Authoritarianism and support for Trump and Clinton in the 2016 primaries

Abstract

Research in the wake of the contentious 2016 presidential primaries contends both Democrats and Republicans were internally divided along psychological lines. Specifically, MacWilliams (2016) finds authoritarian personality was strongly related to Trump support among Republican primary voters, and Wronski et al. (2018) finds authoritarianism was strongly related to Clinton support among Democratic primary voters. In this paper, I reassess the relationships between authoritarianism and 2016 primary candidate preferences for both Republicans and Democrats. I analyze two new large, probability-based surveys and generate random effects estimates using these surveys and two national surveys from Wronski et al. (2018). Overall, I find authoritarianism was moderately associated with support for Clinton over Sanders among Democratic primary voters, but weakly associated with support for Trump among Republican primary voters. My findings indicate authoritarianism may have played a more limited role in shaping Americans’ candidate preferences in the 2016 presidential primary elections than past studies have suggested.

Keywords

Authoritarianism political psychology voting behavior primary elections replication

Introduction

Political psychologists have long been interested in how authoritarian personality shapes political behavior in the United States (Adorno et al., 1950; Hetherington and Weiler, 2009). In two recent accounts, MacWilliams (2016) and Wronski et al. (2018) extend this line of inquiry to intra-party contexts by evaluating how authoritarianism was associated with Americans’ candidate preferences in the 2016 presidential primaries. MacWilliams (2016) argues authoritarianism was associated with support for Donald Trump over other Republican candidates in 2016, while Wronski et al. (2018) find that authoritarianism was associated with support for Hillary Clinton over Bernie Sanders among Democrats, but insignificantly related with support for Trump among Republicans. While these studies diverge somewhat on whether authoritarianism spurred support for Trump among Republicans, they agree authoritarianism played an important role structuring Americans’ intra-partisan political behavior during the 2016 primaries.

In this paper, I reexamine the relationships between authoritarianism and partisans’ 2016 primary candidate preferences. I conduct additional tests of the hypotheses that authoritarianism was associated with greater support for Trump and Clinton among their parties’ respective primary voters using two large, probability-based surveys. I twice find that authoritarianism was neither significantly associated with support for Trump among Republican primary voters, nor support for Clinton among Democratic primary voters (though just barely shy of significance in one case). And while random effects estimates pooling across the four available national samples between this study and Wronski et al. (2018) suggest authoritarianism was moderately associated with support for Clinton among Democratic primary voters (Wronski et al., 2018), authoritarianism seems to have been very weakly associated with Trump support for Republicans (MacWilliams, 2016). My findings indicate authoritarianism played a more limited role shaping Americans’ candidate preferences in the 2016 primaries than past studies have suggested.

Authoritarianism and candidate preferences in the 2016 primaries

I begin by reviewing two published studies that have examined how authoritarianism, measured with self-reported childrearing values, was associated with candidate preferences in the 2016 US primaries: MacWilliams (2016) and Wronski et al. (2018).¹ MacWilliams evaluates the relationship between authoritarianism and support for Trump and other Republican primary candidates using a survey of 540 likely Republican voters fielded in December 2015. Although reproduction data for MacWilliams (2016) is not publicly available, in Table 1 and Figure 1, I directly reproduce MacWilliams’ (2016) results pertaining to authoritarianism and Republican primary candidate preferences. Referencing the results in this table and figure, MacWilliams concludes authoritarianism was significantly associated with support for Trump, but not support for other Republican primary candidates. From Figure 1, the estimated marginal effect of authoritarianism is an approximately 0.30-point change in the predicted probability of Trump support.

Table 1.

Associations of 2016 Republican primary candidate preferences reproduced from MacWilliams (2016). Entries are logistic regression coefficients with standard errors in parentheses. Sample includes only “likely Republican primary voters.” Source: MacWilliams 2016, University of Massachusetts Amherst Political Science Department December 2015 National Survey.

	Trump	Trump	Cruz	Carson	Rubio	Bush
Authoritarianism	0.273	0.347	0.357	−0.028	0.091	−0.060
Authoritarianism	(0.084)	(0.160)	(0.226)	(0.256)	(0.206)	(0.206)
Terror threat	0.150	0.212	0.367	−0.064	−0.074	−0.075
Terror threat	(0.053)	(0.124)	(0.171)	(0.199)	(0.167)	(0.164)
Authoritarianism*Terror		−0.604	−4.008	1.486	-0.049	0.193
Authoritarianism*Terror		(1.099)	(1.637)	(1.714)	(1.624)	(1.542)
Gender	−0.126	−0.126	−0.276	0.089	−0.480	0.443
Gender	(0.200)	0.200	(0.291)	(0.320)	(0.319)	(0.307)
Education	−0.415	−0.423	−0.075	−0.731	0.543	0.040
Education	(0.381)	0.381	(0.554)	(0.625)	(0.594)	(0.581)
Age	−0.013	−0.009	−1.616	1.956	0.183	−1.575
Age	(0.433)	0.433	(0.636)	(0.702)	(0.677)	(0.672)
Evangelicalism	0.025	0.035	0.447	−0.326	−0.024	−0.506
Evangelicalism	(0.214)	(0.215)	(0.308)	(0.338)	(0.341)	(0.361)
Ideology	0.053	0.052	0.239	0.070	0.031	−0.275
Ideology	(0.214)	(0.061)	(0.102)	(0.094)	(0.096)	(0.093)
Church attendance	−0.387	−0.385	0.062	0.947	−0.528	−0.054
Church attendance	(0.220)	(0.220)	(0.303)	(0.324)	(0.355)	(0.345)
Race	0.253	0.253	−0.087	0.234	−0.392	0.644
Race	(0.257)	(0.257)	(0.370)	(0.407)	(0.342)	(0.343)
Income	−0.066	−0.058	−0.251	0.753	0.588	0.682
Income	(0.437)	(0.437)	(0.616)	(0.725)	(0.659)	(0.664)
Constant	−1.917	−2.161	−3.175	−4.278	1.800	0.278
Constant	(0.619)	(0.765)	(1.100)	(1.218)	(1.048)	(1.000)
Observations	540	540	540	540	540	540

Figure 1.

Associations of authoritarianism to support for Republican primary candidates reproduced from MacWilliams (2016). Points are predicted probabilities of support for five Republican candidates with 95% confidence intervals as functions of authoritarianism. Sample includes only “likely Republican primary voters.” Source: MacWilliams 2016, University of Massachusetts Amherst Political Science Department December 2015 National Survey.

Wronski et al. (2018) similarly examine how authoritarianism shaped preferences in the 2016 primary. Although their article focuses on the Democratic contest between Hillary Clinton and Bernie Sanders, Wronski et al. also analyze the Republican primary in appendices. Wronski et al. analyze two national surveys: the 2016 Cooperative Congressional Election Study (CCES) and a 2017 YouGov survey.² Using Wronski et al.’s publicly available reproduction materials, I reproduce their analyses in Table 2 and Figure 2 with linear probability models.³ These analyses show authoritarianism was significantly related with support for Clinton over Sanders among Democratic primary voters, with estimated effects of 0.35- and 0.25-points measured as changes in predicted probability in the CCES and YouGov, respectively. Testing the association between authoritarianism and support for Trump over Cruz, however, the CCES and YouGov samples show non-significant, but positive, effects of 0.10- and 0.20-points, respectively. Wronski et al. thus conclude authoritarianism was associated with Democrats’, but perhaps not Republicans’, candidate preferences in the 2016 primaries.

Table 2.

Associations of 2016 Democratic primary candidate preferences reproduced from Wronski et al. (2018). Entries are linear probability model coefficients with standard errors in parentheses. Partisan groups include independent leaners. Comparisons are Clinton over Sanders for Democrats and Trump over Cruz for Republicans. Source: Wronski et al. 2018, YouGov 2017, CCES 2016.

	Democratic primary		Republican primary
	CCES	YouGov	CCES	YouGov
Authoritarianism	0.352	0.254	0.098	0.198
Authoritarianism	(0.091)	(0.101)	(0.161)	(0.150)
Party identity strength	0.190	0.274	0.006	−0.040
Party identity strength	(0.083)	(0.090)	(0.103)	(0.110)
Ideology (conservative)	0.176	0.474	−0.595	−0.708
Ideology (conservative)	(0.111)	(0.163)	(0.165)	(0.212)
Church attendance	0.059	0.064	−0.123	−0.149
Church attendance	(0.096)	(0.145)	(0.115)	(0.152)
Male	−0.060	−0.065	−0.027	0.162
Male	(0.061)	(0.067)	(0.078)	(0.089)
White	−0.079	−0.118	−0.122	0.047
White	(0.067)	(0.086)	(0.101)	(0.117)
Education	−0.121	−0.030	−0.344	0.176
Education	(0.109)	(0.123)	(0.167)	(0.197)
Income	0.588	0.190	0.194	−0.212
Income	(0.167)	(0.191)	(0.228)	(0.286)
Union	−0.045	−0.168	−0.140	−0.019
Union	(0.060)	(0.072)	(0.097)	(0.100)
Married	−0.043	0.036	−0.055	−0.023
Married	(0.066)	(0.079)	(0.091)	(0.098)
Age	0.040	0.029	0.017	0.026
Age	(0.017)	(0.023)	(0.026)	(0.031)
South	0.079	0.052	0.005	−0.073
South	(0.060)	(0.076)	(0.075)	(0.104)
Need for closure		−0.106		−0.076
Need for closure		(0.185)		(0.243)
Social dominance		−0.202		−0.104
Social dominance		(0.203)		(0.280)
Racial resentment		0.028		0.153
Racial resentment		(0.145)		(0.244)
Constant	0.026	0.148	1.384	0.932
Constant	(0.142)	(0.188)	(0.245)	(0.369)
Observations	260	195	145	119
R-squared	0.221	0.267	0.152	0.159

Figure 2.

Associations of authoritarianism to support for Clinton and Trump in 2016 primary reproduced from Wronski et al. (2018). Points are predicted probabilities of support for Clinton over Sanders among Democrats (blue) and Trump over Cruz among Republicans (red) with 95% confidence intervals as functions of authoritarianism. Partisan groups include independent leaners. Source: Wronski et al. 2018, YouGov 2017, CCES 2016.

In this paper, I reevaluate the associations of authoritarianism and support for Trump and Clinton in the 2016 primaries. Why might these relationships warrant further assessment? First and foremost, MacWilliams (2016) and Wronski et al. (2018) reach different conclusions about how authoritarianism was associated with Republicans’ preferences in the 2016 primaries. Widely publicized media claims following MacWilliams (2016) suggest there was something psychologically distinct about Trump’s 2016 primary supporters relative to standard Republican identifiers, and that Trump was uniquely attractive to these authoritarian individuals among many Republican candidates. These popular accounts often imply psychological authoritarians are something of a lurking danger in the mass polity, and that 2016 was an example of this voting bloc collectively rearing its head.⁴ Yet my reproductions of Wronski et al. (2018) suggest that non-authoritarian Republicans supported Trump’s primary campaign at rates similar to authoritarians; thus, it is unclear whether the credit (or blame) for Trump’s success in the 2016 primary can be mostly attributed to psychologically authoritarian Republicans.

Second, there are potential issues with the samples analyzed by Wronski et al. (2018) and MacWilliams (2016). The national samples analyzed in Wronski et al. are quite small: whereas MacWilliams’ December, 2015 survey includes 540 likely Republican primary voters, Wronski et al.’s CCES includes 260 Democratic and 145 Republican primary voters after listwise deletion, and their YouGov survey includes 195 Democratic and 119 Republican primary voters. These samples are concerning because samples with fewer than approximately 250 observations often generate unreliable correlational estimates (Schönbrodt and Perugini, 2013). Further, perhaps due to their small samples, Wronski et al. exclude Kasich and Rubio voters, who comprise 14% and 11% of the Republican primary electorate, from their study. Analyses of large samples would be beneficial towards reducing sampling error, increasing statistical power (Arel-Bundock et al., 2022), and allowing supporters of many major Republican primary candidates to be analyzed. Another potential concern with Wronski et al. and MacWilliams’ samples is that they are drawn from opt-in web panels. The generalizability of estimates derived from non-probability panels is the subject of ongoing debate, with some studies finding discrepancies to favor probability-based approaches (Malhotra and Krosnick, 2007; Meng, 2018; Yeager et al., 2011) and others finding approximately similar results across recruitment methods (Berrens et al., 2003; Clifford et al., 2015; Vitriol et al. 2019). Verifying that the associations identified by Wronski et al. (2018) and MacWilliams (2016) also emerge in probability-based samples would bolster these studies’ claims.

Finally, I argue MacWilliams (2016) does not provide sufficient evidence to sustain the claim that authoritarianism was associated with support for Trump in the 2016 primaries, but not other Republican primary candidates. MacWilliams (2016) claims that “Trump supporters are also distinct in their attitudes from the followers of the other Republican candidates.” MacWilliams reports the association of authoritarianism with Trump support (which is positive and significant), but does not report comparable coefficients for the other four Republican candidates. Instead, as shown in Table 1, MacWilliams (2016) presents models for these other candidates while interacting authoritarianism and terrorism threat. Without further information about the unconditional (or “main”) effects of authoritarianism, which is unavailable, it is impossible to conduct significance tests on the difference in the coefficients for authoritarianism across the five candidates. Notably, Figure 1 also shows a positive association between authoritarianism and support for Cruz, which could be consistent with Wronski et al.’s (2018) analyses. However, whether Trump and Cruz supporters significantly differ in authoritarianism in MacWilliams (2016) is unknown. As it stands, I argue MacWilliams’ claim that authoritarianism was associated with support for Trump, but not other Republican primary candidates, is insufficiently supported by the reported analyses.

Data and methodology

In this paper, I reassess the relationships between authoritarianism and candidate preferences in the 2016 primaries. I leverage two large, probability-based surveys: the 2016 American National Election Study (ANES) Time Series and a Public Religion Research Institute (PRRI) poll fielded on the AmeriSpeak panel. In Table 3, I outline key details regarding these surveys compared to those of Wronski et al. (2018) and MacWilliams (2016). Since the PRRI was fielded amidst the 2016 primaries, I limit my main analyses of the PRRI by excluding respondents who reported not voting if their state’s primary had already occurred and those who reported no intention to vote if their state’s primary was forthcoming.⁵ Usefully, both the ANES and PRRI boast larger samples than those analyzed by Wronski et al. and MacWilliams. I weight my analyses of the PRRI and ANES, and account for the ANES’s complex sampling design when calculating standard errors. Further details about the sampling methodologies of these surveys are available in Appendix A.

Table 3.

Sample descriptions. PRRI and ANES data are weighted. Sample sizes are after listwise deletion. Source: MacWilliams 2016, University of Massachusetts Amherst Political Science Department December 2015 National Survey, Wronski et al. 2018, YouGov 2017, CCES 2016, 2016 ANES Time Series, 2016 Public Religion Research Institute (NORC AmeriSpeak panel).

Published Surveys	Dec 2015 survey			2016 CCES		2017 YouGov
Dependent variable	Candidate preferences			Vote choice recall		Vote choice recall
Field dates	12/10/2015–12/15/2015			09/28/2016–12/14/2016		09/29/2017–10/10/2017
Sampling design	Non-probability, online			Non-probability, online		Non-probability, online
Partisan subsample	Democrats	Republicans		Democrats	Republicans	Democrats		Republicans
Authoritarianism (mean)	-	-		0.48	0.62	0.41		0.61
Supports: Hillary Clinton	-	-		63%	-	56%		-
Supports: Bernie Sanders	-	-		37%	-	44%		-
Supports: Donald Trump	-	-		-	74%	-		64%
Supports: Ted Cruz	-	-		-	26%	-		36%
Supports: John Kasich	-	-		-	-	-		-
Supports: Marco Rubio	-	-		-	-	-		-
Observations	-	540		260	145	195		119
New Surveys	2016 PRRI (AmeriSpeak panel)				2016 ANES Time Series
Dependent variable	Candidate preferences				Vote choice recall
Field dates	04/04/2016–05/02/2016				09/07/2016–01/08/2017
Sampling design	Probability-based, online and telephone				Probability-based, online and face-to-face
Partisan subsample	Democrats		Republicans		Democrats		Republicans
Authoritarianism (mean)	0.56		0.71		0.45		0.60
Supports: Hillary Clinton	55%		-		63%		-
Supports: Bernie Sanders	45%		-		37%		-
Supports: Donald Trump	-		44%		-		52%
Supports: Ted Cruz	-		34%		-		23%
Supports: John Kasich	-		22%		-		13%
Supports: Marco Rubio	-		-		-		11%
Observations	911		651		731		594

I model Democrats’ primary preferences using linear probability models with a dependent variable that takes values of 0 (Sanders) and 1 (Clinton). I model Republicans’ primary preferences using multinomial probit regression with a dependent variable that takes values of 1 (Trump), 2 (Cruz) and 3 (Kasich) in the PRRI, plus 4 (Rubio) in the ANES.⁶ I also use all four estimates from Wronski et al.’s CCES and YouGov surveys, the ANES, and the PRRI survey in random effects models to estimate authoritarianism’s overall associations with Clinton and Trump support.⁷ To facilitate random effects modeling of the Republican primary, I collapse the dependent variables into binary Trump/non-Trump outcomes and estimate these associations using linear probability models. However, the random effects estimates should be interpreted with caution, and always alongside the estimates from individual surveys, because these models aggregate over potentially important differences in the surveys’ recruitment methodology, timing, and analytic modeling.

I assess authoritarianism with the standard four-item childrearing values measure, which is the same authoritarianism measure used by Wronski et al. (2018) and MacWilliams (2016). I use the same set of controls as Wronski et al.’s CCES model⁸: age, gender, ethnicity, education, income, marriage, church attendance, southern residency, partisan identity, ideology, and union membership (which is not available in the PRRI survey). The covariates in this model differ from those used in MacWilliams (2016) to analyze the Republican primary, most importantly in that MacWilliams controls for terrorism threat but not partisan identity strength. However, in Appendix C, I show that my key findings are not substantively changed when using MacWilliams’ alternative model to analyze the Republican primary.

Results

I begin by examining the 2016 Democratic primary. In Table 4, I regress Democrats’ candidate preferences (PRRI) or vote choice recalls (ANES) on authoritarianism and the controls such that positive coefficients indicate associations with support for Clinton over Sanders. I find authoritarianism is positively, though non-significantly, associated with Clinton support in both samples (p_ANES=0.30, p_PRRI=0.06). As shown in Figure 3, the predicted probabilities of support for Clinton increase by 0.08- and 0.12-points in the ANES and PRRI, respectively, as functions of authoritarianism.⁹ In the random effects model including estimates from the ANES, PRRI, CCES, and YouGov surveys, I estimate a 0.11-point association between authoritarianism and Clinton support among Democrats (p=0.01). Although authoritarianism was associated with Clinton support in the 2016 primaries, tests using large, probability-based samples show that the strength of this association was approximately one-third to half the size estimated by Wronski et al. (2018).

Table 4.

Associations of 2016 Democratic primary candidate preferences. Entries are linear probability model coefficients with standard errors in parentheses. Sample includes only Democrats and Democratic-leaning independents. Data are weighted. Standard errors are adjusted for the complex sampling design in the 2016 ANES. Source: 2016 ANES Time Series, 2016 Public Religion Research Institute (NORC AmeriSpeak Panel).

	2016 PRRI	2016 ANES
Authoritarianism	0.124	0.077
Authoritarianism	(0.065)	(0.074)
Party identity strength	0.185	0.309
Party identity strength	(0.047)	(0.051)
Ideology (conservative)	0.247	0.462
Ideology (conservative)	(0.098)	(0.100)
Church attendance	−0.022	0.009
Church attendance	(0.067)	(0.049)
Male	−0.032	−0.051
Male	(0.043)	(0.037)
White	−0.041	−0.106
White	(0.045)	(0.043)
Education	−0.118	0.163
Education	(0.071)	(0.129)
Income	0.017	0.038
Income	(0.093)	(0.078)
Married	0.108	0.064
Married	(0.044)	(0.045)
Age	0.521	0.022
Age	(0.097)	(0.043)
South	0.081	0.572
South	(0.043)	(0.085)
Union		0.077
Union		(0.041)
Constant	0.101	−0.127
Constant	(0.095)	(0.129)
Observations	911	731
R-squared	0.154	0.278

Figure 3.

Predicted probability of Clinton vote in 2016 Democratic primary. Points are predicted probabilities of support for Clinton over Sanders with 95% confidence intervals as a function of authoritarianism. Sample includes only Democrats and Democratic-leaning independents. Data are weighted. Standard errors are adjusted for the complex sampling design in the 2016 ANES. Source: 2016 ANES Time Series, 2016 Public Religion Research Institute (NORC AmeriSpeak panel).

Next, in Table 5, I provide results for the multinomial probit regressions of Republicans’ primary candidate preferences. Here, positive coefficients indicate a variable is associated with support for non-Trump candidates over Trump, the baseline category. I perform significance tests using the average marginal effects of authoritarianism, and show the predicted probabilities of support across candidates as functions of authoritarianism in Figure 4.¹⁰ In the ANES, from its minimum to maximum values, authoritarianism is associated with 0.12-points greater support for Trump (p = 0.16), 0.00-points less support for Cruz (p = 0.98), 0.02-points greater support for Rubio (p = 0.74), and 0.13-points less support for Kasich (p = 0.01). In the PRRI, authoritarianism is associated with 0.02-points greater support for Trump (p = 0.76), 0.03-points greater support for Cruz (p = 0.73), and 0.05-points less support for Kasich (p=0.41). Results from linear probability models are similar; the associations of authoritarianism with Trump support are 0.13-points (p_ANES = 0.14) and 0.04-points (p_PRRI = 0.67). In the random effects model, I estimate a 0.05-point association between authoritarianism and Trump support that, although statistically significant (p = 0.01), is much weaker than the approximately 0.30-point association in MacWilliams (2016).

Table 5.

Multinomial probit regression of 2016 Republican primary candidate preferences. Entries are multinomial probit regression coefficients with standard errors in parentheses. Trump is the baseline category of comparison. Sample includes only Republicans and Republican-leaning independents. Data are weighted. Standard errors are adjusted for the complex sampling design in the 2016 ANES. Source: 2016 ANES Time Series, 2016 Public Religion Research Institute (NORC AmeriSpeak panel).

	2016 PRRI		2016 ANES
	Cruz	Kasich	Cruz	Kasich	Rubio
Authoritarianism	0.023 (0.338)	−0.248 (0.353)	−0.199 (0.388)	−1.088 (0.422)	−0.136 (0.371)
Party identity strength	−0.205 (0.203)	−0.070 (0.213)	−0.002 (0.254)	−0.940 (0.285)	−0.654 (0.254)
Ideology (conservative)	2.599 (0.522)	0.034 (0.457)	2.577 (0.736)	−1.276 (0.816)	0.745 (0.549)
Church attendance	0.723 (0.321)	0.403 (0.311)	0.792 (0.252)	0.388 (0.285)	0.430 (0.256)
Male	−0.232 (0.190)	−0.069 (0.206)	−0.356 (0.217)	−0.044 (0.210)	−0.426 (0.226)
White	−0.447 (0.241)	0.183 (0.296)	−0.470 (0.300)	0.350 (0.394)	−0.055 (0.377)
Education	0.294 (0.345)	1.165 (0.431)	0.862 (0.731)	3.948 (0.956)	2.088 (0.868)
Income	−0.472 (0.438)	1.311 (0.425)	0.492 (0.395)	0.538 (0.409)	1.217 (0.529)
Married	0.095 (0.208)	−0.192 (0.235)	−0.306 (0.229)	0.161 (0.284)	−0.219 (0.254)
Age	−1.485 (0.462)	−0.012 (0.496)	−0.884 (0.430)	−0.409 (0.518)	0.059 (0.486)
South	−0.140 (0.188)	−0.421 (0.214)	0.380 (0.187)	−0.753 (0.238)	0.768 (0.218)
Union			−0.974 (0.344)	−0.471 (0.379)	−0.902 (0.462)
Constant	−1.107 (0.524)	−1.957 (0.585)	−2.630 (0.876)	−2.300 (1.047)	−3.539 (1.000)
Observations	651	651	594	594	594

Figure 4.

Predicted probabilities of candidate preferences in 2016 Republican primary. Points are predicted probabilities of support for each candidate with 95% confidence intervals as functions of authoritarianism. Sample includes only Republicans and Republican-leaning independents. Data are weighted. Standard errors are adjusted for the complex sampling design in the 2016 ANES. Source: 2016 ANES Time Series, 2016 Public Religion Research Institute (NORC AmeriSpeak panel).

Discussion

Before concluding, it is worth directly addressing the major limitations of my analysis. I use cross-sectional survey data to estimate the effects of authoritarianism on 2016 primary candidate preferences. As such, I cannot empirically discern the direction(s) of causality that generate correlations between authoritarianism and primary candidate preferences. A foundational assumption of political psychology is that personality occurs causally prior to political behavior. However, this assumption is seen as increasingly untenable. In recent studies, Bakker et al. (2021) find political attitudes and identities affect personality traits, and Luttig (2021) similarly finds Trump support causes Republicans to report higher levels of authoritarianism (but see also Engelhardt et al., 2021 who find authoritarianism is exogeneous to political attitudes/identities). The estimated effects of authoritarianism on primary candidate preferences in this study risk being biased by reverse causal influences, so they should be viewed as strictly associational.

A second issue is that it is difficult to handle partisan identity strength and ideology with a regression-based design. Partisan identity and ideology causally follow from authoritarianism (Hetherington and Weiler, 2009; Luttig, 2017), so controlling for these variables risks attenuating the estimated effect of authoritarianism. However, if authoritarianism is endogenous to partisan identity and ideology, and partisan identity and ideology cause primary candidate preferences, excluding these variables risks confounding authoritarianism’s relationship with primary candidate preferences. In supplemental analyses (Appendix D), I drop these covariates and find the random effects estimated association of authoritarianism with support for Clinton increases modestly from 0.11-points to 0.17-points, while the association with Trump support barely decreases from 0.05-points to 0.04-points.

Conclusion

Prior research contends that psychological authoritarianism divided American partisans during the 2016 presidential primaries, spurring support for Hillary Clinton among Democrats (Wronski et al., 2018) and support for Donald Trump among Republicans (MacWilliams, 2016). Towards confirming how authoritarianism was related to primary candidate preferences in 2016, I examined two new large, probability-based surveys. I twice failed to find authoritarianism was significantly associated with support for Clinton among Democrats or support for Trump among Republicans in the 2016 primaries. Using a random effects model to pair my analyses with two reproduced from Wronski et al. (2018), I found a moderate association between authoritarianism and support for Clinton among Democratic primary voters. However, I also found MacWilliams’ (2016) claim that authoritarianism was strongly associated with support for Trump among 2016 Republican primary voters to be considerably overstated.

That I find authoritarianism had weaker effects than past studies is not especially surprising; social scientific replication studies generally find smaller effects than original studies (Camerer et al., 2018). Ideally, however, we would know what caused these discrepancies. In my view, four factors could explain these discrepancies: differences in (1) sample size; (2) survey timing; (3) survey recruitment; and/or (4) modeling decisions.

Sample size is very likely one reason why I find smaller effects than past studies. The typical political science study is underpowered and, therefore, noisy (Arel-Bundock et al., 2022); there is an inverse relationship between sample size and effect size in published social science research (Gerber et al., 2001; Kühberger, Fritz, and Scherndl, 2014); and the publication of large effects when analyzing small samples is more or less mechanical when publication biases favor significant findings because only large effects can attain significance in small samples (Esarey and Wu, 2016; Loken and Gelman, 2017). Large, out-of-sample tests are useful in this regard because they reduce noise and offer opportunities for mean reversion to emerge. In addition, the surveys’ timings likely affected the estimated associations between authoritarianism and primary candidate preferences. Studies assessing these associations have used surveys spanning 2015–2017, and there is no guarantee these associations were stable over this period. Specifically, although authoritarianism seems to have been weakly associated with Trump support during the primaries and afterwards in vote choice recalls, authoritarianism could have been more strongly associated with Trump support immediately before the primary’s onset when MacWilliams’ study was fielded. Finally, further discrepancies could be caused by differences in sample recruitment. Since the goal of this inquiry is to identify the associations of authoritarianism with primary candidate preferences in the electorate, generalizability to this population is important. Claims of generalizability are typically seen as more credible when analyzing probability-based surveys like the ANES and PRRI, even though non-probability surveys can still be informative.

Although differences in survey size, timing, and recruitment strike me as plausible causes of the discrepancies between my estimates and those of past studies, I do not believe differences in modeling decisions explain these discrepancies. The main regression model I use is admittedly different from the one MacWilliams (2016) used to examine the Republican primary; however, I also find weaker associations using MacWilliams’ preferred model (Appendix C). My estimates also remain relatively weaker when controls for partisan identity strength and ideology are dropped (Appendix D), using different likely voter screens (Appendix E), and using logistic regressions (Appendix F). Across multiple model specifications and samples, authoritarianism is moderately associated with Clinton primary support and weakly associated with Trump primary support.

This study makes several important contributions. My analyses add to an emerging line of scholarly inquiry concerned by the intra-partisan consequences of psychological differences (MacWilliams, 2016; Wronski et al., 2018; Luttig, 2017). Although I find authoritarianism likely mattered less in shaping 2016 presidential primary preferences than extant research suggests, this does not necessarily mean psychological traits are unimportant in intra-partisan contexts; rather, I hope my findings prompt further inquiry into whether and how psychological differences manifest in intra-party divisions. However, this study does show the claim that authoritarianism “provided the fuel for Trump’s [primary] campaign” is overstated (MacWilliams, 2016). To be clear, this is not to say Trump was not authoritarian himself—his dehumanizing rhetoric, nativist appeals, and calls for political opponents to be jailed suggest otherwise (Jardina and Piston, 2021). But despite Trump’s authoritarian proclivities, the distribution of his primary support in authoritarianism was similar to other Republican candidates. My findings thus support accounts that argue other social and economic transformations better explain Trump’s electoral success than psychological authoritarianism (Gordon, 2016). As such, it would not be surprising if authoritarian and non-authoritarian Republicans rally behind Trump’s recently announced 2024 primary campaign.

Supplemental Material

Supplemental Material - Authoritarianism and support for Trump and Clinton in the 2016 primaries

Supplemental Material for Authoritarianism and support for Trump and Clinton in the 2016 primaries by Trent Ollerenshaw in Research & Politics

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Trent Ollerenshaw

Supplemental Material

Supplemental material for this article is available online

Notes

References

Adorno

Else

Levinson

, et al. (1950) The Authoritarian Personality. New York City, NY: Harper and Brothers.

Altemeyer

(1981) Right-Wing Authoritarianism. Winnipeg: University of Manitoba Press.

Vincent

Ryan

Hristos

, et al. (2022) Quantitative political science research is greatly underpowered. https://osf.io/7vy2f/ (accessed 12 July 2022).

Bakker

Lelkes

Ariel

(2021) Reconsidering the link between self-reported personality traits and political preferences. American Political Science Review 114: 1–17.

Berrens

Bohara

Hank

, et al. (2003) The advent of internet surveys for political research: a comparison of telephone and internet samples. Political Analysis 11(1): 1–22.

Bizumic

Duckitt

(2018) Investigating right wing authoritarianism with a very short authoritarianism scale. Journal of Social and Political Psychology 6(1): 129–150.

Camerer

Dreber

Holzmeister

, et al. (2018) Evaluating the replicability of social science experiments in nature and science between 2010 and 2015. Nature Human Behaviour 2(9): 637–644.

Clifford

Ryan

Waggoner

(2015) Are samples drawn from mechanical Turk valid for research on political ideology? Research and Politics 2(4): 2053168015622072.

Engelhardt

Feldman

Hetherington

(2021) Advancing the measurement of authoritarianism. Political Behavior 45: 537–560.

10.

Esarey

(2016) Measuring the effects of publication bias in political science. Research and Politics 3(3). Available at: https://doi.org/10.1177/2053168016665856.

11.

Feldman

(2003) Enforcing social conformity: a theory of authoritarianism. Political Psychology 24(1): 41–74.

12.

Gerber

Green

Nickerson

(2001) Testing for publication bias in political science. Political Analysis 9(4): 385–392.

13.

Gordon

(2016) The authoritarian personality revisited: reading adorno in the age of trump. Boundary 2 44: 31–56. https://www.boundary2.org/2016/06/peter-gordon-the-authoritarian-personality-revisited-reading-adorno-in-the-age-of-trump/ (acessed October 19 2022).

14.

Hetherington

Weiler

(2009) Authoritarianism and Polarization in American Politics. Cambridge: Cambridge University Press.

15.

Jardina

Piston

(2021) The effects of dehumanizing attitudes about black people on whites’ voting decisions. British Journal of Political Science 52: 1–23.

16.

Kühberger

Fritz

Scherndl

(2014) Publication bias in psychology: a diagnosis based on the correlation between effect size and sample size. PLoS One 9(9): e105825.

17.

Loken

Gelman

(2017) Measurement error and the replication crisis. Science 355(6325): 584–585.

18.

Ludeke

Klitgaard

Joseph

(2018) Comprehensively-measured authoritarianism does predict vote choice: the importance of authoritarianism’s facets, ideological sorting, and the particular candidate. Personality and Individual Differences 123: 209–216.

19.

Luttig

(2017) Authoritarianism and affective polarization: a new view on the origins of partisan extremism. Public Opinion Quarterly 81(4): 866–895.

20.

Luttig

(2021) Reconsidering the relationship between authoritarianism and Republican support in 2016 and beyond. The Journal of Politics 83(2): 783–787.

21.

MacWilliams

(2016) Who decides when the party doesn’t? Authoritarian voters and the rise of Donald Trump. PS: Political Science and Politics 49(4): 716–721.

22.

Malhotra

Krosnick

(2007) The effect of survey mode and sampling on inferences about political attitudes and behavior: comparing the 2000 and 2004 ANES to internet surveys with nonprobability samples. Political Analysis 15(3): 286–323.

23.

Meng

X-L

(2018) Statistical paradises and paradoxes in big data (I): law of large populations, big data paradox, and the 2016 US presidential election. Annals of Applied Statistics 12(2): 685–726.

24.

Schönbrodt

Piston

(2013) At what sample size do correlations stabilize? Journal of Research in Personality 47(5): 609–612.

25.

Stenner

(2005) The Authoritarian Dynamic. New York: Cambridge University Press.

26.

Suurmond

Rhee

Hak

(2017) Introduction, comparison, and validation of meta-essentials: a free and simple tool for meta-analysis. Research Synthesis Methods 8(4): 537–553.

27.

Vitriol

Larsen

Ludeke

(2019) The generalizability of personality effects in politics. European Journal of Personality 33(6): 631–641.

28.

Wronski

Bankert

Amira

, et al. (2018) A tale of two Democrats: how authoritarianism divides the Democratic Party. The Journal of Politics 80(4): 1384–1388.

29.

Yeager

Krosnick

Chang

, et al. (2011) Comparing the accuracy of RDD telephone surveys and internet surveys conducted with probability and non-probability samples. Public Opinion Quarterly 75(4): 709–747.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.38 MB