On Anti-Corruption Tools in Developing Countries

(1)

On Anti-Corruption Tools in Developing Countries

Arieda Muço

(2)

i

Dissertation for the Degree of Doctor of Philosophy, Ph.D., in Economics. Stockholm School of Economics, 2017

On Anti-Corruption Tools in Developing Countries

ISBN 978-91-7731-063-1 (pdf)

Front cover illustration:

Python cloud. Mostly common words used in the abstracts of the papers in this thesis, plotted on the map of Brazil.

Printed by:

Ineko, Gothenburg, 2017

Keywords:

Audit Program, Corruption, Complaint, Local and National government, Neighbouring, Party, Politician, Principal Component Analysis, Machine Learning, Media Market, Whistleblowing

(3)

ii Doctoral Dissertation

Department of Economics Stockholm School of Economics

Abstract

Learn from thy Neighbor: Do Voters Associate Corruption with Political Parties? This paper exploits a randomized audit program to document how information about a corrupt politician causes electoral spillovers on his party. I focus on two types of spillovers: spillovers across types of elections (cross-electoral spillovers) and spillovers across jurisdiction borders (cross-border spillovers). Using detailed data on radio antenna location and coverage, I identify neighboring areas in the same media market. I uncover how information about corruption shapes voting decisions through the struc- ture and geography of the media market. I show that voters hold the party of the incumbent politician accountable in four distinct ways. In municipalities where corruption occurs, voters punish parties in (1) local and (2) national elections. Most importantly, I show that news of a politician’s corruption affects his party in neighboring municipalities that share the same media market, and these spillovers affect both (3) local and (4) national elections.

Ruling out other potential mechanisms, I show that these findings are con- sistent with electoral accountability.

Corruption, Intimidation and Whistleblowing: An Empirical Ap- proach. I exploit an anti-corruption program in Brazilian local governments to investigate the reported wrongdoing by whistleblowers. The program consists of frequent random audits. I argue that the program indirectly improves whistleblower protection against potential retaliation. I show that after an audit is announced, reported wrongdoing increases. The likelihood of placing

(4)

iii a complaint peaks the month of the audit announcement. Areas with less social capital and those where higher corruption prevails are more likely to file in a complaint if an audit is announced. These findings have important policy implications and contribute to the growing literature on incentives and motivations of whistleblowers.

A Novel Approach to Measuring Corruption using Machine Learn- ing Tools. This paper proposes a data-driven approach to create an index that ranks municipalities according to their corruption level. From text data, I extract several features that point at irregularities implementing Machine Learning and Text-Analysis tools. To classify corruption, I use an unsupervised learning technique, namely a principal component analysis (PCA). The use of the PCA has two rationales: i) it is a widely recognized tool for dimensionality reduction, ii) the different features extracted from the text data reflect different shapes of corruption. Thus, latent common factor should capture the underlying corruption. Finally, I test the performance of the al- gorithm with hand-coded measures already available in the literature.

(5)

iv

To Antonio and Reno

(6)

v

(7)

vi

...

We say “corruption chokes”, Then why do we breathe it?

We say “corruption kills”

Then why do we breed it?

...

from Hiral Soni

(8)

vii

(9)

viii

Foreward

This volume is the result of a research project carried out at the Depart- ment of Economics at the Stockholm School of Economics (SSE).

This volume is submitted as a doctoral thesis at SSE. In keeping with the policies of SSE, the author has been entirely free to conduct and present her research in the manner of her choosing as an expression of her own ideas.

SSE is grateful for the financial support provided by the Jan Wallander and Tom Hedelius Foundation, which has made it possible to carry out the project.

Göran Lindqvist Director of Research

Stockholm School of Economicss

Tore Ellingsen, Professor

Head of the Department of Economics Stockholm School of Economics

(10)

ix

(11)

x

Acknowledgments

The Ph.D. has been an incredible journey. During these six years, I have learned a great deal, academically and not. I have also learned to fail. Through- out these years, I have been immensly lucky to have met great people and researchers who taught me a lot. All of them, in one way or another, helped me get back on my feet and deserve a special mention in this section of my Ph.D. thesis.¹

First of all, I would like to thank my advisors Erik Lindqvist and David Strömberg. I would also like to thank my “other advisors" and members of my Ph.D. committee: Diego Puga and Giancarlo Spagnolo. These amazing researchers have all contributed in shaping me and my research in different ways.

I will always be thankful to Erik for agreeing to be my main advisor even though my research interests were slightly far from his. He has always been extremely supportive of me and my research ever since. During one of our first meetings, Erik told me that those that go the extra mile make the difference. This piece of advice stuck with me. So have many other pieces of precious advice Erik has given me throughout these years. He is a very careful and detail-oriented researcher, traits which made the feedback I received to be very thorough, for projects included in this thesis, those that were not, or just pieces of ideas (logical, incoherent, or disorganized) which Erik patiently helped me outline and think more clearly. I have always been very at ease talking with Erik, sometimes it almost felt like talking to a friend. I will miss our long conversations about research and non-strictly-research related topics.

In the second year of my Ph.D., I was exposed to the courses of Political Economy co-taught by David Strömberg and Torsten Persson. These courses set me on a path that I am happy to walk. David’s work on Media and

1Please note that while I have attempted to thank everyone (my) words do not make justice of the deep gratefulness that I have for everyone mentioned in this session.

(12)

xi Political Economy has had enormous influence on me. This thesis fully reflects that. I am thankful to David for his generous advice, great tips, and for setting the bar, always, very high. During our meetings, David tirelessly asked questions about each idea or result I brought to him. His disposition to see both the big and the small picture helped me to get the best out of every project. I am also very grateful for his constant support, including the job market experience. I will, for sure, miss our lunch meetings.

I spent my third and fourth year of the Ph.D. at another wonderful insti- tution, at the Center for Monetary and Financial Studies (Cemfi), in Madrid.

At Cemfi, I met with Diego Puga who became another member of my Ph.D.

committee and supervisor during my stay at Cemfi. Diego, with his ability to see feasibility and potential of each project helped discard the bad ideas.

This thesis would have been very different without Diego’s constant input and support.

The Ph.D. journey would not have started without Giancarlo Spagnolo.

Giancarlo accepted to be my master thesis supervisor in Rome and since then we have gone a long way. Today, Giancarlo is not only a supervisor but also a co-author. My research has moved away from the fields of IO and Procurement which I first came to pursue in Stockholm. However, we still have many overlapping research interests and I continue to turn to him for advice.

I am also deeply grateful to Manuel Arellano, Tore Ellingsen, and Torsten Persson, which have also been very generous with their time and feedback. I feel very lucky to have been able to interact with these wonderful researchers during my Ph.D. years.

Tore is a research enthusiastic which makes any research related discussion very easy. He has a great intuition and always updated about the research frontier. Tore, always has spot on advice where one should focus his/her energies and gives the right feedback regardless of the research field. I was no exception.

(13)

xii Torsten has also been another source of great influence and inspiration throughout these years. Since the second year, we discussed research during class breaks or after class. Torsten always found the time to listen to my ideas, shaped them, and always suggested relevant literature. Torsten’s ability to recollect important literature, facts, and events never ceases to fascinate me.

I am extremely thankful for his encouragement and help to pursue interesting questions in the field.

Manolo opened to me Cemfi’s doors and mentored me during the first months of my visit there. Throughout my stay at Cemfi I always felt welcome to talk with Manolo. This was also very easy given that his office door was always open (literally). I have learned a great deal from the interactions with Manolo and I treasure his advice.

I would also like to thank Anna Dreber Almenberg, Anna Tompsett, An- dreas Madestam, Johanna Wallenius, Kerem Coşar and Martina Björkman Nyqvist who have all provided extensive feedback during my Ph.D. years and especially during the job market experience. Rickard Sandberg for helping me during the first years of the Ph.D., Mark Voorneveld and Jörgen Weibull for organizing the researcher’s lunch.

With special thanks to Mónica Martínez-Bravo, Guillermo Caruana, Samuel Bentolila for providing extensive feedback during my stay at Cemfi. This list is not complete without the addition of Rosario Crinò and Claudio Michelacci, Guillermo Caruana and Josep Pijoan-Mas who, at different time periods, lead the reading group at Cemfi, taught us how to present, discuss, and constructively criticize papers on the current research frontier.

Stockholm is a cluster of great institutions. SITE is one of those and is located just one floor above the department of Economics at SSE. This gave me the possibility to frequently interact and befriend Elena Paltseva and Maria Perotta-Berlin. I also benefited from conversations with other researchers at SITE such as Erik Meyersson, Cloé le Choq, Anders Olofsgård, and Jesper Roine.

(14)

xiii Gustavo Piga and Stefano Gagliarducci have both been great sources of inspiration during my bachelors and masters in Rome. I thank Gustavo for teaching economics with a great passion. If I wouldn’t have had a great teacher such as Gustavo during my BA courses, my path would have been a different one. Stefano was my master thesis co-supervisor. Not only is he a passionate researcher and teacher, but during the master thesis, he gave me a “trial version” of what applied research would look like with it’s ups and downs. I thank Stefano greatly for always finding time to discuss research and provide broad feedback throughout these years, in Rome, Madrid, or over Skype.

Stockholm also gave me many good friends. Luca Facchinello, is one of them. Luca was not only my colleague during the Ph.D. years but also a flatmate for several months. We realized that we also share a lot of common interests (music, research, books, tv-series...) during our long conversations, while cooking, having a coffee, a meal, and the long hours it took to assemble my IKEA furniture.

At Cemfi I shared a common Ph.D. room with many great people. I would especially like to thank Alessandro, Big Julio, Elena, Felipe, Gustavo, Ines, Jesus, Luca, Lucciano, Little Julio, and Sandra. Thank you all for being passionate about research and fun to have around. I cannot think of better people to share office space.

My job market buddies Erik and Nathan. Thanks for sharing your desks and the unique job market experience and making it fun and enjoyable.

Mounir Karadja for always finding time to discuss research and not only. I am looking forward to seeing our project finished.

The dark days in Stockholm were brighter thanks to the girl power of Audi (the other Italian speaking but non-Italian), Amanda, Evelina, Eleonora, Marta, Nadiia, and Pamela. My first two years would have been less bearable without the (frequent) knocking on the door of the office where Anna Maria, Paola, and Spyros where sitting.

(15)

xiv My latest office mates Andreas and Bengt who made our office a great place to work. My cohort at SSE: Erik, Evelina, Niklas, Jun, Markus. I also learned a lot from both Andrea (the Italians), Aljoscha, Benedetta, Benni, Christopher, Domenico, Elle, Max V., Roman, Jose, Julian, Karl, Miri, Nana, Saman, Seryashi, Sirus.

To my family: my mother Eli, father Duka, and mostly my brother Reno.

I took the decision to pursue the Ph.D. in Europe mainly because I wanted to stay close to you. However, during these years, I never managed to be back home as much as I wanted to. None of you ever complained about this.

I dedicate to you this achievement as I dedicate to you all achievements in my life. Reno, je dhurata më e bukur që më është bërë në jetë. Te dua shumë!

To Antonio: Thank you for your constant love and support. For always pushing me to be a better person, researcher, and programmer. You are the first one to hear my joys, achievements, difficulties, and setbacks and you always have the words of wisdom. Thank you for never holding me back, for helping me to pursue my dreams and giving me the courage that I sometimes lack. Above all, thank you for not hesitating to start a new life across Europe:

Italy, Spain, Sweden and in this new adventure in Hungary. I love you and I am looking forward to this new chapter with you.

(16)

Chapter 1 Introduction

How can we fight corruption? This is among the oldest question in economic literature with seminal contributions by Rose-Ackerman (1975) to Shleifer and Vishny (1993). The first generation of anti-corruption studies and it’s implication on economic outcomes has focused on cross-country comparisons, mostly using corruption perception indices as a proxy for real levels of corruption.¹ The second generation of corruption studies, instead, uses microe- conometrics techniques and lab experiments to analyze corruption at the micro-level (individual, household, firm).²

This thesis belongs to the latter approach and consist of three self-contained papers each tackling a different question in the anti-corruption literature.

While Chapter 2 and Chapter 3 answer questions directly related to anti- corruption tools and derive clear policy implications, Chapter 4 is method- ological paper which uses machine learning and text analysis to construct a corruption index which is used by both previous chapters as a means to deeper understand the mechanisms of each question posed by the previous two.

1Such as Treisman (2000), Wei (1997), Fisman and Gatti (1999), Persson and Tabellini (2005).

2Some examples are Fisman and Miguel (2007), Olken and Barron (2009), Fisman and Wei (2004), Olken (2007), Hsieh and Moretti (2006), Reinikka and Svensson (2004).

1

(19)

Introduction 2 All chapters in this thesis deal with tools to fight malfeasance using the same unique policy program, the random audits in Brazilian local governments which was initiated in early 2000 by the Brazilian federal government with the objective of fighting corruption in Brazilian municipalities.

The program – which continues to date – is implemented by the Contro- ladoria Geral da Uniaõ (CGU) and consists of random selection, for audit purposes, of Brazilian local governments. Once a municipality is selected for audit purposes, a team of auditors shows up in the municipality and inves- tigates transfers allocated by the federal government to each municipality.

After several days of investigation in the municipality, the auditors compile a detailed report of their findings which is made available for download on the CGU website.

In Chapter 2 I ask whether providing information about the corruption level of a local politician effects, across different dimensions, the support of the political party the politician is associated with. News about the corruption level of the local politician is provided to the voters via the audit program. In this chapter, I show that several electoral spillovers emerge. First, I document cross-electoral spillovers, which I define as spillovers from local elections to national elections, such as presidential elections. Second, I document cross-border spillovers, which I define as spillovers from a politician to a local government in a neighboring area, not only in local elections but also presidential elections. Taken together, findings in this chapter show that electoral punishment is higher if more corruption is revealed.

Findings emerging from this paper are important for several reasons: i) These results are relevant towards better understanding the role of political parties as controlling authorities. ii) Moreover, these result can also be interpreted as a direct test of theory by Tirole (1996) which suggest that the reputation of the organization is affected if one of its members’ misbe- haves. ii) From a policy perspective, these findings are also relevant as they suggest that electoral accountability might be an important anti-corruption

(20)

Introduction 3 tool. This is mainly important in developing countries where corruption is widespread and electoral punishment is the highest form of punishment that corrupt individuals and parties face. Thus, providing information to voters, which they in turn use to oust parties associated with corrupt politicians, is another step towards ensuring better electoral accountability.

Naturally, this chapter raises other interesting questions to be answered which are part of my future research agenda. The first and the most natural question that comes to mind is related to if and how the political parties themselves incorporate these electoral spillovers. If it is true that corrupt individuals affect negatively party’s reputation, we expect that in the short run parties run cleaner candidates. Preliminary findings suggest that parties associated with corrupt candidates are more likely to run "cleaner" candidates i.e more female politicians are likely to enter the politics and more educated candidates.³ Another research line would be to investigate what happens to corrupt individuals after these corruption scandals emerge. Are these individuals are more likely to be out of the political arena? Are they more or less likely to stick to their own party? Is their future income affected as a result of the corruption scandals?

In Chapter 3 of this dissertation, I move to another important question in the anti-corruption literature: "How do we motivate whistleblowers?". En- couraging individuals to report wrongdoing ("or blow the whistle"), and pro- tecting them when they do, is an important part of the anti-corruption fight in both the public and private organizations (OECD (2016)).⁴ Employees are usually the first to recognize wrongdoing in the workplace, so empowering them to speak up without fear of reprisal can help authorities both detect

3Education is perceived as a signal of quality. For example, Besley, Montalvo and Reynal-Querol (2011) show having a more educated leader affects the rate of economic growth. From the economics and the political science literature, we know that female candidates are not only perceived more trustworthy but also shown to be less corrupt than men, i.e Dollar, Fisman and Gatti (2001) or Brollo and Troiano (2016).

4The report can be found here: http://www.oecd.org/corruption/committing-to- effective-whistleblower-protection-9789264252639-en.htm.

(21)

Introduction 4 and deter violations.⁵

Blowing the whistle bears high costs, moral, financial as well as endangers the career and the life of whistleblowers and their family members. Thus, to providing whistleblowers with the right incentives, several hypotheses have been brought forward. For example, Zingales (2004) and Butler, Serra and Spagnolo (2017) suggest that monetary incentives should be provided. Chas- sang and Pardo i Miguel (2014) suggest that whistleblowers are more likely to blow the whistle if they receive protection and this protection must come in the form of plausible deniability. Despite this important debate in the literature and among policy members, empirical evidence of the effects of policies on whistleblower motivation is scarce.⁶

I credibly show that when provided with plausible deniability, whistleblowers are more likely to report wrongdoing. To do so, I use the audit program and argue that the random audits serve as a means to provide plausible deniability to whistleblowers. Indeed, after an audit is announced in a municipality, reported wrongdoing in that municipality increases. The likelihood of placing a complaint peaks at the month after the audit announcement.

Furthermore, I show that areas, where the underlying corruption is high and less social capital, are more likely to file a complaint if an audit is announced. These findings have important policy implications and contribute to the growing literature on the incentives and motivations of whistleblowers.

In Chapter 4 I propose a novel methodology to measure corruption using a data-driven approach. I use the same policy experiment as in in the previous two chapters and create a corruption index that ranks Brazilian municipalities on the severity of infractions committed. To do so, I use the reports that the auditors write after each inspection. Previous studies before this have

5In Article 33 of the UNCAC, the UN encourages signatory countries to incorporate into their domestic legal systems measures to protect whistleblowers and witnesses from any unjustified treatment.

6There is, however, some experimental evidence such as Butler, Serra and Spagnolo (2017) among others.

(22)

Introduction 5 hand-coded audit reports to create corruption measures.⁷ Differently from previous literature I first automatize the procedure of extracting the text from the audit reports. Moreover, I use unsupervised learning techniques to reduces the high dimensionality of text data and create the corruption index.

I test the performance of my procedure with hand-coded measures already available in the literature. The corruption index is highly correlated with the hand-coded measures but it has several advantages over those: i) it does not rely on subjective decision-making, ii) it’s replicable, and iii) cost effective.

In the same spirit as recent economic literature, such as Olken (2007), Hsieh and Moretti (2006), Reinikka and Svensson (2004), Fisman and Miguel (2007), Olken and Barron (2009), Fisman and Wei (2004), Di Tella and Schar- grodsky (2003), and others, this thesis shows that audits are an important tool to fight wrongdoing.⁸

This thesis takes a step forward, with respect to current literature, show- ing that audits not only generate spillovers that should be taken into account when considering cost-benefit analysis of such a costly anti-corruption policy, but they go further. First, they provide information to voters about the individuals and parties that govern them both at the local level and higher levels of government. Second, this thesis showed that random audits can also be seen as a tool to provide plausible deniability to whistleblowers.

While policy implications brought forward by each of these separate papers arise from empirical evidence focused on a specific country, Brazil, using a specific experiment there are important lessons to be learned on anti- corruption measures. After all, some facets of corruption are the same in developing and developed countries alike.

7See for example Ferraz and Finan (2008, 2011); Brollo et al. (2013); Larreguy, Marshall and Snyder Jr (2014); Bobonis, Fuertes and Schwabe (2016).

8See Zitzewitz (2012) and Olken and Pande (2012) for a literature review.

(23)

BIBLIOGRAPHY 6

Bibliography

Besley, Timothy, Jose G Montalvo, and Marta Reynal-Querol. 2011.

“Do educated leaders matter?” The Economic Journal, 121(554).

Bobonis, Gustavo J, Luis Cámara R Fuertes, and Rainer Schwabe.

2016. “Monitoring Corruptible Politicians.” The American Economic Re- view, 106(8): 2371–2405.

Brollo, Fernanda, and Ugo Troiano. 2016. “What happens when a woman wins an election? Evidence from close races in Brazil.” Journal of Development Economics, 122: 28–45.

Brollo, Fernanda, Tommaso Nannicini, Roberto Perotti, and Guido Tabellini. 2013. “The Political Resource Curse.” The American Economic Review, 103(5): 1759–1796.

Butler, Jeff, Danila Serra, and Giancarlo Spagnolo. 2017. “Motivating Whistleblowers.”

Chassang, Sylvain, and Gerard Pardo i Miguel. 2014. “Corruption, Intimidation, and Whistle-blowing: a Theory of Inference from Unverifiable Reports.” National Bureau of Economic Research.

Di Tella, Rafael, and Ernesto Schargrodsky. 2003. “The role of wages and auditing during a crackdown on corruption in the city of Buenos Aires.”

The Journal of Law and Economics, 46(1): 269–292.

Dollar, David, Raymond Fisman, and Roberta Gatti. 2001. “Are women really the “fairer” sex? Corruption and women in government.”

Journal of Economic Behavior & Organization, 46(4): 423–429.

Ferraz, Claudio, and Frederico Finan. 2008. “Exposing Corrupt Politi- cians: The Effects of Brazil’s Publicly Released Audits on Electoral Out- comes.” The Quarterly Journal of Economics, 123(2): 703–745.

(24)

BIBLIOGRAPHY 7 Ferraz, Claudio, and Frederico Finan. 2011. “Electoral Accountability and Corruption: Evidence from the Audits of Local Governments.” Amer- ican Economic Review, 101(4): 1274–1311.

Fisman, Raymond, and Edward Miguel. 2007. “Corruption, norms, and legal enforcement: Evidence from diplomatic parking tickets.” Journal of Political economy, 115(6): 1020–1048.

Fisman, Raymond, and Roberta Gatti. 1999. “Decentralization and corruption: Cross-country and cross-state evidence.” Unpublished manuscript.

Washington: World Bank.

Fisman, Raymond, and Shang-Jin Wei. 2004. “Tax rates and tax eva- sion: evidence from “missing imports” in China.” Journal of political Econ- omy, 112(2): 471–496.

Hsieh, Chang-Tai, and Enrico Moretti. 2006. “Did Iraq cheat the United Nations? Underpricing, bribes, and the oil for food program.” The Quar- terly Journal of Economics, 121(4): 1211–1248.

Larreguy, Horacio A, John Marshall, and James M Snyder Jr. 2014.

“Revealing malfeasance: How local media facilitates electoral sanctioning of mayors in Mexico.”

Olken, Benjamin A. 2007. “Monitoring corruption: evidence from a field experiment in Indonesia.” Journal of political Economy, 115(2): 200–249.

Olken, Benjamin A, and Patrick Barron. 2009. “The simple economics of extortion: evidence from trucking in Aceh.” Journal of Political Econ- omy, 117(3): 417–452.

Olken, Benjamin A, and Rohini Pande. 2012. “Corruption in developing countries.” Annu. Rev. Econ., 4(1): 479–509.

(25)

BIBLIOGRAPHY 8 Persson, Torsten, and Guido Enrico Tabellini. 2005. The economic

effects of constitutions. MIT press.

Reinikka, Ritva, and Jakob Svensson. 2004. “Local capture: evidence from a central government transfer program in Uganda.” The Quarterly Journal of Economics, 119(2): 679–705.

Rose-Ackerman, Susan. 1975. “The economics of corruption.” Journal of public economics, 4(2): 187–203.

Shleifer, Andrei, and Robert W Vishny. 1993. “Corruption.” The quarterly journal of economics, 108(3): 599–617.

Tirole, Jean. 1996. “A theory of collective reputations (with applications to the persistence of corruption and to firm quality).” The Review of Eco- nomic Studies, 63(1): 1–22.

Treisman, Daniel. 2000. “The causes of corruption: a cross-national study.”

Journal of public economics, 76(3): 399–457.

Wei, Shang-Jin. 1997. “Why is corruption so much more taxing than tax?

Arbitrariness kills.” National Bureau of Economic Research.

Zingales, Luigi. 2004. “Want to stop corporate fraud? Pay off those whistleblowers.” Washington Post (outlook section).

Zitzewitz, Eric. 2012. “Forensic economics.” Journal of Economic Litera- ture, 50(3): 731–769.

On Anti-Corruption Tools in Developing Countries