Regular Article
The political economy of health epidemics: Evidence from the Ebola outbreak

https://doi.org/10.1016/j.jdeveco.2021.102651Get rights and content

Highlights

  • I investigate whether political incentives affect the government's response during a health epidemic.

  • The government prioritized villages hit by the epidemic, but misallocated relief efforts toward politically swing villages.

  • Voters' support for the national incumbent party increased in areas with higher resource misallocation.

Abstract

This paper investigates whether political incentives affect the government's response during a health epidemic and the subsequent effects on citizens' voting behavior. Leveraging novel data, I study this question in the context of the 2014 Ebola outbreak in Liberia. The national incumbent government appropriately prioritized the allocation of resources to villages affected by the epidemic. By building a spatiotemporal epidemiological model that estimates the ex-ante optimal allocation of relief efforts, there is also evidence that resources were misallocated toward electoral swing villages. Instead, no resources were diverted toward core supporters or co-ethnic villages. Voters, in turn, reacted by rewarding the national incumbent party in areas where additional resources were misallocated.

Introduction

Governments can promote economic development and citizens' welfare through an efficient allocation of national resources. This is especially true in developing countries where the population mostly relies on the state for basic services. However, political incentives may induce distortions at the expense of citizens' welfare. Although it is well documented that distortions exist in the provision of public goods and social welfare programs (Golden and Min, 2013), evidence on how governments allocate relief efforts at the time of disasters is still scarce. Disasters, however, are a unique test of governmental accountability. On the one hand, governments have the opportunity to influence voters, signaling through their responsiveness how well they can perform. On the other hand, citizens have a chance to learn about the government's capacity and thus ensure electoral accountability at the time of voting. Therefore, understanding the efficacy of, and motives behind, the response to disasters is of paramount policy relevance to provide better incentives to governments to act appropriately during crises.

Most of the distributive politics literature builds on theoretical models explaining under which circumstances politicians target core (Cox and McCubbins, 1986) or swing (Lindbeck and Weibull, 1987; Dixit et al., 1996; Stokes, 2005) voters. The models predict that, when no party can deliver benefits with particular efficiency to any group of voters, then politicians favor swing voters, defined as those ideologically indifferent to the alternatives. However, results from the empirical literature testing this hypothesis are mixed. Recent studies conducted in developing-world democracies, which use individual survey data, do not always find evidence of swing voters receiving party largesse (Stokes et al., 2013; Diaz-Cayeros et al., 2015). However, when using aggregated definitions of swing voters, rather than analyzing individual behavior as the theory predicts, research around the world uncovers frequent political targeting of swing districts (Stokes et al., 2013).1

In addition, past research in developing countries—and especially in Africa—shows that politicians allocate more resources to co-ethnic groups (Kramon and Posner, 2013).2 Empirical evidence on how political motives drive the allocation of relief efforts at the time of disasters (Garrett and Sobel, 2003; Reeves, 2011; Gasper and Reeves, 2012) is also primarily limited to the United States, given the inherent difficulty in gathering data in high-risk settings.

Given the mixed empirical results, a key question remains: How do politicians allocate public goods in developing countries at the time of disasters? The 2014 West Africa Ebola outbreak (Ebola Virus Disease, EVD hereafter) in Liberia provides a unique context to fill this gap. In just over one year more than 28,000 cases were reported—almost half of which resulted in fatalities. Estimates of the total cost of the epidemic reach up to $25 billion (World Bank Group, 2014a). By September 2014, an international state of emergency was declared and $4 billion of foreign aid was channeled to the three most affected countries: Guinea, Liberia, and Sierra Leone. Schools and markets closed; in Liberia, the Senatorial election was postponed due to increased fear. Taking advantage of this health epidemic, the goal of this study is twofold. First, I empirically test whether political incentives drove the allocation of resources (primarily foreign aid), especially toward a distinct type of voters. Second, I investigate whether the epidemic affected citizens’ voting behavior during the Senatorial election at the end of 2014.

In fragile, post-conflict democracies, such as Liberia, where patronage politics and corruption are common (Catholic Relief Services, 2016; Transparency International, 2019), political incentives could play a role in getting the government to respond appropriately to disasters. However, given how lucrative holding public office is in the country (Pailey and Harris, 2020), electoral incentives could also lead politicians to strategically allocate resources toward voters who do not need relief efforts the most, but who are politically important in winning seats. If that is the case, then, political incentives might lead to distortions in the allocation of resources. This analysis focuses on local politicians, i.e., senators up for re-election in 2014, who represent the 15 counties of Liberia. In this context, the misallocation of relief efforts could derive from the central, incumbent government deciding, at the national level, how to allocate resources down to the counties, or it could be the result of local politicians being able to attract more national resources in areas in which their electoral incentives are the highest. This study aims to shed some light on these open questions.

The first step is the creation of a comprehensive village-level database on the EVD outbreak in Liberia, encompassing the entire set of 9686 villages in the 15 counties of the country. The database combines (i) proprietary patient-level data from the Liberian Ministry of Health (MOH), which contains every person tested (ii) data on relief effort measures, such as the construction of Ebola Treatment Units (ETUs) and Community Care Centers (CCCs), and the deployment of burial teams; (iii) publicly available data on the 2011 and 2014 elections from the National Election Commission (NEC); (iv) the 2008 National Population and Housing Census from the Institute of Statistics and Geo-Information Services (LISGIS).

Using this novel dataset, I compare the reactions of the national incumbent government and citizens in villages affected by EVD—primarily defined as villages where at least one probable case of EVD was recorded—with villages in which no cases were registered, within the same county and controlling for a large set of observable pre-outbreak characteristics.3 However, villages that were directly hit by EVD are different in terms of other unobserved characteristics in comparison to villages that were not hit. For example, the prior level of trust in the government might have affected how the government was able to effectively provide relief efforts, and may thus have influenced the recording of cases. The analysis then implements an alternative instrumental variable approach, which uses distance by road to the EVD place of origin to address these concerns. The validity assumption is that similar villages, in terms of a defined set of observable characteristics, are less likely to be affected by EVD if they are farther away from the place of EVD origin. I also explore heterogeneous effects across swing, core, or co-ethnic villages. I finally investigate whether similar effects by incumbent senators different from the national incumbent party are present. Since there exist few differences across these types of villages, I control for those observable geographical and socio-demographic characteristics in the analysis. When data were available before the disaster, such as for voting behavior, a difference-in-differences approach (combined with coarsened matching) is implemented in order to minimize the influence of unobserved time-invariant determinants.

I first establish that the national incumbent government prioritized resources to villages affected by EVD compared to unaffected ones. I then find evidence that more resources, such as CCCs and ETUs, were allocated toward swing villages affected by the epidemic. By constructing a spatiotemporal epidemiological model that identifies the villages that were ex-ante more likely to be affected by EVD, I quantitatively measure misallocation as the difference between how much the government spent and how much it should have spent given the predicted number of EVD cases from the model, and I examine whether more resources were allocated toward swing villages hit by EVD than those needed. From the estimation, I find that on average $38,886 were misallocated toward each swing village affected by the epidemic, compared to non-swing affected villages, diverting about 9.4% of the total governmental expenses. Based on the model predictions, I do not find any evidence that resources were diverted toward core supporters or co-ethnic villages. Similarly, there is no misallocation toward villages where incumbent senators were from political parties different from the national incumbent party. Overall, the national incumbent government was able to respond to the epidemic in the affected areas. However, more relief efforts were diverted toward swing villages. The findings suggest that where senators from the national incumbent party faced the highest competition to win seats at the 2014 Senatorial election, political incentives led different levels of the government to divert resources away from those in need.

Turning to the reaction of the citizens, turnout was consistently higher (4 percentage points) in the 2014 Senatorial election for citizens affected by EVD, and even higher (5.2 percentage points) for citizens affected by EVD living in swing villages, compared to unaffected villages. This confirms that individuals' experiences with the epidemic highlighted the government's response to the crisis (Fair et al., 2017). On average, senators from the national incumbent party did not lose vote share in areas affected by EVD compared to unaffected ones. Yet, they gained vote share (8.1 percentage points) and increased their margin of votes (9.9 percentage points) in swing villages affected by EVD, where resources were diverted. This gain in political support completely overcame the loss in non-swing villages affected by EVD, which potentially experienced a worse response. Overall, senators from the national incumbent party successfully won four out of 15 seats in the 2014 Senatorial election, especially in those counties in which more resources were misallocated. However, this political allocation of resources did not come without cost for citizens' welfare. A simple static cost-benefit analysis exercise estimates that, without political (mis)allocation of resources, 8.6% of lives could have been saved.

This study contributes to three strands of the literature. First, it builds on research investigating politicians' electoral incentives to allocate public funding (Brollo and Nannicini, 2012; Finan and Mazzocco, 2016), to manipulate national policies (List and Sturm, 2006; Brollo et al., 2015), and to allocate resources after disasters in the United States (Garrett and Sobel, 2003; Reeves, 2011; Gasper and Reeves, 2012).4 This study is also in line with the theoretical literature showing how electorally accountable policymakers may rationally under-invest in preparation for rare and potentially catastrophic events, and how they instead provide more palliative, visible effort to pander to the public (Cole and Werker, 2008; Canes-Wrone et al., 2001; Maskin and Tirole, 2004; Choe and Raschky, 2016).5 Previous research focuses mainly on the United States, so this paper adds to the limited evidence on political misallocation in response to disasters in low-income countries (Cole et al., 2012), which might be due to the difficulty in gathering data.6 It takes advantage of a relatively unexplored type of disaster—namely, a health epidemic—when the presence of contagion effects might alter the ex-ante trade-offs politicians face to strategically allocate resources, as the government's response is a key determinant of the ultimate size of the disaster. In addition, the dynamic nature of health epidemics implies that the costs to citizens may be especially severe if the government's response is politically distorted. It contributes by quantitatively measuring the misallocation of resources and its cost to citizens' welfare. This study also adds to the empirical literature in developing countries testing the swing-voter hypothesis (Stokes et al., 2013 for a review). It provides evidence that palliative resources were diverted toward swing villages, allowing senators from the national incumbent party to win seats at the election.7

Second, it contributes to a large body of literature studying the impact of negative economic shocks on political perceptions (Fair et al., 2017; Andrabi and Das, 2017) and voting behavior (Malhotra and Kuo, 2008; Healy and Malhotra, 2010; Gasper and Reeves, 2011; Healy and Malhotra, 2013; Ashworth and Bueno de Mesquita, 2014).8 The closest study is Fluckiger et al. (2018), which shows more political support for the incumbent government in areas in which it responded more intensively to the epidemic. This study complements the findings in the rational voters’ literature by analyzing how citizens reacted to the political misallocation of resources.9

Overall, this study adds to both of these two strands of the literature, which tend to analyze, separately, the government's resource allocation decisions and citizens' reactions to those allocations. It provides a novel empirical examination of both the politicians' allocation behavior and the voters' reaction to it, in the context of governmental relief efforts at the time of an epidemic.

Finally, it contributes to a recent set of studies that investigate the role of politics during the EVD epidemic and the current COVID-19 pandemic. Campante et al., (2020) exploits the location of the cases in the United States to describe how the fear of EVD led to a lower incumbent share (Democratic) in the 2014 midterm elections. The opposition also strategically exploited this fear with meaningful impacts on elections.10 Instead, current work on the COVID-19 pandemic investigates the relationship between partisanship and responses to the epidemic (Adolph et al., 2020) or compliance with preventive measures, mainly in the United States (Allcott et al., 2020; Andersen, 2020; Barrios and Hochberg, 2020; Milosh et al., 2020). This study provides useful insights into government responses to epidemics in developing countries. More importantly, while most of the emerging research argues that COVID-19 could be exploited as a natural experiment, the widespread impacts of this worldwide pandemic make it difficult for researchers to define precise control and treatment groups to investigate the economic, political or health effects of the contagion. This study contributes methodologically by taking advantage of the spatial variation of a disease that did not affect an entire country to isolate the political economy of a health epidemic.

The paper is structured as follows: Section II provides background on the EVD outbreak and the context of politics in Liberia. Section III describes the data sources and descriptive statistics. Section IV elaborates on the empirical models and the main results for the response of the government to the epidemic. Section V does similarly for the reaction of the citizens. Section VI discusses additional results, and Section VII concludes.

Section snippets

Background

This section provides an overview of the 2014 EVD outbreak in Liberia, including information on how the government and the international community responded to the contagion. It also provides background on the political environment in the country, focusing on the Senatorial elections.

Data

I combine multiple sources of data, that were manually coded and matched at the village level for the entire set of 9686 villages in the 15 counties of Liberia. First, I seek access to proprietary patient databases from the MOH and burial teams to construct measures of EVD and the deployment of burial teams; I also access other publicly available data sources to construct measures of relief efforts, such as the location of ETUs and CCCs. Second, I use the 2008 National Population and Housing

The political response of the government

The first part of the analysis investigates the political response of the national incumbent government to the EVD outbreak. I initially compare the provision of relief efforts in villages hit and not hit by EVD, and then explore heterogeneous effects toward electoral swing villages. I use a county fixed-effects model, controlling for a set of pre-outbreak geographical and socio-demographic characteristics. I then implement an instrumental variable approach, which uses the distance (by road) to

The reaction of the citizens

The second part of the analysis investigates the reaction of the citizens to the EVD outbreak. I implement a difference-in-differences empirical strategy, combined with coarsened matching, to explore citizens' voting behavior post-outbreak. First, I outline the empirical approaches. Second, I show how the individuals' exposure to the epidemic highlighted the government's response, and lead to higher electoral participation. I also describe how the national incumbent party (UP) maintained its

Other results

This section complements the main findings on the response of the national incumbent government and the reaction of the citizens to the epidemic, by estimating how costly (in terms of human lives) to citizens, and how beneficial (in terms of seats in the Senate) to the national incumbent party (UP), was the (mis)allocation of resources.

Understanding how costly to citizens the political distortion of relief efforts was at the time of the health epidemic remains of policy interest to better align

Conclusion

I took advantage of the unique setting given by the evolution of the 2014 West Africa Ebola outbreak, combined with the timing of the Senatorial election in Liberia, to show that there exist political incentives in the allocation of (foreign aid) resources at the time of a health epidemic. By means of a comprehensive database gathered at the time of the disaster, the analysis shows that the national incumbent government appropriately responded to the epidemic in the areas affected. However,

Author statement

Elisa M Maffioli: Conceptualization; Data collection; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Writing - original draft.

References (127)

  • Eva Johansson

    Intergovernmental grants as a tactical instrument: empirical evidence from Swedish municipalities

    J. Publ. Econ.

    (2003)
  • Daron Acemoglu et al.

    The colonial origins of comparative development: an empirical investigation

    Am. Econ. Rev.

    (2001)
  • Christopher H. Achen et al.

    Blind Retrospection Electoral Responses to Drought, Flu, and Shark Attacks

    (2004)
  • Claire L. Adida et al.

    Ebola, Elections, and Immigration: How Politicizing an Epidemic Can Shape Public Attitudes

    (2018)
  • Christopher Adolph et al.

    Pandemic Politics: Timing State-Level Social Distancing Responses to COVID-19

    (2020)
  • Hirotugu Akaike

    A new look at the statistical model identification Automatic Control

    IEEE Transactions

    (1974)
  • Kathleen Alexander

    What factors might have led to the emergence of Ebola in West Africa?

    PLoS Neglected Trop. Dis.

    (2015)
  • Hunt Allcott et al.

    Polarization and Public Health: Partisan Differences in Social Distancing during the Coronavirus Pandemic

    (2020)
  • Joseph G. Altonji et al.

    An evaluation of instrumental variable strategies for estimating the effects of catholic schooling

    J. Hum. Resour.

    (2005)
  • Martin Andersen

    Early Evidence on Social Distancing in Response to COVID-19 in the United States

    (2020)
  • Tahir Andrabi et al.

    In aid we trust: hearts and minds and the Pakistan earthquake of 2005

    Rev. Econ. Stat.

    (2017)
  • Scott Ashworth et al.

    Is voter competence good for voters?: information, rationality, and democratic performance

    Am. Polit. Sci. Rev.

    (2014)
  • Mike Baiocchi et al.

    Building a stronger instrument in an observational study of perinatal care for premature infants

    J. Am. Stat. Assoc.

    (2010)
  • Nicola Barban et al.

    Assortative Mating on Education: A Genetic Assessment

    (2019)
  • John M. Barrios et al.

    Risk Perception through the Lens of Politics in the Time of the COVID-19 Pandemic

    (2020)
  • Alec T. Beall et al.

    Infections and elections: did an Ebola outbreak influence the 2014 US federal elections (and if so, how)?

    Psychol. Sci.

    (2016)
  • Alexandre Belloni et al.

    High-dimensional methods and inference on structural and treatment effects

    J. Econ. Perspect.

    (2014)
  • Liberia Country Report

    (2012)
  • Sonia Bhalotra et al.

    Education and freedom of choice: evidence from arranged marriages in vietnam

    J. Eur. Econ. Assoc.

    (2019)
  • Robert Blair et al.

    Public health and public trust: survey evidence from the Ebola Virus Disease epidemic in Liberia

    Soc. Sci. Med

    (2016)
  • Alix Boucher

    Supporting, Non-standard Mission Role: US Operations in Liberia 2014-2015 that Enabled the US and UN Response to the EVD Outbreak

    (2018)
  • Fernanda Brollo et al.

    “Tying your enemy's hands in close races: the politics of federal transfers in Brazil

    Am. Polit. Sci. Rev.

    (2012)
  • Fernanda Brollo et al.

    The Political Economy of Enforcing Conditional Welfare Programs: Evidence from Brazil

    (2015)
  • John M. Brooks et al.

    Effect of dialysis center profit-status on patient survival: a comparison of risk-adjustment and instrumental variable approaches

    Health Serv. Res.

    (2006)
  • Robin Burgess et al.

    The value of democracy: evidence from road building in Kenya

    Am. Econ. Rev.

    (2015)
  • Filipe Campante et al.

    The Virus of Fear: the Political Impact of Ebola in the U.S.

    (2020)
  • Brandice Canes-Wrone et al.

    Leadership and pandering: a theory of executive policymaking

    Am. J. Polit. Sci.

    (2001)
  • Bryan Caplan

    The Myth of Rational Voter

    (2007)
  • David Card

    Using Geographic Variation in College Proximity to Estimate the Return to Schooling

    (1993)
  • State of Peace, Reconciliation and Conflict in Liberia

    (2016)
  • Damian Clarke et al.

    Practical considerations for questionable IVs

    STATA J.

    (2018)
  • Shawn Cole et al.

    The political economy of natural disasters

    J. Conflict Resolut.

    (2008)
  • Timothy G. Conley et al.

    Learning about a new technology: pineapple in Ghana

    Am. Econ. Rev.

    (2010)
  • Timothy G. Conley et al.

    Plausibly exogenous

    Rev. Econ. Stat.

    (2012)
  • Gary W. Cox et al.

    Electoral politics as a redistributive game

    J. Polit.

    (1986)
  • Czado et al.

    Predictive model assessment for count data

    Biometrics

    (2009)
  • Matz Dahlberg et al.

    On the vote-purchasing behavior of incumbent governments

    Am. Polit. Sci. Rev.

    (2002)
  • Giacomo DeLuca et al.

    Ethnic favoritism: an axiom of politics?

    J. Dev. Econ.

    (2014)
  • David Denemark

    Partisan pork-barrel in parliamentary systems: Australian constituency-level grants

    J. Polit.

    (2000)
  • Alberto Diaz-Cayeros et al.

    Strategies of Vote Buying: Democracy, Clientelism, and Poverty Relief in Mexico

    (2015)
  • Cited by (13)

    • Elections and COVID-19 benefit payments

      2023, Journal of Asian Economics
    • Mobile phone coverage and infant mortality in sub-Saharan Africa

      2023, Journal of Economic Behavior and Organization
    • Ministers Engage in Favoritism Too

      2022, Journal of Public Economics
      Citation Excerpt :

      Aid flows from China (Dreher et al., 2019) and US humanitarian aid (Bommer et al., 2022) are directed mostly to the head of state’s birth region. Maffioli (2021) finds that resources were misallocated towards electoral swing villages during the 2014 Ebola outbreak in Liberia. We distinguish ourselves from previous work on favoritism by our systematic focus on ministers rather than country leaders.

    • The Virus of Fear: The Political Impact of Ebola in the United States

      2024, American Economic Journal: Applied Economics
    View all citing articles on Scopus

    I am indebted to Erica Field, Robert Garlick, Manoj Mohanan and Duncan Thomas for invaluable advice. I thank Nicolas-Aldebrando Benelli, Maria Carreri, Emanuele Colonnelli, Amanda Grittner, Timur Kuran, José Martinez, Gabor Nyeki, Robert M. Gonzalez, Mounu Prem, Javier Romero Haaker, Edoardo Teso, Kate Vyborny, Xiao Yu Wang, Erik Wibbels, David Yang, seminar participants at Duke University and at the Duke Global Health Institute, and conference participants at NEUDC (2017), at APPAM (2017), at the Annual Congress EEA/ESEM (2017), at the Annual International Conference of the Research Group on Development Economics (University of Gottingen, Germany, 2017) and at DEVPEC (2017) for thoughtful discussions and comments. I gratefully acknowledge financial support from the International Growth Center, Duke Global Health Institute, and Duke Sanford School of Public Policy.

    View full text