The Relationship Between Follower Affect for President Trump and the Adoption of COVID-19 Personal Protective Behaviors.

In the current series of studies, we draw upon implicit leadership theories, social learning theory, and research on decision making to investigate whether affect toward President Trump explains U.S. residents' evaluations of his leadership during the COVID-19 crisis, as well as the likelihood that that residents engage in personal protective behaviors. A meta-analysis using 17 nationally representative datasets with a total of 26,876 participants indicated that participants who approve of President Trump tend to approve of his leadership regarding the COVID-19 pandemic and were less likely to engage in personal protective behavior (PPBs; i.e., hand washing, wearing a mask or other face covering in public, and social distancing). On the other hand, those disapproving of President Trump also tended to disapprove of his leadership during the COVID-19 crisis and were more likely to engage in PPBs. In a second study, using an established measure of leader affect (leader affect questionnaire) and controlling for political party, we replicated and extended these results by demonstrating that expending cognitive effort toward understanding the COVID-19 crisis attenuated the relationship between affect toward President Trump and (1) approval of his leadership during the COVID-19 crisis and (2) engagement in some, but not all, PPBs.

I could stand in the middle of 5th Avenue and shoot somebody and I wouldn't lose voters.

—Donald J. Trump, President of the United States

Introduction

Crisis leadership is at the forefront as the COVID-19 pandemic grips the world. One could argue that no issue has created a more divided U.S. citizenry than how to handle the COVID-19 crisis. In addressing the needs of the nation in light of this pandemic, there have been two opposing forces: the health and well-being of Americans and the health and well-being of the American economy. In the long run, healthy residents and a healthy economy go hand in hand (McKee & Stuckler, 2020), but in the short term, however, they appear to be inversely related. During the COVID-19 pandemic, physical health and economic health have become politicized. Although both parties clearly favor a healthy citizenry and a healthy economy, Republicans seem to prioritize actions designed to bolster the ailing COVID-19-impacted economy whereas Democrats seem to prioritize actions designed to address safety and disease reduction.

According to Sprunt and Turner (2020), President Trump is the most salient symbol of this divide through actions such as urging states to move more rapidly toward re-opening the economy and schools and threatening to pull federal aid from schools that do not reopen (Baker et al., 2020). With limited exceptions, he has famously refused to wear a face mask in public (Blake, 2020) despite recommendations from the Centers for Disease Control (CDC; CDC Press Release, 2020). The divisiveness created by COVID-19 and President Trump's actions has exacerbated American residents’ extreme views of an already polarizing the President. A study by the Brookings Institution found that President Trump is the most polarizing president in the U.S. history (Eady et al., 2018). In other words, American residents tend to have either very strong positive affect (i.e., liking) toward President Trump or very strong negative affect (i.e., disliking).

Using President Trump as the prototype, we examine the role of affect toward a leader in predicting Americans’ approval of his handling of the COVID-19 pandemic, as well their adoption of personal protective behaviors (PPBs). We hypothesize that individuals who like President Trump will tend to approve of his leadership during the COVID-19 pandemic. Moreover, we expect that these opinions will translate into behavioral responses such that those who like President Trump will be less likely than others to engage in common PPBs, including washing one's hands, wearing a mask in public, and social distancing. Likewise, we hypothesize that those who dislike President Trump will tend to disapprove of his leadership and be more likely to engage in PPBs. Through our examination of these hypotheses, we contribute to our overall understanding of how followers make decisions about their leaders by examining how affect toward President Trump relates to their decisions regarding COVID-19 and related PPBs.

In the following sections, we provide a theoretical rationale for our predictions, relying on the research on implicit leadership theories and social learning theory (Bandura, 1986). We then test our hypotheses across two studies. First, we meta-analyze the results of 17 nationally representative datasets that ask U.S. residents about their perceptions of President Trump and their PPBs. After presenting results of the meta-analysis, we describe the results of our own study, which replicates and extends the findings from the meta-analysis by (1) utilizing an established measure of leader affect, (2) introducing the research on decision making by proposing cognitive effort as a moderator, and (3) controlling for political party.

Implicit Leadership Theory

Research regarding the validity of subordinates’ evaluations of their leaders has a long history (Junker & van Dick, 2014). Four decades ago, researchers began to question the accuracy of leadership ratings and postulated they were a function of the raters’ implicit theories, as opposed to actual leader behaviors (Eden & Leviatan, 1975; Rush et al., 1977; Weiss & Adler, 1981). A comprehensive series of studies followed that investigated the factors that affect the accuracy of leadership ratings, including implicit theories (Lord & Shondrick, 2011; Lord et al., 1978, 1984; Shondrick et al., 2010) and follower affect (Allen et al., 2008; Bono & Ilies, 2006; Hunter et al., 2007; Naidoo et al., 2010). These studies suggested that followers have preconceived notions about the characteristics of good and bad leaders. Once evaluators judge leaders as good or bad, their responses to questions regarding the leaders’ characteristics and behaviors tend to follow their implicit theories rather than leaders’ actual behaviors.

More recently, in an attempt to integrate these studies, Martinko et al. (2018) found that the majority of the variance in followers’ survey responses assessing leader behaviors can be attributed to follower affect toward the leader. Specifically, they argued that followers’ affective reactions to their leaders (i.e., liking) are the primary mechanism by which their implicit leadership theories are triggered. To support this contention, they designed a series of 10 studies, in which they developed and validated measures of leader affect (i.e., the leader affect questionnaire [LAQ]) and demonstrated that the LAQ accounted for the majority of the variance in measures of abusive supervision, authentic leadership, leader–member exchange, transformational leadership, and ethical leadership. Thus, if subordinates like their leaders, they will report that their leaders are not abusive, have good leader–member relations, and are authentic, transformational, and ethical. As a result, they concluded that leader affect cues subordinates’ responses to questions assessing their leaders’ behavior by activating implicit leadership theories. Positive implicit theories are cued when the leader is liked, resulting in positive leader evaluations and vice versa when the leader is disliked. We argue that this same process is enacted when U.S. residents are asked about their approval of Trump as President and their perception of how he is handling the COVID-19 pandemic. Thus, we hypothesize:

Hypothesis 1: There will be a positive relationship between approval/liking of President Trump and U.S. residents’ approval of his handling of the COVID-19 crisis.

Social Learning Theory and Leaders’ Behavior Modeling

Research on social learning theory (Bandura, 1986) suggests that followers observe and emulate attractive role models. Leaders who “walk the talk” and display desired behaviors will serve as role models for followers, increasing their desire to emulate those leaders. In the present study, we were interested in how liking/approving of President Trump may impact engaging in PPBs during the COVID-19 pandemic. An interesting exemplar of this idea is a study conducted by Kessler et al. (2020), who found that when leaders complied with safety measures, their followers were also more likely to comply, which, in turn, reduced accident rates. This study serves as a strong parallel to the COVID-19 situation because it focuses on manufacturing, construction, and transportation employees who face a wide range of potentially dangerous situations in their work. Similarly, the safety precautions recommended by the CDC (e.g., hand washing, face covering, and social distancing) are PPBs (i.e., safety behaviors), with President Trump serving as a prominent role model to the U.S. population. Although Kessler et al. (2020) did not assess the role of leader affect, we extend their thinking by suggesting that followers who like their leaders are more likely to emulate their behavior than those who do not. In the present example, President Trump has famously and continuously refused to wear a face mask (Blake, 2020), so we expect those who have strong positive affect toward President Trump to be less likely to wear a face mask (and, by extension, engage in other PPBs) than those who do not.

Hypothesis 2: Approval/liking of President Trump will be negatively related to (a) wearing a mask, (b) washing one's hands, and (c) engaging in social distancing behaviors.

Study 1: Methods

Study Search and Inclusion Criteria

Seventeen nationally representative datasets including 26,876 participants were used in this study (see Table 1 and Supplemental Appendix A). To identify eligible studies, we conducted a search for datasets via Roper Center for Public Opinion Research (https://ropercenter.cornell.edu/). The Roper Center contains a database of public opinion data. We conducted the search using the terms “COVID” and “COVID-19” and filtered by downloadable datasets. As of July 2, 2020, this yielded 55 datasets. Two of these datasets were immediately eliminated because they did not generalize to the U.S. population (i.e., one was from Russia and the other consisted of only New Jersey residents).

Table 1.

Overview of Datasets Included in Study 1.

Dataset	Dates collected	N
1. Monmouth University National Poll: March 2020	March 18, 2020–March 22, 2020	851
2. Pew Research Center: American Trends Panel Wave 64	March 19, 2020–March 24, 2020	11,537
3. The March 2020 AP-NORC Center Poll	March 26, 2020–March 29, 2020	1,057
4. CNN Poll: April 2020 Coronavirus	April 3, 2020–April 6, 2020	1,002
5. Monmouth University National Poll: April 2020	April 3, 2020–April 7, 2020	857
6. ABC News/Ipsos Poll: 2020 Coronavirus Wave 4	April 8, 2020–April 9, 2020	512
7. SSRS: April 2020 Coronavirus Opinion Panel Survey	April 8, 2020–April 13, 2020	1,001
8. Washington Post-University of Maryland Coronavirus Poll Week 1	April 14, 2020–April 19, 2020	1,013
9. ABC News/Ipsos Poll: 2020 Coronavirus Wave 5	April 15, 2020–April 16, 2020	514
10. The April 2020 AP-NORC Center Poll	April 16, 2020–April 20, 2020	1,057
11. Monmouth University National Poll: May 2020 Coronavirus	April 30, 2020–May 4, 2020	808
12. CNN Poll: May 2020 Coronavirus	May 7, 2020–May 10, 2020	1,112
13. Kaiser Family Foundation Poll May 2020 Kaiser Health Tracking Poll	May 13, 2020–May 18, 2020	1,189
14. The May 2020 AP-NORC Center Poll	May 14, 2020–May 18, 2020	1,056
15. Fox News Poll: Coronavirus/Economy, 2020 Election: May 2020	May 17, 2020–May 20, 2020	1,207
16. Monmouth University National Poll: June 2020 Coronavirus	May 28, 2020–June 1, 2020	807
17. Kaiser Family Foundation Poll: June 2020 Kaiser Health Tracking Poll	June 8, 2020–June 14, 2020	1,296

Note. Additional demographic information is available in Supplemental Appendix A.

All datasets were screened for items that focused on President Trump's general approval (i.e., a proxy for affect), his management of the COVID-19 pandemic, and three PPBs (i.e., washing hands, wearing a mask or other face covering in public, and social distancing). Datasets were retained if they contained at least two of these variables, yielding 19 available datasets. One dataset was eliminated because although the codebook indicated items were collected, data were not available in the downloadable data at the time. One additional dataset was eliminated because the mask wearing item was double barreled (i.e., “have you bought or worn a protective mask, or not?”). This yielded 17 usable datasets.

Analyses

Items asking about four variables (i.e., President Trump's general approval/favorability, approval of his management of the COVID-19 pandemic, mask wearing, and hand washing) were straightforward and easy to combine across studies. However, this was not the case for the social distancing items. The seven studies inquiring about social distancing varied in the way questions were asked as well as the number of questions posed. Therefore, to avoid the “apples and oranges” comparison issue (Cortina, 2003), similar items were combined to yield four broad types of social distancing behaviors: avoiding large groups, avoiding groups in general, maintaining six feet of social distance, and avoiding friends and neighbors.

All datasets were weighted to account for sampling biases using the weight indicator provided in each study codebook. Descriptive statistics and correlations among variables of interest were calculated for each dataset. Once these correlations were calculated, we used Hunter and Schmidt’s (2004) meta-analytic techniques to weight the study's sample size as we estimated the average correlation across all available studies.

Study 1: Results/Discussion

Meta-analytic results are available in Table 2 and the corresponding coding information for their estimation is provided in Supplemental Appendix B. Hypothesis 1 was supported; individuals who approved of President Trump also strongly approved of his handling of the COVID-19 pandemic (). Because this meta-analytic correlation was so strong, we decided to test Hypothesis 2 by conceptualizing both of these variables (i.e., approval of Trump as President and approval of President Trump's handling of the COVID-19 pandemic) separately. Hypotheses 2(a)–2(c) proposed that individuals who approved of President Trump would be less likely to engage in PPBs. As can be seen in Table 2, the pattern of relationships between both general approval of President Trump and approval of President Trump's COVID-19 response as they relate to PPBs were similar. Hypothesis 2(a) was supported as those who approved of President Trump were less likely to wear masks or face coverings in public (). Although the relationships were much smaller in magnitude, Hypothesis 2(b) was also supported as those who approved of President Trump were less likely to wash their hands more as a result of the COVID-19 pandemic (). Finally, Hypothesis 2(c) was supported for all four types of social distancing. That is, participants who approved of President Trump were less likely to engage in social distancing behaviors ().

Table 2.

Study 1 Meta-analysis Results.

Independent variables	Dependent variables	Weighted N (total number of respondents)	k (number of samples)	r¯ (weighted mean correlation)
General approval	COVID-19 approval	21,572	9	.808
General approval	Wear a mask	7,477	4	−.324
COVID-19 approval		9,183	7	−.281
General approval	Wash hands	2,102	2	−.079
COVID-19 approval		3,143	3	−.088
General approval	Avoid large groups	2,102	2	−.173
COVID-19 approval		3,143	3	−.149
General approval	Avoid people (generally)	2,102	2	−.233
COVID-19 approval		4,144	4	−.168
General approval	Maintain 6 feet of social distance	1,001	1	−.068
COVID-19 approval		—	—	—
General approval	Avoid friends and neighbors	12,729	1	−.118
COVID-19 approval		2,041	2	−.090

Note. Weighted N is indicated by using the sample size determined by the weighting procedure specified for each dataset. This is not necessarily the actual number of participants surveyed.

Study 2: Introduction

The results of Study 1 supported the hypotheses that approval of Trump as President strongly related to participants’ evaluations of his performance (i.e., his handling of COVID-19) and of individuals’ PPBs (Hypotheses 1 and 2). In Study 2, we conducted a constructive replication of Study 1 (Köhler & Cortina, 2021), retesting Hypotheses 1 and 2 by exploring how affect towards President Trump, as measured by the LAQ (Martinko et al., 2018) relates to approval of his management of the COVID-19 pandemic, his administration's policies regarding COVID-19, and individuals’ PPBs. To do this, instead of using the single-item presidential approval measure used in most polls, we utilized a validated measure of affect, the LAQ (Martinko et al., 2018) with President Trump as the referent. We then examined whether the single-item approval measure was a meaningful and appropriate surrogate for affect toward President Trump using the LAQ. We also attempted to rule out any group-based identification explanation of follower behavior by controlling for political party. Finally, we extended Study 1 by testing two additional hypotheses that explored whether individuals’ cognitive effort toward understanding COVID-19 would moderate the relationship between liking President Trump and both approval of his handling of the COVID-19 crisis and U.S. residents’ PPBs. Below, we describe our rationale for the role of cognitive effort.

Decision making and Cognitive Effort

As Martinko et al. (2018) proposed, followers’ responses to surveys asking them to evaluate their leaders’ behavior can be viewed as a series of decisions. According to the dual processing model (Baumeister & Bushman, 2008), when decisions are not particularly important or stressful, individuals rely on heuristics (i.e., cognitive short cuts) to conserve cognitive resources (Fiske & Taylor, 1991). On the other hand, when decisions are important or stressful, individuals are more likely to expend the type of cognitive effort exemplified by rational decision-making models such as those proposed by Weiner (1986) as well as Lazarus and Folkman (1984).

Within the context of the dual processing theory, we argue that some people will spend considerable cognitive effort to understand the risks and scientific facts surrounding COVID-19, whereas others will be less concerned and not put forth much cognitive effort. When concerns about the COVID-19 crisis are not viewed as particularly important, we expect that heuristics, such as implicit leadership theories, will be activated by affect toward President Trump. We expect this affect (i.e., liking) toward President Trump will automatically signal approval of his handling of COVID-19. On the other hand, when respondents view the COVID-19 crisis as important, they will apply more cognitive effort to learning about it. In this case, we expect that more cognitive effort will attenuate the positive relationship between affect toward President Trump and perceptions of his handling of the pandemic. In other words, we expect the influence of affect toward President Trump on his management of the COVID-19 crisis to be stronger when little effort is put into understanding COVID-19 and more attenuated when the respondents expend cognitive effort toward understanding the pandemic.

Hypothesis 3: Cognitive effort will moderate the positive relationship between affect toward President Trump and approval of his handling of the COVID-19 crisis such that higher cognitive effort will weaken the relationship.

Similarly, as previously argued, we expect that the negative relationship between liking of President Trump and followers’ PPBs, based on the role modeling tenets that are central to the predictions of social learning theory, will be attenuated for those who have spent more cognitive effort to understand the issues related to COVID-19. Thus, we hypothesize:

Hypothesis 4: Cognitive effort will moderate the negative relationships between affect toward President Trump and (a) wearing a mask, (b) washing one's hands, and (c) engaging in social distancing such that higher cognitive effort will weaken the relationships.

Study 2: Methods

Participants and Procedure

We collected data from 522 participants between June 29, 2020 and July 2, 2020, using Qualtrics research services. We excluded four participants due to missing data, yielding a final sample of 518 participants. Eligible participants were at least 18 years of age and residents of the United States. The average age of the participants was 41.49 years (standard deviation = 15.84) and 48.10% were male. Additional demographic information is available in Table 3. All scales were randomized so that participants received them in a different order.

Table 3.

Sample 2 Demographic Information.

Category	Characteristic
Race/ethnicity[a]	White = 72.0%; Black = 15.3%; Latino = 9.8%; Asian = 5.2%; Native America/American Indian = 1.2%; Middle Eastern/North African = .4%; other = 1.4%
Marital status	Married = 46.9%; single = 38.8%; divorced = 9.3%; separated = 1.2%; widowed = 3.9%
Household income	Below $30,000 = 26.1%; $30,000–$49,999 = 22.0%; $50,000–$69,000 = 14.9%; $70,000–$89,999 = 10.4%; $90,000–$149,000 = 14.9%; above $150,000 = 11.9%
Social class	Not sure = 3.9%; poor = 11.4%; working class = 26.6% middle class = 38.4%; upper middle class = 15.4%; upper class = 4.2%
Education (highest degree earned)	Some high school = 3.5%; high school or equivalent = 23.4%; some college = 23.2%; associate's degree = 9.3%; bachelor's degree = 23.6%; master's = 14.1%; doctoral degree = 3.1%
Current work status	Full time = 63.6%; part time = 17.9%; furloughed = 6.9%; recently unemployed = 11.7%
Employment status as a result of COVID-19	Loss of a primary job by household member = 25.1% (separately, 21.8% of participants reported a pay cut and 15.6% reported a furlough as a result of COVID-19).
Political party affiliation	Democratic = 40.3%; Republican = 33.8%; unaffiliated = 23.2%; other = 2.7%

Participants could report more than one category and therefore these total over 100%.

Measures

Affect Toward President Trump

We adapted the 5-item LAQ from Martinko et al. (2018) with Donald Trump as the referent rather than an employee's immediate supervisor. A sample item was “I feel positively about Donald Trump.” Participants responded to items using a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree; Cronbach's alpha = 0.98).

Approval of Donald Trump as President

Following national polls, we asked respondents to indicate their approval of the job Donald Trump is doing as President using the item, “I approve of the job Donald Trump is doing as President.” Participants responded along a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Approval of President Trump's Handling of COVID-19

For this measure, we used a single item (i.e., “to what extent do you approve of the manner in which Donald Trump is handling the COVID-19 situation?”) similar to that used in the various polls included in Study 1. Participants responded using a 7-item Likert scale ranging from 1 (strongly disapprove) to 7 (strongly approve).

Approval of the Trump Administration's Handling of COVID-19

We used four items adopted from several polls (e.g., Gallup) to measure various elements of the administration's response to COVID-19. These items were more specific in nature and focused on tangible action items. A sample item was, “when it comes to implementing measures to limit the spread of COVID-19, how satisfied are you with the Trump Administration?” The other three items referred to “providing those who have lost a job or otherwise been negatively impacted with financial assistance,” “providing Americans with reassurance and direction during this uncertain period,” and “ensuring that there are enough medical supplies to treat COVID-19 (i.e., ventilators and protective gear for doctors, nurses and other first responders).” Participants responded using a 7-item Likert scale ranging from 1 (strongly disapprove) to 7 (strongly approve; Cronbach's alpha = 0.95).

Personal Protection Behaviors

To assess PPBs, we focused on three behaviors: wearing a face mask in public, washing one's hands, and social distancing. To assess mask wearing behaviors, we followed the protocol from several polls (e.g., [1] Fox News Poll: Coronavirus/Economy/2020 Election: May 2020 and [2] Kaiser Family Foundation Poll: May 2020 Kaiser Health Tracking Poll) and asked participants the following question: “How often do you wear a face covering in public?” Participants responded using a 6-point Likert frequency scale ranging from 1 (never) to 6 (very frequently).

Hand washing and social distancing were assessed using 1 and 9 items, respectively. We used a 5-point Likert scale ranging from 1 (a lot less) to 3 (about the same) to 5 (a lot more) for respondents to indicate the extent to which their behavior had changed “compared to before the COVID-19 pandemic.” To assess hand washing, we asked participants the following: “compared to before the COVID-19 pandemic, how frequently do you engage in the following activities?—Wash your hands.” Finally, to examine the full range of social distancing behaviors, we designed a measure of social distancing that extended beyond what has been included in national polls. To do so, we asked participants the extent to which they engage in a series of social distancing behaviors using the same stem “compared to before the COVID-19 pandemic….” Examples include “attend family gatherings,” “invite others inside your own home,” and “stay at home” (reverse-scored). This 9-item scale had strong reliability (Cronbach's alpha = 0.95).

Cognitive Effort

We used a 6-item measure of cognitive effort adapted from Krell (2017). A sample item was, “I have made an intellectual effort to understand the issues surrounding COVID-19.” Participants responded to this 6-item measure using a Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree; Cronbach's alpha = 0.73).

Political Party

To control for the possibility that political party allegiance, rather than affect toward President Trump, explained our hypothesized relationships, we controlled for participants’ political party affiliation. We coded political party such that 1 = Democratic Party and 2 = Republican Party. Only 384 of the 518 respondents reported that they fell into one of these two categories (i.e., some reported that they were independent or not affiliated). Therefore, analyses including this control variable have a sample size of 384 instead of 518.

Study 2: Results/Discussion

We used SPSS 25.0 to estimate descriptive statistics, correlations, and regression results. Descriptive statistics and correlations are available in Table 4. To demonstrate that approval of President Trump was a valid proxy for leader affect, we correlated the approval measure with the LAQ. Table 4 shows that the LAQ is strongly related to approval of President Trump (r = .91, p < .01). In fact, this relationship is so strong that study participants did not meaningfully distinguish between approval of President Trump and having positive affect toward him, as assessed by the LAQ. This finding provides strong justification for using the LAQ to test our hypotheses and facilitates comparisons between Study 1 and Study 2 results.

Table 4.

Study 2 Correlations and Descriptive Statistics.

	1	2	3	4	5	6	7	8	9	10
1. Political party	—
2. Leader affect questionnaire (LAQ)	.74**	(.98)
3. Trump approval	.70**	.91**	—
4. Trump COVID approval	.69**	.90**	.87**	—
5. Trump admin. COVID approval	.56**	.78**	.76**	.79**	(.95)
6. Wear a mask	−.24**	−.31**	−.27**	−.28**	−.20**	—
7. Wash hands	−.08	−.09*	−.16	−.04	−.05	.35**	—
8. Social distance	−.15**	−.34**	−.32**	−.34**	−.33**	.32**	.29**	(.95)
9. Cognitive effort	−.14**	−.32**	−.29**	−.32**	−.30**	.30**	.26**	.52**	(.73)
10. LAQ × cognitive effort	.72**	.92**	.86**	.84**	.73**	−.22**	.00	−.15**	.01	—
Mean	1.46	3.43	3.51	3.39	3.89	4.94	4.37	4.00	5.24	17.09
Standard deviation	.50	2.39	2.49	2.32	2.06	1.44	.99	1.03	1.16	12.35

Note. N = 518 for all variables except political party (N = 384). For replication purposes, the moderator term is included. Admin = administration. Political party is coded such that 1 = Democratic Party and 2 = Republican Party.

*p < .05. **p < .01.

Next, we retested Hypotheses 1 and 2 using the LAQ instead of “approval of Trump as President” as the independent variable. Hypothesis 1 proposed that individuals who have positive affect toward President Trump would be more likely to approve of his management of the COVID-19 pandemic. This hypothesis was supported. The LAQ was related to both approval of President Trump's COVID response (r = .90, p < .01) and his administration's response to COVID-19 (r = .78, p < .01). Likewise, Hypothesis 2 was fully supported. Those who reported positive affect toward President Trump (as assessed by the LAQ) were less likely to wear a face covering/mask in public (r = −.31, p < .01), less likely to wash their hands during the COVID-19 pandemic (r = −.09, p < .05), and less likely to engage in social distancing behaviors during the pandemic (r = −.34, p < .01). These results closely mirrored those obtained in Study 1, which used “approval of Trump as President” as the independent variable.

Finally, we conducted a series of moderation analyses to investigate whether cognitive effort moderated the relationships between liking President Trump (i.e., LAQ) and approval of his management of the COVID-19 pandemic (Hypothesis 3) as well as residents’ PPBs (Hypotheses 4a–4c). These results are presented in Table 5. The results controlling for political party are presented in Table 6. The results with and without political party as a control variable are not substantively different, so we report the results without the political party control variable throughout the text below to facilitate accurate interpretation of our findings. Hypothesis 3 was fully supported; cognitive effort moderated the relationship between affect toward President Trump (i.e., the LAQ) and approval of his handling of the COVID-19 pandemic (β = .676, p < .01; see Figure 1). As shown in Figure 1, the simple slopes for the lower (i.e., one standard deviation below the mean; b = .634, t = 4.726, p < .001) and higher (i.e., one standard deviation above the mean; b = .520, t = 11.628, p < .001) slopes were both significant.

Table 5.

Moderated Regression Results.

Dependent variables	Trump COVID-19 approval	Wearing a mask	Washing hands	Social distancing
Step 1
Leader affect questionnaire (LAQ)	.758**	−.238**	−.005	−.192**
Cognitive effort	−.057	.222**	.261**	.460**
R 2	.606	.140	.069	.305
Adjusted R2	.604	.137	.065	.302
Step 2
LAQ	.062	−.307	−.083	−.691**
Cognitive effort	−.283**	.200**	.236**	.298**
LAQ × cognitive effort	.676**	.067	.075	.485**
R 2	.628	.140	.069	.316
Adjusted R2	.626	.135	.064	.312
R2 change	.022**	.000	.000	.011**

Note. N = 518. We report standardized regression coefficients (β).

*p < .05. **p < .01.

Table 6.

Moderated Regression Results While Controlling for Political Party.

Dependent variables	Trump COVID-19 approval	Wearing a mask	Washing hands	Social distancing
Step 1
Political party	.689**	−.238**	−.076	−.145**
R 2	.474	.057	.006	.021
Adjusted R2	.473	.054	.003	.018
Step 2
Political party	.024	−.087	−.114	.168**
Leader affect questionnaire (LAQ)	.888**	−.165*	.107	−.338**
Cognitive effort	−.045*	.209**	.302**	.454**
R 2	.849	.127	.086	.343
Adjusted R2	.848	.120	.079	.338
R2 change	.375**	.070**	.081**	.322**
Step 3
Political party	.021	−.088	−.117	.163**
LAQ	.597**	−.241	−.175	−.814**
Cognitive effort	−.142**	.184*	.209*	.295**
LAQ × cognitive effort	.284**	.074	.275	.465*
R 2	.853	.127	.090	.353
Adjusted R2	.852	.118	.080	.346
R2 change	.004**	.000	.003	.010*

Note. N = 384. We report standardized regression coefficients (β).

*p < .05. **p < .01.

Figure 1.

Study 2 depiction of cognitive effort moderating the relationship between the leader affect questionnaire (LAQ) and Trump COVID approval.

Although not formally hypothesized, we note that the correlations in Table 3 indicate that those putting forth more cognitive effort to understand COVID-19 are significantly more likely to (a) wear masks (r = .30, p < .01), (b) wash their hands (r = .26, p < .01), and (c) engage in social distancing (r = .52, p < .01) than those putting forth less effort. Finally, we tested whether cognitive effort moderated the relationship between liking President Trump (i.e., LAQ) and PPBs (i.e., Hypothesis 4). Cognitive effort moderated the relationship between the LAQ and (c) social distancing behaviors (β = .485, p < .01), but not hand washing or mask wearing. Figure 2 indicates that those who exerted greater cognitive effort to understand the COVID-19 pandemic engaged in approximately the same (high) amount of social distancing behaviors regardless of how they felt about President Trump. However, those who did not exert cognitive effort to understand the pandemic were less likely to engage in social distancing, especially those who have positive affect toward President Trump. Additionally, as shown in Figure 2, the simple slopes for the lower (i.e., one standard deviation below the mean; b = −.258, t = −4.079, p < .001) and higher (i.e., one standard deviation above the mean; b = −.338, t = −3.779, p < .001) slopes were both significant.

Figure 2.

Study 2 depiction of cognitive effort moderating the relationship between the leader affect questionnaire (LAQ) and social distancing behaviors.

General Discussion

Results of nationally representative polls and our primary study indicate that measures of general affect/approval of President Trump explained a majority of the variance in ratings of approval of his leadership during the COVID-19 pandemic. Importantly, this affect was related to individuals’ adoption of PPBs. More specifically, individuals who like President Trump were less likely than others to engage in social distancing behaviors, wash their hands, and wear masks in public. This finding remained strong even after controlling for political party affiliation. This finding is important because political party could have provided the alternative explanation that group identity, rather than leader affect, was the underlying explanation for perceptions of the President's handling of the pandemic and PPBs. As mentioned in the opening paragraph, Democrats and Republicans seem to have different priorities pertaining to tradeoffs between the economy and controlling the spread of COVID-19, leading to potentially different conclusions about the President's handling of the pandemic. Our results show that leader affect has a significant effect on residents’ perceptions and behaviors above and beyond party affiliation.

Further, we expected that cognitive effort expended toward understanding the COVID-19 pandemic could affect these relationships, but we only found partial support for this notion. That is, cognitive effort attenuated the negative relationship between the LAQ and social distancing, but not mask wearing or hand washing. Interestingly, a high cognitive effort was related to greater social distancing, even when respondents had positive affect toward President Trump. For those putting forth little cognitive effort, positive affect toward President Trump was negatively related to social distancing.

It was interesting that cognitive effort was not strong enough to influence the relationships between liking President Trump and hand washing or mask wearing. In retrospect, it makes sense that hand washing behaviors were less impacted by affect for President Trump because hand washing is a familiar and unobtrusive behavior. However, we were surprised that cognitive effort did not impact the relationship between liking President Trump and wearing a mask in public because mask wearing has become politicized, with President Trump famously refusing to wear a mask in public (Blake, 2020). Consistent with social learning theory (Bandura, 1986), President Trump is a salient role model and his uncovered face is highly visible, whereas social distancing is harder to observe. We found that followers who have strong, positive affect toward President Trump were less likely to wear masks regardless of the amount of cognitive effort they may have put into understanding COVID-19. Thus, affect for President Trump apparently obfuscated whatever influence cognitive affect had on mask wearing.

In addition, evidence suggests that there is a growing antiscience bias in the United States (Sizemore, et al., 2019), meaning that many U.S. residents are increasingly distrustful of scientists and their advice. This bias may have become even more activated at the apparent “reverse course” that U.S. scientists exhibited when they changed the mask recommendation for the general population based upon new evidence (Cornell Alliance for Science, 2020).

Alternatively, the research on cognitive dissonance (Festinger, 1962) provides a competing perspective on the relationship between affect toward President Trump and mask wearing. According to dissonance theory, individuals who receive information that challenges their beliefs may respond by discounting that information and seeking confirming information. Thus, if individuals have positive affect toward President Trump, they will be more likely to discount negative information (i.e., news sources which challenge his handling of COVID-19), and seek information that supports his behavior and policies (i.e., news sources that support his handling of COVID-19). This confirmation bias will create the desire to “dig in” and more strongly defend/emulate President Trump. As a result, whether or not individuals educate themselves about COVID-19, participants may respond in a way that is consistent with their liking of President Trump, both in terms of their ratings of his management of the COVID-19 pandemic and their own PPBs. Thus, the effects of affect may be further reinforced through the respondents’ behavior to reduce cognitive dissonance as they engage in the decision-making process that is central to our theorizing.

Implications

There are several important implications for leadership in our study that extend beyond the COVID-19 pandemic. Specifically, our study's implications include: (1) individuals’ affect toward leaders explains significant variance in leaders’ evaluations; (2) followers’ affect for leaders is related to individuals’ behaviors; and (3) cognitive effort sometimes acts as a substitute for leadership.

First, our study contributes to the recent literature on leader affect by providing a compelling case for the existence of a general, underlying construct that explains the majority of variance in followers’ evaluations of their leaders (Banks et al., 2018; Yammarino et al., 2020). Approval of President Trump in the polls and the LAQ explained 65% and 85% of the variance in respondents’ approval of his handling of the COVID-19 pandemic, respectively. These results represent the upper end of the range reported by Martinko et al. (2018), whose findings revealed that affect, as measured by the LAQ, explained between 45% and 85% of the variance in leader ratings. Using a completely different context (i.e., national politics) and different measures (i.e., polling measures of Presidential approval across numerous sources), our findings increase our confidence that we can generalize the proposition that affect toward leaders accounts for the majority of the variance in leader ratings.

These results suggest that being liked is a critical factor in leader evaluations. Thus, all things equal, it is an asset to be liked, especially when followers have the power to influence leader outcomes, such as promotion, productivity, and compensation. Social psychological research suggests that similarity breeds liking (Byrne, 1961; Collisson & Howell, 2014), so leaders who wish to be evaluated favorably should seek opportunities to demonstrate similarities in the interests, attitudes, and personality of their followers (Montoya et al., 2008).

A second implication of our study is that, besides influencing followers’ evaluations of their leaders, affect is also related to followers’ behaviors. There is only limited evidence to date that the LAQ has influence beyond leader ratings and attitudinal variables (e.g., organizational identification). For instance, Martinko et al. (2018) examined the effects of the LAQ on employee deviance (replicating Tepper et al., 2009). The correlations between the LAQ and employee deviance explained between 10% and 20% of the variance in organization-directed and supervisor-directed deviance. In the present study, both the meta-analytic data and the results of our primary study reveal that the LAQ explains at least some of the variance in PPBs related to COVID-19. This finding is noteworthy not only because it directly links how one feels about a leader to one's behavior, but because those behaviors, by definition, protect the individual and those around him or her.

Finally, we provide evidence that knowledge, acquired through cognitive effort, can act as a substitute for leadership (Kerr & Jermier, 1978). Indeed, those individuals who put forth more cognitive effort toward understanding COVID-19 were more likely to wear masks, wash their hands, and engage in social distancing than those not putting in the same level of effort. Moreover, cognitive effort moderated the relationship between the LAQ and social distancing, but not the other PPBs. This result suggests that when individuals expend cognitive effort, they are less likely to be influenced by leader affect, resulting in more independent decision making. Thus, ideally, leaders, followers, and evidence-based practice benefits from a knowledgeable and educated populace in order to enhance individual decision making.

Limitations

There are a couple of limitations to note. First, the data in Study 2 were cross-sectional, so temporal order cannot be inferred. Second, some items used in the current studies were single-item measures. Although multi-item measures are preferable because they increase variability, this did not seem to be much of a problem in our study given that the LAQ measure in Study 2 was so highly correlated with single-item measures of approval. Moreover, even though concerns have been raised about conclusions that can be drawn from cross-sectional studies and single-item measures, recent research (e.g., Dierdorff & Ellington, 2008; Nagy, 2002; Spector, 2019) suggests that these concerns have been overstated. Moreover, the 17 datasets included in Study 1 are all representative of the U.S. population and typically use single-item measures. We argue that the benefits of these samples and measures outweigh shortcomings of the study design.

Conclusion

President Trump famously said: “I could stand in the middle of 5th Avenue and shoot somebody and I wouldn't lose voters.” Although this statement certainly was an exaggeration and not meant to be taken literally, President Trump's underlying philosophy that his supporters have such strong affect toward him that he can do whatever he pleases appears to be well founded. Results of the current line of research provide empirical support that if leaders are liked, they tend to be evaluated positively regardless of their policies and behaviors. It is also important to consider the reciprocal: leaders who are not liked will tend to be evaluated negatively, regardless of their policies and behaviors. Either way, affect appears to be a significant predictor of followers’ behavior, to their benefit or detriment.