Running Head: AFTA Effects Disaggregated

The Effects of AFTA: A Disaggregated Analysis David Cheong† Johns Hopkins University, SAIS Bologna Center

This Version: August 2008

†Johns Hopkins University, SAIS Bologna Center, Via Belmeloro 11, 40125 Bologna, Italy. Tel: 051-2917824, Email: [email protected]

Abstract

In this paper, using a gravity model, I study changes in trade patterns of ASEAN (Association of South-East Asian Nations) at the Harmonized System (HS) six-digit level in the period 2001 to 2003. These three years saw the acceleration of trade liberalization among ASEAN countries as they moved towards the ASEAN Free-Trade Area (AFTA). To avoid inconsistent estimates due to a log-linear specification of the gravity model, I use the Fixed Effects Poisson Quasi-Maximum Likelihood estimator. The estimates show that: in general, ASEAN preferential margins had a trade-creating effect at the product level; the majority of ASEAN countries benefited significantly from this trade creation; and there was net trade creation in the HS4 and HS8 product categories. These results suggest that ASEAN trade liberalization in the early 2000’s had positive welfare effects.

JEL Codes: C23, F15

Keywords AFTA, ASEAN, free trade area, gravity model, panel data

Introduction The number of preferential trading arrangements (PTAs) has increased fourfold since the early 1990’s (World Bank, 2005). This proliferation has occurred in parallel to, and perhaps out of frustration with, attempts at multilateral trade liberalization within the World Trade Organization. The fact that there are fewer participants in a PTA facilitates reciprocal reduction in tariffs and non-tariff barriers and provides a framework for coordination of trade-related activities such as customs operations and transportation networks within the PTA. By creating a more integrated trading area, a PTA could have other positive effects like higher foreign direct investment. As a commitment mechanism, a PTA could also improve the ability of member countries to reform their domestic policies and enhance the overall economic climate. As shown by Viner (1950), the discriminatory nature of a PTA creates one major pitfall. The margins of preference that PTA members enjoy could induce trade diversion, when a PTA member substitutes its imports from non-member countries with imports from within the PTA. Trade diversion creates a negative welfare effect because imports are not sourced from the most efficient supplier-country. In contrast, a PTA could also induce trade creation, when a member country substitutes its domestic production with imports from member countries. This creates a positive welfare effect because PTA members alter their production structures to fit better with their respective comparative advantages. In this paper, I use a gravity model to assess the impact of trade liberalization according to the ASEAN Free-Trade Agreement (AFTA). The main research question is whether preferential tariffs have created or diverted trade. Most empirical studies of the effects of preferential trading arrangements are conducted with aggregate data. The innovation in this paper is to conduct the analysis at the product level. By doing so, I am able to avoid the aggregation bias in estimating gravity models as noted by Anderson and Wincoop (2004). The rationale for studying the period 2001-2003 is that ASEAN countries, in response to the shocks of the Asian Crisis, decided in 1998 to accelerate tariff reductions towards AFTA, thus creating a natural experiment for studying the trade effects of AFTA. In the next section, I describe the evolution of AFTA and the trade liberalization provisions within this ASEAN agreement. In the third section, I discuss the phenomenon of production sharing in the ASEAN region and its impact on intra-regional trade. In the fourth section, I describe the empirical model used for my assessment of AFTA and discuss the results. In the last section, I provide some conclusions and policy implications. AFTA and Intra-Regional Trade Liberalization History At the fourth ASEAN summit in Singapore in 1992, the original six ASEAN members (i.e., Brunei, Indonesia, Malaysia, Philippines, Singapore, and Thailand) launched the regional trade liberalization project known as AFTA. The member states’ primary aim was to accelerate trade liberalization and improve upon the provisions in the old

1

agreement on ASEAN Preferential Trading Arrangements signed in 1977. The belief was that a free-trade area would benefit member states by increasing trade, investment, production opportunities, and foreign exchange earnings. At the heart of the AFTA declaration was the Agreement on the Common Effective Preferential Tariff (CEPT) Scheme. Article 4 of the CEPT agreement stated that member states would progressively reduce tariff rates on manufactured goods to 0-5% by 2008.i Fifteen product groups (i.e., vegetable oil, cement, chemicals, pharmaceuticals, fertilizers, plastics, rubber products, leather products, textiles, ceramic products, glass products, gems and jewelry, copper cathodes, electronics, and wooden and rattan furniture) were chosen for the fast-track program, which would reduce tariffs to 5% or less by 2003. Other products included in the CEPT scheme were put on the normal track, whereby tariffs would fall to 5% or less by 2008. To qualify for preferential tariffs, at least 40% of the value-added of a product on an importing ASEAN member’s inclusion list had to be of ASEAN origin. Once a product was included in the CEPT scheme, quantitative restrictions had to be eliminated immediately while other non-tariff barriers had to be removed within five years. ASEAN members could exclude products from the CEPT in three cases: 1) Temporary exclusions, 2) Sensitive agricultural products, and 3) General exceptions. Temporary exclusions referred to products for which tariffs would ultimately be lowered to 0-5%, but which were being protected temporarily by a delay in tariff reductions.ii Sensitive agricultural products included agricultural raw materials or unprocessed products covered under Chapters 01-24 of the Harmonized System (HS), and similar agricultural raw materials or unprocessed products in other related HS categories. General exceptions referred to products which a country deemed necessary for the protection of national security, public morals, the protection of human, animal or plant life and health, and protection of articles of artistic, historic, or archeological value. The original scheme was revised in 1994 at the twenty sixth ASEAN Economic Ministers meeting. The tariff-reduction schedule was accelerated so tariff rates would be 0-5% by 2000 for fast-track products and by 2003 for normal-track products. At the same meeting, the ASEAN Economic Ministers also agreed to implement the AFTA Council’s recommendations to bring some agricultural products into the CEPT and to transfer products from the temporary exclusion list to the inclusion list in five stages. In 1993, the trade value of all products included in the CEPT covered 85% of total intra-ASEAN and 86% of total ASEAN trade. Malaysia and Singapore began implementing the CEPT agreement starting in 1993, Brunei Darussalam in 1994, Indonesia and Thailand in 1995, and the Philippines in 1996. It was expected that the simple average tariff of CEPT products would decline from 13.40 % in 1993 to 2.63 % in 2008.iii Between 1994 and 1998, ASEAN member countries further enhanced trade liberalization in the region by eliminating non-tariff barriers and quantitative restrictions; harmonizing customs nomenclature, valuation, and procedures; and developing common productcertification standards. In 1995, ASEAN signed framework agreements for the intraregional liberalization of trade in services and for regional cooperation in intellectual

2

property rights. The AFTA project became more formal and binding with the signing of the Protocol on ASEAN Dispute Settlement Mechanism (DSM) in 1996. The ASEAN DSM would provide an expeditious and transparent way of settling disputes among member states regarding the implementation of ASEAN economic agreements. In the same year, the ASEAN Industrial Cooperation Scheme (AICO), whose main objective was to promote joint manufacturing industrial activities between ASEAN-based companies, was launched.iv The expansion of ASEAN to include Vietnam in 1995, Laos and Myanmar in 1997, and Cambodia in 1999 widened the AFTA project as all four countries were required to sign on to the AFTA agreement in order to join ASEAN. In light of their later accession to the CEPT Agreement, the new member countries (the so-called CLMV countries) were scheduled to meet their tariff reduction obligations a few years later than the original ASEAN members. Vietnam was expected to realize AFTA in 2006, Laos and Myanmar in 2008, and Cambodia in 2010. Although the CEPT scheme had operated with strictly reciprocal preferences, the original ASEAN members later agreed in 2001 to unilaterally extend tariff preferences to ASEAN's new members from 2002 onwards so that integration of the CLMV countries would be accelerated. Spurred by the Asian Financial Crisis, ASEAN Leaders agreed at the sixth ASEAN Summit in 1998 to bring forward the establishment of AFTA to the year 2002 instead of 2003 for the six original signatories to the CEPT Agreement. Members also agreed to adopt a harmonized tariff nomenclature by 2000. In addition, to encourage intra-regional investment and production that would complement the AICO, a framework agreement on the ASEAN Investment Area was signed. AFTA effects on ASEAN Trade To manage preferential trade for ASEAN exporters, AFTA initially relied on the 40% regional content rule, but the alternative criterion of a tariff-classification change at the HS 4-digit level was added in 1995. The evidence for the utilization of ASEAN trade preferences is mixed. Baldwin (2006) reports that the utilization rates of AFTA were very low at around 3% of intra-regional trade, but a study by the Thai Office of Industrial Economics (2006) reports that, among products with CEPT rates below the MFN (MostFavored Nation) rates, the average utilization rate in 2006 was 51.2% of the value of Thai exports to other ASEAN countries while it was 26.9% of the value of Thai imports from other ASEAN countries. These findings may be consistent in that Baldwin’s figure was ASEAN-wide, and, since Singapore dominates intra-ASEAN trade and has zero MFN tariffs on almost all products, preferences are irrelevant for much of Singaporean trade and, by extension, much of intra-ASEAN trade. So, in effect, the non-use of preferences in Singaporean trade and its heavy weight in intra-regional trade may have diluted the high Thai utilization rate and resulted in a low region-wide average utilization rate. Beside the effect of the Singaporean trade regime on trade preference utilization, there are also other reasons for a low region-wide average utilization rate. Manchin and Pelkmans-Balaoing (2007) note three problems related to the CEPT’s rules of origin that

3

possibly explain a low average utilization rate of the CEPT scheme. Firstly, the CEPT scheme is relatively complex with MFN and CEPT tariff rates varying by product and ASEAN importing country. However, to reduce uncertainty in preferences, ASEAN countries have published their MFN and CEPT tariffs three years in advance.v Secondly, ASEAN exporters have some difficulty cumulating the required 40% local content.vi Thirdly, the CEPT’s administrative and compliance requirements are cumbersome as they lack transparency and uniformity.vii However, these authors find that preferential tariffs stimulated intra-ASEAN trade when they were at least 25% lower than MFN tariffs. According to the ASEAN Statistical Yearbook (2006), in 1999, intra-ASEAN trade was US$132 billion and by 2003, it was US$206 billion, which represents a 56% increase. As a share of total ASEAN trade, intra-regional trade was 21% in 1999 and 25% in 2003. As such, it appears, at the aggregate level, that AFTA may have increased intra-regional trade in the early 2000s. Trade Liberalization and Production Fragmentation in ASEAN Production fragmentation is a prominent feature of cross-border economic activity in East Asia. It is likely that multilateral and regional reductions in official trade barriers (tariffs, quotas, etc.) encouraged the integration of ASEAN countries into the East Asian production network. Other possible factors include innovation in technologies that allowed the fragmentation of production; falling transportation and distribution costs; and improvements in communications. The logic of production sharing entails a rise in both trade of final goods and, particularly, of parts and components between countries in a production network. Ng and Yeats (1999) found that, just prior to the Asian Financial Crisis, trade in parts and components accounted for about a fifth of East Asian manufacturing trade, having grown much faster in the preceding decade than other product groups. In addition, East Asian trade in parts and components was concentrated in just a few items, namely, telecommunications equipment, office and adding machinery, switchgear parts, motorvehicle parts and accessories, and electronic components. Ando (2006) computes that intra-ASEAN exports among four ASEAN countries (Indonesia, Malaysia, Philippines, and Thailand) as a share of world trade were 3.6%, 4.1%, and 7.5% in 1981, 1991, and 2001 respectively. In other words, controlling for global export trends, ASEAN trade was increasingly intra-regional in the last two decades of the 20th. century. Ando also computes the same shares of exports from China, the NIEs (i.e., Taiwan, Korea, Hong Kong, and Singapore) and Japan to the ASEAN-4. She finds that these shares increased for China and the NIEs but fell for Japan. Therefore, China and the NIEs became increasingly important as sources of ASEAN imports while Japan became less important in the same period. The type of trade, in which ASEAN countries engage, has also changed. Ando also shows that one-way trade has given way mainly to vertical intra-industry trade.viii This is particularly true in the machinery, transport equipment, and electronics sectors.

4

In theory, how do PTAs affect trade when production fragmentation exists? In a 3country general-equilibrium model, Arndt (2004) shows that, unless the non-member country has the lowest cost for every activity in the production chain, production sharing between member countries reduces the trade-diverting elements of preferential trade liberalization. This is because, with a finer division of the value chain, it is more likely that member countries of the preferential trading arrangement have a comparative advantage in some production activities. So, reduced tariffs, albeit just within the region, creates trade in these activities rather than diverts it as member countries engage in a finer specialization of activities along the value chain. Indeed, Athukorala and Yamashita (2006) provide evidence that vertical trade was very responsive to the formation of AFTA but not in other PTAs around the world. Although trade diversion may be less likely in the context of fragmented production, PTAs may induce other negative effects. Chen et al. (2004) show that tariff reductions in intermediate-goods markets may lead to strategic outsourcing, and rival firms in final goods may be more inclined to collude and increase import prices of both final and intermediate goods.

Empirical Model and Data There exist many studies at the aggregate level on the trade impact of PTAs with mixed evidence for trade creation or diversion (see Baier and Bergstrand, 2007; Glick and Rose, 2002; Ghosh and Yamarik, 2004). Adams et al. (2003) provide an extensive review of the empirical findings of such aggregate-level studies. They also summarize the results of seven studies that consider the trade effects of AFTA, noting that all these studies find more trade creation than trade diversion. However, in studies not considered by them, two do not find positive results for AFTA (Sharma and Chua, 2000; and Elliott and Ikemoto, 2004), while two do find evidence for greater intra-ASEAN trade (Hapsari and Mangungsong, 2006; and Bun, Klassen, and Tan, 2007). To contribute to the empirical literature on AFTA, I use the gravity model (see Anderson, 1979; and Bergstrand, 1985) for theoretical foundations) as the framework for my empirical analysis. The gravity model is an expenditure equation that basically relates one country’s import expenditure on another country’s output to the importing and exporting countries’ GDPs and trade costs between the two countries. Trade costs capture all barriers that impede trade and include all border-related (e.g. tariff) costs, transportation and distribution costs, and exchange rate costs. Although the model is often estimated with total bilateral trade, the underpinning theory actually explains unidirectional trade from one country to another. Furthermore, as Anderson and Wincoop (2004) point out, estimating the gravity model at an aggregated level is likely to lead to biased estimates because border barriers and elasticities of substitution vary by product. In addition, as Medvedev (2006) finds, an estimation of the gravity equation with aggregate data and PTA dummy variables may detect little benefit from preferential trade liberalization because the PTA may initially exclude a significant number of products and have long phase-in periods. With disaggregated data on trade and preference margins, I am able to control for unobserved heterogeneity at the import-exporter-product level. So,

5

to explain imports (Mijk) of country i from country j of product k in year t, I use the following estimating equation:

ln M ijkt = β 0 + β 1 ln Prefmarg ikt * ( D jNONASEAN = 1) + β 2 ln Prefmarg ikt * ( D jASEAN = 1) +

β 3 ln GDPit + β 4 ln GDPjt + α ijk + α t + ε ijkt The first two explanatory variables are the natural logarithm of the preference margin offered by country i (ln Prefmargikt) interacted with dummies for nonASEAN trade partners and ASEAN trade partners. The preference margin is calculated as the difference between the simple average MFN tariff and the simple average CEPT tariff. The set of importer (i) countries considered is only ASEAN. The coefficient on the PrefmargnonASEAN interaction will capture trade diversion, if negative, while that on the Prefmarg-ASEAN interaction will measure trade creation, if positive. The difference between the coefficients will provide a measure of the net welfare effects of AFTA. The next two explanatory variables are those conventionally used in gravity equations, i.e. GDP of the importing and exporting countries. To capture all time-invariant effects of any pair of trading countries, which includes the distance variable, I include importerexporter-product fixed effectsix. Besides distance, these fixed effects capture variables such as remoteness, exchange rate regimes, colonial history, common languages, prior industry expectations, and other fixed industry-level characteristics. I also include year effects. As Santos and Senreyro (2006) note, estimation of the log-linear specification of the gravity model with OLS in the presence of heteroskedasticity will produce inconsistent estimates of the elasticities. This result, which applies generally to log-linearized models, obtains because the non-linear transformation of the estimating equation makes the conditional expectation of the error depend on higher moments of the data. They recommend using the Poisson Pseudo- (or Quasi-) Maximum Likelihood (PQML) estimator to solve this problem. Another benefit of this estimator is that zero observations of the dependent variable can be included in the estimation to avoid truncation bias. Given panel data, I estimate the model with fixed-effects PQML. The PQML estimator is perfectly robust to heteroskedasticity if the conditional variance is proportional to the conditional mean. However, as this assumption is unlikely to hold, I compute robust standard errors as derived by Wooldridge (1999) for the fixed-effects PQML estimator. The import data is taken from COMTRADE, MFN and preferential tariff rates are sourced from the ASEAN Secretariat, and GDP data are from the World Bank’s Development Indicators. As tariff-rate data is published at the HS 9-digit level, I compute the tariff rates for each HS 6-digit category as the simple average tariff. This is necessary as the most disaggregated import data available for ASEAN countries from COMTRADE are at the HS 6-digit level. The sample is limited to the period 2001 to 2003 for three reasons. First, CEPT tariffs during this period were predetermined by ASEAN countries in 1998 following the Asian Crisis. Therefore, CEPT tariffs in 2001 to 2003 are less likely to be endogenous and neither are preferential margins as MFN tariffs did not change from 2001 to 2003.

6

Second, data on ASEAN preference margins are available to the public only from 2001 onwards. Third, there was a significant change in 2002 of the HS trade-classification system, which was subsequently implemented in 2003 onwards by ASEAN countries. As such, by ignoring data after 2003, I avoid attributing any AFTA effects spuriously to the change in trade classification. In addition, this study does not consider foreign direct investment (FDI) and non-tariff trade barriers as explanatory variables although these are important factors in determining trade flows. The main reason for their exclusion is data availability – which is very rarely at a disaggregated level – and data quality. Results

Table 1 shows the results for a sample of all ASEAN imports with positive preferential margins both from ASEAN and non-ASEAN partners. The specifications in Columns 1 and 2 do not include a variable to capture trade creation. In Column 1, the results from the OLS within estimator show significant trade diversion with an elasticity of 0.08. However, in Column 2, when the Fixed Effects PQML estimator is used, the sign is reversed, implying that higher preferential margins increased trade with non-ASEAN countries at the HS 6-digit level. This effect is likely to be the result of increasing production fragmentation even within a HS 6-digit category following intra-regional trade liberalization. The estimated elasticity of intra-regional trade creation is 0.26. Given that the average increase in preference margins during the period was 8%, the average preferred HS 6-digit category experienced an increase in intra-regional ASEAN imports of 2% and even in extra-regional ASEAN imports of 0.7%. Using the average value of about US$395,000 in preferred ASEAN imports for a HS 6-digit category in 2001, this would translate to an increase in US$8,000 in intra-regional imports and US$2,500 in extra-regional imports at the product level. With over 150,000 HS 6-digit products involved, the expansion in trade is around US$1.5 billion. The GDP variables are estimated with the wrong sign, but they are not statistically significant at the 1% level. Table 2 performs the estimations by ASEAN country. The estimated trade diversion elasticities are all positive, indicating increased trade with non-ASEAN partners. However, they are statistically significant at the 5% level for only Brunei, Cambodia, Indonesia, the Philippines, and Thailand and range from 0.17 for Thailand to 2.7 for Cambodia. As for the intra-regional trade creation elasticities, these are also positive for all countries but significant at the 5% level only for Cambodia, Malaysia, Thailand, and Vietnam. These range from 0.22 for Thailand to 4 for Cambodia. This country-level evidence may be consistent with a story of intra-regional trade liberalization leading to further production fragmentation, finer specialization, and increased intra- and extraregional trade. In Table 3, the estimation results are by the HS 1-digit category. In contrast to the previous results, the trade diversion elasticities are negative for half the industry categories, and significantly negative at the 5% level for HS1 (Vegetable Products and Foodstuffs) and HS4 (Raw Hides, Skin, Leather and Fur, and Wood and Wood Products). As HS1 represents industries with relatively little production fragmentation, it is not surprising that AFTA had trade-diversion effects in this category. As for the trade-

7

creation elasticities, again, about half the industry categories show a negative sign, but there are statistically significant results only among the positive estimates (i.e. HS4 and HS8). As such, the net trade creation effect for HS4 was 0.07 (=0.212-0.141) and for HS8, it was 0.213 (=0.283-0.077). The HS8 category includes industries such as Metals, Machinery, Electrical Products, and Transportation, which are mainly industries with complex and intense production fragmentation. HS8 is also the largest HS 1-digit category in terms of import value for ASEAN, accounting for about 45% of total ASEAN imports. As such, this result implies that AFTA had a positive and significant welfare effect on the bulk of ASEAN trade. As robustness checks, I first test for the presence of heteroskedasticity by using RESET and find that the conditional distribution of the data is indeed heteroskedastic. As such, I am justified in not using the OLS Within estimator but the Fixed Effects PQML estimator. Next, I use a Park-type test (see Santos-Silva and Tenreyro, 2006, p.646 ) to check whether OLS estimation of my empirical model would be consistent. The test rejects the use of OLS estimation, implying that the PQML estimator should be used for consistency.x Conclusion

This paper has found that AFTA had net trade creation effects at a disaggregated level. The positive effects of trade creation were felt mainly in the industries covered by the HS8 1-digit category. Industries in this category are known to have complex production networks and a high level of product differentiation, both vertical and horizontal. As such, AFTA appears to have intensified production fragmentation and intra-industry trade. The paper’s main finding has one key policy implication: ASEAN should proceed on its path towards an ASEAN Free Trade Area by liberalizing trade in sectors with a higher propensity to fragment production as this tends to create trade and, thus, have a more positive welfare effect. The trade-diversion effects may be sufficiently mitigated by production fragmentation to make the preferential trading arrangement a good secondbest solution. So, if a PTA is the only politically-feasible alternative, policy makers may make it a second-best solution by prioritizing sectors that are part of production chains However, this paper does not imply that ASEAN or other regional groupings should continue to pursue PTAs at the expense of multilateral liberalization as there is still a risk that these PTAs induce trade diversion.

8

References

Adams, Richard, Philippa Dee, Jyothi Gali, and Greg McGuire (2003), “The trade and investment effects of preferential trading arrangements—old and new evidence.” Working paper, Productivity Commission, Canberra. Anderson, James E. (1979), “A theoretical foundation for the gravity equation”, American Economic Review, v69, p106-116. Anderson, James E. and Eric van Wincoop (2004), “Trade Costs”, Journal of Economic Literature, v42, p691-751. Ando, Mitsuyo (2006), “Fragmentation and vertical intra-industry trade in East Asia”, The North American Journal of Economics and Finance, v17(3), p257-281. Arndt, Sven (2004), “Trade Diversion and Production Sharing”, Journal of Economic Assymetries, v1(1). ASEAN Statistical Yearbook (2006), ASEAN Secretariat, Jakarta: Indonesia. Athukorala, Prema-chandra and Nobuaki Yamashita (2006), “Production fragmentation and trade integration: East Asia in a global context”, The North American Journal of Economics and Finance, v17(3), p233-256. Baier, Scott L. and Jeffrey H. Bergstrand (2007), “Do free trade agreements actually increase members' international trade?”, Journal of International Economics, v71(1), p7295. Baldwin, Richard (2006), “Managing the Noodle Bowl: The Fragility of East Asian Regionalism”, CEPR Discussion Paper No. 5561. Bergstrand, Jeffrey H. (1985), “The gravity equation in international trade: Some microeconomic foundations and empirical evidence”, The Review of Economics and Statistics, v67, p474-481. Bun, Maurice, Franc Klaassen, and Randolph Tan (2007), “Free trade areas and intraregional trade: the case of ASEAN”, Econometrics Discussion Paper: 2007/04, Universiteit van Amsterdam. Chen, Yongmin, Jota Ishikawa, and Zhihao Yu (2004), “Trade liberalization and strategic outsourcing”, Journal of International Economics, v63(2), p419-436. Elliott, Robert J. R. and Kengo Ikemoto (2004). “AFTA and the Asian Crisis: Help or Hindrance to ASEAN Intra-Regional Trade?”, Asian Economic Journal, East Asian Economic Association, v18(1), pages 1-23.

9

Ghosh, Sucharita and Steven Yamarik (2004). “Are regional trading arrangements trade creating?: An application of extreme bounds analysis”, Journal of International Economics, v63(2), p369-395. Hapsari, Indira M. and Carlos Mangunsong (2006), “Determinants of AFTA Members’ Trade Flows and Potential for Trade Diversion”, Asia-Pacific Research and Training Network on Trade Working Paper Series, No. 21 Manchin, Miriam and Annette O Pelkmans-Balaoing (2007), “Rules of Origin and the Web of East Asian Free Trade Agreements”, World Bank Policy Research Working Paper No. 4273. Medvedev, Denis (2006), "Preferential trade agreements and their role in world trade," World Bank Policy Research Working Paper Series 4038. Ng, Francis and Alexander Yeats (1999),“Production sharing in East Asia : who does what for whom, and why?”. Policy Research Working Paper Series 2197, The World Bank. Santos Silva, J. M. C. and Silvana Tenreyro (2006), “The Log of Gravity”, The Review of Economics and Statistics, v88(4), p641-658. Sharma, S. C., and S. Y. Chua (2000), “ASEAN: economic integration and intra-regional trade”, Applied Economics Letters, v7, p165-169. Thai Office of Industrial Economics (2006), http://www.oie.go.th/project/fta/03.pdf Viner, Jacob (1950), “The Customs Union Issue”, New York: Carnegie Endowment for International Peace. Wooldridge, Jeffrey M. (1999), “Distribution-free estimation of some nonlinear panel data models”, Journal of Econometrics, v90(1), p77-97. World Bank (2005), “Global Economic Prospects: Trade, Development, and Regionalism”, Washington.

10

Summary Statistics Variable ASEAN Country i's Import Value of HS6 Product k (in US$) ASEAN Importer Country's GDP (in US$) Partner Country's GDP (in US$) ASEAN Country's Preference Margin (in %) on Product k

Obs 413118

Mean 408583.4

Std. Dev. 7662207

Min 0

Max 2.29E+09

413118

1.16E+11

5.22E+10

3.98E+09

2.35E+11

413118

1.38E+12

2.46E+12

4.50E+07

1.09E+13

413118

11.62184

9.040214

0.083333

146.5962

11

Table 1: The Effects of AFTA on ASEAN Imports (2001 to 2003)

Dependent Variable: ASEAN Country i's Import Value of Product k from Country j in Year t (1) (2) (3) (4) Explanatory Variables ln(Country's Preference Margin (in %) on Product k in Year t) * -0.079** 0.137** -0.075** 0.149** (DNONASEAN=1) (0.023) (0.039) (0.023) (0.041) ln(Country's Preference Margin (in %) on Product k in Year t) * 0.072 0.260** (DASEAN=1) (0.045) (0.063) ln(Country i's GDP in Year t) -0.639** -0.523* -0.657** -0.573* (0.089) (0.256) (0.089) (0.262) ln(Country j's GDP in Year t) 0.374** -0.126 0.378** -0.132 (0.053) (0.082) (0.053) (0.082) year 2002 0.079** 0.153** 0.080** 0.154** (0.012) (0.033) (0.012) (0.033) year 2003 0.216** 0.193** 0.216** 0.196** (0.022) (0.052) (0.022) (0.053) Observations 333247 413118 333247 413118 Number of groups (by HS6 153305 153305 153305 153305 classification) Estimation Method Fixed Fixed Fixed Fixed Effects Effects Effects Effects Poisson Poisson Notes: 1) Statistical significance indicated as: + significant at 10%; * significant at 5%; ** significant at 1% 2) For the Fixed-Effects Estimator, the dependent variable is the natural logarithm of import value, and the reported standard errors are robust. 3) For the Fixed Effects Poisson estimator, the dependent variable is the level of import value, and the reported standard errors are robust and calculated according to Wooldridge (1999). 4) All regressions exclude Laos, Myanmar, and Singapore. Singapore is not included because Singapore's MFN tariffs were all zero during the sample period. 5) The sample period is 2001-2003.

12

Table 2: The Effects of AFTA on ASEAN Imports by Country (2001 to 2003)

Explanatory Variables ln(Country's Preference Margin (in %) on Product k in Year t) * (DNONASEAN=1) ln(Country's Preference Margin (in %) on Product k in Year t) * (DASEAN=1) ln(Country j's GDP in Year t) year 2002

Dependent Variable: ASEAN Country i's Import Value of Product k from Country j in Year t (1) (2) (3) (4) (5) (6) (7) Brunei Cambodia Indonesia Malaysia Philippines Thailand Vietnam 1.970* (0.903)

2.693** (0.842)

0.265** (0.090)

0.180 (0.109)

0.201* (0.086)

0.174** (0.055)

0.002 (0.117)

0.770 (1.198) -2.051 (1.580) 0.398* (0.172)

5282 2641

4.039** (0.696) -4.471** (1.034) -0.758** (0.220) -0.960** (0.372) 6406 2655

0.215+ (0.120) -0.080 (0.457) -0.050+ (0.027) -0.198 (0.135) 69564 26848

0.669** (0.138) 0.099 (0.273) 0.081* (0.037) -0.064 (0.063) 81548 29389

0.283+ (0.152) 0.267 (0.222) -0.021 (0.024) -0.092* (0.039) 66739 24443

0.220* (0.099) -0.095 (0.094) 0.136** (0.023) 0.189** (0.034) 149934 54546

0.491* (0.194) -0.530** (0.176) 0.151** (0.058) 0.326** (0.069) 33645 12783

Fixed Effects Poisson

Fixed Effects Poisson

Fixed Effects Poisson

Fixed Effects Poisson

Fixed Effects Poisson

Fixed Effects Poisson

Fixed Effects Poisson

year 2003 Observations Number of groups (by HS6 classification) Estimation Method

Notes: 1) Statistical significance indicated as: + significant at 10%; * significant at 5%; ** significant at 1% 2) Reported standard errors are robust and calculated according to Wooldridge (1999). 3) As these regressions are by ASEAN country, the GDP variable for Country i is subsumed into the year variables. 4) The model for Singapore was not estimated because Singapore's preferential margins were all zero in the sample.

13

Table 3: The Effects of AFTA on ASEAN Imports by HS 1-digit Category (2001 to 2003)

Explanatory Variables ln(Country's Preference Margin (in %) on Product k in Year t) * (D NONASEAN =1) ln(Country's Preference Margin (in %) on Product k in Year t) * (D ASEAN =1) ln(Country i 's GDP in Year t) ln(Country j 's GDP in Year t) year 2002 year 2003 Observations Number of groups (by HS6 classification) Estimation Method

(1) HS0

Dependent Variable: ASEAN Country i 's Import Value of Product k from Country j in Year t (2) (3) (4) (5) (6) (7) (8) HS1 HS2 HS3 HS4 HS5 HS6 HS7

(9) HS8

(10) HS9

0.112 (0.330)

-0.390** (0.128)

-0.037 (0.106)

-0.053 (0.044)

-0.141* (0.062)

0.037 (0.096)

-0.028 (0.159)

-0.002 (0.061)

-0.077 (0.094)

0.057 (0.132)

1.379 (1.270) -0.914* (0.446) -0.082 (0.134) -0.000 (0.253) 0.574 (0.466) 11987 4481

0.010 (0.164) 2.778** (0.700) 0.156* (0.065) -0.113+ (0.066) -0.295* (0.145) 11046 4046

-0.127 (0.123) -0.619+ (0.363) -0.243* (0.124) 0.021 (0.064) 0.239* (0.102) 21118 7718

0.072 (0.073) -0.673** (0.191) 0.100 (0.147) 0.182** (0.029) 0.364** (0.054) 44591 16100

0.212+ (0.117) -0.269 (0.335) 0.271 (0.179) 0.086* (0.037) 0.186* (0.075) 38235 14020

-0.006 (0.327) -1.710** (0.278) 0.007 (0.168) 0.173** (0.036) 0.323** (0.064) 46819 17042

0.213 (0.402) -0.256 (0.526) 0.299 (0.489) -0.034 (0.079) 0.064 (0.126) 51008 18988

-0.104 (0.124) 0.285 (0.352) 0.006 (0.219) 0.037 (0.047) 0.155+ (0.094) 51215 18553

0.283* (0.124) 0.078 (0.448) -0.225 (0.266) 0.080 (0.058) 0.193* (0.098) 104409 40143

-0.275 (0.188) -0.510 (0.608) 0.625* (0.287) 0.121* (0.060) 0.236+ (0.143) 32690 12214

Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Poisson Poisson Poisson Poisson Poisson Poisson Poisson Poisson Poisson Poisson

Notes: 1) Statistical significance indicated as: + significant at 10%; * significant at 5%; ** significant at 1% 2) For the Fixed-Effects Estimator, the dependent variable is the natural logarithm of 3) For the Fixed Effects Poisson estimator, the dependent variable is the level of import 4) All regressions exclude Laos, Myanmar, and Singapore. Singapore is not included 5) The sample period is 2001-2003.

14

Endnotes i

Although the initial AFTA framework agreements were general, senior officials formulated specific details over the course of the year and drew up three implementation documents: 1) The Operational Procedures for CEPT, 2) Rules of Origin for CEPT, and 3) Interpretative Notes to the Agreement on the CEPT Scheme for AFTA. The AFTA approach of working out details after launching the project has been characterized as “Agree First Talk After”. ii For example, in 2000, Malaysia invoked the Protocol Regarding the Implementation of the CEPT Scheme Temporary Exclusion List and delayed tariff reductions on completely-built-up automobiles and automobile knock-down kits in order to protect its local auto industry. iii Joint Press Statement by the 5th. Meeting of the AFTA Council, Chiang Mai, Thailand, September 21, 1994. iv The AICO was intended to be an important feature of ASEAN economic cooperation. It was designed to encourage technology-based investments in ASEAN, and was open to any ASEANbased company meeting the following requirements: 1) incorporated in and operating in an ASEAN country, 2) a minimum of 30 percent ASEAN equity, 3) the company engaged in some form of resource sharing (such as sharing of technology, market sharing, or consolidated purchases of raw materials). Raw materials, intermediate products, and output of approved AICO projects would enjoy 0-5% tariffs immediately. Such products would also enjoy local content accreditation and non-tariff incentives. v However, given the annual publication of tariff rates, some updating may have taken place between the publication year and the year the tariffs came into effect. vi Manchin and Pelkmans-Balaoing illustrate this by showing that the import content (sum of direct and indirect inputs) of key manufacturing sectors in ASEAN in 2001 ranged from 31% (other manufacturing) to 67.8% (electronics). vii To benefit from preferences, ASEAN exporters are required to submit a cost analysis of the product to be exported with accompanying invoices for scrutiny and approval by the local trade ministries. If approved, the local trade ministry provides the exporter with Form D, which is a certificate that the product qualifies for trade preferences. The exporter may then export the good to another ASEAN country at the preferential rate, but the goods are still subject to random checks. It is likely that, instead of facilitating intra-regional trade, the process has turned out to be more timeconsuming and costly in practice relative to ordinary customs clearance. viii Intra-industry trade, which is the overlap of imports and exports within a product category, is classified as vertical if the unit value of imports in the product category differs by at least 15% from the unit value of exports in the same product category. ix The importer-exporter-product fixed effects also help control for all factors related to import decisions made prior to 2001. These could be, for example, preannounced tariff rates and procedural changes in customs or FDI flows into an industry that were planned before 2001 but implemented after. x Test results from these robustness checks are not reported but are available from the author ([email protected]).

15