canadian Customs Tariff schedule include specific, compound (combining both specific and ad valorem components) and technical rates (duty dependent on the input content). 8

portrait of the protection structure (especially in the limited number of agricultural and food product lines where non ad valorem rates are concentrated). Composition and significance of trade flows The 2003 data indicate that around 74% of total imports from developing countries entered Canada under MFN treatment, 15% via GPT and 0.7% via LDCT treatment (Table 4). Corresponding shares computed on the basis of dutiable imports are respectively 64%, 33% and 0%. The differences are explained by the shares of duty-free imports within each of these treatments. In 2003, around 72% of imports from developing countries entering under MFN treatment were duty free while the corresponding share for the GPT and LDCT were 41% and 100% (Table 7). It is worth pointing out that imports from developing countries entering Canada under United States Tariff originated from Mexico, Brazil, China, Ecuador, Pakistan, Morocco, Thailand and South Korea. In order to shed more light on the importance of the Canadian market for developing countries exports Tables 4 and 5 present exports to Canada under all schemes (including MFN) (Table 4) and all preferential schemes (Table 5) as a share in beneficiary s exports to all trading partners. 8 Taking 2003 as a reference year, the shares concerning all (preferential and non-preferential) exports were significant and reached up to 30% for some developing countries. Nevertheless, exports under preferential schemes typically accounted for less than 1% of total beneficiary exports. In 2003, this ratio was higher than 1 percent for only nine developing countries (Bangladesh, Mexico, Cambodia, Haiti, People s Democratic Republic of Laos, Maldives, Barbados, Trinidad and Tobago, and Lesotho). It is worth noting that for six of these countries significant export increases took place in 2003, most likely, as a result of the introduction of the new LDC initiative. The remarkable reliance of some developing countries on non-preferential access to Canada s market suggests their obvious interest in further liberalisation of the MFN regime. However, for a few countries, mainly LDCs trade under preferential schemes accounts for the bulk of their exports to Canada. This suggests that the latter group of countries would have no major interest (apart, of course, from preventing any negative effects of preference erosion) in lowering MFN tariffs. These contrasting situations point to potential divisions in developing countries positions over the issue of lowering Canadian MFN tariffs. Table 8 presents the 2003 product structure of imports from developing countries entering Canada under LDCT, GPT and MFN treatments. Imports under the LDCT scheme exhibit a heavy concentration in 5 HS-2 digit textile and clothing chapters (61-65). These chapters accounted for 98% of imports entering under the LDCT scheme. Exports under the GPT scheme were more diversified. Still, 6 top HS chapters accounted for more than a half of imports under this treatment. Furniture products (HS2 chapter 94) accounted for 13% of imports under GPT scheme; Electrical machinery and equipment (HS chapter 85)11%; Toys, games and sports requisites (HS2 chapter 95) 7.5%; Plastics and articles thereof (HS2 chapter 39) 7.4%; Articles of leather (HS chapter 42) 6.2%; Nuclear reactors, boilers (HS chapter 84) 5.3%). Structure of imports from developing countries under MFN scheme was also relatively concentrated with more than 50% of imports under MFN treatment were in 4 HS2 chapters: Electrical machinery and equipment (HS chapter 85) 18.2% of imports under MFN treatment; Nuclear reactors, boilers (HS chapter 84) 16.5%; Mineral fuels, oils and products thereof (HS chapter 27) 14.3% and Vehicles other than railway or tramway rolling stock (HS chapter 87) 5.7%. 8 Data on total exports of a given developing country were collected from the WITS database. 9

Preferential tariff rates Substantial shares of imports from developing countries enter Canada via duty-free or low MFN tariff rates. In 2003, 72% of imports from developing countries entering Canada under MFN treatment were free of duty (Table 6) and simple average MFN tariff rate on lines with imports from developing countries was 5.8% (down from 6% in 2002). On a simple average basis, LDCT offered 12 percentage point advantage over MFN rates on LDCT eligible-lines. This was significantly higher than GPT (2 percentage points), CARIBCAN (3.5 percentage points) or various country-specific tariffs (see columns 5 and 6 of Table 3). Notably, the GPT rates were available on many more lines than the other Canadian preferential arrangements. As far as individual tariff lines were concerned, preferential margins reached up to maximum 20 percentage points for LDCT treatment, 18.5 percentage points for GPT and 12.5 for CARIBCAN. Nevertheless, the bulk of preferential trade associated with these schemes occurred on lines with medium and small preferential margins (see Figure 1). Simple averages of preferential margins calculated at the HS chapter level for all and each of the LDCT, GPT and CARIBCAN schemes presented in Table 9 indicate a marked dispersion across products. For the three schemes assessed together, in 2003 highest average margins (up to 13 percentage points) were observed in textile & clothing products (HS chapters 61-64) which was largely an implication of margins enjoyed by LDCT beneficiaries (up to 18 percentage points). GPT margins in these chapters were also consistently higher than the average. However, the four chapters accounted for only 6% of imports entering under GPT, LDCT and CARIBCAN schemes. Figure 1. Count of tariff lines with positive trade flows under LDCT, GPT and CARIBCAN treatments and associated preferential margins 120,000,000 25 value of imports 100,000,000 margin 20 Value of imports, US$ 80,000,000 60,000,000 40,000,000 15 10 associated preferntial margin (% points) 20,000,000 5 0 Spectrum of 19,147 tariff lines with positive trade values 0 10

One of the possible solutions to the preference erosion problem that may be considered in the DDA negotiations is a simultaneous lowering of MFN and corresponding preferential rates so that preferential margins are preserved. Discussion of the linear formula case provided in the Annex shows how this could be done. Certainly, such a procedure can only apply to tariff rates where existing preferential rates are nonzero. Additionally, in order to preserve initial preferential margins, proportional cuts to preferential rates would need to be deeper than those concerning MFN rates and, as is discussed in the annex, for lines where initial difference between MFN and preferential rates is large such strategy is only feasible for relatively shallow cuts of MFN rates. In this context it is worth presenting the structure of Canadian preferential tariffs distinguishing between the relative size of the preferential and corresponding MFN rates. Figure 2 offers such a presentation where all 19,147 records with positive trade values are divided into a subset of lines where the problem of preference erosion associated with MFN liberalisation could in principle be dealt with by simultaneous reductions of preferential rates and lines for which this is not feasible (rates where preferential rates are already zero). In addition a distinction is made between 7,382 lines where both preferential and MFN rates are zero and 4,138 lines where only preferential rates are zero. This distinction is important because the latter set of lines would be affected by declining margins under MFN liberalisation while the former would not. Overall, simultaneous reductions of MFN and preferential rates would in principle be feasible on up to 7,582 lines 9 covering 66% of total imports entering under GPT, LDCT or CARIBCAN schemes. 7,382 lines covering 24% of preferential trade would not be affected by MFN cuts since on these lines both the MFN and preferential rates are already zero. Declines of preferential margins would be inevitable on 4,183 lines covering 20% of preferential trade including all preferential trade with LDCs. 9 The exact number of lines for which this is feasible depends on the ambition of the associated cut to MFN rates. 11

Figure 2. Count of tariff lines with positive trade flows under LDCT, GPT and CARIBCAN treatments and associated ratio of preferential to MFN tariff 120,000,000 120 ratio of tariffs value of imports 100,000,000 100 Value of imports, US$ 80,000,000 60,000,000 40,000,000 4,183 lines with positive MFN and 0 preferential 7,382 lines with 0 MFN and 0 preferential rate 80 60 40 Preferntial tariff / corresponding MFN tariff * (100) 20,000,000 20 0 Spectrum of 19,147 tariff lines with positive trade values 0 Coverage, utilisation and utility of preferences The literature highlights three dimensions of preferential programmes and the associated trade flows: coverage, utilisation and utility. Inama (2003) provides an overview of these dimensions, although there is some variation in the way these concepts are applied by various authors. In this assessment, Product coverage is calculated as the ratio between imports that are covered by a preferential trade arrangement and total imports from beneficiary countries. While this ratio gives an indication of the extent of eligibility for preferences it also reflects the structure of imports from beneficiary countries. Coverage ratios by beneficiary for LDCT, GPT and CARIBCAN schemes are presented in Table 10. With some exceptions, these ratios tend to be high (often close to 100%) and consistent across time for CARIBCAN beneficiaries. Across GPT beneficiaries there is no clear tendency across countries or in time. In fact 54% of GPT beneficiaries record a decrease in coverage ratios between 1998 and 2002. This observation is not easily reconciled with the 1999 extension of the GPT and LDC product coverage by some 220 product lines and lowering of the GPT tariff to two thirds of the corresponding MFN rates (see An overview of Canadian tariff preferences above). A similar drift is observed in LDC coverage ratios with the difference that all (apart from Afghanistan) LDC beneficiaries move to 100% coverage ratio in 2003. Undoubtedly, this is a result of the almost complete removal of all remaining tariff and quota restrictions on imports from LDCs in 2003. The expansion of exports under the LDC scheme, mostly in textile and clothing products which are not covered by the GPT, has also certainly affected the coverage ratios calculated for the GPT scheme. Utilisation rates are computed as ratios between imports actually entering under a preferential scheme and imports covered by the scheme. They are presented for each beneficiary in Table 11. Utilisation rates give an indication of the take up by importers of the offer of preferential access. It is argued that this take up rate depends crucially on the stringency and complexity of rules of origin and ancillary requirements 12

(Inama, 2003). Utilization rates calculated for the GPT indicate a very uneven take up and there is no consistent tendency in time. In fact an increase in utilisation rate from 1998 to 2002 was observed in only 45% of beneficiaries, an increase from 2002 to 2003 in 31% of countries and an overall increase between 1998 and 2003 in 37% of beneficiaries. The structure of LDC utilisation rates seems very dichotomous. For 35 LDC countries utilisation rates have remained consistently at zero in the three considered years. In most remaining LDC beneficiaries a positive tendency of increasing utilisation rates was observed. In particular, the change of regime in 2003 seems to have had a significant impact on utilisation rates of Bangladesh (89% utilisation rate in 2003), Cambodia (89%), People s Democratic Republic of Laos (89%), Madagascar (16%), Malawi (16%), Maldives (47%), Nepal (51%). As is explained in the next section these substantial increases can be almost entirely attributed to increased trade flows in textiles and wearing apparel products. A comment on utilisation of alternative preferential schemes is warranted. Why do some LDC beneficiaries continue to use both GPT and LDC schemes even though the LDC scheme seems more generous? Such a situation arises often if both the GPT and LDC tariff rates are zero. Additionally, access under the GPT scheme may be chosen if the LDC ROOs or other forms of compliance verification constitute more of a hurdle than those applicable under the GPT. As indicated in Overview of Canadian tariff preferences above, the 2003 new LDC initiative included a modification of ROO for textiles and clothing products. To be eligible under the new ROOs the cloth has to be cut and sewn or fabric woven from yarn produced in the eligible country. It is difficult to establish whether the ROO introduced in 2003 have indeed been particularly restrictive. Nevertheless, such a possibility would be suggested by postulates put forward by the Canadian clothing industry which opposed the initiative during the consultation period and urged the government if it went ahead to have strict rules of origin and controls to avoid transhipment (Weston, 2003). The new ROOs for textiles and clothing products are one potential explanation of utilisation of both GPT and LDCT for shipments of the same products. This utility rate is calculated as the ratio of imports actually receiving preference and all imports (covered or not) and gives an indication of the importance of preferences in relation to all trade (with Canada). It is a combined indication of the generosity of a given preferential scheme, simplicity of its use as well as of export capabilities of beneficiaries. The calculated utility rates are presented in Table 12. Similarly to utilisation rates, utility rates of LDCT exhibit a certain dichotomy with 35 LDCs recording consistent zero utilisation rates and other LDCs recording marked increases in 2003. The GPT utility rates are also very dispersed with a significant number of countries displaying marked reliance on preferences in their trade with Canada and others for which preferences seem unimportant. Value of preferences The value of preferential programmes of Canada is estimated by calculating the difference between the MFN rate and the corresponding preferential rate and weighting by the value of imports. Total values of preferences calculated in this way for all preferential schemes and years are presented in Table 13. Remarkably, the value of preferences granted to Mexico under Mexican and Mexican-United States tariff exceeded the combined value of GPT and LDC preferences in 2003. Up until 2002 the value of GPT preferences was close to 300 times the value of LDC preferences. In 2003 the value of LDC preferences multiplied by factor of 100 and accounted for one third of the value of GPT scheme. The value of CARIBCAN preferences, was more than a double of LDC preferences in years 1998 and 2002 and only approximately one twentieth in 2003. As far as the individual beneficiaries are concerned, the largest beneficiaries include several more advanced countries such as Mexico, South Korea, China, India, Brazil, Israel, Malaysia, Hong Kong-China as well as some LDCs such as Bangladesh or Cambodia. Expressed as percentage of total beneficiary s 13

exports the value of preferences did not typically exceed 1%. 10 Nevertheless, several LDCs exhibited higher than average ratios of value of preferences to total exports (0.52% in Bangladesh, 0.43% in Cambodia, 0.28% in Haiti, 0.23% in People s Democratic Republic of Laos, 0.22% in Maldives, 0.20% in Lesotho). Remarkably, the corresponding ratios for 1998 and 2002 were significantly lower. The striking increase in the value of preferences in 2003 observed for many LDC countries does not reflect an error in data or calculations but rather a marked increase in the utilization of preferences by LDCs and preferential margins on textile and clothing products that were up to this date excluded from the duty-free treatment. As argued above, these changes are clearly related to the introduction of the new Canadian LDC initiative. The percentage increases in the value of preferences in 2003 relative to the average for 1998 and 2002 are very significant: 4000% for Bangladesh, 5500% for Cambodia, 8890% for Laos, 4658% for Maldives, 5061% for Madagascar or a stunning 1.5 million % for Lesotho. Focusing on Lesotho s case, this increase can be entirely attributed to changes on 11 HS 8-digit tariff lines all of them within the two 2-digit HS chapters Articles of apparel and clothing accessories, not knitted or crocheted (61) and Articles of apparel and clothing accessories, not knitted or crocheted (62). Prior to the introduction of the 2003 scheme each of these 11 tariff lines faced an import duty of 19 or 18%. In 2003 all these lines gained duty-free and quota-free access and the corresponding value of trade increased by 80%. Value of preferences by tariff line level Which tariff lines carry the highest value of preferences granted under GPT and LDCT schemes? An answer to this question may help decide whether exclusion of certain lines from MFN liberalisation could alleviate the bulk of negative effects of preference erosion. Additionally, and perhaps more importantly, identification of such lines will be a necessary step in the assessment of costs and benefits of such potential exclusion or in designing any prospective compensation schemes. Tables 17 and 18 present the importance of individual tariff lines in the value of LDC and GPT preferences. As far as the LDCT scheme is concerned, there were altogether 390 8-digit tariff lines for which positive trade values were recorded in 2003. However, for some lines these values were very small. A criterion of at least 1 US$ million in terms of revenue forgone at the tariff line level was used to identify the tariff lines where the value of LDC preferences was greatest (in Table 17). There were 13 such lines: 6 in Chapter 61 (Art of apparel & clothing access, knitted or crocheted) of HS classification 6 in Chapter 62 of HS classification (Art of apparel & clothing access, not-knitted or crocheted) and 1 in Chapter 63 (Other made-up textile articles). These tariff lines accounted for 64% of total value of LDCT preferences in 2003. At the same time the associated values of preferences for each or these individual lines did not in any case exceed US$ 8.6 million and accrued consistently to a small group of beneficiaries such as Bangladesh, Cambodia, Laos, Lesotho, Madagascar, Malawi, Maldives, Nepal, Niger and Haiti (see Table 17). As far as the GPT scheme is concerned, there were 2000 8-digit tariff lines for which positive trade values were recorded in 2003. Table 18 adopts the same criterion 11 that was used for the assessment of the LDC scheme to identify the tariff lines carrying the greatest value of GPT preferences. There were 32 such lines: 7 in HS chapter 94 (Furniture; bedding, mattress, matt), 4 in each HS chapter 39 (Plastics and articles thereof.) and 42 (Articles of leather; saddlery and harnesses, travel goods, handbags), among others. These tariff lines accounted for 38% of the total value of GPT preferences in 2003. The noticeable difference is that for each of the identified tariff lines the number of beneficiaries was typically larger than in the LDC case. However, the highest value of preferences calculated at the tariff line level was lower than in the LDC case (around US$ 5 million) and typically the value of GPT preferences accounted for 10 An exception are Turks and Caicos Islands - an overseas territory of the UK. 11 Minimum of 1 US$ million in terms of revenue forgone at the tariff line level. 14

smaller shares of beneficiaries exports. A striking feature of the value of GPT preferences that transpires from Table 18 is that for the majority of the identified tariff lines most of the value of preferences was associated with imports from China (see shares in fourth column of Table 18). Overall, China accounted for 64% of value of Canadian GPT preferences in 2003. High tariff line concentration of value of preferences in the case of LDCT and, to a lesser extent, GPT schemes suggests that exclusion of relatively few lines from MFN liberalisation could alleviate the bulk of negative effects of preference erosion in the Canadian market. However, the benefits of such an option would have to be weighed against the costs of unrealized liberalisation. For example, the value of LDCT preferences on HS-8 digit line 62046200 (US$ 8.6 million see Table 17 ) should be compared with benefits (to Canada and its trading partners) that could be obtained by reducing protection on the US$ 257 million worth of Canadian non-ldc imports on this line. Economic value of preferences in 2001 - a general equilibrium assessment This section reports on the results of a model assessment of value of Canadian preferences. The analysis employs the standard GTAP model 12 and the version 6 of the GTAP database. The model approach allows a fuller economic assessment of the value of preferences which depends on consideration of the interaction of impacts across preference receiving and other sectors within a given beneficiary country as well as an assessment of impacts across trading partners. Changes in market access conditions for one product category are linked to developments in other sectors through goods and factors markets. Representation of such inter-sectoral linkages permits accounting for the reality that while some producers in selected preference-receiving sectors may be affected negatively, the resources that are freed from that sector can be employed in other sectors that may gain better access to world markets or be simply more productive. One disadvantage of this approach is its high level of product and country aggregation that may mask some individual effects. This shortcoming is nevertheless partially offset by insights gained from the more detailed analysis of Canadian preferences presented in the first part of this paper. Additionally, in contrast to earlier iterations of the GTAP database, version 6 of GTAP database used here takes into account an exhaustive range of tariff preferences and therefore offers an opportunity to verify some of the general hypotheses that arise from the analysis of Canadian preferences structure. The scenario considered here involves an equalisation of product-level bilateral ad valorem measures of protection (see next Section for an explanation) with the pre-shock average calculated across all trading partners (a proxy for the MFN rate). Admittedly, this is not a very realistic scenario in that it is unrelated to any multilateral tariff liberalisation that may be agreed in the DDA negotiations. Instead, this scenario mimics a situation in which, starting from the 2001 base, all preferential access to the Canadian market is removed. This allows us to gauge the value of Canadian preferences. Preferential access to Canadian market in the GTAP framework The dataset used for the simulations comprises data with a base year of 2001 and covering 57 broad economic sectors and 87 countries. For our purposes, the database has been aggregated to 44 countries/regions and 22 sectors. The choice of country aggregation was aimed at maintaining the maximum number of individual developing as well as key industrialized countries (including a separate treatment of all major preference granting countries). The choice of sectoral aggregation has been made so as to permit a comparison of sectors according to differing protection patterns. 12 The most comprehensive description of the model is in Hertel (1997). Information more recent developments can be accessed at www.gtap.org 15

The distinct advantage of protection data in the employed database is that it fully integrates the information on bilateral ad valorem tariffs (both MFN and preferential), ad valorem equivalents of specific tariffs (MFN and preferential), as well as tariff rate quotas from CEPII/ITC Market Access Maps (MAcMaps) database. 13 The resulting ad-valorem equivalent measure of applied protection is thus an all inclusive measure of protection that is consistent across all bilateral trade flows. In the employed model and database, the tariff line information on protection measures is aggregated to broad product categories by means of trade weighting. Hence, necessarily, the analysis presented here masks the detail present in the underlying tariff-line data and is incapable of shedding light on impacts on very specific product categories. Due to the level of data aggregation, each bilateral ad valorem equivalent measure of applied protection included in the database by necessity combines the information on MFN and preferential market access as well as the actual composition of trade within the product category. Thus, bilateral protection rates for a given product category vary from one trading partner to another. To illustrate the structure of protection and preferential access to the Canadian market for each product category and trading partner, differences between trade-weighted averages of ad valorem rates of protection across all trading partners and the rate applied to imports from a specific trading partner are calculated. These equivalent measures of applied protection are then used in the simulation described in the following sub-sections. It is worth mentioning that the difference between the average trade-weighted ad-valorem equivalent and bilateral rates may provide a better indication of preferential market access as compared to the difference between the official MFN and bilateral rates. As discussed in the part reviewing the detailed structure of Canadian preferential programmes, MFN rates typically apply only to a share of the actual trade flow making the trade-weighted average of ad valorem equivalent measures a better indication of the average trade restrictiveness. 14 Table 19 presents the differences (measured as percentage points) between average and bilateral rates in the Canadian market. The positive and negative values shown in the table reflect respectively lower-than-average and higher-than-average restrictiveness of access to a given destination market for each source country and product category. The presented margins provide an indication of the extent of preferential treatment enjoyed in a particular product by beneficiary countries. 15 They take into account the preferential situation of a given exporter relative to all the exporters to the Canadian market, including other developing countries and not simply those countries exporting under MFN rates. For example, Canada offers duty-free access to its market for Paddy rice margins are nil across all source countries indicating that none of the partners receives a preferential treatment (Table 19). In contrast, in Dairy products, where protection is particularly high, a number of Canada s trading partners benefit from preferential margins reaching up to 50 percentage points (e.g. Singapore and Uruguay). At the same time, several other countries (including a number of developing countries) face ad valorem rates that are several percentage points higher than the average. 13 The dataset is documented in detail in Bouët, A., Fontagné, L., Mimouni, M., and F. Von Kirchbach (2002), Market Access for GTAP: A Bilateral Measure of Merchandise Trade Protection by GTAP Resource #1045, available at the following web address (confirmed on 8 January 2005): http://www.gtap.agecon.purdue.edu/resources/res_display.asp?recordid=1045. 14 It has to be pointed out that this approach has its weaknesses. By calculating the average rate for each product category we run a risk of imposing a lower counterfactual tariff rate (since certain trading partners pay tariff rates higher than the average). An alternative would be to use, for a given product, the maximum tariff rate paid by a trading partner as the proxy for the counterfactual MFN rate. The latter approach would result in a higher counterfactual rate. However, it would also mean a change in the pattern of Canadian tariff protection across broad product categories an effect that is absent when the average rate for each product is used. 15 As indicated above, because of the level of aggregation, these margins may also reflect different structures of trade within aggregated product categories. 16