Political Trust and American Public Support for Free Trade

Abstract

Debates over trade liberalization vs. protectionism have becoming increasingly relevant as the world moves through a contentious era of economic globalization. This is particularly true in the United States, where an elite consensus on the merits of free trade has fractured in recent years. While we know a good deal about the economic and cultural determinants of trade opinion, we know little about how attitudes toward government may matter. Here, I address this oversight by examining the relationship between political trust and trade support. I do this with cross-sectional and panel data from the American National Election Studies (ANES) and the National Annenberg Election Surveys (NAES), and a survey experiment fielded through Amazon’s Mechanical Turk (MTurk). Overall, I find that there is a positive and substantively significant relationship between political trust and mass support for free and open “pro-trade” policies. I attribute this to greater citizen confidence that government will pursue trade deals in the national interest and mitigate any perceived risks associated with free trade. These findings help us to better understand the determinants of public opinion toward trade policy and underscore the consequences of political trust.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11109-023-09858-x.

Debates over trade liberalization versus protectionism have become especially relevant as the world continues to move through a contentious era of globalization.1 In particular, the economic effects of trade have important implications for U.S. electoral politics (e.g., Autor et al., 2020; Baccini & Weymouth, 2021; Feigenbaum & Hall, 2015; Jensen et al., 2017; Margalit, 2011). Given the importance of trade in contemporary U.S. politics, we have more to learn about the factors that shape American public opinion toward trade policy, particularly when compared to what we know about the micro-level determinants of opinion toward policies regarding social welfare spending, immigration, and the so-called “culture wars.”2

Existing work on trade opinion has long focused on how much economic self-interest matters relative to sociotropic concerns and various psychological predispositions and social identities such as nationalism, cosmopolitanism, ideology, and partisanship (e.g., Hainmueller & Hiscox, 2006; Mansfield & Mutz, 2009; Rankin, 2001; Scheve & Slaughter, 2001). Past work has also focused on how issue framing, and more specifically on who is emphasized as the purported beneficiaries of free trade, shape public attitudes toward trade policy (e.g., Ardanaz et al., 2013; Brutger & Rathbun, 2021; Hiscox, 2006; Mutz & Kim, 2017; Rho & Tomz, 2017; Spilker Nguyen & Bernauer, 2020).3

While we have learned a good deal about the economic and cultural determinants of trade opinion (Ehrlich, 2018; Guisinger, 2017; Mutz, 2021), we know little about how attitudes toward government, the actor most responsible for crafting, implementing, and managing trade policy, may matter. Here, I address this oversight by examining the relationship between political trust and American public opinion toward trade policy. I do this by using cross-sectional survey data from the 2016 American National Election Study (ANES), pooled cross-sectional data from the 2004–2008 National Annenberg Election Surveys (NAES) and 1988–2012 Cumulative ANES, panel data from the 2016–2020 ANES, and a survey experiment fielded through Amazon’s Mechanical Turk (MTurk).

Overall, I find that there is a positive and substantively significant relationship between political trust and support for pro-trade policies, specifically for joining free trade agreements with other countries and for not imposing new limits on foreign imports. These findings help us to better understand the micro-level determinants of trade opinion while also underscoring the political consequences of political trust. They also help us to better understand why, despite the macroeconomic benefits that free trade can provide (e.g., Congressional Budget Office, 2016; Johns et al., 2015), many Americans do not strongly support free trade policies. Aside from ignorance about trade’s benefits (e.g., Rho & Tomz, 2017) or anti-foreigner sentiment (e.g., Mutz, 2021), anti-government sentiment may also play a role.

Why Political Trust Matters for Free Trade Support

Political trust, defined here as attitudes toward government as a whole, has important implications for democratic politics (Zmerli & van der Meer, 2017).4 Political trust promotes compromise and facilitates effective governance (Hetherington & Rudolph, 2015), in part by affording domestic governments with the leeway and flexibility to pursue costly and risky policies, i.e., those that are perceived to entail risks and to burden certain segments of the population with costs and risks that are not offset by clear, widespread, and/or tangible benefits (Hetherington, 2005).

Previous work (Hetherington, 2005, Chapter 5) provides a useful framework for thinking about the kinds of policies that are shaped by political trust (vs. those that are not). This framework argues that political trust is “activated,” that is, made to significantly matter in terms of shaping peoples’ subsequent evaluations, when the policy under consideration entails potential risks without also conferring widely distributed benefits, i.e., if people need to bear costs that are not sufficiently offset by clear benefits. In contrast, policies that entail few risks and/or demand few sacrifices from the mass public while also conferring benefits clearly and widely, are unlikely to “activate” political trust, i.e., trust will not significantly predict peoples’ attitudes toward said policies. Examples of policies that fall into the former category include affirmative action, welfare spending, and immigration (Hetherington & Globetti, 2002; Macdonald, 2021), while policies that fall into the latter category include spending on the environment, crime prevention, and Social Security (Hetherington, 2005; Rudolph & Evans, 2005). I argue that free trade falls into the former category, i.e., as a policy that will “activate” political trust.

I argue that this relationship (between political trust and trade support) should result because of the perceived risks and unclear benefits associated with free trade. When policies entail risks without providing clear, widespread, and/or tangible benefits, people should be less likely to support said policies. However, if they are more politically trustful, then they should be more likely to support domestic governments crafting, enacting, and implementing potentially risky public policies (e.g., Macdonald, 2021; Ryan et al., 2022). This is due, in part, to greater confidence that government, the actor most responsible for crafting, implementing, and managing public policies, will act in the national interest and mitigate any perceived risks associated with the policy under consideration. As such, politically trustful citizens should be more likely to afford government the leeway to pursue policies that entail perceived risks without clear, widespread, and tangible benefits, e.g., trade liberalization.

American political elites have long highlighted the benefits of free trade, claiming that it will increase economic prosperity and opportunities. For example, in his 1988 State of the Union Address, President Ronald Reagan advocated for continued trade liberalization saying “protectionism is destructionism” and that “America’s jobs, America’s growth, America’s future depend on trade—trade that is free, open, and fair.” Upon signing the North American Free Trade Agreement (NAFTA) in 1993, Bill Clinton said “NAFTA means jobs, American jobs, and good-paying jobs,” and argued that trade liberalization was crucial to America’s economic success in “an era in which commerce is global.” In 2000, President Clinton used similar rhetoric when he, along with many congressional Republicans, pushed for China’s entry into the World Trade Organization (WTO) arguing that it “creates a win-win result for both countries.” Clinton’s successors used similar rhetoric. In 2005, George W. Bush said, upon signing the Central American Free Trade Agreement (CAFTA), that “the agreement was good for America.” In 2015, Barack Obama argued that Trans-Pacific Partnership (TPP) “puts American workers first and will help middle-class families get ahead.”5

In contrast to these five recent U.S. presidents, other political elites have demonized U.S. free trade policies. Most notably, Donald Trump ran on protectionism in 2016, promoting the aggressive use of tariffs and claiming that NAFTA was “the worst trade deal maybe ever signed anywhere but certainly ever signed in this country.” In 2015, Bernie Sanders, who has long been critical of U.S. free trade agreements, said that “since 2001, we have lost almost 60,000 factories and millions of good-paying jobs” and claiming that “trade is not the only reason, but it is a significant reason why Americans are working longer hours for low wages and why we are seeing our jobs go to China and other low-wage countries.” Ross Perot similarly ran against trade liberalization, saying, in a 1992 presidential debate, that NAFTA would mean “you’re going to hear a giant sucking sound of jobs being pulled out of this country” and that “we don’t have good trade agreements around the world.” This echoes sentiments from other political elites and organized interest groups who have long warned that trade liberalization benefits multinational corporations at the expense of American workers and that the economic gains of free trade agreements are unequally distributed (e.g., Beehner, 2005; Behr, 1993; Lee, 2016).

While trade liberalization purports to increase economic opportunity and open up new markets, these are, for ordinary people, abstract and non-tangible benefits. Furthermore, it may be difficult for ordinary Americans to understand how free trade agreements benefit the masses, particularly during a time of growing economic inequality and sluggish wage growth among the lower and middle classes (e.g., Bartels, 2016; Saez & Zucman, 2020). In contrast, the closure of factories and shipping of previously high-paying jobs to other countries is very clear and tangible.

Indeed, Table 1 shows that only a minority of Americans think that free trade agreements provide clear, tangible, and “understandable” benefits to the United States. For example, only about one third of Americans think that free trade agreements lead to lower prices, one in four think that it leads to economic growth, and only about one in ten think that they lead to higher wages and more jobs for American workers. I do not wish to imply that free trade does not yield objective macroeconomic benefits nor that a large majority of Americans explicitly think that free trade is a “bad thing.” Rather, I argue that many Americans do not think that clear, tangible benefits result from free trade.6

Table 1.

Perceptions of Free Trade’s Societal Consequences, 2006–2015

	Percentage who think free trade agreements lead to...
	Lower prices	Higher wages	More jobs	Economic growth
2006	32%	11%	12%	28%
2008	29%	8%	9%	19%
2009	32%	11%	13%	25%
2010	31%	8%	8%	19%
2015	36%	11%	17%	31%
Mean	32.0	9.8	11.8	24.4

Shows the percentage of respondents who think that free trade agreements have positive consequences (vs. negative, neutral, or don’t know/refused to answer). Sources are five Pew Research Center polls. These were accessed by searching for “free trade” in the Roper Center’s iPoll archives (https://ropercenter.cornell.edu/ipoll/). The samples are weighted to be representative of the national adult U.S. population. N ranges from 1502 to 2002

Because of the imbalance between (perceived) benefits and risks, with the latter appearing to outweigh the former in many Americans’ minds, and because free trade agreements lack, for most people, clear material or symbolic benefits, e.g., a check in the mail, improved infrastructure in one’s community, or a clear political victory for one’s partisan or ideological “team,” trade is the type of policy that should serve to “activate” political trust, i.e., make it relevant when people are determining their support for that particular policy. Formally, I hypothesize the following: political trust should be positively associated with support for free and open “pro-trade” policies.

Data and Methods

Here, I describe the design for my main observational analysis, which uses cross-sectional data from the 2016 American National Election Study (ANES). I supplement this analysis with pooled cross-sectional data from the 2004 and 2008 National Annenberg Election Surveys (NAES), pooled cross-sectional data from the 1988–2012 Cumulative ANES, panel data from the 2016–2020 ANES, and a survey experiment fielded through Amazon’s Mechanical Turk (MTurk). I discuss the main (2016 ANES) design here and the designs associated with additional analyses in detail later on in the paper.

My dependent variable is support for free trade. I measure this in the 2016 ANES by combining three questions (all asked post-election) into an index ($\alpha$ = 0.637). These questions ask respondents (1) have increasing amounts of trade with other countries been good for the United States? (1 = bad; 2 = neither good nor bad; 3 = good), (2) do you you favor the U.S. making free trade agreements with other countries? (1 = oppose; 2 = neither favor nor oppose; 3 = favor), and (3) do you favor placing new limits on foreign imports? (1 = favor; 2 = haven’t thought much about it; 3 = oppose). I code responses to these three questions in the “pro-trade” direction so that higher (lower) values indicate higher (lower) support for free and open “pro-trade” policies. This yields a 9-point scale that I re-code to range between 0 and 1 (mean = 0.547, sd = 0.296).7

My main independent variable is political trust. I measure this in the 2016 ANES by combining four questions (all asked pre-election) into an additive index ($\alpha$ = 0.635). These four questions, which have long been used as a valid measure (Hetherington, 2005), ask respondents their attitudes regarding: (1) how often they can trust the federal government to do the right thing? (1 = never; 2 = some of the time; 3 = about half the time; 4 = most of the time; 5 = always), (2) whether government is run by a few big interests or for the benefit of all the people? (1 = few big interests; 2 = benefit of all), (3) how much tax money government wastes? (1 = a lot; 2 = some; 3 = not very much), and (4) how many people in government are corrupt? (1 = all; 2 = most; 3 = about half; 4 = a few; 5 = none). I code responses to these five questions in the “trusting” direction so that higher (lower) values indicate higher (lower) levels of political trust. This yields a 12-point scale that I re-code to range between 0 and 1 (mean = 0.348, sd = 0.178).

I also include a battery of theoretically appropriate variables that have been identified by past work as important explanations of Americans’ trade attitudes (e.g., Ehrlich & Maestas, 2010; Essig et al., 2021; Hainmueller & Hiscox, 2006; Johnston, 2013; Mansfield & Mutz, 2009). Collectively, this combination of variables should help to account for the role of economic self-interest, sociotropic economic concerns, political identities, out-group attitudes, and cosmopolitanism. This can help to reduce omitted variable bias but does not, of course, address all of the issues associated with observational data.

In addition to political trust, my 2016 ANES models also account for: gender, race, marital status, home ownership, union household status, education, age and income. I also control for survey mode to account for differences between people who completed the 2016 ANES in-person vs. whether they were a part of the online sample. In addition to these demographics, which are intended to help account for economic self-interest and life socialization, I also control for partisanship (strong Democrat $\to$ strong Republican) and ideological self-placement (extremely liberal $\to$ extremely conservative).8 I also include measures of both personal and sociotropic economic evaluations, specifically asking about peoples’ personal financial situation and their perceptions of the national economy over the past year (much worse $\to$ much better). I also control for authoritarianism (4-item index; $\alpha$ = 0.646; low $\to$ high), traditionalism (4-item index; $\alpha$ = 0.711; low $\to$ high), social trust (people can never be trusted $\to$ people can always be trusted), and feeling thermometer ratings (cold $\to$ warm) of Big Business and of Donald Trump.

Authoritarianism, a predisposition associated with a preference for social order, certainty, and security, can help account for risk aversion (Ehrlich & Maestas, 2010; Johnston, 2013). Traditionalism, a core value associated with opposition to new ideas and social change, can help account for psychological predispositions such as cosmopolitanism, nationalism, and out-group attitudes (Hainmueller & Hiscox, 2006; Mansfield & Mutz, 2009). Controlling for social trust is important because it helps ensure that my measure of political trust is not simply capturing generalized trust. Controlling for feelings toward “Big Business” is important because multinational corporations have long been powerful advocates of trade liberalization (e.g., Goodman, 2016; Swanson, 2021). Furthermore, trade attitudes may be shaped by whether people trust large corporations (rather than the federal government) to act in the national interest and to behave honestly; this variable helps to account for this. Finally, controlling for feelings toward Donald Trump is important because Trump’s populist and protectionist rhetoric in 2016 could have shaped peoples’ political trust and trade opinion and also forged a relationship between these two attitudes. Moreover, feelings toward Trump, in addition to partisanship, can help to ensure that my measure of political trust does not merely reflect attitudes toward the incumbent [Obama] administration.

I also include state fixed effects (dummy variables for each respondent’s state). Because these data are cross-sectional, this can account for all state-level factors that may shape Americans’ trade attitudes. This includes, but is not limited to, objective trade-related jobs gains/losses in a respondent’s state. With the exception of age and household income, all of these variables are either categorical or are scaled to range from 0 to 1.9

Main Results

I present the main (2016 ANES) results in Table 2. Overall, these results show that political trust is positively and significantly associated with greater support for “pro-trade” policies, i.e., favoring free trade over protectionism. Consistent with past studies, I find that education, labor union affiliation, economic evaluations, social trust, authoritarianism, and traditionalism are all significant correlates of free trade support. I also find that feelings toward “Big Business” matter as well. Trade attitudes also appear to strongly driven by a “Trump effect” (Essig et al., 2021), with anti-trade (pro-trade) individuals supporting (opposing) the candidacy of Donald Trump in 2016 or, more likely, the result of individuals “following the leader” (Lenz, 2012) and adjusting their trade attitudes to bring them in line with their feelings toward Trump.

Table 2.

Political Trust and Mass Support for Free Trade, 2016

	DV = Free Trade Support Scale
	coef	std error	coef	std error
Political Trust	0.171	(0.024)***	0.153	(0.025)***
Female	$-$0.020	(0.010)*	$-$0.026	(0.011)**
White	0.001	(0.012)	0.017	(0.011)
Married	0.006	(0.011)	0.008	(0.011)
Homeowner	$-$0.002	(0.015)	0.001	(0.014)
Union Household	$-$0.047	(0.011)***	$-$0.043	(0.011)***
College Degree	0.076	(0.010)***	0.068	(0.010)***
Online Survey Mode	$-$0.001	(0.011)	$-$0.007	(0.011)
Age	$-$0.000	(0.000)	$-$0.000	(0.000)
Household Income	0.001	(0.001)	0.001	(0.001)
Personal Finances Perceptions	0.080	(0.019)***	0.074	(0.018)***
National Economy Perceptions	0.126	(0.019)***	0.090	(0.020)***
Partisanship	$-$0.070	(0.021)***	0.003	(0.023)
Ideology	$-$0.010	(0.027)	0.014	(0.026)
Social Trust	0.060	(0.023)**	0.062	(0.023)***
Authoritarianism	$-$0.079	(0.019)***	$-$0.067	(0.019)***
Traditionalism	$-$0.158	(0.027)***	$-$0.126	(0.026)***
Feeling Thermometer: Big Business	0.082	(0.020)***	0.106	(0.022)***
Feeling Thermometer: Donald Trump			$-$0.188	(0.023)***
Constant	0.456	(0.039)***	0.478	(0.038)***
State Fixed Effects	$✓$		$✓$
Observations	3,156		3,132
R${}^2$	0.208		0.230

Dependent variable ranges 0–1 (higher values = greater support for free trade). OLS coefficients with robust standard errors clustered by state in parentheses (ordered probit models yield very similar results). Source is the 2016 ANES

*** p<0.01, ** p<0.05, * p<0.1, two-tailed

Column 1 of Table 2 shows that a shift in political trust from its minimum to its maximum (0 $\to$ 1) is associated with an approximately 0.17 shift in free trade support, from 0.49 to 0.66.10 Column 2 of Table 2 shows that this result is not simply a product of Donald Trump’s 2016 campaign that both demonized government’s trustworthiness and harshly criticized free trade. Indeed, political trust is substantively and significantly associated with support for a more “open” and “free” trade policy, net of how people felt about Donald Trump, with a minimum to maximum shift (0 $\to$ 1) associated with an approximately 0.15 shift in trade support, from approximately 0.49 to 0.64. The magnitude of this relationship (between political trust and trade support) is also substantively significant, matching or exceeding the magnitude of analogous shifts (0 $\to$ 1) in national economic evaluations, authoritarianism, traditionalism, and feelings toward “Big Business” and feelings toward Donald Trump.

Robustness of Findings

In the following sections, I supplement my main (2016 ANES) results with additional data. First I show, using pooled cross-sectional data from the 2004 and 2008 National Annenberg Election Surveys (NAES) as well as the 1988–2012 Cumulative ANES, that the results hold, i.e., the trust-trade relationship is positive and statistically significant, across different datasets and election years. Second, I show, using panel data from the 2016–2020 ANES, that political trust appears to drive trade attitudes rather than the reverse. Finally, I show, using data from a survey experiment conducted through Amazon’s Mechanical Turk (MTurk), that political trust and trade support appear to be causally related to one another.

Evidence from Additional Surveys

In Tables 3 and 4, I respectively use pooled cross-sectional data from the 2004–2008 National Annenberg Election Surveys (NAES) and data from the 1988–2012 Cumulative ANES to further examine the trust-trade relationship. Each of these datasets only include a single question to measure trade attitudes. However, they are still useful because they can show that the results hold in years when trade was less politically salient than in 2016. Indeed, Hetherington and Husser Hetherington and Husser (2012) show that political trust matters more (for relevant policy preferences) when the issue under consideration is more salient in political discourse and in media coverage. As such, it is important to demonstrate that they results are not simply due to the heightened salience of trade during the 2016 election. Moreover, these analyses also show that the results hold during both Democratic and Republican presidencies and across different surveys, i.e., that there is not something unique about the 2016 ANES that yields the observed results.

Table 3.

Political Trust and Support for Free Trade Agreements, 2004–2008

	DV = Favor Free Trade
Political Trust	0.193***
	(0.012)
Additional Controls	$✓$
State Fixed Effects	$✓$
Year Fixed Effects	$✓$
Observations	27,930
R${}^2$	0.068

Dependent variable is dichotomous (0 = opposed to/neutral toward more free trade agreements; 1 = favor more free trade agreements). OLS coefficients with robust standard errors clustered by state in parentheses (a probit model yields very similar results). Sources are the 2004 and 2008 NAES

***p < 0.01, **p < 0.05, *p < 0.1, two-tailed. See Appendix Table B1 for the full model

Table 4.

Political Trust and Support for Limiting Foreign Imports, 1988–2012

	DV = Favor Import Limits
Political Trust	−0.108***
	(0.016)
Additional Controls	$✓$
State Fixed Effects	$✓$
Year Fixed Effects	$✓$
Observations	12,628
R${}^2$	0.062

Dependent variable is dichotomous (0 = oppose new limits on foreign imports/haven’t thought much about it; 1 = favor new limits on foreign imports). OLS coefficients with robust standard errors clustered by state in parentheses (a probit model yields very similar results). Source is the 1988–2012 Cumulative ANES

***p < 0.01, **p < 0.05, *p < 0.1, two-tailed. See Appendix Table B2 for the full model

In Table 3, I use pooled cross-sectional data from the 2004 and 2008 National Annenberg Election Surveys (NAES). Both of these NAES surveys ask (in nearly identical format) a question about trade attitudes that I use here as my dependent variable. This question asks respondents whether the federal government should negotiate more free trade agreements like NAFTA? I code responses as follows (0 = strongly oppose/somewhat oppose/neither favor nor oppose/don’t know; 1 = somewhat favor/strongly favor).11

Both of these NAES surveys also include a single question to measure political trust. This question, which I use here as my main independent variable, asks respondents the following: thinking about the federal government in Washington, how much of the time do you think you can trust the federal government to do what is right?. Responses are coded as follows (1 = never; 2 = some of the time; 3 = most of the time; 4 = always) and are re-scaled to range between 0 and 1 (mean = 0.379; sd = 0.198).12

Unfortunately the NAES lacks several important determinants of trade attitudes that are available in the ANES. While not ideal, I can still include a battery of relevant demographics (gender, race, marital status, union household status, education, age, income, how long the respondent has lived at their current address, and whether they live in an urban area), partisanship and ideological self-placement, and favorability toward George W. Bush. I also include year fixed effects (a dummy variable for the survey year) and state fixed effects (dummy variables for each respondent’s state). These can respectively account for election-specific factors and the objective state of the national economy, as well as respondents’ objective subnational (state-level) economic conditions. The aforementioned limitations notwithstanding, these NAES analyses should bolster the main (2016 ANES) results by showing that the findings are similar across different datasets and election years.

Overall, the results in Table 3 show that political trust is positively and significantly associated with support for free trade. In particular, a shift in political trust from its minimum to its maximum (0 $\to$ 1; from “never” to “always” trusting the federal government to do what is right) is, holding the aforementioned controls variables constant, associated with an approximately 0.19 increase in explicit free trade support (vs. a neutral/don’t know position or explicit opposition), from 0.34 to 0.53.13 In short, the totality of evidence thus far demonstrates that political trust and support for free and open “pro-trade” policies appear to be linked in a substantively significant manner.

In Table 4, I use pooled cross-sectional data from the 1988–2012 Cumulative ANES. Until 2016, most ANES surveys only included a single question to measure trade attitudes. This question was first asked in 1988 and was included in most subsequent ANES survey years. Respondents are presented with the following text: some people have suggested placing new limits on foreign imports in order to protect American jobs. Others say that such limits would raise consumer prices and hurt American exports. Do you favor or oppose placing new limits on imports, or haven’t you thought much about this? I code responses as follows (0 = oppose new limits/haven’t thought much about it; 1 = favor new limits) so that a score of “1” reflects an explicit desire for a protectionist shift in trade policy and a score of “0” reflects a lack of support for such a change in trade policy.

I measure political trust by using an index consisting of the same four questions from the 2016 ANES (how often government can be trusted to do the right thing, whether government looks out for a few big interests, how much money government wastes, and how many people in government are crooked). This index ranges from 0 to 100 in the Cumulative ANES (higher values = greater political trust), I re-scale it to range from 0 to 1.

In addition to a measure of political trust, I also include a battery of socio-demographic and attitudinal variables that broadly control for economic self-interest, sociotropic concerns, and cosmopolitanism/out-group attitudes. In order to maximize the sample size (including every presidential election year from 1988 through 2012), I was unable to include all of the same variables as in Table 2 (using 2016 ANES data). Nevertheless, this model still accounts for: gender, race, marital status, labor union affiliation, education, age, income, evaluations of one’s personal financial situation over the past year, evaluations of the national economy over the past year, partisanship, ideological self-placement, traditionalism, and feelings toward Big Business. I also account for state and year fixed effects, which helps to control for various election-specific factors such as candidate rhetoric and state-level factors such as trade-related job losses/gains that may shape trade attitudes.

The results in Table 4 show that higher levels of political trust are associated with greater opposition to protectionist policies that seek to restrict foreign imports. On average, a shift in political trust from its minimum to its maximum (0 $\to$ 1) is, controlling for the various other aforementioned variables, associated with a 0.11 decrease in explicit support (vs. a neutral “haven’t thought much about it” position or explicit opposition to this policy) for new restrictions on foreign imports, from approximately 0.41 to approximately 0.30. In short, the results in Table 4 provide additional evidence to support the robustness of the trust-trade relationship, showing that it holds over time and across surveys.

In the Supplemental Appendix (Tables B3, B4, and B5), I explore possible heterogeneity in the trust-trade relationship. I do so with data from the 2016 ANES, the 2004–2008 NAES, and the 1988–2012 Cumulative ANES. To test whether the observed relationship between political trust and support for pro-trade policies holds across groups, I separately interact political trust with: race (White, non-Hispanic vs. not), gender (male vs. female), and education (4-year college degree or higher vs. not). Overall, the results from these models suggest that political trust matters for trade attitudes broadly across the population, i.e., across race, gender, and educational attainment.

Panel Data

No cross-sectional, observational research design can address all endogeneity issues, even with a robust battery of control variables. In Table 5, I seek to assuage concerns about one specific type of endogeneity, that being “reverse causality.” By this I mean, determining whether political trust drives trade attitudes or whether the reverse is true.

Table 5.

Panel analyses of Political Trust and Free Trade, 2016–2020

	Trade Support (2020)	Political Trust (2020)
Trade Support (2016)	0.317***	$-$0.002
	(0.019)	(0.006)
Political Trust (2016)	0.197***	0.392***
	(0.054)	(0.019)
Observations	2,579	2,740
R${}^2$	0.108	0.163

Dependent variables range 0–1 and 0 vs. 1 (higher values = greater support for free trade and greater political trust). OLS coefficients with robust standard errors in parentheses (a probit model for column 1 yields very similar results). The constant terms are not displayed here. Source is the 2016–2020 ANES Panel Study

***p < 0.01, **p < 0.05, *p<0.1, two-tailed

There are several reasons why political trust should drive trade attitudes rather than the reverse. For one, most people do not have exceptionally strong attitudes toward most political issues (e.g., Freeder et al. 2019). Furthermore, trade has not typically occupied as central a part of the national political agenda as have issues regarding, for example, healthcare, race, immigration, gay rights, and abortion. As such, trade attitudes should occupy a less central place in mass belief systems and thus be less likely to shape other attitudes and predispositions.

Previous research from the United States has also shown, using panel data, that political trust shapes attitudes toward policies regarding race (Hetherington & Globetti, 2002), immigration (Macdonald, 2021), and foreign affairs (Hetherington & Husser, 2012), rather than the reverse. As such, it seems reasonable to expect that political trust drives support for free trade instead of the reverse. However, it is prudent to test this empirically rather than simply assuming the directionality of this relationship. This is impossible to do in a cross-sectional analysis, where all of the variables are measured at the same time. As such, I use data from the 2016–2020 ANES panel study.

I measure political trust in each panel wave (2016 and 2020) by using the same four questions used in the main (2016 ANES) analyses. This yields two 12-point scales (in 2016 and 2020) that I re-code to range from 0 to 1. I measure free trade support with one question that was asked identically in both waves (2016 and 2020). This questions asks respondents the following: (1) do you favor the U.S. making free trade agreements with other countries?. I code responses as follows (0 = oppose/neither favor nor oppose; 1 = favor). In short, I have, for the same survey respondents at two different points in time (2016 and 2020), valid measures of both political trust and attitudes toward trade policy.14

Table 5 presents the results of two cross-lagged regression models. The first column shows that political trust (measured in 2016) significantly predicts future (2020) free trade support, holding previous (2016) trade attitudes constant. In contrast, the second column shows that trade attitudes (measured in 2016) do not significantly predict political trust four years later (in 2020), holding respondents’ prior (2016) political trust constant. Cross-lagged models are not a panacea (e.g., Bellemare et al., 2017), but are useful here as a means of testing the direction of the trust-trade relationship. Overall, the results in Table 5 should help assuage concerns about one type of endogeneity, that being “reverse causality.” This should serve to bolster the validity and robustness of my cross-sectional results.

Experimental Evidence

The evidence thus far has demonstrated that political trust and trade support are linked in a substantively significant manner. However, because these data are observational in nature, they are potentially vulnerable to omitted variable bias. I address concerns about this in Table 6 by using data from a survey experiment fielded through Amazon’s Mechanical Turk (MTurk).15 By randomly assigning respondents to read an article about how people can (and should) trust the federal government, I can demonstrate that political trust and free trade support are causally related. This should serve to assuage additional concerns about endogeneity and bolster the validity of the main observational results.

Table 6.

Experimental Analyses of Political Trust and Free Trade, 2021

	DV = Encourage Trade
Treatment vs. Control	0.085*
	(0.048)
Received Treatment vs. Control		0.097**
		(0.049)
Constant	0.650***	0.650***
	(0.036)	(0.036)
Observations	376	346
R${}^2$	0.008	0.011

Dependent variables are dichotomous (0 = discourage/neither encourage nor discourage international trade; 1 = encourage international trade). OLS coefficients with robust standard errors in parentheses (probit models yield very similar results). Source is a 2021 MTurk survey experiment. ***p<0.01, **p<0.05, *p<0.1, two-tailed

I fielded this experiment online in September, 2021 on Amazon’s Mechanical Turk (MTurk).16 In general, MTurk samples are considerably younger, more “tech-savvy,” and better educated than the general population (Berinsky et al., 2012). Despite these demographic differences (relative to the general adult population) and the fact that MTurk surveys use convenience rather than probability sampling, MTurk respondents are still more representative than college student samples. Importantly, survey experiments with MTurk samples have been shown to yield similar results as compared to those conducted with more representative samples (Coppock, 2019). In short, this survey experiment, its limitations notwithstanding, should complement and bolster the observational (ANES and NAES) analyses, both of which use high-quality nationally representative samples.17

This is a simple randomized experiment with two groups, treatment and control. All respondents were asked a small series of pre-treatment questions (gender, race, age, and education), and were then asked their attitudes about free trade. I measured trade attitudes here by using one question that asks respondents the following: overall, do you think the policy of the United States should be to encourage, discourage, or neither encourage nor discourage international trade?. This variable is coded as follows (0 = discourage/neither encourage nor discourage; 1 = encourage).

The treatment group differs from the control group in that they first read a fictitious news article intended to prime and increase political trust. This follows the approach of past work that has used broadly similar (fictitious) articles to prime/increase political trust (Faulkner et al., 2015; Macdonald, 2021). The (fictitious) article that “treatment group” respondents read was entitled “We Can Still Trust Government” and was published online in The Economist.18 It was written by a (fictitious) former civil servant named Paul Martin whose biography notes that he worked for the federal government for 30 years, under both Democratic and Republican administrations. In this article, the author acknowledges that political trust is low, but claims that this is due, in large part, to negative media coverage that hyper-focuses on a relatively small number of scandals and instances of corruption. The author reminds readers about what government does well (effectively administering Medicare, Medicaid, and Social Security and guaranteeing clean air and water) and also notes that vast majority of government officials and politicians are “decent, honest, and hard-working people.” The author concludes by arguing that “we can still trust government.”19

I expect that respondents in the “treatment group” will, as a result of having read the aforementioned “pro-trust in government” article, be more supportive of free and open “pro-trade” policies than respondents in the “control group.” Because people were randomly assigned to have read this article (vs. not), I can be confident that any observed differences in trade attitudes between the treatment and control groups are attributable to the news article that “primed” political trust and not some other unobserved factor.20

The results in Table 6 show that individuals in the “treatment” group were significantly more likely to support a U.S. policy of encouraging trade (vs. not) as compared to their counterparts in the “control” group. The first column compares the difference in trade support between the control group (did not see the aforementioned news article) and the treatment group (saw the aforementioned “pro-trust” news article). The second column compares the difference in trade support between the control group and individuals who actually “received the treatment.” By this, I mean people in the treatment group who read and absorbed the message of the news article. I measure this with a simple attention check that asks treatment-group respondents to recall what the article they just read was about.21 Overall, this survey experiment should, by demonstrating that political trust and trade support appear to be causally related to one another, serve to complement and bolster the nationally representative observational analyses.

Conclusion and Political Implications

Using cross-sectional, panel, and experimental survey data from the United States, I have shown that political trust is positively and significantly linked with support for free and open “pro-trade” policies. All of these data and designs have limitations, but the totality of evidence points in the same direction and strongly suggests that higher levels of political trust boost citizen support for free trade. In short, trade opinion is not solely driven by economic interests, cultural antipathy, nor social/political identities. I argue that people are also consider whether government, the actor most responsible for crafting, implementing, and managing public policies, will pursue trade agreements in the national interest and effectively mitigate the societal risks, be they real or perceived, associated with free trade. Survey data shows that people are less supportive of free trade than expected, based on the macroeconomic benefits it yields. Reasons for this disconnect could include ignorance about trade’s distributional benefits (e.g., Tomz & Rho, 2017), anti-foreigner sentiment, and perceptions that the U.S. tends to “lose” more often than it “wins” from free trade (e.g., Mutz, 2021). I argue that anti-government sentiment plays an important role as well.

There are several potentially fruitful avenues for future research. First, it would be valuable to explore potential heterogeneity in the trust-trade relationship, perhaps by examining whether it is moderated by individual-level predispositions such as employment sector or cosmopolitanism, and/or by geographic factors such as local job losses. Second, future work should also look beyond the United States by examining the trust-trade relationship in other countries. Indeed, public skepticism about free trade’s societal benefits is not confined to the U.S. and there appears to be an interesting divide in public opinion about trade’s benefits across advanced, emerging, and developing nations (Pew Research Center, 2014). Third, future work would do well to examine the possible impact of free trade on political trust, potentially by examining whether trade-related job losses (gains) in one’s geographic area are associated with lower (higher) levels of political trust. If a relationship is found, this could suggest a possible dynamic (and deleterious) relationship in which trade-related job losses reduce political trust, which in turn, decreases political trust and hinders domestic governments’ ability to pursue new trade agreements. Finally, future work could explore whether low (high) political trust leads people to vote for (against) parties and candidates who support protectionism and restrictions on international trade. Indeed, protectionist candidates may find fertile ground among politically distrustful electorates.

An interconnected and interdependent world is highly unlikely and practically unable to abandon large-scale international trade any time soon, particularly given trade’s potential to reduce poverty and increase economic growth (e.g., Congressional Budget Office, 2016; Johns et al., 2015). However, recent trade agreements have also produced clear economic “winners” and “losers” and has elicited a significant political backlash (e.g., Naio 2020; Walter, 2021). Domestic governments can and should do more to alleviate the negative societal effects of free trade, possibly by reforming existing trade agreements and/or forging new trade agreements with other countries. These tasks will be far easier to accomplish with a more politically trustful mass public.

Supplementary Information

Below is the link to the electronic supplementary material.