ssrn-5502798

Who Wants Geopolitics? Explaining Varying Business Support for Supply Chain Resilience∗ Rodrigo Fagundes Cezar† 18 September 2025 Abstract Which business actors defend geoeconomic measures in trade policy? This paper argues that support for supply-chain resilience measures in response to geopolitical threats depends jointly on sectoral exposure and firm-level integration into global value chains. Drawing on submissions to the U.S. Trade Representative’s 2024 consultation on supply-chain resilience, I test my hypotheses via a combination of qualitative and quantitative text analysis using large language models (LLMs). The results show that firms and associations invoke geopolitics when they stand to benefit from sanctions and inducements. Individual firms adopt sharper geopolitical framings than associations. Cross-sectoral associations, in turn, display meaningful cleavages that complement the expectations of the “new” new trade theory (NNTT). Overall, the findings demonstrate that sectoral cleavages re-emerge as a key variable in the geopolitics of trade, occasionally producing unexpected business alliances. They also underscore both the potential and the limitations of LLMs for scaling business positions and preferences. Keywords: Large Language Models. Trade Preferences. Business Lobbying. Bradley-Terry Model. ∗I would like to thank Pedro Rolim, Paula Corradi Rabello and Christian Brand ˜ao for their excellent research assistance. I also thank the participants of the workshop “Geopoliticization and the Changing International Political Economy of Global Value Chains” (Trento, 2025), specially Jappe Eckhardt, Arlo Poletti and Ferdi De Ville. I acknowledge financial support from FAPESP under project number 2024/01500-1. †Fundac¸˜ao Getulio Vargas, Brazil rodrigo.cezar@fgv.br, https://www.rfcezar.com/. 1 -- 1 of 49 -- Introduction There is a surge of works examining recent changes in the strategic behavior of countries in a geopo- litically fractured world, with a few key papers focusing on the European Union (Bauerle Danzman and Meunier 2023; Meunier and Nicolaidis 2019; Fiott 2024). Fewer works address business interests and their responses to the geopolitics of trade (D ¨ur et al. 2025; Eckert 2024; Aiyar et al. 2024; Doppen et al. 2024), although research in the field of business administration has examined how and why multinational corporations (MNCs) adapt their strategies to new geopolitical realities (Lubinski and Wadhwani 2020; Moura et al. 2025; Mavreti ´c and Vangeli 2024; Mol et al. 2023). Since business interests are likely to remain influential even in a geopolitical context (D ¨ur et al. 2025; Eldes et al. 2025b), understanding how they react to the geopolitics of trade is highly relevant. This paper takes up that task and seeks to explain variation in the use of geopolitics to support business lobbying for supply-chain resilience. To what extent are existing theories sufficient to account for firms’ responses to geopolitics? Geopolitics alters the incentives for firms to lobby and create new coalitions (McNamara 2024) that may differ from those typically expected by the “new” new trade theory (Madeira 2016). Will the traditional pro-trade coalition composed of international firms and cross-sectoral organizations (Osgood 2021) resist geopolitical measures (Lee and Osgood 2021)? What business coalitions will arise, and who will be the new allies? What are the implications for lobbying through trade associations? This paper’s key argument is straightforward: firms position themselves according to an economic evaluation in which sectoral and firm-level characteristics are both relevant at the same time. The resulting level of support for geoeconomic measures reflects firms’ strategic adaptation to their context (Nicoll Victor 2007). Geopolitics is therefore akin to what social movement theory calls a political opportunity structure (Van Der Heijden 1997), as it helps firms that stand to benefit most from certain measures to mobilize and improve their chances of securing favorable outcomes. Using Bauerle Danzman and Meunier (2024)’s categorization of offensive and defensive sanctions and inducements, I argue that strategic sectors and less integrated firms should accrue concentrated benefits from geoeconomic measures and are therefore expected to frame their positions on supply- chain resilience more strongly in terms of geopolitical rivalry. Sanctions include the Section 301 tariff, designed to investigate and impose trade measures against foreign countries engaged in unfair practices that harm U.S. producers. Inducements include the CHIPS and Science Act, designed to boost 2 -- 2 of 49 -- domestic semiconductor manufacturing and advanced technology R&D. Although multinationals are generally expected to be the least favorable toward geoeconomic policies due to fears of supply-chain distortion (Lee and Osgood 2021), I expect them to occupy a middle ground between those strongly opposed and those strongly supportive, reflecting the balance of costs and benefits they derive from sanctions and inducements. As Robert Gilpin argued (Gilpin 1976), MNCs and the U.S. government will thus share some degree of converging interests. At the bottom of the distribution, one should expect “strange bedfellows” that benefit neither from sanctions nor inducements, such as providers of logistical services, some agricultural organizations, international firms primarily exporting to the U.S., and retailers. At the top should be U.S.-based manufacturing firms and mining companies. I also expect geopolitics to have implications for association lobbying. Trade associations are likely to adopt more aggregated and moderate stances, while individual firms may articulate sharper positions. Moreover, because both sectoral and firm-level cleavages matter in the geopolitics of trade, cross-sectoral associations should not fully behave as expected by NNTT. According to NNTT, it should be easier for cross-sectoral associations that represent multinationals (e.g., the Retail Industry Leaders Association, Business Roundtable, and National Foreign Trade Council) to reach a common position (Cezar 2023). In the geopolitics of trade, however, instead of unified positions, I would expect some cleavage. For instance, in the US, I would expect IBM to generally have similar positions as the Business Roundtable. I show empirically that while this held before 2020, it no longer did in 2024. The implications are twofold: cross-sectoral associations that would otherwise behave similarly will behave differently; and firms that are members of such organizations will sometimes prefer to submit individual positions in complement to, or in substitution for, their associations’ positions. I test these arguments qualitatively and quantitatively using human coding and large language models (LLMs). I rely on position papers submitted by business interests to the United States Trade Representative (USTR) consultation on “Promoting Supply Chain Resilience,” held in March 2024. I then evaluate their absolute and relative positions according to the strength of a geopolitical framing in support of supply-chain resilience tools. Qualitative analysis indicates that firms across the board often use geopolitical language, adapting to the strategic context in which they operate. However, there are clear differences in the strength of the geopolitical framing across business groups. That variation aligns well with my initial expectations. In terms of coalition lobbying, however, the level of cleavage in cross-sectoral associations is somewhat lower than I anticipated, though still present. 3 -- 3 of 49 -- The results indicate that the hierarchy between sectoral and firm-level characteristics expected by NNTT is weakened in geopolitical contexts. In other words, geopolitics brings sectors back into the study of trade politics. These results also contribute to understanding the shifting trade coalitions (McNamara 2024) and how interest groups lobby on trade issues (Hanegraaff et al. 2024, 2023). Finally, the paper presents some of the potential of LLMs for studying business position-taking. The remainder of this paper is organized as follows. First, I review the literature on the geopolitics of trade and its effects on business strategies, drawing on both international business and political science scholarship. Second, I outline the theoretical framework and hypotheses, followed by the materials and methods. I then present a qualitative overview of the findings, before turning to the quantitative LLM analysis, cross-validated using human coding. I also report a series of robustness tests and discuss some apparently deviant results of the text analysis. Finally, I conclude with a discussion of the implications of the findings and avenues for future research. Literature Review The geopolitical turn in trade policy points to a new paradigm in which geopolitical tensions and power politics become the guiding factors behind state behavior (Fiott 2024), motivated by the pursuit of relative gains (Gilpin 1975; Haroche 2024). This literature has developed particularly quickly in documenting and explaining the EU’s geopolitical turn (McNamara 2024; De Ville et al. 2023; Eckert 2024; Bauerle Danzman and Meunier 2024; Meunier and Nicolaidis 2019; Eliasson and Garcia-Duran 2023). Since state power and corporations are deeply intertwined (Babic et al. 2017; Gilpin 1976; Fichtner 2017) and mutually dependent, recent scholarship has also examined the connection between the geopolitics of trade and firm behavior. Existing works can be grouped into three main strands: (1) those showing evidence that geopolitics affects business strategies (Mol et al. 2023; Moura et al. 2025); (2) those categorizing business responses to geopolitics (Doppen et al. 2024; D ¨ur et al. 2025); and (3) those attempting to explain the sources of variation in these responses (Eldes et al. 2025a). Yet the literature still has limitations. For instance, the international business field is strongly focused on MNC behavior. Besides, political science has so far fallen short of systematically explaining and measuring variation across firms, despite some initial efforts. Geopolitics influences firm behavior in multiple but nuanced ways. Aiyar et al. (2024) show that geopolitical pressures shape the geographical footprint of bilateral investments. Geopolitics 4 -- 4 of 49 -- also affects firms’ decisions to engage in friend-shoring depending on the diversity of their value chains (Eldes et al. 2025a). Scholars have also investigated how state–firm interactions may influence geopolitical responses (Doppen et al. 2024), and how trust relations and firms’ political capabilities (e.g., political connections and networks) condition their strategic responses to geopolitical pressures (Moura et al. 2025; Sallai et al. 2024). However, despite widespread claims about “de-globalization,” MNCs are not retreating globally but are instead hedging (Linsi and Gristwood 2024; K ¨oncke and Graaff 2024; Babic and Linsi 2025). Thus, global value chains (GVCs) remain a central factor in shaping responses to the geopolitics of trade (Curran and Eckhardt 2023), sometimes leading firms to oppose geopolitical measures (Kinderman 2021; Lee and Osgood 2021). Lubinski and Wadhwani (2020) further show that MNCs often engage in geopolitical jockeying to delegitimize competitors, instrumentalizing geopolitics. Firms may also use geopolitics to gain access to subsidies, and multinationals may shift their discourse to present themselves as geopolitically responsible actors (Cui et al. 2023; Mavreti ´c and Vangeli 2024; Mol et al. 2023). The literature has also begun categorizing business responses to the geopolitics of trade. Choer Moraes and Wigell (2022) propose a threefold typology: “business as usual,” which involves resisting politicization and maintaining pre-existing operations; “one company, two systems,” whereby firms adapt business models to operate across opposing strategic blocs; and “patriotic capitalism,” which entails aligning with state geoeconomic policies to secure economic and strategic benefits. Eckert (2024) show that business actors have rationales to lobby against, to limit, or to support geoeconomic measures. In other words, there is variation in interest group responses to geopolitics. Herranz-Surrall ´es et al. (2024), in turn, argues that the depth of geopoliticization varies with ideas, interests, and institutions, leading to superficial, reluctant, or deep geopoliticization. Cui et al. (2023) combine risk management and institutional characteristics of MNCs to create a two-by-two typology of firm responses. D ¨ur et al. (2025) focus on the concentrated costs and benefits of restrictive, conflictive, and promotional instruments. But what exactly explains variation in firms’ response to the geopolitics of trade? Explanations vary across perceptions of the costs and benefits of specific instruments (D ¨ur et al. 2025), firms’ institutional characteristics (Cui et al. 2023), firm–state connections (Doppen et al. 2024; Moura et al. 2025), the availability of allied markets and supply-chain diversification (Eldes et al. 2025a), and supply-chain linkages with rivals (Lee and Osgood 2021). Yet we still fall short of fully explaining and 5 -- 5 of 49 -- testing variation in business responses to the geopolitics of trade. First, the international business literature is particularly focused on MNCs, whereas the objective here is to account for other types of business actors as well. Second, existing works overlook differences between how individual firms and trade associations position themselves in geopolitical contexts. Third, firm-level supply-chain characteristics alone cannot fully account for firm behavior, as I will demonstrate throughout this paper. Fourth, the literature has not yet provided an analysis of the relative positions of business groups on the geopolitics of trade. Still, knowing relative positions is particularly relevant because it allows us to identify which actors are close enough to coordinate, which are positioned as outliers, and where the fault lines within or across associations are most likely to appear. Theoretical Argument Firms may support or oppose geopolitical and geoeconomic measures in response to cost calculations (Eckert 2024; Curran and Eckhardt 2023; Lee and Osgood 2021), and may adopt different responses to the geopolitics of trade (Herranz-Surrall ´es et al. 2024; Eckert 2024; Choer Moraes and Wigell 2022). I here build upon existing classifications and seek to explain what accounts for the variation therein. First and foremost, I expect firms not to detract from their economic preferences (D ¨ur et al. 2025). On the other hand, however, firms seek to fully align their economic goals with the geopolitical challenges posed so as not to go against the government agenda (Lubinski and Wadhwani 2020), therefore adapting their lobbying efforts to the strategic context in which they are inserted (Nicoll Victor 2007). Cezar (2022b), for instance, shows that U.S. business interests have adopted intricate strategies to strike a balance between their own economic interests and the shifting U.S. context with regard to non-trade issues in trade agreements. Those more superficial changes preceded deeper changes in positions and strategies that came later on (Cezar 2022a). Figure 1 illustrates the variation in geopolitical emphasis in the position of business groups. Business-as-usual (Moura et al. 2025) indicates that firms and associations keep an efficiency framing in response to geoeconomic measures. In a partially aligned framing, it is in the firms’ interests to push for a geopolitical framing, although they can also reap concentrated costs from certain geopolitical actions (Lee and Osgood 2021; Eldes et al. 2025a), so they alternate between a geopolitics and an efficiency framing. In some cases, geopolitics appears as a central framing — even trumping efficiency considerations at first — but geopolitical concerns are fully aligned with what one would expect 6 -- 6 of 49 -- from pure commercial self-interest, making it difficult to disentangle both (e.g., firms competing against imports will frame China as a security threat when their position might as well be driven by purely economic motivations). In a context of strategic lobbying, I would expect many firms to try and reach such a level of alignment to at the same time leverage geopolitical threats to advance objectives that are observationally indistinguishable from pure economic preferences. Finally, a geopolitics-first frame prioritizes geopolitics over cost-efficiency. The closest to business-as-usual, the fewer references to geopolitics and the closer to a geopolitics first frame, the greater such references. BUSINESS- AS-USUAL Maintains traditional commercial positions, with no geopolitical language. Focus on free trade, efficiency, and predictability, or on protecting jobs and leveling the playing field, with few or no security reasons mentioned. PARTIALLY ALIGNED FRAMING Adds selective geopolitical references, but reverts to efficiency and cost arguments when geopolitical and economic objectives clash. FULLY ALIGNED FRAMING Integrates geopolitics as a central argument, but aligned with immediate commercial self-interest, in a way that is hard to disentangle both. GEOPOLITICS FIRST Prioritizes geopolitical objectives over commercial logic, accepting higher costs or market losses for strategic aims. Figure 1: Illustrative geopolitical frames But what explains variation in the use of geopolitical frames? I will now flesh out the distributive effects of geopolitics on domestic interests (D ¨ur et al. 2025). I focus on the categories created by Bauerle Danzman and Meunier (2024). Defensive sanctions shield economies from foreign leverage through investment screening, export controls, trade remedies, and anti-coercion tools, while offensive sanctions project power outward via export controls on chokepoint technologies, outbound investment restrictions, and reciprocity instruments. Defensive inducements promote resilience and diversification, such as industrial policies and mechanisms like the EU carbon border adjustment mechanism (CBAM), whereas offensive inducements seek dominance through subsidies and strategic industrial policy. I would not expect firms to engage in purely offensive sanctions and will instead privilege defensive measures and inducements. That is because export controls, for instance, create concentrated benefits for a few industries and indirect benefits for downstream producers (D ¨ur et al. 2025) who might, instead, prefer to secure a diversified supply rather than depend on domestic sourcing. Instead, I would mostly expect support for defensive sanctions and 7 -- 7 of 49 -- inducements, support which is defined by the balance of concentrated costs and benefits reaped by firms, which is in turn explained by those firms’ sectoral and supply-chain characteristics. Governments will offer inducements and sanctions to promote and protect certain sectors. They will “delineate small areas of the domestic economy that have substantial security implications and then prevent foreign actors from gaining access to these areas” (Bauerle Danzman and Meunier 2024). Such an approach is often referred to as “small yard, high walls” and comprises areas such as critical minerals, artificial intelligence, telecommunications, defense industry, and semiconductors. Producers in such sectors would thus reap concentrated benefits from sanctions and inducements coming their way. This should, in principle, encompass even multinational corporations in the high-technology manufacturing sector, as long as they have operations in their home country. For instance, the global manufacturer Intel operates large fabrication plants in Chandler (Arizona), Hillsboro (Oregon), and Rio Rancho (New Mexico), with two additional plants under construction in Ohio, all central to its wafer production and packaging capacity. In such contexts, there is a “close identity of interest between the multinational corporations and their home governments” (Gilpin 1976), whereby the former seek to assure relative gains and improve national security and the latter aim to gain competitiveness. However, the literature has shown that multinationals often reject geoeconomic measures due to their supply-chain linkages with political rivals (Eldes et al. 2025a; Lee and Osgood 2021). I therefore expect sectoral considerations to be moderated by firms’ level of supply-chain integration (Curran and Eckhardt 2023). For instance, certain domestic firms that compete more than they depend on foreign imports may support both defensive sanctions and inducements. As they want both, they have incentives to strategically defend the urgency of geopolitical measures as a justification for supply- chain resilience, resulting in a stronger geopolitical framing. On the other hand, more international firms will oppose any action that involves raising tariffs against foreign producers, especially in Asia, as well as local content rules (Eckhardt and Lee 2018). In such a context, international firms in “small yard” sectors will want inducements but will reject sanctions, leading to fewer geopolitical references or an admixture of geopolitical and efficiency frames (Figure 1). The expectations above complement the “new” new trade theory (NNTT). Whereas firm-level characteristics should be the defining feature of trade politics, strong and focused government incentives (“industrial policy”) bring sectoral characteristics back to the forefront of trade policy. 8 -- 8 of 49 -- This leads me to the following hypotheses, whereby H1.1 accounts for the sectors that should benefit the most from sanctions and inducements and H1.2 for the ones that should lose the most from them (Table 1): Hypothesis 1.1 Firms and associations representing firms in “small yard” sectors and with lower supply-chain integration will be most supportive of justifying supply-chain resilience using geopolitics. Hypothesis 1.2 Firms and associations representing firms outside of “small yard” sectors and with greater supply-chain integration will be least supportive of justifying supply- chain resilience using geopolitics. Benefits (inducements) Loses (inducements) Benefits (sanctions) Strongly supportive Somewhere in between Loses (sanctions) Somewhere in between Not supportive Table 1: Levels of support for a geopolitical frame to supply-chain resilience Which exactly are those firms and associations? The agricultural sector, some service providers, and international firms exporting to the US should neither accrue concentrated benefits from geoeconomic sanctions nor inducements, whether offensive or defensive. For one, the agricultural sector is oftentimes not directly benefited by geoeconomic measures, which generally follow the “small yard, high walls” formula. In the recent geopoliticization wave, the targeted sectors have been industry, high-technology, and critical materials, not so much agriculture. In turn, some groups, like the National Cattlemen’s Beef Association (NCBA), have raised concerns about the potential impacts of certain geoeconomic measures on agriculture. The NCBA cautions against using U.S. trade policies as a platform to erect climate-related non-tariff trade barriers, such as those seen in the EU’s “subjective model,” – as NCBA calls it – namely the EU Deforestation Regulation (EUDR). Moreover, many agricultural organizations in the US and beyond rely on machinery and fertilizers imported from other countries, which could be especially sensitive to government restrictions on imports from rivals. These short- and medium-term cost evaluations should not compensate for the longer-term expectation of more reliable US suppliers. Domestic telecommunication providers can largely benefit from the US imposing bans on international providers such as Huawei and also benefit from inducements (Bauerle Danzman 9 -- 9 of 49 -- and Meunier 2024). However, the services sector is heavily globalized and depends on regulatory coordination across borders to operate efficiently. Although telecommunications is at the forefront of certain geoeconomic disputes, services related to finance, transportation, and logistics are not the main sectors to benefit from US inducements. Disruptions caused by geoeconomic measures (such as new regulations, tariffs, or supply-chain constraints) could create regulatory mismatches or complicate cross-border service provision, adding costs and reducing efficiency. Given the high global integration of services, firms are likely to resist policies that create uncertainty or regulatory fragmentation, as they would undermine their ability to serve international clients or maintain global operations. Besides them, firms that export to the U.S. — regardless of the sector — are highly dependent on the openness of trade channels. Geoeconomic measures that impose trade barriers or disrupt global supply-chains would directly impact their ability to meet U.S. demand. These firms can overcome difficulties by installing upstream or downstream operations in the U.S., thus the hypothesis is mostly relevant for those interests that have not yet done so. On the flipside, the mining sector and manufacturing firms and associations relying more heavily on U.S.-based supply-chains should benefit both from sanctions and from inducements. For mining companies with upstream activities in the U.S., competition against exports from China and Russia aligns with the hub of U.S. geopolitical tensions, making a geopolitical framing around defensive sanctions and inducements highly advantageous, as it reinforces domestic production and strengthens the U.S.’ strategic autonomy in critical industries. By the same token, these sectors strongly benefit from the CHIPS Act, the Inflation Reduction Act (IRA), the Bipartisan Infrastructure Law, and other incentives for supply-chain resilience. The U.S.-based manufacturing sector, including steel, aluminum, and automobile producers, should also benefit from “Buy American” incentives, or see a clear opportunity to demand increased domestic content requirements to expand their market share. In the case of the transformation industry, firms compete against final products from abroad and are particularly sensitive to supply-chain risks, even when they would otherwise profit from cheaper intermediate imports. They should therefore support inducements and sanctions to limit cheaper imports of products that compete against theirs and to increase resilience against disruption. As indicated above, international manufacturers like Intel and IBM are highly integrated into global supply-chains, but also maintain strong ties to the U.S. As such, they occupy a middle position in their support for geoeconomic measures aimed at supply-chain resilience. On the one hand, 10 -- 10 of 49 -- they could be negatively impacted by sanctions. On the other hand, these companies might support geoeconomic policies when they are offered, for example, cheaper access to credit and other incentives. This positions such firms somewhere between the extremes of H1 (least supportive) and H2 (most supportive). Figure 2 summarizes the logic of the theoretical framework. Strongest geopolitical frame Strong dependence on specific foreign suppliers; competition against im- ports and dependence on exports; calls for subsidies, protective mea- sures, or tariff relief to assure domestic production and autonomy. Typical examples: Domestic or export-oriented manufacturers and mining producers, specially upstream mining firms. Instrument of choice: defensive sanctions and inducements Moderate geopolitical frame (“Somewhere in between”) Preference for maintaining status quo in open trade, but willing to accept targeted incentives to incentivize stronger domestic production; avoid GVC disruption and incentivize diversification. Typical examples: Highly-integrated firms in the manufacturing sector, with downstream operations in the US. Instrument of choice: inducements but no defensive sanctions Weakest geopolitical frame Emphasis on expanding free trade agreements (FTAs) without calls for protection, although import-competing interests might support specific restrictions; incentives for domestic manufacturing incentives can lead to problems such as distortions, retaliation risks, or higher costs. Typical examples: agricultural exporters and importers. Instru- ment of choice: neither defensive sanctions nor inducements Figure 2: Firms positions on trade, incentives, and protection across sectors. The conditions above illustrate when we could expect – informal – coalitions among some strange bedfellows, such as agricultural producers and logistics service providers (e.g. US Meat Institute and FedEx or UPS); between multinational mining companies and local producers of medical devices (e.g. Rio Tinto and the American Medical Manufacturers Association). But besides these informal connections, what are the implications for lobbying via associations? As expected by NNTT, there should be some sort of cleavage within associations given intra- industry differences in firm-level characteristics such as size and GVC integration. Building upon that, my expectations are twofold. First, I generally expect that trade associations will reflect aggregated positions, resulting in more moderated or conciliatory demands, while individual companies will position themselves more assertively (Cezar 2023). That assertive position could be either in favor of a geopolitical framing or against it. As part of a strategic lobbying behavior, however, I would 11 -- 11 of 49 -- expect firms to capitalize on the government’s concerns around geopolitics and to actively avoid blatantly opposing them. Therefore, if the context indicates to firms that geopolitics is key to the government, opponents should do one of two things: (1) not weigh in on the issue of geoeconomic measures if their opposition is not particularly strong; or (2) “hide” by lobbying via associations, further diluting the average position of such associations around geoeconomic measures. Second, cross-sectoral associations of firms with similar firm-level characteristics (e.g., the Retail Industry Leaders Association, Business Roundtable, and National Foreign Trade Council) should generally adopt similar positions on trade both among themselves and vis- `a-vis their members (Cezar 2023). Yet, in the geopolitics of trade, because both sectoral factors and supply-chain integration matter, even these associations are expected to exhibit some degree of cleavage stemming from the diverse sectors of their members, independently of whether such members are only MNCs, for instance. This leads to new expectations regarding when firms will lobby alone or as part of their associations (Hanegraaff et al. 2023, 2024). In sum: H2.1. On average, individual firms will be more supportive of justifying supply-chain resilience using geopolitics than associations. H2.2. Cross-sectoral associations representing multinationals will have distinct positions on geopolitics and supply-chain resilience vis- `a-vis each other and vis- `a-vis their members, leading to cleavage. Materials and Methods This paper relies on business position papers submitted during the USTR consultation on “Supply Chain Resilience,” held in early 2024 (Docket USTR-2024-0002). I scraped the Regulations.gov website to collect only submissions with attachments, resulting in 264 files. In the same call, USTR also invited expressions of interest in participating in a U.S. Congress hearing, to which many interest groups responded with a separate file submission. I excluded those, as they contained no substantive content on policy positions, only expressions of willingness to testify. I also excluded non-business actors such as labor unions, NGOs, and think tanks, keeping only firms and business associations. This yielded a final sample of 169 submissions. For H2, I further classified the organizations as either individual firms or trade associations. The USTR consultation on “Supply Chain Resilience” was 12 -- 12 of 49 -- chosen for three reasons. First, the United States is a hub for the diffusion of trade norms (Baccini et al. 2015) and an influential case in trade politics (Seawright and Gerring 2008). Second, given the logic of strategic lobbying (Nicoll Victor 2007; Lubinski and Wadhwani 2020), a consultation explicitly framed in geopolitical terms should push firms either to conform to that framing or to abstain, making it a hard case to uncover variation in geopolitical stances—except in specific circumstances (Lee and Osgood 2021)—which is precisely the focus of this paper. Third, USTR provided specific questions that all submissions were asked to address, reducing the risk that observed variation reflects format or drafting style rather than substantive differences in position. The consultation took place in 2024, thus before Trump became president. On the one hand, that timing is likely to somewhat limit the external validity of the findings, given the more recent shifts in U.S. trade politics triggered by Trump. On the other hand, by March 2024 geopolitics and supply-chain resilience were already deeply enmeshed in U.S. governmental discourse and policy action. For that reason, I expect the results to be relevant beyond the Biden administration. Another potential limitation is that participation in consultations is not mandatory; therefore, firms self-select into them. However, rather than producing a systematic bias that could threaten validity, submissions to U.S. consultations should be seen as a scope condition, or “circumstances under which the relationship expressed in a hypothesis is expected to hold true” (Foschi 1997).1 The U.S. has a well-known and easily accessible system for submitting consultation responses through the regulations.gov website. In addition, the consultation call remains open for a considerable period of time (three months in the case of the consultation serving as the source for this paper). Drafting a position paper requires relatively few resources and is therefore accessible to interest groups of different sizes and capacities. As a result, interest groups that do not submit positions in a U.S. trade consultation are here assumed either not to have sufficient interest in the topic or to participate indirectly via their business associations. With the final list of organizations at hand, I categorized them in three blocks, those expected to be gain the most from sanctions and inducements, those expected lose the most for sanctions and inducements, and those somewhere in between (e.g. gain from inducements but lose from sanctions). That classification was done qualitatively, based on the submissions by the business interests and on 1Only around 10% of the universe of business groups (firms and associations) that submitted positions during USTR calls on U.S. trade agreements also submitted papers during the supply-chain resilience call. In turn, around 51% of the organizations that submitted position papers during the supply-chain resilience call had never submitted on regulations.gov during other USTR calls related to trade negotiations. 13 -- 13 of 49 -- their publicly available data. That classification process was twofold. First, I checked if the firms or associations operate in one of the following sectors: mining of critical minerals and rare earths, steel and titanium manufacturing, semiconductors, electric vehicles, medical technology, solar panels and energy, specially renewable energy and national defense. Then, I evaluated the extent to which these groups are domestic-based and import-competing, on the one hand, or integrated to value chains and export-oriented on the other. The final list of organizations is available in the appendix. After doing the above, I qualitatively examined each position paper to identify two separate frames — geopolitical and efficiency framings of supply-chain resilience — as well as the general positions on offensive and defensive sanctions (Section 301, Section 302, export controls, extraterritorial sanctions, etc.) and on offensive and defensive inducements (CHIPS Act, Inflation Reduction Act, Bipartisan Infrastructure Act, etc.). A geopolitical framing included references to “China risk,” “strategic autonomy,” “friend-shoring,” “supply-chain resilience,” “critical minerals as security assets,” and “deterring adversaries.” An efficiency framing included references to “competitiveness,” “predictability,” “level playing field,” “avoiding supply disruptions,” and “minimizing adjustment costs.” In addition to this first layer of analysis, I also relied on human coding to assess the extent to which geopolitics was used as a justification for supply-chain adjustment and resilience. The qualitative coding was performed by a human research assistant, who was instructed to rank the submissions on a five-point scale according to the emphasis placed on geopolitics as a justification for supply-chain adjustment. The research assistant was provided with examples of submissions that would typically be ranked lowest, highest, and intermediate. I then used OpenAI’s Generative Pre-Trained Transformer (GPT) to examine the submissions. I did so using a relative ranking technique following Cezar (2025), whereby dyads of submissions are chosen by GPT based on which one is more favorable of geopolitics. This technique has multiple benefits over absolute point-wise ranking. Ordinal rankings fail to convey the magnitude of difference between items, while direct scoring does not guarantee comparability across cases and demands careful calibration, which is often difficult (Qin et al. 2024). That is not ideal, since the hypotheses H1-H2 require understanding relative business positions (one in comparison to the other). The prompt guiding the LLM relative ranking was: “which organization is more supportive of adjusting supply-chains to reduce dependency on adversarial countries?”. That question is aimed at prompting GPT to answer based specifically on the interest groups’ framing of supply-chain resilience as a 14 -- 14 of 49 -- geopolitical issue, as opposed to only an efficiency issue. Besides, it considers geopolitics as ”synonym for power-driven political rivalry and conflict” (Hurrell 2024). The LLM prompt is applied to submission dyads using GPT-4.1-mini. The results are then analyzed using a Bayesian Bradley-Terry (BT) model, leading to fully comparable business positions. An obvious trade-off is that, in the BT model, the number of comparisons increases quadratically, making full scale human validation particularly hard. Still, dyadic comparisons are widely used in political science and are particularly well-suited for LLMs, as they reduce cognitive load by simplifying the task (Wu et al. 2023; Leo et al. 2025; Qin et al. 2024). Besides, works have shown that LLMs and the BT model deliver solid results. LLM use is particularly advanced in ideological scaling (Mens and Gallego 2025; Leo et al. 2025; O’Hagan and Schein 2024; Kato et al. 2024; Wu et al. 2023; Kim et al. 2025; Cezar 2025). Yet another strength of using BT model is that one can apply Bayesian inference to create credibility intervals (Cezar 2025; Leo et al. 2025), which should make the analysis more robust by quantifying uncertainty around estimated positions. To increase confidence in the findings, a human research assistant was tasked with assessing 200 random paired comparisons, which were contrasted against the results of GPT’s assessment.2 As indicated above, the human research assistant also performed a qualitative 1-5 scaling of all submissions which I used as robustness. With both the list of organizations according to whether they should be expected to gain the most or the least from sanctions and inducements, as well as the BTM scores, I run ANOVA and tukey-tests, along with regressions to check if the results align with the hypotheses. I did as indicated in Figure 3. I first regressed the category of the organizations, defined ex ante according to their sectoral and supply chain characteristics against their explicit support or opposition for sanctions and inducements. A human research assistant did so by categorizing each submission into whether they explicitly supported/opposed “Section 301” or “Section 232” (sanctions) and “Inflation Reduction Act”, “CHIPS and Science Act”, “Defense Production Act”, “Opportunity to Invest Act” and “Bipartisan Infrastructure Law” (inducements). Absence of explicit references to those instruments in specific was coded as soft refusal (Findley et al. 2015) to promote sanctions and inducements. I used multinomial logit for that task. I then regressed the categories of support and opposition above against the BTM scores using weighted least squares (WLS). To test H2.1, I used t-tests and WLS to regress BTM scores against whether the submission is from a firm or association. Since large language models can introduce bias in pairwise comparisons by relying on superficial 2To be done. 15 -- 15 of 49 -- Firm and sector-level characteristics Support for sanctions and inducements Geopolitical framing Figure 3: Testing Hypotheses 1.1 and 1.2 attributes (Jeong et al. 2025; Tripathi et al. 2025), I control for whether comment length could be one such factor. In addition, I use the credibility intervals from the Bayesian BT model as weights to test whether the results are sensitive to estimation uncertainty. The standard errors are clustered by the 2-digit NAICS, in order to account for a key assumption of the BT model: independence of observations. This assumption means that the outcome of a comparison between organizations A and B should not be influenced by the presence of other comparisons involving A, B, or any third organization. Violations can arise, for example, when multiple comparisons are derived from the same underlying text, or when organizations are institutionally linked (e.g., a firm and an association it belongs to). For H2.2, I used direct comparisons between estimates to assess the plausibility of the hypothesis. I also compare the relative positions (calculated using the BT model) of business interests in USTR’s “Supply Chain Resilience” consultation (March 2024) to the positions of business interests that supported or opposed the protection of foreign investments during the negotiation of U.S. trade agreements between 2007 and 2020. By comparing these two settings, I can illustrate the results in an experiment-like fashion, with estimates before and after the “treatment” (the U.S. government’s response to COVID-19 and the Russia–Ukraine war) and across sectors that should respond more strongly or more weakly to the geopolitics of trade. Finally, I performed a series of additional robustness tests to increase confidence in the findings. This includes performing a product-of-experts (PoE) assessment (Cezar 2025; Liusie et al. 2024) and testing for sensitivity of the results to superficial attributes (Jeong et al. 2025; Tripathi et al. 2025). I also explored a few meaningful cases ex post in which differences were initially puzzling but could be explained by the theory, as well as cases that 16 -- 16 of 49 -- remained unexplained and therefore had to be acknowledged. Who Wants Geopolitics? An Overview of the Positions An overview of business positions on supply-chain resilience indicates three clearly distinguishable ways of invoking geopolitics as a justification. At one end are submissions where geopolitics are at the forefront; in the middle are those where economic efficiency shares space with security considerations, though to varying degrees; and, finally, those where security arguments downplayed and at times explicitly opposed in favor of economic justifications (e.g., to avoid job losses or loss of competitiveness). Overall, these positions align with the expectations of the theoretical framework, as outlined below. Both Inducements and Sanctions I expect that the greater the economic benefits a group secures from inducements and sanctions, the more likely it is to push for supply-chain adjustment using geopolitics as justification. This pattern is particularly evident among mining and manufacturing producers that rely heavily on domestic supply-chains while competing against foreign imports, especially in sectors such as automotive, steel, clean technologies and aluminum in the United States. These industries profit from “small yards, high walls” policies, which restrict access around critical minerals and infrastructure. A clear example is the Battery Materials and Technology Coalition (BMTC), an industry group composed of U.S. and Canadian companies across the battery supply-chain, including miners, processors, and technology developers. The BMTC’s core objective is to build secure, domestic alternatives to Chinese dominance in battery materials. Since its members are not dependent on China as suppliers, but rather face Chinese producers as competitors, their agenda is geared toward substituting Chinese imports and reinforcing U.S.–Canada supply-chain integration. Westwin Elements, in turn, describes itself as the first major U.S. nickel refinery and competes directly against imports from China. The company urges USTR to adopt aggressive trade measures to strengthen U.S. critical mineral supply-chain resilience, particularly in nickel, cobalt, and manganese. Westwin supported raising Section 301 tariffs on Chinese refined minerals from 25% to 40%, ending de minimis privileges for tariffed goods from China, and creating sector-specific trade agreements with allied countries. It also proposes an export price support strategy to shield U.S. producers from 17 -- 17 of 49 -- Chinese undercutting and oversupply. Framing critical mineral independence as both an economic and national security imperative, Westwin positions itself as the first major U.S. nickel refinery compliant with Inflation Reduction Act domestic sourcing rules, leveraging allied feedstock to drive American energy independence and reduce reliance on China. The Uranium Producers of America, in turn, is composed of U.S. mines located mainly in Texas and Wyoming, such as Cameco, enCore, and Strata Energy. USA Rare Earth is based in Texas, while AMG Vanadium is listed as a domestic producer of vanadium products. These groups also defended supply-chain adjustments explicitly on geopolitical grounds. In some cases, producers of strategic minerals have significant operations in the U.S. but are also international. Still, they should profit from sanctions to the extent that their operations are already located in “friendly” countries. Arcadium Lithium, one of the world’s largest integrated lithium producers formed through the 2023 merger of Livent and Allkem, is a clear example, as its production is highly internationalized. The company argued that strengthening U.S. supply-chain resilience for critical minerals—particularly lithium—depends on three pillars: expanding domestic processing capacity, deepening cooperation with Western Hemisphere allies, and reducing reliance on China, which currently dominates global refining. While the company supports U.S. mining, it emphasizes that South American lithium—especially from its large operations in Argentina—remains indispensable in the short term to supply American processing facilities. To secure this flow, Arcadium advocates renewing the Generalized System of Preferences (GSP) to remove tariffs on lithium inputs, negotiating a Critical Minerals Agreement with Argentina so that its production qualifies under IRA tax credits, and establishing a “Lithium Alliance for the Americas” to coordinate investment, infrastructure, and sustainability standards. For Arcadium Lithium, the emphasis on “friendly” countries is therefore quite notable. The same logic that applies to Arcadium Lithium also applies to Sibanye-Stillwater. In its April 2024 submission to USTR, the company frames its U.S.-based platinum group metals mining and recycling operations in Montana as both an economic and national security asset, urging policy changes to improve competitiveness against cheaper foreign supply. It cites the United States as its operational base, Russia as the dominant global palladium producer with 40% of supply, and China as a key source of critical mineral dependence to be reduced, while highlighting Montana’s stringent tax, regulatory, and labor standards as a model. Although Sibanye-Stillwater operates globally — 18 -- 18 of 49 -- including in South Africa, Zimbabwe, Canada, Brazil, Australia, Finland, and France — it has no operations in Russia, which it portrays solely as a competitor with lower environmental and labor costs. Its geopolitical framing thus supports a primarily commercial goal: securing U.S. incentives, tax credits, and enforcement measures that would sustain domestic production for security reasons. Beyond mining, the manufacturing sector with a mostly domestic supply-chain also strongly frames its position in terms of geopolitics. The Alliance for American Manufacturing (AAM), for instance, argues that “an innovative and growing manufacturing base is vital to America’s economic and national security” 3 and supports both the continuation of Section 301 sanctions and incentives for domestic production such as the CHIPS and Science Act. Although AAM mentioned China’s export controls on rare earths against Japan in 2010, it did not itself advocate for similar controls or for stockpiling rare metals. In some cases, however, where I would have expected strong advocacy for sanctions and inducements, support for a geopolitical framing was more nuanced than initially anticipated. DENTEC, for example, is a U.S.-based producer of personal protective equipment (PPE) such as masks and respirators. Its emphasis falls mainly on the lessons from COVID. The American Medical Manufacturers Association (AMMA), of which DENTEC is a member, did not present as extensive a geopolitical framing as AAM and mining companies. Still, it framed security issues in very assertive terms: “we face a national security risk that must be addressed.”4 Some other nuances are worth noting. Overall, mining and transformation companies have used heavier geopolitical language than other domestic manufacturers. Some mining companies, such as Bekaert, have more complex and diverse supply-chains. The company produces steel wire and downstream industrial inputs and depends on global inputs and foreign export markets. Yet it still employs an aggressive and pervasive geopolitical language vis- `a-vis other producers that rely more heavily on domestic supply-chains. This seems linked to strategic lobbying decisions that do not contradict the company’s commercial preferences. Because of competition against downstream products from India, Bekaert has decided to “choose its fights,” promoting a more coherent position before the government — one in favor of keeping “equal or greater tariffs to finished imported transmission conductors”5 rather than simultaneously requesting lower tariffs in some cases and higher in others. In some cases, both a geopolitical and an efficiency framing coexist, and differences are somewhat harder to distill qualitatively. They may include the breadth of references as well as 3USTR-2024-0002-0047. 4USTR-2024-0002-0149. 5USTR-2024-0002-0029 19 -- 19 of 49 -- the depth of references (e.g., “must” versus “should”). All in all, however, these initial results align with the theoretical framework. Neither Inducements, Nor Sanctions In some cases, groups are neither favorable to sanctions nor inducements and instead favor opening new markets while avoiding retaliation from trade partners. These should generally be export-oriented actors that strongly depend on U.S. action with its trade partners to assure market access in heavily protected foreign markets. Although some of these groups also compete against imports (e.g., beef and pork producers), they are even more concerned about losing market opportunities abroad. They will therefore use a nuanced language and detach themselves from geopolitical rhetoric. Rather than justifying the closing of domestic markets based on geopolitical rivalry—which would reinforce a frame counter to their interests—these groups may defend a “science-based” approach to imports. Government incentives will also be seen as potentially distorting, since they do not directly benefit from them. Croplife, for instance, indicates that “USTR plays an important role in contributing to supply-chain stability, but this is best done through trade liberalization and facilitation policies while preventing harmful non-market policy distortions”.6 In result, I would expect these groups to avoid mentioning geopolitics or to cast their preferences primarily in terms of efficiency. The National Pork Council (NPC), for instance, does not cite national security issues. It does, however, adopt some of the language used in the USTR consultation, such as “geopolitical friendly nations,” probably to avoid appearing too disconnected from the context of the consultation, but it is far less emphatic than mining companies and domestic U.S. manufacturers. The same applies to the Meat Institute. It urged USTR to strengthen supply-chain resilience by expanding and preserving market access through reduced tariffs, eliminating non-tariff barriers, and enforcing existing trade agreements. It warned that competitors like China and the EU are rapidly securing preferential access to key markets, threatening U.S. exporters. The Institute emphasized that both exports and certain imports are vital to the industry’s stability, called for coordinated federal action to address infrastructure needs and foreign animal disease risks, and advocated for industry-led, science-based sustainability approaches over “prescriptive measures” – as they call it – such as the EU’s Farm to Fork strategy and the EUDR. 6USTR-2024-0002-0116 20 -- 20 of 49 -- The U.S. export-oriented agricultural sector shares common positions with some unconventional potential allies. Logistics company United Parcel Services (UPS) is one of them. UPS and the Express Association of America (EAA), whose members include FedEx, UPS, and DHL, both avoid mentioning geopolitics in their submissions other than to remain minimally in sync with the language used by USTR. EAA explicitly shows skepticism vis- `a-vis “carrots” and “sticks” approaches to on-shoring and friend-shoring.7 Other strange bedfellows include foreign firms and associations of firms that primarily export final and intermediate products to the U.S. This includes Autos Drive America (ADA), which represents 13 international automakers and suppliers. ADA hardly mentions geopolitics beyond using the term “friendly” nations, also present in USTR’s call for submissions. ADA is clearly against sanctions and adopts a skeptical tone toward inducements: “Domestic incentive measures can also play an important role in reducing costs and promoting investment, but they will work at cross purposes with market-closing trade and investment policies that raise costs and increase uncertainty.”8 ADA, NPC, and EAA all reject protectionism, warning that tariffs and re-shoring alone undermine resilience. They emphasize that open trade and investment policies, supported by strong FTAs, are essential. Each highlights the need for diversified sourcing rather than over-reliance on domestic or single foreign suppliers. Finally, all stress partnerships with allies and “like-minded” countries as the foundation for resilient supply-chains. Adding up to the list of “strange bedfellows” above are firms and associations in the retail sector, especially in textiles and footwear, and foreign companies exporting to the U.S. The National Retail Federation (NRF), U.S. Fashion Industry Association (USFIA), and Footwear Distributors and Retailers of America (FDRA) all held positions broadly aligned with those of logistics and agricultural interests. FDRA, for instance, sought to distance itself from strategic sectors to avoid having sanctions imposed on imported machinery: “The footwear sector, however, does not raise national security concerns. It is a relatively low-value sector.”9 When it comes to incentives, these groups also detached themselves from the geopolitical debate: “we should . . . dedicate resources to high-end manufacturing . . . industries such as aerospace, advanced technology, automobiles, and renewable energy, to name a few.”10 As a result, their framing of geopolitics is limited, focusing instead on an economic, business-as-usual rationale. Foreign companies of all sectors that export 7USTR-2024-0002-0180. 8USTR-2024-0002-0180. 9USTR-2024-0002-0213. 10USTR-2024-0002-0213. 21 -- 21 of 49 -- to the U.S. also integrate that broad coalition. This includes Auto Drives America, a coalition of car manufacturers, such as Mercedes, which depend a lot on intra-industry trade and exports of final products to the U.S. An interesting case stands out in agriculture. The sector depends heavily on fertilizer imports. Although some agricultural producers sought to distance themselves from geopolitics to send a clearer message to USTR, given their dependency on imported machinery, the Agricultural Retailers Association (ARA) behaved more like the firms and associations detailed in the next section, which benefit more directly from domestic incentives. ARA advocated for both free trade and the inclusion of phosphate and potash in the list of critical minerals, invoking geopolitical language. This suggests an effort to balance uncertainties and risks. On the one hand, there is the risk of decoupling geopolitics from economics and failing to align with the national security framing that dominated the consultation. On the other, there is the risk of appearing overly dependent on restrictions and market distortions, which could undermine the sector’s defense of open markets and cheaper imports of machinery. Some actors therefore resorted to blending the two registers, invoking both geopolitical necessity and economic liberalization. Even then, the empirical analysis conducted below indicate that these groups rank low on their support for supply-chain adjustment based on geopolitical threats. That is likely because incentives to strengthen the domestic supply-chain of fertilizer are likely to deliver only on the mid- to long-range, while the costs of sanctions are felt immediately. Somewhere in Between There are groups that explicitly mention geopolitics but also connect it to an efficiency framing, which moderates the strength of their geopolitical stance. I would mostly expect such groups to reap concentrated benefits from inducements but not from sanctions. This should include highly integrated manufacturing sectors with operations in the U.S., such as assembly and technological development. Overall, these groups’ arguments are nuanced. They consider that the best approach to improve U.S. security is to negotiate more and stronger trade agreements to reduce dependency on China. They further emphasize the need to avoid reliance on any single source, including domestic ones. First and foremost, however, they advocate abandoning sanctions, in particular Section 301 tariffs. In some cases, they justify avoiding sanctions on the grounds that such tariffs are not linked to strategic sectors and harm U.S. consumers, sometimes framing this in geopolitical terms, or else 22 -- 22 of 49 -- simply stating that “it is simply not economically feasible for the U.S. to manufacture these types of inputs domestically.”11 To these interests, geopolitics remains relevant, and they often praise measures such as the IRA and the CHIPS Act. However, alignment with like-minded countries appears as a secondary consideration, achieved through new trade agreements rather than pursued as a goal in itself. In solar energy, a sector benefiting from inducements, several firms adopt similar positions. The Canadian Solar U.S. Module Manufacturing Corporation supports inducements and on-/near- shoring but opposes Section 301 tariffs, stressing Chinese inputs as “critical.” Heliene likewise favors inducements and near-shoring to Canada, citing labor and environmental protections, yet opposes Section 301 on cost grounds: “some products and machinery simply do not exist outside of China,” which keeps it competitive. Heliene employs somewhat stronger geopolitical language than other firms in the sector, in part because it has a large share of its operations in Canada. Maxeon demands tariffs on final product imports (sometimes justified with forced labor concerns) but rejects tariffs on intermediate goods, while supporting inducements and export incentives. These stances are echoed by the Solar Energy Manufacturers for America (SEMA), which promotes combating China’s dominance and welcomed the IRA and the so-called credits for Advanced Manufacturing, but urged that Section 301 tariffs be scrapped and replaced with inducements. Interestingly, in some cases the solar energy sector does defend restrictive action against “unfairly” imported products, yet it still largely rejects Section 301 specifically. As such, in the continuous scale to be produced during the quantitative text analysis, I would expect them to fall below, but still relatively close to, the top tier of mining and purely domestic U.S. manufacturers. Still as part of the broad “somewhere in between” category are multinational corporations in manufacturing sectors that can profit from the IRA, CHIPS and Science Act, and Bipartisan Infrastructure Law. IBM, for instance, maintains operations in Yorktown Heights and Albany (New York), Poughkeepsie (New York), and Rochester (Minnesota). Out of context, parts of its submission may appear as a purely geopolitical stance: “these public-private initiatives are critical for creating the regulatory and financial incentives to help re-shore industry sectors crucial to national security.”12 In context, however, these references are strongly tempered by efficiency considerations. When countering sanctions against countries supplying key intermediate products, IBM argued that “it is 11USTR-2024-0002-0089 12USTR-2024-0002-0089. 23 -- 23 of 49 -- simply not economically feasible for the U.S. to manufacture these types of inputs competitively.” This blend is expected due to the need to balance uncertainties, leading to a strategic framing that positions such actors between U.S.-based mining sectors and logistical operators. These groups are also strong supporters of trade liberalization, noting that “FTAs make it easier for U.S. companies to regionalize their manufacturing capacity in countries susceptible to less geopolitical risk.”13 At first glance, some associations of multinationals, such as the National Foreign Trade Council (NFTC), seem to refer more to geopolitics than IBM, hinting at what H2 proposes. Yet these differences are nuanced and will be further developed in the quantitative text analysis. Are there strange bedfellows in the “somewhere in between” group? Generally, I would expect it to be a somewhat cohesive group formed mostly of highly integrated manufacturing producers and their associations. This should typically encompass groups that traditionally align in trade politics, such as the Semiconductor Industry Association (SIA), Computer and Communications Industry Association (CCIA), the App Association, Business Roundtable, and NFTC (Cezar 2025). Within this group, I would also expect to find another class of actors that defend sanctions but not inducements. These include U.S.-based manufacturers that compete against imports but have limited incentives or capacity to export. Such groups should generally fall outside of the “small yard” sectors, such as textiles and clothing. On the one hand, they depend heavily on cheap imports of manufacturing machinery, particularly from China; on the other hand, I could not identify any organization that fit this profile. A strong candidate would have been the National Council of Textile Organizations (NCTO), but it did not submit a position paper, only a request to testify. Quantitatively, such groups should be somewhat comparable in terms of the strength of their geopolitical framing vis- `a-vis highly integrated U.S. manufacturers, but their preferred policy tools should consistently differ (sanctions vs. no sanctions, inducements vs. no inducements). For this reason, they should not become bedfellows. Evaluating H1 Using Quantitative Text Analysis The qualitative analysis above indicates that H1.1 and H1.2 are plausible. Given the nuances in positions, the qualitative analysis did not allow for a clear first check into H2.1 and H2.2, which are probed quantitatively in upcoming sections. Figures 5 and 6 show the results of the Bayesian BT model, based on pairwise comparisons generated by GPT. Some results stand out. First, the 13USTR-2024-0002-0158. 24 -- 24 of 49 -- general findings discussed above are confirmed. Mining companies top the ranking, with coalitions defending U.S.-based manufacturing interests also appearing among the top. This includes steel manufacturers, domestic producers of PPE (Blue Star NBR, DENTEC, and the American Medical Manufacturers Association), and coalitions of domestic manufacturers such as the Coalition for a Prosperous America, the American Automotive Policy Council, and the Alliance for American Manufacturing. Exiger and Premier deviate from my expectations, as they are service providers specializing in supply-chain risk management, a services sector I had not accounted for. I also did not expect to find the Trade Alliance for Health in the top quartile. As suspected, although SEMA and Heliene are solar producers fighting against Section 301 tariffs (a defensive sanction), they are positioned close to the top quartile because they nonetheless support some restrictions and inducements (see previous section). Over the middle toward the bottom of the distribution, one begins to see fewer mining companies and U.S.-based manufacturing firms. Instead, representatives of large multinationals appear, such as the National Foreign Trade Council (NFTC), IBM, Retail Industry Leaders Association (RILA), U.S. Chamber of Commerce, and Business Roundtable. There are also sectoral associations representing “small yard” sectors but that remain hesitant vis- `a-vis sanctions, such as the Semiconductor Industry Association, the Global Business Alliance, and the Information Technology Council. Some unexpected “intruders” also emerge, such as the Wire Rod Coalition, which I would generally expect to be near the top of the distribution, and the Mexican Association of Producers, which I would expect to be closer to the bottom. As indicated above, solar manufacturers occupy an interesting position. Some of them, such as the SEMA and Heliene, rank relatively high in the distribution (estimated positions of 2.722 and 1.447, respectively), while SEMA and Maxeon appear closer to the mid-to-lower part of the distribution (1.001 and -0.447, respectively). I further explore these nuances in the following sections. As expected, toward the bottom of the distribution are agricultural importers, such as the Cheese Importers Association of America; retail industry associations, such as the National Retail Federation and Meyer Natural Foods; agricultural exporters, such as the U.S. Dairy Export Council; logistics and transportation groups, such as the American Trucking Association, the Express Association of America, United Parcel Service (UPS), and Mercantile Logistics; and foreign firms exporting to the U.S., such as EGA America, Autos Drive America, and Verband Deutscher Maschinen- und 25 -- 25 of 49 -- Anlagenbau (VDMA). I ran a post-ANOVA Tukey test indicating that the differences in BT scores are statistically significant. The reference category is the set of business interests in the “somewhere in between” group, or those benefiting from inducements but not sanctions. I stress-tested the ANOVA results by accounting for clustering, using credibility intervals as weights, and introducing superficial characteristics as controls. The results hold, as indicated in Table 2. Table 2: Regressions on BTM Estimates Multinomial Logit WLS (clustered SE) (1) Defend (2) Reject (3) BTM (4) BTM (5) BTM Small yard + domestic 0.858∗∗∗ −0.006∗∗∗ 1.067∗∗∗ (0.000) (0.000) (0.393) Non-small yard + international −12.503∗∗∗ 1.830∗∗∗ −1.357∗∗∗ (0.000) (0.000) (0.340) Defend sanctions + inducements 0.963∗∗∗ (0.322) Reject sanctions + inducements −1.533∗∗∗ (0.297) Individual firm 1.120∗∗∗ (0.257) Length 0.000 0.000 0.000 0.000∗ 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Observations 148 148 148 148 148 R2 0.356 0.262 0.091 Adjusted R2 0.343 0.247 0.078 Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01 Figure 4 illustrates the ways I tested for hypotheses 1.1 and 1.2. As expected, Model 3 indicates that firms expected to be more in favor of/against both sanctions and inducements indeed frame their submissions more/less in line with geopolitics relative to the reference category (firms that are somewhere in between, benefiting from one but not the other). However, this analysis can be fleshed out further. Is there a correlation between the business interests selected ex ante and their actual position on sanctions and inducements (Model 1 and 2)? And did support for those instruments translate into broader support for geopolitics (Model 4)? The multinomial logit results underscore a sharp divide that is consistent with the theoretical expectations. For domestic-oriented “small yard” sectors, the coefficient is strongly positive for the likelihood of defending sanctions and inducements relative to the reference category (0.858, 𝑝 < 0.01). The effect of being in this group on the probability of rejecting sanctions is negative and significant (−0.006, 𝑝 < 0.01), suggesting 26 -- 26 of 49 -- that domestic, small-yard firms prefer to avoid opposing sanctions and industrial policy. In contrast, internationally exposed firms outside of small-yard sectors exhibit the opposite pattern. Compared to the reference category, they are much less likely to support sanctions or inducements (−12.503, 𝑝 < 0.01). Instead, these groups are strongly associated with rejection of both sanctions and inducements (1.830, 𝑝 < 0.01). This is in line with my theory. Firm and sector-level characteristics Support for sanctions and inducements Geopolitical framing Model 1 and 2 Model 4 Model 3 Figure 4: Multiple stages of testing Hypotheses 1.1 and 1.2 in Table 2 The above results indicate that being more/less internationalized and belonging to certain sectors as opposed to others influence the explicit demand for sanctions and inducements. A second step is to check if the explicit demand/opposition to sanctions and inducements influence geopolitical stances. The linear regression results indicate that positions on sanctions and inducements are systematically related to BTM scores. Business interests which submissions explicitly defended sanctions and inducements score on average 𝛽 = 0.963 points higher than the reference category (𝑝 = 0.0415). In contrast, those explicitly or implicitly rejecting sanctions and inducements score 𝛽 = −1.533 points lower than the reference category (𝑝 < 0.001). Overall, the BTM score captures the expected divide: defenders of sanctions and inducements score higher, opponents of sanctions and inducements score lower, and intermediates occupy the middle ground. Evaluating H2 Using Quantitative Text Analysis Do firms and associations behave differently on the geopolitics of trade? H2.1 expects the answer to be yes. Associations should adopt a more moderate position on the geopolitics of trade because they reflect the average positions of their members. As part of their strategic lobbying, firms 27 -- 27 of 49 -- Figure 5: BT Scores. 28 -- 28 of 49 -- Figure 6: BT Scores (continuation). 29 -- 29 of 49 -- are unlikely to confront the government directly when geopolitics is a central concern. Instead, they would be expected to frame their positions in ways that do not appear to openly challenge the government’s priorities. When the context signals that geoeconomic measures are crucial for policymakers, opponents generally have two options. The first is to remain silent and avoid weighing in on the matter altogether. The second is to “hide” behind their associations, channeling their lobbying through collective platforms and thereby diluting the association’s overall stance. To test H2.1, I first conducted a simple t-test to determine whether the difference in mean BT scores between firms and associations is statistically significant. As expected, the t-test indicated that firms (M = 0.63) scored higher on the Bradley–Terry model than associations (M = –0.45), with 𝑝 < 0.01. Table 2 further estimates a weighted least squares model using the inverse of the variance as weights, as well as average submission length and cluster-robust standard errors. The results hold under those conditions. Figure 7: Comparison between firms and their sectoral associations based on Bradley-Terry scores. Does the geopolitics of trade generate new cleavages not expected under NNTT? Once more, I would expect so based on H2.2, especially around cross-sectoral coalitions. NNTT predicts sectoral cleavages; however, firms in cross-sectoral associations are often more able to unite (Cezar 2023; Osgood 2021). Cross-sectoral associations are typically composed of large firms from different sectors that nevertheless align around key issues due to similar firm-level characteristics (e.g., the Retail Industry Leaders Association, which includes Walmart and Microsoft). Yet because sector matters under “small yard, high walls” policies, I would expect cleavages within and across cross-sectoral 30 -- 30 of 49 -- coalitions. Large firms within those associations will have different geopolitical preferences depending on whether they benefit from inducements. This should lead to (1) marked differences in how these associations behave vis- `a-vis each other and their members, and (2) many large members choosing to go solo. The results generally corroborate these expectations, though with nuances. In Figure 8, one can see that the positions of Business Roundtable, NFTC, RILA, and the U.S. Chamber of Commerce are generally not significantly different from one another or from IBM when considering positions on investment protection before 2020. With the geopolitics of trade, however, things shift. RILA diverges from NFTC and Business Roundtable, as expected. RILA is composed of large and highly GVC-integrated firms such as Gap Inc., Home Depot, Walgreens, CVS Health, and Walmart. NFTC and Business Roundtable also include large, highly GVC-integrated firms such as Microsoft, IBM, and Caterpillar. Yet Business Roundtable and NFTC hold positions that are statistically different from those of IBM and RILA. This provides initial support for H2.2. IBM distances itself from its main cross-sectoral associations, which represent firms that benefit both more and less from geopolitics than IBM itself, creating incentives for the firm to go solo. For RILA, which is mostly composed of firms that benefit neither from sanctions (given their import dependence) nor from inducements (since they are not “small yard” sectors), it makes sense to adopt a more conservative stance vis- `a-vis geopolitics compared to IBM (which benefits from inducements) and NFTC/Business Roundtable (which have many members in strategic sectors such as semiconductors and energy). However, I found considerably fewer firms going solo than expected based on H2.2. The two largest cross-sectoral associations representing multinationals, namely the Business Roundtable and NFTC, count hundreds of members each, yet only IBM positioned itself individually. Why? The variation among associations’ positions, while statistically significant, is not dramatic. Many large firms are simultaneously members of Business Roundtable, NFTC, and other associations such as RILA, CCIA, the App Association, and AdvaMed. These associations differ in emphasis—some leaning toward free trade, others supporting selective incentives—but none deviate radically from a middle-ground stance. By belonging to several at once, firms can rely on the diversity of association positions to reflect their preferences indirectly, hedging risks of misrepresentation of preferences and thus reducing the need to speak out individually. And why not submit individual positions in order to show “patrioc capitalism” (Moura et al. 2025) or “patriotic lobbying” (Ryu and Stone 2018)? Part of the 31 -- 31 of 49 -- answer lies on the fact that multinationals are often just as hesitant towards geoeconomic measures as they are supportive of certain selective incentives in some sectors, resulting in a potentially incoherent framing (Figure 1). Thus, it is unlikely for them to gain much by going solo, instead of via their associations. Figure 8: Comparison of positions in two issue areas. Explaining Potentially Deviant Cases The robustness tests indicate that the coalition dynamics specified in the hypotheses account for roughly 36% of the variance in the data, which is a reasonable level of explanatory power given the complexity of the subject matter. At the same time, there are nuances that deserve closer examination, as well as deviant cases that do not fully conform to the expected patterns. For example, Figure 9 highlights firms and associations in the solar energy sector that exhibit substantial variation in their support for measures aimed at reducing dependency on adversarial countries. The American Alliance for Solar Manufacturing (AASM) represents many domestic producers that compete against imports and therefore advocated for policies to promote U.S. production. The alliance favors inducements and defensive trade instruments such as countervailing duties (CVD) and anti-dumping (AD) measures, supports the restoration of Section 201, and emphasizes the importance of maintaining a domestic 32 -- 32 of 49 -- supply-chain. Its position combines criticism of China with a defense of multilateralism. AASM members include Convalt, First Solar, Meyer Burger, Mission Solar, Qcells, REC Silicon, and Swift Solar. By contrast, Maxeon, Heliene, and Suniva are more internationally oriented firms that are not members of AASM, and consequently articulate positions that are less strongly aligned with a geopolitical framing. Figure 9: Comparison of selected organizations There is also a significant difference between the Computer and Communications Industry Association (CCIA) and the Semiconductor Industry Association (SIA). CCIA has a broader mem- bership than SIA, including e-commerce companies such as eBay, cloud service providers such as Cloudflare, and transportation services such as Uber. By contrast, SIA is much more focused on semiconductor manufacturing, which is concentrated in a few geopolitical flashpoints, notably Taiwan and South Korea. In these regions, physical infrastructure, specialized capabilities, and regional security dynamics make geographic and alliance considerations unavoidable. In contrast, digital and communications service supply-chains are more distributed, virtual, and modular, relying less on single physical sites and more on globally dispersed networks. This makes them less vulnerable to territorial chokepoints and reduces the centrality of physical control over production locations in 33 -- 33 of 49 -- resilience planning. There is also meaningful variation within the retail and medical/pharmaceutical sectors. As expected, retail associations display very low BT scores overall. However, RILA’s scores are sig- nificantly higher than FDRA’s. One potential explanation is that RILA members include, in some cases, high-technology manufacturers such as Apple, which may benefit from inducements. FDRA’s members, by contrast, are heavily import-dependent and belong to a sector that does not benefit from inducements, as discussed earlier. Producers of pharmaceuticals and medical devices also vary in their positions. AMMA’s stronger geopolitical framing is easier to explain, given its largely domestic membership. PhRMA, in turn, represents global members such as Pfizer. The same applies to AdvaMed, whose members include large multinational device producers such as Abbott. These organizations are more exposed to geopolitical disruptions and therefore more likely to stress alliances and adversarial dependencies. PhRMA, by contrast, relies primarily on intellectual property, patents, and regulatory frameworks rather than critical inputs. Its discourse centers on IP protection, trade rules, and regulatory stability, which reads as commercial/legal rather than geopolitical. Exiger and Premier are companies which I did not expect to necessarily find at the top of the classification. An ex post evaluation shows that their classification makes sense though. Both are companies that position themselves as service providers deeply embedded in supply-chain management and risk monitoring. Their business models depend on identifying vulnerabilities, promoting resilience, and framing supply-chains as matters of strategic concern, which naturally aligns them with a geopolitical discourse on trade and national security. In some cases, in turn, it has been difficult to assign a category to the organizations that submitted positions during the consultation. This includes companies such as Applied DNA, Coalition for Fair Trade in Hardwood Plywood, Consumer Brands Associations, Forging Industry Association and Trade Alliance for Heath. In some other cases, the results did not conform to what I expected, as exemplified by the Specialty Steel Industry having a low BT score. Generally speaking, however, these have been exceptions rather than the rule. Overall, the theory seems to capture well the variation I expected to find. As part of an explanation that assumes firms will not detract from their economic preferences when advocating for geoeconomic measures, I expected a “geopolitics first” frame (Figure 1) to be rare. The same should apply to offensive sanctions, which I also expected to be uncommon. Indeed, both are rare, with some potential exceptions. Perryman Company is a U.S. producer of titanium mill 34 -- 34 of 49 -- products that relies on imported titanium sponge, particularly from Japan, while advocating a careful mix of trade liberalization, sanctions, and domestic incentives. The firm supports eliminating duties on imports of titanium sponge to secure access to inputs but insists on maintaining offensive sanctions on Russia and defending against Chinese encroachment. Its position combines inducements—such as incentives for domestic sponge production and the creation of a strong national stockpile—with both defensive and offensive measures directed at adversarial suppliers. Robustness Tests Table 2 shows that even when accounting for clustering and dependency among observations, the results hold. Comment length has no effect, suggesting that superficial characteristics do not influence LLM decisions. I rely on a human research assistant to cross-validate a random sample of the pairwise comparisons performed by GPT and used to create the BT model. At the time of writing, the cross-validation is still ongoing. The human research assistant was also tasked with creating a 1–5 pointwise assessment of the level of geopolitical references in the submissions. On this scale, the average score of firms that benefited the most from sanctions and inducements is 2.13 according to human coding and 2.37 according to LLM coding. For those that benefited the least, the average is, respectively, 1.39 and 1.72, while the intermediate group has an average of 2 and 2.37. These results are consistent with the initial expectations. On the other hand, if we compare the human 1-5 ranking and ask GPT to carry out the same task, the squared Cohen’s Kappa of 0.234 (𝑝 < 0.01) indicates only fair agreement. The similarity is statistically significant, but GPT struggles to consistently reproduce human distinctions. Part of the low concordance may reflect cognitive load, as humans integrate multiple cues differently than GPT’s pattern recognition. This reinforces the relevance of pairwise comparisons, as they simplify categorization tasks performed by LLMs Qin et al. (2024). To further assess whether the BT model estimates can be trusted, I test the transitivity assumption, which expects that if organization A is preferred over B, and B is preferred over C, then A should also be preferred over C. This assumption enables the model to infer a coherent latent ranking from noisy pairwise data. I found that 12.4% of the data violates the transitivity assumption. This value is likely inflated by cases in which the same organization submitted multiple entries under a shared name, or where collective submissions were split into separate actors with identical texts. Both situations artificially multiply potential inconsistencies in the pairwise structure. Overall, the results can be 35 -- 35 of 49 -- considered acceptable, and it is unlikely that the transitivity violations threaten the validity of the results in this case. Future versions of this paper will also include a comparison of outcomes using subtle variations in prompting, to evaluate the sensitivity of both LLMs and the BT model to small prompt changes (Cezar 2025). They will also include a product-of-experts (PoE) analysis of model outputs for further robustness. Conclusion In the geopolitics of trade, sectoral cleavages matter once again, complementing existing firm-level expectations in trade politics. This paper has shown that the distribution of concentrated costs and benefits across sectors and supply-chain characteristics explains which business actors adopt stronger or weaker geopolitical framings. It also argued that firms and industry associations display behavior that complements NNTT expectations. Once selective inducements target “small yards,” even superstar exporters and highly integrated firms and multinationals begin to use geopolitical references to defend supply-chain resilience. This is exemplified by the Business Roundtable, a cross-sectoral coalition of large firms, which adopts a language that navigates between an efficiency and a geopolitical framing of supply-chain resilience. The paper also expected to find some cleavage among cross-sectoral associations that would otherwise be expected to adopt aligned positions (Cezar 2023; Osgood 2021). To test my hypotheses, I employed a combination of qualitative and quantitative text analysis using LLMs. First, I conducted a close reading of the submissions to identify efficiency- and geopolitics-based framings, complemented by systematic human coding that classified each text along a five-point scale of geopolitical emphasis. Second, I used LLMs both for pointwise classification (mirroring the human- coded scale) and for pairwise comparisons between submissions, applying a Bradley–Terry model to estimate organizations’ relative positions. This two-pronged strategy allowed me to triangulate findings. By cross-validating LLM results against human coding and applying robustness checks such as weighted least squares and Bayesian inference, I ensured that the findings are not just idiosyncrasies but instead reflect consistent patterns in how firms and associations articulate their positions. I also ran multiple additional robustness tests to boost the confidence in my findings. The tests confirmed hypotheses H1.1, H1.2, H2.1, and—partially—H2.2. As expected by H1.1 and H1.2, US-based mining and manufacturing consistently supported defensive sanctions and 36 -- 36 of 49 -- inducements, framing their positions in national security terms, while agriculture, retail, and logistics emphasized efficiency and market openness. Multinationals, in turn, blended both registers, supporting selective inducements but remaining cautious toward broad sanctions. The analysis has shown that the broad coalition hesitant of geopolitics brings together some usual partners, reaffirming the effects of geopolitics on new trade coalitions (McNamara 2024). In line with H2.1 and H2.2, the results also highlight the differentiated role of individual firms and trade associations. Firms articulated sharper and more assertive geopolitical framings, whereas associations moderated their language, often reflecting the average preferences of heterogeneous memberships. In several cases, firms appeared to strategically delegate lobbying to associations or abstain from voicing strong opposition. Cross-sectoral associations, such as the Retail Industry Leaders Association and the National Foreign Trade Council, exhibit cleavage, but there have been fewer firms going solo than expected by H2.2. It seems that firms managed to hedge across associations, allowing firms to convey nuanced preferences without going solo. All in all, the paper confirms that heterogeneity in business geopolitical positioning aligns with instrument design. In turn, firms’ preferences for each instrument are explained by their expected distributive consequences. This finding reinforces existing works showing that responses to the geopolitics of trade are varied and pragmatic (D ¨ur et al. 2025; Eldes et al. 2025a; Lee and Osgood 2021; Lubinski and Wadhwani 2020). The paper also confirms that a theory based on sector × firm-level characteristics explains well how firms behave in the geopolitics of trade. Finally, the paper reasserts the relevance of LLMs for analyzing business positions (Cezar 2025), though with the need for caution. Future research could use LLMs to extend the analysis carried out here to other countries. It may also explore which sectors are most likely to become “strange bedfellows” in the geopolitics of trade. Word embeddings could also help explore in further detail the patterns of discursive alignment among unusual partners in a geopolitical setting. 37 -- 37 of 49 -- References Aiyar, S., Malacrino, D., and Presbitero, A. F. (2024). Investing in friends: The role of geopolitical alignment in FDI flows. European Journal of Political Economy, 83:102508. Babic, M., Fichtner, J., and Heemskerk, E. M. (2017). States versus Corporations: Rethinking the Power of Business in International Politics. The International Spectator, 52(4):20–43. Babic, M. and Linsi, L. (2025). Mapping Corporate Investments Between China and Europe in an Era of Geoeconomic Competition. JCMS: Journal of Common Market Studies, 63(3):932–963. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jcms.13672. Baccini, L., D ¨ur, A., and Haftel, Y. (2015). Imitation and Innovation in International Governance: The Diffusion of Trade Agreement Design. In D ¨ur, A. and Elsig, M., editors, Trade Cooperation: The Purpose, Design and Effects of Preferential Trade Agreements. Cambridge University Press, Cambridge, United Kingdom. Bauerle Danzman, S. and Meunier, S. (2024). The EU’s Geoeconomic Turn: From Policy Laggard to Institutional Innovator. JCMS: Journal of Common Market Studies, 62(4):1097–1115. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jcms.13599. Bauerle Danzman, S. and Meunier, S. (2023). Na¨ıve no more: Foreign direct investment screening in the European Union. Cezar, R. (2025). Ranking Business Trade Preferences using GPT. Cezar, R. F. (2022a). Dispute settlement, labor and environmental provisions in PTAs: When will business interests shift positions? Business and Politics, 24(2):221–240. Publisher: Cambridge University Press. Cezar, R. F. (2022b). The Intricacies of Firms’ Support for Labor Provisions in US Trade Agreements. Global Policy. Cezar, R. F. (2023). When do business associations want a hard trade-sustainability nexus? A framework of analysis and the EU case. New Political Economy, 0(0):1–17. Publisher: Routledge eprint: https://doi.org/10.1080/13563467.2023.2238627. 38 -- 38 of 49 -- Choer Moraes, H. and Wigell, M. (2022). Balancing Dependence: The Quest for Autonomy and the Rise of Corporate Geoeconomics. In Babi ´c, M., Dixon, A. D., and Liu, I. T., editors, The Political Economy of Geoeconomics: Europe in a Changing World, pages 29–55. Springer International Publishing, Cham. Cui, V., Vertinsky, I., Wang, Y., and Zhou, D. (2023). Decoupling in international business: The ‘new’ vulnerability of globalization and MNEs’ response strategies. Journal of International Business Studies, 54(8):1562–1576. Curran, L. and Eckhardt, J. (2023). The EU’s COVID-19 policy response and the restructuring of global value chains. Global Policy, 14(S3):30–39. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/1758-5899.13165. De Ville, F., Happersberger, S., and Kalimo, H. (2023). The unilateral turn in EU trade policy? The origins and characteristics of the EU’s new trade instruments. EUROPEAN FOREIGN AFFAIRS REVIEW, 28:15–34. Doppen, F., Calcara, A., and De Bi `evre, D. (2024). Balancing security and economics: domestic state-firm relations and investment screening mechanisms in Europe. Re- view of International Political Economy, 31(6):1637–1661. Publisher: Routledge eprint: https://doi.org/10.1080/09692290.2024.2368601. D ¨ur, A., Mateo, G., and Visart, L. (2025). Geopolitics meets business interests: the EU and European economic security. Journal of European Public Policy, 0(0):1–28. Publisher: Routledge eprint: https://doi.org/10.1080/13501763.2025.2537776. Eckert, S. (2024). Business Power and the Geoeconomic Turn in the Single Euro- pean Market. JCMS: Journal of Common Market Studies, 62(4):973–992. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jcms.13604. Eckhardt, J. and Lee, K. (2018). Global Value Chains, Firm Preferences and the Design of Preferential Trade Agreements. Global Policy, 9(S2):58–66. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/1758-5899.12612. Eldes, A., Lee, J., and Osgood, I. (2025a). Allied import options available? Finding friendly trade partners amidst decoupling from China. The Review of International Organizations. 39 -- 39 of 49 -- Eldes, A., Lee, J., and Osgood, I. (2025b). Trade lobbying works (for big firms): Evidence from the China trade war. Business and Politics, pages 1–35. Eliasson, L. J. and Garcia-Duran, P. (2023). New is old? The EU’s Open, Sustainable and Assertive Trade Policy. Global Policy, 14(S3):9–18. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/1758- 5899.13183. Fichtner, J. (2017). Perpetual decline or persistent dominance? Uncovering Anglo-America’s true structural power in global finance. Review of International Studies, 43(1):3–28. Findley, M. G., Nielson, D. L., and Sharman, J. (2015). Causes of Noncompliance with International Law: A Field Experiment on Anonymous Incorporation. American Journal of Political Science, 59(1):146–161. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/ajps.12141. Fiott, D. (2024). From Liberalisation to Industrial Policy: Towards a Geoeconomic Turn in the European Defence Market? JCMS: Journal of Common Market Studies, 62(4):1012–1027. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jcms.13600. Foschi, M. (1997). On Scope Conditions. Small Group Research, 28(4):535–555. Publisher: SAGE Publications Inc. Gilpin, R. (1975). U.S. Power and the Multinational Corporation: The Political Economy of Foreign Direct Investment. Basic Books. Gilpin, R. (1976). The Political Economy of the Multinational Corporation: Three Contrasting Perspectives. American Political Science Review, 70(1):184–191. Hanegraaff, M., Poletti, A., and Aizenberg, E. (2023). Economic globalization and the fracturing of business interest representation in the European Union. Business and Politics, 25(4):353–369. Hanegraaff, M., Poletti, A., and Van Ommeren, E. (2024). Firms and Trade Policy Lobbying in the European Union. JCMS: Journal of Common Market Studies, 62(3):629–652. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/jcms.13520. Haroche, P. (2024). Geoeconomic Power Europe: When Global Power Competition Drives EU Integration. JCMS: Journal of Common Market Studies, 62(4):938–954. 40 -- 40 of 49 -- Herranz-Surrall ´es, A., Damro, C., and Eckert, S. (2024). The Geoeconomic Turn of the Single European Market? Conceptual Challenges and Empirical Trends. Hurrell, A. (2024). Geopolitics and global economic governance. Oxford Review of Economic Policy, 40(2):220–233. Jeong, H., Park, C., Hong, J., Lee, H., and Choo, J. (2025). The Comparative Trap: Pairwise Comparisons Amplifies Biased Preferences of LLM Evaluators. arXiv:2406.12319 [cs]. Kato, K., Purnomo, A., Cochrane, C., and Saqur, R. (2024). L(u)PIN: LLM-based Political Ideology Nowcasting. arXiv:2405.07320 [cs]. Kim, J., Evans, J., and Schein, A. (2025). Linear Representations of Political Perspective Emerge in Large Language Models. arXiv:2503.02080 [cs]. Kinderman, D. (2021). German Business Mobilization against Right-Wing Populism. Politics & Society, 49(4):489–516. Publisher: SAGE Publications Inc. K ¨oncke, P. and Graaff, N. d. (2024). Chinese Multinationals and Europe’s Geoeconomic Turn: The De-Globalization of the Chinese ICT and Automotive Industry? | Article | Politics and Governance. Lee, J. and Osgood, I. (2021). Firms Fight Back: Production Networks and Corporate Opposition to the China Trade War. In Solingen, E., editor, Geopolitics, Supply Chains, and International Relations in East Asia, pages 153–172. Cambridge University Press, Cambridge. Leo, R. D., Zeng, C., Dinas, E., and Tamtam, R. (2025). Mapping (A)Ideology: A Taxonomy of European Parties Using Generative LLMs as Zero-Shot Learners. Political Analysis, pages 1–8. Linsi, L. and Gristwood, E. (2024). The Myth of Deglobalization: Multinational Corporations in an Era of Growing Geopolitical Rivalries | Article | Politics and Governance. Liusie, A., Raina, V., Fathullah, Y., and Gales, M. (2024). Efficient LLM Comparative Assessment: a Product of Experts Framework for Pairwise Comparisons. arXiv:2405.05894 [cs]. Lubinski, C. and Wadhwani, R. D. (2020). Geopolitical jockeying: Economic nationalism and multinational strategy in historical perspective. Strategic Management Journal, 41(3):400–421. eprint: https://sms.onlinelibrary.wiley.com/doi/pdf/10.1002/smj.3022. 41 -- 41 of 49 -- Madeira, M. A. (2016). New trade, new politics: intra-industry trade and domestic political coalitions. Review of International Political Economy, 23(4):677–711. Publisher: Routledge eprint: https://doi.org/10.1080/09692290.2016.1218354. Mavreti ´c, A. and Vangeli, A. (2024). Engaging with Geopolitical Uncertainty: Discursive Riskwork and Western MNEs’ Divestment Announcements at the Onset of the Ukraine War. Economic and Business Review, 26(4):265–283. McNamara, K. R. (2024). Transforming Europe? The EU’s industrial policy and geopolitical turn. Journal of European Public Policy, 31(9):2371–2396. Mens, G. L. and Gallego, A. (2025). Positioning Political Texts with Large Language Models by Asking and Averaging. Political Analysis, 33(3):274–282. Meunier, S. and Nicolaidis, K. (2019). The Geopoliticization of European Trade and Investment Policy. Journal of Common Market Studies, 57(Supplemental Issue):103–113. Mol, M. J., Rabbiosi, L., and Santangelo, G. D. (2023). Should I Stay or Should I Go? How Danish MNEs in Russia Respond to a Geopolitical Shift. AIB Insights, 23(1). Publisher: Academy of International Business. Moura, S. T. G., Lawton, T. C., and Tobin, D. (2025). How do multinational enterprises respond to geopolitics? A review and research agenda. International Journal of Management Reviews, n/a(n/a). eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/ijmr.12399. Nicoll Victor, J. (2007). Strategic Lobbying: Demonstrating How Legislative Context Affects Interest Groups’ Lobbying Tactics. American Politics Research, 35(6):826–845. Publisher: SAGE Publications Inc. O’Hagan, S. and Schein, A. (2024). Measurement in the Age of LLMs: An Application to Ideological Scaling. arXiv:2312.09203 [cs]. Osgood, I. (2021). Vanguards of globalization: Organization and political action among America’s pro-trade firms. Business and Politics, 23(1):1–35. Publisher: Cambridge University Press. 42 -- 42 of 49 -- Qin, Z., Jagerman, R., Hui, K., Zhuang, H., Wu, J., Yan, L., Shen, J., Liu, T., Liu, J., Metzler, D., Wang, X., and Bendersky, M. (2024). Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting. arXiv:2306.17563 [cs]. Ryu, J. and Stone, R. W. (2018). Plaintiffs by proxy: A firm-level approach to WTO dispute resolution. The Review of International Organizations, 13(2):273–308. Sallai, D., Schnyder, G., Kinderman, D., and N¨olke, A. (2024). The antecedents of MNC political risk and uncertainty under right-wing populist governments. Journal of International Business Policy, 7(1):41–63. Seawright, J. and Gerring, J. (2008). Case Selection Techniques in Case Study Research: A Menu of Qualitative and Quantitative Options. Political Research Quarterly, 61(2):294–308. Publisher: [University of Utah, Sage Publications, Inc.]. Tripathi, T., Wadhwa, M., Durrett, G., and Niekum, S. (2025). Pairwise or Pointwise? Evaluating Feedback Protocols for Bias in LLM-Based Evaluation. arXiv:2504.14716 [cs]. Van Der Heijden, H. (1997). Political opportunity structure and the institutionalisation of the environmental movement. Environmental Politics, 6(4):25–50. Publisher: Routledge eprint: https://doi.org/10.1080/09644019708414357. Wu, P. Y., Nagler, J., Tucker, J. A., and Messing, S. (2023). Large Language Models Can Be Used to Estimate the Latent Positions of Politicians. arXiv:2303.12057 [cs]. 43 -- 43 of 49 -- Appendix: List of Business Groups Categories I here list all business organizations according to whether they are expected to gain the most or the lose the most from sanctions and inducements. The list has been created based on the organizations’ self-description of their activities and supply-chain connections both in submissions and in their websites. Concentrated Benefits • AGRU America • Altana • AMG Vanadium LLC • ATI Inc. • Alliance for American Manufacturing (AAM) • Aluminum Association • Aluminum Extruders Coalition • Aluminum Extruders Council • Aluminum Extrusions Fair Trade Committee • American Alliance for Solar Manufacturing • American Automotive Policy Council (AAPC) • American Chemistry Council (ACC) • American Council on Renewable Energy (ACORE) • American Institute of Steel Construction (AISC) • American Iron and Steel Institute (AISI) • American Medical Manufacturers Association (AMMA) • American Mold Builders Association (AMBA) • Applied DNA Sciences, Inc. • Arcadium Lithium • Arkema Inc. • Blue Star NBR, LLC • Century Aluminum 44 -- 44 of 49 -- • Coalition for a Prosperous America (CPA) • Consumer Technology Association (CTA) • Dentec Safety Specialists • Exiger • Forging Industry Association (FIA) • Hemlock Semiconductor Operations • Howmet Aerospace • Initiative for Responsible Mining Assurance (IRMA) • Insteel Industries Inc. • Liberty Steel USA • MP Materials Corp. • Medline Industries, LP • NV Bekaert SA • National Defense Industrial Association (NDIA) • National Mining Association (NMA) • Nucor Corporation • Olin Corporation • Optimus Steel LLC • Outokumpu Stainless USA, LLC • PC Strand Coalition • Perryman Company • Premier Inc. • Recycled Materials Association (ReMA) • Retractable Technologies, Inc. • Rio Tinto • Sibanye-Stillwater • Sotera Health • South Star Battery Metals Corp. • Specialty Steel Industry of North America (SSINA) • Sumiden Wire Products Corp. 45 -- 45 of 49 -- • TIMET • The Battery Materials and Technology Coalition (BMTC) • Transformer Manufacturing Association of America (TMAA) • USA Rare Earth, LLC • United States Steel Corporation (U.S. Steel) • Uranium Producers of America (UPA) • Wacker Chemical Corporation • Westwin Elements, Inc. • Wincom, Inc. • Wire Mesh Corp. • Wire Rod Coalition “Somewhere in Between” • App Association • AETER Data Analytics, Inc. • Advanced Medical Technology Association (AdvaMed) • Business Roundtable • Canadian Solar US Module Manufacturing • Chemours Company • Coalition for Economic Partnerships in the Americas (CEPA) • Coalition for Fair Trade in Hardwood Plywood • Coalition of Services Industries (CSI) • Computer Communications Industry Association (CCIA) • Epsilon Advanced Materials (EAM) • FDP Virginia Inc. • GS1 US • Glen Raven • Global Business Alliance (GBA) • Global Data Alliance (GDA) • Global ReEnergy Holdings • Heliene USA Inc. 46 -- 46 of 49 -- • IBM Corporation • Information Technology Council (ITI) • J.R. Simplot Company • MECO Corporation • Maxeon Solar Technologies • National Association of Manufacturers (NAM) • National Electrical Manufacturers Association (NEMA) • National Marine Manufacturers Association (NMMA) • Nuclear Energy Institute (NEI) • Pharmaceutical Research and Manufacturers of America (PhRMA) • Semiconductor Industry Association (SIA) • Solar Energy Manufacturers for America (SEMA) • Suniva • U.S. Chamber of Commerce Concentrated Costs • Agricultural Retailers Association (ARA) • Alliance for Fair Sugar Policy (AFSP) • American Feed Industry Association (AFIA) • American Forest Paper Association (AFPA) • American Trucking Associations, Inc. (ATA) • Animal Health Institute (AHI) • Asociaci´on Mexicana de Horticultura Protegida, A.C. • Asociaci´on de Productores de Hortalizas del Yaqui y Mayo • Autos Drive America • Azucareros del Istmo Centroamericano (AICA) • Becton, Dickinson and Company (BD) • Bracane Company • Chese Importers Association of America (CIAA) • Confederaci´on de Asociaciones Agr´ıcolas del Estado de Sinaloa, A.C. • Consejo Agr´ıcola de Baja California, A.C. 47 -- 47 of 49 -- • Consumers Brands Association (CBA) • CropLife America (CLA) • EGA America, Inc. • Express Association of America (EAA) • Footwear Distributors Retailers of America (FDRA) • Fresh Produce Association of the Americas (FPAA) • Global Coalition for Efficient Logistics (GCEL) • IMS Worldwide, Inc. (IMSW) • ImportGenius • International Dairy Foods Association (IDFA) • Meat Institute • MedSource Labs • Mercantile Logistics International Trade, Inc. • Meyer Natural Foods • National Association of Foreign-Trade Zones (NAFTZ) • National Cattlemen’s Beef Association (NCBA) • National Customs Brokers Forwarders Association of America, Inc. (NCBFAA) • National Fisheries Institute (NFI) • National Foreign Trade Council (NFTC) • National Milk Producers Federation (NMPF) • National Pork Producers Council (NPPC) • National Retail Federation (NRF) • National Wooden Pallet and Container Association (NWPCA) • Pacific Seafood Processors Association (PSPA) • Parkdale Mills, Inc. • Remake • Responsible Industry for Sound Environment (RISE) • Retail Industry Leaders Association (RILA) • Seafood Producers • Sistema Producto Tomate [Mexican Tomatoes Association] 48 -- 48 of 49 -- • The Software Data Alliance (BSA) • Trade Alliance for Health (TAFH) • U.S. Dairy Export Council (USDEC) • U.S.Fashion Industry Association (USFIA) • United Parcel Service (UPS) • Verband Deutscher Maschinen- und Anlagenbau (VDMA) 49 -- 49 of 49 --