Ethnic diversity and political trust at the onset of the COVID-19 pandemic

Abstract

Current research faces challenges in explaining how contextual factors account for variations in the rally effect in political trust during the COVID-19 pandemic. While systematic explanations of country-level differences are hard to establish by means of cross-sectional comparisons, we propose to compare subnational areas within a country to learn more about the role of contextual factors. In this research note, we argue that ethnic diversity is a crucial contextual factor that helps researchers understand differences in political trust at the onset of the pandemic. Specifically, we propose that the rally effect should be restricted to ethnically more homogeneous contexts. An analysis of geocoded household panel data from the Netherlands reveals a strong rally effect in ethnically homogenous areas, while political trust in ethnically diverse contexts appears not to respond to the pandemic. This suggests an entrenched geography of political trust, which is associated with ethnic divides and is even maintained under crisis.

Introduction

Pandemics require swift and decisive collective action to contain the spread of infections. This poses fundamental challenges for liberal democracies on several fronts. The COVID-19 pandemic led to important shifts in state-society relations, to which political science research has recently dedicated much attention. In many societies, a crucial shift in public opinion at the onset of the pandemic was a rally-around-the-flag effect, leading to exceptionally high levels of trust in politics as COVID-19 case numbers rose and lockdown measures were implemented (Bol et al., 2021; Devine et al., 2021; Schraff, 2021). Yet, research also documented pronounced differences in this dynamic across different Western democracies (e.g., Kritzinger et al., 2021). In response to this, this research note adds to our understanding of how societal context shapes the degree of the rally effect on political trust during the first wave of the COVID-19 pandemic.

Existing studies document that contextual differences across countries, such as pre-existing political trust, political polarization, and the design of democratic institutions have shaped public opinion and pandemic policies (Altiparmakis et al., 2021; Engler et al., 2021; Nielsen and Lindvall, 2021). The rally effect in political trust varied substantially across countries, with, for example, a pronounced rally effect in Austria, but an absence of a rally effect in France (Kritzinger et al., 2021). This suggests that context matters for our understanding of a society’s response to the pandemic. However, we still lack more insights into the different ways of how context shapes the rally effect in political trust.

As comparative explanations of contextual effects on the country level are hard to investigate due to their structural, cross-sectional nature, we propose that looking at subnational contextual effects can make a helpful contribution to the debate (Snyder, 2001). Indeed, recent evidence suggests that subnational variation in trust and polarization are associated with a successful response to the pandemic (Charron, et al., 2023). Nevertheless, what remains unclear is how societal context can condition the rally effect, which is considered an important question to understand successful collective action during sudden crisis events (Goldstein and Wiedemann, 2022).

We propose that individuals’ belief in successful collective crisis response is deeply rooted in their social environment. While existing research points to more political factors, such as political polarization and democratic institutions (Altiparmakis et al., 2021; Engler et al., 2021), we propose that more antecedent social context conditions, characterizing the makeup of society are key to completing the picture. Ethnic diversity is one crucial contextual factor that has been shown to fundamentally shape a society’s level of trust and polarization (Dinesen et al., 2020; Schraff and Sczepanski, 2022). While ethnic diversity matters on many different levels of analysis, it has been shown to be most relevant for individuals’ trust at the subnational level, referring to people’s immediate social surroundings (Dinesen and Sønderskov, 2015). Based on this, we argue that individuals living in ethnically diverse areas have worse pre-crisis perceptions of government responsiveness and may thus have less reason to believe in a society’s ability to respond to the pandemic successfully. This, in turn, should make it less likely that individuals living in these areas increase their trust during collective crisis events. Ethnic diversity, therefore, counters the rally effect, which is then restricted to ethnically more homogenous contexts.

We draw on two waves of geocoded household panel data among a nationally representative sample of around 1,600 respondents covering the time before and during the first COVID-19 wave in the Netherlands. An analysis of individual-level changes in political trust reveals a strong increase in citizens’ trust in politics as the pandemic unfolded, but only for respondents residing in ethnically more homogenous areas. By contrast, individuals living in ethnically fragmented contexts maintain a lower level of trust as COVID-19 cases rose. These findings are robust across a range of statistical specifications and alternative levels of spatial aggregation, illustrating a deeply entrenched geography of political trust, which is shaped by ethnic divides and is even maintained under exceptional crisis events.

In this research note, we provide a novel argument on how ethnic diversity can shape a society’s response to crisis, adding to more structural arguments on the societal implications of ethnic diversity (Dinesen et al., 2020; Trounstine, 2016). Moreover, our analyses of fine-grained geocoded data provide insights into within-country contextual variation in the public response to the pandemic, going beyond existing studies on cross-country variation (e.g., Altiparmakis et al., 2021; Kritzinger et al., 2021; Nielsen and Lindvall, 2021; Plümper and Neumayer, 2022; Toshkov et al., 2022). Finally, we provide an important qualification to the rather optimistic conclusion that trust strongly rose at the onset of the pandemic (Devine et al., 2021). The rally-around-the-flag effect is fundamentally conditioned by societal context and has materialized to a lesser extent in ethnically more diverse environments.

Ethnic diversity and the rally effect at the onset of the COVID-19 pandemic

Democracies’ successful response to a crisis event requires a certain level of trust between citizens and their political institutions to overcome the collective action problems associated with effective crisis response (Goldstein and Wiedemann, 2022). Trust provides the glue keeping societies together in the face of existential threats (van der Meer and Zmerli, 2017). Crucially, social and political trust can serve as important pre-conditions for successful collective action to tackle societal threats, such as pandemics (Harring et al., 2021). At the start of the COVID-19 pandemic, several studies observed a sharp increase in citizens’ trust in politics, diagnosing a rally-around-the-flag effect in trust toward public institutions and satisfaction with democracy (Bol et al., 2021; Esaiasson et al., 2021; Schraff, 2021). Related to this, country-specific case studies also pointed toward decreasing political polarization (Merkley et al., 2020) and increasing government voting (Leininger and Schaub, 2024). The mechanism behind this dynamic is often argued to reflect public satisfaction with the policy responses enacted by national governments (Baekgaard et al., 2020; Bol et al., 2021), drawing on a more general argument that portrays citizens as individuals who rationally reward institutions with more trust for passing policies intended to provide swift crisis relief (Bechtel and Hainmueller, 2011; Colaresi, 2007). Following this rational account, the rally-around-the-flag effect can be explained by citizens’ positive evaluations of governmental performance during the first wave of the COVID-19 pandemic. However, this approach struggles to understand the immediate rise of political trust at the start of the pandemic, as many policy responses were not yet in place and uncertainty about their suitability was high. Indeed, studies of rally effects in political trust within the first few weeks of the pandemic suggest that pandemic policies, such as lockdowns, were unrelated to the rise in trust while rising infection rates correlated strongly with the rally effect (Schraff, 2021).

Therefore, another strand of research on the rally effect in political trust proposes psychological mechanisms as an alternative explanation. For instance, Esaiasson et al. (2021) maintain that the rally effect is shaped by the extent of anxious arousal the pandemic evokes. People attach themselves to political institutions in the immediate unfolding of a crisis, with the institutions serving as lifebuoys in times of high uncertainty. This argument is supported by psychological studies that demonstrate anxious arousal in response to the pandemic (Tabri et al., 2020). In this regard, the COVID-19 crisis has temporally altered political trust formation, as standard explanations of trust in politics seem to have lost relevance with the arrival of the pandemic (Schraff, 2021). Indeed, across several country contexts, fear of health risks has been shown to be strongly associated with increased political trust (Erhardt et al., 2023). This rally effect, however, only altered political trust formation for a limited period, as panel studies show a normalization of political trust as the pandemic dragged on (Johansson et al., 2021). However, the initial rally effect in the first period of the pandemic appears crucial to ensure effective collective action and popular support for the far-reaching pandemic policies enacted. It, therefore, is important to further study the contextual conditions under which such rally effects in trust emerge.

Considering these two theoretical accounts, we suspect that a combination of rational policy evaluation and emotional affect shapes the public response to the onset of the COVID-19 pandemic. In line with the psychological account, we expect that high uncertainty at the beginning of the pandemic does not provide a lot of room for cognitive evaluations of the measures enacted by governments. In the first few weeks of the pandemic, the potential outcomes of the measures were hard to predict. A combination of high uncertainty paired with an immanent existential threat generated a strong potential for an anxiety-driven rally effect (Erhardt et al., 2023; Schraff, 2021). ¹

However, we suggest that this potential for a rally effect in political trust did not materialize in all social contexts to the same extent, as people’s evaluations of the social and political context do not always make it a credible strategy to take a leap of faith in existing public institutions. In light of this, we expect that people’s residential context continues to condition individual beliefs about the trustworthiness of political institutions. Put differently, we suggest that locally induced pre-crisis perceptions of government responsiveness moderate the trust people are willing to invest in their government’s problem-solving capacity (Chatagnier, 2012). These perceptions, we argue, are the result of the immediate social environment an individual lives in, offering strong cues of how to think about and behave within a society (Bisgaard et al., 2016). In particular, the degree of ethnic diversity within an individual’s social context is an important factor that shapes people’s willingness to trust each other and entrust politics with tackling crises. Contexts with higher ethnic diversity often come with increased policy skepticism and lower generalized trust, which depresses people’s willingness to take a leap of faith in the face of collective threats.

Indeed, ethnic diversity has been shown to be a crucial contextual explanation of social trust (Dinesen et al., 2020; Dinesen and Sønderskov, 2015) and people’s willingness to cooperate (Vollan and Ostrom, 2010). A prominent explanation for this relationship is based on social identity theory. Accordingly, it is argued that group-based identities structure cooperation and trust as people prefer to cooperate with perceived similar others while simultaneously displaying heterogeneity aversion (Alesina and La Ferrara, 2002; Tyler and Blader, 2001). This suggests that group-based identities increase people’s willingness to trust and cooperate within homogenous groups but suppress inclinations to extend trust within heterogeneous groups. This lack of in-group trust within ethnically more heterogeneous contexts has similar consequences for political trust by weakening a sense of community and the connection between citizens and political elites (McLaren, 2017).

We argue that, as COVID-19 cases grow exponentially, people in ethnically diverse contexts are less willing to extend trust and rally around institutions due to their higher pre-existing policy- and polity-skepticism. Put differently, individuals in ethnically more diverse localities will face more challenges to overcome collective action problems posed by collective threats. Even under crisis, other individuals and the policies enacted to counter the crisis will be faced with greater skepticism in ethnically diverse settings compared to more homogeneous contexts. On the side of natives, this context-induced skepticism is strongly driven by the local experience of immigration and natives’ beliefs that concerns about immigration have not been addressed sufficiently at the national level (McLaren, 2012). Alternatively, highly diverse and segregated contexts can also lead to a lower sense of collectivism as groups are less willing to support the joint provision of public goods (Trounstine, 2016). Regarding the behavior of migrants in ethnically diverse areas - especially those living in disadvantaged neighborhoods - some studies are pointing to a higher level of alienation from politics, as migrants often exhibit lower rates of political participation (Maxwell, 2010). Furthermore, migrants who are confronted with discrimination in their host countries express considerably lower levels of political trust (Adman and Strömblad, 2015; Heath et al., 2013).

This context-induced skepticism is a challenge for successful collective action and is therefore likely to neutralize the rally effect in political trust within ethnically more diverse contexts. Indeed, recent experimental evidence suggests that the rally effect in political trust depends on people’s beliefs that political institutions are capable of coping with crises (Erhardt et al., 2022). Putting it this way, ethnically diverse contexts are likely to come with lower trust in democratic institutions’ ability to tackle the pandemic. Ethnically homogeneous contexts, by contrast, come with prior conditions that are more favorable for the rally effect to unfold. ²

Overall, our theoretical argument proposes that a rally effect in political trust at the onset of a collective crisis is conditional on the level of ethnic diversity in an individual’s residential surroundings. As crisis hits, people in high-trust contexts take a leap of faith and collectively rally around political institutions. By contrast, people residing in ethnically more diverse contexts will face more challenges in overcoming collective action problems. They will, therefore, be less likely to extend more political trust as a response to a collective threat. We thus hypothesize that the arrival of the COVID-19 pandemic increases political trust more strongly in ethnically homogenous areas, while a rally effect in political trust is less pronounced in ethnically heterogeneous areas.

Data and methods

We test our hypothesis with geocoded household panel survey data covering the period before and during the first COVID-19 wave. Specifically, we use two waves of the Dutch Longitudinal Internet Studies for the Social Sciences (LISS) panel with a nationally representative sample of around 1,600 respondents. ³ These two-wave panel data allow us to investigate individual-level changes in political trust as the COVID-19 pandemic arrived in the Netherlands. Data for our first wave were collected toward the end of 2019, constituting our pre-pandemic baseline. As a second wave, we have fielded a special survey over March 2020, the time the first COVID-19 wave hit the Netherlands. This special COVID-19 wave was fielded with a representative subsample of the LISS panel (N=1,600), which was then matched with the same respondents’ data from the official 2019 LISS wave. The LISS panel allows geolocating respondents within one of the about 4,000 Dutch postcode areas and 350 municipalities. This is crucial for studying ethnic diversity, as the link between diversity and trust seems to only exist on low levels of spatial aggregation (Dinesen and Sønderskov 2015). As such, data requirements for this project are demanding.

Our dependent variable measures respondents’ political trust in the national parliament on a scale from 0 to 10, with low values indicating low trust and high values indicating high trust. ⁴ While much of the rally effect literature traditionally focuses on government support, manifold studies from the pandemic documented a rally effect extending to a broad understanding of institutional trust and political satisfaction. Using panel data, Reeskens et al. (2021), for example, show that the arrival of the pandemic is systematically associated with a rise of trust in parliament, government, social trust, trust in the educational system, trust in the healthcare system, political satisfaction, and trust in experts. Reeskens et al. (2021) even collapse trust in government and parliament to a joint indicator due to the strong commonalities in the measures. We are, therefore, confident in speaking to the broader rally effect literature with our dependent variable.

Our main independent variable is the rising number of COVID-19 infections to capture the effect of the pandemic on political trust. We measure the impact of the pandemic by using data on daily COVID-19 incidents reported at the municipal level since our pandemic wave was fielded over the whole month of March 2020. Individuals surveyed in our pandemic wave at the beginning of March 2020 were interviewed in a different context (low case numbers, no lockdowns) compared to respondents who were interviewed toward the end of March 2020 (exponentially increased case numbers and lockdowns). Therefore, we can proxy the pandemic’s dynamic by measuring the increase of COVID-19 case numbers from the pre-pandemic wave to the individual-specific interview date in the pandemic wave. We merge daily COVID-19 incident numbers with our individual-level survey data by the day of the fieldwork. In this regard, our variable for COVID-19 incidents is always zero in our pre-pandemic wave, but varies by location at the municipal level and by the date of the pandemic wave. ⁵ This provides us with an indication of the intensity of the pandemic on the day of the interview during the pandemic wave. All panel data models specified below use the cumulative number of COVID-19 infections, allowing us to account for the increasing severity of the pandemic.

We argued above that the pandemic-induced rally effect on political trust could be moderated by local ethnic diversity. We operationalize ethnic diversity based on data about the local population shares of ethnic groups in terms of country of origin. Following Dinesen et al. (2020), we calculate a local ethnic fragmentation index that takes larger values as an area’s population is more dispersed across ethnic groups. ⁶ We measure ethnic fragmentation on two levels of aggregation. First, we capture ethnic fragmentation on the level of the 350 Dutch municipalities. Second, we calculate ethnic diversity for the Dutch postcode areas, which are rather small neighborhoods with a median population of 2,835 inhabitants. Figure 1 presents a map of our postcode-level ethnic fragmentation index showing considerable variation across space.

Figure 1. Ethnic fragmentation across Dutch postcode areas, 2019. Authors’ own graph. Data from Statistics Netherlands.

Our measure of the pandemic effect (i.e., cumulative COVID-19 case numbers) is interacted with our variable on local ethnic diversity to model how a pandemic-induced increase in political trust is moderated by local ethnic diversity. Omitted variable bias is an obvious challenge to any research design investigating social context variables, such as ethnic diversity. However, analyzing subnational contextual differences within a country already holds constant more confounding factors than cross-country comparisons. We do have a limited number of variables available on our fine-grained geographical levels of analysis. A crucial control variable is local wealth. The effects of the pandemic might vary across poorer and wealthier areas, while local wealth might also be a determinant of trust. We, therefore, include measures of average household income for the Dutch municipalities and postcode areas in our models. Moreover, local unemployment may play a similar confounding role. As a robustness check, we re-estimate our models with local unemployment instead of wealth. ⁷ We further control for pre-crisis support of the incumbent government, as this might both be related to political trust and ethnic diversity. We do so by including a variable on municipal vote shares in the 2017 general election for the government Rutte III. Another variable we can add is local population size, providing some control for the variation in population density and urbanization, as the mere population size of a local area, rather than its ethnic diversity, might drive the effects we observe. Lastly, studies on ethnic diversity and support for radical right parties have pointed to a so-called ‘halo effect’, where voting behavior is not driven by the ethnic composition of an area itself, but rather by the composition of places surrounding that area (Martig and Bernauer, 2018; Rydgren and Ruth, 2013). We account for this possibility by controlling for the ethnic diversity of neighboring postcodes as a robustness check. ⁸

Our statistical estimation strategy needs to consider within-individual temporal change and the hierarchical clustering of our data (individuals nested in postcodes and municipalities). Our preferred strategy to estimate the within-individual change in political trust applies individual-level random effects to estimate the change in political trust from the pre-pandemic to the pandemic survey wave. The individual random effects allow us to include individual-level control variables, such as gender, age, education, income, and migrant background. We can replicate all our models with alternative approaches to modeling temporal change, such as two-way fixed effect models as well as first differences/change models, where our dependent variable is the difference in political trust between the pre-pandemic and pandemic wave. The substantive findings remain the same. ⁹ We have also considered additional random effect specifications to account for the hierarchical data structure. However, hierarchical multi-level models clustering individuals within municipalities report a municipality-level intraclass correlation close to zero. Modeling variation across municipalities in our data is, therefore, not relevant or, at least in our case, the survey data we employ might not be rich enough to measure local variation in political trust meaningfully. Modeling variation across postcode areas is similarly irrelevant and comes with the additional limitation that most respondents are uniquely identified by a postcode area. ¹⁰ Yet, this does not mean that local context measures, such as ethnic diversity, cannot play a crucial role in people’s public opinion response to the pandemic, as we document below.

Results

Table A2 in the online Appendix presents estimates of individual-level changes in political trust with context variables measured at the postcode level, while Table A3 replicates the identical models with context variables measured at the municipal level. Our main interest lies in the interaction between the dynamic of the pandemic and ethnic diversity. In line with our main argument, the interaction terms of the number of cumulative COVID-19 cases and the ethnic fragmentation index show that the rally effect is more pronounced in ethnically homogenous areas and weaker in ethnically more diverse contexts. Put differently, with infection numbers surging, average levels of political trust within individuals increase, yet, at a much higher rate for individuals living in areas scoring low on ethnic fragmentation.

The interaction is illustrated by Figure 2 displaying the marginal effects of cumulative COVID-19 cases on political trust conditional on ethnic fragmentation at the postcode (panel A) and the municipal level (panel B). With increasing ethnic fragmentation of a respondent’s locality, the rally effect weakens. On average, an individual’s level of political trust in homogenous contexts increases by about 0.6 points on an 11-point scale at the start of the pandemic. This effect vanishes for respondents living in environments that are ethnically very diverse. This finding is robust at both the postcode and the municipal level. Furthermore, all interaction effects remain statistically significant when we add critical context-level covariates, such as an interaction between infection rates and local wealth as well as incumbent government support. Lastly, our findings also hold when we add province fixed effects and control for an area’s population size.

Figure 2. Interaction effect of COVID-19 incidences and ethnic diversity on political trust.

Conclusion

At the start of the first COVID-19 wave, a rally-around-the-flag effect boosted citizens’ trust in politics (Bol et al., 2021). This rally effect, however, varied substantially across contexts (e.g., Kritzinger et al., 2021). In this research note, we argue that one explanation for the contextual variation in rally effects can be found in the ethnic diversity of residential contexts. Drawing on geocoded household panel data from the Netherlands, our analysis reveals that respondents from ethnically more homogenous areas strongly rally around political institutions. Respondents from ethnically diverse contexts, by contrast, retain a more critical stance. We proposed that this finding is explained by more skeptical pre-crisis evaluations of government responsiveness and trustworthiness in ethnically more diverse areas. Individuals in ethnically diverse contexts face more challenges to overcome collective action problems, which makes local ethnic diversity an important factor for the legitimacy of public action at the start of the COVID-19 pandemic.

Considering the geographically polarized rally effect together with the negative mid-term consequences of lockdown policies on civic attitudes (Marbach et al., 2020) highlights that rising trust can be a double-edged sword in the fight against the pandemic. While the rally effect provides much-needed legitimacy for decisive government action, it leaves behind people from low-trust contexts. As such, ethnic composition appears to be a crucial contextual factor that has to be considered by comparative studies on societal responses to sudden collective crisis events.

The findings of our study have clear limitations by being purely associational. They are, therefore, subject to omitted variable bias stemming from temporally varying variables and contextual moderators besides ethnic diversity. Against this backdrop, there is much room for experimental research designs to contribute further to debates about ethnic diversity, collaboration, and trust. Future research should also focus more on how government policies, especially during a crisis, affect trust in diverse contexts and among ethnic minorities, to further improve our understanding of the conditions under which trust can be increased in ethnically diverse settings. Moreover, additional research will be required to arrive at a more comprehensive understanding of context-dependent rally effects, going beyond this study’s focus on ethnic diversity.