Trade politics at the checkout lane: ethnocentrism and consumer preferences

As international trade flourishes, Americans can choose from an increasing number of foreign products even at their local grocery stores, allowing consumers to directly experience the consequences of globalized trade in a simple and intuitive way that does not require much political expertise. Yet, most prior scholarship on political consumerism assumes that consumers are aware of the political and economic implications of their choices at the checkout lane. We move away from this assumption, focusing instead on more fundamental psychological predispositions such as ethnocentrism that may guide daily consumer choices. Using a discrete choice conjoint experiment, we show that Americans, on average, exhibit ethnocentric consumer preferences, with demand for products falling as they are produced in more culturally and ethnically distant places. Additionally, we show that this effect is more pronounced among those with higher levels of ethnocentrism. Our results provide evidence for a “ naïve ” form of political consumerism.

We replicate past country-of-origin findings, showing that Americans, on average, prefer to consume domestically produced goods over otherwise identical foreign goods and are willing to pay a premium to do so. Second, we show that while Americans on average have a home bias in consumption, ethnocentric Americans prize domestically produced goods over foreign goods significantly more than their less ethnocentric compatriots. Finally, we show that this effect is larger for countries that are relatively more culturally and ethnically distinct from the United States. Our findings are similar when we employ alternative measures of ethnocentrism and after controlling for education, partisanship, gender, age, and income. Taken together this pattern of results suggests a dispositional effect of ethnocentrism that is independent of partisan or elite cues.

Non-economic and psychological origins of consumer choices
Traditionally, political consumerism is defined as a "…consumer choice of producers and products on the basis of attitudes and values that concern issues of […] ethical and political assessment of favorable and unfavorable business and government practice" (Micheletti, 2002). Indeed, recent work on political consumerism shows that consumers rely on a variety of non-economic preferences when choosing which products to buy and sometimes consciously use their purchasing power to express their political or policy preferences. It is, thus, not surprising that scholars increasingly consider political consumerism a form of political participation. As Stolle et al. (2005) put it: "The study of political consumerism and other such action repertoires therefore forces us to expand the number and types of political targets citizens choose for their engagement" (p. 263). These targets oftentimes include, but are not limited to, corporations as well as general labor and production practices (Stolle et al., 2005).
In the United States, for example, Nike was the target of a boycott in the 1990s because of poor labor conditions in their production facilities (Bennett, 2004). This boycott was part of a larger, international, anti-sweatshop movement that involved, among others, consumer organizations, labor unions, and international human rights organizations (Micheletti and Stolle, 2007). More recently, Chick-fil-A found itself the target of a boycott due to its support of charities with anti-LGBTQ stances (Becker and Copeland, 2016). Similarly, consumer choices can reflect general partisan preferences, whereby consumers reward or punish corporate political activity , oftentimes initiated by party elite messaging . Consumer choices are also impacted by international politics: Pandya and Venkatesan (2016) show that Americans responded to the French refusal to support the 2003 US invasion of Iraq with largely disorganized boycotts of products with "French sounding" brand names. There is also a long line of research in marketing documenting similar "non-performance" country-of-origin effects (Bilkey and Nes, 1982;Verlegh and Steenkamp, 1999;Wilcox, 2005), finding evidence that economic and political tensions with particular countries drive consumers away from goods produced in those same countries (see Kock et al., 2019). 2 Political consumerism of this kind, however, requires levels of political knowledge and interest (Endres and Panagopoulos, 2017;Copeland and Boulianne, 2020) that many consumers do not possess-a notion that is consistent with the broader public opinion literature showing that many individuals lack coherent political ideologies (Converse, 2006;Kinder and Kalmoe, 2017) and political awareness (Delli Carpini and Keeter, 1996). Survey data, for example, reveal that only a quarter of Americans have stopped using products from a company because of its political leanings (Ipsos, 2017). At the same time, the small share of Americans who do actively engage in political consumerism-either in the form of buycotting or boycotting-tend to be more educated and politically more knowledgeable (Baek, 2010). Thus, most Americans may not be explicitly motivated by politics when making daily consumer decisions. Yet, we argue that politically relevant predispositions like ethnocentrism influence consumers even in more or less apolitical contexts. Americans' attitudes toward free trade are connected to psychological predispositions, often at least as much as they are to more general economic factors (Mansfield and Mutz, 2009;Mansfield et al., 2016). Relevant predispositions include American identity, nationalism, and ethnocentrism-all of which correlate with hostile-free trade attitudes. We test whether these relationships hold true for concrete consumer choices and not just abstract economic policy preferences.
From this vantage point, we study a "naive" form of political consumerism that might not be driven by explicitly political considerations but rather by psychological predispositions that have political implications. For example, firms strategically choose their production locations and decide to emphasize or de-emphasize a product's origin through their advertising efforts, thereby shaping people's perceptions of the product, its country of origin, as well as the nature of interdependent, global economies. At the same time, ethnocentrism is increasingly politicized (i.e., "American First") even without any concrete grievances against certain brands and companies. As such, understanding how these predispositions shape behavior is important because they affect the strategies, frames, and appeals that political leaders themselves choose (Mutz, 2021).
The above discussion leads us to two hypotheses about how ethnocentrism shapes consumer behavior. First, when choosing between otherwise identical goods, consumers will have a "home bias" such that, on average, they will prefer goods produced in the United States to those produced abroad (H1). Second, we predict that ethnocentrism will moderate this preference, with those exhibiting relatively high levels of ethnocentrism having a greater home bias in consumption compared to those with relatively low levels of ethnocentrism (H2).

Data and methods
The data were collected as a two-wave panel survey by the research firm Bovitz, Inc. 3 Bovitz recruits and maintains an opt-in panel of potential survey participants, similar to firms like YouGov. Participants who opt-in to a survey are compensated for their participation, but this compensation does not require them to complete the survey. Bovitz matched our sample to the United States population on the dimensions of race, ethnicity, gender, household income, and age. Overall, our sample is more educated, more Democratic, and younger than the national average (see online Appendix Table A1).
We rely on the panel approach to minimize concerns that pre-treatment measures of moderators and controls would affect our estimation of treatment effects. Wave 1 (N = 1619) was fielded in early April 2020 and contained the questions we used to measure ethnocentrism. These questions were adapted from a scale proposed and validated by Neuliep and McCroskey (1997). This scale included items such as "People in other cultures could learn a lot from people in my culture" and "Lifestyles in other cultures are not as valid as those in my culture" (see online Appendix for full scale). In the analyses that follow, we estimate the conditional effect of ethnocentrism using a binning estimator approach similar to Hainmueller et al. (2019). 4 Wave 2 (N = 995, see Table A2), fielded in early June 2020, contained a conjoint experiment to assess how ethnocentrism affects purchasing decisions (see Appendix for a brief justification of our design choice). The conjoint procedure began by informing participants that they would be asked to choose between two similar products. This description also informed them that each product comparison would include a brief description of the product, its price, country of origin, and quality rating. 3 Bovitz, Inc. is used in other political science research (e.g., Druckman and Levendusky 2019). 4 The estimator relaxes the strong assumption of linear interaction effects by assigning respondents to bins based on which third of the ethnocentrism distribution they happen to fall into (low, medium, high) and interacting indicator variables for each bin with the treatment independently.
In each round, respondents were asked, "if you had to purchase one of the products above, which one would it be" and they selected between "Product A" and "Product B" (see Figure A1 for an example of the choice task). The products varied randomly on the following dimensions: country of origin, price, and a rating of quality/consumer satisfaction. Country of origin could be the United States, Germany, China, or "a country outside of the United States". We selected Germany for its relative cultural similarity to the US, and China for its relative cultural dissimilarity. We expect ethnocentrists to be sensitive to these nuances in cultural similarity. The general category "a country outside of the United States" was included to obtain an estimate of the average effect for any foreign product rather than a country-specific effect. The quality/satisfaction rating could be three out of five, four out of five, or five out five stars, breaking any ex ante link between country of origin and perceptions of quality or reliability. Price markup could be 0, 25, 50, or 100 percent above the baseline price of a product.
The design included 12 products from four of categories: food, household supplies, appliances, and products used in previous research on consumer choices. Table A3 in the Appendix contains a list of each product and the corresponding baseline (0 percent price markup) price used in the experiment. We restricted our design to allow a respondent to evaluate a given product, e.g., a microwave, only once. Each respondent completed ten choice tasks, exposing each respondent to a random subset of ten of the 12 products. In all, the experiment yielded 19,910 individual product choices. All product attributes (country of origin, price, and quality rating) were fully randomized. This design allows for a direct test of country-of-origin effects relative to other relevant product attributes and the extent to which the pre-treatment measure of ethnocentrism moderates the effect of a product's country of origin on consumer choice. The Appendix includes a number of robustness checks. We find no evidence that respondents changed how they chose products based on the specific type of product they evaluated (see Figure A2). In addition, our results are robust across rounds and product categories (see Figures A3 and A4).

Results
Overall, participants prefer higher quality and lower priced goods, suggesting that they are reacting to the experiment reasonably (see Table A4). Figure 1 presents the results for our country of origin and ethnocentrism analyses. In panel 1, following recent work on interpreting conditional effects in conjoint designs (Leeper et al., 2020), we calculate the marginal mean or predicted probability that a respondent would choose a given product from a given country relative to all other potential countries of origin. The red squares represent the sample-wide estimated marginal mean while the green circles and blue diamonds represent marginal mean estimates for the sub-samples of high (>66th percentile) and low (<33rd percentile) ethnocentrists respectively. The results show that the public has a strong relative preference for domestically produced goods (H1). The strength of home bias preference appears greatest when the alternative product is produced in a more culturally distant country. The marginal mean probability of selection is about 0.67 (95 percent CI 0.65-0.68) for US goods, 0.52 (95 percent CI 0.51-0.54) for goods made in Germany, and 0.33 (95 percent CI 0.31-0.34) for goods made in China. Factors other than cultural proximity vary considerably across China and Germany, thus more conclusive evidence of the effect of ethnocentrism comes from comparing the buying choices of those with high levels of ethnocentrism (green squares) to those exhibiting low levels of ethnocentrism (blue diamonds). Here we see strong evidence of ethnocentric consumption (H2): Those exhibiting high levels of ethnocentrism have much stronger relative preferences for domestically produced goods and against goods produced in culturally distinct countries like China compared to their less ethnocentric compatriots. This combination of in-group bias and out-group antipathy is precisely what theories of ethnocentrism anticipate.
The marginal means provide initial evidence that those exhibiting high levels of ethnocentrism are put off relatively more by goods that are made abroad and that this effect is magnified as the cultural distance between the United States and the place of production increases. 5 We formally test this conjecture in panel 2 of Figure 1. Here we present estimates of the average marginal component effect (AMCE) of country of origin and the AMCE of country of origin conditional on respondent ethnocentrism. These quantities represent how much the probability of consumption would decline if the product's country of origin changed from the United States to Germany, outside the United States, or China averaging over the other product features. The magnitude of the AMCE of country of origin increases as the country becomes increasingly culturally distinct from the United States. As above, we see that this effect is moderated by respondents' level of ethnocentrism. Demand for goods drops off much more dramatically among high ethnocentrists than among low ethnocentrists as we move from a culturally similar country like Germany to a culturally distinct country like China.
To further illustrate these effects, we present marginal willingness to pay (MWTP) estimates relative to the baseline of a domestically produced product in panel 3 of Figure 1 (see online Appendix for the calculation details and disaggregated estimates in Figure A5). The results show that Americans are willing to pay more for domestically produced goods, but how much more depends on their level of ethnocentrism. Among those with low levels of ethnocentrism, a product from the United States would need to cost about 64 percent more than an otherwise identical good produced in China to make respondents indifferent between the two. This quantity is 34.4 (95 percent CI 19.6-51.3) percentage points higher for those with high levels of ethnocentrism. Products produced in Germany and "outside the United States" yield more modest estimates of MWTP for domestic production, but suggest similar support for our predictions; the difference in MTWP between high and low ethnocentrists is 25.3 (95 percent CI 12.4-39.8) percentage points and 17.7 (95 percent CI 5.4-31.2) percentage points for the respective treatment conditions. To put our results in more specific context, a recent incentive compatible field experiment on "made in USA" labels found that such labels increase the price that consumers are willing to pay by about 28 percent when shopping for cell phone screen protectors online (Kong and Rao, 2020). In our closest comparison to that analysis, we estimate the MWTP for a cell phone 5 The marginal means analysis requires that we subsample by level of ethnocentrism, but in the conditional AMCE analysis we directly estimate the interaction between ethnocentrism and country of origin. screen protector made in the United States relative to one made "outside the United States" to about 39 percent, suggesting that our experiment returns estimates of country-of-origin effects that are not exceptionally far outside the range of similar incentive compatible studies.
We pre-registered and fielded a conceptual replication of this experiment in February of 2021. The replication employed the same products, prices, and product ratings, but employed a different set of countries. 6 The new countries were the United States, Canada, Japan, China, and India. In this extension, we show that high ethnocentrists prefer produced in the United States over those from India and China and are willing to pay a greater price premium for domestic products compared to those from China or India (see online Appendix Figure A6). Reassuringly, in light of our theory, we found weaker effects for goods from Canada, a near cultural and ethnic twin of the United States. Notably, however, goods from Japan were not penalized as much as goods from China and India, suggesting that ethnocentric consumption is responsive to change in the political and economic features of the exporting country in addition to change in cultural proximity, as others have documented (Chen et al., 2021).
The similar results for China and India also alleviate concerns about the selection of China given the COVID-19 pandemic. If our results were driven solely by the pandemic, the results for China and India would likely not have aligned with each other. While Study 1 was conducted between April and June of 2020-the beginning of the pandemic-the conceptual replication was fielded in February of 2021. Despite this time gap and the corresponding changes in people's perception of the COVID-19 pandemic as well as a US presidential election, our results remain robust.
In sum, we find that high ethnocentrists are generally more sensitive to a product's country of origin than their less ethnocentric compatriots and, as such, are willing to pay more to consume domestically produced goods. Taken together, our results suggest robust evidence for the notion that ethnocentric predispositions drive consumer behavior. The findings we report here are robust to controlling for standard demographic covariates (e.g., party identification and income), when sub-setting our analyses by education level, with other measures of ethnocentrism, and with different distributional assumptions (see online Appendix Figure A7 and Tables A5-A8).

Discussion
Consumer decisions are one of the most direct, frequent, and individually consequential ways in which Americans "vote" on globalization. Using a conjoint experiment and two different samples, we found that across a wide range of products and price points, respondents showed a strong preference for goods from countries that are culturally and ethnically similar to the US. Crucially, this effect is magnified among those exhibiting relatively high levels of ethnocentrism. In short, when it comes to consumption, ethnocentrists appear quite willing to act on their protectionist preferences.
Critics might take issue with our experimental design which allowed for choices that are unlikely to be common in the real world (e.g., butter or cheese prominently labeled or marketed as being produced in China). Indeed, if our interests were in the precise real-world economic effects of ethnocentrism, we might have been better off focusing on observed consumer behavior (e.g., Helms et al., 2020) or a representative sample of real products that consumers are likely to encounter at the store. However, given that firms strategically choose what goods to produce, where to produce them, and how to market them, tests of the effects of ethnocentrism on consumer buying decisions using real-world products almost certainly suffer from strategic selection 6 In this replication, we also included a more detailed preamble into the experiment which included the statement, "All products meet all relevant U.S. consumer product health and safety standards", in order to dissuade participants from inferring that certain types of products were more dangerous when sourced from particular countries. If one considers biased beliefs about the health and safety of goods to be one effect of ethnocentrism, this preamble should also serve to make the experiment a harder test by dampening at least this form of ethnocentric consumer perceptions.
bias. To avoid this bias and generate estimates of the total effect of ethnocentrism on consumer preferences, we chose instead to expose respondents to a broad range of products that could be made either domestically or abroad. The same rationale motivated us to forgo any product brands. Admittedly, brands oftentimes convey useful information to consumers in the real-world, including a product's country of origin (e.g., Bosch, Rolex, and Prada). While our study examined the country-of-origin effect for generic products, future research in this area could explore whether or under what conditions a brand's reputation or advertising efforts are capable of overcoming consumer choices rooted in ethnocentrism (or other predispositions). In fact, some foreign brands may be able to take advantage of certain (positive) cultural or national stereotypes in product marketing (e.g., Swiss watches, French wine, or Belgian chocolate), potentially reducing the normally pernicious effects of ingroup favoritism. Future work might investigate whether the prevalence of this kind of "Swiss watch effect" varies with ethnocentrism as well.
The fundamental contribution of this study, however, is to provide evidence for a "naïve" political consumerism whose effects appear large enough to plausibly shape how firms choose production locations and/or market goods produced abroad for domestic consumption. Thus, the politics of consumption is not limited to contemporary political controversies of which only consumers with high political knowledge are aware but is embedded in the behavior of consumers in even apolitical contexts in which partisan or elite cues about the political implications of particular consumer choices are ill-defined or absent entirely.