![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
ABSTRACT
During his campaign and subsequent presidency, Donald Trump staked out and implemented an isolationist foreign policy agenda that sought to put “America First” and curtail U.S. participation in international treaties and trade agreements. Isolationism represented a dramatic turn away from the internationalism of all postwar presidencies, and Trump’s radical foreign policy changes raise two novel questions: Did isolationism resonate with American voters, and if so, how did this shape the outcome of the 2016 presidential election? We analyze how attitudes toward isolationism changed in the American electorate from 1992 to 2016, and whether aggregate changes in isolationism conferred any electoral advantages on Trump in 2016. The results indicate that while isolationism tends to be supported by a relatively small portion of the electorate, aggregate levels of support increased noticeably in 2016. Controlling for established factors, particularly partisanship, isolationism disposed voters to support Trump’s candidacy. These two developments delivered a measurable benefit to Trump, one that is of theoretical significance in going beyond partisanship to influence both Democratic and Republican identifiers. Future shifts in levels or political relevance carry with them a potential to shape elections and policy-making. We discuss implications of findings for scholarship in political behavior and political sociology.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Notes
1. Indeed, to our knowledge the most recent scholarly analysis of trends in preferences for isolationism is Page and Bouton’s (Citation2006) study.
2. A further theoretical contrast of relevance involves the role of emotions, and the degree to which affect-driven beliefs have durable versus short-lived effects on behavior. In Sears (Citation1988) work on the origins of symbolic racism, for instance, childhood exposure to modern racial stereotypes and anti-black affect lead to racial attitudes that persist into adulthood. In contrast, “rally round the flag” attitudes formed during periods of war and national security crises may be potent yet shorter-lived effects, as the affect-driven events that drive voters tend to expire over time or be replaced by politicians and media frames conveying new patterns of affect (Feinstein Citation2020).
3. The missing items in question are a battery of four questions used to measure racial resentment.
4. The ANES also incorporates an internet-based sample, but we restrict the analysis to the in-person data to avoid the risk of mode effects on estimates.
5. We note that our measure of isolationism should not be conflated with protectionism, understood as support for restrictive trade policies. In supplemental analysis, we found no systematic evidence of an association between isolationism and attitudes toward restrictive trade policies – for example, limiting imports, opposing outsourcing, and protecting jobs.
6. Recently, Bonikowski, Feinstein, and Bock (Citationforthcoming) have turned their attention to examining the relevance of nationalism for U.S. elections and partisanship. While the ANES has not consistently fielded items that measure nationalism, two measures are available in the 2016 release: one that asks if the world would be a better place if people from other countries were more like Americans and another that asks if seeing the American flag makes the respondent feel good, bad, or neither good nor bad. A Wald test for the equality of coefficients indicates that our estimate for isolationism is largely unchanged when these items are included as a pair (p < .45) in comparison to a model that excludes them. The same is also true when the items are included as a scale (p < .42). This suggests that the estimate for the effect of isolationism are likely to be unaffected by the absence of these items from the over-time analysis.
7. The data weights are provided by ANES, and we use the weights that are appropriate for analyzing the in-person sample only – specifically: V927000, V000002, V040101, V080101, weight_ftf (for 2012), and V160102f. The models were estimated with Stata and require using the “svy:” prefix.
8. The data for these estimates reflect the presence of weights, and we note that estimated trends for the unweighted data lead to substantively similar estimates.
9. In supplemental analysis, we explored models that included both isolationism × partisanship interactions (to evaluate whether the influence of isolationism depended on partisanship) and election year × isolationism × partisanship interactions (to evaluate whether the influence of isolationism depended on partisanship and election year). None of these models yielded evidence that the influence of isolationism was conditional on party identification.
10. This model includes economic insecurity and spending preferences as control variables. In supplemental analyses, we re-estimated both the baseline and expanded models with these items excluded. The estimates for the influence of isolationism across all the models were largely unchanged.
11. In supplemental analyses, we also examined whether isolationist preferences were associated with switching from supporting Obama in 2012 to supporting Trump in 2016. In a model that has switching as its dependent variable (switching from Obama to Trump = 1; 0 = else) and includes the controls found in the expanded model, the coefficient for isolationism is positive and statistically significant (b = 2.20; se = .98). This provides further evidence that Trump’s isolationist rhetoric aided his electoral prospects.
12. Comparing the influence of binary variables to continuous variable is not straightforward. We follow Gelman (Citation2008), who suggests comparing the 0/1 change in a binary variable to a two standard deviation change in a continuous variable. It should be emphasized that these changes are not strictly equivalent; instead, they are approximations and the comparisons should be taken as such.
13. These scenarios are based on the expanded model that allows the influence of isolationism to vary across election context. Importantly, the estimates for this model indicate the influence of isolationism was highest in 2016. The calculations for the counterfactual scenarios assume that the election under consideration is the 2016 election and that the influence for isolationism was at its peak. Mechanically, this implies the isolationism × 2016 interaction term is activated, while all other interaction terms are deactivated.
14. These counterfactuals can be estimated separately for Republicans and Democrats. Among Republicans, if isolationism had stayed at its 2004 level, voting for Trump would have declined by 3.0 percentage points, and for Democrats, voting for Trump would have declined by 2.2 percentage points.
15. In supplemental analyses, we explored whether the influence of isolation in 2016 was likewise limited to certain demographic characteristics, including with respect to gender, race, education, age, and region. These models failed to yield evidence of statistically significant interactions.
Additional information
Notes on contributors
Kyle Dodson
Kyle Dodson is associate professor of sociology at the University of California-Merced. His interests are in the areas of political sociology, social movements, and social stratification.
Clem Brooks
Clem Brooks is rudy professor of sociology at Indiana University. His interests are in the areas of political sociology, political psychology, and quantitative methods.