Biased Media or Biased Citizens? How Information Shapes the Perception of Corruption

Biased Media or Biased Citizens? How Information Shapes the Perception of Corruption By Ana María Montoya ana.m.montoya@vanderbilt.edu Diana Orcés diana.m.orces@vanderbilt.edu Vanderbilt University (Please do not circulate) Abstract: What is the relationship between information and perceptions of corruption? The literature on corruption suggests that voters support corrupt politicians when they lack information about their corrupt behavior. However, we know very little about how individuals process information about corruption. In this paper, we explore whether perceptions of corruption are influenced by the media and how this relationship varies across countries. Using data from 23 countries across Latin America and the Caribbean, this study employs a multilevel analysis to understand how information both at the national and individual level matter for the perception of corruption. We find that the more citizens consume news via television, radio, newspapers and internet, the higher their levels of perceived corruption at the individual level. We also find in countries with more press freedom, citizens show lower levels of corruption perception, whereas in countries with less freedom of the press, citizens perceive higher levels of corruption, controlling for other factors. For this research, we rely on survey data provided by the AmericasBarometer and the Reporters Without Borders (RWB). Paper prepared for presentation at the 2014 annual meeting of the Midwest Political Science Association Chicago Illinois April 3-6, 2014 1

Introduction Corruption is one of the most salient and ubiquitous challenges in the developing world, not only because of its well-known pernicious effects on economic growth, but also because it weakens the belief in the legitimacy of the political system (e.g., Mauro 1995, Seyf 2001, Cooper, Krickhaus, Lusztig, 2006; Sandholtz and Koetzle, 2000; Treisman 2000; Morris 1999, Seligson 2002, Anderson et al. 2003, Chang and Chu 2006). Moreover, corruption corrodes the rule of law and weakens democratic institutions by violating important principles of democratic governance such as transparency and equal treatment for all (Douglas, 1990). At the same time, freedom of the press is believed to be one of the most effective institutions to combat corruption in contemporary democracies. The media provides a great amount of information that may influence citizens perceptions of public officials performance and in turn affect their accountability (Bruns and Himmler 2011, Reinikka and Svensson, 2005, Stromberg, 2004, Snyder and Stromberg 2010). In sum, the role of the media is significant for uncovering the misuse of political power 1. This raises important questions: What is the relationship between citizens exposure to the news and their perception of corruption? What is the relationship between the freedom of the press at the national level and citizens perceptions of corruption? There is a growing body of literature based on experiments that establishes causality between information on corruption and how it affects the decision to vote (Ferraz abd Finan, 2008; Banerjee et al. 2010; De la O et al, 2010; Winters and Weitz-Shapiro, 2013; de Figuereido, Hidalgo, and Kasahara, 2011; Klasnja and Tucker, 2012). Moreover, non-experimental studies focus on media consumption and its effect on electoral campaigns and presidential approval (García and Wills, 2012; Arriagada, Navia, Schuster, 2010; Boas, 2005; Lawson, 1999; Lawson and McCann, 2005). However, we know very little about the role of the media in shaping 1 Corruption has generally been defined as the improper use of public office in exchange for private gain. 2

citizens perceptions of corruption. To the best of our knowledge, no studies have explored the relationship between media consumption and perception of corruption. In this paper, we seek to understand that relationship in the context of Latin America and the Caribbean. Most Latin American countries are new democracies that experienced democratic transitions in recent decades and still face some challenges in their process of consolidation, among which corruption remains significant. Evaluating the determinants of corruption perception is especially relevant in a region still prone to political and economic instability. The main contribution of this work is to assess the media as a factor that potentially shapes Latin American opinions of corruption at both the national and individual levels. By employing a multilevel regression model, with the help of survey data from the 2012 AmericasBarometer, data from the Reporters Without Borders (RWB), and data from the World Bank, we examine the effects of theoretically relevant variables on citizens perceptions of corruption. We find that the more citizens consume news via television, radio, newspapers and internet, the higher their levels of perceived corruption at the individual level. We also find that in countries with more press freedom, citizens show lower levels of corruption perception, whereas in countries with less freedom of the press, citizens perceive higher levels of corruption, net of other factors. The first section of this paper provides a brief account of the sources of citizens perceptions of corruption and the role of the media in shaping these perceptions in Latin America, paying special attention to the case of Colombia. It is followed by a discussion of the data and the empirical strategy employed in this paper, and then concludes with a presentation of the main results and corresponding discussion. 3

Previous empirical research Access to reliable information is key to holding public official accountable for their behavior. Thus, the media plays a fundamental role in fighting corruption, not only because it functions as a direct vehicle for citizens to get information about corrupt practices, but also because independent media outlets have proven to be the most effective institutions to uncover trespassing by corrupt politicians and government officials (Brunetti and Weder, 2001, Waisbord, 2008). In fact, cross national studies show that a free press is associated with lower levels of corruption (Brunetti and Weder, 2001, Lindstedt and Naurin 2010, Chowdhury, 2004, Svensson, 2005). In short, the media works as a means by which citizens become aware of corrupt bad practices by public officials. A recent and growing body of literature offers important insights into the relationship between information and corruption to answer the key question of whether voters will use information to hold their elected officials accountable (Ferraz and Finan, 2008, Zechmeister and Zimzumbo, 2013, de Figuereido, Hidalgo, and Kasahara 2011). Although experimental research is a powerful tool for understanding the casual mechanisms underlying the relationship between information about political corruption and political attitudes and behaviors, when assessing the specific impact of the media, studies have been limited in number. There is evidence that exposure to political corruption information will make voters more likely to punish corrupt politicians at the ballot box (Ferraz and Finan, 2008), while other studies show that it depresses turnout, which suggests less government accountability (De la O et al, 2010). Others studies find that the impact of corruption information varies by social class (Winters and Weitz-Shapiro 2013), party affiliation (de Figuereido, Hidalgo, and Kasahara 2011) and the state of the economy (Klasnja and Tucker 2012). All of these studies have been experimental, and though 4

they highlight the important causal relationship between corruption information and perception and how it affects the decision to vote, we know very little about whether or not this relationship holds up across different contexts. Moreover, a few cross national studies show the impact of information freedom on corruption ((Brunetti and Weder, 2001, Lindstedt and Naurin 2010, Chowdhury, 2004, Svensson, 2005, Escaleras, Lin and Register 2010), some focusing specifically on highly unequal societies (Peisakhin 2012). Furthermore, a few country-specific studies explore media effects on successful electoral campaigns (Boas, 2005; Lawson, 1999; Lawson and McCann, 2005), presidential approval (García and Wills, 2012; Arriagada, Navia, Schuster, 2010), and democratic governance (Boas, 2013). To the best of our knowledge, no studies have explored the specific relationship between media consumption and perception of corruption utilizing data collected at the national and the individual level. In this paper, we seek to understand that relationship in the context of Latin America and the Caribbean, with a particular attention paid to Colombia. Why is news consumption important at the individual level? Latin American countries have faced several problems in their process of democratic consolidation, such as climbing unemployment, rising crime rates, macroeconomic difficulties, and generally inferior governance levels. The lack of accountability, both vertical and horizontal, highlights the disconnect between citizens and government. One of the results of this lack of accountability is the persistence of corruption in the region (Altman and Pérez-Liñán 2002; Hagopian and Mainwaring 2005; O'Donnell 1999). 5

High levels of corruption have important effects on citizens opinions and attitudes, such as the erosion of interpersonal trust (Morris and Klesner 2010) and political trust, or the legitimacy of democratic institutions (Canache and Allison 2005; Seligson 2002). Even though there has been an explosion of studies that explains various ways to fight corruption, few have focused on the vital role that the media plays in shaping corruption perceptions. On the one hand, the media makes a wide range of information accessible to the public immediately by reducing the costs of gathering information themselves by supplying relevant information related to public officials bad behavior. Citizens only need to watch the news, listen to the radio, or read the newspaper to receive a wide range of information previously undisclosed (Lindstedt and Naurin 2010). Thus, the role of the media becomes a significant factor in solidifying governmental accountability as citizens increasingly make demands for better governance, all within contexts of a free and competitive media. In fact, independent media outlets are important to prevent corrupt and incompetent governments from staying in power (Petrova 2011, 790). On the other hand, under an environment of restricted information, corrupt practices are less likely to be revealed, hindering governmental accountability and making public officials more prone to bad behavior (Brunetti and Weder, 2003). The media may also itself be corrupt or may for various reasons be unwilling to report some facts about public officials. As Di Tella and Franceschelli (2011, 124) noted in their study, in many places the government has proven to be the largest advertiser in the media and could either help or diminish the freedom of information. According to the authors, for instance, there is the idea of affinity, which suggests that governments provide more allocations to media outlets that are perceived to be closer to the government in exchange for the media diminishing the coverage of negative news about the government. There is also the notion of collusive 6

equilibrium, which maintains that corruption scandals, when reported, are costly for governments, while for newspapers result in reduced circulation. In short, significant transfers from the government to a media outlet are associated with distortions in media coverage. In fact, Di Tella and Franceschelli (2011) find that a one standard deviation increase in monthly government advertising (0.26 million pesos of 2000) is associated with a reduction in the coverage given to government corruption scandals (2011:144). Unfortunately, this arrangement benefits the government and the media outlet, but has pernicious effects on citizens access to accurate political information. Consequently, the role of a free and competitive media is crucial for increasing citizens access to political information and for creating political discussion (Mondak 1995). One of the main mechanisms by which the media influences the opinions of individuals is through the selection of relevant topics (McCombs and Shaw, 1972). Moreover, citizens give more importance to some topics more than others (Iyengar, Ansolabehere et al.1992). For example, if the press highlights unemployment, citizens will vote for politicians that make this topic the center of their electoral campaigns. Thus, if the media reports on corruption scandals, it is more likely that corruption will become more salient and relevant for the citizenry. We argue that the more citizens consume news, the higher the probability these citizens will have access to information about government performance, and in turn, the higher their perceptions of corruption, especially under conditions of high political corruption messaging. According to the RAS (Receive-Accept-Sample) model on public opinion formation by Zaller (1992), individuals are exposed to information from governments, civil society and the media. When individuals receive a message, they can accept or reject these messages, be it because of their predispositions (such as party identification) or their previous levels of political 7

information. In this sense, identification with an incumbent government will have a negative effect on the perception of corruption because citizens will be predisposed to not believe the incriminating information regarding their preferred leader (e.g., Chavez, Ortega, Correa). Consequently, higher presidential approval leads to lower perceptions of corruption vis-à-vis the incumbent administration. One of the difficulties in capturing the effect of media messages on political opinions is related to the competitiveness and variety of messages that an individual receives. In that way, a measure to capture if the message was successfully received is if the individual exhibits political interest or has political awareness (Zaller 1992). The more this individual is interested in political affairs, the more attention this individual will pay to the news and the higher the probability that she is a recipient of news messages. Moreover, education also strengthens people s capacity to receive and process information from the media. When citizens consume news provided by a free media outlet, they become more cognizant of their governments actions, making political figures more accountable for their bad practices (Bruns and Himmler 2011, Reinikka and Svensson, 2005, Stromberg, 2004, Snyder and Stromberg 2010). Even if news consumers care more about sports, showbiz, clubs, etc., they will nonetheless receive political information as a result of the simple act of perusing their sources of information (Bruns and Himmler 2011), and with more education, citizens are better able to process political information (Markus and Lupia, 2008). We expect, then, that at the individual level, the more citizens access the news be it via television, radio, newspapers or internet the higher their perceptions of corruption, even after taking into account their levels of education and political interest. 8

Furthermore, individuals take information more seriously when it comes from credible and trusted sources (Druckman, 2001b; O Keefe, 2002; Lupia and McCubbins, 1998). In fact, the credibility of news sources matters for corruption. Botero et al. (2013) found in the Colombian context that corruption accusations coming from the leading national newspaper drive down levels of support and trust for corrupt politicians relative to identical accusations made against the same candidates by NGOs and the judiciary. Consequently, we hypothesize that the more citizens trust the media, the higher their perceived levels of corruption. Why is freedom of the press important at the national level? Taking into account context, the extent to which the media reaches citizens varies significantly between nations and regions. We know for instance that affluent countries experience less corruption (Treisman 2000) and that economic development reduces incentives and opportunities to utilize corrupt practices (Sandholtz and Koetzle 2000). More specifically, in prosperous and open societies where equal treatment applies to everyone regardless of the group to which one belongs (Mungiu 2006, 88), the opportunity for corrupt practices decreases and become less common. Affluent nations are more likely to enjoy working institutions, such as a free and competitive media. In these countries, government officials are more accountable to the public, while the public constantly evaluates the conduct of elected officials and punishes them for corrupt behavior (Snyder and Stromberg 2010, Bruns and Himmler 2011). In short, a free media restricts the relative power of special interest groups and facilitates public discussion (Lindstedt and Naurin 2010). In addition, the free circulation of information can improve government performance especially when laws are implemented to assure greater public access to information on governmental activities. Prime examples of laws that enhance public access to information are 9

the Freedom of Information Acts, known as FOIs (Escaleras et al 2010, Peisakhin 2012). Finally, journalists in countries with a free media are willing and able to make available to the public previously undisclosed information without the fear of retaliation, thus inhibiting corrupt behavior by public officials (Brunetti and Weder 2003). Poorer nations, conversely, are well known for their institutional deficiencies and inadequacy at fighting corruption. In countries where incomes are quite low, the economy produces minimal wealth for citizens, creating certain structural incentives for corrupt behaviors (Sandholtz and Koetzle, 2000, 36; Van Rijckeghem and Weder 2001). Furthermore, a restricted media will encourage unaccountability of elected officials and may perpetuate the vicious cycle of corruption as citizens in these countries will have less capacity to access and process information, making less likely the enforcement of sanctions for bad behavior. Moreover, as noted, governments may fund media outlets that distort information readily available to citizens (Brunetti and Weder 2008, Di Tella and Franceschelli 2011). Therefore, where free press is present, corruption is less likely, leading citizens in these countries to reveal lower levels of corruption perception (Chowdhury, 2004, Freille et al, 2007). In nations where the press is censored and journalists are often persecuted, citizens will perceive more corruption. In this paper, we explore the relationship between freedom of the press, at both the individual and national levels, and perception of corruption. Next we present the data and methods. Data and Methods In order to assess the effects of the media at the individual and national levels on citizens perceptions of corruption, we rely on a multilevel analysis. Even though there is an emerging availability of multi-country survey data, the majority of studies have depended on either 10

individual level or aggregate analyses, preventing the simultaneous analysis of the effects of both individual and country level characteristics on citizens perceptions of corruption. In this study, we hope to address these limitations by employing multilevel methods that take into account both the individual and country level nature of the data (Raudenbush and Bryk 2002). The data we employ in this project comes from the AmericasBarometer 2012, involving face-to-face interviews conducted in most of Latin America and the Caribbean, totaling 24 nations. 2 The project used national probability samples of voting-age adults. The merged file includes close to 39,000 respondents. 3 For the national level characteristics, we employ data from Reporters without Borders (Freedom Press Index) and The World Bank (GDP). The dependent variable in the model employed later is an individual s response to the following question: Taking into account your own experience or what you have heard, corruption among public officials is very common, common, uncommon or very uncommon? Perception and experience are largely unrelated because they reflect different types and patterns of corruption (Morris 2010, Weber Abramo 2008). On the one hand, experience relates to streetlevel corruption or with petty levels of bureaucratic corruption (Rose-Ackerman, 1999). On the other hand, perception relates mainly to corruption at the top level. We focus in this paper on perception of corruption of public officials rather than experience with corruption, as perceptions have been shown to be closely related to attitudes towards the political system and citizens behaviors (Morris 2009, Canache and Allison 2005). 2 Argentina, Belize, Bolivia, Brazil, Canada, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, El Salvador, the United States, Uruguay and Venezuela. In this paper we exclude Canada and the United States. 3 It corresponds to around 1,500 respondents per country, with the exception of Bolivia and Ecuador that have around 3,000 respondents. For that reason, we include a weight to compensate for those countries. The surveys took place under the sponsorship of the Latin American Public Opinion Project (LAPOP) at Vanderbilt University as well as support by the United States Agency for International Development. These surveys address topics related to views on democracy, national and personal economy, political and social tolerance, voting behavior, immigration, social and political participation, and ethnicity, among others. See, www.lapopsurveys.org 11

Figure I shows cross-national variation on perceptions of corruption in the Latin American and Caribbean region. We first note that the countries with the highest levels of corruption perception are Colombia and Trinidad and Tobago, with averages above 80 points on a 0-100 scale. At the other end, Suriname exhibits the lowest levels of perception of corruption with 39 points on the same scale. Other countries in the sample with low levels of corruption perception are Chile, Nicaragua and Uruguay (less than 65 points on the same scale). Moreover, most of the countries situate themselves close to the regional average of 71 points, showing that Latin American citizens perceive, on average, high levels of corruption. Colombia Trinidad & Tobago Argentina Guyana Panama Dom. Rep. Peru Honduras Mexico Venezuela Jamaica Costa Rica Paraguay Guatemala Bolivia Ecuador Belize Haiti El Salvador Brazil Chile Nicaragua Uruguay Suriname 38.8 81.7 80.9 79.5 79.4 78.4 78.1 76.9 76.7 76.5 75.3 75.2 74.0 73.0 69.0 68.2 68.1 67.5 66.4 65.6 65.4 64.9 62.2 61.8 0 20 40 60 80 Perception of Corruption 95% Confidence Interval (Design-Effects Based) Source: AmericasBarometer by LAPOP Figure 1. Perception of Corruption across LAC We also illustrate national-level freedom of the press as measured by the Press Freedom Index by Reporters Without Borders, an organization that evaluates (1) the number of journalists murdered, expelled, or harassed; (2) the presence of a state monopoly on TV and radio; and (3) 12

the extent to which a country has censorship and/or self-censorship in the media. In sum, it scrutinizes the overall independence of the media and possible challenges facing foreign reporters. Keep in mind that higher numbers denote more freedom of the press; lower numbers, conversely, represent less freedom. Figure 2 demonstrates the marked differences in freedom of the press among nations in Latin America and the Caribbean. For instance, Jamaica and Costa Rica are at the higher end while Mexico and Colombia are found at the lower end, and the differences are striking. Jamaican and Costa Rican freedom of the press scores are 76 and 77 times higher, respectively, than that of Mexico; and 70 times higher than that of Colombia. Jamaica Costa Rica Suriname Uruguay El Salvador Argentina Trinidad & Tobago Haiti Guyana Nicaragua Paraguay Chile Dom. Rep. Guatemala Brazil Ecuador Bolivia Panama Peru Venezuela Honduras Colombia Mexico 76.7 75.9 74.7 69.4 64.4 59.7 58.7 58.0 54.2 49.3 44.7 44.7 40.4 38.7 38.3 35.7 33.7 28.0 22.4 18.7 12.7 7.2 1.0 0 20 40 60 80 Press Freedom Index 2011/2012 Source: Reporters Without Borders Figure 2. Press Freedom Index 2011/2012 across LAC 13

Data from each of these countries revealed perceived corruption levels above the regional average of 71 points. Moreover, we note that while Colombia has become a strong Latin American economy in the past few years, it remains one of the countries with the weakest freedom of the press. Colombian citizens show the highest levels of corruption perception. Does a weak freedom of the press explain Colombians high perceptions of corruption? We will explore this question in more depth in the following pages. These results show preliminary support for our expectations that in countries with more independent media, citizens perceive less corruption, whereas in countries with less freedom of the press, citizens show high levels of corruption perception. We now proceed to the analysis of these questions to better understand the factors that help shape an individual s perception of corruption. Analytical Strategy We describe three ways by which the media might influence citizens perceptions of corruption. First, we begin with various indicators that previous research suggests influence perceptions. We include a respondent s age (measured in years), sex (recoded into female=1), level of education (years of education), and a binary measure of area of residency (recoded into urban=1). All of these seek to control for socio-demographic characteristics. A second group of variables examines political factors such as respondents support for the current president. It is likely that individuals who support the president (and the government more generally) will express lower levels of corruption perception. This question asks: Speaking in general of the current administration, would you rate the job performance of [President of country] very good, good, neither good nor bad (fair), bad, very bad? 14

In addition, we incorporate measures representing views of external controls of corruption. As Brunetti and Weder (2003) indicate, a court system where corrupt bureaucrats can be easily and effectively sued sharply reduces the potential rewards of corruption (p. 1803). We then include two variables that control for citizens views of their judicial system that may also influence their levels of perceived corruption. The items asked: To what extent do you trust the justice system? and To what extent do you think the courts in [country] guarantee a fair trial? The final political variable we think is important to assess citizens perception of corruption is political interest. Informed individuals will express higher levels of political interest, and in turn, display higher perceived levels of corruption (Bohn, 2012). This item was measured by the following item: How much interest do you have in politics: a lot, some, little or none? With the inclusion of these multiple political measures, we hope to capture the impact of these factors on perceptions of corruption. The final group of variables seeks to capture the role that the media plays in shaping people s opinions on corruption. We specifically want to know if citizens exposure to the news leads to higher levels of perceived corruption. The items asked: About how often do you pay attention to the news, whether on TV, the radio, newspapers or the internet: daily, a few times a week, a few times a month, rarely, never? We expect, at the individual level, that the more frequently citizens consume the news, the more pronounced their perceptions of corruption will be. Furthermore, we also believe it is important to understand to what degree citizens trust in the media affects their views on corruption. Thus, we incorporate the following item: To what extent do you trust the mass media? All independent variables are scaled to run from 0-100 for ease of interpretation. 15

In this first portion of the analysis, besides our traditional socioeconomic variables, we focus on individual level factors already described above and their relationship to perceptions of corruption. This individual-level model is illustrated as follows: Y ij = 0j + 1j *(TRUST IN THE MEDIA) ij + 2j *(NEWS CONSUMPTION) ij + 3j *(POLITICAL INTEREST) ij + 4j *(PRESIDENTIAL APPROVAL) ij + 5j *(TRUST IN COURTS) ij + 6j *(FAIR TRIAL) ij + 7j *(EDUCATION) ij + 8j *(AGE) ij + 9j *(FEMALE) ij + 10j *(WEALTH) ij 11j *(URBAN AREAS) ij + R ij Where Y ij represents average levels of perception of corruption for an individual i in country j, and 0j is the individual level intercept. 1j 12j are the effects of the variables included in our model. Finally, R ij is the error term. Our second analytical strategy pertains to the effects of country level factors. Specifically, we expect that, irrespective of individual characteristics, Latin American citizens living in wealthy nations and with more freedom of the press will show lower levels of corruption perception. The specific country-level equation for the model intercept is as follows: 0j = γ 00 + γ 01 *(FREEDOM OF PRESS) ij + γ 02 *(ECONOMIC DEVELOPMENT) ij + U 0j Where γ 00 is the country level intercept and γ 01 and γ 02 are the effects of press freedom and national wealth on the model intercept ( 0j ) 16

Our expectation with all of these measures is that at the individual level, exposure to the media will lead to higher levels of perceived corruption, while at the national level, a free and competitive media will lead to lower levels of perceived corruption. Results for LAC To test our expectations, we present two models. The first column in Table 1 displays the results from our individual level model (Model I), in which we assess the direct relationship between frequency of news consumption and perception of corruption with a basic model for the pooled data of the 23 countries in the AmericasBarometer. Next, we introduce the country-level model (Model II) that accounts for national level differences, specifically, freedom of the press and economic development. When we look at the results of our baseline model, the predicted level of perception of corruption is 74.57 holding the rest of variables at their mean. 4 Among our traditional socio-economic variables, all variables yield statistically significant results. Not surprisingly, wealthy and educated individuals and those who live in urban areas tend to perceive more corruption. In particular, more educated individuals are more likely to be aware of the political system within which they live than those who are less informed and, therefore, are more likely to be in a position to be critical of the government s bad behavior. These results may reflect that higher levels of sophistication generate more critical citizens (Norris 1999), thus increasing their levels of perceived corruption. In fact, education has been closely linked to news consumption (Markus and Lupia, 2008). Moreover, these individuals tend to live in urban/metropolitan areas where more public institutions are available and where the incidence of corrupt practices is higher. In addition, males and older individuals exhibit higher 4 Multilevel models presented in this paper were computed using the statistical package Stata 12. For a more detailed discussion on multilevel modeling, see Raudenbush and Bryk. 17

levels of corruption perception. As we mentioned at the beginning of this paper, socio-economic characteristics play a role in determining perceptions of corruption. Table 1. Explaining Factors of Perception of Corruption Model I Individual-Level Model II Country-Level Intercept 74.57*** (1.237) 79.93*** (4.617) Socioeconomic Education 2.015*** (.2875) 1.735*** (.4605) Age.0710*** (.0110).0718** (.0231) Female -.9789*** (.3035) -.8443* (.4256) Urban Areas 1.734*** (.5332) 1.005 (.6258) Wealth.3621* (.1563).5857** (.1937) Political Factors Political Interest.0045 (.0058).0054 (.0069) Presidential Approval -.1068*** (.0091) -.0977*** (.0153) Courts Guarantee Fair Trial -.0625*** (.0077) -.0623*** (.0121) Trust in Court System -.1099*** (.0076) -.0947*** (.0082) The Role of the Media Trust in Media.0069 (.0069).0175 (.0144) Frequency of News Consumption.0244** (.0078).0222+ (.0124) (TV, the radio, newspapers or the internet) Country-Level Intercepts Effects Press Freedom Index (RWB) -.1667* (.0824) GDP per capita (current US$) -.00005 (.00031) Note: Multilevel mixed-effects linear regression coefficients with estimated standard errors in parentheses. + p <.10, *p <.05, **p <.01, ***p <.001 Number of Countries = 23 Number of Cases = 30677 Sources: AmericasBarometer by LAPOP 2012, Press Freedom Index 2011/2012 by Reporters Without Borders, GDP per capita (current US$) 2012 by World Bank Our next set of variables show that political factors also influence the extent to which citizens perceive corruption. As we expected, individuals who approve of the president s performance and who trust institutions that serve as external controls of corruption such as courts perceive less corruption. In addition, citizens who believe that the courts guarantee a fair trial also show lower levels of perceived corruption. Surprisingly, political interest did not reach statistically significant results. 18

When assessing the role of the media in shaping citizens opinions toward corruption, we observe that the more individuals pay attention to the news, whether on TV, the radio, newspapers, or the internet, the higher their levels of perceived corruption. These results provide preliminary evidence for the significant role that the media plays in shaping citizens perceptions of corruption. Trust in the media, on the other hand, did not reach statistically significant results in our individual level model. The final step for our analytical strategy is the introduction of our country-level variables for the model intercepts freedom of the press and economic development. It shows that at the national level, the only statistically significant variable that matters is freedom of the press. In countries with a more independent media, levels of perceived corruption among the citizenry are lower. The effect remains strong even after controlling for the wealth of nations. A free and competitive media has been closely linked to the ability of civil society to control government performance (Brunetti and Weder, 2003, Freille, 2007). This evidence suggests that the freedom of the media matters more than the wealth of nations when explaining corruption perceptions. In the following pages, we explore in more detailed this relationship, paying special attention to the case of Colombia. CASE OF COLOMBIA In the last decade, Colombia has made improvements in terms of the rule of law and governance. A testament to this country s efforts is the measure approved last year for Colombia s addition to the OECD by 34 of its members. Moreover, according to the World Bank s Worldwide Governance Indicators (WGI), the country has improved from 25.8 in 2002 to 27.9 in 2011. However, other indicators of the same index show that Colombia ranks in the lower half with a 19

score of 46.4 in terms of control of corruption (World Bank, 2012).5 Most recently, perception of corruption has dramatically increased at various levels of government. According to the Global Corruption Barometer (2010/2011), 56 percent of Colombians perceive that corruption levels have increased in the last 3 years6 and data from the AmericasBarometer 2012 reveal Colombia as the country with the highest perception of corruption in the region. Furthermore, Transparency International 2012 reports that Colombia received the worst score in 10 years, shifting its rank from 57 in 2002 to 94 out of the 176 countries assessed in 2012 on the Corruption Perception Index.7 Although high levels of perceived corruption are worrisome, they do not always correspond to high (or rising) levels of real or objective corruption from a comparative perspective. It is possible that with increased media coverage and the media s role in uncovering high-level government corruption in recent years, citizen perceptions may have increased amid heightened public attention to corruption scandals. Thus, in this section we analyze the impact of news consumption in shaping Colombians views of corruption. Thanks to additional questions on news consumption broken down by different types of sources of information such as radio, television, and newspapers, we are better able to assess the impact of each source on perceptions of corruption. The media in Colombia reflect historical patterns of market concentration and ownership by political and economic elites (Boas 2013). At the same time, the media have been one of the main targets of the political violence that has impacted the country on several fronts: drug trafficking, guerrillas, paramilitaries, and even the State. Usually the most affected are members of the media at the local level, given that the Colombian state has little control of the most 5 See, http://info.worldbank.org/governance/wgi/index.aspx#home 6 See, http://www.transparency.org/gcb201011 7 See, http://www.transparency.org/cpi2011/results 20

remote and inaccessible municipalities in the jungle region, for instance. However, there are still threats to the mainstream media in Bogota. In 2013, a journalist from the magazine Semana who was investigating corruption in the military was attacked. 8 His car was shot several times, but he was fortunately unhurt. In spite of the fact that Colombia has proven to be one of the countries with the highest risk for investigating journalism (FLIP, 2012), in the past decade, journalists have become especially active in investigating corruption cases. Corruption can be found at several levels of the state in Colombia. Corruption manifests itself in various forms, including widespread financial and political corruption, patronage, and misuse of power. Recently, local public officials and technocrats have been prosecuted. In 2011, the Mayor of Bogotá was suspended for failing to fulfill his public duty and for irregularities in awarding contracts which were part of a $500 million bribery scheme, involving a former member of Congress and a large construction group, as well as other businessmen, politicians, and high-level urban planners (López and Romero, 2007 ). Also, in 2011, investigations over an alleged plot to embezzle billions of dollars in public funds between 2004 and 2009 led to the arrest of 12 staff members from the Office of Taxation and Customs. Before and during the 2011 regional and municipal elections, Colombian media reported widespread vote rigging and vote buying across large areas of Colombia tied to the financial support of narco-paramilitary groups with revenues from organized crime (López, 2011; Gutiérrez, 2013). In the following section, we examined the impact of news consumption on perception of corruption in the Colombian context. 8 See, http://www.semana.com/nacion/articulo/el-atentado-contra-ricardo-calderon-repudiable/341900-3 21

Results To test our expectations in the context of Colombia, we present two models. The data we employ in this section comes from the AmericasBarometer 2006-2013, involving face-to-face interviews with Colombian citizens for a total of 12,000 cases. We incorporate the same variables we did for the full sample of Latin American and Caribbean countries, with the sole exception being that we add political knowledge as another key predictor of corruption perception. Because the Colombian data set offers consistent inclusion of political knowledge indicators across time, we constructed an index of political knowledge which asks: (1) What is the name of the current president of the United States of America? (2) How long is the presidential term of office in Colombia? and (3) How many representatives does Congress have? The Colombian data set is also unique in the sense that it allows to differentiate the effects of different types of media outlets on the perception of corruption across time. Those include: (1) Do you listen to the news by radio? (2) Do you watch news on the TV? (3) Do you read news in Newspapers and (4) Do you read or listen to the news via Internet? Model I in the first column of Table 1 resembles our baseline model for the full sample, with the addition of political knowledge and education broken down by levels. 9 Once again, we notice that socio-economic variables matter. Wealthy Colombians and those who live in urban areas perceive higher levels of corruption. Moreover, Colombians who hold secondary and tertiary levels of education display higher levels of corruption perception compared to Colombians who do not have any education, indicating again the significant role that education plays when assessing public officials political performance (Miller and Krosnick, 2000). In the case of Colombia, there are no statistically significant differences between females and males in 9 In our full sample with all LAC countries, we run education broken down by levels but the results did not show statistically significant differences between primary, secondary and tertiary education compared to no education, therefore, we decided to leave the original variable as it is in the full sample. 22

their levels of perceived corruption in contrast to Latin Americans in general, where males show higher levels of this perception. Moving on to our next set of political variables, we again see that Colombian citizens who approve of President Santo s performance exhibit lower levels of perceived corruption and the same is true for those who trust the court system and believe that the courts guarantee a fair trial. In this model, we see the important role that political knowledge plays in evaluating the performance of public officials. 23

Table 1. Explaining Factors of Perception of Corruption Model I Model II Intercept 57.61*** (3.749) 54.96*** (3.286) Socioeconomic Education Primary a 1.745 (2.361).7274 (1.867) Secondary 4.606+ (2.522) 4.934** (1.940) Superior 6.854** (2.628) 6.834*** (2.026) Age.2253*** (.0236).2453*** (.0208) Female -.7630 (.6542) -.7744 (.6124) Urban Areas 3.123** (1.140) 2.621*** (.8133) Wealth 1.122*** (.2449) 1.059*** (.2031) Political Factors Political Interest.0157 (.0105).0134 (.0084) Political Knowledge.1104*** (.0142).1189*** (.0123) Presidential Approval -.0560***(.0136) -.0415***(.0111) Courts Guarantee Fair Trial -.0271+ (.0147) -.0363** (.0125) Trust in Court System -.0645*** (.0134) -.0532***(.0121) The Role of the Media Trust in Media -.0402** (.0136) -.0246* (.0115) Frequency of News Consumption (TV,.0474* (.0197) --- the radio, newspapers or the internet) Radio ---.1016 (.2633) Television ---.6054 (.4241) Newspapers ---.2665 (.3016) Internet ---.7823* (.3445) Years b 2006 --- -2.657+ (1.400) 2007 --- -1.763 (1.347) 2008 --- -3.384** (1.249) 2010.7908 (1.228).8703 (1.227) 2011 2.886* (1.183) 3.100** (1.167) 2012 4.543***(1.382) 4.650***(1.351) 2013 -.7328 (1.455) -.5142 (1.401) Note: OLS with Robust Standard Error in parentheses that account for the complex sample features of the AmericasBarometer data. + p <.10*, p <.05, **p <.01, ***p <.001 Number of Cases = 10,428 All models include the standard AmericasBarometer probability weights at the individual level Baseline: a. No education b. year 2009 (start year of statistically significant higher levels of perceived corruption) 24

One way to measure political awareness is through the extent to which individuals possess knowledge about political issues of the day. Not surprisingly, we then see that the more Colombians exhibit political knowledge the higher their perceptions of corruption. Moreover, Colombians who pay attention to the news, whether on TV, the radio, newspapers or the internet, the higher their perceived levels of corruption. Because the Colombian data set offers a unique opportunity to break down these sources of information, we were able to run a second model that differentiates the linkage between various types of media sources and perceptions of corruption. In our second model in Table 2, we find that the frequency of news consumption through the internet is the only statistically significant predictor of corruption perception compared to other news sources such as television, radio, and newspapers, despite these being in the expected direction. This could be explained due to the differences that exist between being exposed and effectively receiving news messages (Zaller 1992) that is, we argue that individuals who consume news through the internet would be radically different from those who consume news via television or radio, for instance. In the case of the internet, and to a lesser extent the newspapers, individuals play a more active role in seeking information by themselves. That is, individuals self-select into looking for specific types of information, whereas in the case of the Television or the radio, individuals could be exposed to all sorts of information and not necessarily pay attention to the sources. All of the effects for the other variables remain virtually the same when adding news consumption broken down by different sources of information. Our final variable of interest, trust in the media, goes against the expected direction in the Colombian context. This finding is particularly interesting as media outlets that decisively pursue the exposure of corrupt practices increase citizens perceptions of corruption especially when these sources are regarded by citizens as credible and legitimate (Botero el. 2013, Miller and 25

Kronisck, 2000). Of course, the extent to which citizens trust the media depends on the effectiveness of this institution. According to Przeworski et al. (1999), the media and political opposition are important because they provide information about government performance by the incumbent, and with this information citizens decide to punish or reward public officials. In the case of Colombia and the rest of Latin America and the Caribbean, we do not find support for this argument. We find instead that the more Colombians trust the media, the less they perceive corruption. We do not know why this is the case, but we speculate that Colombians who trust the media and perceive it as legitimate, they may also exhibit more general positive political views toward their system, thus showing, in turn, lower levels of perceived corruption. Conclusion This paper provides empirical evidence for the relationship between news consumption and perceptions of corruption. Our findings reveal that, at the individual level, the more frequent attention Latin Americans pay to the news via Television, radio, newspapers or internet, the higher their perceived levels of corruption. Interestingly, trust in the media does not yield any significant effects on the perception of corruption across Latin America. In addition, individuals who trust more judicial institutions show, on average, lower levels of these perceptions. This finding is particularly significant because perceived levels of corruption as well as low levels of real corruption depends on the effectiveness of the judicial system in punishing cases of corruption. Finally, Latin American citizens who approve of the incumbents government performance, exhibit low perceptions of corruption. It would be interesting to explore further how presidential approval may be a predisposition that conditions the effects of the media. This is particularly significant in a region such as Latin America where populist practices have created 26

an environment of impunity toward scandals of corruption (Ryan, Love and Martínez Gallardo, 2014). National level findings show that the conditions under which the media is competitive and free, citizens in these countries exhibit lower levels of corruption perception, while in countries where journalists have been murdered, expelled, or harassed, where there is a presence of a state monopoly on TV and radio and where the media has been censored (RWB), perceptions of corruption are higher. In other words, context matters when explaining individual level opinions on corruption. When analyzing the case of Colombia in particular, a country that shows the highest perceived levels of corruption in the region in 2012 and also is one of the countries with the least freedom of the press after Mexico, according to RWB, we find again that Colombian citizens who consume more frequently the news express higher levels of perceived corruption and when broken down by specific sources of information, the internet surfaces as the most significant vehicle to increase these perceptions. We speculate that internet users play a more active role in seeking information, that is, they self-select into looking for specific types of information thus more likely to sort through corruption scandals and, in turn, be influenced by these messages. Moreover, the role of the media in Colombia in recent years has been instrumental in putting the issue of corruption at the top of its agenda. It has been a key player in researching different types of corruption. Future research should concentrate efforts, not only in taking into account the conditions of press freedom in general but also the ability of the media in exposing corruption, to better capture the effects of whether the information provided in the news influences corruption perceptions among the citizenry, and also, offer a better assessment for the quality of democracy more generally. 27