This study analyzes the impact of COVID-19 deniers on the spread of COVID-19 in Germany. In a first step, we establish a link between regional proxies of COVID-19 deniers and infection rates. We then estimate the causal impact of large anti-lockdown protests on the spread of COVID-19 using an event study framework. Employing novel data on bus stops of travel companies specialized in driving protesters to these gatherings, and exploiting the timing of two large-scale demonstrations in November 2020, we find sizable increases in infection rates in protesters’ origin regions after these demonstrations. Individual-level evidence supports the main results by documenting that COVID-19 deniers engage less in health protection behavior. Our results contribute to the debate about the public health costs of individual behavior that has detrimental externalities for the society.

Societal cohesion is key in confining the outbreak of diseases that threaten collective survival. If some part of the society does not comply with public health precautions aimed at stopping the
spread of a deadly disease, the effectiveness of public policy and preventative efforts undertaken by others are substantially constrained. The ongoing COVID-19 pandemic is a case in point.
Even with over 100 million people infected by the disease worldwide and more than 2 million associated deaths, a notable segment of the population denies the threat posed by-or even the existence of-the novel coronavirus SARS-CoV-2. According to survey data from YouGov Cambridge (2020), 13 percent of people in the U.S. believe the coronavirus probably or definitely does not exist. Similar figures are reported for Germany and France (10 percent).

How does this group contribute to the spread of COVID-19? Do COVID-19 deniers behave differently than the majority of the population? And do large protests of this group affect the transmission of the disease? In this study, we show that the spread of COVID-19 can be substantially increased by individuals who downplay the disease’s public health threat and reject both recommended and mandatory behaviors aimed at curtailing its transmission.

In a first step, we establish a link between different proxies for the regional presence of COVID19 deniers and infection rates in Germany. Motivated by the observation that COVID-19 deniers disproportionally support populist parties and largely oppose vaccinations, we use regional information on vote shares of the largest populist party in Germany and the share of children vaccinated against measles as proxies for a potentially high share of COVID-19 deniers. We find a significant and sizable correlation between these proxies and COVID-19 infection rates, which suggests that a higher share of COVID-19 deniers in a region may facilitate the spread of the coronavirus.

Building upon this observation, we show in a second step that COVID-19 deniers indeed contribute significantly to regional disparities in COVID-19 infection rates. Specifically, we estimate the causal impact of large anti-lockdown protests organized by COVID 19 deniers on the spread of COVID-19. For identification, we exploit the particularity that an alliance of bus companies has specialized in transporting anti-lockdown protesters to rallies across Germany.

Using web-scraped data on all possible points of departure offered by this alliance allows to identify the home regions of protesters. This information is used in an event study framework where we compare the development of infection rates in regions with and without such bus stops in the aftermath of two large-scale demonstrations in November 2020. Our results show a significant increase in new COVID-19 cases in home areas of protesters after the demonstrations.

The effects are most pronounced in regions where bus stops exist even in small towns with fewer than 20,000 residents. This finding is in line with the interpretation that regions with the highest demand for transportation to the demonstrations see the highest increases in COVID-19 infections after the protests. We estimate that those areas faced a 35.9 percent higher infection rate by the end of 2020.

Our results are robust to a number of sensitivity checks. Allowing county-level characteristics associated with the spread of COVID-19 to have differential effects on our outcome variables over time does not change our results. For example, we include interactions between time dummies and the infection rate just before the protests, nursing home capacities, population density, GDP 2 per capita, and bus stops from the largest commercial bus travel operator. Our results also do
not depend on states that see a particularly high increase in infection rates or share a border with highly affected neighboring countries.

Moreover, using different measures of COVID-19 infections does not change our main findings qualitatively. Finally, we complement our regional evidence with individual-level survey data from the
beginning of the pandemic. Regression results from this survey suggest that individuals who downplay the risk of COVID-19 infections also engage less in COVID-19 mitigation strategies, exhibit lower trust in the government and in public health institutions, and are less likely to acquire information about COVID-19 from established media sources.

To the best of our knowledge, this is the first study to quantify the impact of COVID-19 deniers on the local spread of this disease. There are, however, a number of studies analyzing related questions. One strand of the economics’ COVID-19 literature focuses on the influence of large-scale events or gatherings on the spread of COVID-19. For instance, Dave et al. (2020) investigate the spread of COVID-19 as a consequence of a large U.S. gathering of motorcycle
enthusiasts that took place without any infection mitigation strategies. They estimate that counties with the highest share of event attendees experienced 6.4 to 12.5 percent higher COVID19 cases than counties without attendees. In a similar vein, Harris (2020) investigates the outbreak of COVID-19 at the University of Wisconsin and concludes that a cluster of bars facilitated the spread of the coronavirus. Large gatherings do not automatically contribute to
an increase of COVID-19 cases, as shown by Dave et al. (2020). Their study looks at the impact that Black Lives Matter gatherings have on the spread of COVID-19 and find that these demonstrations -at which participants largely complied with coronavirus mitigation strategies- have no or only a marginal effect on COVID-19 infections.

By analyzing the effect of large-scale political protests where participants deliberately disobeyed health care regulations, our study directly relates to the trade-o when civil liberties and public health policies seem at odds. Citizens’ right to protest is a corner stone of modern democracies. In times of public crisis, such as a pandemic, local authorities may, however, deem such protests as too dangerous for society. As Alsan et al. (2020) show, individuals are heterogeneous in their willingness to trade these civil liberties for uncertain public health improvements. Our results quantify the public health costs associated with this trade-off.

Our study also relates to the finding that partisanship is a major driver of health outcomes and behavior in the U.S. during the current pandemic (see Allcott et al., 2020; Clinton et al., 2020; Gadarian et al., 2020; Grossman et al., 2020; Makridis and Rothwell, 2020, among others). Democrats are usually more willing to reduce mobility and voluntarily engage in social distancing; Republicans are less likely to do so. Less is known, however, about COVID-19 containment
behavior along political lines in less politically polarized countries. Barbieri and Bonini (2020) and Mellacher (2021) show evidence that individual mobility and COVID-19 related deaths are higher in areas where populist parties enjoy larger vote shares. Our study documents this relationship for Germany, the largest economy and democracy in Europe and a country where the biggest populist party systematically downplays the threat of COVID-19.

Our findings also contribute to studies that document the spatial influence of social capital on the spread of COVID-19. Individuals living in areas of the U.S. or Europe with higher social capital reduce their mobility much more than individuals living in regions with lower social capital (Bargain and Aminjonov, 2020; Brodeur et al., 2020; Barrios et al., 2021; Durante et al., 2021). Similarly, Bartscher et al. (2020) show that COVID-19 spreads more slowly in European regions with higher social capital. Social capital is mainly measured by trust in politicians or institutions, blood or organ donations, and electoral turnout. By validating that skepticism about COVID-19 is associated with less trust in institutions, we add a novel measure of trust in public health care-that is, regions with a high share COVID-19 deniers-to the existing spatial measures of institutional trust.

Moreover, this study is related to research that is concerned about how media coverage of COVID-19 influences public health. We document that individuals who believe COVID-19 poses no great threat to their personal health nor to public health are less likely to turn to established media for information about COVID-19. Given German TV, radio, and newspapers’ rather homogenous coverage of risks associated with the coronavirus, COVID-19 deniers must use other (social media) channels to access and disseminate their views about the pandemic. This evidence is in line with findings of Bursztyn et al. (2020) and Simonov et al. (2020) who show that the downplaying of the coronavirus threat by Fox News cable programs increases infection and death rates associated with COVID-19.

This study continues as follows. We present background information about the spread of COVID-19 in Germany and discuss its relationship to COVID-19 deniers and their protest movement. The section thereafter introduces the data employed for the empirical analysis. In Section 4, we present our main descriptive and causal event study estimates including several sensitivity analyses. In the subsequent section, our empirical analysis is complemented by individual evidence that offers potential explanations for our main results. Finally, in Section 6, we discuss the broader relevance of our findings.