Research

Incentives, Audits, and Procurement: Evidence from a District-Level Field Experiment in Ghana. (With Elaine Denny, Ngoc Phan, and Erik Wibbels)

Abstract. There is growing affinity for audits as a tool to promote political accountability and reduce corruption. Nevertheless, knowledge about the mechanisms through which audits work remains limited. While most work on audits shows that they can work via citizen sanctions of bad performers, we emphasize that audit effects can also run through prospective incentives, i.e., the desire to avoid poor audit results in the first place. We distinguish audits’ impact on prospective incentives and sanctions using a field experiment in Ghana; districts were randomized into audit treatment conditions targeting district procurement and oversight of development projects. We assess the effect of audits on political officials using survey experimental data and show that officials respond powerfully to prospective incentives. In districts treated with top-down audits, in-party favoritism falls from 60 percent at baseline to 20 percent at midline, and rates remain at 19 percent at endline. This suggests that the audit’s main effect occurred before the audit results were made public, and that prospective mechanisms play an important role in audit efficacy.

Forthcoming at Governance.

Bureaucratic Capacity and Favoritism in Public Procurement

Abstract. Government contracts are huge business and, in many countries, are associated with considerable corruption. Much research emphasizes bureaucratic improvements as a means to reduce corruption. In this paper, I draw a sharp distinction between the extent to which a bureaucracy is politically controlled and its technical capacity. I argue that in politically controlled bureaucracies, stronger technical capacity facilitates corruption. In such contexts, more capable bureaucrats utilize their skills to shield favored firms from competition using complex strategies that minimize the risk of detection. I test the argument on a novel dataset of 781,425 municipal contracts in Guatemala and 40,570 firm-politician ties. In line with the argument, I find that more capable bureaucracies increase the likelihood of well-connected firms winning contracts through less competitive processes. This paper delivers important policy lessons, an original, widely applicable, measure of political networks and new insights into the sources of corruption.

R&R at Comparative Political Studies.

Unpacking Bribery: Petty Corruption and Favor Exchanges.

Abstract. The incidence of petty corruption in public service delivery varies greatly across citizens and geography. This paper proposes a novel explanation for citizen engagement in collusive forms of petty corruption. It is rooted in the social context in which citizen-public official interactions take place. I argue that social proximity and network centrality provide the two key enforcement mechanisms that sustain favor exchanges among socially connected individuals. Bribery, as a collusive arrangement between a citizen and a public official, relies on the same enforcement mechanisms. Using an original dataset from a household survey conducted in Guatemala, the analysis shows that social proximity and centrality allow citizens to obtain privileges through implicit favor exchanges and illicit payments. These effects go beyond simply increasing the frequency of contact with public officials and are not driven by better access to information about the bribery market.

Under review. This paper received the 2019 Outstanding Preliminary Paper Award (Political Science Department, Duke University) and the 2020-2021 Quality of Government Best Paper Award (Quality of Government Institute, University of Gothenburg)

Bribery, Service Quality and Attitudes Towards Government

Abstract. Citizens’ exposure to corruption in public service delivery has hitherto been associated with negative attitudes towards government. This paper argues that the two components of corruption experiences, namely whether a payment is required and whether an illegal advantage is granted, affect citizen’s perceptions of state capacity in different ways. The act of paying a public official informs a citizen of the state’s inability to prevent its workers from engaging in corruption. In contrast, the experience of obtaining illegal advantages from public officials informs a citizen of the state’s ability to provide expedited service delivery. To test the implications of this argument, I rely on data from a large household survey conducted in Guatemala. The findings show that, experiencing extortion by public officials has a negative effect on individuals’ perceptions of the government’s capacity to provide services. Furthermore, exposure to illegal advantages obtained through favor exchanges positively impacts perceptions of state capacity, while bribery, defined as an exchange of money for access to illegal advantages, has no effect on such perceptions.

Information, Civil Society and Corruption: Evidence from a Randomized Control Trial in Guatemala.

Abstract. In the past 10 years, corruption scandals, particularly those related to public procurement, have caused the removal of presidents in Brazil, Guatemala and Perú. Ongoing corruption scandals continue to rattle civic space across many countries. Considerable programming and research has focused on disseminating information about government malfeasance with an eye toward promoting transparency. The efficacy of such information likely varies depending on the way in which it is delivered as well as its intended audience. We propose to examine alternative ways to disseminate information on government malfeasance in public procurement in Guatemala’s municipalities. In particular we ask: Is it more effective to provide information on past corruption to citizens themselves or the bureaucrats who are directly engaged in procurement? To answer this question, we are in the process of implementing an information intervention with two treatment arms directed to local civil society and municipal bureaucrats, respectively, in collaboration with the Guatemalan think tank ASIES. Randomization was conducted by first matching 141 municipalities into sets of triplets, and then assigning program condition randomly within each matched triplet. Individuals within municipalities assigned to either treatment will receive information about the incidence of single bidding (a popular measure of risk of corruption in public procurement) and political favoritism in the allocation of municipal contacts, and how it compares to the national average. We will measure changes in procurement practices through surveys of civil society leaders and municipal bureaucrats as well as by monitoring procurement outcomes scraped from Guatecompras, Guatemala’s e-procurement system. Data collection will conclude on September 30, 2023. We expect that information given to local leaders will decrease the incidence of political favoritism in public procurement and the number of contracts allocated, and should have a stronger impact in more competitive municipalities. This is so because civil society leaders that receive the information treatment will increase their oversight and political participation, putting pressure on incumbent mayors to change their procurement practices. On the other hand, we expect information to have no effect on public procurement outcomes when given to municipal bureaucrats. Since hiring at the municipal level is highly clientelistic, bureaucrats are better off following the guidelines set by politicians regarding which contractors to select.

Pre-Analysis Plan.

Paying the dues? Access, Congestion and Bribery. (With Serkant Adiguzel and Marco Morucci)

Abstract. Bribery in public service delivery, regardless of its welfare consequences, is a fact of life for citizens in many developing countries. The existing literature on bribery and corruption has argued that citizens with low access to public services are more likely to pay bribes to make up for their lack of access. We argue that sometimes the opposite might be true, with individuals that have better access to public services being more likely to engage in corrupt exchanges with public officials, both because they are socially closer to the public officials, and because their baseline cost for accessing the public service is lower. Using administrative and survey data from Guatemala, we show that individuals that have easier access to public services are more likely to engage in bribery in several ways, as well as more willing to pay higher bribes, and less likely to report public officials for corrupt behavior. Our results imply that policy efforts to improve access to public services in developing countries might have the unexpected negative effect of increasing corruption if they are not accompanied by civil service reform.

High-Frequency Evidence on Corruption in 53 Countries: New Data from the MLP Project. (With Jitender Swami and Erik Wibbels)

Abstract. Corruption is associated with everything from poor governance outcomes and low-quality public services to poor economic growth and political instability. Corruption can have negative effects on civic space as it reduces political participation and generalized trust. However, corruption scandals have also motivated powerful protest movements and civic action in countries as diverse as Brazil, Guatemala and Moldova in recent years. Despite the centrality of corruption scandals to civic space dynamics in many countries, our capacity to understand when, why and where corruption elicits civic responses is sharply limited by constraints inherent in standard corruption data. Most such data, whether in the form of expert (e.g., Transparency International's Corruption Perceptions Index) or citizen (e.g., Afrobarometer) surveys, provide annual snapshots that preclude answering key questions such as: When do corruption scandals evoke protests, civic activism, legal changes and/or the collapse of governments? And what are the implications of corruption for the evolution of civic space more generally? In this project we introduce a new big data approach to measuring corruption that allows researchers and analysts to address these kinds of questions. Our corruption measure relies on the Machine Learning for Peace project's infrastructure, which has collected and classified over 90 millions articles published by international and local newspapers every day from 2012 until last month for nearly 60 countries. By measuring the share of monthly news reporting on corruption, we provide data on its salience. This measure does a good job of identifying corruption scandals and provides a tool for monitoring corruption in near real-time. To elucidate one potential use of the data, we analyze the relationship between corruption scandals and anti-corruption protests across 53 countries. We complement the cross-national evidence with case studies of Guatemala and Ghana, two cases with relatively high incidences of corruption, but where civic responses have varied a lot.

Autocratization and Media Responses to Government Repression of Journalism: Machine Learning Evidence from Tanzania. (With Serkant Adiguzel and Erik Wibbels)

Abstract. One crucial feature of the ongoing global wave of democratic backsliding is that aspiring autocrats seek to influence the media, oftentimes through legal restrictions on the press and social media. Yet little research has examined how formal and social media respond to those legal restrictions targeting the free flow of information. We develop an original argument linking key characteristics of media sources to the regulatory environment and examine how the content and sentiment of their coverage responds to restrictive media laws. We test our claims using an enormous corpus of electronic media in Tanzania and employ two state-of-the-art neural network models to classify the topics and sentiment of news stories. We then estimate diff-in-diff models exploiting a significant legal change that targeted media houses. We find that critical news sources censor the tone of their coverage, even as they continue to cover the same issues; we also find that international news sources are unable to fill the hole left by a critical domestic press. The paper sheds light on the conditions under which the press can be resilient in the face of legal threats.

Under review.

The Impact of Legal Repression on Citizen Online Behavior: Evidence from Tanzania's Jamii Forums. (With Serkant Adiguzel and Erik Wibbels)

Abstract. How does the restrictive legislation on online behavior impact citizen posting on the web? Although many welcomed the internet as a liberating technology, especially after the Arab Spring, governments quickly adopted various surveillance and repression strategies to monitor, regulate, and steer online citizen behavior. Given the global rise of informational autocrats it is critical to understand the impact of such legal repression on citizens' online behavior. Previous research has shown that the internet, particularly social media, has a significant impact on political participation in the form of voting behavior and protests. While some research has examined how repressive regimes use propaganda and repress information systems, there is little to no evidence on the impact of legal repression on citizens' online behavior. We provide such evidence from the context of Tanzania, where we analyze the impact of the 2015 Cybercrimes Law on posts on political threads in Jamii Forums –a widely used, citizen-driven online platform in the country. Our initial analysis of more than 11 million individual posts reveals four key findings. First, there is increased activity on political threads during the period leading up to the Cybercrimes Act going into full effect (September 2015). Second, the number of new and inactive individual accounts increases dramatically during the period between the parliamentary approval of the Cybercrimes Act (April 2015) and when it went into full effect. One possibility is that this indicates citizens using new accounts to hide their identities. Third, consistent with that possibility, new accounts that are opened in the lead up to the law's enforcement have the same sentiment towards government actors relative to accounts that go inactive. This suggests that the law failed to reduce the incidence of critical posting against the government by citizens. Fourth, there is also a spike in inactive accounts after the law goes into effect. This spike does not correspond with a wave of new accounts and more likely reflects citizens withdrawing from online posting.

Machine Learning for Peace: Civic Insights from New, High-Frequency Data. (With Serkant Adiguzel, Jeremy Springman, Donald Moratz, Mateo Villamizar Chaparro and Erik Wibbels)

Abstract. Democratic rights are expanding and contracting on a daily basis as democratic and autocratic forces contest over fundamental rights. This paper introduces the `Machine Learning for Peace' (MLP) project, which provides high-frequency data on many features of day-to-day `civic space'. To do so, MLP applies machine learning to classify an unprecedented corpus of digital news every day from 2012 to the present for 47 developing countries into 20 distinct civic space event types. These data allow researchers and practitioners to track trends in civic space at high frequency and allow us to build predictive models to inform policy and practice.

Fake Media, Corruption and Democratic Backsliding in El Salvador. (With Erik Wibbels and Ernesto Calvo)

Abstract. Nayib Bukele came to power in 2019, presenting himself as a political outsider –despite having been elected mayor with FMLN before temporarily switching to GANA for the 2019 presidential elections— promising to fight corruption and to implement an ambitious infrastructure plan. While many important infrastructure projects have been completed since Bukele came to power, he has captured the institutions of justice, harassed the press and even attempted a coup on February 9th, 2020. In the independent media, there are many reports of corruption in the government’s infrastructure projects, but contrary to his campaign promises Bukele has rejected attempts to set up an independent body to investigate and prosecute corruption. Indeed, Bukele broke his 2019 agreement with the Organization of American States to set up a commission—CICIES—tasked with helping local prosecutors investigate corruption cases once those prosecutors began targeting members of his government. Instead of prosecuting corruption, Bukele has gone after journalists for their coverage of corruption scandals in effort to silence their work. His government has started investigations against elFaro for alleged money laundering, and he regularly criticizes established print newspapers (El Diario de Hoy and La Prensa Gráfica) as well as digital outlets (Revista Factum and Gato Encerrado). A key feature of the success of Bukele’s “Millennial Authoritarianism” (Meléndez-Sánchez, 2020) is his manipulation of the media environment through the effective use of social media to both promote his image and spread positive, fake news. Bukele’s own popularity started on TikTok, and his campaign relied heavily on social media to promote his image. Once in power, Bukele has continued to rely on social media to control the information environment in two different ways: First, through the use of bots and fake accounts to combat criticism in social media; and second, through the dissemination of fake news from newly created newspapers such as the official Diario El Salvador through social media platforms, particularly Facebook and Twitter. In this project, we explore whether and how these social media-based strategies work, i.e. if, when and how they crowd out serious reporting on corruption in El Salvador. In order to accomplish this, we will scrape articles from propaganda outlets and compare their content to that of traditional outlets; we will focus specifically on corruption, but we will also compare how the two types of media cover the full range of civic space, including the 18 civic space event types the Machine Learning for Peace project is already collecting data on. First, we will focus on content and relative coverage, i.e. whether a given outlet publishes articles about government corruption and the share of its daily, weekly and monthly coverage devoted to those stories. To study whether articles published by propaganda outlets actually crowd out the information environment, however, we proposed to rely on data from Facebook and Twitter. We will measure the number of times articles are shared/read to assess the degree to which: (a) propaganda travels across information networks—one important output of the project will be to map the social media network surrounding the president and his supporters; and (b) propaganda actually crowds out real news. Second, we will study Twitter activity –given the fact that Bukele himself is a big user— to see whether fake news and content generated by propaganda outlets is deployed by Bukele-aligned Twitter accounts around corruption scandals as one more tool to change the narrative. To accomplish this, we have partnered with a group of journalists from Revista Gato Encerrado, an online investigative journalism outlet from El Salvador.

Economic Insecurity and Deportees’ Decision to Re-migrate in a COVID-19 Era. (With Elaine Denny, David Dow and Erik Wibbels)

Abstract. For potential immigrants, the COVID-19 pandemic has reduced economic opportunities and increased risks both at home and abroad. We seek to understand how COVID-19 has impacted the calculations that govern one's decision to emigrate. Leveraging a unique panel survey of Guatemalans recently deported from the United States, we explore how COVID-19 has affected deportees' economic well-being and the intent to re-migrate. We find that while COVID-19 does not measurably decrease deportees' (already poor) current economic conditions, the pandemic increases expectations of bein g worse off in the next year and uncertainty about future economic conditions. Furthermore, the pandemic also increases uncertainty about whether deportees intend to re-migrate in the coming year. This increase in uncertainty reflects the increased difficulty potential migrants face in weighing relative opportunities and risks during a transnational crisis, even as one's expectations about economic well-being in the home country become more pessimistic.

Journal of Political Institutions and Political Economy.

The Human Impact of Deportation: Resettlement and Remigration. (With Elaine Denny, David Dow, Gabriella Levy, Wayne Pitts, Juan F. Tellez, Mateo Villamizar, Erik Wibbels, and Pamela Zabala)

Abstract. What determines the resettlement and remigration preferences and decisions of deportees? We argue that returnees seek locations in their country of origin which can provide them with physical safety, employment, and social and familial ties. When returnees are unable to achieve these goals within their country of origin, they look to remigrate. Experimental and observational evidence from an original survey of over 1,300 deportees from the United States to Guatemala suggests that deportees are focused on security, jobs, and social networks when deciding where to go within Guatemala. But, in follow up surveys, fewer than half are employed, and many are living in considerably insecure locations. Indeed, unstable economic and security conditions in Guatemala compared to the United States are key determinants of remigration preferences. These findings provide concrete evidence that deportation does not successfully halt migration in the absence of improvement in conditions in home countries.

The Market in Smugglers: Survey Experimental Evidence on the Choice of Coyotes in Guatemala. (With Mateo Villamizar and Erik Wibbels)

Abstract. Illicit migration is big business. Each year it involves millions of potential migrants, many of whom hire human smugglers to facilitate perilous journeys. Yet very little is known about this illicit market or how migrants choose who to hire. Faced with very high costs and uncertain prospects of making it to their target destination, migrants face a high-stakes, complex choice when identifying a smuggler to work with. In this paper, we provide descriptive evidence on how the market for coyotes (i.e. human smugglers in Latin America) works and survey experimental evidence on the key dimensions impacting the choice of which coyote to hire. Our forced choice experiment allows us to isolate the impact of referrals, reputation, reliability, safety, and price on this high-stakes choice. Our evidence draws on two original sources: a panel survey of deportees from the U.S. to Guatemala and a recent household survey (n=1,500). We find that referrals and a reputation for successful, safe journeys are key drivers when choosing a smuggler. We also find that respondents are unresponsive to price, which provides more solid micro-foundations for recent findings elsewhere that hardening borders raises the cost of migration but fails to reduce its incidence.

State Absence, Vengeance, and the Logic of Vigilantism in Guatemala. (With David Dow, Gabriella Levy and Juan F. Tellez)

Abstract. Across the world, citizens sidestep the state to punish offenses on their own. Such vigilantism can help communities provide order, yet it raises concerns about public accountability and the rights of the accused. While prior research has identified the structural correlates of vigilantism, an open question is in which scenarios citizens prefer vigilantism over conventional policing. To make sense of these preferences, we delineate two potential logics of vigilantism -state substitution and retribution- by drawing on research in criminology and psychology on punishment motives. Using survey data from over 9,000 households across Guatemala, we provide estimates of the effect of varying crime scenarios on people's preference for vigilantism over policing. The results highlight how support for vigilantism can vary substantially across crimes, suggest that a desire for retribution is a strong determinant of attitudes towards vigilantism, and ultimately raise concerns about the viability of `informal' forms of policing.

Forthcoming at Comparative Political Studies.

Public Services and the Geography of Citizen Perceptions of Government in Latin America and the Caribbean. (With Gary Bland, Derick Brinkerhoff, Anna Wetterberg, and Erik Wibbels)

Abstract. One of the linchpins of democratic accountability runs through service quality. Because citizens are sensitive to the quality of basic services, they can translate (dis)satisfaction into assessments of incumbent politicians. Yet, although previous research has shown that both access to, and the quality of, basic services decline in rural settings, this seems not to translate into increased dissatisfaction with incumbents. In this paper we seek to understand why. We theorize four potential mechanisms that might underpin the weaker accountability for poor service outcomes in more remote settings. To test these mechanisms, we use data from 34,514 geocoded survey respondents across 19 countries in Latin America. We show that the likelihood of translating dissatisfaction with services into discontent with elected officials decreases as distance to urban centers increases. We find some evidence that a low sense of political efficacy and deference to hierarchy mediate the relationship between remoteness, service quality and accountability. Nevertheless, some of the direct relationship between distance and attitudes towards elected officials persists in the face of our mediation analysis, suggesting that more work needs to be done on the relationship between remoteness, service quality and accountability.

Political Behavior.

Trust and Public Institutions: Reporting of Gender-based Violence in Guatemala. (With David Dow and Livia Schubiger)

Abstract. Gender-based violence remains one of the most pressing human rights issues of our time, and accountability for these crimes remains weak. Gender-based violence is also starkly under-reported. Possible reasons for under-reporting include social norms, psychological costs, and a lack of trust in, as well as limited access to, relevant institutions. How can the issue of under-reporting be addressed? We investigate whether the expansion of specific offices and agencies belonging to the Ministerio Público (General Attorney's office) increased the reporting of gender-based violence in Guatemala. These recently expanded institutions are comparatively well-trusted points of access for reporting crimes, and thus enable us to study the impact of policy efforts to mitigate under-reporting, one of the enduring challenges for prosecuting gender-based violence.