To assess the effect of politically targeted trade retaliation on public attitudes, we employ a survey experiment that allows us to isolate the effects of different types of trade retaliation. We fielded our survey using Lucid Theorem on a diverse sample of over 3,500 respondents. The study was fielded between October 20th and 25th in 2020 and targeted respondents to resemble the demographics of US adults based on gender, age, geographic, and racial distributions. Survey samples from Lucid are increasingly used in social science research, including numerous articles published in top political science journals.Footnote 28
While Lucid allows researchers to access a diverse sample of respondents, recent research finds that data quality declined as a result of the COVID-19 pandemic, so we employed attention checks and asked respondents to pledge to pay attention to address quality concerns. Approximately 36 percent of recruited participants failed the attention checks and were not included in the sample, which is generally consistent with broader trends for the time.Footnote 29 Even though response quality declined during the pandemic, Peyton et al. (2020) find that studies conducted throughout the pandemic successfully replicated earlier studies and should generalize beyond the pandemic, though treatment effects from experiments fielded during the pandemic were somewhat more conservative given reduced attention of respondents. We also implemented the recommendation of Burleigh et al. (2018), and blocked respondents from participating if they were located outside of the US or were flagged for using a Virtual Private Server (VPS). The sample resulted in a diverse sample that closely mirrored the national population on demographics of age, partisanship, and gender, as shown in the demographic breakdown in Appendix B. Like most online surveys, our sample skews lower on income and is somewhat more likely to be college educated than the national population.
In designing the experiment, we chose a relatively hard test of our theory. The first critical decision was whether to focus the experimental vignette on trade retaliation by China, another country, or to leave the country unspecified. We wanted the scenario to be realistic as well as plausible, so we chose to use a real country that had retaliated against the US during the trade war.Footnote 30 However, we also chose to make the test a hard test for the theory by focusing on the European Union, rather than China. At the time we fielded the survey, suspicions that China would engage in election interference were high. Both President Trump and Vice President Pence had accused China of attempting to interfere in US elections. Furthermore, they had also explicitly described China’s trade policy as an attempt at election interference. In addition to previously-mentioned remarks by Pence in October 2018, Trump also made the link explicit during a speech at the U.N. Security Council:
Regrettably, we found that China has been attempting to interfere in our upcoming 2018 election, coming up in November, against my Administration... They do not want me, or us, to win, because I am the first president ever to challenge China on trade, and we are winning on trade. We are winning at every level. We don’t want them to meddle or interfere in our upcoming election.
In order to make the test a hard one, we opted to focus on the trade policies of the European Union, a trading partner that had not been as strongly accused of using its trade policies to interfere in American elections. The test is also a harder one because most EU countries are allies of the United States who are not generally thought to interfere in US elections. Additionally, in the 2020 election, EU members had not publicly taken as strong of a stance against Trump, compared to the 2016 election. By contrast, China is viewed by many as a threat to the US, and it has engaged in cyber attacks and espionage against the US.Footnote 31 Furthermore, public opinion toward China has become quite negative, with about two-thirds of Americans having an unfavorable view toward China,Footnote 32 whereas a majority of Americans have a favorable view toward the EU.Footnote 33 Thus, we would expect that respondents would be more sensitive to additional information about political targeting from a geostrategic adversary like China, compared to the EU, making ours a conservative test of the theory.Footnote 34
Our survey experiment randomly assigned respondents to one of four conditions, each of which varied what the respondent read about the trade war. In the control condition, respondents read a few short lines about the trade war.
Control: In 2018 and 2019, the Trump Administration started a trade war by imposing tariffs on imports of steel and aluminum from the European Union.
These tariffs are meant to lower imports of steel from Europe into the United States and to convince the European Union to change its trade policies.
In response to President Trump’s tariffs, the European Union retaliated with tariffs against a variety of products that it imports from the United States.
The control treatment does specify that retaliation occurred, but does not specify any particular regional or political target of the retaliation. This allows us to differentiate between the public’s response to retaliation in general, versus retaliation that is politically targeted in the subsequent treatments.
Our study included three separate treatment conditions that described particular features of the retaliation. For respondents not assigned to control, they read the same information as in the control, and were randomly assigned to read one additional piece of information, which was either the Base, Swing, or Placebo treatment. We modeled the wording of these treatments after media reporting on the tariffs, so that our treatment in the survey experiment resembles the “treatment” received by respondents when they read information about the trade war or heard arguments from elites.
Base: ... The European Union intentionally chose products that come from states that voted for Trump in the last election. The European Union hopes that punishing these states will hurt President Trump politically and convince him to end the trade war.
Swing: ... The European Union intentionally chose products that come from “swing states” that will be important in the upcoming US election. The European Union hopes that punishing these states will hurt President Trump politically and convince him to end the trade war.
Placebo: ... The European Union announced the list of products in a press release and also communicated their decision to the United States Trade Representative through diplomatic channels. The official notification contained further details about the tariffs.
The Base and Swing conditions are the most important for our study. They each contain information about the politically-targeted nature of the European Union’s retaliation.Footnote 35 The Base condition emphasizes how retaliation targeted states that supported President Trump in previous elections, and the Swing treatment emphasizes how retaliation targeted more electorally competitive areas in battleground states. Both treatments explain the intentionality behind the EU’s targeting, describing how the retaliation is meant to affect Trump’s decision making. When compared to the control condition, they allow us to isolate the effect of trade retaliation being politically targeted at the base or swing states, beyond the mere existence of retaliation itself.
As previously mentioned, we sought to design the treatments in a manner that made for a relatively hard test of the theory. One option would have been to model our treatments to mirror language used by President Trump or Vice President Pence that explicitly describes PTTR as a form of electoral interference. Rather than using such an explicit treatment, we chose more moderate language that only referenced the political nature of the trade retaliation (i.e., that the tariffs were targeted with political considerations in mind), but we avoided specifically linking the tariffs to electoral interference. The treatment thus presents respondents with facts about PTTR, but leaves it up to the respondent to decide on their own whether the fact that tariff policies were politically targeted should cause them to be worried about electoral interference.
The final treatment is a placebo which has a nearly identical word count to the Base and Swing treatments and a similar structure, but contains little information that would affect a respondent’s attitude about the trade war. We included this Placebo treatment to ensure that effects of the Base and Swing treatments resulted from the informational content of those treatments, as opposed to simply having additional information on the page. Like the Control treatment, the Placebo treatment describes the presence of retaliation but does not attribute any political motivations.Footnote 36
We generally see balance across treatment groups in the characteristics of the respondents, as expected with random assignment. To formally assess balance, we used the approach described in Hansen and Bowers (2008) to compare each treatment with the control group. In each comparison, we fail to reject the null hypotheses of balance. We see some small differences between control and treatment in some covariates.Footnote 37 For example, there are 5% more respondents identifying as Democrats in the Control group, compared to the Base treatment. Imbalance is very unlikely to influence our results. For starters, the estimates presented below do not change much when we include or exclude respondent characteristics.Footnote 38 We also conducted sensitivity testing to show that imbalance in some other respondent characteristic would have to be many orders of magnitude larger than our currently observed levels of imbalance in order to affect our results.Footnote 39
After reading about the trade war, respondents were presented a bullet-point recap of the key details of the treatment, which remained at the top of the survey screen as they answered post-treatment questions. For example, a respondent assigned to the Base treatment would have read:
The US put tariffs on imports from the EU to get them to change their trade policies.
The EU retaliated with tariffs of their own on imports from the US.
The EU’s retaliatory tariffs targeted President Trump’s base.
Our primary outcome measure comes from response to the following question:
I am worried the European Union’s retaliatory tariffs are an attempt to interfere with the upcoming US Presidential election.
Respondents could choose from five responses, ranging from “Strongly agree” to “Strongly disagree” with “Neither agree nor disagree” as a middle option. We deliberately chose the wording of this outcome measure. The term “election interference” has taken on a significant, negative connotation in the US context.Footnote 40 This term, as used in media reports, has come to encompass things like leaking emails, hacking voting machines, or other malicious acts of election meddling. The outcome measure thus allows the respondent to independently make a connection between trade policy—an act that is not inherently electoral—and election interference, a much more serious and negative issue that directly impacts the well-being of democratic politics. If politically targeted trade policy makes respondents more likely to express worry about election interference, then that is more than just an answer that factually “connects the dots” from treatment to outcome. To agree with the prompt, respondents indicate more than just an acknowledgement that retaliation was politically targeted; they associate trade retaliation with the same term that is used to describe direct challenges to the foundations of democracy.
The wording of our dependent variable also raises the bar, since it requires that respondents express worry or anxiety about PTTR as a form of election interference. This is substantively important, since “political anxiety triggers engagement in politics” (Albertson and Gadarian, 2015, p. 1). Feeling worried or anxious about politics also leads people to seek out more information, especially when the information is threatening, such as PTTR (Gadarian and Albertson, 2014). Thus, if respondents are worried that PTTR is an attempt to interfere in the election, it implies that they not only believe the EU is attempting to interfere, but that the respondent is likely to find it politically salient, since the information induces a sense of anxiety or worry.