When firms oppose protecting their own industry

The classic literature on trade policy and protectionism does not predict firms opposing trade barriers on their own products. In the “producers versus consumers” model,Footnote ¹¹ uncompetitive firms and industries are the demanders of trade protection to keep out foreign-made goods. The Ricardo–Viner model instead emphasizes the competition between competitive and uncompetitive industries,Footnote ¹² but again, all firms in comparative disadvantage industries would benefit from trade protection. The literature on firms in trade politics, some long-standing and some much more recent, has a different vision.Footnote ¹³ We focus on three parts: the rise of global production strategies, the utility of these strategies, and firm heterogeneity in the ability to take advantage of these opportunities.

Contemporary firms have developed a rich toolkit of strategies to globalize or “offshore” production of their own products.Footnote ¹⁴ They can produce their products overseas in their own foreign subsidiaries, branches, or plants. Keeping offshoring within the boundaries of the firm limits issues around incomplete contracts and monitoring, but at the expense of greater startup costs.Footnote ¹⁵ As an alternative, firms can directly contract with foreign manufacturers. Though by no means easy, locating, training, hiring, and coordinating with foreign contract manufacturers may be cheaper than purchasing or creating a foreign subsidiary. Of course, this strategy comes with risks of holdup, theft of intellectual property (IP) or trade secrets, and run-of-the-mill misexecution. As a final alternative, some companies may indirectly source their final products overseas, choosing to purchase at arm's length from a wholesaler. This is the least likely of the three paths for manufacturers wanting control over their products’ development and specs, but it does happen.

Each of these strategies requires an opportunity to take advantage of the resources, endowments, productivity, or other comparative advantage of a foreign trade partner. For example, offshoring won't make sense if labor is more expensive, managerial experience is lacking, or supply chains are underdeveloped in foreign markets. Thus, each of these strategies of foreign production are mostly restricted to cases in which a firm's industry is at a comparative disadvantage, globally or in relation to a specific trade partner. It doesn't make sense for a firm to offshore production if it is located in the best market for producing that good, or even a marginally less competitive market. Only a large, fundamental, and multidimensional lack of competitiveness could plausibly motivate firms to pay the significant costs associated with offshoring production.

These costs are key to understanding the final aspect of offshoring production in the firm-centered trade politics literature: firm heterogeneity.Footnote ¹⁶ Moving production offshore, whether to a subsidiary or a foreign manufacturer, has large startup costs. These fixed costs mean that only firms producing at sufficiently large scale find it profitable to move overseas. Thus, offshoring is more likely to be concentrated among large firms. In a similar way, only firms that are sufficiently productive (with high enough margins per unit sold) will find the per-unit costs of offshoring bearable. Thus, offshoring is concentrated among highly productive firms.Footnote ¹⁷

These three pieces—(1) offshoring of final production, (2) by large, productive firms (3) in comparative disadvantage industries—fundamentally change the predictions from the classic literature, that is, that firms in uncompetitive industries should be united in strong support for trade protection. Instead, firms that have offshored final production strongly oppose tariffs in their own industry.Footnote ¹⁸ Those tariffs will raise the landed price of their foreign-made goods, to the advantage of producers that are producing at home or in another foreign market not covered by tariffs. Note, moreover, that the greater the comparative disadvantage of the local industry, the higher the intensity of the preference against tariffs among producers that have gone offshore. This different vision of producers’ preferences is important because it means that industries that otherwise might be united in the search for protection are undermined from within. This makes it harder to secure protection because the industry doesn't speak with one voice.

It is important to contrast this argument with three other ideas about when firms will support trade liberalization, one of which is complementary but distinct and two others that have different mechanisms.

First, support for trade liberalization among big firms also occurs due to the global sourcing of inputs.Footnote ¹⁹ Unlike final products, inputs, parts, and components are clearly upstream of the product a company produces. However, like final production, global sourcing of inputs generate frictions between the larger firms that can exploit opportunities for input sourcing abroad, and smaller firms that can't.Footnote ²⁰ Note that firm support for reducing tariffs on inputs has distinct empirical implications. For example, firms should seek tariff reductions on clearly upstream goods (not their own product) and the intensity of the preferences is stronger where upstream foreign suppliers are more competitive.Footnote ²¹

Second, firms in strongly export-competitive industries facing no import competition might support liberalization of their own industry.Footnote ²² Certainly, they have no motive to oppose liberalization, but they may actively support trade liberalization of their own industry if they view such policy move as part of a strategy to cut tariffs in global negotiations. Whether trade partners would fall for such an empty concession is a different question, but at least it provides an alternative explanation for why exporting firms or industries would support trade. Note that this explanation should be present among firms of all sizes in strongly export-competitive industries and depends on a surrounding context of reciprocity.

Finally, big firms in industries with significant intraindustry trade might support reciprocal liberalization.Footnote ²³ As with offshoring and importing, economists have documented that exporting is often dominated by a small number of large firms.Footnote ²⁴ Amidst intraindustry trade, small firms that don't export will oppose trade and strongly support protection; larger firms may support liberalization of their own home market as the price for gaining access to a foreign market. As in the case described in the preceding text, reciprocity is key to unlocking large, exporting firms’ support for liberalization in industries with intraindustry trade.Footnote ²⁵

The state of the literature

The firm-centered model of trade politics has been tested in two ways. First, industries with certain characteristics are more divided over broadly liberalizing agreements. Industries where products are differentiated or intraindustry trade is high are more likely to have internal divisions over US trade agreements.Footnote ²⁶ Industries with greater related-party trade and input sourcing are also more divided.Footnote ²⁷ Second, scholarship has shown that large firms and firms with foreign subsidiaries are more supportive of trade liberalization.Footnote ²⁸ But again, the focus has been on Preferential Trade Agreements (PTAs) and other broad forms of liberalization.

This literature is valuable but it has several gaps in precisely testing the argument that industrial opposition to trade is hollowed out by firm heterogeneity in the ability to offshore final production. First, comprehensive forms of trade liberalization, like PTAs or GATT rounds, cut tariffs across huge arrays of products. So, firms might come out to support a trade deal because they want to offshore their own final product, because they source inputs from overseas, or because they are exporters. Second, comprehensive forms of trade liberalization, like PTAs or the creation of the World Trade Organization (WTO), have many implications across a wide variety of policy instruments, some of which are barely connected to trade.Footnote ²⁹ If we observe that large firms in uncompetitive industries support trade agreements, this activity is liable to multiple interpretations, including that they are benefiting from the myriad of policy changes involved in contemporary trade agreements. Finally, many PTAs may not make specific and meaningful adjustments to trade barriers because trade is often quite free already under MFN rules. Firms may still come out to support the agreements to lock in tariff cuts or deals, but the stakes may be somewhat low.

So, what do we need to effectively test the idea that offshoring of final production leads firms to oppose tariffs in their own industry? First, we want to unambiguously identify that firms support lowering tariffs on their own final products and are not just motivated by domestic tariff cuts on inputs or foreign tariffs cuts that will improve their export access. Second, we would like to find a setting where firms are clearly motivated by tariffs or other forms of clear trade protection and not by other policies that are included as part of large package negotiations.Footnote ³⁰ Finally, we would like to find a setting where the stakes are high, and a demand for liberalization in one's own industry could seriously impact trade flows. We argue that the exclusion process associated with Section 301 tariffs meets all of these criteria.

The US-China trade dispute and the tariff exclusion process

The United States’ economic relationship with China is complex, interdependent, and contentious. Trade volumes between the United States and China expanded dramatically after China joined the WTO. Trade volumes peaked in 2017, with US imports of goods from China reaching $505 billion, 21.4 percent of total merchandise imports. US exports of goods to China peaked at $130 billion (8.6 percent of goods exports). Despite China's plans for technological catchup, the two economies maintain significant complementaries in their comparative advantages, so that sharp industry-by-industry trade imbalances prevail.Footnote ³¹ As China became the world's manufacturing hub, American businesses took advantage of better production costs and manufacturing efficiencies in China. Many US companies shifted manufacturing to China, contracting for inputs or having final goods assembled and sourced back to the United States.Footnote ³²

China's dominance in world manufacturing and supremacy in global supply chains also became a source of friction in US-China relations.Footnote ³³ Domestically, there have been growing concerns about job losses due to outsourcing and import competition with China.Footnote ³⁴ There have been multiple accusations of China improperly intervening in foreign exchange markets and manipulating its currency to boost exports. State-led industrial programs such as “Made in China” have been criticized as unfair competition due to the use of government subsidies, contested IP practices, and state ownership.Footnote ³⁵

Trade relations with China became particularly salient during the 2016 presidential elections, and dramatic changes were in store once Donald Trump was elected.Footnote ³⁶ Under Trump's direction, the USTR initiated an investigation of China under Section 301 of the Trade Act of 1974. In a 215-page study of China's IP policies, the USTR determined China's practices are “unreasonable or discriminatory, and burden or restrict US commerce.” Beginning in 2018, the Trump administration implemented successive rounds of tariffs.

The initial phase consisted of “List 1” tariffs of 25 percent placed on $34 billion worth of Chinese goods in July 2018. “List 2” tariffs, also at 25 percent, were placed on a further $16 billion worth of Chinese goods in August 2018. “List 3” tariffs of 10 percent (later increased to 25 percent) were placed on $200 billion worth of Chinese imports. Then, “List 4A” tariffs of 15 percent were placed on $120 billion worth of Chinese goods. Plans for “List 4B” tariffs on a final $160 billion worth of Chinese goods were called off and the 4A tariffs were reduced to 7.5 percent in February 2020, as a result of the Phase One trade deal. Considering that the simple average MFN applied tariffs of the United States against the world was 3.4 percent in 2018, these additional tariffs on Chinese imports were substantial in terms of level and breadth.

The scope of the Section 301 tariffs is seen in Figure 1. Each horizontal line in the figure represents one of 1,222 four-digit HTS (“Harmonized Tariff System”) industries. For example, 0409 (in “Animal Products”) is honey while 5905 is textile wall coverings. Each of the four columns represents one of the tariff lists; the first subcolumn in each (colored in light gray and labeled “Product Covered by the List”) represents the share of eight-digit HTS codes that is covered by a tariff within the four-digit industry. For example, no rubber or plastic products were covered by List 1, but subsequent lists (especially List 2) covered most of those products. The figure illustrates the evolution of the trade war, from the advanced manufactured goods and rubber/plastics of the first two lists, to the much broader and eventually comprehensive coverage of Lists 3 and 4A.

Figure 1: Tariff coverage and exclusion requests across industries.

For each new round of tariff adjustments, the USTR solicited comments from stakeholders. The volume of submissions was unprecedented. Producers, mainly protrade firms and associations, raised concerns about harm to their businesses and broader US interests. In their submissions, producers predominantly mentioned disturbance to global supply chains (87 percent) as their primary reason for taking a position on Section 301 tariffs. Import competition from China (7.9 percent), threat of Chinese retaliation on exports (4.9 percent), and Chinese intellectual property and technology transfer practices (2.2 percent) were mentioned only sparingly.Footnote ³⁷

In response, the USTR created a process for interested parties to petition for specific products to be excluded from the Section 301 tariff increases.Footnote ³⁸ Through the Federal Register, the USTR announced the procedures for tariff exclusions and began accepting exclusion requests for List 1 tariffs in July 2018 and List 2 tariffs in September 2018. Lists 1 and 2 exclusion requests were received through Regulations.gov. Later, the USTR developed a separate online portal to accept exclusion requests for Lists 3 and 4A.Footnote ³⁹ Through these platforms, the USTR asked requestors to provide a detailed product description along with the applicable 10-digit HTSUS product category and a rationale for the exclusion.

Figure 1 illustrates the breadth of the response from firms seeking exclusion. Recall that in the figure each line represents a four-digit HTS industry, and each main column describes one of the four lists of tariffs. The middle subcolumn in each of these (colored in medium gray and labeled “Exclusion Request Made on Product”) is filled in according to the share of eight-digit tariffs in each four-digit industry that had at least one request for exclusion filed with the USTR. In other words, the medium gray of the middle subcolumns displays the extent to which firms opposed the Section 301 tariffs. For example, requests for exclusion were submitted for virtually all of the tariffs applied to machinery and appliances industries covered under List 1; but relatively few of the animal and vegetable products covered in List 4A. The figure illustrates that at least one request for exclusion was registered across a wide array of tariffs.

Figure 2 shows the depth of firms’ requests for exclusion. For each list, the distribution of the number of exclusion requests falling under any given eight-digit tariff code is represented in gray. For example, 41 percent of eight-digit tariffs in List 1 had 0 exclusion requests filed; while about 7 percent had 50 or more exclusion requests filed. The coverage of exclusion requests looks strikingly more comprehensive when we weigh exclusion requests by total US imports from China in the eight-digit HTS industry. For example, 20–30 percent of US imports on Lists 1, 2, and 3 had more than 50 exclusion requests.

Figure 2: Depth of requests for exclusions across tariffs and lists.

We conclude by noting that each request ultimately resulted in a ruling by the USTR on whether to exclude the tariff. These rulings were based on a detailed set of criteria and information supplied by the requestors. We leave investigation of the rulings to future research and focus here on firms’ decisions on whether to request exclusion.

Testable theoretical implications

The exclusion request process provides a unique opportunity to precisely test the idea that offshoring firms oppose protection for their own industry. Firms requesting that products be excluded from the Section 301 tariffs have to request a specific 10-digit HTS code for exclusion. We can then trace back whether that code relates to a product that the firm produces or if it is more likely to be an input. Moreover, the exclusion process is clearly focused on one issue: a US tariff being placed on a product. Requests for exclusions therefore focus on a clear and well-defined policy end and support for that exclusion cannot be conflated with other policy issues or other trade barrier reductions. Finally, the tariffs—especially in Lists 1, 2, and 3—were big. A 25 percent tariff is a significant cost, especially on a final product, so the stakes in the tariff war were high.

What do we expect to see if offshoring firms are coming out to support reductions in tariffs in their own industries? Particular types of firms will request exclusions. They should be larger and so able to undertake the significant costs of offshoring production. They should be in industries where the US imports a great deal from China. And some of these firms will own subsidiaries in China. Importantly, we expect these firm-level characteristics to especially drive requests for exclusion among firms whose own product or products are covered by a tariff. This interaction provides the crucial evidence that offshoring firms are seeking to end protection for their own products. To see this, consider if we found an association between owning a Chinese subsidiary and making an exclusion request. That might be consistent with a firm seeking exclusion for their own product but it could also be that the firm offshored input production or just doesn't like the trade war because of the frictions with a key host market. Showing that firms that own subsidiaries and whose final products are covered by a US tariff are especially likely to request exclusion is therefore key.

We operationalize our theory with three testable hypotheses. First, firms that are larger and whose final products are covered by a Section 301 tariff are more likely to file an exclusion request (H1). Second, firms that own subsidiaries in China and whose final products are covered by a Section 301 tariff are more likely to file an exclusion request (H2). Third, firms in industries that face more imports from China and whose final products are covered by a Section 301 tariff are more likely to file an exclusion request (H3). Put another way, if firms offshoring production are requesting tariff relief, any positive association between exclusion-requesting and firm size (or ownership of Chinese subsidiaries or imports from China) should be particularly strong for firms whose final product industry is covered by a tariff.

We highlight three points on these hypotheses. First, each of these hypotheses focuses on the interaction between a firm-level characteristic and whether the firm is covered by a Section 301 tariff for its own final product. Focusing on the interaction terms’ coefficients is therefore critical for testing the hypotheses. For example, we expect that the coefficient on the interaction term between a firm size measure and a Section 301 tariff coverage measure will be positive. Of course, the theory also predicts that the net effect of firm size (and coverage by tariffs) on exclusion-requesting will be positive as well, so we must also be mindful of the size and sign of lower-order terms relative to the interaction term. Second, each of the hypotheses helps to distinguish the idea of firms offshoring final production opposing tariffs from alternative explanations focused on input-sourcing and exporting. For example, input-sourcing firms would have no concern if their final product was covered by a tariff and should not be swayed by own-industry imports, but rather imported inputs. Exporters fearing retaliation should also not be (positively) influenced by imports from China and ownership of subsidiaries should not drive requests for exclusion.Footnote ⁴⁰ We highlight, however, that support for the final product offshoring mechanism does not preclude that other mechanisms are simultaneously active. Finally, and relatedly, it is important that the above predictions hold conditional on firms’ interests in input-sourcing from China and exporting to China, the leading alternative explanations for when firms oppose tariffs. This is because it is plausible that input-sourcing, importing, and exporting might be correlated the reason being they occur in similar industries or among the same firms.

Collection of the exclusion requests data

As described in the preceding text, the USTR established a process by which US stakeholders could request that products be excluded from Section 301 duties as they unfolded across the four lists. Because firms’ requests for exclusion lie at the center of our analysis, we supply technical details and statistics on the lists and the exclusion process in Table 1. Each list began with an announcement in the Federal Register providing notice of the tariffs. The volume, issue, page number, and date in the Federal Register are given in the rows of the table labeled “Notice of Action” and “Notice Date.” Each notice had an associated docket, and stakeholders and the public were offered the opportunity to provide comments on the proposed action at the Federal eRulemaking portal www.regulations.gov. These comments were a lively site of feedback though we do not investigate them in this article.

Table 1: Technical information and statistics on tariff lists and exclusion process

The notices in the Federal Register included the most pertinent information on the tariff actions. The proposed additional ad valorem tariff was listed for each action. For example, Lists 1–2 goods were to be subject to a 25 percent additional tariff on all goods covered under the action, while List 3 goods were initially slated for a 10 percent tariff but then ultimately raised to 25 percent, following a setback in US-China trade negotiations.Footnote ⁴¹ Each notice included an Annex or Annexes setting out the eight-digit HTS codes to be covered by a tariff, and any good falling within one of those codes would be required to pay the additional tariff if imported from China to the United States. We record in the table the number of eight-digit codes for each list and the approximate amount of imports from China covered by the proposed tariffs, as well as the level of the final tariff and its starting date.

Procedures for exclusion requests were set out in subsequent notices. Exclusion requestors were asked to provide their organization name, a 10-digit HTS code and description of the product, and answers to a variety of questions including whether they use the product as a final product or input. These requests were made public under a separate docket (listed in the table as “Excl. req. docket”). The first two lists’ dockets were made available through www.regulations.gov; the second two lists’ dockets were provided at a newly created USTR Comments Portal https://comments.ustr.gov/s/. We scraped the sites to collect all requests for exclusion.

We record the number of exclusion requests available on the websites for each list and describe the number of unique 10-digit HTSUS codes as well as the number of unique organizations (almost all firms) requesting exclusion. As can be seen, the number of requests were significant. Because single companies in some cases filed numerous requests it is also useful to look at the number of HTS codes covered and the number of unique requestors.Footnote ⁴² Looking at the number of unique eight-digit HTS codes filed in exclusion requests in the row “Unique HTS codes (8)” is especially illuminating in comparison to the number of eight-digit HTS codes covered by the tariffs (row “8-digit HTS codes”).

Descriptive findings on requesting exclusion

Before we move to our formal hypothesis testing, we provide some descriptive information on the role of final products offshoring in seeking exclusion. These descriptive results illustrate in various ways the importance of final product producers in opposing tariffs in their own industries.

We first examine what share of the requested exclusions were made by firms producing the good on which they requested exclusion. First consider the exclusion requests for List 1. The docket had 10,814 unique requests for exclusion; out of these, 1,721 of the requests were for tariffs that map back to one of the six-digit industries of the firm making the request. That is an overall rate of 15.9 percent. Using a less stringent approach, 35.0 percent of requests were for tariffs that map to one of the three-digit industries of the firm making the request. Because many products do not involve significant sourcing of parts/inputs from within the three-digit industry, this larger figure may be closer to the truth. The figures are similar for List 2. A total of 13.3 percent and 27.3 percent of requests map to a six- or three-digit NAICS code of the requestor, respectively. For lists 3 and 4A, the equivalent figures were 8.7 percent and 17.3 percent of requests mapping to one of the requestor's six-digit industries, and 16.0 percent and 34.8 percent mapping to a three-digit industry. In other words, for a typical list, at least 9–17 percent of requests came from a firm who almost certainly produce the product that they requested exclusion for, and perhaps as many as 16–35 percent were likely producers of the product.

The exclusion request data afford us a second, independent chance to illustrate the importance of final products. On the exclusion request forms, requestors were asked to report whether the request concerned primarily a “final product” or an “input.” Of course, not all firms stating “final product” are producers of that product (as with wholesalers and retailers). So, we then consider firms that are themselves producers of goods (i.e., in NAICS industries beginning with 11, 21, or 31–33). For example, on List 1, 13.6 percent of requestors report that the firm is requesting exclusion of a “final product” and are producers of goods. It is likely that most of these firms were therefore requesting exclusion of their own product. The same figures for the other dockets are even higher: 29.3 percent, 18.8 percent, and 34.7 percent.

The figures from these two approaches are strikingly similar across the dockets, though one is based on tariff coverage and detailed industry codings from Orbis and the other is based on firms’ self-reports. It seems that for most of the lists, a sizable minority of firms varying between 14–35 percent came to request exclusion for their own products. Thus, requesting exclusion on one's own final product was not the majority of exclusion-requesting, but it was a significant minority of the exclusion requests.

We further illustrate this point by showing the breadth of exclusion-requesting in Figure 1. In the figure, the third subcolumn in each major column (colored in dark gray and labeled “Request by Final Product Producer”) represents the share of tariff codes that had an exclusion request by at least one firm with a six-digit NAICS industry related to that tariff. Thus, the third subcolumns are a subset of the exclusion requests that are represented in the second subcolumns and represent only those requests that were made by firms that produce the product that was covered by tariffs. As can be seen, coverage of the tariffs with exclusion requests from “final producers” is extensive, particularly across the first two dockets and in the industries where it is more plausible that US firms would offshore final production to China. These data therefore again illustrate that exclusion requests by producers of final products were a significant feature of the exclusion request process. Of course, these are not formal hypothesis tests and they neither permit us to explore the mechanisms underlying offshore producers’ opposition to tariffs nor to control for alternative explanations. For these tasks, we must turn to our regression-based tests.

Sample of firms and covariates

To formally test our theory of what leads firms to request exclusion of their product from Section 301 tariffs, we need a representative sample of US firms including firms that never requested an exclusion. We turn to the Orbis dataset supplied by Bureau Van Dijk. From Orbis, we downloaded the names, identifying information, and other covariates of all American firms in goods-producing industries (NAICS codes 11, 21, and 31–33) that Orbis classifies as medium, large, or very large (these are 195,911 firms). We also downloaded a random sample of 100,000 US goods firms that Orbis classifies as small. We use weights in all our regression models to upweight the smaller firms in our subpopulation so that our weighted sample correctly reflects the population of firms in Orbis. Note two further points. First, we focus on goods-producing firms because these are the only firms whose products are covered by tariffs. Of course, wholesalers, retailers, and other services firms might request exclusion of products because they are intermediaries or buy inputs. Second, we include all goods firms that requested an exclusion within our sample; weights are adjusted to account for this. We match exclusion-requesting firms to firms in Orbis by name.

The unit of analysis for our panel data is the firm-list. We represent firms in our sample with the subscript i and represent the lists with the subscript l. Our main outcome is a dichotomous variable equal to 1 if a firm requested 1+ exclusion on a given list. We call this Exclusion requested _il. This is constructed from the exclusion request data described in the previous subsection and using our matching of exclusion-requesting firms to Orbis records.

Orbis provides us a host of additional information on firms. Most important of these are codings of firms’ industries (i.e., codings of the final products that are produced by firms) that are supplied at the six-digit NAICS level. We used these codings to construct a variable called Covered _il, which is equal to 1 if any one of a firm's final products was covered by a particular list of tariffs. To construct this measure, we start with all of the six-digit NAICS industries for a firm supplied by Orbis. We map these six-digit NAICS industries to a corresponding set of eight-digit HTS codes using the 2019 imports concordance from the US Census Bureau.Footnote ⁴³ We then determine if any of those HTS codes was covered by a tariff for a given list l. If so, then we say that a firm was “covered” by the list's tariffs, and we set Covered _il = 1. We examine alternative measures of coverage to show that our findings are robust to alternative approaches.

We also coded a similar variable for whether the inputs used by firms are covered by Section 301 tariffs, which is a continuous measure called Input coverage _il. This measure is proportional to the share of the inputs that a firm uses to make its final products that are covered by a tariff. To construct this measure, we start with all the six-digit NAICS industries for a firm supplied by Orbis and map those back to NAICS input vectors using input-output tables.Footnote ⁴⁴ The input vectors describe the share of a product's value that originates in various inputs described at the six-digit NAICS level. We map the input industries to a corresponding set of eight-digit HTS codes using the 2019 imports concordance from the US Census Bureau, and then determine if any of the HTS codes have a tariff for a given list.Footnote ⁴⁵

In most models we include measures of firm size, foreign subsidiaries/branches, and trade data with China. Orbis supplies roughly coded firm sizes of small, medium, large, and very large. For simplicity of presentation, we recode this into a variable called Large _i that equals 1 if a firm is large or very large. We subset our data to examine only large and very large, or only very large firms, at a few points. Additionally, in the following text we use operating revenues to measure firm size in a subsample of publicly traded firms where revenue data are available. To measure firms’ foreign operations in China, we count the number of China-located subsidiaries owned by a firm according to Orbis. We use a dichotomous version of this variable, which is equal to 1 if a firm has any subsidiaries in China, called China subsidiary _i.

To measure firms’ trade exposure to China, we use data from the US Census Bureau's “NAICS Related Party Database” at the six-digit NAICS level. We examine US trade with China summed across all years from 2010–14 to average over year-to-year variation. We use these data to measure US imports from China; US exports to China; and inputs imported to the United States from China for each NAICS industry, which are calculated using the approach described in Osgood (Reference Osgood2018).Footnote ⁴⁶ For each of these trade flows, we sum the total trade flows across all of a firm's six-digit NAICS industries. We call the resulting variables Imports _i, Exports _i, and Imported inputs _i.

We control for firms’ general profile in global production with dummy variables for whether the firm owns a non-Chinese foreign subsidiary or foreign branch, both constructed using data from Orbis. We include a dummy variable for whether a firm is currently public, which is taken from Orbis. We also include separate intercepts for each list. In addition to these measured covariates, we include in our main models three-digit NAICS industry fixed effects. These fixed effects partial out all unmeasured industry-level factors that might conduce toward filing an exclusion request.

We use in all instances a weighted linear probability model, that is, weighted OLS with a dichotomous outcome. We do so because generalized linear models are biased in the presence of fixed effects due to the incidental parameters problem. Linear models also make interpretation of the interaction terms on which our hypothesis testing relies much more straightforward. Nonlinear models imply interactions even in the absence of interaction terms; linear models do not.

Our three hypotheses focus on the interaction of firm-level characteristics (relating to size, trade exposure, and multinationality) and whether the firm's final product was covered by the tariffs on a particular list. We test these ideas with a linear predictor called F _il because it refers to offshoring of final production:

$$\matrix{ &{F_{il}} = {\beta _1\cdot {\rm Covere}{\rm d}_{il} + \beta _2\cdot {\rm Larg}{\rm e}_i + \beta _3\cdot {\rm Larg}{\rm e}_i\cdot {\rm Covere}{\rm d}_{il} } \; + & {} \cr \qquad {} {} & {\beta _4\cdot {\rm China\;subsidiar}{\rm y}_i + \beta _5\cdot {\rm China\;subsidiar}{\rm y}_i\cdot {\rm Covere}{\rm d}_{il} \; + } & {} \cr {} {} & {\beta _6\cdot \ln {\rm \;Import}{\rm s}_i + \beta _7\cdot \ln {\rm \;Import}{\rm s}_i\cdot {\rm Covere}{\rm d}_{il}} & {} \cr } $$

We expect that interaction term coefficients β ₃, β ₅, and β ₇ will all be positive. If so, the effect of the respective firm-level variable will be larger for a covered firm; and the effect of coverage will be greater among larger, more multinational firms in import-competing industries. Secondarily, we also expect that the lower-order β coefficients numbered 2, 4, and 6, will be positive or, if negative, noticeably smaller in absolute value than their corresponding interaction coefficient, that is, β ₂ + β ₃ > 0, β ₄ + β ₅ > 0, and β ₆ + β ₇ > 0.Footnote ⁴⁷ If these inequalities hold, then the effect of the firm-level feature on exclusion-requesting is positive among covered firms.

It is valuable to show that our main tests are robust to controlling for the two prime alternative explanations for firm opposition to tariffs: the sourcing of inputs from overseas and fears of retaliation among exporters. We operationalize those alternatives in a linear term calledA _il:

$$\matrix{ &{A_{il}} = {\gamma _1\cdot {\rm Input\;coverag}{\rm e}_{il} + \gamma _2\cdot {\rm Larg}{\rm e}_i\cdot {\rm Input\;coverag}{\rm e}_{il} + \gamma _3\cdot {\rm China\;sub} ._i\cdot {\rm Input\;coverag}{\rm e}_{il} } \; + & {} \cr {} {} & {\gamma _4\cdot \ln {\rm \;Imported\;input}{\rm s}_i + \gamma _5\cdot \ln {\rm \;Imported\;input}{\rm s}_i\cdot {\rm Input\;coverag}{\rm e}_{il}} \; + & {} \cr {} {} & {\gamma _6\cdot \ln {\rm Export}{\rm s}_i + \gamma _7\cdot \ln {\rm Export}{\rm s}_i\cdot {\rm Covere}{\rm d}_{il}} & {} \cr } $$

The first two lines of A _il tests the input-sourcing explanation for firms’ opposition to Section 301 tariffs, while the third line of A _il tests the notion that exporters will oppose the tariffs.Footnote ⁴⁸

Finally, we also include a supplemental matrix of controls in all models called X_il:

$$\matrix{ {X_{il}} = {\delta _1\cdot {\rm Foreign\;subsidiar}{\rm y}_i + \delta _2\cdot {\rm Foreign\;branc}{\rm h}_i + \delta _3\cdot {\rm Publicly\;trade}{\rm d}_i + \mu _l + \mu _{{\rm Industry}}} & {} \cr } $$

where μ _Industry refers to three-digit industry fixed effects. Ultimately, our most comprehensive model (which corresponds to model 2 in Table 2) is the following:

$$\matrix{ {{\rm Exclusion\;requeste}{\rm d}_{il}} = {F_{il} + A_{il} + X_{il} + \varepsilon _{il}} & {} \cr } $$

where ɛ _il represents the error term. Note that we suppress the estimates for variables that do not directly relate to our hypotheses to facilitate presentation.

Table 2: Own-industry tariff coverage and firms’ exclusion requests

Note: All models are weighted OLS with weighted OLS standard errors. ***p < 0.001; **p < 0.01; *p < 0.05; ⁺p < 0.1.