Is Nationalism Rising in Times of the COVID-19 Pandemic? Individual-Level Evidence from the United States

Abstract

Following the advent of the COVID-19 pandemic, analysts have noted a global rise of nationalism as countries have engaged in a number of nationalist moves in response to the pandemic. However, the implication of policy changes at the individual-level remains unclear: do citizens support those nationalist government responses? More importantly, do people tend to be more nationalistic following the outbreak? Building on terror management theory (TMT), this article examines whether and how ideological beliefs affect individuals’ support for nationalist policies during the COVID-19 pandemic. According to TMT, to cope with death anxiety, people are predisposed to ideological defense, resulting in cohesion with individuals who validate their beliefs and hostility toward those who threaten them. Thus, we argue that when states’ nationalist policies are aligned with their ideology, people tend to support them; yet, when states’ nationalist policies contradict their ideology, people tend to withdraw their support. Specifically, this study found that as non-conservatives (compared to conservatives) are more concerned with the virus, they are more likely to show an inclination of ideological validation. Given that their ideology advocates more tolerance, non-conservatives are less likely to support nationalistic policies. To test the hypotheses, we applied structural equation modeling to a March 2020 CNN Poll (nationally representative US data about COVID-19). The statistical analysis demonstrated strong support for our arguments.

Introduction

As the COVID-19 pandemic continues to rage, countries have behaved in increasingly selfish ways: they shut down their borders, fight for medical supplies, and blame each other for the sources of the disease. The president of United States, Donald J. Trump, suspended entry of aliens who supposedly present a risk to the US labor market to protect American workers. Thus, a number of observers see a new epoch in the making, one in which nationalism tends to rise globally [1, 14, 23, 27, 51, 62–64, 66]. It is tempting to make inferences regarding the broader trends in the post-pandemic world based on those specific observations. However, it is possible that the observed increase in nationalism might prove to be temporary government responses to the pandemic, rather than reflect actual changes in the underlying attitudes or sentiments shared by people. Put differently, the implication of policy changes at the individual-level remains unclear: do people support those nationalist government responses? More importantly, did people tend to become more nationalistic following the pandemic?

It has long been hypothesized that nationalism thrives in times of crisis [7, 14, 16, 54]. Scholars have noted that a strengthening of nationalism happened after the Wall Street crash in 1929 and the global financial crisis in 2008, as states prioritized self-interest and took many nationalist moves [14, 54]. To explain the causal link between crisis and nationalism, the crisis-nationalism literature usually relies on social identity theory (SIT): insecurity and uncertainty caused by crisis predispose individuals to powerful identification with groups [32, 35, 61]. Meanwhile, crisis boosts the primacy of state as people rely on states to manage the threatening situation [7, 16, 18]. These twin imperatives will lead to a rise of nationalism that emphasizes the symbiotic relationship between nationhood and statehood [68]. According to scholars, health crises, similar to other crises, can arouse individuals’ nationalist emotions and trigger greater cohesion within countries and hostility between them [25]. However, this hasty conclusion neglects the peculiarity of the pandemic in arousing individuals. Unlike economic or political crises, what a global health crisis (the COVID-19 pandemic) directly threatens are not assets or rights, but people’s lives. As many social theorists have suggested, death anxiety is a critical driving force of people’s attitudes, beliefs, and behaviors [49]. The intensified concerns about death during health crises can trigger a distinct psychological mechanism of response and thereby complicate nationalism resulting from this type of crisis.

This study relied on terror management theory (TMT) to assess the impact of death awareness on nationalism during the COVID-19 pandemic. TMT posits that in the face of crisis when people’s lives are at stake and death is made momentarily salient, individuals tend to bolster their self-esteem to reduce death anxiety, which leads to cohesion among people who share similar worldviews and hostility toward those with alternative worldviews [30]. In this sense, strong group identification in response to the pandemic might be intertwined with a defense of worldviews. Building on TMT, this article examines whether and how ideological beliefs affect individuals’ support for nationalist policies during the outbreak. We argue that during the pandemic, people’s concern with their own mortality may arouse a strong desire to defend their own ideological beliefs. As a result, ideological validation, not national identification, is primary to uncertainty reduction needs. When states’ nationalist policies are aligned with their ideology, people tend to support them. In contrast, when states’ nationalist policies contradict their ideology, people tend to withdraw their support. Specifically, the study found that compared to conservatives, non-conservatives are more concerned with the virus. The pandemic’s death prime predisposes them to ideological validation. As their ideology advocates more tolerance, they are less likely to support nationalistic policies. To test the hypotheses, we used Cable News Network (CNN) survey data from the US to explore how people from different ideological groups have responded to one of the primary US nationalist policies, the international travel ban.

The reminder of this article begins with a discussion of existing theories on crisis and nationalism. By focusing on a mortal crisis (the COVID-19 pandemic), TMT is first introduced to set the theoretical stage for future analysis. Unlike SIT, TMT with a focus on death anxiety is a theoretical tool to understand the psychological origin of individual response to death-causing crises. Building on TMT, this article further investigates the interplay among the pandemic, nationalism, and ideology and generates testable hypotheses. Next, an analysis of individual-level survey data from the US is reported examining how individuals with different ideological beliefs perceive current US nationalist policies. Then a brief discussion and conclusion follows.

Literature Review

Uncertainty reduction and anxiety management have been the focus of many studies on human social behavior [5, 11, 35]. Anxiety and threat in times of crisis could have a variety of psychological consequences. Researchers agree that nationalism is one of the consequences triggered by individuals’ motivation to alleviate anxiety [7, 14, 49, 54]. However, they disagree on the mechanisms that link anxiety with nationalism.

Conventional wisdom has identified two psychological pathways people may follow to protect themselves against threat and anxiety in times of crisis: group affiliation and ideological validation. The first pathway, introduced primarily by SIT, suggests that it is an evolutionary imperative that individuals tend to affiliate with a group to reduce uncertainty and threats caused by crisis [32, 35]. In other words, individuals are predisposed to powerful identification with groups that could better reduce their anxiety. Therefore, people may demonstrate strong national identification as their state assumes the major responsibility to protect them during a crisis [7, 18]. This causes the rise of nationalism during a crisis. SIT also incorporates ideologies into the framework. For instance, SIT has recognized the association between ideological beliefs (for example, authoritarianism and conservatism) and conditions of uncertainty amid crisis [34]. Michael A. Hogg reported that in some situations, people might strive to belong to groups with extreme ideologies, since these groups, being distinctive in their nature, are thought to be able to effectively reduce uncertainty [33, 36]. Thus, “to the extent that such belief systems are tied to group memberships, identification may mediate the link between social uncertainty and ideology” ([34]: 121). For SIT, it is possible that people simultaneously affiliate with different groups, such as racial, national, and ideological groups. In this sense, ideological beliefs and nationalism have no direct relationship and are both consequences of group affiliation. Specifically, ideology matters only when it fulfills people’s affiliation needs, and nationalism is the manifestation of their desire to side with the nation-state. In general, SIT may predict that regardless of their ideologies, people are more likely to be nationalistic in times of crisis as it fulfills their affiliation needs.

The second pathway, proposed by TMT indicates that the existential concerns arouse not only a desire to affiliate with a group but more importantly a desire to defend one’s ideologies. Unlike SIT with an interest in general anxiety, TMT focuses specifically on death anxiety. Inspired by the writings of Ernest Becker [5, 6], social psychologists Jeff Greenberg, Tom Pyszczynski, and Sheldon Solomon developed TMT to understand how people manage their existential concerns [30, 31]. Anchored in evolutionary theory, TMT begins with the broad assumption that although human beings share with other species a self-preservation instinct, only human are aware of their inevitable personal mortality. The clash of a desire for life with the awareness of inevitable death imposes an immense potential for paralyzing terror. To manage this potential, people develop a dual-component cultural anxiety buffer that consists of a cultural worldview and self-esteem [53]. According to TMT, cultural worldview is a set of shared beliefs of reality that imbues the world with meaning, order, stability, and permanence. Cultural worldviews also provide standards for how people should behave and the promise of literal and/or symbolic immortality to those who comply with the standards of value prescribed by the worldview. Self-esteem is defined as beliefs that one is valuable and living up to standards of value set by the cultural worldview with the hope of transcending death and attaining immortality.

TMT postulates that self-esteem and cultural worldviews help buffer people against death-related terror and anxiety. Based on this, a key hypothesis, the mortality salience hypothesis, is proposed in TMT. Specifically, reminders of mortality (that is, mortality salience) will increase the need to bolster one’s self-esteem and affirm faith in one’s worldview, which results in increased aggression toward those with different worldviews and attachment toward those who validate their beliefs [60]. A handful of studies confirmed this hypothesis [15, 29, 38, 52, 57, 65]. For instance, Burke et al. (2010) conducted a meta-analysis on empirical trials investigating the mortality salience hypothesis of TMT [10]. The study further examined 164 articles with 277 experiments and found that the hypothesis is robust.

Admittedly, “pondering one’s demise is likely to have a variety of consequences on the individual’s cognition, emotion, and behavior” ([13]: 604). Thus, it could be possible and compatible that death primes simultaneously lead to both strong national identification and defense of ideology, as some studies suggested that the defense of worldviews is aligned with strong national identification because mortality-salience-induced biases can extend to nations that in most cases represent groups sharing similar worldviews ([49]: 885). However, it is also possible that the two mechanisms and subsequent outcomes might contradict each other [70]. In particular, when the national group advocates values contrary to people’s ideological beliefs, it raises the question: which one would individuals be predisposed to, ideological validation or national group affiliation?

Without additional conditions, it is difficult for SIT to answer this question. The incompetency of SIT is due to two reasons. First, without making a distinction between existential anxiety and non-existential anxiety, SIT has neglected some special psychological consequences of mortal crisis, such as ideological validation proposed by TMT. According to evolutionary psychologists, existential concerns are among the major driving forces of human behaviors [5, 6, 30]. Thus, existential concerns may predispose people to different behaviors that have been missing in SIT.

Second, the priority of SIT is to confirm the existence of affiliation need when people are faced with anxiety, not the interplay among different affiliation choices. For SIT, both ideological validation and national group affiliation are manifestations of people’s affiliation needs. Although SIT advocates have recognized the possibility of multiple loyalties, they have not investigated how cross-cutting or multiple identities may change the face of nationalism. In contrast, TMT provides a clear answer to the question: ideological defense outweighs group affiliation to reduce death anxiety. In other words, it is the ideology that mediates the link between uncertainty and nationalism. People are not necessarily more nationalistic in times of crisis. Whether one becomes more nationalistic is contingent on their ideology: if a person’s ideological belief aligns with their national identification, they will become more nationalistic; however, if their ideological belief contradicts their national identification, they are more likely to withdraw support for the nation-state.

Given that the COVID-19 pandemic is a mortality reminder in which it is possible that people’s nationalistic inclination contradicts their ideological beliefs, the following discussion will rely on TMT to investigate in more detail the interplay among the pandemic, nationalism, and ideology.

The Pandemic, Nationalism, and Ideology

As the coronavirus spread rapidly across the world causing more than 500,000 deaths, the pandemic is a mortality reminder and causes considerable death anxiety among individuals. This anxiety soon triggered some public and national responses. At the beginning of the outbreak, street violence against Asians occurred that extended to other nationals, such as Iranians and Italians, in late February. On March 16, President Donald Trump suddenly referred to the COVID-19 coronavirus by the “Chinese virus” intentionally, which ushered in the propaganda war between the US and China [58]. A number of analysts were concerned that these responses manifest a striking trend: nationalism rising globally [1, 14, 23, 27, 63, 64, 66].

However, there is a lack of support for the hypothesis at the individual level, leaving a major gap in our understanding of nationalism during the pandemic. Despite the general application of psychological theories to build a link between the pandemic and nationalism, analysts tend to use empirical evidence at the group level, as previously mentioned, to support the claim that the outbreak has accelerated the advancement of nationalism [1, 14, 64]. This inconsistency implies the conceptual confusion of “nationalism” in general. Nationalism can be expressed in numerous forms with different meanings relating to various levels of analysis: nationalism as an xenophobic policy, an ideology, a movement, and individual sentiment [19]. The variety of expressions makes it difficult to conclude that the previously mentioned responses reflect a rise of nationalistic sentiment at the individual level because these responses are either from the government or from a small group. As Druckman (1994) correctly pointed out, “Although granting that nationalism is a political, economic, and sociological phenomenon, it becomes a social-psychological phenomenon to the extent that individuals develop attitudes about their own and other nations” ([20]: 43). In this case, however, it remains unclear whether individuals have become more nationalistic as a result of the pandemic.

To understand individuals’ attitudes and responses to a certain event, it is important to know first how they perceive it [37]. Researchers note that the pandemic has become a highly politicizing issue in a number of countries, such as the US [9]. Specifically, people have revealed an ideological divide in perceptions of the pandemic: conservatives in the US are less concerned about the virus than those with other ideologies [17].1 A recent Quinnipiac University national poll revealed that only 35% of conservatives (compared to 68% of liberals and 57% of moderates) were concerned that they or someone they know will be infected with the coronavirus [46]. Due to the considerable divergence on the perception of the virus threat between conservatives and non-conservatives, the following sections focus on the comparison between the two groups. According to TMT, to cope with anxiety, people are predisposed to ideological defense, resulting in cohesion with individuals who validate their beliefs and hostility with those who threaten them. Thus, if conservatives are less concerned about the virus, it suggests that they are less likely to be threatened by the mortality reminder. For them, the coronavirus may only present a medium threat level not as severe as a death threat, so the threat is less likely to activate conservatives’ ideological defense mechanism. This leads to the first two hypotheses:

1. Higher threat levels of the COVID-19 pandemic will be less likely to induce conservatives to be more conservative.

2. Higher threat levels of the COVID-19 pandemic will be more likely to induce non-conservatives to be more liberal.

As Rothgerber et al. (2020) argued, “Political ideology not only represents shared beliefs, opinions, and values held by an identifiable group or constituency, but also endeavors to describe and interpret the world and envision the world as it should be” ([56]: 4–5). Thus, political ideology provides people with a lens through which they process information and thereby influence their political attitudes and policy preferences [39]. Given the ideological divide in perceptions of the virus, we could expect an ideological divide in their nationalistic attitudes.

Liberal and conservative as concepts have multiple dimensions [39]. Central to the notion of conservativism, and therefore to the distinction between conservatism and non-conservatism, is the concept of intolerance [13, 40]. Generally, conservatives demonstrate more of a preference for exclusive or discriminatory policies. As conservatives are less likely to be aroused by the pandemic’s death prime, their attitude toward nationalistic policies are less likely to be affected by the death threat of the virus, but rather reflect conservatives’ core political belief (that is, intolerance) or the national affiliation need as proposed by SIT. Therefore, stemming from TMT, one could expect that the death prime would only enhance tolerance among non-conservatives. For non-conservatives, as the nationalistic policies contradict the core of their ideology (that is, tolerance [69]), they will tend to defend their ideology. Accordingly, non-conservatives are less likely to support nationalistic policies, such as the international travel ban. Fig. 1 illustrates the major causal mechanism.

Fig. 1.

Causal mechanism linking the pandemic to individuals’ attitudes toward nationalist policies

Based on these discussions, three more hypotheses are proposed.

• Hypothesis 3:

  With other conditions equal, conservatives, compared to non-conservatives, are more likely to support nationalistic policies, such as the international travel ban.

• Hypothesis 4:

  For conservatives, the increase of the COVID-19 threat has no significant impact on their support for nationalistic policies, such as the international travel ban.

• Hypothesis 5:

  For non-conservatives, with the increase of the COVID-19 threat, they tend to be more liberal, which results in a decrease of support for nationalistic policies, such as the international travel ban.

Research Design

Data

To test the hypotheses, we utilized data from a March 2020 CNN Poll [12]. This was the latest nationally representative dataset about COVID-19 that was available when this article was written. The poll phone interviewed US adults over age 18 from March 4 through March 7, 2020. Random digital dial (RDD) methodology was adopted to produce an equal probability selection method (EPSEM) sample of residential telephone numbers that identified the respondents. The same techniques were applied to both landline and cell phone numbers nationally. In total, 1211 US adults were included in the survey. This national poll collected data on individual-level attitudes of a range of policies, enabling researchers to link these attitudes with individuals’ coronavirus threat.

Methods

We used structural equation modeling (SEM) for the analysis. SEM can simultaneously estimate different equations/relationships, enabling both the multiple group analysis and the mediation analysis [42, 48]. The multiple group analysis in SEM is best suited for analyzing the differences between conservatives and non-conservatives, and the mediation analysis facilitates the estimation of how the threat of COVID-19 influences support for the international travel ban through political ideology. Another advantage of SEM is that it has a convenient and powerful technique of handling missing data, the full information maximum likelihood (FIML) method [21, 22]. The FIML and multiple imputation are two mainstream methods of handling missing data [3, 45]. However, compared with multiple imputation, the FIML yields results that are unaffected by the imputation model and asymptotically efficient [2]. Thus, SEM is an appropriate method for this study.

In SEM, the independent variables should be continuous variables. For a categorical independent variable, a series of dummy variables is often created and then included in the model. The ordinal dependent variables were treated as continuous variables in our SEM analysis for the parsimoniousness of modeling and ease of interpretation. To address the potential challenges to our modeling strategy, particularly the use of ordinal variables as continuous variables, we also conducted a series of robustness checks at the end of the analysis, utilizing ordered and multinomial logistic models for categorical dependent variables and the Karlson–Holm–Breen (KHB) method for mediation analysis [59]. Multiple imputation was employed for missing data in robustness checks.

Key Variables

This article focuses on individuals’ attitude toward nationalist policies. Although the CNN Poll dataset does not directly measure it, we can make up for the lack of ideal dependent variable by focusing on one prominent policy adopted by the government in response to the pandemic: the international travel ban. As previously mentioned, the international travel ban has been perceived as a major nationalist move following the pandemic, so people’s preference for the policy could be seen as a robust indicator of their overall attitude toward nationalist policies. Therefore, the dependent variable used herein is individuals’ support for the international travel ban.2 It has three different levels: 0 represents undecided, 1 denotes screening all international travels and admitting those without COIVD-19 symptoms, and 2 stands for banning all international travelers. The “undecided” attitude is included as the lowest level of support for the international travel ban because it is indeed an expressed attitude that is different from missing values while it imposes the fewest restrictions on travelers.

The mediator variable is political ideology, which serves as both an independent variable predicting the attitude toward nationalist policies and a dependent variable that is affected by the threat of COVID-19. It has five gradients ranging from 1 to 5: very conservative, conservative, moderate, liberal, and very liberal. For the ease of analyzing group differences, respondents are also divided into two groups according to their political ideology: conservatives (that is, an ideology that is very conservative or conservative) and non-conservatives (that is, an ideology that is moderate, liberal, or very liberal).

The key independent variable is the degree of the COVID-19 threat, which is evaluated by the likelihood of someone in the respondent’s local community being infected by COVID-19. It is coded to range from 1 to 4 denoting not likely at all, not too likely, somewhat likely, and very likely, respectively.3

Control Variables

The respondents’ demographic characteristics and socioeconomic status are included as control variables as they are considered able to influence public attitudes by existing studies [24, 50]. Specifically, we control for people’s age, gender (1 = male and 0 = female), and race (1 = whites and 0 = other races) as demographic characteristics, and people’s household annual income, education, and employment status as indicators of socioeconomic status. As these three indicators of socioeconomic status are not continuous variables in the dataset, they are transformed into dummy variables suitable for SEM. Income is coded as whether household annual income equals or surpasses $50,000 (1 = yes and 0 = no). Education is coded as whether respondents have a graduate degree or not (1 = yes and 0 = no). Employment status is converted into two dummy variables: being employed (1 = yes and 0 = no) and being retired (1 = yes and 0 = no).4 These two are included because they are distinct employment statuses in the special situation of COVID-19.

The descriptive statistics of all the variables are listed in Table 1.

Table 1.

Descriptive Statistics of Variables Used in Models

	Mean	SD/Proportion	Min	Max	Observations
Key Variables
Support for Travel Ban	1.256	0.526	0	2	1211
Degree of COVID-19 Threat	2.620	1.039	1	4	1190
Political Ideology	2.803	1.174	1	5	1176
Conservative (1 = Conservative, 0 = Non-conservative)	0.349		0	1	1211
Control Variables
Gender (1 = Male, 0 = Female)	0.548		0	1	1211
Age	54.194	19.124	18	96	1192
Race (1 = White, 0 = Other)	0.704		0	1	1211
Household Annual Income (1 = $50 K or above, 0 = below$50 K)	0.664		0	1	1211
Education (1 = Graduate Degree, 0 = No Graduate Degree)	0.156		0	1	1202
Being Employed	0.522		0	1	1211
Being Retired	0.340		0	1	1211

Results

Main Results

Table 2 shows how the degree of the COIVD-19 threat affects political ideology differently for conservatives and non-conservatives (Model 1). For conservatives, the degree of the COVID-19 threat has no significant effect on their political ideology. But for non-conservatives, the degree of the COVID-19 threat is positively related to their political ideology, that is, non-conservatives become more liberal as the degree of the COVID-19 threat increases. These findings support Hypothesis 1 and Hypothesis 2, illustrating that the COVID-19 threat only activates the ideological validation mechanism of non-conservatives. It should be noted that such a difference in statistical significance between conservatives and non-conservatives is not attributed to different sample sizes. Typically, a larger sample is more likely to yield statistically significant estimates because a large sample reduces standard errors. However, this is not the case in this study. Although there are fewer conservatives than non-conservatives in the sample (423 vs 788), the standard error for conservatives is smaller than for non-conservatives. Thus, the difference between conservatives and non-conservatives is not an artifact due to their difference in sample size.

Table 2.

Maximum Likelihood Estimates for the Relationship between the Degree of the COVID-19 Threat and Political Ideology (Model 1)

Parameter	Unstandardized Coefficients
	Conservatives	Non-conservatives
Political Ideology
← Degree of COVID-19 Threat	0.019 (0.023)	0.066* (0.028)
← Gender (1 = Male, 0 = Female)	−0.005 (0.052)	−0.142** (0.054)
← Age	−0.001 (0.002)	−0.008*** (0.001)
← Race (1 = White, 0 = Nonwhite)	−0.114† (0.060)	0.090 (0.059)
← Household Annual Income (1 = $50 K or above, 0 = below$50 K)	−0.052 (0.056)	−0.077 (0.059)
← Education (1 = Graduate Degree, 0 = No Graduate Degree)	−0.043 (0.083)	0.205** (0.072)
← Being Employed	−0.039 (0.082)	−0.050 (0.082)
← Being Retired	−0.098 (0.093)	0.052 (0.105)
Observations	423	788

Note: Goodness of fit test statistics are not reported since the model is a saturated model

Standard errors in parentheses.

^†<0.1, * < 0.05, ** < 0.01, *** < 0.001

Table 3 outlines the differences in support for the international travel ban across the respondents’ various conditions (Model 2). Compared with non-conservatives, being a conservative is associated with a 0.177 unit increase in support for the international travel ban after controlling for demographic characteristics and socioeconomic status. The degree of the COVID-19 threat has no significant direct impact on the level of support for travel. This evidence confirms Hypothesis 3, indicating that although the virus threat did not activate conservatives’ ideological validation mechanism, their ideological belief predisposes them to support the travel ban.

Table 3.

Maximum Likelihood Estimates for the Relationship between Conservative Ideology and Support for the International Travel Ban (Model 2)

Parameter	Unstandardized Coefficient
Support for Travel Ban
← Conservative	0.177*** (0.032)
← Degree of COVID-19 Threat	0.005 (0.015)
← Gender (1 = Male, 0 = Female)	0.003 (0.031)
← Age	0.002 (0.001)
← Race (1 = White, 0 = Nonwhite)	−0.040 (0.034)
← Household Annual Income (1 = $50 K or above, 0 = below$50 K)	−0.115*** (0.033)
← Education (1 = Graduate Degree, 0 = No Graduate Degree)	−0.065 (0.043)
← Being Employed	0.040 (0.047)
← Being Retired	−0.014 (0.058)
Observations	1211

Note: Goodness of fit test statistics are not reported since the model is a saturated model

Standard errors in parentheses.

^†<0.1, * < 0.05, ** < 0.01, *** < 0.001

Table 4 shows the results from the model simultaneously estimating the effects of a series of variables on both political ideology and support for the international travel ban (Model 3). Consistent with the findings presented in Table 2, an increased degree of COVID-19 has no significant impact on conservatives’ political ideology but significantly makes non-conservatives more liberal. In addition, among conservatives, being less conservative is associated with reduced support for the international travel ban. In other words, the most conservative have the greatest support for the international travel ban. Similarly, among non-conservatives, being more liberal is associated with less support for the international travel ban. These results regarding the relationship between political ideology and support for the international travel ban are consistent with the findings in Table 3.

Table 4.

Maximum Likelihood Estimates for the Mediating Effect of Political Ideology on the Relationship between Degree of COVID-19 Threat and Support for Travel Ban (Model 3)

Parameter	Unstandardized Coefficients
	Conservatives	Non-conservatives
Support for Travel Ban
← Political Ideology	−0.117* (0.053)	−0.091*** (0.025)
← Degree of COVID-19 Threat	0.019 (0.025)	0.005 (0.019)
← Gender (1 = Male, 0 = Female)	0.080 (0.056)	−0.049 (0.036)
← Age	0.003 (0.002)	−0.000 (0.001)
← Race (1 = White, 0 = Nonwhite)	0.029 (0.066)	−0.064 (0.039)
← Household Annual Income (1 = $50 K or above, 0 = below$50 K)	−0.205*** (0.061)	−0.083* (0.039)
← Education (1 = Graduate Degree, 0 = No Graduate Degree)	−0.025 (0.090)	−0.064 (0.049)
← Being Employed	−0.004 (0.089)	0.060 (0.054)
← Being Retired	−0.145 (0.101)	0.061 (0.069)
Political Ideology
← Degree of COVID-19 Threat	0.019 (0.023)	0.066* (0.028)
← Gender (1 = Male, 0 = Female)	−0.005 (0.052)	−0.139* (0.054)
← Age	−0.001 (0.002)	−0.007*** (0.001)
← Race (1 = White, 0 = Nonwhite)	−0.114† (0.060)	0.088 (0.059)
← Household Annual Income (1 = $50 K or above, 0 = below$50 K)	−0.051 (0.056)	−0.075 (0.059)
← Education (1 = Graduate Degree, 0 = No Graduate Degree)	−0.043 (0.083)	0.206** (0.072)
← Being Employed	−0.039 (0.082)	−0.051 (0.082)
← Being Retired	−0.098 (0.093)	0.052 (0.105)
Support for Travel Ban ← Political Ideology ← Degree of COVID-19 Threat	−0.006* (0.003)	−0.002 (0.003)
Observations	423	788

Note: Goodness of fit test statistics are not reported since the model is a saturated model

Standard errors in parentheses.

^† < 0.1, * < 0.05, ** < 0.01, *** < 0.001.

Most importantly, at the bottom of Table 4 is the mediating effect of political ideology on the relationship between the degree of the COVID-19 threat and support for the international travel ban. Although the degree of the COVID-19 threat has no significant direct effect on support for the international travel ban, it can indirectly influence support for the international travel ban by affecting political ideology among non-conservatives. Specifically, a one-unit increase in the degree of the COVID-19 threat, via affecting non-conservatives’ political ideology, reduces their support for the international travel ban by 0.006 unit. This indirect effect is essentially the product of the effect of the degree of the COVID-19 threat on political ideology and the effect of political ideology on support for the international travel ban, both of which are less than 1. Two factors less than 1 multiplied by each other will yield a product much smaller than themselves. This explains why this indirect effect is small in effect size. Despite the relatively small effect size, it reveals a critical difference between conservatives and non-conservatives since such an indirect effect of the degree of the COIVD-19 threat on support for the international travel ban, no matter how small in size, is only statistically significant among non-conservatives rather than among conservatives (even if both have the same standard error). This evidence confirms Hypotheses 4 and 5.

Robustness Check

One major challenge to the results is that the dependent variables in our models, support for the international travel ban and political ideology, are essentially ordinal variables, while SEM assumes that they are continuous. To account for this challenge, we conducted a series of robustness checks by running logit models (ordered or multinomial logit models) using the Karlson–Holm–Breen (KHB) method for mediation analysis. Multiple imputation was employed to handle missing data. To obtain unbiased and efficient point estimates as well as replicable standard errors not varying much across different imputations, 200 imputations were computed, which are more than necessary as suggested by the current literature [28, 67].

The results of the robustness checks are listed in Table 5. For Model 1 (in Table 2) predicting political ideology, the ordered logit model yields the same result as SEM. Conservatives are indifferent to the degree of the COVID-19 threat while non-conservatives become more liberal with the increase of the degree of the COVID-19 threat.

Table 5.

Robustness Check

	Unexponentiated Coefficients
	Conservatives	Non-conservatives
Robustness Check for Model 1 Ordered Logit Model of Predicting Political Ideology
Degree of COVID-19 Threat	0.080 (0.096)	0.216** (0.080)
Control Variables	Included	Included
Observations	423	788
Robustness Check for Model 2 Predicting Support for Travel Ban
Option A: Ordered Logit Model
Conservative	0.721*** (0.132)
Control Variables	Included
Observations	1211
Option B: Multinomial Logit Model (1 = Screening as reference, 2 = Ban as Outcome)
Conservative	0.749*** (0.139)
Control Variables	Included
Observations	1211
Robustness Check for Model 3
KHB Method of Estimating the Indirect Effect of Degree of COVID-19 Threat on Support for Travel Ban Through Political Ideology
Option A: Ordered Logit Model
Indirect Effect	−0.009 (0.012)	−0.029* (0.015)
Control Variables	Included	Included
Observations	423	788
Option B: Multinomial Logit Model (1 = Screening as reference, 2 = Ban as Outcome)
Indirect Effect	−0.011 (0.014)	−0.031* (0.016)
Control Variables	Included	Included
Observations	423	788

Note: Multiple imputation (200 imputations) is used to handle missing data

Standard errors in parentheses.

^†<0.1, * < 0.05, ** < 0.01, *** < 0.001

Regarding Model 2 (in Table 3) that predicts support for the international travel ban, an ordered logit model is first estimated. Consistent with the previous finding, being conservative is found to be more supportive of the travel ban. To address a potential concern that the attitude “undecided” is non-parallel to “screening” and “ban” and thus support for international travel ban should be regarded as a purely nominal variable, a multinomial logit model is also estimated with “screening” as the reference category. Of note, being conservative is still associated with an increased possibility of supporting the travel ban in contrast to screening only.

Finally, regarding Model 3 (in Table 4) that estimates the indirect effect of the degree of the COVID-19 threat on support for the international travel ban, we used the KHB method to conduct mediation analysis. The KHB method is a popular technique that estimates the mediating effects in non-linear models (for example, the logit models) [8, 41, 43]. As shown at the bottom of Table 5, the indirect effect represents the indirect effect of the COVID-19 threat on support for the international travel ban through the mediator political ideology. No matter whether we used ordered logit model or multinomial logit model, an increase in the degree of the COVID-19 threat always indirectly reduces support for the international travel ban among non-conservatives, while among conservatives, there is no significant link between the degree of COVID-19 and support for the international travel ban. These results are consistent with the previous SEM findings.

In summary, although the SEM coefficients in our major analysis and the logit models’ coefficients in the robustness checks cannot be directly compared as they are in different models, they are the same in statistical significance and effect direction. The conclusions drawn from these results are also the same. Therefore, our findings are not contingent on specific models and are thus robust enough.

Conclusion

With the significance and potential impact of the COVID-19 pandemic far beyond the health realm, the pandemic and its aftermath have been seen as a game-changer for the world. As an article in Foreign Policy concluded, “Like the fall of the Berlin Wall or the collapse of Lehman Brothers, the coronavirus pandemic is a world-shattering event whose far-ranging consequences we can only begin to imagine today” [1]. Therefore, from the beginning of the outbreak, analysts have continued to predict what a post-coronavirus world will look like. A global rise of nationalism appears to be a popular but hasty answer to the question. Observing a number of nationalist moves taken by states, Farage (2020) directly pointed out that “we are all nationalists now” [23]. However, without robust theoretical and empirical support, this prediction could be untenable and misleading.

First, nationalism is not new but has been ubiquitous since it was constructed [4, 47, 55]. As Bieber (2020) noted, “Nationalism is deeply engrained into the global system and most societies around the world. As such, the global coronavirus pandemic and state responses are not going to fundamentally alter this reality” ([7]: 10). Second, given that nationalism has a variety of expressions, it will lead to a misunderstanding that various forms of nationalism share a common logic. In particular, nationalistic policies during the pandemic, such as the closure of borders, may be rational temporary responses and thus not reliable to predict the post-pandemic world. In contrast, nationalism, as a group ideology or sentiment shared by individuals, may have a relatively long-lasting impact on the world after the crisis.

Thus, we investigated whether people have become more nationalistic following the outbreak. Relying on the TMT, we argued that the fear of death triggered by the pandemic made people cling to their ideology, rather than making everybody more nationalistic: when states’ nationalist policies are aligned with their ideology, people tend to support them; yet, when states’ nationalist policies contradict their ideology, people tend to withdraw their support. Specifically, we found that as non-conservatives (compared to conservatives) are more concerned with the virus, they are more likely to show an inclination of ideological validation. Given that their ideology advocates more tolerance, non-conservatives are less likely to support nationalistic policies.

Although this study’s findings lend little support for the claim (that is, a rise of nationalism) at the individual level, they should not be seen as a source of optimism either. As both SIT and TMT predict a divided world in times of crisis, it is not controversial that the pandemic will usher in a new era of global division. However, people disagree on how the pandemic divides the world. This study demonstrated that the pandemic exacerbates not only the divides along national borders, but more importantly the divides among ideologies. Despite the possibility that ideological validation may mitigate the effect of national affiliation, it is also feasible that nationalism is aligned with people’s ideology. Therefore, a most concerning post-coronavirus world will be one in which national borders and ideological boundaries are congruent. In that case, global division and tension are more difficult to assuage. Over the long term, threatening nationalism and ideological division will further strengthen hostility among nations with different ideologies, accelerating retreat from globalization [26, 44], and eventually remaking the world.

Biographies

Ruolin Su

(Ph.D., University of Pennsylvania) is assistant professor of international relations in the School of International and Public Affairs at Shanghai Jiao Tong University. Her research focuses on international security, foreign policy analysis, and political psychology.

Wensong Shen

(Ph.D., University of Pennsylvania) is assistant professor in the Department of Sociology at the Chinese University of Hong Kong. His research interests center on social stratification, education, and quantitative methods.