Global Markets, Corporate Assurances, and the Legitimacy of State Intervention

Matthew Amengual Associate Professor, Saïd Business School University of Oxford [email protected]

Tim Bartley Professor of Sociology Washington University in St. Louis [email protected]

January 6, 2021

*** DRAFT *** Abstract Collective perceptions of harm and impropriety channel the evolution of capitalism, as shown by research on the moral boundaries of markets. But how are boundaries perceived when markets cross borders, harms are distant, and individuals must navigate competing claims from advocacy organizations and corporations? These conditions are particularly salient in global supply chains, which have become central modes of production and exchange. Extending research on situational perceptions of objectionable markets, we analyze data from a conjoint survey experiment in which a national sample of American adults considered a range of labor and environmental problems in global supply chains. We find that claims from advocacy organizations increase the perceived importance of state intervention, though not for all parts of the political spectrum. Corporations’ voluntary assurances of responsibility have a widespread effect of undermining support for the state, though more symbolic prosocial statements do not. Moreover, different problems evoke distinct construals of harm. These findings move beyond stylized attitudes about international trade to reveal the contours of moralized global markets, the role of assurances in greasing the wheels of objectionable markets, and the legitimacy of state intervention in a period of withering neoliberalism.

Keywords: economic sociology, globalization, environment, labor

Acknowledgments: For comments on the research plan and/or earlier drafts of the paper, we thank Marius Busemeyer, Greg Distelhorst, Julian Garritzmann, Sebastian Koos, Akshay Mangla, Ethan Michelson, Ariela Schachter, Christi Smith, Emmanuel Teitelbaum, and audiences at the University of Amsterdam Center for Inequality Studies, University of Konstanz Cluster on the Politics of Inequality, and Society for the Advancement for Socio-Economics annual conference. We thank Kim Heinser for research assistance.

1 INTRODUCTION How do Americans perceive the moral boundaries of global markets? Despite four decades of globalization and repeated exposés of labor exploitation and environmental degradation, it is not clear how people decide what counts as objectionable and what can be dismissed as distant, negligible, or the normal working of the market. News about the poisoning of workers making smartphones in China or the burning of forests in Brazil to make leather footwear, for instance, could be perceived as unfortunate but inevitable, as best left to others to solve, or as requiring intervention. Under what conditions will a particular response predominate? When will “sympathy fatigue” (Hochschild 2016) be high, and when will a sense of global interdependence materialize? A large literature has revealed the economic and socio- cultural bases of “protectionist” attitudes (Mansfield, Mutz and Brackbill 2019), but this research reads popular perceptions of globalization narrowly through the lens of trade openness versus closure. Economic sociologists argue that markets are interpreted through morally-infused lenses. Objections to child labor, prostitution, pollution, life insurance, debt, and the trade in human organs and eggs have historically led to a range of laws to restrict markets or channel their growth (Anderson 2018, Fourcade 2011, Healy and Krawiec 2017, Quinn 2008, Zelizer 2013). Yet objectionable forms of market activity can be made more palatable through arrangements that obfuscate exchange by adding intermediaries or charitable frames (Almeling 2007, Chan 2009, Hoang 2018, Schilke and Rossman 2018, Wherry, Seefeldt and Alvarez 2019). Though rarely examined from this perspective, current architectures of global production raise questions about how markets are judged when harms are distant (Bair 2019), structures of production are fragmented (Davis 2009), and observers face competing claims about severity, responsibility, and benevolence (McDonnell, King and Soule 2015). Global supply chains organize large swaths of international trade but are most often studied from structural and firm-centered perspectives (Gereffi 2018). In this paper, we transpose a theoretical interest in the fluidity of popular judgments about market restriction with a focus on global supply chains as predominant but evolving modes of production. Specifically, we ask about the conditions under which individuals perceive state intervention as necessary to combat various forms of environmental degradation and labor exploitation in global supply chains. State intervention is a direct and intuitive way of restricting the boundaries of markets (Evans 1997, Kenworthy and Pontusson 2005, Zelizer 2013), and states have become more interventionist vis-à-vis in global supply chains over the past decade (Evans 2020, Horner 2017, Rodrik 2018). But the legitimacy of state mandates continues to be in question, particularly amidst polarized perceptions of evidence (Eyal 2019) and the rise of private authority beyond the state (Sassen 2006). We treat perceptions of state intervention into global markets as potentially flexible, situational, and multi-dimensional, consistent with the moral markets literature (Bandelj 2020, DiMaggio and Goldberg 2018). Rather than measuring generalized “free trade” versus “protectionist” attitudes, we investigate how judgments depend on particular threats or injustices, framings of severity, and assurances of responsibility. Methodologically, we use a conjoint survey experiment, a useful tool for analyzing multi-dimensional perceptions of socio-political phenomena (Flores and Schachter 2018), to extend research on popular perceptions of markets (Hahl, Zuckerman and Kim 2017, Schilke and Rossman 2018).

2 A range of environmental, labor, and human rights concerns have inspired a similar form of transnational advocacy, in which activists highlight egregious damage, press well-known corporations to take responsibility, and lobby for state intervention (Longhofer et al. 2016, Seidman 2007). Rather than bracketing these specific concerns or leaving them segregated in different literatures, we follow Polanyi (1944) in seeking to generalize across labor and environmental issues, while also asking if there are distinctive logics at play (Evans and Kay 2008). In addition, we connect the moral markets literature to research on corporations and “private politics” (Carlos and Lewis 2018, McDonnell, King and Soule 2015, Walker 2014) to ask whether assurances (not only obfuscating structures) grease the wheels of objectionable markets. In particular, we examine the effects of corporate assurances of responsibility (Lim and Tsutsui 2012) and prosocial messaging (McDonnell and King 2013) in a crowded discursive arena. Recent studies that strongly draw respondents’ attention to corporate actions have produced mixed findings about corporate voluntary initiatives’ effects on public opinion (Dana and Nadler 2019, Kolcava, Rudolph and Bernauer 2020, Malhotra, Monin and Tomz 2019). We seek a more complete portrayal, in which firms’ promises are one piece of a larger landscape, which also contains messages from advocacy groups, as theorized in the literature on transnational advocacy networks (Keck and Sikkink 1998). We proceed by developing hypotheses about advocacy frames, corporate assurances, and distant labor/environmental problems and then describe the design and administration of our survey experiment. After analyzing the results in our full sample and assessing heterogenous effects across political orientations, we incorporate qualitative evidence from an open-ended question, which provides insight into the logics of judgment. THEORY AND PRIOR EVIDENCE Moral markets research emerged with accounts of how popular perceptions of markets are sensitive to particular forms of commodification (Zelizer 2013). Scholars then identified arrangements that mute concerns by encouraging observers to see the activity as something other than a market transaction—such as structures of obfuscated exchange that mask the purchase of political influence (Schilke and Rossman 2018) or scripts of gift-giving that frame the selling of human eggs as a loving donation (Almeling 2007). The fundamental insight that markets are interpreted through situational moral lenses need not be limited to the primordial question of whether an exchange is marketized or not. The lenses through which markets are understood can also shape how economic valuation is carried out (Fourcade 2011) or how individuals view profit-seeking and inequality (Davis and Robinson 2006, DiMaggio and Goldberg 2018). In the context of global supply chains, buying-selling relationships may be apparent rather than obscured, but the controversies encountered may nevertheless be difficult for observers to square with their moral codes. This raises questions about the factors that shape these judgments and the ability of various actors to strategically amplify or minimize concerns. We therefore turn to research on advocacy networks and corporate responsibility, which highlight frames and assurances that are especially salient in global supply chains.

3 Advocacy Frames and the Structuring of Concern Research on transnational advocacy networks suggests that perceptions of distant problems are malleable, rather than being rooted purely in ideological commitments or self- interest. In addition to the initial work of selecting, naming, and drawing media attention to particular problems (Carpenter 2007), advocacy organizations mobilize particular forms of information to make distant problems into pressing and policy-relevant local concerns (Evans and Kay 2008, Keck and Sikkink 1998). For one, advocacy organizations often draw on the legitimating power of science to demonstrate the severity of the problem. Scientific frames are most commonly noted in research on environmentalism (Longhofer and Schofer 2010) and public health (Best 2012), but scientific legitimation has also proven powerful in advocacy focused on global poverty (de Souza Leão and Eyal 2019), occupational health and safety (Hilgert 2013), and corporate exploitation of vulnerable populations (Quark 2016). A second approach is to mobilize personal testimony from affected individuals. This can help to build sympathy or solidarity (Keck and Sikkink 1998, Polletta 2014) and demonstrate the deservingness of victims (Best 2012). Testimony might be conveyed in person through visits and speaking tours or in a more mediated fashion through documentary films (Vasi et al. 2015) or press quotations. For some problems, transnational advocates emphasize corporate irresponsibility and greed, which can heighten publicity and appeal to anti-corporate sentiment in various parts of the political spectrum. Scholars have argued that the legitimation of claims about global labor rights, environmentalism, and human rights has rested in part on critiques of well-known corporations (e.g., Nike, Walmart, Shell) and momentary crises of corporate legitimacy (Bair 2019, Meyer, Pope and Isaacson 2015). It can be particularly potent to counterpose the power and profits of corporations with the desperation of workers and communities in their supply chains (Bartley and Child 2014). At the same time, some activists and scholars have come to worry that criticizing corporations fuels the “private politics” of corporate social responsibility rather than strengthening state capacities (Seidman 2007, Soule 2009). Advocacy groups may therefore explicitly push for state intervention as a solution to corporate malfeasance. This approach has contributed to a move from private, voluntary initiatives to state regulation of corporate responsibility in several instances (Evans 2020, Leipold et al. 2016). Taken together, these lines of research portray perceptions of global problems as open to influence from advocacy groups, and they specify particular frames that are likely to be potent. Though these insights emerge largely from case-based and comparative research, they can be generalized into a first set of expectations about popular perceptions: Advocacy hypothesis: Statements from advocacy groups that mobilize science, testimony, or corporate critique increase the perceived importance of state intervention in global supply chains. In addition, anti-corporate statements have a larger positive effect on the perceived importance of state intervention when explicitly calling for state action rather than merely criticizing corporations.

4 Corporate Promises and the De-legitimation of the State A growing body of research shows that perceptions of state regulation are open to a different type of influence—based on companies’ assurances of responsibility and promises of voluntary reform (Lim and Tsutsui 2012, McDonnell and King 2013). If corporate assurances are perceived as a substitute for state action or otherwise ameliorate concern, then companies may be able to pre-empt state regulation with voluntary initiatives (Boghossian and Marques 2019, Malhotra, Monin and Tomz 2019). When problems in global supply chains are exposed, consumer-facing corporations often adopt new codes/standards for their suppliers and audit compliance. This is a standard form of “private regulation” in apparel, electronics, and food/agriculture industries (Bartley 2018, Locke 2013) and the most common private regulatory practice among global firms (Thorlakson, de Zegher and Lambin 2018). But this kind of “first party” private regulation leaves companies open to charges of untrustworthy greenwashing or fairwashing (Appelbaum and Lichtenstein 2016, Balsiger forthcoming). Companies often seek greater credibility by working with NGOs and nonprofit organizations in multi-stakeholder initiatives that oversee or certify compliance (Boström 2006, Cashore 2002). Multi-stakeholder initiatives have proliferated over the past two decades, with nearly 20% of publicly listed food, clothing, and forest products firms claiming to follow multi-stakeholder standards (Thorlakson, de Zegher and Lambin 2018). A third response is for corporations to cut off recalcitrant suppliers, usually after unsuccessful attempts at remediation (Amengual, Distelhorst and Tobin 2020). This punitive form of private regulation is rarer than the others and sometimes controversial. Case based research portrays these forms of private regulation as shields against state action—but imperfect shields that do not necessarily block pro-regulatory efforts (Boghossian and Marques 2019, Evans 2020, Kinderman 2020). During the expansion of private regulation, scholars have documented increased incorporation of environmental and labor conditionalities in trade agreements (Morin, Dür and Lechner 2018, Raess and Sari 2018), growing public skepticism about international trade (Mansfield, Mutz and Brackbill 2019), and a “hardening of soft law” in some environmental policy arenas (Bush and Oosterveer 2019, Leipold et al. 2016). Looking at public opinion data in , Burgoon and Fransen (2018) find at best contingent effects of private regulation on attitudes about state action with regard to domestic redistribution and (to a lesser extent) foreign aid. Given countervailing tendencies, do corporate assurances of private regulation actually undermine support for the state? Two recent studies suggest that they can, at least in particular domains and when highlighted for respondents. Respondents in Switzerland became less supportive of government regulation of companies’ foreign activities when told that Swiss companies had “voluntarily committed themselves to protect people and the environment at their operating sites abroad” (Kolcava, Rudolph and Bernauer 2020). Effects were especially clear when companies were in sectors with a history of controversy (e.g., gold, oil, coffee) and were partnering with “independent non-profit organizations” as an indicator of credibility. Research in the U.S. found that support for government regulation of environmental problems (e.g., banning genetically modified foods or the sale of bluefin tuna) declined when respondents were told of firms’ voluntary efforts, especially when more firms participated (Malhotra, Monin and Tomz 2019). However, the depth of the voluntary action did not influence respondents’ views,

5 resulting in corporations being able to reduce public support for environmental regulation with relatively meager actions. These lines of research converge on a second set of hypotheses about perceptions of state intervention: Private regulation hypothesis: Corporate assurances that suppliers are being privately regulated reduce the perceived importance of state intervention in global supply chains. In addition, some parts of the literature suggest that corporate assurances of private regulation reduce the perceived importance of state intervention more when corporations partner with nonprofit organizations than when corporations take unilateral actions. Some research goes further in suggesting that symbolic statements from companies, even if they do not amount to assurances of private regulation, can undermine more stringent interventions. McDonnell and King (2013) find that firms in the midst of controversy often make prosocial claims that “show a firm’s commitment to socially acceptable norms, beliefs, and values and protect its image by diluting, rather than refuting, the negative claims made by activists” (p.388). Carlos and Lewis (2018) argue that prosocial claims especially buffer firms from further criticism if they avoid the specific issue under contention, since this reduces charges of hypocrisy. Some research suggests that largely irrelevant information about local corporate philanthropy decreases concern about international trade (Kerner and Sumner 2020), but there is also evidence that the public is discerning of empty symbolism (Darnall, Ji and Vázquez-Brust 2018). Therefore, we test the following claims: Prosocial hypotheses: (a) Corporations’ prosocial statements reduce the perceived importance of state intervention, and (b) the effect of prosocial statements is smaller than assurances of private regulation. It is worth noting that by some accounts, assurances of private regulation could have the opposite effect—that is, to legitimate state intervention. Private regulation could help to inform observers about poorly understood problems, demonstrate that alternatives are feasible, and create business coalitions for more widespread reform (Dana and Nadler 2019, Tzankova forthcoming). While most of these processes are not visible or salient to average citizens, Dana and Nadler (2019) find evidence that firms’ voluntary actions can legitimate state regulation among conservatives, who would otherwise be opposed. Beyond this specific finding, their research speaks to the importance of examining political heterogeneity in the effects of corporate promises. Labor, Environment, and the State in Global Supply Chains Global supply chains are at the center of many current tensions between morals and markets. The “supply chain revolution” of the 1980s and 1990s allowed production processes to be split into smaller pieces, outsourced, and offshored to an unprecedented extent, with tight coordination across organizational boundaries and geographical distances (Bair 2008, Gereffi 2018). Such orchestrated value chains now account for nearly half of global trade (World Bank 2020). Research and advocacy have made it clear that global supply chains frequently rely on, produce, or exacerbate distant forms of labor exploitation and environmental degradation. This includes child labor, forced labor, dangerous work, withheld wages, and repression of unions in

6 supply chains for many consumer products (Chan, Ngai and Selden forthcoming, Collins 2003, Kerrissey and Schuhrke 2016); as well as deforestation, land degradation, water pollution, toxic waste, and violence against local communities in places where natural resources are extracted or processed (Jorgenson 2006, Shandra, Leckband and 2009, Smith, Sonnenfeld and Pellow 2006). Social scientists know surprisingly little about how people on the consuming end of global supply chains perceive this array of distant “upstream” harms. One body of research examines how attitudes about tariffs and trade are shaped by occupation, education, ethnocentrism, and economic shocks (Mansfield and Mutz 2013, Rodrik and Di Tella 2020). But with its exclusive focus on the effects of international trade on domestic labor markets, this research treats objections to distant labor exploitation or environmental degradation as either marginal or epiphenomenal. Other research examines how concerned people are as consumers, producing evidence that some consumers will pay premiums for products labeled as Fair Trade (Hainmueller, Hiscox and Sequeira 2015) or as sustainable (Teisl, Roe and Hicks 2002). Yet comparable estimates for labor and environmental problems are rare, leaving observers to speculate on which problems best motivate conscientious consumption (Bartley et al. 2015, Johnston and Baumann 2010). While it is difficult to predict precisely which problems will elicit the most interest in state intervention, it is likely that violations of fundamental labor rights (ILO 1998) will strike observers as highly objectionable, especially when they also violate sacred categories of childhood and personhood (Zelizer 2013). This suggests a high degree of support for intervention against child labor and forced labor—with the latter commonly referred to as “modern slavery” (LeBaron and Rühmkorf 2019). In practice, concerns about forced labor have inspired aggressive forms of state intervention in recent years, including actions by U.S. Customs and Border Protection from 2016 to 2020 to exclude garments and wigs from the Xinjiang region of China, cotton from Turkmenistan, seafood from certain Taiwanese fishing boats, and several other products traceable to forced labor (Sands 2019). The 2019 U.S.-Mexico-Canada Agreement, which replaced NAFTA, also allows for import bans of products made with forced labor.1 Yet other labor concerns have also sparked widespread outrage, such as dangerous work following the death of over 1,100 garment workers in the Rana Plaza factory collapse in Bangladesh (Kerrissey and Schuhrke 2016) and unpaid wages as supplier factories had orders cancelled in the wake of COVID-19-related lockdowns. Ultimately, the salience of these concerns—relative to each other or to widespread environmental problems—must be seen as an empirical question. An additional rationale suggests that support for state intervention will be heightened when problems are described as criminal violations. Theorists of crime and markets have argued that criminality is unique as a domain in which there is widespread support for strong state intervention even amidst market fundamentalism (Harcourt 2010). This is clearest in the discourse of global environmental crime, which fuses environmental and penal rationales (Pellow and Brehm 2013, Song et al. forthcoming). Interventions against illegal extraction/harvesting of natural resources provide prominent examples. In 2008, the U.S. banned the sale of products that originate with illegal logging anywhere in the world, inspiring similar policies in the EU, Japan, and Australia (Leipold et al. 2016). Illegal fishing has provoked

1 Chapter 23, Article 26.3I

7 analogous interventions by the EU, as well as aggressive crackdowns elsewhere (Hopewell 2020). This research leads us to expect that when a problem is defined as illegal or criminal, support for state regulation will be higher than when a similar problem is described in other terms. Though rooted in environmental research, this tendency may also extend to labor issues, where activists have begun to use criminal framings, such as using “wage theft” to describe unpaid wages. Of course, other justifications for state intervention could come to the fore as observers consider distant environmental and labor concerns. The fragmentation of existing research underscores the importance exploring what justifications emerge from a range of problems. RESEARCH DESIGN AND DATA Survey experiments help to uncover shared logics of judgment that can be obscured in standard survey research (Wallander 2009). Economic sociologists have used them to estimate the effects of status (Hahl, Zuckerman and Kim 2017), morality (Schilke and Rossman 2018), and inequality (McCall et al. 2017) on perceptions of markets. Conjoint survey experiments, in which a variety of elements are randomized simultaneously, are commonly used to study multi- dimensional perceptions of socio-political phenomena, including immigration (Flores and Schachter 2018), terrorism (Huff and Kertzer 2018), and political candidacy (Hainmueller, Hopkins and Yamamoto 2014). In a conjoint design, respondents select and/or rate profiles that contain a randomized array of information (Hainmueller, Hopkins and Yamamoto 2014). This approach has two main benefits for understanding perceptions of state intervention in global supply chains. First, it allows us to include a variety of labor and environmental concerns, rather than being restricted to a single domain. Second, a conjoint approach allows us to construct scenarios that approximate the noisy ways in which information about problems, advocacy frames, and corporate promises is ordinarily conveyed in news reports. For example, a recent article on labor abuses in the Ethiopian garment industry included both a statement from the corporation H&M (“We take seriously any allegations of violations of labor standards and will continue to follow up with suppliers and implement our programs addressing working conditions and workers’ rights”) and a statement from an NGO, the Worker Rights Consortium (“Unfortunately . . .there is a yawning gap between the brands’ ethical pretentions and the workplace reality for the people sewing their clothes”).2 Our design approximates the experience of reading news stories like this, while using statements that can generalize across problems and products. Outcome To measure the perceived importance of state intervention in global supply chains, we ask respondents when and to what degree it would be important for the U.S. government to “ban items from being sold in the U.S. if they are made in especially harsh or damaging ways.” A ban is a simple and straightforward way of restricting the market to prevent morally objectionable practices, as illustrated by other research in economic sociology (Healy and Krawiec 2017, Zelizer 2013). It is also an intuitively meaningful way for the U.S. government to intervene in global supply chains, without delving into the complexity of trade agreements or extended legal

2 Reuters, “Tommy Hilfiger and Calvin Klein probe 'labor abuses' in Ethiopian factories,” April 16, 2019.

8 liabilities. As illustrated above, bans have become increasingly plausible forms of state intervention to address global labor exploitation and environmental degradation. After showing respondents a pair of scenarios containing our treatments (described below), we asked for two types of judgments. The first asks respondents to make a prioritizing choice about when it would be most important to impose a ban, choosing between the two scenarios shown. Comparing a pair of scenarios tends to increase respondents’ engagement (Hainmueller, Hangartner and Yamamoto 2015), and it allows us to examine judgments in a constrained decision environment, in which respondents cannot simply revert to overarching defaults (i.e., toward a crackdown in every case or a laissez-faire approach in every case). A second measure focuses on respondents’ degree of support or opposition to a ban in each scenario. Here, respondents registered their opinion on whether a ban should be imposed on a 7- point scale from “definitely should not” to “definitely should.” This question allows preferences for a ban in both or neither of the two scenarios, as well as gradations of support or opposition, enabling us to examine the extent to which beliefs about state regulation are situational. Popular perceptions are not only indicators of the “cultural life of market exchange” (Zelizer 2013:371), they are resources and constraints for policy entrepreneurs (Manza and Brooks 2012). Measuring prioritization is especially apt if one thinks of policy entrepreneurship as finding messages that nudge public choices down one path or another. Measuring what "moves the needle” of support/opposition is highly relevant if one sees policy entrepreneurship as tapping into strong feelings. We examine both outcomes. If a treatment shapes just one outcome, we take it as a more suggestive finding that may be unique to a particular form of judgment. We also gathered qualitative evidence by asking respondents an open-ended question about the reasons for their choice, which we explore after presenting the quantitative results. Treatment Dimensions Each scenario consisted of information on (1) a product being made through global supply chains and consumed in the U.S., (2) a specific labor or environmental problem encountered, (3) a statement from an advocacy group, (4) a statement from companies in the industry, and (5) a description of which companies made this statement. Research suggests that five dimensions is well within the range that conjoint survey respondents can process (Bansak et al. 2019), and we have taken several additional steps (described below) to enhance attention and comprehension. We designed our scenarios such that all elements could be fully randomized without needing to restrict logically implausible combinations. (See Table 1 for the full list.) (1) Products: To make the scenarios concrete, realistic, and applicable to a variety of current controversies, they refer to the production of clothing, electronics, or seafood. Serious labor and environmental abuses have been documented in the supply chains for each of these products. For instance, fishing boats may be relying on forced labor, polluting waters with leftover plastic nets, and over-harvesting threatened species (Bush and Oosterveer 2019). Apparel and electronics factories may have unsafe working conditions and emit toxic chemicals into nearby skies and waterways; and the inputs for these products may endanger workers and degrade environments in farming and mining areas (Kerrissey and Schuhrke 2016, Locke 2013). We deliberately chose products that allow full randomization with the problems and to ensure that our findings are not driven by how or how often products are consumed.

9 (2) Problems: We focus on nine problems that are well-documented in the literature on labor, environmental, and human rights concerns in global supply chains and applicable to each of our three products. These include pollution of rivers and oceans, emission of toxic chemicals, dangerous work, child labor, and modern slavery. For two sets of problems, we varied whether the problem was framed in terms of illegality (i.e.., “overuse of natural resources” vs. “illegal use of natural resources;” and “unpaid wages” vs. “wage theft by employers”) to probe whether an illegality/criminality frame increases support for state intervention. (3) Advocacy frames: We developed four statements to be conveyed by “a major advocacy group” with reference to the problems at hand. Following the theoretical guidance above, these used science, personal testimony, corporate critique, or an explicit call for state action to underscore the problem and legitimate intervention. We also developed a baseline statement, which merely restated the commonness of the problem, to serve as a reference category while maintaining narrative consistency. To ensure that the statements were comprehensible and plausible, we refined based on open-ended responses from pre-testers. (4) Corporate promises: We developed four statements that companies could make in response to the problem in their industry, as well as a baseline statement that the company was simply “weighing the evidence.” Our prosocial statement makes references to ethical conduct, community, and charity, which are common elements in prosocial corporate discourse (McDonnell and King 2013). The other statements captured the three assurances of private regulation described above—(1) a unilateral response in which companies impose “strict rules” for their suppliers and audit compliance, (2) partnership with a nonprofit organization to certify suppliers’ compliance, and (3) a decision to cut ties to problematic suppliers. Again, we crafted these statements based on the theoretical arguments made above and the need for comprehension and plausibility across our problems and products, as aided by pretests. (5) Which companies made the statement: We also included information about which companies voluntarily chose to respond to the problem by issuing a statement. This serves several purposes. First, excluding this information might lead respondents to assume from the statement which type of firm made it—a problem referred to as masking (Bansak et al. 2019). Second, this allows us to see whether the presence of particular firms or larger/smaller groupings alters perceptions of state intervention, as it might if perceptions of firms are highly politicized or sensitive to industry unity. Third, this allows us to perform supplemental tests of whether corporate promises carry more weight when made by larger groups of companies (Malhotra et al. 2019) or by specific well-known companies (McDonnell and King 2018) that sell each of our three products.3 As a baseline, we include an unnamed “large retailer” as a single generic company.

3 Walmart and Target are large, multi-product retailers that have similarly high scores on Fortune’s “Most Admired Companies” ranking, which is McDonnell and King’s (2018) measure of status (#18 and #22 respectively in the 2020 rankings).

10 Table 1: Attributes and Levels of Scenarios

Product Clothing Electronics Seafood

Problem Child labor Modern slave labor Dangerous working conditions Unpaid wages Wage theft by employers Pollution of rivers and oceans Emission of toxic chemicals Overuse of natural resources Illegal use of natural resources

Advocacy frame Common problem “This is a common problem in this industry. We are looking into how it can best be addressed.” Corp coverup “Corporations might say that they are taking responsibility. But they only want to cover up their relentless pursuit of profits, which is the root of the problem.” Testimony “Listen to the voices of affected people like J.N., who told us: ‘This has caused great suffering for me and my community. We need a solution now.’” Science “The scientific evidence is clear: This is a pressing problem that severely harms many people around the world. We need to take action now.” Do not trust; Need gov “We should not trust corporations to solve a problem as important as this. It is time for the government to act.” Corporate promise No comment “We are weighing the evidence about this issue, and we cannot comment further at this time.” Prosocial “We believe in ethical conduct and good corporate citizenship. We will continue to promote these values through our charitable and community engagement initiatives.” Strict rules “We have adopted strict new rules to prevent this problem and are closely monitoring our suppliers' compliance.” Partnership to certify “We are partnering with independent non-profit organizations to certify that our suppliers meet strict standards for responsible production.” Cut supplier “We have decided to stop doing business with any suppliers that exhibit this problem.” Which firms A large retailer Target Walmart A group representing half of the companies in this industry A group representing nearly all the companies in this industry

11 Data Collection and Analysis We conducted the survey experiment online from mid-February to early-March of 2020 with a sample of 4,489 respondents provided by Dynata (formally Survey Sampling International).4 This is a national sample of U.S. adults, intended to approximate the population of the U.S. in terms of region, age, gender, income, and education. While not a probability sample, it provides a varied set of respondents with which to test our hypotheses. In a pre- treatment question, 33.8% of respondents described themselves as slightly to extremely conservative and 34.1% described themselves as slightly to extremely liberal. Respondents were 48.5 years old on average (s.d.=17.3), and 52% were women. Although people with at least some college experience were over-represented, 25% of the sample had only a high school diploma or less. The location of respondents across U.S. states is similar to the states’ percentage share of the full U.S. population. Part 1 of the Appendix provides more information on the sample and the degree to which it matches various characteristics of the U.S. population. Before fielding the survey, we registered our research design, hypotheses, and analysis plan with the Open Science Framework (OSF).5 To maximize comprehension and engagement, we first presented each scenario in a narrative, bullet-point style, approximating a news story in abbreviated form. (See Figure 1.1.) To maintain the logic of the narrative, we started with the problem and product and then moved to the business and advocacy statements—randomizing the order of these statements (consistently for each respondent throughout the survey), to ensure that any differences in these theoretically important effects are not driven by order of presentation. Respondents who read the narrative descriptions very quickly were reminded that most people take longer and asked if they wanted to go back to review (Zhang and Conrad 2016). In addition, after the instructions but prior to the descriptions, respondents were asked a simple multiple-choice question to probe whether they recalled the relevant policy, and those who failed were reminded of the correct answer, similar to Hahl, Kim, and Zuckerman (2018). After seeing two such scenarios in narrative form, respondents were shown a comparison table, as is common in conjoint surveys. (See Figure 1.2.) This allowed respondents to recall and contrast details before registering their judgments about (1) the scenario that they would prioritize for a ban by the U.S. government and (2) the degree to which they support or oppose a ban in each scenario. Attention checks were included after each pair of scenarios.

4 Opt-in samples have been found to perform similarly to population-based samples when performing survey experiments (Coppock 2019, Mullinix et al. 2016), and Dynata’s sample has fewer problems of misrepresentation than samples drawn from Amazon Mturk (Ahler, Roush and Sood 2020). 5 Anonymized registration posted on the Open Science Framework: https://osf.io/jwrty/?view_only=281cb290e3d54426bf4ab3d18783c5a0

12 Figure 1.1: Example in Narrative Form Scenario 1

Problem and Product:

illegal use of natural resources during the production of seafood sold by many companies

Of all the American businesses that sell seafood, what have some chosen to say about illegal use of natural resources?

A group representing half the companies in this industry chose to issue a statement saying:

“We have decided to stop doing business with any suppliers that exhibit this problem.”

What is a major advocacy group saying about illegal use of natural resources in the production of seafood?

“This is a common problem in this industry. We are looking into how it can best be addressed.”

Figure 1.2: Example Table of Scenarios

Here is a table summarizing the two scenarios you just read about. Remember, these refer to products that are made abroad and sold by many companies in the U.S.

Scenario 1 Scenario 2

Problem: illegal use of natural resources modern slave labor

Product: seafood sold by many companies clothing sold by many companies

“We have decided to stop doing “We have adopted strict new rules to Business business with any suppliers that prevent this problem and are closely statement: exhibit this problem.” monitoring our suppliers' compliance.”

Which businesses A group representing half of the chose to make A large retailer companies in this industry this statement:

“Corporations might say that they are taking “This is a common problem in this Advocates’ responsibility. But they only want to cover up industry. We are looking into how it statement: their relentless pursuit of profits, which is can best be addressed.” the root of the problem.”

13 As is standard in conjoint designs, each respondent was shown two pairs of scenarios, generating 17,956 observations in total (Bansak et al. 2018). Following established practice, we cluster standard errors by respondent to account for dependence among their four responses. We follow recommendations in the literature to be minimally exclusionary of low-attention respondents in our main analyses (Berinsky, Margolis and Sances 2014) and make further use of attention checks in robustness tests.6 Our analyses focus on the Average Marginal Component Effect (AMCE) of each randomized element on the perceived importance of state intervention. The AMCE can be interpreted as the causal effect of an attribute, conditional on the joint distribution of all other attributes (Hainmueller, Hopkins and Yamamoto 2014). We present the results in graphical form, with 95% confidence intervals around the point estimates. While the interpretation of AMCEs can be complicated when reference categories are arbitrary (Leeper, Hobolt and Tilley 2019), our key dimensions include non-arbitrary reference categories (baseline statements). Supplemental models using alternate specifications produce consistent findings. RESULTS While the mean of prioritization is 0.5 by design, the average level of support for state intervention (mean=5.05, s.d.=1.6, median=5, 1-7 scale) suggests a fairly high degree of concern about distant labor and environmental problems. Below, we examine the effects of each of our treatments on these judgments in the full sample, then examine heterogeneity by political orientation, and finally explore open-ended justifications. Advocacy Frames Figure 3 shows ACMEs of advocacy frames and corporate promises on the likelihood of a scenario being prioritized for state intervention (left panel) and the degree of support for state intervention (right panel). (Results are based on a model with all treatments but isolated for ease of presentation; we report the results in table form in the appendix (Table A3).) Consistent with claims from the transnational advocacy literature, advocacy frames do shape judgments, though some more consistently than others. When the advocacy statement contained an anti-corporate critique—accusing companies of wanting to cover up their relentless pursuit of profit—the likelihood of that scenario being prioritized for state intervention increased by an average of 2.8 percentage points (p=0.016, 95% CI[0.5 - 5.1]) compared to the baseline statement, which only observed that the problem is common. The anti-corporate critique also influenced the degree of support for regulation, generating an average increase of 0.11 points on our 7-point scale (p=0.004, CI[0.04 - 0.18]). Similarly, the statement about scientific evidence of serious harms increased the likelihood of that scenario being prioritized by 4.3 percentage points (p<0.001, CI[2.0 - 6.5]) and increased support for state intervention by 0.078 points on average (p=0.043, CI[0.002 - 0.15]). While these effects do not represent dramatic shifts in opinion, they are notable given that the advocacy statements were embedded in a mix of other information.

6 We rejected 334 responses that showed clear evidence of complete disengagement with the survey or automated processing, defined as satisfying all of the following conditions: failing the instruction check, failing both problem/product attention checks after the scenarios, and answering the open-ended rationale questions with nonsense characters or words with no bearing on the question (e.g., “very good,” “okay fine”). Dynata added additional respondents at our request.

14 Figure 3: Effects of Advocacy Frames and Corporate Promises on the Perceived Importance of State Intervention

Advocacy statements that mobilized personal testimony or explicitly called for government intervention shaped the likelihood of a scenario being prioritized, but they did not have discernible effects on the degree of support for state intervention. The testimony statement increased the likelihood of a scenario being prioritized by 2.4 percentage points (p=0.040, CI[0.1 - 4.7], and the call for government action increased the likelihood of prioritization by 3.2 percentage points (p=0.007, CI[0.9 - 5.5])—both relative to the baseline statement. Contrary to the claim that explicit foregrounding of the state should provide a powerful alternative to private politics, we find no evidence that the state-centered advocacy statement has a larger effect than the simple corporate critique. In fact, the corporate critique has a larger effect than the call for government action on the degree of support (F(1,4488)=12.96, p<0.001), though there is no difference for prioritization (F(1,4488)=0.09, p=0.758). Corporate Promises All three corporate assurances of private regulation negatively affect the likelihood of a scenario being prioritized for state intervention and the degree of support for state intervention. When companies promised to set rules for their suppliers and audit compliance, it reduced the likelihood of a scenario being prioritized for state intervention by an average of 6.0 percentage

15 points (p<0.001, 95% CI[-8.4 - -3.7]) compared to the baseline corporate statement, in which companies simply say they are weighing the evidence. The promise of rules and auditing decreased support for state intervention by an average of .12 points on our 7-point scale (p=0.002,CI[-0.19 - -0.04]). A seemingly more credible form of private regulation, in which companies partner with nonprofit organizations to certify suppliers’ compliance, had similar effects. It reduced the likelihood of a scenario being prioritized for state intervention by 6.7 percentage points (p<0.001, CI[-9.2 - -4.5]) and decreased support for state intervention by .17 points (p<0.001, CI[-0.24 - -0.09]). When companies promised to cut ties to problematic suppliers, it reduced the chances of the scenario being prioritized by 5.1 percentage points (p<0.001, CI[-7.4 - -2.3]) and reduced support for state intervention by .11 points (p=0.002, CI[- 0.19 - -0.04]). These results support the hypothesis that corporate assurances of private regulation reduce the perceived importance of state intervention. Note that these effects, which are highly significant but modest in magnitude, are conditional on the uniform distribution of all other treatments, including two advocacy statements that explicitly cast doubt on the credibility of corporate promises. Contrary to the idea that partnering with nonprofit organizations should increase the credibility of assurances of private regulation over unilateral corporate actions, we do not find any significant differences among the private regulatory actions. (See Appendix Table A5 for all pairwise comparisons.) While scholars have extensively theorized credibility in private regulatory initiatives, these distinctions did not register in the court of public opinion. Individuals do differentiate, though, between assurances of private regulation and prosocial statements. The prosocial statement, in which companies espouse ethical conduct, corporate citizenship, charity, and community, did not have a discernible effect on either the likelihood of prioritization (p= 0.156) or the degree of support for state intervention (p=0.304), compared to the baseline corporate statement. In addition, compared with this prosocial statement, each of the private regulatory actions—unilaterally setting rules, partnering with nonprofits, and cutting ties to problematic suppliers—had a significantly greater negative effect on prioritization and support for state regulation. (See Appendix Table A5.) This pattern of effects is consistent with the claim that assurances of private regulation are more influential than prosocial statements but not with the claim that prosocial statements reduce support for state intervention. Figure 3 also shows that which corporations made a statement did not affect the likelihood of prioritization or degree of support. The one exception is that scenarios in which half of firms in the industry spoke were more likely to be prioritized for state intervention, compared to scenarios in which a single generic large retailer issued the response (b=2.5, p=0.029, CI[0.3 - 4.8]) or “nearly all” businesses issued the response (F(4,488)=3.86, p=0.049). To probe further, we examined interactions between these elements (estimating Average Component Interaction Effects (ACIEs)). (See Appendix Table A6.) Given the large number of combinations, these estimates are less precise and should be taken with caution. In general, we do not find evidence that corporate assurances of private regulation have larger effects when made by larger groups of firms. Nor do we find clear evidence that statements are more influential when made by a well-known firm—Target or Walmart—than by an unnamed large retailer. The single exception is that Walmart’s prosocial statements reduce the likelihood of prioritization (ACIE=-0.074, p=.047) and degree of support (ACIE=-0.244, p=0.035) to a greater extent than prosocial statements made by an unnamed large retailer. This raises intriguing

16 questions about how respondents judged the credibility of prosocial statements. We also examined whether corporate assurances might be less influential when accompanied by advocates’ critiques of corporations. These interactions are imprecise and do not reveal any clear evidence of this kind of countervailing effect.7 Advocacy frames and corporate promises both resonated, but their combination was not so salient as to drive perceptions amidst the array of information respondents were considering. Problems & Products Specific labor and environmental problems clearly affected perceptions of when and to what degree state intervention is needed, as shown in Figure 4. Compared to the reference category in which the problem is dangerous working conditions (roughly in the middle of the distribution of problem effects), we find several problems that significantly increase both prioritization and support for state intervention. Figure 4: Effects of Problems and Products on the Perceived Importance of State Intervention

7 We find only one significant interaction effect between advocacy and corporate statements (Table A7 in Appendix).

17 Child labor, modern slave labor, emission of toxic chemicals, and water pollution generated the most support for state intervention, with a bit of differentiation within this group. When the problem was child labor, the likelihood of the scenario being prioritized for state intervention was 18.1 percentage points higher (p<0.001; CI[14.9 - 21.2]) and the degree of support for state intervention was .40 points higher (p<0.001, CI[0.30 - 0.50]) than when the problem was dangerous working conditions. Modern slave labor had a similarly large effect on the degree of support for state intervention (b=0.33, p<0.001, CI[0.23 - 0.43]) and a slightly smaller effect on the likelihood of prioritization (b=11.3, p<0.001, CI[8.2 - 14.4]). Emissions of toxic chemicals inspired a similar likelihood of a scenario being prioritized for state intervention (b=11.0, p<0.001, CI[7.8 - 14.1]) and increased support for state intervention to a slightly lesser extent (b=0.21, p<0.001, CI[0.11 - 0.31]). Pollution of rivers and oceans had a similar effect on support for state intervention (b=0.19, p<0.001, CI[0.09 - 0.29]) and also increased the likelihood of prioritization (b=8.2, p<0.001, CI[5.1 - 11.3]).8 Note that the problem effects are larger in magnitude than effects of private regulation and advocacy frames, underscoring questions about how respondents perceive distant problems, which we address in a later section. Other problems generated much less interest in state intervention. Most notably, overuse of natural resources reduced the likelihood of a scenario being prioritized by 7.6 percentage points (p<0.001, CI[-10.8 - -4.5]) and reduced support for state intervention by .20 points (p<0.001, CI[-0.39 - -0.19]) compared to the reference category (dangerous work). When this was defined as illegal use of natural resources, the likelihood of prioritization and degree of support for state intervention was significantly higher (F(1,4488)=8.6, p=0.003 and (F(1,4488)=22.9, p<0.001, respectively), though still comparatively low. This partially supports the claim that an illegality frame increases the perceived importance of state intervention. Note, however, that the contrast between unpaid wages and its illegal analogy—wage theft by employers—worked in the opposite direction (F(1,4488)=21.1, p<0.001 for prioritization and F(1,4488)=6.58, p=0.010 for degree of support). Based on our reading of the open-ended responses, we suspect that the problem of wage theft was unfamiliar or sometimes misunderstood as theft by employees (despite the treatment reading “wage theft by employers”). This is perhaps suggestive of the practical difficulties of applying a criminal frame to workplace problems. Finally, the product in question did not shape when or to what degree state intervention was seen as important. Here we see precise null effects for electronics and seafood compared to clothing, suggesting that few respondents judged scenarios based on their taste or perception of different products. POLITICAL IDEOLOGIES AND DIVERGENT EVALUATIONS While recognizing widespread objections and approvals, some scholars of moral markets have emphasized population heterogeneity in perceptions of market processes and state controls (DiMaggio and Goldberg 2018). Given the politicization of knowledge claims and trust in

8 For prioritization, the effect of child labor is significantly larger than the effects of modern slavery and toxic emissions, which are significantly larger than the effect of water pollution. For degree of support, child labor and modern slavery are indistinguishable from each other but different from toxic emissions and water pollution, which are indistinguishable from each other.

18 institutions, political orientations would seem to play an especially important role in evaluating advocacy frames and corporate promises. Public opinion research suggests that conservatives should be less responsive to advocacy frames and more trusting of business statements than are liberals and moderates. Scientific expertise, testimony, and corporate critique are not necessarily anathema to conservative advocacy (Gross, Medvetz and Russell 2011, Whittier 2014), but conservatives on average have lower trust in science (Gauchat 2012, O’Brien and Noy 2015), environmental organizations (Brewer and Ley 2013), and labor unions (Newman and Kane 2017) than do individuals with moderate or liberal views. Conservatives also tend to have more trust in business (Leibrecht and Pitlik 2019). In the 2018 General Social Survey, strong or very strong conservatives were nearly twice as likely as those elsewhere on the political spectrum to express a great deal of confidence in business (33% vs. 17%).9 Principles of small government and voluntary provision of public goods may also make private regulation especially attractive to those on the right (Vandenbergh and Gilligan 2017).10 Attitudes about government regulation should also affect the degree to which people see private action by corporations as a reasonable substitute. The rise of private regulation has been fueled by skepticism about “command and control” regulation (Burgoon and Fransen 2018) and projects to “reinvent government” in search of efficiency (Abbott and Snidal 2009). If this is reflected in public opinion, then individuals with ambivalent or negative views of government regulation should be especially keen to let the private sector substitute for the state. In contrast, “true believers” in government regulation should be less swayed by private regulation and companies’ prosocial statements. To examine these conditional effects, we asked pre-treatment, pre-instructions questions about political ideology (using a scale of 1 for extremely liberal to 7 for extremely conservative) and attitudes about government regulation in general (based on a 7-point scale for agreement/disagreement with the statement that “government regulation of business is generally good for society”). To capture stark differences, we contrast strong conservatives (6 or 7 on our scale, 24% of the sample) with all others; and “true believers” in government regulation (6 or 7 on our scale, 27% of the sample) with all others (as specified in our pre-analysis plan).

9 A similar contrast between strong/very strong conservatives and others can be seen in questions about blaming Wall St. for poor economic conditions (American National Elections Survey, 2012) and confidence in organized labor (GSS, 2018). 10 There is mixed evidence of partisanship conditioning results in prior survey experiments (Druckman et. al. 2019; Malhotra et. al. 2019), but clear differences in views towards advocates and business in the public opinion literature.

19 Figure 5: Effects of Advocacy Frames and Corporate Promises by Political Orientations

Prioritization Degree of Support

Advocacy Frames

Common problem

Corp coverup

Testimony

Science

Do not trust Need gov

Corporate Promises

No comment

Prosocial

Strict rules

Partnership to certify

Cut supplier

-.1 0 .1 -.4 -.2 0 .2 Average Marginal Component Effect

Liberal/Moderate Conservative Pro-Regulation Not Pro-Regulation

Figure 5 shows the effects of advocacy frames and corporate promises in these split samples (full models in Appendix Table A8). Starting with advocacy frames, we find some evidence that conservatives are distinctive. In the analysis of prioritization, the liberal/moderate group is responsive to all advocacy frames, while the conservative group is not, but formal tests

20 show the interaction effects to be non-significant.11 In the degree of support, though, we see stark and significant differences between conservatives and others. Here, there are negative ACIEs for the combination of conservative and the explicit call for state intervention (p<0.001), testimony (p=0.012), and the corporate coverup message (p=0.043).12 In addition, advocacy statements with explicit calls for state regulation and testimony from affected individuals had negative effects on conservatives’ degree of support for state intervention—that is, leading them to greater opposition than when advocacy groups made only a neutral statement. Thus, when respondents were asked to register the intensity of their preferences, conservatives responded quite differently to most advocacy statements. Reactions to the scientific frame, however, were less clearly rooted in political ideologies.13 Turning to corporate promises, we find more similarity than difference, contrary to expectations. For instance, both conservatives and liberals/moderates became less likely to prioritize state intervention when companies promised to enforce strict rules for their suppliers or partner with nonprofits to certify compliance. In addition, support for state intervention decreased to a similar magnitude among both groups when companies partnered with nonprofits to certify compliance, despite conservatives’ skeptical views of advocacy organizations as reported above. In general, we find no significant interactions between political ideology and corporate promises.14 This result suggests that even groups that have comparatively low levels of trust in business give credence to corporate promises to rectify labor and environmental problems in the global economy. Similarly, those who are “true believers” in government regulation are not significantly different from others in their responses to private regulation and prosocial statements. These pro-regulation individuals became less supportive of state intervention (across both outcome measures) when companies promised to partner with nonprofits or cut problematic suppliers; and less likely to prioritize state intervention even when companies only promised to set strict rules for their suppliers. The lack of heterogeneous treatments here is confirmed with non-significant ACIEs (see Appendix Table A11) and omnibus F tests (prioritization: F(22,4488)=0.83, p=0.694; rating: F(22,4488)=1.25, p=0.196) and also persists with lower thresholds for being a supporter of government regulation (Table A11). Thus, although scholars and proponents often portray private regulation as a remedy for those with little faith in the state, we find no evidence of this. The appeal of private regulation in the global economy is much broader and not conditioned by people’s views about government regulation in general. Overall, these results reveal an asymmetry in responses to the crowded discursive space in which the global economy is debated. Only liberals and moderates respond to advocacy

11 Tests of the of ACIEs between ideology and advocacy statements do not show significant effects on prioritization. Omnibus test (F(22,4488)=0.99, p=0.478. All interaction terms in Appendix Tables A9 and A10. 12 Following Leeper et al. (2019), we use an omnibus F test, which shows that allowing the effects of advocacy frames and corporate promises to vary across conservatives and others increases the fit of the model (F(22,4488)=2.00, p=0.004). 13 The cut-point reflects our theoretical expectations and pre-registered analysis plan. Supplemental analyses with the threshold lowered by one and two points produce identical results with respect to the state-centered advocacy statement (Appendix Table A12) and suggest heterogeneous effects of the science frame if the definition of conservative is expanded. 14 As shown in the appendix, the only exception when splitting groups differently is that liberals are more responsive that moderates/conservatives to the corporate promise to cut off problematic suppliers.

21 frames by becoming more supportive of state intervention. But all groups respond similarly to corporate statements. Robustness Checks A series of analyses probe the robustness of our findings. First, we used two open-ended attention checks to increase engagement and identify inattentive respondents. Specifically, after rating a pair of scenarios, respondents were asked to type the name of one of the two problems/products shown in an open question, with “problems” or “products” randomly assigned and alternated for the second pair. We hand-coded to identify incorrect answers and re-ran analyses (consistent with our pre-registered plan) with the subsample who passed at least one of the checks (90% of the full sample). The estimates are substantively identical to those reported above (see Appendix Figure A4a), indicating that our results are not driven by noise from low- attention respondents. Similarly, the results were substantively identical when we excluded speeders, defined as those at the below 10th percentile in time for completion (Appendix Figure A4b). Second, although our sample approximates the U.S. population more closely than in many experimental studies, lack of lack of representativeness could bias our estimates if under/over-represented characteristics interact with the treatments (Mutz 2011). Our sample provides a reasonable approximation of the U.S. population in terms of geography (state of residence), gender, and age, but it over-represents those with higher levels of education and those who identify as white. (See Appendix, Tables A1-A2 for a full comparison.) We ran additional analyses using entropy balancing (Hainmueller 2012) to weight the sample to match key characteristics of the U.S. population (age, education, gender, and race/ethnicity, as specified in our pre-analysis plan). The estimates for the main models are somewhat less precise, but lead to identical conclusions about problems, products, and corporate promises (Appendix Figure A2). Two effects of advocacy frames (the effect of corporate critique on prioritization (p=0.054) and science on degree of support (p=0.193)) are less discernible using the weighted estimates. This suggests that the effects of advocacy frames may only generalize for particular types of judgments, with science and the explicit call for state intervention shaping priorities within a fixed choice set, while corporate critique most clearly moves the needle of support vs. opposition. The weighted version of the models of conditional effects (political ideology and support for regulation) produce identical conclusions. Third, we probed the assumptions of the conjoint analysis in several ways. To check balance on observables, we follow the recommended approach to regress treatments on a set of respondent attributes—gender, age, political ideology, education, and employment. We examine whether the features of the vignettes are jointly significant using omnibus F-tests (Hainmueller, Hopkins and Yamamoto 2014) and find no significant results, suggesting the randomization was effective. In addition, we ran analyses that controlled for these and other respondent characteristics. The effects of the treatments remained identical, while also revealing effects of gender, political ideology, and general attitude about regulation on support for state intervention (Appendix Figure A5 and Table A4). We report further robustness checks, including carryover and order effects, in the Appendix.

22 JUSTIFICATIONS FOR STATE INTERVENTION: EVIDENCE FROM OPEN-ENDED RESPONSES Further insight comes from responses to our open-ended question, which asked respondents to explain why they prioritized a particular scenario. This qualitative data adds depth to the findings and points toward new questions about logics of judgment and construals of harm. As a number of theorists have argued, discursive justifications reveal socially salient vocabularies of motive and worth (Boltanski and Thévenot 2006, Mills 1940, Tilly 2012). While theorists of justification have focused largely on broad logics of social order (e.g., Boltanski and Thévenot 2006), our respondents’ justifications tended to be narrowly focused on responsibility- taking and the perceived severity of different problems. For instance, one respondent justified state intervention based on the weakness of companies’ prosocial statement in one scenario, as contrasted to concrete action in the opposing scenario: “While both are important concerns, scenario 2 shows industry is taking steps to address the problem. Charity and ‘community engagements’ are nice—but again, it doesn't address the problem.” In other cases, respondents discussed why a distant problem was (or was not) particularly deserving of U.S. government intervention. For instance, in comparing pollution of rivers and oceans with dangerous working conditions, one person explained, “To me pollution is something that affects everybody. [T]hat is why I chose Scenario 2. [T]he other is more concentrated into the industries.” We developed a thematic coding scheme, starting inductively, and hand-coded a random sample of 1800 open-ended responses (20% of the total).15 (See Appendix.) As shown in Table 2, 14% of codable justifications argued for state intervention by noting the insufficiency of the voluntary business response. For example, respondents were concerned that corporations’ prosocial statement “is completely without personal responsibility” or “needs government intervention because the companies do not use an independent group to verify the standards.”16 This lends further credence to the quantitative finding that state intervention was seen as more important when companies did not provide concrete assurances of private regulation.17 The open-ended responses also show some people differentiating between more and less credible forms of private regulation, though this was too infrequent to come through in the estimates of ACMEs. As shown in Table 2, some respondents also discussed the proper role of government and the practicalities of implementing a ban, as well as a few respondents who described corporate actions as legitimating government intervention (Dana and Nadler 2019), though this was a tiny slice.

15 All coding by authors. 71.6% of the sampled statements were clear and detailed enough to be coded. We excluded unclear statements in analyses. Borderline statements were assigned codes only if both authors agreed that the statement was clear. Codes were not mutually exclusive. Statements were assigned an average of 1.1 codes (minimum 0, maximum 3). 16 Because we asked respondents to explain why state intervention was needed in the prioritized case (rather than about the non-prioritized case), respondents tended to discuss the lack of private regulation rather than its presence. 17 Additional analyses using LPMs confirm that the insufficiency of the business response was more likely to be cited as a justification for state intervention when the scenario included the “no comment” corporate statement (p<0.001) or the vague prosocial statement (p<0.001) as compared to the private regulatory actions.

23 Table 2: Justifications for State Intervention: Themes in Open-Ended Responses

Theme Example (% of codable responses)

Insufficiency of the business “Target's response does not seem to have a strict plan/regulations in response place, while Walmart seems to have implemented new rules and is (14%) following up with suppliers.”

Problem’s Consequences for “Dangerous work conditions hurt and maim people, often Victims (7.3%) preventing them from working further to support their families.”

Problem Affects Americans “Low wages and wage theft by employers causes unfair

(4.2%) comp[etition] with wage earners in this country and has the effect of lowering wages in the U.S.A. for American workers.” Problem Affects Everyone “The entire planet will suffer from the overuse of natural resources, (32.6%) (8.3%) not just one community. I feel that the greater good of many outweighs the suffering of the few.” Other Problem “Emission of toxic chemicals is far more dangerous comparing the Consequences (14.2%) two scenarios and I would certainly not like to be surrounded by LOGIC OF CONSEQUENCES it.” Moral Terms used to “Child labor is immoral and it is disgusting. If corporations won't Describe the Problem do anything about it then the government should impose their (11.5%) power on this horrible problem.”

Problem Deserves “Seafood companies should never over use natural resources so Categorical Prohibition they can profit.” (5.5%)

(23.5%) Other Statement about “I believe in a fair wage for your labor, I don't believe it's is are LOGIC OF Inappropriateness of the [our] right to tell other countries how they use their natural Problem (7.7%) resources. On wages it must be freedom of choice to work for these APPROPRIATENESS companies, but if people are force[d] to work then our government must act.” Other problem-centered “The environment should be top priority right now.” rationales (20.7%) Proper role of government “I generally believe in a small government but children cannot fight (5.2%) for themselves and in my opinion that is an area the government should step in.” Practical reasoning about “I feel that emissions of chemicals are easier to regulate because feasibility (2.9%) they can be quantified.” Business legitimates gov. “Difficult choice. When a business group [s]ays it's a problem, it (0.2%) must be a big problem.” Other justification (7.2%) “My ex-husband’s family lived in a location that involved a Superfund site that was created by a company that produced electronics.”

24 More frequently, respondents discussed the particular labor or environmental problems at hand, and in doing so articulated a variety of rationales for state intervention. 7.3% of justifications described how the problem harmed people at distant points of production (e.g., workers or community members threatened by dangerous settings). One respondent reasoned, for example, that “Chemicals released into the air seem more problematic to me because it would be impossible for people living in the area to avoid breathing.” (See Table 2 for additional examples.) Less commonly, 4.2% of justifications described how the problem could spill over to harm people within the U.S.—whether as workers (e.g., “wage theft … has the effect of lowering wages in the U.S.A. for American workers”) or as consumers (e.g., “Toxic chemicals can also harm consumers as well as production workers”). Given the centrality of material interests in theories of trade protectionism, it is striking that these themes appeared so rarely. In 8.3% of justifications, respondents suggested that the problem could spill over to harm larger collectivities (e.g., “all of us,” “everyone around the globe”), essentially justifying state intervention on the basis of cross-border collective goods/bads. This included concerns that “the entire planet will suffer from the overuse of natural resources, not just one community” or that “the risk of spillover to workers in other communities as a result of abusive employers gaining a competitive advantage is tangible and serious.” In many other cases, respondents used explicitly moral terminology (e.g., “immoral,” “unconscionable”) or broader injunctions (e.g., “unacceptable anytime and anywhere,” “should not exist”), often without specifying concrete consequences for those affected. 11.5% of justifications described problems using an explicitly moral language of right vs. wrong, moral vs. immoral, good vs. bad/evil. This included statements that “child labor is immoral and it is disgusting” or that it is “majorly wrong in my book” if a person “does a hard job that takes most of their day and they do not get any sort of compensation.”18 Similarly, roughly half as often, justifications invoked broad categorical prohibitions, stating for instance that “child labor is unacceptable anytime and anywhere” or that “modern slave labor should never even be talked about!” These statements were often made in definitive ways that implied that further justification was not needed. Others invoked categorical distinctions (e.g., “Overuse and illegal use are different”), normative principles (e.g., “I believe in a fair wage for your labor”), or judgments of impropriety (e.g., “Companies make enough profit without stealing from their workforce”). After starting inductively, we turned to March and Olsen’s (2008) distinction between a “logic of consequences” and a “logic of appropriateness” to make sense of responses to problems. A logic of consequences involves some estimation of benefits or harms, while a logic of appropriateness draws on “prescriptions of what is socially defined as normal, true, right or good, without, or in spite of, calculation of consequences and expected utility” (March and Olsen 2008:690). This distinction, which is frequently invoked in research on organizations and institutions (e.g., Clemens 1993), helps to organize the vocabularies used by our respondents, while also allowing for variation within logics—such as the varying answers to the “consequences for whom?” question mentioned above. While scholars of international trade

18 We do not mean to imply that consequentialist logics do not have moral foundations. In this context, though, explicit moral language and consequentialist logics rarely appeared together (in only 17 cases). We developed a working list of moral terms based on widely used dichotomies (e.g., right/wrong, good/bad, fair/unfair) and normatively loaded terms found in our responses (e.g., unconscionable, horrible, despicable, abhorrent). Our full coding guide is available upon request.

25 have contrasted self-interest and sociotropic logics underlying policy preferences, our research suggests that consequentialism and appropriateness hold promise for explaining perceptions of distant problems. As shown in Table 2, approximately 1/3 of codable justifications discussed the problem using a logic of consequences of some sort. This commonly involved discussing harms at the point of production or harms that spill over to “all of us,” or discussing specific harms without specifying precisely where they are felt. Nearly a quarter of justifications described problems using a logic of appropriateness of some sort. This often meant using explicit moral terminology without specifying concrete harms or invoking non-consequentialist normative principles. Figure 6 shows that these logics are distributed unevenly across the problems presented in the scenarios. For each problem in a prioritized scenario, we calculated the proportion of all codable justifications that utilized each logic, following the codes in the table. (Appendix Figures A6 and A7 provide a disaggregated mapping.) As seen on the left side of the figure, a logic of appropriateness was far more likely than a logic of consequences to be used when justifying state intervention into problems of child labor, modern slave labor, and unpaid wages. While some people did employ a logic of consequences to explain why child labor is problematic (e.g., “children need to attend school and get educated”), this was relatively rare. Similarly, while respondents sometimes used a logic of consequences to justify intervention against unpaid wages (e.g., “their families are suffering because of this”), they more frequently drew on less consequentialist language of morality and fairness (e.g., “majorly wrong in my book,” “I don’t think people should have to be unpaid while others are making tons of money off their unpaid labor”).19 Figure 6: Logics and Problems in Open-Ended Justifications

19 Note that there were exceptions, in which labor abuses were described as spilling over to everyone (e.g., “Natural resources of another country should not be an issue for America. But slave labor prospering around the globe is an issue for everyone around the globe”).

26 As seen on the right side of the figure, a logic of consequences was far more likely to be used when justifying state intervention to address all four environmental problems in our study— water pollution, overuse or illegal use of natural resources, and emission of toxic chemicals. This often involved claims about how the problem could harm “all of us” or “the globe as a whole,” but it also included concerns about consequences for Americans (e.g., “Pollution in one country can spread around the world, so it can affect the U.S.A. and its inhabitants”) or others (e.g., “Pollution has many implications for drinking water and the health of many people”). In contrast, it was exceedingly rare for respondents to use a logic of appropriateness to justify state intervention into these problems. Despite the potential for viewing it as immoral or unfair to externalize environmental damage to distant sites of production (Clapp 2015), respondents rarely talked about environmental problems this way. Note that it is not exclusively environmental problems that evoked a logic of consequences. Dangerous working conditions were also more likely to be discussed with a logic of consequences (e.g., “a dangerous work place is going to result in unchecked deaths and dismemberment of employees”) than a logic of appropriateness. Nevertheless, the qualitative evidence reveals both a diverse array of justifications for state intervention and a striking divergence of vocabularies. CONCLUSION Global supply chains distance consumers from much of the environmental damage and labor exploitation experienced in the production process. Over the past three decades, transnational corporations have constructed vast supply chains, while activists have sought to expand the moral boundaries of markets by bridging these distances and evoking feelings of sympathy, solidarity, or interdependence. The peak globalization of the early 2000s is now faltering, as supply chains are being reconfigured (Gereffi 2018) and neoliberal scripts are withering under the weight of trade wars, nationalism, and domestic discord (Beckert 2020, Hopewell 2020). While states have long been central to the construction of global supply chains, a new wave of contestation is likely to center more fully on nation-states as active managers of global flows and enforcers of the moral boundaries of markets. Our study, rooted in an integrative analytical framework, reveals key situational components of popular judgments about this type of market restriction. Most notably, we find that corporate assurances of private regulation have the power to undermine support for state intervention, even in a crowded discursive arena. The effects of corporate assurances are apparent across more and less credible forms of private regulation and across ideologically and politically distinct groups. By contrast, the effects of advocacy frames are much more contingent. Advocacy frames focused on the corporate quest for profit or scientific evidence of serious harm do increase the perceived importance of state intervention, but other advocacy frames have more mixed effects, and none are effective for political conservatives. Corporate voices resonate more clearly, but our results also demonstrate that corporations do not gain political capital for mere symbolism: Prosocial language of corporate ethics, charity, and community does not decrease support for state intervention. At the same time, perceptions of state intervention vary substantially across different labor and environmental problems, independent of any corporate promises. It is not only shocking abuses such as child labor and forced labor but also problems of toxic emissions and water pollution that most prompt support for state intervention, especially in contrast to problems

27 like unpaid wages and the depletion of natural resources. Of the issues we examined, neither labor nor environmental issues dominate in importance. Yet the logics that Americans use to justify state intervention appear to differ for labor and environmental problems. A logic of consequences predominates when justifying intervention into environmental problems, while a logic of appropriateness is more central for most labor issues. While previous research has found divergent strategies in the negotiation of global labor and environmental standards (Evans and Kay 2008), our findings suggest that there are also differences in the public imagination of these issues. Labor abuses seem to trigger a language of immorality and unfairness that is strikingly absent for environmental abuses. Conversely, cross-border spillovers and concrete threats to well-being are more salient in the public imagination of environmental problems, in spite of evidence that the degradation of working conditions can also spill across borders (Blanton and Blanton 2016). In addition to raising questions about global imaginaries, this finding may carry implications for the study of practical moralities (Hitlin and Vaisey 2013). The moral lenses that Americans use when judging markets are varied and situational, but they may also be structured by implicit assumptions about when to use consequentialist or deontological conceptions of ethics. In sum, Americans perceive the moral boundaries of global markets with sensitivity to a subset of both labor and environmental concerns, deference to corporate promises of reform, selective attention to advocacy organizations, and through lenses that vary depending on the problem at hand. As policymakers renegotiate the rules for global trade, our results reveal a unique terrain of public opinion that could serve as a resource or constraint on reform: Observers of objectionable markets rarely reject strong state interventions outright or discuss them solely through a language of self-interest. But even across political divides, they tend to be pragmatic in wanting someone to address the problem, even if that means letting corporations take the helm. To expand the role of the state, policy entrepreneurs may need to either leverage private action (for instance, by encouraging corporate responsibility but penalizing performance failures) or render corporate assurances of private regulation the equivalent of symbolic prosocial statements. Of course, it remains to be seen what is unique to the U.S. and what would generalize to other affluent countries, including those that have institutionalized tighter connections between states and markets. Further research should also explore whether advocacy and corporate statements condition one another, looking further into whether advocacy organizations can effectively defuse corporate assurances with messages about hypocrisy or credibility. The moral markets literature has most often focused on commodification and obfuscated exchange within domestic markets. Our findings show the promise of moving in two further directions. First, when taken to the global level, the moral markets approach can provide a unique sociological alternative to dominant accounts of protectionist attitudes. At a minimum, it suggests that the standard way of measuring attitudes about globalization—asking about limiting imports to protect domestic jobs—is insufficient to capture either the range of salient concerns or the situational character of judgments. Notably, when asked to justify U.S. government intervention to address distant labor or environmental problems, respondents in our study rarely (less than 4% of the time) mentioned the protection of American jobs.

28 Second, our research demonstrates that it is not only obfuscating structures and gift- giving frames that reduce the objectionability of markets. Assurances of reform from market actors themselves diminish the perceived need for external market restriction. In our case, relatively modest assurances from corporations selling apparel, electronics, or seafood played this role, but a focus on assurances can also be further integrated into moral markets research. Potentially objectionable markets might be made more palatable, for example, by assurances from adoption agencies that compensation for birth-mothers is used for legitimate expenses (see Schilke and Rossman 2018) or assurances from unions of sex workers that their members are not exploited. Formal assurances of trust have proliferated in many arenas, as projects to trace and rate become more sophisticated (Power 2019). These assurances may be shaping not only market segmentation and status (Espeland and Sauder 2016) but also conceptions of moral order within evolving markets.

29 REFERENCES Abbott, Kenneth and Duncan Snidal. 2009. "Strengthening International Regulation Through Transnational New Governance: Overcoming the Orchestration Deficit." Vanderbilt Journal of Transnational Law 42:501-78. Ahler, Douglas J, Carolyn E Roush and Gaurav Sood. 2020. "The micro-task market for lemons: Data quality on Amazon’s Mechanical Turk." Working paper, 7/29/2020, presented at the Midwest Political Science Association meetings, 4/6/2019. Almeling, Rene. 2007. "Selling Genes, Selling Gender: Egg Agencies, Sperm Banks, and the Medical Market in Genetic Material." American Sociological Review 72(3):319-40. Amengual, Matthew, Greg Distelhorst and Danny Tobin. 2020. "Global Purchasing as Labor Regulation: The Missing Middle." Industrial and Labor Relations Review 73(4):817-40. Anderson, Elisabeth. 2018. "Policy Entrepreneurs and the Origins of the Regulatory Welfare State: Child Labor Reform in Nineteenth-Century Europe." American Sociological Review 83(1):173-211. Appelbaum, Richard P. and Nelson Lichtenstein. 2016. Achieving Workers' Rights in the Global Economy. Ithaca: Cornell University Press. Bair, Jennifer. 2008. "Analysing global economic organization: embedded networks and global chains compared." Economy and Society 37(3):339-64. ---. 2019. "Dialectics of dissociation." Dialogues in Human Geography 9(1):68-72. Balsiger, Philip. forthcoming. "The dynamics of ‘Moralized Markets’: a field perspective." Socio-Economic Review. Bandelj, Nina. 2020. "Relational Work in the Economy." Annual Review of Sociology 46. Bansak, Kirk, Jens Hainmueller, Daniel J. Hopkins and Teppei Yamamoto. 2018. "The Number of Choice Tasks and Survey Satisficing in Conjoint Experiments." Political Analysis 26(1):112-19. ---. 2019. "Beyond the breaking point? Survey satisficing in conjoint experiments." Political Science Research and Methods:1-19. Bartley, Tim and Curtis Child. 2014. "Shaming the Corporation: The Social Production of Targets and the Anti-Sweatshop Movement." American Sociological Review 79(4):653- 79. Bartley, Tim, Sebastian Koos, Hiram Samel, Gustavo Setrini and Nikolas Summers. 2015. Looking Behind the Label: Global Industries and the Conscientious Consumer. Bloomington: Indiana University Press. Bartley, Tim. 2018. Rules without Rights: Land, Labor, and Private Authority in the Global Economy. Oxford and New York: Oxford University Press. Beckert, Jens. 2020. "The exhausted futures of neoliberalism: from promissory legitimacy to social anomy." Journal of Cultural Economy 13(3):318-30. Berinsky, Adam J., Michele F. Margolis and Michael W. Sances. 2014. "Separating the shirkers from the workers? Making sure respondents pay attention on self-administered surveys." American Journal of Political Science 58:739-53. Best, Rachel Kahn. 2012. "Disease Politics and Medical Research Funding: Three Ways Advocacy Shapes Policy." American Sociological Review 77(5):780-803. Blanton, Robert and Shannon Lindsey Blanton. 2016. "Globalization and Collective Labor Rights." Sociological Forum 31(1):181-202.

30 Boghossian, Johnny and José Carlos Marques. 2019. "Saving the Canadian fur industry’s hide: Government’s strategic use of private authority to constrain radical activism." Organization Studies 40(8):1241-67. Boltanski, Luc and Laurent Thévenot. 2006. On Justification: Economies of Worth. Translated by C. Porter. Princeton, NJ: Princeton University Press. Boström, Magnus. 2006. "Regulatory Credibility and Authority through Inclusiveness: Standardization Organizations in Cases of Eco-Labelling." Organization 13(3):345–67. Brewer, Paul R. and Barbara L. Ley. 2013. "Whose Science Do You Believe? Explaining Trust in Sources of Scientific Information About the Environment." Science Communication 35(1):115-37. Burgoon, Brian and Luc Fransen. 2018. "Might Corporate Social Responsibility Hollow Out Support for Public Assistance in Europe?". British Journal of Industrial Relations 56(1):128-63. Bush, Simon R. and Peter Oosterveer. 2019. Governing Sustainable Seafood. London and New York: Routledge. Carlos, W. Chad and Ben W. Lewis. 2018. "Strategic Silence: Withholding Certification Status as a Hypocrisy Avoidance Tactic." Administrative Science Quarterly 63(1):130-69. Carpenter, R. Charli. 2007. "Setting the Advocacy Agenda: Theorizing Issue Emergence and Nonemergence in Transnational Advocacy Networks." International Studies Quarterly 51:99-120. Cashore, Benjamin. 2002. "Legitimacy and the Privatization of Environmental Governance: How Non-State Market-Driven (NSMD) Governance Systems Gain Rulemaking Authority." Governance 15(4):503-29. Chan, Cheris. 2009. "Creating a market in the presence of cultural resistance: the case of life insurance in China." Theory and society 38(3):271-305. Chan, Jenny, Pun Ngai and Mark Selden. forthcoming. Dying for an IPhone: Apple, Foxconn, and the Lives of China's Workers. Chicago: Haymarket Books. Clapp, Jennifer. 2015. "Distant agricultural landscapes." Sustainability Science 10(2):305-16. Clemens, Elisabeth S. 1993. "Organizational Repertoires and Institutional Change: Women's Groups and the Transformation of U.S. Politics, 1890-1920." American Journal of Sociology 98(4):755-98. Collins, Jane L. 2003. Threads: Gender, Labor, and Power in the Global Apparel Industry. Chicago: University of Chicago Press. Coppock, Alexander. 2019. "Generalizing from survey experiments conducted on Mechanical Turk: A replication approach." Political Science Research and Methods 7(3):613-28. Dana, David A and Janice Nadler. 2019. "Regulation, Public Attitudes, and Private Governance." Journal of Empirical Legal Studies 16(1):69-93. Darnall, Nicole, Hyunjung Ji and Diego A. Vázquez-Brust. 2018. "Third-Party Certification, Sponsorship, and Consumers’ Ecolabel Use." Journal of Business Ethics 150(4):953-69. Davis, Gerald F. 2009. Managed by the markets: How finance re-shaped America: Oxford University Press. Davis, Nancy J. and Robert V. Robinson. 2006. "The Egalitarian Face of Islamic Orthodoxy: Support for Islamic Law and Economic Justice in Seven Muslim-Majority Nations." American Sociological Review 71(2):167-90. de Souza Leão, Luciana and Gil Eyal. 2019. "The rise of randomized controlled trials (RCTs) in international development in historical perspective." Theory and society 48(3):383-418.

31 DiMaggio, Paul and Amir Goldberg. 2018. "Searching for Homo economicus: Variation in Americans’ construals of and attitudes toward markets." European Journal of Sociology/Archives Européennes de Sociologie 59(2):151-89. Espeland, Wendy Nelson and Michael Sauder. 2016. Engines of Anxiety: Academic Rankings, Reputation, and Accountability. New York: Russell Sage Foundation. Evans, Alice. 2020. "Overcoming the Global Despondency Trap: Strengthening Corporate Accountability in Supply Chains." Review of International Political Economy 27(3):658- 85. Evans, Peter. 1997. "The Eclipse of the State? Reflections on Stateness in an Era of Globalization." World Politics 50:62-87. Evans, Rhonda and Tamara Kay. 2008. "How Environmentalists Greened Trade Policy: Strategic Action and the Architecture of Field Overlap." American Sociological Review 73:970-91. Eyal, Gil. 2019. The Crisis of Expertise: Polity Press. Flores, René D and Ariela Schachter. 2018. "Who are the “illegals”? The social construction of illegality in the United States." American Sociological Review 83(5):839-68. Fourcade, Marion. 2011. "Cents and Sensibility: Economic Valuation and the Nature of “Nature”." American Journal of Sociology 116(6):1721-77. Gauchat, Gordon. 2012. "Politicization of Science in the Public Sphere:A Study of Public Trust in the United States, 1974 to 2010." American Sociological Review 77(2):167-87. Gereffi, Gary. 2018. Global value chains and development : redefining the contours of 21st century capitalism. Cambridge, : Cambridge University Press. Gross, Neil, Thomas Medvetz and Rupert Russell. 2011. "The Contemporary American Conservative Movement." Annual Review of Sociology 37(1):325-54. Hahl, Oliver, Ezra W Zuckerman and Minjae Kim. 2017. "Why elites love authentic lowbrow culture: Overcoming high-status denigration with outsider art." American Sociological Review 82(4):828-56. Hahl, Oliver, Minjae Kim and Ezra W Zuckerman Sivan. 2018. "The authentic appeal of the lying demagogue: Proclaiming the deeper truth about political illegitimacy." American Sociological Review 83(1):1-33. Hainmueller, Jens. 2012. "Entropy Balancing for Causal Effects: A Multivariate Reweighting Method to Produce Balanced Samples in Observational Studies." Political Analysis 20:25-46. Hainmueller, Jens, Daniel J. Hopkins and Teppei Yamamoto. 2014. "Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Experiments." Political Analysis 22:1-30. Hainmueller, Jens, Dominik Hangartner and Teppei Yamamoto. 2015. "Validating vignette and conjoint survey experiments against real-world behavior." Proceedings of the National Academy of Sciences 112(8):2395-400. Hainmueller, Jens, Michael J Hiscox and Sandra Sequeira. 2015. "Consumer demand for fair trade: Evidence from a multistore field experiment." Review of Economics and Statistics 97(2):242-56. Harcourt, Bernard E. 2010. "Neoliberal penality A brief genealogy." Theoretical Criminology 14(1):74-92. Healy, Kieran and Kimberly D. Krawiec. 2017. "Repugnance Management and Transactions in the Body." American Economic Review 107(5):86-90.

32 Hilgert, Jeffrey. 2013. Hazard or hardship: Crafting global norms on the right to refuse unsafe work. Ithaca: Cornell University Press. Hitlin, Steven and Stephen Vaisey. 2013. "The new sociology of morality." Annual Review of Sociology 39:51-68. Hoang, Kimberly Kay. 2018. "Risky Investments: How Local and Foreign Investors Finesse Corruption-Rife Emerging Markets." American Sociological Review 83(4):657-85. Hochschild, Arlie Russell. 2016. Strangers in Their Own Land: Anger and Mourning on the American Right. New York: The New Press. Hopewell, Kristen. 2020. Clash of Powers: U.S.-China Rivalry in Global Trade Governance. New York: Cambridge University Press. Horner, Rory. 2017. "Beyond facilitator? State roles in global value chains and global production networks." Geography Compass 11(2):e12307. Huff, Connor and Joshua D Kertzer. 2018. "How the public defines terrorism." American Journal of Political Science 62(1):55-71. International Labor Organization Declaration on Fundamental Principles and Rights at Work (Adopted June 18, 1998, Geneva 1998). Johnston, Josée and Shyon Baumann. 2010. Foodies: Democracy and Distinction in the Gourmet Foodscape. New York: Routledge. Jorgenson, Andrew. 2006. "Unequal ecological exchange and environmental degradation: A theoretical proposition and cross‐national study of deforestation, 1990–2000." Rural Sociology 71(4):685-712. Keck, Margaret and Kathryn Sikkink. 1998. Activists Beyond Borders: Trans-National Advocacy Networks in International Politics. Ithaca: Cornell University Press. Kenworthy, Lane and Jonas Pontusson. 2005. "Rising inequality and the politics of redistribution in affluent countries." Perspectives on Politics 3(3):449-71. Kerner, Andrew and Jane L. Sumner. 2020. "Salvation by good works? Offshoring, corporate philanthropy, and public attitudes toward trade policy." Economics & Politics 32(1):1-27. Kerrissey, Jasmine and Jeff Schuhrke. 2016. "Life chances: Labor rights, international institutions, and worker fatalities in the Global South." Social forces 95(1):191-216. Kinderman, Daniel. 2020. "The challenges of upward regulatory harmonization: The case of sustainability reporting in the European Union." Regulation & Governance 14(4):674-97. Kolcava, Dennis, Lukas Rudolph and Thomas Bernauer. 2020. "Voluntary business initiatives can reduce public pressure for regulating firm behaviour abroad." Journal of European Public Policy:1-24. LeBaron, Genevieve and Andreas Rühmkorf. 2019. "The domestic politics of corporate accountability legislation: struggles over the 2015 UK Modern Slavery Act." Socio- Economic Review 17(3):709-43. Leeper, Thomas J., Sara B. Hobolt and James Tilley. 2019. "Measuring Subgroup Preferences in Conjoint Experiments." Political Analysis 28(2):207-21. Leibrecht, Markus and Hans Pitlik. 2019. "Is confidence in major companies rooted in generalized social trust, or regulatory quality, or both?". Journal of Institutional Economics:1-17. Leipold, Sina, Metodi Sotirov, Theresa Frei and Georg Winkel. 2016. "Protecting “First world” markets and “Third world” nature: The politics of illegal logging in Australia, the European Union and the United States." Global Environmental Change 39:294-304.

33 Lim, Alwyn and Kiyoteru Tsutsui. 2012. "Globalization and Commitment in Corporate Social Responsibility." American Sociological Review 77(1):69-98. Locke, Richard M. 2013. The Promise and Limits of Private Power: Promoting Labor Standards in a Global Economy. New York: Cambridge University Press. Longhofer, Wesley and Evan Schofer. 2010. "National and global origins of environmental association." American Sociological Review 75(4):505-33. Longhofer, Wesley, Evan Schofer, Natasha Miric and David John Frank. 2016. "NGOs, INGOs, and Environmental Policy Reform, 1970–2010." Social forces 94(4):1743-68. Malhotra, Neil, Benoît Monin and Michael Tomz. 2019. "Does Private Regulation Preempt Public Regulation." American Political Science Review 113(1):19-37. Mansfield, Edward D and Diana C Mutz. 2013. "US versus them: Mass attitudes toward offshore outsourcing." World Politics 65(4):571-608. Mansfield, Edward D, Diana C Mutz and Devon Brackbill. 2019. "Effects of the Great Recession on American attitudes toward trade." British Journal of Political Science 49(1):37-58. Manza, Jeff and Clem Brooks. 2012. "How Sociology Lost Public Opinion: A Genealogy of a Missing Concept in the Study of the Political." Sociological Theory 30(2):89-113. March, James G. and Johan P. Olsen. 2008. "The Logic of Appropriateness." in The Oxford Handbook of Public Policy, edited by R. E. Goodin, M. Moran and M. Rein. Oxford: Oxford University Press. McCall, Leslie, Derek Burk, Marie Laperrière and Jennifer A Richeson. 2017. "Exposure to rising inequality shapes Americans’ opportunity beliefs and policy support." Proceedings of the National Academy of Sciences 114(36):9593-98. McDonnell, Mary-Hunter and Brayden King. 2013. "Keeping up Appearances:Reputational Threat and Impression Management after Social Movement Boycotts." Administrative Science Quarterly 58(3):387-419. McDonnell, Mary-Hunter, Brayden G King and Sarah A Soule. 2015. "A Dynamic Process Model of Private Politics." American Sociological Review 80(3):654-78. McDonnell, Mary-Hunter and Brayden G King. 2018. "Order in the court: How firm status and reputation shape the outcomes of employment discrimination suits." American Sociological Review 83(1):61-87. Meyer, John W., Shawn M. Pope and Andrew Isaacson. 2015. "Legitimating the Transnational Corporation in a Stateless World Society." Pp. 27-72 in Corporate Social Responsibility in a Globalizing World, edited by K. Tsutsui and A. Lim. Cambridge: Cambridge University Press. Mills, C Wright. 1940. "Situated actions and vocabularies of motive." American Sociological Review 5(6):904-13. Morin, Jean-Frédéric, Andreas Dür and Lisa Lechner. 2018. "Mapping the trade and environment nexus: Insights from a new data set." Global Environmental Politics 18(1):122-39. Mullinix, Kevin J., Thomas J. Leeper, James N. Druckman and Jeremy Freese. 2016. "The Generalizability of Survey Experiments." Journal of Experimental Political Science 2(2):109-38. Mutz, Diana Carole. 2011. Population-based survey experiments. Princeton, N.J. ; Oxford: Princeton University Press. Newman, Benjamin J. and John V. Kane. 2017. "Economic Inequality and Public Support for Organized Labor." Political Research Quarterly 70(4):918-32.

34 O’Brien, Timothy L. and Shiri Noy. 2015. "Traditional, Modern, and Post-Secular Perspectives on Science and Religion in the United States." American Sociological Review 80(1):92- 115. Pellow, David N. and Hollie Nyseth Brehm. 2013. "An Environmental Sociology for the Twenty-First Century." Annual Review of Sociology 39(1):229-50. Polanyi, Karl. 1944. The Great Transformation. Boston: Beacon Press. Polletta, Francesca. 2014. "“It Was Like A Fever…” Narrative and Identity in Social Protest." Social Problems 45(2):137-59. Power, Michael. 2019. "Infrastructures of traceability." Pp. 115-30 in Thinking infrastructures (Research in the Sociology of Organizations), edited by M. Kronenberger, G. Bowker, J. Elyachar, A. Mennicken, P. Miller, J. Nucho and N. Polloc: Emerald. Quark, Amy A. 2016. "Ratcheting up protective regulations in the shadow of the WTO: NGO strategy and food safety standard-setting in ." Review of International Political Economy 23(5):872-98. Quinn, Sarah. 2008. "The Transformation of Morals in Markets: Death, Benefits, and the Exchange of Life Insurance Policies." American Journal of Sociology 114(3):738-80. Raess, Damian and Dora Sari. 2018. "Labor Provisions in Trade Agreements (LABPTA): Introducing a New Dataset." Global Policy 9(4):451-66. Rodrik, Dani. 2018. Straight Talk on Trade. Princeton: Princeton University Press. Rodrik, Dani and Rafael Di Tella. 2020. "Labour Market Shocks and the Demand for Trade Protection: Evidence from Online Surveys." The Economic Journal 130(628):1008-30. Sands, Geneva. 2019. "US halts import of goods suspected to have been made with forced labor." CNN, Oct. 1, 2019, https://edition.cnn.com/2019/10/01/politics/cbp-trade-forced- labor/index.html. Sassen, Saskia. 2006. Territory, Authority, Rights: From Medieval to Global Assemblages. Princeton and Oxford: Princeton University Press. Schilke, Oliver and Gabriel Rossman. 2018. "It’s only wrong if it’s transactional: Moral perceptions of obfuscated exchange." American Sociological Review 83(6):1079-107. Seidman, Gay. 2007. Beyond the Boycott: Labor Rights, Human Rights and Transnational Activism. New York: Russell Sage Foundation/ASA Rose Series. Shandra, John M, Christopher Leckband and Bruce London. 2009. "Ecologically unequal exchange and deforestation: A cross-national analysis of forestry export flows." Organization & Environment 22(3):293-310. Smith, Ted, David A. Sonnenfeld and David Naguib Pellow, eds. 2006. Challenging the Chip: Labor Rights and Environmental Justice in the Global Electronics Industry. Philadelphia: Temple University Press. Song, Andrew M., Joeri Scholtens, Kate Barclay, Simon R. Bush, Michael Fabinyi, Dedi S. Adhuri and Milton Haughton. forthcoming. "Collateral damage? Small-scale fisheries in the global fight against IUU fishing." Fish and Fisheries n/a(n/a). Soule, Sarah A. 2009. Contention and Corporate Social Responsibility: Cambridge University Press. Teisl, Mario F, Brian Roe and Robert L Hicks. 2002. "Can Eco-Labels Tune a Market? Evidence from Dolphin-Safe Labeling." Journal of Environmental Economics and Management 43(3):339-59. Thorlakson, T., J. F. de Zegher and E. F. Lambin. 2018. "Companies' contribution to sustainability through global supply chains." Proc Natl Acad Sci U S A 115(9):2072-77.

35 Tilly, Charles. 2012. Why? What Happens when People Give Reasons...and Why. Princeton: Princeton University Press. Tzankova, Zdravka. forthcoming. "Can private governance boost public policy? Insights from public–private governance interactions in the fisheries and electricity sectors." Regulation & Governance. Vandenbergh, Michael P. and Jonathan M. Gilligan. 2017. Beyond Politics: The Private Governance Response to Climate Change. New York: Cambridge University Press. Vasi, Ion Bogdan, Edward T. Walker, John S. Johnson and Hui Fen Tan. 2015. "“No Fracking Way!” Documentary Film, Discursive Opportunity, and Local Opposition against Hydraulic Fracturing in the United States, 2010 to 2013." American Sociological Review 80(5):934-59. Walker, Edward T. 2014. Grassroots for hire: Public affairs consultants in American democracy. New York: Cambridge University Press. Wallander, Lisa. 2009. "25 years of factorial surveys in sociology: A review." Social Science Research 38:505-20. Wherry, Frederick F., Kristin S. Seefeldt and Anthony S. Alvarez. 2019. "To Lend or Not to Lend to Friends and Kin: Awkwardness, Obfuscation, and Negative Reciprocity." Social forces 98(2):753-93. Whittier, Nancy. 2014. "Rethinking Coalitions: Anti-Pornography Feminists, Conservatives, and Relationships between Collaborative Adversarial Movements." Social Problems 61(2):175-93. World Bank. 2020. "World Development Report 2020: Trading for Development in the Age of Global Value." World Bank. Zelizer, Viviana. 2013. Economic Lives: How Culture Shapes the Economy. Princeton: Princeton University Press. Zhang, Chan and Frederick G. Conrad. 2016. "Intervening to Reduce Satisficing Behaviors in Web Surveys: Evidence From Two Experiments on How It Works." Social Science Computer Review 36(1):57-81.

36 Appendix 1. Sample description and weighting 2. Outcome measures 3. Robustness checks on main conjoint analysis 4. Additional analyses and interactions 5. Additional analyses of open-ended responses 6. Full transcript of survey instructions 7. Data availability

1 1. Sample

Our sample is a national diverse sample recruited through Dynata (formally Survey Sampling International). Dynata recruits participants through a variety of means online and invites them into the panel. There are not strict quotas, but the recruitment is structured to approximate the U.S. population in terms of education, age, geography, income, and gender. Table A1 reports basic descriptive statistics. Figure A1 shows the geographic distribution of the sample at the zip code level. Table A2 provides further data on geographic distribution at the state level.

In Figure A2, we use entropy balancing (Hainmueller 2012) to weight our data to match population characteristics described in Table A1. Following our preregistration, we balance on gender, age, education, and race. As described in the paper, the pattern of results is nearly identical using the weighted sample.

Table A1: Descriptive Statistics

Sample U.S. Population1 Gender Women 52.7% 51.6% Men 47.3% 48.4% Another gender identity 0.3% - Education Less than high school 4.4% 10.6% High school graduate 20.5% 28.3% Some college / Associates degree 34.3% 27.8% Bachelor’s degree 20.9% 21.5 Postgraduate degree 20.0% 12.0% Age 18-24 10.4% 11.6% 25-34 16.8% 18.0% 35-44 16.8% 16.4% 45-54 13.9% 16.2% 55-64 18.9% 16.7% 65-74 18.6% 12.6% 75+ 3.9% 8.5% Race/Ethnicity White 81.2% 63.1% Black or African American 6.2% 11.8% American Indian, Alaska Native, or Pacific 0.6% - 2 Islander Asian 3.8% 6.4% Latino or Hispanic 5.9% 16.5% Another racial/ethnic identity 1.6% 2.2% not listed above Note: N=4,489

1 Population figures come from the Current Population Survey, March 2019, as compiled by NORC. 2 Included in 2.2% other group below.

2 Figure A1: Respondent Locations (by ZIP code)

Note: Darker circles had more respondents per zip code. Maximum four respondents in a zip code.

3

Table A2: Geographic Distribution of Sample

% of % of U.S. % of % of U.S. State sample pop. State sample pop. Alabama 1.40 1.49 Montana 0.33 0.33 Alaska 0.16 0.22 Nebraska 0.58 0.59 Arizona 2.72 2.22 Nevada 1.31 0.94 Arkansas 0.80 0.92 New Hampshire 0.33 0.41 California 8.51 12.04 New Jersey 2.63 2.71 Colorado 1.51 1.75 New Mexico 0.65 0.64 Connecticut 1.14 1.09 New York 7.15 5.93 Delaware 0.33 0.30 North Carolina 3.12 3.20 Dist. of Columbia 0.05 0.22 North Dakota 0.20 0.23 Florida 7.73 6.54 Ohio 4.83 3.56 Georgia 2.90 3.23 Oklahoma 1.02 1.21 Hawaii 0.38 0.43 Oregon 1.83 1.29 Idaho 0.51 0.54 Pennsylvania 4.50 3.90 Illinois 3.88 3.86 Rhode Island 0.27 0.32 Indiana 2.14 2.05 South Carolina 2.12 1.57 Iowa 0.91 0.96 South Dakota 0.18 0.27 Kansas 0.91 0.89 Tennessee 2.70 2.08 Kentucky 1.85 1.36 Texas 5.75 8.83 Louisiana 1.00 1.42 Utah 1.16 0.98 Maine 0.42 0.41 Vermont 0.20 0.19 Maryland 1.38 1.84 Virginia 2.99 2.60 Massachusetts 1.85 2.10 Washington 2.65 2.32 Michigan 3.25 3.04 West Virginia 0.74 0.55 Minnesota 1.51 1.72 Wisconsin 2.07 1.77 Mississippi 0.51 0.91 Wyoming 0.22 0.18 Missouri 2.36 1.87 missing zip code 0.36

4 Figure A2: Weighted versus Unweighted

5 2. Outcome Measures

We use two related outcome measures: a dichotomous measure of prioritization and a rating of degree of support. As expected, on average respondents did support state intervention more highly in the scenarios that they prioritized for intervention. The mean ratings for scenarios that were not selected was 4.6 (S.D. 1.6) and for scenarios that were selected was 5.5 (S.D. 1.49). The two measures are correlated (r=0.27, p<0.001). However, there is variation in ratings across the choices, indicating that these two measures picked up two distinct types of judgment. Figure A3 shows the distribution of responses to the ratings question, separated by the respondents’ selection to the prioritization question. Figure A3: Distribution of Degree of Support (By Prioritization)

Not Selected Selected 30 20 Percent 10 0 1 2 3 4 5 6 7 1 2 3 4 5 6 7 Degree of Support Graphs by Choice

6 3. Robustness checks of estimation of ACMEs

In this section, we report our method of the cleaning the sample, our main results in table form (Table A3) and a series of robustness checks: (1) results with only attentive respondents, (2) order effects, (3) carryover effects, (4) models with controls, and (5) a conditional conjoint analysis of the prioritization outcome variable. Cleaning After a soft launch of the survey (the first 161 respondents), we observed that there were some responses that indicated either automated processing or evidenced complete non-engagement. Therefore, we paused data collection and amended our registration. As described in the paper (endnote 6), we adopted an approach that excluded responses that failed all three attention checks and for which the responses to the open-ended questions were nonsense characters or words that had no bearing on the question being posed (e.g., “very good” or “okay fine”). In total, we identified 334 such response and requested them to be replaced by the survey provider, resulting in our final sample of 4,489.

Attention

We repeat the main analyses above, excluding respondents who failed both of the problem/product checks (as specified in our pre-registration, ~10% of the sample). This rate of failure of attention checks is consistent with what has been found broadly in the literature (Berinsky, Margolis and Sances 2014). These analyses, shown in Figure A4a, reveal no substantive differences compared with the full sample.

In addition, we report analyses of a subsample that eliminates speeders, defined as those whose time to completion was below 10th percentile. Figure A4b shows that the results are unchanged.

7 Table A3: Main Results

(1) (2) Prioritization Degree of Support Advocacy Frame (omitted: Common problem) Corp Coverup 0.0280* 0.110** (0.0117) (0.0381) Testimony 0.0241* 0.0324 (0.0117) (0.0380) Science 0.0426*** 0.0775* (0.0117) (0.0383) Do not trust; Need gov 0.0316** -0.0286 (0.0117) (0.0381) Corporate Promises (omitted: No Comment) Prosocial -0.0167 -0.0377 (0.0117) (0.0367) Strict rules -0.0602*** -0.119** (0.0119) (0.0379) Partnership to certify -0.0685*** -0.166*** (0.0118) (0.0375) Cut supplier -0.0506*** -0.114** (0.0119) (0.0376) Which Corporations (omitted: Large retailer) Target 0.000158 -0.0507 (0.0115) (0.0374) Walmart 0.0102 -0.0339 (0.0118) (0.0380) Half 0.0252* -0.0232 (0.0116) (0.0377) Nearly all 0.00249 -0.0442 (0.0117) (0.0379) Problem (omitted: Dangerous working conditions) toxic chemicals 0.110*** 0.211*** (0.0161) (0.0495) child labor 0.181*** 0.400*** (0.0160) (0.0519) modern slave labor 0.113*** 0.330*** (0.0159) (0.0521) unpaid wages -0.0131 -0.125* (0.0163) (0.0515) wage theft -0.0856*** -0.259*** (0.0160) (0.0516) water pollution 0.0823*** 0.186*** (0.0162) (0.0512) overuse of natural resources -0.0763*** -0.286*** (0.0159) (0.0504) illegal overuse of natural resources -0.0315* -0.0439 (0.0159) (0.0505) Product (omitted: clothing) electronics 0.00425 -0.0355 (0.00903) (0.0299) seafood 0.00673 -0.0196 (0.00907) (0.0299) Constant 0.471*** 5.105*** (0.0180) (0.0590) Observations 17,956 17,956 R-squared 0.035 0.024 Robust standard errors in parentheses clustered by respondent. *** p<0.001 ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests.

8 Figure A4a: All Respondents vs. Attentive Respondents

9 Figure A4b: All Respondents vs. No Speeders

10 Order Because we did not randomize the order of all elements (i.e., the problem and product always came first to make the narratives comprehensible), a full test of order effects is not possible. Instead, we randomized the order of the business and advocacy statements to ensure that conclusions regarding the differences between these two sets of statements were not driven by order. We check the effect of the ordering by interacting a dummy for the order (whether the business statement went first) with all the treatments and testing the joint significance of these interaction terms. We fail to reject the null hypothesis that there are no differences in the order of presentation for both the prioritization (F(22, 4488)=1.03; p=0.424) and the degree of support (F(22, 4488)=0.83; p=0.696) dependent variables. Carryover Although we had fewer choice tasks than is typical with conjoint experiments, we nevertheless test for carryover effects, interacting a dummy variable for the second round of choices with all the treatments and testing the joint significance of these interaction terms. We fail to reject the null hypothesis that there are no differences between order of choice tasks for both the prioritization (F(22, 4488)=1.17; p=0.27) and the degree of support (F(22, 4488)=1.15; p=0.28). Controls As reported in the main text, we did not detect any imbalances on observables. Given our randomized design, controls should not be necessary. Nevertheless, we re-run our analysis with controls for age, race/ethnicity, gender, education, ideology (measured before the treatment), and support for regulation (measured before the treatment). Figure A5 plots the main results with and without controls. As expected, the two are substantively the same. Table A3 reports the models with controls in full.

11 Figure A5: Models with and without Controls

12 Table A4: Full Models with Controls

(1) (2) Prioritization Degree of Support Activist Statements (omitted: Common problem) Corp Coverup 0.0281* 0.0954** (0.0117) (0.0370) Testimony 0.0244* 0.0413 (0.0118) (0.0369) Science 0.0426*** 0.0763* (0.0117) (0.0371) Do not trust; Need gov 0.0318** -0.0201 (0.0117) (0.0367) Corporate Promises (omitted: No Comment) Prosocial -0.0164 -0.0388 (0.0118) (0.0355) Strict rules -0.0602*** -0.125*** (0.0119) (0.0365) Partnership to certify -0.0687*** -0.180*** (0.0119) (0.0360) Cut supplier -0.0507*** -0.122*** (0.0119) (0.0365) Which Corporations (omitted: Large retailer) Target -5.17e-06 -0.0384 (0.0116) (0.0363) Walmart 0.0102 -0.0355 (0.0119) (0.0368) Half 0.0252* -0.0129 (0.0116) (0.0363) Nearly all 0.00246 -0.0418 (0.0117) (0.0366) Problem (omitted: Dangerous working conditions) toxic chemicals 0.110*** 0.207*** (0.0162) (0.0475) child labor 0.180*** 0.390*** (0.0160) (0.0505) modern slave labor 0.113*** 0.330*** (0.0160) (0.0510) unpaid wages -0.0131 -0.123* (0.0163) (0.0497) wage theft -0.0860*** -0.262*** (0.0161) (0.0500) water pollution 0.0819*** 0.183*** (0.0162) (0.0490) overuse of natural resources -0.0766*** -0.292*** (0.0160) (0.0487) illegal overuse of natural resources -0.0318* -0.0382 (0.0159) (0.0487) Product (omitted: clothing) electronics 0.00435 -0.0358 (0.00905) (0.0285) seafood 0.00673 -0.0159 (0.00909) (0.0286)

13 Age (omitted 18-24) Age 25-34 -0.00192 -0.0471 (0.00275) (0.0689) Age 35-44 -0.00118 -0.000804 (0.00280) (0.0681) Age 45-54 -0.00159 -0.0994 (0.00290) (0.0729) Age 55-64 0.00302 -0.0686 (0.00275) (0.0681) Age 65-74 -0.00230 -0.0430 (0.00283) (0.0684) Age 75+ -0.000114 -0.137 (0.00402) (0.0890) Education (omitted: less than high school) High School -0.000796 0.177+ (0.00362) (0.105) Some College / Associates 0.000329 0.147 (0.00352) (0.102) College Degree 0.00196 0.0319 (0.00369) (0.106) Postgraduate 0.00390 0.0289 (0.00374) (0.106) Race/Ethnicity (omitted: White) Black or African American -0.00366 -0.00837 (0.00298) (0.0738) American Indian, Alaska Native, or Pacific Islander -0.00731 0.295 (0.0104) (0.285) Asian 0.00102 0.103 (0.00381) (0.0874) Latino or Hispanic 0.00408 0.127 (0.00324) (0.0774) Bi-racial or Multi-racial -0.00369 0.167 (0.00554) (0.132) Another racial/ethnic identity not listed above -0.00142 0.148 (0.00805) (0.216) Gender (omitted: Men) Women (and other gender identity) 0.000270 0.222*** (0.00145) (0.0357)

Political Ideology (1=ext. lib.- 7=ext. cons) -0.000114 -0.0771*** (0.000438) (0.0113) General Support for Regulation (1-7 scale) -6.59e-05 0.229*** (0.000497) (0.0141) Constant 0.472*** 4.185*** (0.0188) (0.156)

Observations 17,952 17,952 R-squared 0.035 0.094 Robust standard errors in parentheses clustered by respondent. *** p<0.001, ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

14 Model Choice We estimate ACMEs in all of our analyses. Data from conjoint survey experiments with dichotomous outcomes can also be analyzed using a conditional logit model. To ensure that the results of the prioritization results are robust to other types of models, we report the results of a conditional logit model. Shown in Figure A6, the results are substantively unchanged. Figure A6: Conditional Logit Model

15 4. Additional analyses

Table A5: Comparison Among Business Statements Dependent Variable Comparison F p Degree of Support Prosocial Partnership 11.67 <0.01 Strict rules 4.64 0.03 Cut supplier 4.04 0.04 Partnership Strict rules 1.57 0.21 Cut supplier 1.82 0.18 Strict Rules Cut supplier 0.01 0.81 Prioritization Prosocial Partnership 19.35 <0.01 Strict rules 13.93 <0.01 Cut supplier 8.46 <0.01 Partnership Strict rules 0.49 0.48 Cut supplier 2.35 0.13 Strict Rules Cut supplier 0.68 0.41 Note: Wald tests. F(1, 4488). Models shown in main text, Figure 3.

16 We probed interactions among the attributes, specifically all combinations of: 1) business statements and which firms made the statements, and 2) business statements and advocacy statements. As described in the main text, we found little evidence of significant interactions.

Table A6: Business Statements and Which firms

(1) (2) Prioritization Degree of Support

Prosocial * Target 0.00933 -0.0991 (0.0362) (0.112) Prosocial * Walmart -0.0740* -0.244* (0.0373) (0.116) Prosocial * Half -0.0129 -0.0599 (0.0371) (0.116) Prosocial * Nearly All 0.0193 0.0235 (0.0366) (0.117) Strict rules * Target 0.00527 -0.103 (0.0361) (0.113) Strict rules * Walmart -0.0500 -0.153 (0.0374) (0.118) Strict rules * Half 0.00166 0.0384 (0.0369) (0.115) Strict rules * Nearly All 0.00173 -0.110 (0.0378) (0.120) Partnership to certify * Target 0.0391 0.0940 (0.0367) (0.117) Partnership to certify * Walmart -0.0253 -0.0272 (0.0370) (0.118) Partnership to certify * Half 0.0394 0.183 (0.0370) (0.117) Partnership to certify * Nearly All -0.0108 0.0552 (0.0373) (0.119) Cut supplier * Target 0.0188 -0.134 (0.0359) (0.119) Cut supplier * Walmart -0.0550 -0.162 (0.0366) (0.120) Cut supplier * Half -0.0261 -0.0499 (0.0368) (0.117) Cut supplier * Nearly All -0.0397 -0.136 (0.0363) (0.118) Observations 17,956 17,956 Note: Reported are Average Component Interaction Effects (ACIEs) as developed by Hainmueller et al. (2017). Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two- tailed tests

17 Table A7: Business Statements and Advocacy Statements

(1) (2) Prioritization Degree of Support

Prosocial * Corp coverup 0.0327 0.0369 (0.0366) (0.112) Prosocial * Testimony 0.0311 0.0320 (0.0373) (0.114) Prosocial * Science 0.0113 0.242* (0.0373) (0.116) Prosocial * Do not trust; need gov -0.00505 -0.0471 (0.0373) (0.115) Strict rules * Corp coverup 0.00293 0.0637 (0.0363) (0.112) Strict rules * Testimony 0.00617 -0.0225 (0.0368) (0.117) Strict rules * Science -0.0462 0.0620 (0.0365) (0.119) Strict rules * Do not trust; need gov -0.0232 0.0495 (0.0365) (0.114) Partnership to certify * Corp coverup -0.00490 -0.108 (0.0370) (0.117) Partnership to certify * Testimony -0.0426 -0.0213 (0.0370) (0.118) Partnership to certify * Science -0.0247 0.105 (0.0376) (0.120) Partnership to certify * Do not trust; need gov -0.0384 -0.0338 (0.0376) (0.118) Cut supplier * Corp coverup 0.00248 -0.151 (0.0367) (0.117) Cut supplier * Testimony 0.0166 -0.0983 (0.0369) (0.118) Cut supplier * Science 0.00682 0.126 (0.0364) (0.118) Cut supplier * Do not trust; need gov -0.0418 -0.0140 (0.0364) (0.116) Observations 17,956 17,956 Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests. We also explored alternative methods for identifying interactions (Egami and Imai 2019) but did not find evidence of theoretically meaningful combinations.

18 Table A8: Effects by Political Views

(1) (2) (3) (4) (5) (6) (7) (8) Prioritization Degree of Support Liberals / Conservative Pro- Not Pro- Liberals / Conservative Pro- Not Pro- Moderates Regulation Regulation Moderates Regulation Regulation Advocacy Frames (omitted: Common problem) Corp Coverup 0.0346** 0.00856 0.0492* 0.0196 0.150** -0.0444 0.0860 0.122** (0.0134) (0.0240) (0.0221) (0.0137) (0.0426) (0.0828) (0.0673) (0.0451) Testimony 0.0284* 0.0119 0.0195 0.0261+ 0.0813+ -0.157+ -0.0227 0.0648 (0.0136) (0.0234) (0.0229) (0.0137) (0.0425) (0.0824) (0.0698) (0.0444) Science 0.0538** 0.0104 0.0484* 0.0401** 0.111* -0.0336 0.0854 0.0870+ (0.0134) (0.0239) (0.0223) (0.0137) (0.0431) (0.0818) (0.0676) (0.0453) Do not trust; 0.0411** 0.00337 0.0601** 0.0216 0.0572 -0.315** 0.0511 -0.0371 Need gov (0.0134) (0.0238) (0.0223) (0.0137) (0.0426) (0.0809) (0.0698) (0.0444) Corporate Promises (omitted: No Comment) Prosocial -0.0199 -0.00716 -0.0144 -0.0175 -0.0745+ 0.0467 -0.0336 -0.0302 (0.0135) (0.0238) (0.0231) (0.0137) (0.0415) (0.0770) (0.0663) (0.0429) Strict rules -0.0608** -0.0588* -0.0660** -0.0582** -0.144** -0.0619 -0.0585 -0.144** (0.0137) (0.0242) (0.0225) (0.0140) (0.0424) (0.0820) (0.0653) (0.0446) Partnership to -0.0672** -0.0726** -0.0869** -0.0616** -0.168** -0.206* -0.161* -0.183** certify (0.0136) (0.0243) (0.0219) (0.0141) (0.0416) (0.0827) (0.0662) (0.0442) Cut supplier -0.0622** -0.0158 -0.0817** -0.0394** -0.123** -0.116 -0.171* -0.0930* (0.0138) (0.0235) (0.0231) (0.0138) (0.0423) (0.0801) (0.0685) (0.0440) Which Corporations (omitted: Large retailer) Target 0.00511 -0.0142 -0.00621 0.00321 -0.0473 -0.0447 -0.0379 -0.0513 (0.0132) (0.0238) (0.0220) (0.0136) (0.0412) (0.0831) (0.0703) (0.0432) Walmart 0.00390 0.0311 0.0299 0.00291 -0.0761+ 0.100 -0.0901 -0.0173 (0.0135) (0.0245) (0.0232) (0.0138) (0.0422) (0.0840) (0.0691) (0.0446) Half 0.0340* -0.000890 0.0342 0.0224+ -0.0114 -0.0364 0.0338 -0.0475 (0.0132) (0.0240) (0.0225) (0.0135) (0.0415) (0.0837) (0.0691) (0.0434) Nearly all -0.000613 0.0129 -0.00470 0.00581 -0.0765+ 0.0811 -0.0665 -0.0463 (0.0132) (0.0252) (0.0229) (0.0136) (0.0420) (0.0837) (0.0692) (0.0442) Problem (omitted: Dangerous working conditions) toxic 0.113** 0.0992** 0.144** 0.0970** 0.205** 0.242* 0.235** 0.197** chemicals (0.0186) (0.0325) (0.0302) (0.0191) (0.0539) (0.113) (0.0905) (0.0578) child labor 0.180** 0.184** 0.177** 0.182** 0.362** 0.521** 0.313** 0.431** (0.0182) (0.0332) (0.0302) (0.0188) (0.0579) (0.114) (0.0982) (0.0601) modern slave 0.120** 0.0917** 0.120** 0.111** 0.281** 0.474** 0.280** 0.355** labor (0.0183) (0.0327) (0.0298) (0.0188) (0.0580) (0.115) (0.102) (0.0599) unpaid wages -0.0139 -0.0131 0.0266 -0.0281 -0.108+ -0.197+ -0.0700 -0.145* (0.0186) (0.0339) (0.0313) (0.0191) (0.0574) (0.110) (0.0971) (0.0598) wage theft -0.0921** -0.0666* -0.0608* -0.0957** -0.249** -0.282* -0.267** -0.261** (0.0185) (0.0324) (0.0304) (0.0189) (0.0575) (0.112) (0.0973) (0.0596) water 0.0816** 0.0868** 0.108** 0.0728** 0.164** 0.277* 0.306** 0.135* pollution (0.0185) (0.0332) (0.0313) (0.0189) (0.0571) (0.111) (0.0916) (0.0601) overuse of -0.0792** -0.0654* -0.0362 -0.0916** -0.262** -0.382** -0.149 -0.340** natural (0.0182) (0.0333) (0.0307) (0.0186) (0.0557) (0.112) (0.0974) (0.0572) resources illegal overuse -0.0299 -0.0325 -0.0250 -0.0344+ -0.0587 0.0163 -0.00379 -0.0496 of natural (0.0183) (0.0318) (0.0300) (0.0187) (0.0559) (0.113) (0.0974) (0.0577) resources

19 Product (omitted: clothing) electronics 0.0148 -0.0283 0.00403 0.00458 -0.0268 -0.0735 -0.0866 -0.0182 (0.0103) (0.0188) (0.0174) (0.0106) (0.0325) (0.0677) (0.0527) (0.0351) seafood 0.0120 -0.00972 0.0140 0.00382 -0.0473 0.0533 0.0108 -0.0257 (0.0104) (0.0187) (0.0177) (0.0106) (0.0329) (0.0664) (0.0527) (0.0350) Constant 0.462** 0.498** 0.448** 0.480** 5.199** 4.842** 5.568** 4.925** (0.0206) (0.0369) (0.0343) (0.0212) (0.0654) (0.129) (0.106) (0.0693)

Observations 13,684 4,272 4,912 13,044 13,684 4,272 4,912 13,044 R-squared 0.037 0.032 0.035 0.036 0.022 0.039 0.022 0.028 Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

Below we also report ACIE’s to test for heterogenous treatment effects by ideology and support for regulation. Results for the other elements of the scenarios—which firms, problems, and products—are available upon request and included in the replication archive, as are results for interactions with support for regulation.

Table A9: Business Statements and Conservative

(1) (2) Prioritization Degree of Support Conservative * Prosocial 0.0141 0.128 (0.0273) (0.0874) Conservative * Strict Rules 0.00148 0.0788 (0.0278) (0.0921) Conservative * Partnership to Certify -0.00581 -0.0355 (0.0278) (0.0926) Conservative * Cut Supplier 0.0488+ 0.0145 (0.0272) (0.0908) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

20 Table A10: Advocacy Statements and Conservative

(1) (2) Prioritization Degree of Support Conservative * Corp Coverup -0.0278 -0.188* (0.0275) (0.0929) Conservative * Testimony -0.0181 -0.232* (0.0271) (0.0929) Conservative * Science -0.0449 -0.137 (0.0273) (0.0922) Conservative * Do not trust; Need gov -0.0394 -0.369** (0.0273) (0.0915) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

We also explored additional cutoffs for ideology (beyond our pre-registered analyses). Our initial measure of conservatives were those who are above 5 on our 7-point scale (1=extremely liberal, 7=extremely conservative). Below we report interactions with conservatives defined as those who were greater than 4 on our scale, and another measure that separate out those who were greater than 3 on our scale. For both approaches to coding ideology, omnibus F tests suggest that there are heterogenous treatments for degree of support (Ideology(>4): F(22,4488)=1.58, p=0.042; Ideology(>3): F(22,4488)=1.80, p=0.012, but not prioritization (Ideology(>4): F(22,4488)=1.17, p=0.260; Ideology(>3): F(22,4488)=1.33, p=0.141).

21 Looking specifically at business statements, we observe no significant interactions with ideology>4. We do, however, observe a positive interaction between the “Cut Supplier” statement and Ideology>3. This suggests that there is a greater negative effect of business promises to cut suppliers among liberals than among moderates/conservatives. None of the results suggest that liberals or moderates (who tend to express less trust in business than conservatives in survey research) are more skeptical of corporate promises in the current context.

Table A11: Business Statements and Ideology

(1) (2) (3) (4) Prioritization Degree of Prioritization Degree of Support Support Ideology (>4) * Prosocial -0.0130 0.00634 (0.0248) (0.0788) Ideology (>4) * Strict Rules 0.00163 -0.00453 (0.0252) (0.0816) Ideology (>4) * Partnership to Certify 0.00748 -0.0632 (0.0253) (0.0808) Ideology (>4) * Cut Supplier 0.0369 0.0853 (0.0248) (0.0801) Ideology (>3) * Prosocial 0.00607 0.0996 (0.0250) (0.0747) Ideology (>3) * Strict Rules 0.0186 0.0666 (0.0253) (0.0779) Ideology (>3) * Partnership to Certify 0.0265 0.120 (0.0249) (0.0768) Ideology (>3) * Cut Supplier 0.0831** 0.173* (0.0252) (0.0773) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

22 Turning to advocacy statements, consistent with our main analyses, we find that conservatives rate the importance of state intervention lower than others when common advocacy frames are used. Specifically, we find negative interactions between more conservative respondents (>4 or >3) and the “Do not trust; Need gov” advocacy statement. The other advocacy statements do not have consistent effects across the different approaches to measuring ideology. For prioritization, we do not find significant interactions.

Table A12: Advocacy Statements and Ideology

(1) (2) (3) (4) Prioritization Degree of Prioritization Degree of Support Support Ideology (>4) * Corp Coverup -0.0321 -0.116 (0.0247) (0.0818) Ideology (>4) * Testimony -0.00372 -0.115 (0.0246) (0.0820) Ideology (>4) * Science -0.0323 -0.171* (0.0249) (0.0829) Ideology (>4) * Do not trust; Need gov -0.0269 -0.293** (0.0248) (0.0812) Ideology (>3) * Corp Coverup -0.0159 0.00198 (0.0246) (0.0780) Ideology (>3) * Testimony -0.0270 -0.191* (0.0251) (0.0778) Ideology (>3) * Science -0.00193 -0.140+ (0.0245) (0.0777) Ideology (>3) * Do not trust; Need gov 0.00254 -0.168* (0.0243) (0.0786) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

For support of regulation, we also explored different cutoffs below the median. We ran models where support for regulation was defined as those greater than 4 (out of 7) on our scale, and those greater than 3 on our scale. We note that neither of these approaches to defining pro-regulation yielded significant interactions with the business statements. Results available upon request.

23 Table A13: Business Statements and Pro-Regulation

(1) (2) Prioritization Degree of Support Pro-Regulation * Prosocial 0.00244 -0.00308 (0.0268) (0.0789) Pro-Regulation * Strict Rules -0.00765 0.0929 (0.0265) (0.0791) Pro-Regulation * Partnership to Certify -0.0253 0.0216 (0.0260) (0.0797) Pro-Regulation * Cut Supplier -0.0419 -0.0686 (0.0269) (0.0814) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

Table A14: Advocacy Statements and Pro-Regulation

(1) (2) Prioritization Degree of Support Pro-Regulation * Corp Coverup 0.0298 -0.0381 (0.0260) (0.0808) Pro-Regulation * Testimony -0.00682 -0.0903 (0.0266) (0.0826) Pro-Regulation * Science 0.00820 -0.00731 (0.0261) (0.0811) Pro-Regulation * Do not trust; Need gov 0.0369 0.0816 (0.0261) (0.0826) Note: ACIE’s reported. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

24 In addition to the analyses reported in the main text, we also ran analyses that collapsed similar statements into a single variable to examine the combined effect. We did this for the private regulation (strict rules, partnership, and cut supplier) and advocacy statements (corporate coverup, testimony, science, and do not trust; need gov.). For private regulation, the effect is negative and significant in both the prioritization and degree of support (p<0.001). For activist statements, the effect is positive for prioritization (p=0.001), but there is no significant effect for degree of support (p=0.12). These results are congruent with our disaggregated analysis. We also pre-registered hypothesis regarding interactions among elements of the scenarios and collapsed categories. These estimates are more imprecise than we anticipated before collecting data. Nevertheless, we report all pre-registered analyses below. Table A15: Corporate Critiques & Business Responses

(1) (2) Prioritization Degree of Support

Corporate Critique * Prosocial 0.0140 -0.00251 (0.0322) (0.0979) Corporate Critique * Strict rules -0.0101 0.0585 (0.0316) (0.0978) Corporate Critique * Partnership to certify -0.0215 -0.0716 (0.0325) (0.102) Corporate Critique * Cut supplier -0.0198 -0.0834 (0.0318) (0.101) Note: Here the advocacy statements were trichotomized into the baseline statement, those that included a corporate critique (corp coverup and do not trust; need gov), and those that emphasized the issue (testimony and science). ACIE’s for the corporate critique and business statements are shown. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

25 Table A16: All Private Regulation & Advocacy Statements

(1) (2) (3) (4) Prioritization Degree of Prioritization Degree of Support Support

All Private Regulation * Corp coverup -0.0165 -0.0824 (0.0237) (0.0754) All Private Regulation * Testimony -0.0232 -0.0650 (0.0239) (0.0738) All Private Regulation * Science -0.0269 -0.0253 (0.0235) (0.0766) All Private Regulation * Do not trust; Need -0.0329 0.0202 gov (0.0235) (0.0749) All Private Regulation * Corporate Critique -0.0246 -0.0330 (0.0204) (0.0654) Note: ACIE’s for selected interactions shown. For Models 1 and 2, corporate statements are dichotomized into those that promised private regulation (rules, partnership, and cut supplier) versus all others. For Models 4 and 5, the advocacy statements were trichotomized as in table A15. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

Table A17: All Private Regulation & Which Firms

(1) (2) Prioritization Degree of Support

All Private Regulation * Target 0.0153 -0.00114 (0.0236) (0.0741) All Private Regulation * Walmart -0.00574 0.00967 (0.0240) (0.0757) All Private Regulation * Half of all 0.0111 0.0867 (0.0237) (0.0737) All Private Regulation * Nearly all -0.0264 -0.0767 (0.0238) (0.0755) Note: ACIE’s for selected interactions shown. Corporate responses are dichotomized as in table A16. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

26 Table A18: All Private Regulation & Political Orientations

(1) (2) (3) (4) (5) (6) Prioritization Degree Prioritization Degree Prioritization Degree of of of Support Support Support

All Private Regulation * Conservative 0.00802 -0.0424 (0.0178) (0.0608) All Private Regulation * Pro- -0.0266 0.0154 Regulation (0.0172) (0.0530) All Private Regulation * Conservative -0.0180 -0.0438 * Walmart (0.0275) (0.0851) Note: ACIE’s for selected interactions shown. Corporate responses are dichotomized as in tables A16-17. Full models available on request. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

Table A19: Advocacy Statements & Political Ideology

(1) (2) Prioritization Degree of Support

All Advocacy Statements * Conservative -0.0324 -0.2326** (0.0213) (0.7134) Note: ACIE’s for selected interactions shown. Advocacy statements are dichotomized into all advocacy frames versus the baseline. Robust standard errors in parentheses clustered by respondent. ** p<0.01, * p<0.05, + p<0.1 Two-tailed tests

27 5. Additional Analysis of Open-Ended Responses

To explore patterns in the open-ended justifications for state intervention, we first selected several small samples to read and use to develop a set of thematic coding categories. While several of our thematic codes were based on theoretical concerns (e.g., on the insufficiency of the business response), most were developed inductively and through an iterative process that eventually led us to contrast a logic of consequences and logic of appropriateness in assessing the problem at hand. Ultimately, we developed a coding guide detailing 13 possible codes to be assigned, 7 of which could be categorized as demonstrating a logic of consequences or logic of appropriateness. (Table 2 in the paper shows all codes and examples. Coding guide available by request.)

We selected a random sample of 1,800 justifications to code. Each of the two authors coded approximately half of the justifications and flagged uncertain or borderline cases, which were then considered collectively.

Figure 6 in the paper shows how the logics of consequences and appropriateness are distributed across the nine labor and environmental problems in our study. Here, we disaggregate these into their components. Indicative of a logic of consequences: Problem’s Consequences for Victims, Problem Affects Americans, Problem Affects Everyone, Other Problem Consequences. For a logic of appropriateness: Moral Terms used to Describe the Problem, Problem Deserves Categorical Prohibition, Other Statement about Inappropriateness of the Problem.

Figure A7: Logic of Consequences Codes by Problem

28 Figure A8: Logic of Appropriateness Codes by Problem

29 6. Full Transcript of Survey Instructions

Background

We hear a lot these days about how products that are sold in the U.S. are made in harsh working conditions or environmentally damaging ways in places like India, Mexico, China, Vietnam, and other countries.

Some American policymakers have proposed banning items from being sold in the U.S. if they are made in especially harsh or damaging ways. (All companies in the U.S. would be prevented from selling products from operations where there are serious problems.)

This policy might help to reduce these problems. However, it could also be overly strict, reduce consumer choices, or take attention away from more important issues.

The question we want your help with is:

When would it be appropriate for the U.S. government to adopt this kind of policy?

Please continue to the next page, which describes how we will collect your opinions on this.

Instructions

To understand where the government’s priorities should be, we will show you pairs of scenarios describing controversies like the ones you might see in news reports.

These scenarios might differ in several ways, including:

-the Problem that has been found to be widespread

-the type of Product involved

-the Response of American Businesses that voluntarily chose to issue a statement

-Which American Businesses chose to issue this statement (while others in their industry remained silent)

-the Response of Advocacy Groups (such as prominent environmental, labor, or human rights organizations)

Note: You might also see some similar elements. This is normal.

Please read each description carefully. We will then show you a table comparing the two scenarios and ask you the following question:

In which scenario would you most want the U.S. government to take action (to adopt a policy banning the relevant products from being sold by anyone)?

We will then ask some follow-up questions and move on to another pair of scenarios.

It is important to pay attention to all of the information provided so that we can accurately understand your views. If you aren’t sure, please make the best choice you can with the information that is given.

30 6. Data Availability

All data and code will be publicly posted upon publication.

31 Works Cited

Berinsky, Adam J., Michele F. Margolis, and Michael W. Sances. 2014. "Separating the shirkers from the workers? Making sure respondents pay attention on self-administered surveys." American Journal of Political Science 58:739-53. Egami, Naoki, and Kosuke Imai. 2019. "Causal Interaction in Factorial Experiments: Application to Conjoint Analysis." Journal of the American Statistical Association 114(526):529-40. Hainmueller, Jens. 2012. "Entropy Balancing for Causal Effects: A Multivariate Reweighting Method to Produce Balanced Samples in Observational Studies." Political Analysis 20:25-46. Hainmueller, Jens, Daniel J. Hopkins, and Teppei Yamamoto. 2017. "Causal Inference in Conjoint Analysis: Understanding Multidimensional Choices via Stated Preference Experiments." Political Analysis 22(1):1-30.

32