Trump Tariffs and Roundabout Trade^†

A. Literature

Previous studies focusing on this topic, although not necessarily exhaustive, found mostly negative impacts of Trump’s tariffs on the US economy. Using difference-in-difference estimations, Amiti, Redding, and Weinstein (2019) and Amiti, Redding, and Weinstein (2020) showed substantially negative effects of Trump’s tariffs on Chinese exports to the US. More specifically, the US tariffs are almost entirely borne by US firms and consumers (no terms of trade effect). A substantial redirection in trade in response to the 2018 tariffs took place. A 10% tariff is associated with approximately a 10% drop in imports for the first three months, but this elasticity doubles in magnitude in subsequent months. Amiti, Redding, and Weinstein (2019) showed that the cumulative deadweight welfare cost (reduction in real income) from the US tariffs was around $8.2 billion in 2018. In an event study estimation, Fajgelbaum et al. (2020) showed that the US import values of Trump-listed goods from targeted countries decreased by 31.7%, whereas the US total import values for Trump-listed goods, irrespective of the import partner country (regardless of whether the import partners were subject to the Trump tariffs or not) decreased by 2.5%. US firms and consumers who bought Trump-listed imported goods suffered US$51 trillion in real income losses, equivalent to 0.27% of US gross domestic product (GDP). The sum of this loss for US firms and consumers plus the increase in consumer surplus (due to a slight decrease in the border price) and tariff revenues resulted in a loss of US$7.2 trillion, equivalent to 0.04% of US GDP. Studying the tariff pass-through at the levels of US retailers and consumers, Cavallo et al. (2021) also showed findings similar to those by the studies above, most notably the complete pass-through of the Trump tariffs to US domestic prices. For consumer goods such as washing machines, handbags, and refrigerators, they found that the tariff burden fell completely on retailers, not consumers, as they found that the retail prices of these goods did not change.

Trump’s tariffs against China had discernible impacts on the Chinese economy. Chor and Li (2021), using satellite readings of night-time luminosity, showed that locations within China that were more exposed to the US tariffs experienced a greater decrease in night light intensity, pointing to contracted local economic activity. Cui and Li (2021) showed that the US import tariff hikes were associated with relative reductions in Chinese new firm entry rates. He, Mau, and Xu (2021) found that firms that were relatively more exposed to US tariffs in 2019 responded by posting fewer job advertisements in the six months following the tariff increase. The reduction amounted to 2.4%–3.2% fewer advertisements on average per firm.

Trump’s tariffs against China affected third countries as well. Using information at the tariff-line level on sanctions and retaliations and encapsulating this information in a general equilibrium framework featuring imperfect competition, recursive dynamics, and global value chains (MIRAGE-e V2), Bellora and Fontagné (2020) found that, consistent with political economy determinants, these twists of value added were transmitted to production factors, leading to sizeable creation and destruction of jobs and the reallocation of capital to the benefit of protected sectors, mostly at the expense of their clients. Regarding the effects on the activities of Japanese firms, Sun et al. (2019) using information on the activities of Japanese multinationals in China, showed that Chinese affiliates, especially those with high exposure to trade with North America, in general saw a decline in sales after the trade war began. Ito (2022a) found that Trump’s tariffs against China substantially decreased US imports from China, whereas many US import partner countries increased their exports to the US at the expense of China, in a kind of trade diversion effect. The study also found that whilst there was no sign of a decrease in border prices of US imports from China (no terms of a trade improvement), US imports from other partner countries of Trump-listed goods (targeting China) showed a decrease in their border price. Ito (2022b) showed that there was no increase in Japan’s import values/quantities of Trump-targeted goods from China, finding however that the import price decreased slightly. Surprisingly and contrary to a priori expectations, it was also found that Japanese industries that are linked as upstream industries of China’s (downstream) industries subjected to Trump’s tariffs increased their exports to China. To investigate the reason for this unexpected result, the analysis focused on China’s exports of Trump tariff-targeted goods to the world and found that China increased its exports of those goods to the world. Specifically, the increase in China’s exports to countries other than the US more than offset the decrease in its exports to the US. On the other hand, somewhat contrary to Ito’s (2022b) finding, Hayakawa et al. (2022) argued that the decrease in China’s output exports to the US caused by the Trump tariffs reduced China’s input imports from supplier economies, especially from Taiwan.

Tariff dodging by trade through third countries may explain the above-mentioned seemingly increasing exports of Trump-tariff goods from China to countries other than the US. This study investigates whether there is evidence of such roundabout trade using large and detailed trade statistics. Iyoha et al. (2024) presented the same research question in their study. Using Vietnamese customs data at the firm level, they showed that 16.1% of Vietnamese exports to the US were identified as having gone through product-level rerouting, whereas only 1.8% were flagged as having been through firm-level rerouting. This study differs from the work of Iyoha et al. (2024) in the following ways. First, the present study examines evidence of roundabout trade not only for Vietnam but also for nearly all potential rerouting countries. Second, it uses a different definition of roundabout trade. Iyoha et al. (2024) defines trade as roundabout trade when the same HS 8-digit products are imported to Vietnam and exported to the US. This definition is too loose because it is well known that there is substantial two-way trade even among the same HS 8-digit products, mainly owing to product varieties and vertical intra-industry trade. Unlike Iyoha et al. (2024), this study controls for intra-industry trade and thus more precisely defines roundabout trade. The definition of roundabout trade employed in this study is described in the following section. Given the lack of firm-level export and import data for many countries that could be involved in roundabout trade, this study attempts to investigate the issue by using export and import data at the most detailed product level.

A. Definition of roundabout trade

As mentioned in the introduction section, when goods are exported to final destination countries via third countries to circumvent export embargos on the destination countries or to avoid high import tariffs imposed by the final destination countries, such exports are termed roundabout trade or detour trade. There can be many ways in which roundabout trade is accomplished. For example, a Chinese firm can set up its affiliate (foreign direct investment) in, for instance, Vietnam, import goods from the headquarters in China, and re-export them to the US. Or, a firm in a bypass country, such as Vietnam, can import goods from China and then re-export them to the US. In the latter case, the Vietnamese firm in the middle is the one that engages in roundabout trade. The Chinese firm that exported the goods to the Vietnamese firm is unaware of the re-exporting from Vietnam. When goods are exported through third countries that have free trade agreements (FTAs) with the US, such as Canada or Mexico, FTA tariff rates may or may not be applied. When US importers apply for certain FTA tariff rates, they need to show what is known as a rules of origin certificate. However, legitimately relabeling ‘made in China’ to ‘made in Mexico’ and thus making the goods eligible for FTA tariffs incurs a relatively high cost because Mexican firms in the middle must engage in some processing of the products. On the other hand, the benefits gained by the use of FTA rates are relatively small because the gap between the most favored nation (MFN) rates and FTA rates is not substantial, usually only a few percent, such as an MFN rate of 3% and an FTA rate of 0%. As the Trump tariffs were mostly 25%, the benefit gained by roundabout trade changed little between MFN imports and FTA imports. Given this cost-benefit perspective, many firms most likely did not have much incentive to apply to use FTA rates. As mentioned in the introduction section, with trade data only, it is impossible precisely to identify roundabout trade. Instead, this paper aims to deduce whether there are general signs of roundabout trade by studying the unusually rapid increase of imports from China and the simultaneously rapid increase of exports to the US of Trump tariff goods compared to non-Trump tariff goods.

B. Data source

To investigate this issue, we use monthly trade data at the Harmonized System (HS) 6-digit level taken from the Global Trade Atlas. Given the extremely large sample size, which reaches tens of millions of observations, we focus on the major trade partner countries of the US and China that may have engaged in roundabout trade. More specifically, first we obtain yearly export and import data at the HS 6-digit level for the 20 countries that are the top US import partner countries as well as the top Chinese export partner countries. These countries are, in alphabetical order, Belgium, Brazil, Canada, France, Germany, India, Indonesia, Italy, Japan, Malaysia, Mexico, Netherlands, Philippines, the Republic of Korea (henceforth, Korea), Russia, Singapore, Spain, Thailand, the United Kingdom, and Vietnam. Then, by making graphs of these countries’ imports from China and exports to the US, we choose seven countries (Canada, India, Malaysia, Mexico, the Philippines, Korea, and Vietnam) as potential bypass routes for Trump-targeted goods. We obtain monthly trade data pertaining to these seven countries from the Global Trade Atlas.

The list of Trump-targeted goods is taken from official announcements by the Office of the United States Trade Representative (USTR). As shown in Table 1, more than half of products are targeted by Trump tariffs.

C. Methodology

As noted in section 2.1, the unusually rapid increase in imports from China and the simultaneous rapid increase in exports to the US of Trump tariff goods compared with non-Trump tariff goods is a sign of roundabout trade. The following is the estimation equation we employ to identify whether roundabout trade took place:

In this equation, y, m, i, j, and p represent the year, month, import partner, export partner, and product, respectively. Note that we estimate the equation according to each reporter country, such as Mexico, Canada, or Vietnam, to determine if the reporter country in question appears to have engaged in roundabout trade of the Trump-targeted goods. The covariate, lnImp_ymip, captures the association between the import value of the reporter country (e.g., Malaysia) of good p from country i to the corresponding export value of the same good p. The expected sign is positive because when the production of good (or industry) p in the reporter country (e.g., Malaysia) is expanded, the expansion brings about both imports of intermediate goods and exports of processed goods classified in the same category p. In other words, by including this covariate, we can control for general intra-industry trade, which is one of the differences between Iyoha et al. (2024) and this study. RoundaboutDummy takes a value of 1 if imports are from China (i =China) and exports are to the US (j =US), and p denotes Trump targeted goods. Thus, the cross-term lnImp_ymip ∗ RoundaboutDummy accounts for roundabout trade; α is the term representing fixed effects. We include full fixed effects to identify roundabout trade correctly. We estimate the equation with full fixed effects as above for the period of January 2016 to June 2019, which sufficiently covers before and after the Trump tariff shock. Out of the selected countries mentioned above, we especially focus on Canada, Mexico, and certain Asian countries. Canada and Mexico were chosen because they are both included in the United States-Mexico-Canada Agreement and have no import duties and thus are potential stopover places for roundabout trade. The Asian countries selected here are also potential stopover places owing to their proximity to China and to their large and regular export flows to the US. To clarify our identification strategy with regard to the estimation equation above, a sketch of the identification strategy is shown in Figure A1, and a sketch of the data structure is shown in Figure A2 in the appendix. To address potential endogeneity further, estimations using lagged explanatory variables were conducted.