Quantifying the Trade Effects of Technical Barriers to Trade: Evidence from China 1. IntroductionTechnical barriers to trade(TBT)are now widespread and have increasing impacts on international trade. The spread of TBT may have some special reasons.First, it’s legitimate. The WTO members are authorized by WTO TBT/SPS Agreement to take such measures in order to protect human health, as well as animal and plant health, provided that the enforced measures are not disguised protectionism. Second, as Baldwin (1970) emphasized, ―The lowering of tariffs has, in effect, been like draining a swamp. The lower water level has revealed all the snags and stumps of non-tariff barriers that still have to be cleared away‖. Wallner (1998) considered this phenomenon as a ―law of constant protection‖, referring to perfect substitutability between tariff and none-tariff barriers in maintaining a degree of desired domestic protection.Third, with the trade liberalization process, the remaining barriers, like TBT have a more important but not a less important impact due to the ―globalization magnification effect‖, seemingly minor differences in technical norms can have an outsized effect on production and trade (Baldwin 2000). Fourth, the increasing income of importing country and consumer preference may result in a higher demand for product quality, safety and environment protection.Since the proliferation of TBT and its increasing trade-restrictive impacts, OECD (2001) drew attention to TBT and suggested more empirical research on it, becausethe quantitative analysis is an important step in the regulatory reform process and can help inform governments to define more efficient regulations. However, due to the theoretical complexity and data s carcity, TBT have been considered as ―one of the most difficult NTBs imaginable to quantify‖ (Deardorff and Stern 1997)So far, there is not a preferred quantification strategy and claims abound on both sides about―whether such restrictions tend to reduce trade by virtue of raising compliance costsor expand trade by increasing consumer confidence in the safety and quality of imported goods‖ (Maskus and Wilson 2001).Maskus and Wilson (2001), Maskus, Otsuki, and Wilson (2001), Beghin and Bureau (2001), Ferrantino (2006) and Korinek, Melatos and Rau (2008) etc provide comprehensive overviews of key economic issues relating to TBT modeling and measurement. Based on these literatures, quantification techniques can be broadly grouped into two categories. Ex-post approaches such as gravity-based econometric models tend to estimate the observed trade impact of standards. On the other hand, ex ante methods such as simulations involving the calculation of tariff equivalents are usually employed to predict the unobserved welfare impact. No approach is or can bedefinitive. Each methodology offers its own pluses and minuses, depending on a number of factors, including the nature of the technical measure, the availability of data, and the goal of measurement. (Popper et al 2004)Concerning the trade effect1, different from any other trade measures, TBT have both trade promotion and trade restriction effects. Although a unified methodology does not exist, the gravity model is most often used for the evaluation. The gravity model employs a number of different approaches to measure the TBT. The policy indices obtained by survey can be used as proxy for the severity of TBT, and direct measures based on inventory approach are incorporated too. Beghin and Bureau(2001)summarized three sources of information that can be used to assess the importance of domestic regulations as trade barriers: (i) data on regulations, such as the number of regulations, which can be used to construct various statistical indicators,or proxy variables, such as the number of pages of national regulations; (ii) data on frequency of detentions, including the number of restrictions; frequency ratios and the import coverage ratio (iii) data on complaints from the industry against discriminatory regulatory practices and notifications to international bodies about such practices. Besides the above mentioned approach, some studies try to use explicit standards requirements such as maximum residue levels too.There are a considerable number of study combined the variable for the stringency of TBT with gravity model to estimate the direction of the trade impact.Swann, Temple, and Shurmer (1996) used counts of voluntary national and international standards recognized by the UK and Germany as indicators of standard over the period1985–1991, their findings suggest that share standards positively impact exports, but had a little impact on imports; unilateral standards positively influence imports but negatively influence exports. Moenius (2004, 2006) examines the trade effect of country specific standards and bilaterally shared standards over the period 1985-1995. Both papers used the counts of binding standards in a given industry as a measure of stringency of standards.Moenius (2004) focus on 12 OECD countries and found that at aggregate level, bilaterally shared standards and country-specific standards implemented by the importing or exporting country are both trade-promoting on average. At the industry level, the only variation is that importer-specific standards have the expected negative trade effect in nonmanufacturing sectors such as agriculture. In manufacturing industries, importer-specific standards are trade promoting too. Moenius (2006) confirm the result of Moenius (2004) in that bilateral standard in EU has very strong trade promoting effect as to the trade between EU and non-EU members, but harmonization decrease the internal trade of EU. Moenius (2006) distinguish 8 EU members and 6 non-EU developed countries. So he also found that importer specific standard in EU promote trade between EU members, but depress trade between EU members and non-EU members; Exporter specific standard inside EU has little trade promoting effect ,but export specific standard of non-EU members expand their tradewith EU.The paper using frequency or coverage ratio within a gravity model framework include Fontagné, Mimouni and Pasteels (2005) and Disdier, Fontagné, and Mimouni (2007). Both of them use the frequency ratio based on notification directly extracted from the TRAINS database. Fontagné, Mimouni and Pasteels (2005) collect data on 61 product groups, including agri-food products in 2001. Their paper generalized the findings of Moenius (2004): NTMs, including standards, have a negative impact on agri-food trade but an insignificant or even positive impact on the majority of manufactured products. Moreover, they distinguish trade effects among ―suspicious products‖, ―sensitive products‖ as well as ―remaining products‖ according to the number of notifications and distinguish different country group. Based on data covering 61 exporting countries and 114 importing countries, they find that over the entire product range, LDCs, DCs and OECD countries seem to be equally affected. However, OECD agrifood exporters tend to benefit from NTMs, at the expense of exporters from DCs and LDCs. The authors account for tariff and other NTM in the model , so they also find that tariffs matter more than NTMs, particularly foragri-food products on which comparatively high tariffs are levied.Disdier, Fontagné, and Mimouni (2008) estimate the trade effect of standards and other NTMs on 690 agri-food products (HS 6-digit level). Their data covers bilateral trade between importing OECD countries and 114 exporting countries (OECD and others) in 2004. As well as a frequency index, they use a dummy variable that records whether the importing country has notified at least one NTM and ad-valorem tariff equivalent measures of NTMs as two alternative approaches to measure NTMs. They find that these measures have on the whole a negative impact on OECD imports and affect trade more than other trade policy measures such as tariffs. The tariff equivalent shows the smallest effect. When they consider different groups of exporting countries, they show that OECD exporters are not significantly affected by SPS and TBTs in their exports to other OECD countries while developing and least developed countries’ exports are negatively and significantly affected. For the subsample of EU imports, NTMs no longer influence OECD exports positively, but exports from LDCs and DCs seem to be more negatively influenced by tariffs and SPS & TBTs than that of OECD. Finally, their sectoral analysis suggests an equal distribution of negative and positive impacts of NTBs on agricultural trade.Many studies are supportive of using maximum residue levels to directly measure the severity of food safety standards within a gravity model. These studies include Otsuki, Wilson and Sewadeh (2001a, b), Wilson and Otsuki (2004b,c) Wilson,Otsuki and Majumdsar (2003), Lacovone (2003) and Metha and Nambiar(2005). These studies tend to focus on specific cases of standards for particular products and countries. Otsuki, Wilson and Sewadeh (2001a,b) and Wilson and Otsuki (2004b) examine the trade effect of aflatoxin standards in groundnuts and other agriculturalproducts (vegetables,fruits and cereals). The first two papers covered African export data to EU members and the third paper covered 31 exporting countries (21 developing countries) and 15 importing countries(4 developing countries). All three studies show that imports are greater when the importing country imposes less stringent aflatoxin standards on foreign products. Lacovone (2003) also used MRL of aflatoxin and found that there were substantial export losses to Latin-America from the tightening of the aflatoxin standards set by Europe. Similarly, Wilson, Otsuki and Majumdsar (2003) analyze the effect of standards for tetracycline residues on beef trade and find that regardless of the exporter standards, the standards of tetracycline imposed by the importing countries have the same negative trade impact. Wilson and Otsuki (2004c) analyze MRL relating to chlorpyrifos and Metha and Nambiar(2005)analyze the impact of MRL on India’s export of four processing agri-products to 7 developed countries and yield the similar result.Since our paper focus on the trade effect of technical barrier, we will use the most suitable ex post quantification methods. Moreover, while frequency and coverage ratio can give some guidance as to the potential trade impact of a technical measure, econometric model is used to estimate its magnitude.Our paper make contributions to the current literature in the following ways: First, in contrast to the existing empirical studies which exclusively focus on developed countries TBT, this paper focuses on a developing country, China. Second, this paper has a self-constructed trade measure database based on disaggregated data covered all HS2 products, including agricultural and processing food products (HS01-24) and manufacturing products (HS25-97) so that it can identify the sectors/products with predominant negative impacts on trade. Third, tariff data, import licenses and quotas are included as additional explanatory variables, allowing the distinction between the impact of traditional trade barriers and TBT on trade. Fourth, our data covers 43 exporting countries (including 25 developing countries), it helps to distinguish the trade effect of different country groups. Fifth, in contrast to most literature relied on cross-section data1, our paper covers 9 years time series data on TBT, so we can both capture variation across products and variation within products over time, in particular the changing effects before and after China’s entry into the WTO.The rest of the paper is organized as follows. In section 2, we construct a TBT database from 1998 to 2006 and use inventory approach (frequency index and coverage ratio) to quantify the stringency of technical measures in China. In section 3, we present our regression model, discuss all the variables and describe the data. In section 4, we discuss our findings. We make some concluding remarks in section 5. 2. Quantification of TBT2.1 Measurement of NTM: Inventory approachThe inventory approach allows estimates of the extent of trade covered by NTMs or their frequency of application in specific sectors or against individual countries orgroups of countries. Bora etc (2002) reviews various approaches to quantify NTMs and give a detailed instruction on how to construct frequency index and coverage ratio as follows. The percentage of trade subject to NTMs for an exporting country j at a desired level of product aggregation is given by the trade coverage ratio:2.2 China’s NTM database: data description and methodologyFollowed the method described above, we will construct a Chinese NTMs database from 1998 to 2006 by using inventory approach. The data covered 96 HS2 digit level agricultural and manufacturing industries. First, we calculate a series of frequency index at 4-digit-level of the Harmonized System and then aggregate them into import coverage ratio at HS2. In this database, data are collected by tariff item on the application of a range of tariff and NTMs (TBT, license and import quota) against Chinese imports. The main source of the information on the trade control measures in the database is from Chinese government publications. ―Administrative Measures Regarding Impor t & Export Trade of the People's Republic of China‖ published bythe Ministry of Commerce and Custom General Administration of China provide detailed information at HS 8-digit-level on tariff and non-tariff measures.The code list of supporting documents subject to customs control provide detailed name of licenses or instruments of ratification, which helps to identify whether a tariff line product subject to a specific non-tariff barrier. Concerning the technical measures, it includes those government administrative measures for environmental protection, safety, national security and consumer interests. The code subject to TBT control remains almost the same during the 1998-2001. Specificly, the code subject to TBT in 1998 is IRFM, denoting for Import commodity inspection (I), Quarantine control release for animal, plant and thereof product (R), Import food inspection certificate (F) and Medicine inspection certificate (M). The code concerning TBT in 1999-2001 is AMPR, denoting for Import inspection and quarantine (A), Import commodity inspection (M), Import animal, plant and thereof product inspection (P) and Import food hygiene supervision inspection (R). Since 2002, the government revised the code list into details. Although there is some tiny difference between years, the new code list remains quite stable during 2002-2006 (See the code list in Annex1). The code subject to TBT is ACFIPQSWX during 2002-2005 and AFIPQSWX in 2006, each code stands for Certificate of inspection for goods inward (A), Certificate of inspection for goods inward: Civil commodity import inspection (C), Import licencing certificate for endangered species (F), Import or export permit for psychotropic drugs (I), Import permit for waste and scraps (P), Report of inspection of soundness on import medicines (Q), Import or export registration certificate for pesticides (S), Import or export permit for narcolic drugs (W), Environment control release noticefor poisonous chemicals (X).Note that our data on trade control measures do not have a bilateral dimension. TBT measures, import license and import quotas are enforced unilaterally by Chinese government and applicable to all exporting countries. When we calculate coverageratio and frequency ratio, Vi is the total value of imports in product i from the whole world and Mi indicates whether there are imports from the whole world of good i. Hence, in a specific year, NTM variables vary among different sectors but remain the same among different countries. Although we miss the bilateral dimension associated with such measures, still the exporters are differently affected by TBT measures depending on the structure of their exports in terms of products and markets.To be precise, the frequency ratio of TBT (FR-TBT) measures the proportion of product items covered by TBT measures within a product category, which varies between 0% (no coverage) and 100% (all products covered). We first count the number of HS items (defined at the 8 digit level of the HS) covered by the TBT measures and divide it by the maximum number of product items belonging to the product category (defined here at the 4-digit level of the HS). So we get the results of frequency ratio of TBT at HS4 digit level. For example, regarding HS2402 (Cigars, cheroots, cigarillos and cigarettes, of tobacco or of tobacco substitutes), there are 3 product items with codes 24021000 (Cigars, cheroots and cigarillos, containing tobacco), 24022000 (Cigarettes containing tobacco), 24029000(other). Only one of them (HS24022000) is covered by TBT measures, so the corresponding TBT frequency index equals 33.33% (1 / 3). Then we do the same at HS2 digit level.The import coverage ratio(IC-TBT) measures the proportion of affected import of the total import within a product category. Take HS17 (Sugars and sugar confectionery) as an example, there are 4 product items with code HS1701, 1702, 1703 and 1704 respectively. Only three of them (except HS1703) are covered by TBT measures (it means the frequency index for HS1703 equals 0, while the other three are between 0 and 100%), the import value of the TBT affected products sum up to 111.216 million US$, the import valued of HS17 is 182.244 million US$, so the corresponding TBT import coverage ratio equals 66.46% (111.216/182.244).2.3 TBT rocked sectors in ChinaBy calculating frequency index and import coverage ratio of TBT, we can examine which products are the most affected. According to the definition by UNCTAD (1997), those with a frequency ratio and coverage ratio both above 50% are TBT rocked product. In our sample, 34 products(HS01-24; HS30,31,33; HS 41;HS 44-47; HS51 and HS72)are TBT-rocked during the period from 1998-2002. In 2003, two product items (HS 42-43) become TBT-rocked. In 2004, two more products (HS 50 and HS80) added into the category. During 2005-2006, HS78 are included asTBT-rocked products but HS50 is excluded. See Annex2 for the detailed product information of TBT rocked products.There are a significant number of products, particularly agricultural products and processing food widely affected by technical measures (HS01-24). However, enforcement of TBT is not limited to those products, but is spreading to manufacturing products also. The TBT rocked manufacturing products includePharmaceutical products(HS30, Essential oils, perfumes, cosmetics, toiletries (HS33), Raw hides and skins, leather, furskins and articles thereof (HS41-43), Wood and articles of wood(HS44-46), Base metals and articles thereof, like iron and steel, aluminium and tin.( HS72, 76 and 80) etc. They are either labor intensive products or final goods concerning consumer safety, like medicaments in particular. Although TBT rocked sectors cover about 1/3 of total number of products at HS2 digit level, the proportion of affected trade is limited: about 10-16% of total import. However technical barriers are the most frequent type of NTM, the import subject to TBT account for above 90% of Chinese total import except for the rare case in 2001 (77.29%). (see Table 1).3. Model, methodology and data3.1 Model specificationWe use gravity model to examine how TBT imposed by China influence the country’s bilateral trade. To capture the size effect, population of both countries is used as proxy for exporting country’s supply capacities and importing country’s demand capacity. Per capita income of the two countries is included because higher income countries trade more in general. Transport costs are measured using the bilateral distance countries’ cultural proximity. We therefore control for this p roximity by introducing a common language dummy variable. Based on the typical gravity model, we introduce our key variables—tariff and non-tariff trade barriers. Our basic regression model takes the following forms:4. Empirical results4.1 The whole sample resultsTable 2-1 shows the summary statistics of our key variables. Table 2-2 reports the Pearson coefficients of the trade control measure variables. For the frequency index, import license and tariff appear to be negatively correlated. For the coverage ratio, besides import license, TBT seems to be slightly negative correlated with the tariff. Except for the above rare cases, the import control policies are positively correlated to each other. In general, different kinds of import control measures in China seem to be complementary to each other. Among them, import license and import quota have the highest positive coefficient, this accords with the fact that these two measures are sometimes combined together. Normally a country will distribute quota by issuing import license.We use OLS to estimate the gravity model. Regressions are run on pooled data for 9 years (see Table 3 and 4) and on data for each year separately (see Table 5 and 6). Table 3 and 5 report the result using frequency index, while Table 4 and 6 report the result using coverage ratio, both at HS 2-digit-level. For the whole sample regression results in Table 3 and 4, column 1 shows the result of the basic gravity model, column 2 introduces tariff and non-tariff barriers, column 3 tries to identify the difference between developing and developed countries and column 4 adds WTO as an additional control variable. Year-country-product fixed effect is used for all thespecifications.The results for standard gravity explanatory variables are consistent with prior expectations except for Contig as a rare case. The effect of GDPPC, POP and dist is positive and highly significant for all regressions. It implies that a 1 percent increase in the population of exporting country yields a 1.39-1.47 percent increase in the bilateral trade, and a 1 percent increase in the per capita GDP of exporting country yields a 0.91-1.40 percent increase in the bilateral trade. A 1 percent increase in geographic distance between the two trade partners will result a 1.42-1.45 percent decrease in bilateral trade. The effect of POPchina and GDPPCchina is positive and significant in two regressions. If Chinese population or per capita GDP increase 1 percent, Chinese import will increase 10.8-14.1 percent or 2.0-2.8 percent respectively. The coefficient for Comlang is positively significant in all specifications, which implies that if the exporting country share a same language with China, Chinese import will be stimulated by 2.6-3.3 percent. If the exporting market belongs to China, it will increase Chinese import by 0.3 percent. The coefficient for Contig is significantly negative, which implies that if the exporting country and China are contiguous, Chinese import will decrease 0.76-0.99 percent. This result is not consistent to the prior expectation. But the intuition is easily understood because the most important importing markets such as the US, Japan, EU members are not contiguous with China mainland.We then discuss the key explanatory variable, Tariff have a significant negative effect on Chinese import. A 1 percent increase in the MFN tariff will decrease import value by 0.64-0.66 percent. The results of the frequency index of NTM are all significant. A 1 unit increase in FRTBT will decrease import value by 1.1%, a 1 unit increase in FRQ will decrease import value by 1.7%, a 1 unit increase in FRL will increase import value by 4.1%. The results of the coverage ration of NTM are different in some extent with that of frequency index. A 1 unit increase in ICTBT will increase import value by 0.2%, a 1 unit increase in ICL will increase the import value by 2.7%, and the coefficient for ICQ is negative but not statistically significant.Table 5 and 6 give us a clear picture about how the effect of trade control measures change yearly. Tariff remains negatively significant for all 9 years.Moreover, the elasticity for Tariff dramatically increased since 2003. The trade depressing effect of Tariff nearly doubled after China’s entry into the WTO. FRTBT is negatively significant in all year specifications, and the coefficient remains stable through the sample period. FRL is positively significant while FRQ is negatively significant except for three years. In 1998, 1999 and 2001, FRL is insignificant while FRQ is positively significant due to the multicollinearity1. The result of ICTBT is changeable during the sample period. The coefficient of ICTBT is positively significant during 1998-2002, negatively significant in 2003 and insignificant during the remaining years. ICL remains positively significant during all 9 years, plus the elasticity for ICL slightly increased since 2002. ICQ is significant during 1998-2002,but the sign of the coefficient is changeable, and ICQ becomes insignificant since 2003. So ICQ doesn’t affect bi lateral trade value in a systematic way. From the yearly result, we observe that some of the trade control measure change trade patterns in a different way. Does the trade effect change significantly before and after China entry into the WTO? Whether there is any systematic difference in the trade effect between developing and developed countries? To solve these two problems, we add two interactions. Column 5 introduces the interaction between Developing and each of frequency indices of trade measures. Column 6 adds the interaction between WTO and each of coverage ratios of trade measure. As we can see, tariff (Tariff) and import quota (FRQ and ICQ) seem to have no difference between difference country groups. The change of FRTBT will affect Chinese import from developing countries less than that from developed countries. The change of ICTBT will affect Chinese import from developing countries more than that from developed countries. The change in FRL or ICL will have less impact on Chinese import from developing countries than from developed countries. On average, tariff, license have an increasing effect but quota has a decreasing effect after China’s entry into the WTO. The effect of TBT does not change significantly.5. ConclusionThe results of current literature suggest that TBT in importing country has restrictive trade effect and exports of poor countries are affected more. The paper explores whether technical measures imposed by China have restrictive effects for the imports from main exporters all over the world. Our research confirms some of the results reported elsewhere in the literature while differences remain in some aspects.First, in general trade control measures do have import restrictive effect in China. Second, tariff plays an important r ole even after China entry into the WTO. So far it’s still the most efficient policy tool. Third, TBT is the most frequently used NTM in China and cover almost all the imports. TBT do have some trade depressing effect but the effect is relatively small compared to the effect of tariff. Fourth, in contrast to the general belief that TBT works as a substitute to tariff and traditional NTM in developed countries(Thonsbury1998, Abbott 1997 etc), there is no obvious substitution effect between tariff and TBT in China, moreover, the TBT is complementary to tariff in some extent.定量商业作用技术贸易壁垒:证据从中国1. 介绍技术贸易壁垒TBT现在普遍并且增加了对国际贸易的冲击。