Tariff rates in gravity

Abstract

AbstractThis paper considers the case where multiple tariff schemes (e.g. general and preferential schemes) are available between trading countries. We incorporate these tariffs into gravity equations and estimate them by the Pseudo-Poisson maximum likelihood technique. The results show that omitting either tariff type leads to significant estimates biases. If fixed effects to control other tariffs are not included in the model, both preferential and general tariffs are to be introduced in the gravity equation. Indeed, some estimation results for precision metals show that reducing both types of tariffs contributes to significantly increasing trade values. However, reducing general tariffs does not always have a trade-enhancing effect. In leather products, for example, its impact was insignificant. Nevertheless, the reduction of preferential tariffs was again found to increase trade values significantly.KEYWORDS: Gravitytariffsregional trade agreementsJEL Classifications: F15F53 AcknowledgementWe would like to thank Kyoji Fukao, Norikatsu Hiraide, Naoto Jinji, Fukunari Kimura, Yoshimasa Komoriya, Hitoshi Sato, Kenta Yamanouchi, and the seminar participants at Keio University, Kangwon National University, the Economic Research Institute for ASEAN and East Asia (ERIA), the Nagoya International Economics Study Group (NIESG), the Japan Society of International Economics, the Japanese Economic Association, the Western Economic Association International (WEAI), and the Institute of Developing Economies (IDE-JETRO).Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 The earlier studies focused on trade by a single country. The examples include Trefler (Citation2004), Romalis (Citation2007), and Debaere and Mostashari (Citation2010). Also, the recent US-China trade disputes increased the number of studies on the trade effects of tariffs in the US or China (Amiti, Redding, and Weinstein Citation2019; Amiti, Redding, and Weinstein Citation2020; Fajgelbaum et al. Citation2020; Ma, Ning, and Xu Citation2021).2 Specifically, imports in Australia from the United States or Canada are approximately 50%. The share of European Union imports from Mexico is about 80%. A similar share can be found in the case of U.S. imports from Australia.3 In the empirical part, we define products at the tariff-section level, i.e., the industry level.4 Following Helpman, Melitz, and Rubinstein (Citation2008) and Helpman, Melitz, and Yeaple (Citation2004), we assume exporters pay fixed costs for products to every destination while dealing with multiple destinations simultaneously. Hence, they cannot save the total fixed costs based on economies of scale. A similar situation is assumed for the fixed cost for RTA usage. These cases are not examined to ensure the model's tractability and help develop an explicit gravity equation.5 We also employ the assumption used in Chaney (Citation2008) that the total mass of potential entrants in each country in each differentiated sector is proportional to national labor income. As a result, larger and wealthier countries have more entrants.6 In contrast, the case where all exporters utilize RTA is called the homogeneous regime (Demidova and Krishna Citation2008). Under the homogeneous regime, MFN tariffs do not affect bilateral trade on the product-level. We focus on the heterogeneous regime to reveal the role of alternative tariffs in gravity analysis. In fact, the homogeneous regime is rarely observed in the data.7 Equation (A3) in Appendix A1 presents the theoretical condition for realizing the heterogeneous regime. Specifically, the heterogeneous regime takes place when (i) the fixed cost for RTA utilization is high compared to the fixed cost for exporting, (ii) the variable trade cost, i.e., the product of RTA tariff rates and procurement adjustment cost, is not significantly low compared to MFN rates, or (iii) the price elasticity of demand is small. Our focus on the heterogeneous regime is reasonable because our empirical analysis is conducted at an industry-level. At this level, it is relatively rare to observe that the share of imports under the RTA regimes is 0% or 100%. See, for example, Keck and Lendle (Citation2012).8 Equilibrium values are shown in Appendix A. Specifically, to derive (1), we used (A4), (A6), (A7), (A8), and (A9) for wiLi, xijh(φ), φijhM, φijhR>M and π, respectively.9 The number of MFN users tends to be large, and the coefficient for MFN rates tends to be negative if (i) initial levels of MFN rates are low, (ii) initial levels of RTA rates or procurement adjustment cost are high, or (iii) initial level of fixed cost for RTA utilization is high.10 Olivero and Yotov (Citation2012) provide the theoretical foundations for a dynamic gravity model.11 Henderson and Millimet (Citation2008) relax this assumption and estimate gravity models.12 We can normalize wages in all countries to 1 as we employ the numeraire sector and focus on equilibria where all countries produce some of the numeraire product. Therefore, for simplicity, we set wit to 1 hereafter.13 Baier and Bergstrand (Citation2009) showed that a Taylor approximation produces coefficient estimates virtually identical to those obtained from an unbiased nonlinear procedure. See also Anderson, Vesselovsky, and Yotov (Citation2016) and Behar and Nelson (Citation2014) for applying Taylor approximation in the derivation of gravity equations.14 Definitions and signs of the coefficients in equation Equation(5)(5) ln⁡Tijht≅ln⁡T¯h+ΔτM(ln⁡τjhtM−ln⁡τ¯hM)+ΔτR(ln⁡τijhtR−ln⁡τ¯hR)+Δρ(ln⁡ρijh−ln⁡ρ¯h)+Δf(ln⁡fijh−ln⁡f¯h)+ΔfR(ln⁡fijhR−ln⁡f¯hR),(5) are given in Appendix A4.15 As in the usual gravity studies, like Feenstra (Citation2002), third-country effects through multilateral resistance are controlled by our inclusion of import country-year fixed effects.16 From the theoretical perspective, the coefficients for tariff variables should be different across countries and products. However, we do not differentiate those and estimate the average magnitude.17 http://www.cepii.fr/CEPII/en/welcome.asp.18 As pointed out by Teti (Citation2020), there are some things that could be improved in the WITS database, which include missing data and misreporting. We set the missing tariff equal to the nearest preceding observation but do not address the misreporting issue.19 Our study period does not include the pandemic period when many countries imposed various trade-restrictive measures.20 The estimation results for other industries are reported in Table B2 in Appendix B.Additional informationFundingThis work was supported by Chuo University: [Grant Number Chuo University Grant for Special Research]; Japan Society for the Promotion of Science KAKENHI: [Grant Number JP17H02530 and JP20H01518].