Out-of-Control COVID-19 Pandemic Hampers the Nationalism

Introduction

The outbreak of the COVID-19 pandemic is believed to increase people’s level of nationalism. Hartman et al. (2020) conducted a cross-sectional representative survey in the United Kingdom in early March 2020; they found that the perceived threat of the COVID-19 may activate an authoritarianism attitude, which shifts the public opinion toward stronger nationalism and anti-immigration. Similarly, Golec et al. (2020) collected a three-wave survey before and after the outbreak in Poland; they found that the level of right-wing authoritarianism, the quest for national cohesion, and sexual prejudice all increased after the outbreak. Most recently, Bol et al. (2020) conducted a panel survey before and after the lockdown in eight European countries; they found out that respondents increased their support to the incumbent and satisfaction to the democracy 2 weeks after the lockdown.

The psychological mechanism linking pandemic and nationalism is bridged by Aarøe et al. (2017). Aarøe and colleagues suggest that an individual’s physiological sensitivity to disgust can explain people’s attitudes toward immigration. In this scenario, immigrants are perceived as a source of pathogen, which alerts one’s behavioral immune system to reject or avoid the immigrants. In this study, the pathogen risk raised the perceived difference between the in-group and out-group members, and therefore changed one’s attitude toward the immigrants. In a similar study, an anti-immigration attitude is triggered by a simply priming on national identity (Wojcieszak and Garrett, 2018).

The question of how long the pandemic could boost nationalism remains. When will the increase of nationalism wane?

The choice of group identity is partly rational process. People strengthen the cognitive linkage between themselves and a salient group if the group can bring a positive image or a higher social status to their self-identity (Ethier and Deaux, 1994). People may understate the national identity when their country is facing the real threat of military invasion (Wang, 2017).

Following a similar logic, we argue that the current COVID-19 pandemic may increase the nationalism in the beginning, but the effect may reverse if the pandemic continues. Right after the outbreak, individuals tend to render their right and trust to the country, hoping a centralized power can extinguish the crisis soon. People think the source of pathogen is outside the country, which creates the in-group cohesion (Aarøe et al., 2017). But if the pandemic continues, people may lose their patience and trust for the authority. After all, the pathogen is already inside the country. In this scenario, an individual’s motherland becomes the source of pathogen risk; the rampant pandemic also implies the country’s lack of capability to deal with the disease properly. Therefore, an individual will rationally withdraw his psychological attachment to the country.

Moreover, we can expect that the reversed effect of pandemic on nationalism may be stronger among the supporters of the opposition party, for they are much more critical to the incumbent and are much likely to receive the negative information about the pandemic from their party elites. Kam and Ramos (2008) noticed that the surge and decline of the presidential approval after the 9/11 attack can be explained by both the nationalism and partisan identification. Once the political entrepreneurs started to reinstate the partisanship, people from the opposite party started to criticize President Bush on his response to the terrorism. We hypothesize that a similar pattern can be found in this case given the presidential election year in 2020.

Context, Research Design, and Data Collection

To test these hypotheses, we conducted an MTurk survey experiment on 20 April 2020, in the United States. After the first case was confirmed on 20 January 2020, the number of confirmed cases in the United States rapidly increased to 776,093 within 3 months. Meanwhile, all states had implemented some form of lockdown policies, but the rate of increase had not slowed down by 20 April. A YouGov survey shows that the disapproval of the US president increased from 45% to 50% from 11 March to 8 April. ¹ In this scenario, the rally-around-the-flag effect may also be reversed. Our survey experiment was pre-registered on Open Science Framework before the data collection and the institutional review board (IRB) proposal was reviewed by the author’s affiliated institution.

Overall, 606 participants were recruited. ² Of these participants, 62% were male; 82% were White; 70% were between the ages of 18 and 34, and 56% had at least a bachelor’s degree. Regarding partisanship, 49% identified as Democrats, 32% as Republicans, and 29% as Independent and others.

After the informed consent, all subjects were asked questions about political interest, news consumption, political knowledge, and party identification. An attention check item was put before the experiment (Berinsky et al., 2014), and 25 in 606 (4.1%) who did not follow the instruction were dropped from the analysis.

All subjects were then randomly assigned into three groups: (1) Control Group (n = 174) did not receive any treatment; (2) Treatment A “Globally” (n = 193) was asked, “On a scale of 1 to 5, where 1 is not at all concerned and 5 is the most concerned, how concerned are you about coronavirus spreading globally?” (3) Treatment B “United States” (n = 214) was asked, “On a scale of 1 to 5, where 1 is not at all concerned and 5 is the most concerned, how concerned are you about coronavirus spreading in the United States?” We follow Wojcieszak and Garrett’s (2018) method to prime our respondents in the treatment group to think about the ongoing pandemic. Moreover, we manipulate the place in the treatment so that the subjects in the Treatment B United States will think more about the COVID-19 pandemic happening nearby.

All subjects were then asked about their level of national identity by the question used in American National Election Studies (ANES): “On a scale of 1 to 5, where 1 is not at all important and 5 is the most important, how important is being an American to you?” A previous study shows that this question can capture people’s level of national identity and is predictive of group-norm related behavior such as turnout (Huddy and Khatib, 2007).

All subjects were then assigned to participate in other experiments unrelated to this study. At the end of the survey, they were asked about their demographics including age, gender, level of education, income, and race. Subjects who completed the survey were compensated $1 and debriefed. All fixed demographic variables except for the partisanship were asked after the treatments (Klar et al., 2020). The treat to the results will be discussed later.

Results

Before we exploit any advanced test, Figure 1 shows the distribution of nationalism across the experimental groups. The error bars indicate the mean value plus and minus 1.96 standard error. Generally speaking, the majority of the respondents agreed that being American is important to them, with the average 3.86 in the 1–5 nationalism scale. The distribution is close to ANES2016 which 76% chose either very or extremely important. However, when people are primed to think of the COVID-19 in the United States, their perceived nationalism is statistically lower (3.73, n = 214) than the control (3.96, n = 174). Considering the ordinal characteristic of the scale, Mann–Whitney U Test also shows a significant difference between the control group and the Treatment B United States group (w = 20,786, p = 0.039). ³

OPEN IN VIEWER

Figure 1.

Distribution of Nationalism Across Groups (n = 174, 193, and 214).

Unfortunately, the randomization checks failed. Analysis of variance (ANOVA) test shows a significant difference in the proportion of African Americans (p = 0.03). ⁴ The literature suggests that nationalism in the United States correlates with the race (Huddy and Khatib, 2007). Ordinal Probit regression is therefore used to mitigate the bias from the covariates.

Table 1 presents the results of three models explaining nationalism. The two treatments are binary coded, the respondent’s post-treatment self-reported ideology (0–100) is included in model 2, while the respondent’s age, education, income, race, and gender are further included in model 3. The standard error is shown in parenthesis.

OPEN IN VIEWER

Table 1.

Ordinal Probit Regressions Explaining the Level of Nationalism.

		Model 1	Model 2	Model 3
Treatment A Globally		−0.143 (0.113)	−0.210 (0.115)	−0.209 (0.116)
Treatment B United States		−0.223* (0.110)	−0.232* (0.112)	−0.227* (0.113)
Ideology (0−100)			0.013** (0.002)	0.013** (0.002)
Age				0.079* (0.039)
Education				−0.079(0.043)
Income				0.086** (0.018)
African American				0.293(0.173)
Male				−0.095(0.095)
Intercept	1|22|33|44|5	−1.82 (0.12)−1.34 (0.10)−0.52 (0.09)0.30 (0.09)	−1.35 (0.13)−0.81 (0.12)0.09 (0.11)0.96 (0.12)	−1.06 (0.27)−0.47 (0.27)0.47 (0.27)1.36 (0.27)
n		581	576	572
Akaike Information Criterion		1629	1544	1508

*

p < 0.05; **p < 0.01.

In Table 1, Treatment B United States is negatively significant in all models. The estimated coefficients are similar across the models, indicating the robustness of the significance. The negative effect suggests that priming MTurk respondents to think of the COVID-19 pandemic in the United States would significantly decrease their level of nationalism. In contrast, Treatment A Globally is insignificant in all models.

To summarize, both Figure 1 and Table 1 provide evidence that priming MTurk US respondents to think of the pandemic in the United States reduces their level of nationalism, while priming the respondents to think of the global pandemic has no such an effect. The effect does not come from the coronavirus itself; instead, it is likely from people’s perception that the pandemic happened in town.

Who are influenced more by the treatments? Table 2 and Figure 2 illustrate the heterogeneous effect of the treatments by the identifications. In Table 2, we re-run model 3 in Table 1 but separate pre-treatment self-identified Democrat, Republican, and Independent respondents. Across the three models, Treatment B United States is only significant among the Democrats. Meanwhile, Treatment A Globally is insignificant in all three models. Table 2 suggests that the negative effect of the US pandemic priming is taking place mainly among the Democrats.

OPEN IN VIEWER

Table 2.

Ordinal Probit Regressions Explaining Nationalism by Partisanship.

	Model 4Democrats	Model 5Republicans	Model 6Independents
Treatment A Globally	−0.268(0.164)	−0.244(0.211)	−0.186(0.266)
Treatment B United States	−0.324* (0.156)	−0.096(0.220)	−0.228(0.266)
Ideology, African American, Male, Education, Income, Age	YES	YES	YES
n	284	179	109

*

p < 0.05; **p < 0.01.

OPEN IN VIEWER

Figure 2.

Distribution of Nationalism Across Groups by Partisanships.

Furthermore, Figure 2 shows the distribution of nationalism across experiment groups by partisanship. Two-way ANOVA shows that there is a significant difference among the experimental groups by partisanship (F = 18.8, p < 0.001). The t-test between the Democrats and Republicans in the control group is statistically significant (p = 0.0002). Among the experimental groups, the only significant difference can be found among Democrats between the control group and the Treatment United States. In other words, when the pandemic brought the negative image to the US respondents, Democrats are much more likely to be influenced by the priming and lower their level of nationalism.

Discussion

This MTurk experiment shows that priming the US subjects to think of the COVID-19 pandemic in where they are living reduces their level of nationalism, especially among the Democrats. In contrast, priming the global pandemic has no similar effect. Therefore, the negative effect on nationalism cannot be explained by the disease itself, but by how much the disease has influenced their own country. Since the country failed to address the disease timely, people would psychologically lower their attachment to their own country which is now connected with the disease along with its negative consequences.

One threat to the result is that the subject’s party identification was asked before the treatment. However, the impact might be limited. People usually held a positive image to his attached party and their country, so priming on his identified party should enhance the accessibility of national identity in one’s brain (Lodge and Taber, 2005). In this scenario, we underestimate the negative impact of COVID-19 on nationalism since part of the negative effect was canceled out by the priming of partisanship. Hence, our conclusion will not be changed.

Our finding implies that the relationship between the crisis and its impact on the government may change across time. In the previous studies linking COVID-19 and politics, Bol et al. (2020) collected the last respondent in 2 weeks after the lockdown, Golec et al. (2020) conducted the last wave of survey within 1 month after the first confirmed case in Poland, while Hartman et al. (2020) started their data collection on the same day of lockdown announcement. In contrast, this experiment was conducted 3 months after the outbreak in the United States, so respondents had enough time to experience the lockdown and evaluated its effectiveness. Our result is close to the pattern found in the post-911 poll in Kam and Ramos (2008). Future work can be done on tracking the resilience of the rally-around-the-flag effect and how people may reverse their attitude.

Importantly, a survey published in September 2020 offers the external validity to our experiment. A telephone survey conducted by The Chicago Council on Global Affairs in early July (n = 2111) ⁵ shows that the gap on the belief of American Exceptionalism is widened between the Republicans and Democrats. While the majority of Republican supporters still hold the same attitude on American Exceptionalism (81% in July 2019 and 80% in July 2020), Democrat supporters’ belief dropped significantly within a year (47% in July 2019 but 35% in July 2020). Overall, the US citizens have a lower level of American Exceptionalism compared with the previous year (57% in July 2019 and 54% in July 2020). Our experimental result can provide the causal mechanism behind the change—which can be directly attributed to the pandemic in the United States.

In the end, our result also speaks to the literature on democratic accountability in general. In Achen and Bartels’ (2017) work, they render many cases to show that voters tend to irrationally blame the government for the crisis beyond the control of the government, such as shark attacks and pandemics. These crises did not boost the incumbent’s support. Meanwhile, voters may fail to reward the government for good economic performance if the economic boost happened too early from the Election Day (Achen and Bartels, 2017: Chapter 5). To reconcile these findings, this article alludes that the time factor may play an important role in how the public opinion toward the government shifts across time in the era of rapid change.