Pandemics meet democracy:

Experimental evidence from the COVID-19 crisis in Spain∗

Francesc Amat Albert Falcó-Gimeno

University of Barcelona - IPERG University of Barcelona - IPERG

Andreu Arenas Jordi Muñoz

University of Barcelona & IEB University of Barcelona - IPERG

April 5, 2020

Abstract

The COVID-19 outbreak poses an unprecedented challenge for contemporary democ-

racies. Despite the global scale of the problem, the response has been mainly national,
and global coordination has been so far extremely weak. All over the world governments
are making use of exceptional powers to enforce lockdowns, often sacrificing civil liber-
ties and profoundly altering the pre-existing power balance, which nurtures fears of an
authoritarian turn. Relief packages to mitigate the economic consequences of the lock-
downs are being discussed, and there is little doubt that the forthcoming recession will
have important distributive consequences. In this paper we study citizens’ responses to
these democratic dilemmas. We present results from a set of survey experiments run in
Spain from March 20 to March 28, together with longitudinal evidence from a panel survey
fielded right before and after the virus outbreak. Our findings reveal a strong preference
for a national as opposed to a European/international response. The national bias is much
stronger for the COVID-19 crisis than for other global problems, such as climate change or
international terrorism. We also find widespread demand for strong leadership, willingness
to give up individual freedom, and a sharp increase in support for technocratic governance.
As such, we document the initial switch in mass public preferences towards technocratic
and authoritarian government caused by the pandemic. We discuss to what extent this
crisis may contribute to a shift towards a new, self-enforcing political equilibrium.

∗
Jordi Muñoz acknowledges the financial support from ICREA under the ICREA Academia programme.
Francesc Amat acknowledges the funding from ‘La Caixa’ foundation under the Postdoctoral Junior Leader
Fellowship programme.
1 Introduction

Could the COVID-19 outbreak weaken or endanger democracy? The 2020 pandemic poses, in
many respects, an unprecedented challenge for modern democracies. It has produced a massive,
global public health problem. It represents a direct life threat to millions of citizens worldwide,
and is also potentially catastrophic for the economy. Its management poses a wide range of
democratic dilemmas, most notably between globalism and nationalism, between public health
and civil liberties, and between political and technocratic governance. It remains an open
question how, and to what extent, democratic institutions will be able to deal with the virus,
and how citizens’ preferences will evolve alongside the crisis and the governments’ response,
possibly shifting towards a new social and political equilibrium.
The challenge posed by the COVID-19 outbreak can be characterized as a twofold collective
action problem. On the one hand, at the international scale, the policy responses and decisions
of one country affect the evolution of the pandemics elsewhere, and coordination is difficult.
Without internalizing the spillovers, given the economic costs of the lockdowns, every country
will tend to lock down later and less intensely than it would be globally optimal, increasing
the spread of the virus. The fierce competition in the international market of medical supplies,
or the European debate on the fiscal response to the crisis are good examples of this global
coordination problem.1 On the other hand, a decentralized management of the crisis may come
along with responses more closely aligned to each country or region’s needs, and can lead to
experimentation and learning. So far, the response has been mostly led by nation-states. To
what extent is this uncoordinated response aligned with citizens’ preferences? Would they
prefer more international coordination to deal with COVID-19?
But the COVID-19 crisis also poses a massive collective action problem within countries,
because of the highly contagious nature of the virus. This problem can be solved either by
citizens’ voluntary compliance and cooperation in following social distancing, or by resorting
to exceptional government powers, limiting basic civil liberties to impose quarantines and lock-
downs, as it has been the case in many countries. The response has often relied on increasingly
invasive surveillance and coercion, which many fear could outlast the COVID-19 crisis. China
1
For instance, amid a mask shortage, France seized stock from a Swedish producer that was headed for Spain
and Italy. Source: France 24, April 3, 2020. Similar episodes have been reported between Spain and Turkey, or
the US and Germany.

2
has been presented by some as a success story on how a state can effectively control the spread
of the disease using mass surveillance. The case of Hungary, where the Parliament passed a
bill granting the government of Viktor Orban the power to rule by decree with no constrains
and no time limit is a clear example of the democratic perils posed by the COVID-19 crisis.
In many other countries, we are witnessing the use of drones, smartphone-based individual
tracking apps, or the deployment of military forces to control the population’s compliance. To
what extent are citizens willing to trade-off democratic principles and individual freedoms in
exchange for protection against the pandemic? Is this trade-off more or less intense in this case
as compared to other democratic dilemmas, such as security and liberty in the fight against
terror?
On top of this health-freedom democratic dilemma, collective decision-making often involves
a tension between politics and technocracy. Most scholars of democratic politics have portrayed
the years between the great recession and the COVID outbreak as the age of populism. Distrust
of experts and of established politicians has, in many countries, led to the emergence of populist
leaders and parties that claim to represent the ordinary people as opposed to the allegedly
detached elite. However, the nature of the COVID-19 crisis necessarily brings technical expertise
at the forefront of the crisis’ management. The question, therefore, is whether this crisis will
favor citizens’ demand for technocratic rule.
In this paper we address these questions through a set of survey experiments fielded in Spain
in March 2020, right after the outbreak; and by using panel data, tracking within-individual
changes in a number of political preferences between January to March 2020. We show that
the national focus of the crisis management is congruent with citizens’ preferences. As opposed
to other global challenges, such as climate change or international terrorism, the preference for
a national management is particularly strong in the case of the COVID-19 crisis.
Our results also reveal a sizable effect of the COVID-19 crisis on democratic attitudes. We
first show, using panel data with individual fixed effects, a sharp increase in preferences for
technocracy between January and March 2020. Right after the outbreak, citizens expressed
significantly more support for expert rule and a technical approach to the management of
public affairs. This shift towards technocracy was especially intense among those with direct
relatives or close friends infected by the virus. We also find a change in the preferred qualities of

3
politicians: the shock makes citizens value capacity and training more intensely, at the expense
of other qualities such as being honest or approachable.
Likewise, the results show both a widespread willingness to sacrifice basic civil liberties in
order to contain the pandemic, and higher support for strong leadership to face this threat.
These attitudes are much more widespread in the COVID-19 case than in the case of other
external threats, such as terrorism or climate change. Moreover, hierarchical frames appear to
be more effective in gathering citizens’ support for restrictive measures than more horizontal,
cooperative frames.
Overall, our results point to a wide agreement with the national-level response, and extensive
willingness to trade-off ideological preferences and individual freedoms for protection against
the virus. An extreme event such as the COVID-19 pandemic appears to be, at least in the
short run, a fertile ground for the erosion of democratic principles, as public opinion appears
to be relatively willing to accept a technocratic or authoritarian turn. Hence, the COVID
may provide a window of opportunity for incumbents to centralize and accumulate power
and increase surveillance and control. We believe that the effort to document this window
of opportunity - identifying the initial shift in mass public preferences towards technocratic
and authoritarian governance caused by the pandemic– is important to understand better the
subsequent moves by political entrepreneurs and the possible social and political consequences
of this crisis.

2 Theory: Pandemics and democratic preferences

There is little doubt that most democracies in the world are currently undergoing a massive
exogenous health and economic shock. The COVID-19 outbreak will have severe consequences
on a wide range of relevant societal outcomes, and it will represent a sizable negative welfare
shock for a large share of the world’s population.
Most research on the political effects of natural disasters (Healy and Malhotra, 2009) and
economic shocks (Margalit, 2019) has focused on their impact on incumbent support. Evidence
shows that often incumbents get punished by events that produce negative welfare shocks on
the population, even if they are outside the incumbent’s control. The literature disagrees on
the interpretation of these results, and while some scholars take them as examples of irrational,

4
blind retrospection (Achen and Bartels, 2017), others make the case for rational punishment
of incumbents following these exogenous shocks, as these situations provide opportunities for
voters to learn about previously hidden qualities of politicians (Ashworth et al., 2018).
But there are theoretical reasons to believe that some of these extreme, exogenous shocks
may have deeper consequences that go well beyond incumbent support and extend to the regime
level. We know that negative shocks of this scale can be potentially destabilizing, especially for
weak political regimes. Theories of democratization have long posed that exogenous negative
economic shocks have the potential to open windows of opportunity for democratization. Per-
haps one of the most influential versions of this classic argument is Acemoglu and Robinson’s
(2001) theory, in which negative shocks reduce opportunity costs of revolt, and hence induce
autocrats to make economic concessions to prevent revolution. There is empirical evidence
that droughts or floods, for example, can operate as potential democratizing devices (Aidt and
Leon, 2016; Brückner and Ciccone, 2011), and they may be more important than slow-moving
economic transformations, although their effect is probably highly conditional to preexisting
characteristics, such as the level of inequality (Kotschy and Sunde, 2019).
However, not only autocracies may suffer from these extreme events. When a democracy
faces a sudden, exogenous negative shock on citizens’ welfare, this can also have potentially
destabilizing consequences. The mechanisms through which these events may affect democratic
preferences and attitudes are multiple. These shocks provide citizens with new information
about the world, and the ability of a given political system to deliver societal welfare (Ashworth
et al., 2018). As we know from historical cases, weather-induced disasters or earthquakes, for
example, can erode democratic legitimacy and increase support for anti-democratic forces,
especially in new democracies, that need to deliver public goods in order to sustain citizens’
support, in absence of a sufficient reservoir of legitimacy (Carlin et al., 2014; Franck, 2016).
The duration of the effects of one-shot events on political and democratic preferences has
been found to be heterogeneous across case-studies (Rehman and Vanin, 2017), an in some
cases only short-lived effects were found (Katz and Levin, 2016). The magnitude of the shock
is likely to matter here: a shock that is large enough, such as the COVID, may generate a
shift towards a new, self-reinforcing political equilibrium. For instance, a large initial shift
towards less civil liberties could reduce social trust and pave the way for a more authoritarian

5
governance (Xue and Koyama, 2018).
Recent research on the impact of negative economic shocks finds significant detrimental
effects on political trust (Foster and Frieden, 2017; Ananyev and Guriev, 2019), increased polit-
ical polarization and voting for non-mainstream parties (Autor et al., Forthcoming; Colantone
and Stanig, 2018; Algan et al., 2017), and increased authoritarian preferences among voters
(Ballard-Rosa et al., 2018).
Pandemics represent a very large shock to public health and the economy, which in turn gives
citizens the opportunity to learn about their governments, based on their political and policy
responses. Research on the effects of previous pandemics has found important and durable
effects on attitudes. For instance, the Black Death of the 14th century has been shown to
have negatively affected cooperation (Richardson and McBride, 2009) and inter-group relations
(Voigtländer and Voth, 2012).2 Aassve et al. (2020) show that the forced social distance and the
social disruption caused by the so-called 1918 Spanish flu had long-term negative consequences
for social trust. Hence, all these effects are important because they can shift societies towards
different, self-sustained social and political dynamics.
Based on these theoretical and empirical antecedents, how can we expect the COVID-19
outbreak to affect democratic preferences? We argue that the characteristics of the pandemic
make it a potential case for eroding preferences for democratic governance, at least in the short
run.
First, the importance of coordination in dealing with the spread of the virus, both across and
within countries, and of technical expertise, especially in public health, may foster preferences
for strong leadership and technocratic forms of government amidst the COVID-19 crisis.
Second, the nature of this crisis comes along a liberty-public health trade-off. Governments
are being rather invasive of citizens’ private life in order to enforce restrictive shelter-in-place
orders at a large scale over an extended period of time, and achieve effective contact tracing,
testing and quarantine. China, where the use of mass surveillance mechanisms has been claimed
as crucial for control of the pandemic, is probably a case in point, despite the fact that the
nature of the Chinese regime also facilitated in a first stage the expansion of the virus by
disincentivizing (and actually punishing) early alerts (Ang, 2020). In a democratic setting, the
2
It also led long-run economic growth (Jedwab et al., 2016; Alfani and Percoco, 2019), despite evidence of a
positive effect on real wages (see Alfani and Murphy (2017) for a review.

6
much praised, and effective strategy of South Korea was also based on heavy use of surveillance
technology, such as CCTV and tracking of bank card and mobile phone usage to identify at-risk
individuals.
Third, beyond the direct COVID victims, the virus threatens the health of the whole popu-
lation, and its expected economic effects have fostered a high level of economic anxiety (Fetzer
et al., 2020), which may have deep consequences on mass political attitudes.
The COVID-19 crisis poses, therefore, a number of democratic dilemmas, most notably
between globalism and nationalism, between public health and civil liberties, and between
political and technocratic governance. Because this crisis relates to the protection of a highly
valued good (health and life), we may expect citizens to be willing to pay a high cost, increasing
the demand for competent and strong leaders who are capable of effectively managing the crisis,
even if this comes at the expense of basic democratic procedures and freedoms and ideological
preference representation.
In this regard, the COVID crisis may provide a window of opportunity for incumbents
to centralize and accumulate power and increase surveillance and control, as citizens may
be willing to trade-off civil liberties and ideological preference representation in exchange for
protection and efficacy in the response. Democratic societies, as a result, might shift towards
a new equilibrium. Our aim in what follows is to document this window of opportunity by
identifying and quantifying the initial shift in mass public preferences towards national rule
and technocratic and authoritarian governance as a result of the pandemic.

3 Data and empirical strategy

We study citizens’ reactions to the democratic dilemmas posed by the COVID-19 crisis using a
survey fielded in Spain in March 20-28, 2020. During the days of the fieldwork, Spain witnessed
one of the fastest escalations in number of COVID-19 positive tests and deaths in the whole
world. At the same time, Spanish authorities issued a restrictive stay-at-home order, that
entered into force in March 14. The Spanish government made use for the second time since
the transition to democracy of the ‘state of alarm’, a constitutional device that allows the
government to impose strict limitations on civil liberties and centralize power in the Spanish

7
executive 3 . The military started to patrol the streets, and the central government seized power
from the autonomous regional and local authorities in order to impose a unified strategy against
the virus. Spain is a case of a heavily affected country with an established (albeit relatively
young) democracy.
Our data collection was based on an on-line survey that we fielded through the survey
company Netquest. In order to limit self-selection, participants cannot self-register into the
panel, and membership is solely based on invitation. Respondents receive a fixed compensation
to answer each survey.4 Our full sample is composed of 1,600 individuals, that were recruited
using age, gender, region and education quotas in order to mirror the actual distribution of these
variables in the voting-age Spanish population. A reduced sample of 818 of our respondents
had been interviewed two months earlier, before the outbreak hit Europe (January 27-30), and
the rest are new respondents.
Our empirical strategy is twofold. On the one hand, we use observational evidence combining
data collected right before and right after the outbreak for the same individuals on the reduced
sample. Using individual fixed-effect models we can estimate the effect of the outbreak on a
number of attitudes while controlling for inter-individual differences. On the other hand, we
embedded a number of randomized survey experiments within our survey to the full sample.
These randomized experiments were designed to test citizens’ reaction to the various democratic
dilemmas outlined above. We provide details below.

4 Empirics

4.1 Global crisis, national response

We first address the issue of the global collective action problem. The COVID-19 pandemic
is a global problem, and there are obvious international implications of the policy decisions
taken by national governments. Although responses at a national or regional level could be
beneficial by tailoring responses to citizens’ needs and preferences, citizens may over-estimate
these benefits if they fail to internalize the large externalities involved in this crisis.
3
The previous declaration of the state of alarm was in 2010 in response to a strike of air traffic controllers,
that allowed the government to substitute them by military personnel.
4
More information in the Netquest Panel Book

8
 So far, the response to the crisis has been mostly national, and even within the EU, coordi-
nation among countries appears to be extremely weak, and international cooperation has been
very limited. We explore to what extent this is congruent with citizens’ preferences. In order
to do so, we run two survey experiments. In the first one, we study citizens’ willingness to
contribute to the provision of COVID-19 control measures at the regional, national, European
and global levels. More specifically, we ask them to report what part of a (hypothetical) e1,000
bonus they would be willing to pay in order to help control the outbreak in randomly varying
areas: region, nation-state, Europe or Asian countries.
Table 1 presents the results of two OLS regression models in which the declared contribution
is the dependent variable, and the various treatments are the predictors. As it can be seen,
there are no significant differences between the national and regional scope. However, when
asked about other European countries or Asia, respondents were significantly less willing to
contribute. Around e100 less for Europe (20% of the mean contribution) and over e175 less
for Asia. Results in Figure 1 represent this large and significant difference in the willingness to
contribute to the COVID-19 control at the regional/national level, on the one hand, and the
European/global level on the other hand.

Table 1: Share of e1000 bonus for COVID control

(1) (2)
Asia -176.2∗∗∗ -183.0∗∗∗
(25.55) (25.58)

Europe -95.59∗∗∗ -98.22∗∗∗

(25.72) (25.53)

Constant (Spain) 503.6∗∗∗

(18.34)

Region -23.11 -22.56

(26.07) (26.03)
Controls X
Mean Dep. Var 429.8 429.8
N 1606 1606
Robust standard errors in parentheses.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
Controls: gender, age, region FE, education FE.

9
 Figure 1: Geographical scope

Average share of 1000€ bonus willing to contribute to COVID control in...

600

500

400

300

200

Region
100 Spain
Europe
Asia
0

Table A1 in the Appendix reports covariate balance tests across treatment arms. As one
would expect because of randomization, there are almost no statistical differences across groups.
This also shows up in column 2 of Table 1: the estimates are very stable after including a number
of covariates.
In the second experiment we directly address the question of the preferred level of govern-
ment to manage the crisis. In order to have a reference point to evaluate citizens’ response to
the COVID crisis, we devised an experimental design to compare the COVID crisis to other
global threats that share the same features of interdependence and potential externalities across
countries. We presented respondents with a set of questions referred to one of the following
three randomly assigned threats: COVID-19, climate change or international terrorism. Among
other outcome questions, we inquired about the preferred level of response, by asking them to
express agreement on whether power should be transferred to the national government or the
EU in order to combat the threat, in a 0-10 scale.
We present the results of this experiment in Table 2 and Figure 2. We asked respondents
whether they agreed in transferring power to the EU, first, and the national government, second,
in order to combat the threat. The table shows how the COVID is, by far, the threat for which
citizens seems trust less the EU and more the national government. The difference is large
and statistically significant in both cases. Climate change and terrorism are cases for which

10
respondents seem to prefer a EU-wide management over a national response, while the COVID
seems to behave differently. Table A2 in the Appendix reports covariate balance tests across
treatment arms. As one would expect because of randomization, there are almost no statistical
differences across groups, with very few exceptions that may arise because due to sampling and
chance (in this case, the Prime Minister assessment, which happens to be slightly different for
the COVID group). Nevertheless, the estimates in column 2 of Table 2 remain very stable after
including a number of controls.

Table 2: Willingness to transfer powers to:

EU National Govt.
(1) (2) (3) (4)
COVID -0.998∗∗∗ -0.756∗∗∗ 0.894∗∗∗ 1.083∗∗∗
(0.174) (0.181) (0.185) (0.180)

Terrorism 0.238 0.442∗∗∗ 0.154 0.400∗∗

(0.158) (0.161) (0.173) (0.174)

Constant (Climate Change) 5.963∗∗∗ 5.035∗∗∗

(0.113) (0.121)
Controls X X
Mean Dep. Var 5.709 5.709 5.384 5.384
N 1606 1606 1606 1606
Robust standard errors in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
Controls: gender, age, region FE, education FE, PM assessment, left voter,
other parties’ voter.

Figure 2: EU/National response to global threats

Willingness to transfer powers to the EU/ national government

EU Spanish government
7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0

Covid Climate Terror

11
 While this pattern may be endogenous to the dominant policy response to the COVID crisis,
that has been essentially national, it reveals nonetheless congruence with public preferences.
The Spanish public does not seem to demand, on average, an europeanization of the response.

4.2 COVID and democracy

The responses of national governments to the COVID crisis in most countries have required
unprecedented policy interventions. The highly contagious nature of the virus, together with
the large number of asymptomatic infected people and the absence of a vaccine, implies that
social distancing is crucial in dealing with the spread of the virus and avoiding a collapse of
the health system. Social distancing can be achieved with citizens’ cooperation and voluntary
compliance. However, this needs large levels of altruism –internalizing the damage of social
distancing for the benefit of society at large. Such a response also requires high levels of trust
on other citizens’ behavior –that they will comply with instructions issued by public authorities
as well as by health experts– in a period when citizens’ trust on experts and governments has
been declining.
Effective social distancing can also be achieved through coercion, for instance imposing
fines or jail threats to the non-compliers. Effective monitoring of non-compliance, however, re-
quires special powers, often involving privacy violations as well as other punitive legal measures.
Hence, the COVID crisis poses evident collective action and democratic dilemmas which, in
many countries, have come down to stay-at-home orders and large scale lockdowns, with gov-
ernments resorting to exceptional powers and curtailing basic civil liberties to enforce them. In
this section we analyze citizens’ response to these dilemmas.

4.2.1 Political Trust

We start by analyzing to what extent the COVID shock had negative consequences on the
levels of individuals’ political trust and democratic preferences. We investigate this relationship
by correlating personal exposure to the virus with standard measures of political trust and
democratic preferences. Specifically, we analyze first how personal exposure to the virus affects
the levels of trust in the Spanish Government and trust in the European Union. We also
investigate how exposure to the virus affects general support towards democracy, and to what

12
extent individuals support an authoritarian response to the COVID crisis.
We presented respondents the following questions:

1. Trust in Spanish Government: Personally, ¿How much do you trust the following insti-
tutions? Use a scale that goes from 0 to 10, where 0 means I don’t trust at all and 10
means I trust completely. The Spanish Government.

2. Trust in the EU: Personally, ¿How much do you trust the following institutions? Use
a scale that goes from 0 to 10, where 0 means I don’t trust at all and 10 means I trust
completely. The European Union.

3. Democratic Preferences: Regarding having a democratic political system. What is your

opinion about this way of governance? A very bad way of governance (1), A rather bad
way of governance (2), a rather good way of governance (3), a very good way of governance
(4).

4. Authoritarian Response: I believe that in order to enforce the indications on home confine-
ment, special powers should be granted to the authorities, even if this requires sacrificing
democratic principles and individual rights (1). I think we should try to enforce the indi-
cations on home confinement but without sacrificing democratic principles or individual
rights (0).

In order to proxy for individuals’ direct exposure to the COVID shock during the outbreak
we employ a dummy variable that takes value 1 if the survey respondent has some direct relative
or friend that has been tested positive. As such, this variable measures direct personal exposure:
it takes value 1 when someone in the personal acquaintance network of the respondent and with
one degree of distance to her is a confirmed COVID case. The goal is to analyze the shock by
focusing on early exposed individuals, since they are the most affected.
In our survey, which was implemented relatively early on during the outbreak, 11.96% of
the respondents report direct exposure in their personal network to COVID-19. We believe
this is roughly plausible estimate. According to the available estimates5 , on March 25th there
were 40,000 confirmed cases in Spain. This represents approximately 0.1% of the Spanish adult
5
For instance, see Source: El Pais, April 4, 2020.

13
population. Therefore, 12% of COVID exposure would require roughly 120 contacts on average
per individual if we consider only clinically confirmed positives.
In order to study the effect of exposure to COVID on political trust we estimate a set of
OLS models that include individual controls (age, gender and education) as well as a set of
regional fixed effects. The inclusion of individual controls and regional fixed effects is important
for several reasons. First, because individuals’ direct exposure to the virus might be driven by
individual characteristics such as age and education. Second, because the intensity of the virus
outbreak in Spain had significant variation across the regions. Third, there might be a com-
pound treatment effect between the outbreak and the slightly varying political and institutional
responses to combat it across regions.
The results in Table 3 provide correlational, preliminary evidence that the COVID shock has
significant detrimental effects on individuals’ political trust and democratic preferences. Not
only that, it seems that the demand for a more authoritarian response increases significantly
when someone has direct personal exposure to the virus. Admittedly, this evidence is just
suggestive (not causally identified) since it is based on conditional correlations based on self-
reported exposure. These correlations might be confounded by unobserved variables.

Table 3: Political Trust, Democratic Preferences and Exposure to Coronavirus March 2020.

(1) (2) (3) (4)

Trust Trust Democratic Authoritarian
Spanish Gov. EU Preferences Response
COVID exposure -0.425∗∗ -0.459∗∗ -0.107∗ 0.0709∗∗
(0.159) (0.184) (0.0526) (0.0299)
Controls X X X X
Mean Dep. Var 3.557 3.612 3.386 0.582
N 1604 1604 1604 1604
Standard errors clustered at the regional level in parentheses.
Controls: gender, age, region FE, education FE.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Figure 3 plots the coefficients estimated for direct exposure to the virus and the other
individual controls included in Table 3. The coefficients for the regional FEs are not plotted in
this figure but included in the estimations. It is remarkable to see how the estimated negative
effects of exposure to Coronavirus on political trust and democratic preferences are sizable and
comparable in magnitude to the standard effects of education on political trust and democratic
preferences.

14
 Figure 3: Political Trust and Democratic Preferences

Exposure to Coronavirus and Political Trust

Trust in Spanish Government Trust in EU
COVID exposure COVID exposure
Primary education Primary education
Secondary education Secondary education
High School High School
Short college degre Short college degre
Long college degree Long college degree
Masters degree Masters degree
PhD PhD
Age Age
Gender Gender
-1 0 1 2 3 4 -1 0 1 2

Democratic Preferences Authoritarian Response

COVID exposure COVID exposure
Primary education Primary education
Secondary education Secondary education
High School High School
Short college degre Short college degre
Long college degree Long college degree
Masters degree Masters degree
PhD PhD
Age Age
Gender Gender
-.2 0 .2 .4 .6 .8 -.2 -.1 0 .1 .2

All models control for region FE. Standard errors clustered at the regional level. Education base category: no schooling.

The negative correlation between direct personal exposure to COVID and political trust
begs the question of whether it is a causal relationship, and to what extent individuals’ also
switch their preferences towards politicians’ behavior and qualities. Below, we more directly
analyze citizens’ evaluations of the trade-off between liberty and health, and the trade-off
between political preferences and technical expertise by studying preferences for technocratic
government, for strong leadership, and for curtailing civil liberties to curb the pandemic. In
order to provide credible causal estimates, we address these questions both experimentally
and by using panel data on democratic preferences, which allow us to analyze intra-individual
changes from January to March 2020. we use individual fixed effects models, that control for
individual unobserved heterogeneity.

4.2.2 Technocracy

Now we turn to the question of preferences for technocracy. It is easy to see why a crisis such as
the COVID-19 can be a fertile ground for increased demand for technocratic government, un-
derstood as the exercise of political power based on neutral technical expertise and competent

15
management of public affairs, rather than ideological representation of preferences (Bertsou
and Caramani, 2019; Dommett and Pearce, 2019). A pandemic is an extremely complex pub-
lic health issue that requires, to be effectively managed, a great deal of technical expertise.
For citizens to comply with social distancing, trust on health experts, which give technical
instructions that may not be understood by a large fraction of the population, is key.
Has the crisis caused a shift towards technocratic attitudes among the public? Are citizens
more predisposed to trade-off ideological representation by technical expertise in this context?
In order to answer this question, we first exploit an individual panel with two waves, in which
we used three different measures of technocratic attitudes. We presented respondents with the
following statements:

1. Some people prefer to vote for a party that shares their ideas, even if they have not
managed public affairs well, while others prefer to vote for a party that has managed
public affairs well, although they do not share their ideas. What do you prefer?

2. Some people believe that politicians should put aside their political agenda and tackle public
problems from a technical point of view. Do you agree?

3. It is better to have experts, and not politicians, deciding which policies are best for the
country. To what extent do you agree?

The first one refers to the willingness to trade off ideological representation for competent
management, while the other two tap more directly into preferences for technocratic govern-
ment. In all three items, we used seven-point response scales.
The first wave of our panel survey was fielded at the end of January 2020, and the second
one was fielded at the end of March 2020. We look at variation in technocratic attitudes within
individuals over time, using individual fixed-effects models. The identifying assumption is that
there are no time varying confounders: no relevant changes between the end of January and
the end of March other than the COVID outbreak. The short time span, the absence of other
relevant events and the extraordinary salience of the COVID crisis contribute to credibility of
the assumption. Table 4 displays the results of these fixed-effects models.
Results in Table 4 are consistent with the expected technocratic turn. The effect of the
outbreak, if our identifying assumption holds, is positive and strong across the three items.

16
 Table 4: Change in technocratic attitudes January-March 2020.

(1) (2) (3)

Good management Technical approach Experts not politicians
March 2020 0.428∗∗∗ 0.220∗∗∗ 0.305∗∗∗
(0.0793) (0.0707) (0.0504)
Individual FE X X X
Mean Dep. Var 4.754 5.055 4.926
N 1108 1162 1182
Baseline: January 2020.
Standard errors clustered at the individual and regional level in parentheses.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
The size of the effect ranges between 15% and one quarter of a standard deviation of the
dependent variable. In all three cases, the average expected response moves from being slightly
over the neutral point 4, to be around or above 5. This is arguably a sizable effect, taking
into account that we are looking at short-term intra-individual change in attitudes. Moreover,
the short time span between the two waves may contribute to a panel conditioning effect that
biases the estimates downwards.
In order to further investigate the mechanism and lend additional credibility to our inter-
pretation of the estimates, we check if the effect of the outbreak is positively moderated by
individuals’ actual exposure to COVID-19. The expectation is that those directly exposed to
the virus should be specially prone to modify their attitudes. In Table 5 we add an interaction
term between the outbreak (March) and the dummy that measures individuals’ direct personal
exposure to the virus -which is the same one that we employed before when exploring the effects
of COVID exposure on political trust. It is worth noting that the effect is twice as large for the
COVID-exposed individuals than for the less exposed.
The technocratic turn may also be conditional to the previous levels of information. Those
citizens with higher levels of political awareness may a priori be expected to be less likely
to move away from democratic preferences following the outbreak. We also investigate the
moderating role of political information. In order to look at this interaction, we use a composite
index of political knowledge from the first wave of the panel based on three factual knowledge
items.6
In Table 6 we present the same models as above, but including an interaction between
6
The first item asked respondents to recognize the picture of the Spanish transport minister. The second
one asked who was the second prime minister of the Spanish democracy, and the third one asked how many
seats are in the Spanish lower chamber.

17
Table 5: Change in technocratic attitudes and COVID exposure January-March 2020.

Good management
(1) (2) (3)
March 2020 0.394∗∗∗
(0.0758)

March 2020 × COVID exposure 0.347∗ 0.338∗ 0.332

(0.180) (0.179) (0.219)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 4.754 4.754 4.754
N 1108 1108 1108
Technical approach
(1) (2) (3)
March 2020 0.185∗∗
(0.0763)

March 2020 × COVID exposure 0.331∗∗ 0.287∗ 0.343∗∗∗

(0.153) (0.140) (0.111)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 5.055 5.055 5.055
N 1162 1162 1162
Experts not politicians
(1) (2) (3)
March 2020 0.301∗∗∗
(0.0484)

March 2020 × COVID exposure 0.0382 0.0202 0.00650

(0.146) (0.146) (0.156)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 4.926 4.926 4.926
N 1182 1182 1182
Baseline: January 2020.
Standard errors clustered at the individual and regional level in parentheses.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

18
Table 6: Change in technocratic attitudes and previous political information January-March
2020.

Good management
(1) (2) (3)
March 2020 0.484∗∗∗
(0.110)

March × knowledge -0.209 -0.198 -0.256∗

(0.143) (0.133) (0.135)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 4.754 4.754 4.754
N 1108 1108 1108
Technical approach
(1) (2) (3)
March 2020 0.276∗∗∗
(0.0755)

March × knowledge -0.211 -0.130 -0.110

(0.179) (0.171) (0.163)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 5.055 5.055 5.055
N 1162 1162 1162
Experts not politicians
(1) (2) (3)
March 2020 0.366∗∗∗
(0.0583)

March × knowledge -0.236∗∗∗ -0.167∗∗ -0.176∗∗

(0.0627) (0.0774) (0.0798)
Individual FE X X X
March*Individual Controls X X
March*Regions X
Mean Dep. Var 4.926 4.926 4.926
N 1182 1182 1182
Baseline: January 2020.
Standard errors clustered at the individual and regional level in parentheses.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

19
March and political knowledge. In all cases the interaction term has a negative sign, indicating
that those with high levels of knowledge have lower propensity to switch towards technocratic
preferences following the outbreak. The interaction only reaches conventional levels of statistical
significance in the case of the "experts not politicians" outcome, which is the item more directly
measuring the preference for substituting politicians for experts.
We conclude this section by analyzing whether the qualities that citizens value from politi-
cians have changed as a consequence of the pandemic. In the two waves, our respondents
were asked to rank the following four characteristics of politicians from more to less important:
share my ideas, capacity to manage, training, honesty, and approachability to people. Table 7
shows the within-individual variation over time in the first ranked quality. We regress preferred
politicians’ qualities on the outbreak dummy and individual fixed-effects. Consistent with the
idea of the technocratic turn, the demand for capable and trained politicians grows substan-
tially –between 5 and 7 percentage points–, while valence traits like being approachable or, in
particular, honest seem to matter less. The importance of ideology, on the other hand, remains
unaltered.

Table 7: Change in preferred qualities for politicians, January-March 2020.

(1) (2) (3) (4) (5)

Ideology Capacity Training Honesty Approachability
March 2020 -0.00605 0.0651∗∗∗ 0.0499∗ -0.0893∗∗∗ -0.0197∗∗
(0.00937) (0.0165) (0.0238) (0.0281) (0.00823)
Individual FE X X X X X
Mean Dep. Var 0.0303 0.181 0.249 0.498 0.0416
N 1322 1322 1322 1322 1322
Baseline: January 2020.
Standard errors clustered at the individual level and regional level in parentheses.
∗
p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

The COVID-19 outbreak appears to have caused a shift towards technocratic preferences
among Spanish citizens. After the outbreak, they are on average more willing to be ruled
by experts rather than politicians, and favor technical expertise and good management over
ideological representation. This is especially the case for those more exposed to the virus, and
holds irrespective of how we measure these preferences.

20
4.2.3 Strong leadership and civil liberties

A large number of governments have responded to the COVID crisis with emergency powers
that curtail civil liberties, instead of relying on regular powers and citizens’ cooperation and
trust in each other and in the governments’ instructions. In this section, we explore whether
these coercive measures are in line with citizens’ assessment of the health vs. individual freedom
dilemma.
To this aim, we use the experiment that compared the three global threats: COVID, cli-
mate change, and international terrorism, with two additional outcomes. Specifically, we asked
for the level of agreement with the following statements: (1) drastic measures should be taken
to stop [coronavirus/climate change/international terrorism], even if that may entail a limi-
tation of individual liberty and (2) in order to cope with a challenge like [coronavirus/climate
change/international terrorism], we need to unite around a strong leadership. The nature of
the threat was assigned randomly to each respondent.
The main result in Table 8 and Figure 4 is that the COVID-19 crisis triggers a significantly
different response than the other two threats considered in the experiment. This is especially
the case for individual freedom: in this case, citizens are especially willing to support drastic
measures even if they curtail basic individual liberties. The average level of agreement with
this trade-off is extremely high in the case of the COVID, more than two points higher than
in the cases of climate change and terrorism. A similar, but less pronounced effect is found in
the strong leadership outcome. While agreement is high in all cases, the COVID threat seems
to be especially effective at fostering support for a strong leader.

21
 Table 8: Willingness to:

Sacrifice individual freedoms Unite around strong leader

(1) (2) (3) (4)
COVID 2.083∗∗∗ 2.929∗∗∗ 0.595∗∗∗ 1.705∗∗∗
(0.160) (0.171) (0.148) (0.170)

Terrorism -0.623∗∗∗ 0.0966 -0.148 0.796∗∗∗

(0.172) (0.182) (0.145) (0.165)

Constant (Climate Change) 6.132∗∗∗ 7.246∗∗∗

(0.121) (0.101)
Controls X X
Mean Dep. Var 6.619 6.619 7.395 7.395
N 1606 1606 1606 1606
Robust standard errors in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
Controls: gender, age, region FE, education FE, PM assessment, left voter, other parties’ voter.

Figure 4: Global threats and democracy

Willingness to sacrifice individual liberties and union around strong leadership

Liberties Strong Leadership
9 9

8 8

7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0

Covid Climate Terror

This is and interesting result, since climate change and terrorism are also global threats
with large externalities that require coordination across countries - and hence, would benefit
from a strong leadership-, and because many effective steps towards climate change mitigation
and terrorist attacks prevention could come along important liberty reductions. Nonetheless,
citizens are more willing to give up liberty to fight COVID, which could presumably have a
larger direct and immediate impact on their individual welfare.

22
4.2.4 Cooperation or discipline?

The previous results suggest that the COVID-19 threat is especially suited for creating demand
for authoritarian responses. Governments all over the world used warlike vocabulary to frame
the emergency and the expected response from citizens. Such a framing exercise could reinforce
the authoritarian turn and, perhaps, favor compliance. We designed an additional experiment
to explore to what extent an authority/discipline frame of the crisis was able to elicit support
from citizens, as opposed to what we call a cooperation/trust frame. We randomly assigned
respondents to two pairs of statements regarding citizens’ behavior during the crisis, and asked
them to express their level of agreement. The items were phrased as follows:

Cooperation/trust frame

1. To cope with the coronavirus crisis, we must all cooperate with each other
2. We should all trust the indications of the experts to mitigate the effects of the coro-
navirus

Discipline/authority frame

1. In order to face the coronavirus crisis, we must all be disciplined

2. We should all strictly follow the authorities’ orders to mitigate the effects of the
coronavirus

In this case, results show relatively modest but significant differences, as displayed in Table 9
and Figure 5. However, in both cases, we find that on average citizens tend to support the
discipline/authority frame more than the cooperation/trust one. This suggests that, to some
extent, the widely adopted coercive framing have slightly more support than a cooperative
approach. This result points in a similar direction as the previous experiments: the COVID-19
crisis appears to be a fertile ground for obedience and acceptance of a more authoritarian and
hierarchical rule.

23
 Table 9: Cooperation vs discipline

Must cooperate Should trust experts

vs be disciplined vs follow orders
(1) (2) (3) (4)
Discipline/Authority 0.139∗ 0.149∗∗ 0.268∗∗∗ 0.276∗∗∗
(0.0733) (0.0747) (0.0726) (0.0739)

Constant (Cooperation/Trust) 5.045∗∗∗ 4.806∗∗∗

(0.118) (0.116)
Controls X X
Mean Dep. Var 5.253 5.253 5.208 5.208
N 1606 1606 1606 1606
Robust standard errors in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
Controls: gender, age, region FE, education FE, PM assessment, left voter,
other parties’ voter.

Figure 5: Cooperation vs discipline

Agreement with cooperation vs discipline

We must all _ We should _ to mitigate
6 6

5.5 5.5

5 5

4.5 4.5

4 4

3.5 3.5

3 3

2.5 2.5

2 2

1.5 1.5

1 1

cooperate be disciplined trust experts' indications follow authorities' orders

Table A3 in the Appendix reports covariate balance tests across treatment arms. As one
would expect because of randomization, there are almost no statistical differences across groups,
with very few exceptions that may arise because due to sampling and chance (in this case, the
Prime Minister assessment, and left wing and other parties’ voters). Nevertheless, the estimates
in column 2 of Table 9 remain very stable after including a number of controls.

24
5 Conclusions

In this paper we have explored citizens’ political reaction to the democratic dilemmas posed
by the COVID-19 outbreak. A global pandemic that poses immediate health and economic
threats to a large share of the population can be expected to have deep political consequences.
And the nature of this threat makes it potentially a fruitful opportunity for a technocratic and
perhaps authoritarian turn.
We have shown how, following the outbreak, citizens shifted strongly towards a preference
for technocratic governance and strong leadership. They appear to be willing to trade-off
protection against the virus for individual liberties and ideological representation. Of course
this preference change might be temporary. While a short term exchange of civil liberties and
pluralism for protection against the virus may be rational and even relatively innocuous, the
question is whether this shift towards technocratic and authoritarian preferences will have long
lasting consequences for the affected democracies.
There are at least two channels through which the pandemics might endanger democracy: a
direct and an indirect one. The direct channel would be a durable transformation of preferences
for representation. If citizens negatively update their beliefs about the ability of democratic,
representative and liberal systems to protect them against these types of threats, the preference
change we have documented here may be more stable in the long term.
Indirectly, the shock opens a window of opportunity for would-be authoritarian leaders and
institutions to seize and centralize power, limit checks and balances and increase population
control and surveillance. In the first stage they will not encounter resistance from the public,
and then the authoritarian turn may outlast the pandemics. The early shift of preferences
that we have documented in this paper, even if it turns out to be short-lived, could offer an
opportunity for policy changes that may then self-enforce and move some democracies towards
a new political equilibrium of more centralized and less limited power, as well as more invasive
surveillance of the population.
As we have emphasized throughout the paper, the COVID crisis is a very large shock to
public health and to the economy, which in turn gives citizens the opportunity to learn about
their governments, based on their political and policy responses. Further disentangling the

25
relative importance of the health and economic shock is left for future research.

References

Aassve, Arnstein, Guido Alfani, Francesco Gandolfi and Marco Le Moglie. 2020. Epidemics
and Trust: The Case of the Spanish Flu. Working Papers 661 IGIER (Innocenzo Gasparini
Institute for Economic Research), Bocconi University.

Acemoglu, Daron and James A Robinson. 2001. “A theory of political transitions.” American
Economic Review 91(4):938–963.

Achen, Christopher H and Larry M Bartels. 2017. Democracy for realists: Why elections do
not produce responsive government. Vol. 4 Princeton University Press.

Aidt, Toke S and Gabriel Leon. 2016. “The democratic window of opportunity: Evidence from
riots in Sub-Saharan Africa.” Journal of Conflict Resolution 60(4):694–717.

Alfani, Guido and Marco Percoco. 2019. “Plague and long-term development: the lasting effects
of the 1629–30 epidemic on the Italian cities.” The Economic History Review 72(4):1175–1201.

Alfani, Guido and Tommy E. Murphy. 2017. “Plague and Lethal Epidemics in the Pre-Industrial
World.” The Journal of Economic History 77(1):314–343.

Algan, Yan, Sergei Guriev, Elias Papaionnaou and Evgenia Passari. 2017. “The European trust
crisis and the rise of populism.” Working Paper EBRD .

Ananyev, Maxim and Sergei Guriev. 2019. “Effect of Income on Trust: Evidence from the 2009
Economic Crisis in Russia.” The Economic Journal .

Ang, Yuen Yuen. 2020. “When COVID-19 Meets Centralized, Personalized Power.” Nature
Human Behavior forthcoming.

Ashworth, Scott, Ethan Bueno de Mesquita and Amanda Friedenberg. 2018. “Learning about
voter rationality.” American Journal of Political Science 62(1):37–54.

26
Autor, David, David Dorn, Gordon Hanson and Kaveh Majlesi. Forthcoming. “Importing
Political Polarization? The Electoral Consequences of Rising Trade Exposure.” American
Economic Review .

Ballard-Rosa, Cameron, Amalie Jensen and Kenneth Scheve. 2018. “Economic decline, social
identity, and authoritarian values in the United States.” American Political Science Associ-
ation, Boston .

Bertsou, Eri and Daniele Caramani. 2019. “People Haven’t Had Enough of Ex-
perts: Technocratic Attitudes among Citizens in Nine European Democracies.”.
https://ecpr.eu/Events/PaperDetails.aspx?PaperID=45843EventID=123.

Brückner, Markus and Antonio Ciccone. 2011. “Rain and the democratic window of opportu-
nity.” Econometrica 79(3):923–947.

Carlin, Ryan E, Gregory J Love and Elizabeth J Zechmeister. 2014. “Natural disaster and
democratic legitimacy: The public opinion consequences of Chile’s 2010 earthquake and
tsunami.” Political Research Quarterly 67(1):3–15.

Colantone, Italo and Piero Stanig. 2018. “The Trade Origins of Economic Nationalism: Import
Competition and Voting Behavior in Western Europe.” American Journal of Political Science
62(4):936–953.

Dommett, Katharine and Warren Pearce. 2019. “What do we know about public attitudes
towards experts? Reviewing survey data in the United Kingdom and European Union.”
Public Understanding of Science 28(6):669–678.

Fetzer, Thiemo, Lukas Hensel, Johannes Hermle and Chris Roth. 2020. “Coronavirus Percep-
tions and Economic Anxiety.” arXiv preprint arXiv:2003.03848 .

Foster, Chase and Jeffrey Frieden. 2017. “Crisis of trust: Socio- economic determinants of
Europeans’ confidence in government.” European Union Politics .

Franck, Raphaël. 2016. “The political consequences of income shocks: explaining the consoli-
dation of democracy in France.” Review of Economics and Statistics 98(1):57–82.

27
Healy, Andrew and Neil Malhotra. 2009. “Myopic voters and natural disaster policy.” American
Political Science Review 103(3):387–406.

Jedwab, Remi, Noel D. Johnson and Mark Koyamah. 2016. “Bones, bacteria and break points:
the heterogeneous spatial effects of the Black Death and long-run growth.” GMU Working
Paper in Economics No. 16-30 . Available at SSRN: https://ssrn.com/abstract=2800002.

Katz, Gabriel and Ines Levin. 2016. “The dynamics of political support in emerging democracies:
evidence from a natural disaster in Peru.” International Journal of Public Opinion Research
28(2):173–195.

Kotschy, Rainer and Uwe Sunde. 2019. “Income Shocks, Inequality, and Democracy.” The
Scandinavian Journal of Economics forthcoming.

Margalit, Yotam. 2019. “Political responses to economic shocks.” Annual Review of Political
Science .

Rehman, Faiz Ur and Paolo Vanin. 2017. “Terrorism risk and democratic preferences in Pak-
istan.” Journal of Development Economics 124:95–106.

Richardson, Gary and Michael McBride. 2009. “Religion, longevity, and cooperation: The case
of the craft guild.” Journal of Economic Behavior & Organization 71(2):172–186.

Voigtländer, Nico and Hans-Joachim Voth. 2012. “Persecution Perpetuated: The Medieval
Origins of Anti-Semitic Violence in Nazi Germany*.” The Quarterly Journal of Economics
127(3):1339–1392.

Xue, Melanie Meng and Mark Koyama. 2018. Autocratic Rule and Social Capital: Evidence
from Imperial China. Technical report Mimeo, Northwestern University.

28
Appendix

Table A1: Covariate Balance, Geographical Scope treatment

(1) (2) (3) (4) T-test F-test

Region Spain Europe Asia Difference for joint
Variable Mean/SE Mean/SE Mean/SE Mean/SE (1)-(2) (1)-(3) (1)-(4) (2)-(3) (2)-(4) (3)-(4) orthogonality
Female 0.509 0.510 0.505 0.509 -0.001 0.004 0.000 0.005 0.001 -0.004 0.008
(0.025) (0.025) (0.025) (0.025)
Age 45.229 45.885 45.094 45.865 -0.656 0.135 -0.636 0.791 0.020 -0.771 0.288
(0.765) (0.785) (0.764) (0.782)
Less than secondary school 0.429 0.465 0.411 0.436 -0.036 0.018 -0.007 0.054 0.029 -0.026 0.825
(0.025) (0.025) (0.025) (0.025)
Secondary/High School 0.090 0.058 0.089 0.085 0.032* 0.001 0.005 -0.032* -0.027 0.004 1.541
(0.014) (0.012) (0.014) (0.014)
College 0.239 0.217 0.218 0.207 0.022 0.022 0.032 -0.000 0.011 0.011 0.422
(0.021) (0.021) (0.021) (0.020)
Andalusia 0.142 0.145 0.171 0.197 -0.003 -0.029 -0.055** -0.026 -0.052* -0.026 1.846
(0.017) (0.018) (0.019) (0.020)
Catalonia 0.157 0.198 0.158 0.150 -0.040 -0.001 0.007 0.039 0.048* 0.009 1.237
(0.018) (0.020) (0.018) (0.018)
Madrid 0.162 0.170 0.163 0.192 -0.008 -0.001 -0.030 0.007 -0.022 -0.029 0.516
(0.018) (0.019) (0.018) (0.020)
Valencian Community 0.147 0.100 0.099 0.090 0.047** 0.048** 0.057** 0.001 0.010 0.009 2.353*
(0.018) (0.015) (0.015) (0.014)
Other Regions 0.392 0.388 0.408 0.372 0.004 -0.017 0.020 -0.021 0.016 0.037 0.387
(0.024) (0.024) (0.024) (0.024)
Prime Minister Assessment 3.264 3.270 3.300 3.379 -0.006 -0.035 -0.115 -0.030 -0.109 -0.080 0.794
(0.060) (0.060) (0.059) (0.059)
Regional PM Assessment 3.032 2.953 2.928 2.930 0.080 0.104 0.102 0.024 0.022 -0.002 0.853
(0.054) (0.055) (0.050) (0.052)
Generalised trust 4.489 4.268 4.475 4.314 0.221 0.014 0.175 -0.208 -0.047 0.161 0.739
(0.133) (0.124) (0.135) (0.138)
Left-wing voter 0.347 0.325 0.332 0.289 0.022 0.015 0.057* -0.007 0.036 0.042 1.114
(0.024) (0.023) (0.023) (0.023)
Riht-wing voter 0.204 0.225 0.215 0.242 -0.021 -0.011 -0.037 0.010 -0.017 -0.027 0.579
(0.020) (0.021) (0.020) (0.021)
Regional party voter 0.065 0.075 0.057 0.050 -0.010 0.008 0.015 0.018 0.025 0.007 0.794
(0.012) (0.013) (0.012) (0.011)
Other parties/responses 0.384 0.375 0.396 0.419 0.009 -0.012 -0.035 -0.021 -0.044 -0.023 0.603
(0.024) (0.024) (0.024) (0.025)
COVID exposure 0.122 0.113 0.121 0.122 0.010 0.001 0.000 -0.009 -0.010 -0.001 0.087
(0.016) (0.016) (0.016) (0.016)
N 401 400 404 401
F-test of joint significance (F-stat) 0.967 0.671 1.461 0.626 0.862 0.403
F-test, number of observations 801 805 802 804 801 805
Notes: Values for t-tests refer to cross-group mean differences. Robust standard errors in
parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.

29
 Table A2: Covariate Balance, Global threats treatment

(1) (2) (3) T-test F-test

Covid Climate Terror Difference for joint
Variable Mean/SE Mean/SE Mean/SE (1)-(2) (1)-(3) (2)-(3) orthogonality
Female 0.508 0.507 0.509 0.002 -0.001 -0.003 0.004
(0.022) (0.022) (0.022)
Age 45.809 45.240 45.504 0.569 0.306 -0.264 0.185
(0.660) (0.663) (0.688)
Less than secondary school 0.452 0.443 0.410 0.009 0.042 0.033 1.079
(0.022) (0.021) (0.021)
Secondary/High School 0.095 0.074 0.071 0.021 0.024 0.003 1.152
(0.013) (0.011) (0.011)
College 0.207 0.218 0.236 -0.010 -0.028 -0.018 0.638
(0.018) (0.018) (0.018)
Andalusia 0.168 0.171 0.152 -0.003 0.017 0.020 0.447
(0.016) (0.016) (0.016)
Catalonia 0.179 0.149 0.169 0.030 0.011 -0.020 0.945
(0.017) (0.015) (0.016)
Madrid 0.191 0.158 0.167 0.032 0.024 -0.008 1.030
(0.017) (0.016) (0.016)
Valencian Community 0.088 0.121 0.118 -0.033* -0.030 0.003 2.027
(0.012) (0.014) (0.014)
Other Regions 0.374 0.400 0.395 -0.027 -0.021 0.005 0.447
(0.021) (0.021) (0.021)
Prime Minister Assessment 3.198 3.326 3.386 -0.128* -0.188*** -0.060 3.487**
(0.052) (0.052) (0.050)
Regional PM Assessment 2.985 2.952 2.946 0.033 0.039 0.006 0.210
(0.047) (0.044) (0.045)
Generalised trust 4.366 4.348 4.446 0.018 -0.079 -0.097 0.200
(0.115) (0.112) (0.118)
Left-wing voter 0.340 0.309 0.320 0.031 0.020 -0.011 0.602
(0.021) (0.020) (0.020)
Riht-wing voter 0.198 0.229 0.238 -0.031 -0.040 -0.009 1.393
(0.017) (0.018) (0.018)
Regional party voter 0.052 0.067 0.066 -0.015 -0.013 0.001 0.650
(0.010) (0.011) (0.011)
Other parties/responses 0.409 0.395 0.376 0.015 0.033 0.018 0.611
(0.021) (0.021) (0.021)
COVID exposure 0.135 0.106 0.118 0.028 0.017 -0.012 1.025
(0.015) (0.013) (0.014)
N 535 537 534
F-test of joint significance (F-stat) 1.150 1.540* 0.404
F-test, number of observations 1072 1069 1071
Notes: Values for t-tests refer to cross-group mean differences. Robust standard errors in
parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.

30
 Table A3: Covariate Balance, Cooperation vs. Discipline treatment

(1) (2) T-test F-test

Cooperation Discipline Difference for joint
Variable Mean/SE Mean/SE (1)-(2) orthogonality
Female 0.504 0.512 -0.009 0.124
(0.018) (0.018)
Age 45.424 45.611 -0.187 0.058
(0.548) (0.546)
Less than secondary school 0.442 0.429 0.013 0.273
(0.018) (0.018)
Secondary/High School 0.079 0.081 -0.002 0.018
(0.010) (0.010)
College 0.218 0.223 -0.004 0.040
(0.015) (0.015)
Andalusia 0.167 0.160 0.007 0.164
(0.013) (0.013)
Catalonia 0.166 0.165 0.001 0.005
(0.013) (0.013)
Madrid 0.161 0.182 -0.021 1.268
(0.013) (0.014)
Valencian Community 0.115 0.102 0.013 0.686
(0.011) (0.011)
Other Regions 0.390 0.390 -0.000 0.000
(0.017) (0.017)
Prime Minister Assessment 3.352 3.254 0.099* 2.758*
(0.043) (0.041)
Regional PM Assessment 2.950 2.971 -0.021 0.157
(0.038) (0.037)
Generalised trust 4.297 4.478 -0.181 1.866
(0.092) (0.095)
Left-wing voter 0.301 0.345 -0.044* 3.477*
(0.016) (0.017)
Riht-wing voter 0.220 0.224 -0.004 0.040
(0.015) (0.015)
Regional party voter 0.058 0.065 -0.007 0.310
(0.008) (0.009)
Other parties/responses 0.421 0.366 0.054** 4.979**
(0.017) (0.017)
COVID exposure 0.115 0.124 -0.008 0.267
(0.011) (0.012)
N 806 800
F-test of joint significance (F-stat) 0.723
F-test, number of observations 1606
Notes: Values for t-tests refer to cross-group mean differences. Robust
standard errors in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.

31