The global transparency bandwagon is large and growing. The Open Government Partnership, launched in September 2011 with eight founding governments, had grown to seventy governments by mid-2016.Footnote 1 Among other things, member states commit to ‘increase the availability of information about governmental activities’ as part of the effort to ‘foster a global culture of open government that empowers and delivers for citizens, and advances the ideals of open and participatory 21st century government’.Footnote 2 Civil society organizations have promoted and joined this movement through networks such as the Partnership for Transparency Fund, which promotes openness in developing coountries and foreign aid donors, including the World Bank.Footnote 3 The World Bank, in turn, first developed a governance strategy promoting transparency in 2007 in which it argued that ‘building “capable, transparent and accountable” country institutions will be fundamental to ensuring sustainable development’.Footnote 4 Government officials in powerful states have been important cheerleaders for transparency, claiming, for example, that ‘transparency can be transformative. It can help build trust, efficiency and save lives’.Footnote 5
Despite these substantial efforts and claims about positive outcomes, we know relatively little about the actual effects of transparency and of the information transmitted by transparency efforts. In this article, we use a series of survey experiments conducted on the streets of Lima, Peru, to investigate a fairly simple question: what are the effects of government transparency and information on attitudes regarding support for the Peruvian political system? In the experiments, we asked subjects to watch short videos that highlighted information about Peru culled from online transparency portals sponsored by the Peruvian government. We then questioned respondents about their evaluations of the Peruvian system generally, their views of the regime’s performance, their trust in regime institutions, and their trust in local government. These four factors may be conceptualized as constituting important dimensions of system support or legitimacy, terms that we use interchangeably.Footnote 6
Concerns about the legitimacy of democratic governments in Latin America and other developing countries are widespread. Third-wave democracies remain wobbly in many parts of the world, unconsolidated or ‘partial’ and prone to democratic backsliding or reversals,Footnote 7 a process that could be caused in part by low or declining system support.Footnote 8 Low system support can become a self-perpetuating cycle because it prevents the government from marshaling resources to accomplish its goals, which then decreases trust levels further.Footnote 9 Such a situation is dangerous because ‘whereas autocratic or hybrid systems can survive for extended periods on the basis of enforced popular acquiescence or the distribution of rewards, democratic regimes depend centrally on the creation and constant renewal of popular legitimacy’.Footnote 10 Concerns about legitimacy are not confined to the developing world; they also arise in wealthy, developed democracies, usually centered on indicators of trust in institutions.Footnote 11 While scholars have sometimes labeled these problems ‘legitimacy crises’, the difficulties seem small when compared with those of developing countries.Footnote 12
Peru has particularly large problems with system support. Government performance – measured by levels of governance, democracy, rights protection and development – is frequently a good predictor of system support.Footnote 13 In Peru, however, government performance is quite a bit higher than system support. As noted by the 2010 AmericasBarometer survey, ‘public opinion in Peru has reached a state of deep misgiving and discontent with the country’s political institutions, which then influences attitudes towards democracy and its associated principles’.Footnote 14 Peruvians have a higher tolerance for military coups than all but four other Latin American countries, and they have the fourth-lowest level of support for the political system in the region. Trust in Peru is so low that one scholar characterized it as an ‘absence’ of trust in political institutions.Footnote 15 Despite promising economic projections, ‘recent public opinion polls show that most Peruvians exhibit a degree of discontent with their political institutions that is usually associated with situations involving civil strife and economic stagnation’.Footnote 16 In sum, ‘In almost all the attitudes that would be conducive to the establishment of a stable democracy, Peru is at extremely low levels when compared to other countries in the region’.Footnote 17
Does increased transparency alter this situation? What are the effects of the information conveyed by that transparency? A large number of studies have found that government performance affects system support; we explore the possibility that information about performance, as conveyed by transparency, alters system support. We find that transparency itself had little impact on political attitudes, but that some information provided by transparency increased system support. In particular, introducing Peruvians to an overview of the government’s transparency website had no effect on their attitudes. However, the source and framing of the information they encountered on the website mattered for attitudes about system support. If the Peruvian government was identified as the source of the information, the information made no difference in respondents’ political attitudes. If a little-known Peruvian Non-Governmental Organization (NGO) was identified as the source of the information, there was also no effect. If the U.S. Agency for International Development (USAID) was identified as the source of the information, however, respondents’ support for the Peruvian political system increased substantially. We attribute these findings to the credibility of the source,Footnote 18 though a different causal pathway is possible. With respect to framing, information suggesting that respondents are relatively well off compared to those in other communities also increased system support. We expect this is due to framing effects that associate the government with improved socioeconomic outcomes, but other pathways are possible. We also found that information about poor socioeconomic well-being did not decrease system support compared to the control group, perhaps because baseline expectations of government performance were already so low. Regardless of the particular mechanism, this result suggests an important refinement of the common finding that government performance influences system support.Footnote 19 Much of that debate has focused on whether voters are narrowly self-interested or are influenced by the general economy. Our findings suggest that scholars should think about the ways in which citizens make judgments about the well-being of their local communities.Footnote 20
In the first section, we lay out scholarly debates about system support and its causes and address the potential importance of transparency and of the nature and content of information revealed by transparency. In the succeeding section, we develop two hypotheses about information revealed by transparency efforts. The first suggests that information will affect system support when it is endorsed by a credible source and the second that the information will affect system support when it provides a strong positive frame about local, relevant socioeconomic outcomes. We then lay out our experimental design and methods. The final section summarizes and reports our findings.
SYSTEM SUPPORT, TRANSPARENCY, AND INFORMATION
While there are a variety of ways to approach political attitudes toward the government, we adopt the conceptual framework of ‘system support’. Utilizing groundbreaking work by Easton, Norris defined system support as ‘reflect[ing] orientations toward the nation-state, its agencies, and its actors’.Footnote 21 As the definition indicates, system support is a multi-dimensional concept that does not simply refer to citizen support for the political system as a whole, but rather related areas of support at different governance levels. In contrast to Easton’s famous distinction between diffuse and specific system support,Footnote 22 Norris identified five levels: support for the nation-state or ‘political community’, the principles that the government embodies, government performance, public confidence in the institutions of government, and support for specific office-holders.Footnote 23 In a sophisticated analysis of public opinion in eight Latin American countries, Booth and Seligson generally confirmed the validity and importance of these dimensions but also found a local government dimension to be salient.Footnote 24 Other scholars have confirmed the existence of this local government dimension as well.Footnote 25
In this study, we examine four dimensions of system support: support for the broad national political community, perceptions of regime performance, confidence in regime institutions, and perceptions of local government institutions. Our factor analysis, discussed in more detail later, suggests that these four dimensions emerged from our survey of political attitudes among Peruvians in ways that are similar to the dimensions identified by Booth and Seligson.Footnote 26 Our dependent variable, which includes all four of these dimensions, is thus broader than, but related to, the ‘trust in government’ issue that many scholars have examined.Footnote 27 Our ‘confidence in institutions’ dimension is most similar to their conception of ‘trust in government’. Scholars also note that questions about citizen ‘satisfaction with democracy’ are frequently correlated with system support,Footnote 28 so we also draw on this literature.
System support is important because it can influence regime stability and the success of government programs.Footnote 29 In a study on whether political trust matters, Marien and Hooghe concluded that stability is undermined when trust in government is low.Footnote 30 This is a direct result of public unwillingness to abide by the laws enforced by the government, including tax laws.Footnote 31 Morris and Klesner found that decreasing trust increases perceptions of corruption in a self-perpetuating cycle that effectively prevents governments from making much progress in the fight against corruption.Footnote 32 Mishler and Rose found that institutional trust contributes ‘in important ways to democratic values and to citizen involvement in politics.’Footnote 33 In the relatively new democracies in the developing world, then, we might rightly be concerned about levels of system support and institutional trust.
What determines the extent of system support and trust in government? Many scholars point to government performance and especially to perceptions of that performance.Footnote 34 Many different types of perceived performance appear to have an influence, including corruption,Footnote 35 general evaluations of the national economy,Footnote 36 clean elections and civil rights,Footnote 37 social assistance programsFootnote 38 and decentralization.Footnote 39
If perceptions of government performance are so influential, then information about that performance should also be important because information can change perceptions.Footnote 40 Government-sponsored transparency provides a potentially important source of information on government performance. Although the definition of transparency is widely debated, it generally centers on information provision.Footnote 41 In fact, Grant and Keohane,Footnote 42 defined it simply as the ‘widespread availability of information’. Florini defined it in a similar way: ‘the degree to which information is available to outsiders that enables them to have informed voice in decisions and/or to assess the decision made by outsiders’.Footnote 43 Government-sponsored transparency thus refers to government efforts to make information available and to the extent of that information. Those efforts might encompass a variety of activities (such as freedom of information laws or open government meetings), but in this paper we focus on Peru’s effort to publish information about government activities and outcomes on prominent websites. While the other sources of transparency may also be important, more citizens are likely to encounter information on a government website than are likely to attend meetings or to utilize freedom-of-information laws. The information revealed by these transparency sites obviously has a number of dimensions including source and content, which we discuss at greater length below.
Why might transparency efforts impact system support? Ferejohn developed an answer using principal–agent theory.Footnote 44 When voters (principals) evaluate politicians (agents), voters frequently face severe information problems because they have difficulty either knowing the nature of government performance or being able to know whether any observed outcomes are the result of government effort or of some other factor. Where transparency is low, voters have little incentive to trust the government. As transparency increases, voters can increase certainty about politicians’ actions and the role they play in the observed outcomes. As long as the information provides positive signals, citizens are likely to increase their trust in government. Another possibility concerns people’s preferences for procedural fairness. As Licht et al. summarized it, ‘Social psychology research has indicated that people are more likely to accept decisions that are arrived at by a procedure that is considered to be fair, and are more satisfied with authorities and institutions using procedures that are considered to be fair …’.Footnote 45 Information may provide a sense of procedural fairness by conveying a sense of openness, impartiality or respectful treatment of those affected by the decision.
At the same time, it is difficult for new information to change citizens’ perceptions because people often discount that information if it disagrees with their preexisting views.Footnote 46 As Lupia and McCubbins describe it, ‘persuasion in political contexts can be difficult. Differing ideologies and competition for scarce resources give political actors a reason to mistrust one another’.Footnote 47 Most people do not have a tabula rasa on which transparency information can be inscribed with clear effects. This makes attitude changes difficult.
With competing theoretical expectations about the role of information on legitimacy, what does the evidence show? Until the last few years there was ‘little or no empirical examination of the impact of information and knowledge on political trust’.Footnote 48 A couple of studies have shown a positive relationship between transparency and trust.Footnote 49 Others have found no relationship or a negative relationship between transparency and trust.Footnote 50 The most important theme in this literature, however, is that the impact of information on trust depends on a large number of contextual factors. Among these are factors like source and content of the information. Grimmelikhuijsen found that the extent of positive or negative spin in information about the government’s air pollution efforts had an important effect on citizen trust in government.Footnote 51 Licht found that an independent journalist’s endorsement of information increased trust more than perceived level of transparency.Footnote 52
Thus, we expect no direct relationship between transparency itself (referring to government efforts to make information available) and trust in government. While some citizens may value transparency as an end in itself, those citizens are unlikely to be numerous enough to influence overall levels of trust when encountering transparency. Rather, we expect dimensions of the information (revealed by transparency efforts), such as source and framing, to influence citizen trust. Consequently, we expect transparency efforts themselves to matter less than the nature and features of the information revealed by those efforts.
Source Credibility
Since political persuasion has proven to be difficult, information must have certain characteristics if it is to change political attitudes. In fact, whether citizens accept a particular message may have more to do with external cues than it does with the content of the message itself.Footnote 53 Scholars have identified credibility as perhaps the most important external cue that makes for communicating persuasive messages.Footnote 54 Since persuasion is difficult and requires some level of trust, messages have no hope of being persuasive and changing political attitudes without credibility.Footnote 55 Citizens will not be persuaded by a politician or campaign that they perceive as dishonest or manipulative.
Therefore, citizens are more likely to trust and be persuaded by a source they believe to be credible.Footnote 56 Despite a scholarly consensus on the importance of information sources, the actual elements that contribute to a source’s credibility are contested. Many researchers cite the bias associated with a source as responsible for its credibility. Some have found that the political party and/or ideology associated with a source affects citizen perceptions of the information provided by that source.Footnote 57
Other researchers note the importance of source status in determining credibility. Druckman, for example, found that people held a significantly higher opinion of an issue if it was attributed to the New York Times rather than the National Enquirer and to Colin Powell rather than Jerry Springer.Footnote 58 Similarly, Greer and Pornpitakpan find that high-credibility sources, such as the New York Times, are more influential than lower-credibility sources.Footnote 59 It also has been argued that when an issue is debated, independent sources are more credible than sources associated with the issue.Footnote 60
In the case of government transparency and system support, governments that report positive performance are similar to corporate public relations departments. People distrust public relations practitioners because they assume that information selected and reported benefits only the company.Footnote 61 People do not accept messages from sources that, although often trustworthy, are lacking in credibility because of a conflict of interest.Footnote 62 Attorneys also understand this concept and, therefore, seek to show that, although the testimony stands up in every other way, the witness is biased because of self-interest.Footnote 63 The public understands that affiliated government agencies, like public relations spokespersons and self-interested witnesses, have biased reporting.Footnote 64 This is the reason why research suggests that positive information coming from a source that is directly affiliated with a company is less persuasive than information that is attributed to an unaffiliated organization.Footnote 65 Callison reports information attributed to an internal organization, such as public relations, has no chance of receiving a positive audience reaction.Footnote 66 Studies show that while citizens have a favorable opinion of government workers and programs with which they interact, they also tend to distrust the government and public employees in the abstract.Footnote 67 It is likely, therefore, that if a government provides generally positive information about its performance, that information is not as credible as it would have been if a third party had advocated the same position. This may be especially true in countries such as Peru, where government credibility is quite low, as previously noted. As a result, we suggest:
Hypothesis 1: Positive information about government performance revealed by transparency efforts will increase system support when it is endorsed by credible third parties and will have no effect otherwise.
Framing Effects: Local Socioeconomic Well-being
Our second hypothesis focuses on the framing of the information being provided. The information on transparency websites can frame government institutions and behavior in particular ways. Informally, ‘framing refers to the process by which people develop a particular conceptualization of an issue or reorient their thinking about an issue’.Footnote 68 Frames orient citizen understandings by providing interpretations of events and issues. For example, many more people support a hate group holding a political rally if framed by the value of ‘free speech’ rather than the ‘risk of violence’.Footnote 69 Citizens are, of course, not blank slates on which governments and political groups inscribe particular opinions, but citizen attitudes are comprised of a large number of underlying beliefs, ideas, and values. Where communicative efforts increase the salience of some of those underlying judgments, citizen attitudes about complex topics can shift as those judgments rise to active consideration in citizens’ thought processes.Footnote 70
Information provided through government transparency is, therefore, unlikely to create fundamentally new attitudes; citizens have already formed opinions on issues like system support and trust in government. However, different types of information may increase the saliency and weight of a particular attitude related to system support. If we find evidence that information about government performance alters system support in an experimental context, then we increase the likelihood that the correlations found by other scholars between government performance and system support are causally related. We also would have evidence that system support can be altered through elite communication efforts, a possibility that previous Latin American research has not examined carefully.
While no piece of information is completely value-free, not all information on transparency websites communicates a coherent frame. In fact, much of it is probably difficult to interpret or could be interpreted in a large number of ways. What does it mean if the information reveals that the government paid a little-known contractor 1,000 Peruvian soles for an office-supplies product? Or if it reveals that the subject of a recent government meeting was traffic problems? Such information probably does not provide a frame capable of affecting political attitudes because it offers no coherent interpretation of government institutions or processes.
Which particular frames are likely to have an effect on opinions about system support? Research on system support in Latin America has found strong correlations between personal well-being, especially as it is associated with government services, and system support. Hiskey and Seligson found that, in Bolivia, ‘citizens with more positive views of local government services had higher levels of system support than those who viewed the quality of local government services as poor’.Footnote 71 Similarly, a study in the Dominican Republic found that ‘the single most important factor explaining levels of trust in institutions was citizens’ evaluation of the provision of basic services, such as education, health, and transportation’.Footnote 72 These findings are supported by research conducted in Argentina, which shows that citizens’ ‘evaluations of their nation’s and city’s political and economic situations’ were significantly correlated with measures of system support.Footnote 73 In El Salvador, Córdova and Layton found that local levels of economic inequality influence trust in government.Footnote 74 In Peru, a study of AmericasBarometer survey data found that perceptions of government efforts on economic issues like poverty and unemployment played an important role in predicting government legitimacy.Footnote 75
In all of these situations, citizens were evaluating not (only) the general economic health of the country, but more specific and localized benefits from government programs. Therefore, we suggest:
Hypothesis 2: Information revealed by transparency efforts will increase system support when it suggests the government is improving local socioeconomic well-being and will decrease system support when it suggests the government is decreasing local socioeconomic well-being.
EXPERIMENTAL DESIGN AND METHODS
We developed three separate, but related, experiments aimed at understanding how transparency and the information revealed by transparency affects the political attitudes of Peruvian citizens. The Peruvian government has spent much time and expense in recent years developing elaborate websites that contain a wealth of information about government activities and outcomes. The genesis of these efforts may be traced back to the scandals that rocked Pres. Alberto Fujimori’s final year in power in 2000. Most prominently, Fujimori’s intelligence chief was caught on camera apparently trying to bribe an opposition congressman and the video became widely viewed in Peru. Soon after, Fujimori went into exile and Congress declared him ‘morally unfit’ to govern. In 2003 the new government passed a law on transparency and access to information intended to help address corruption problems.Footnote 76 In 2010, the government created a centralized Transparency Portal that offers a standard method and website for providing government information to the public.Footnote 77
More than 2,500 distinct government entities – including the office of the president, various other entities in the executive branch, the legislature, the Supreme Court and other courts, autonomous organizations (for example, universities), regional governments, and various types of local governments – all provide information about themselves and their activities on the Portal. The main websites of the national government and many government entities provide prominent and permanent links to the Portal. In addition to these standardized web pages, some governmental entities have produced their own independent transparency websites. In this study, we relied on the independent transparency site produced by the municipality of Lima, where the experiments took place, and on a transparency site produced by the Peruvian Controller, a government agency, in cooperation with USAID and a local Peruvian NGO.Footnote 78 We chose them because they presented theoretically relevant information in graphically attractive ways. Around a quarter of our subjects indicated they had some familiarity with Lima’s transparency website. We did not ask about familiarity with the jointly sponsored website but we believe it would have been low as the website was new and was associated with the Controller, an agency that few Peruvians would have had any reason to interact with. Our theoretical reasoning and hypotheses do not depend on previous familiarity with the sites or its absence.
The information on these portals is vast. With little effort we were able to find information such as yearly expenses on particular public works projects, lists of government contractors and payment amounts, and the make and model of the personal vehicle of a particular city council member, among many other things. Information revealed by transparency can be either procedural or substantive.Footnote 79 That is, it can inform citizens about the way in which decisions are made or about the substance of those decisions. We focus in this article on the substantive dimension of information because that is the focus of the transparency portals we accessed.
Our experiments were designed to understand whether transparency websites and the information provided on them might produce changes in Peruvians’ political attitudes. The treatments consisted of showing respondents videos ranging in length from one-and-a-half to four minutes that displayed the websites and their information. Each video began by emphasizing the Peruvian government’s efforts to build these websites and hence we have no treatments where the information was divorced from the subject of transparency. Subjects viewed these videos on tablet computers and then answered questions posed by hired Peruvian surveyors. Twenty-five of these questions probed political attitudes while the remaining ten questions covered demographic information. We selected all the questions from the most recent editions of the AmericasBarometer and Transparency International surveys. These surveys are some of the most well-known and well-respected efforts to collect data on political attitudes, thus enabling us to make use of established, field-tested questions and to engage in comparisons with their findings. We conducted these experiments with a diverse array of people in Peru’s capital city of Lima in public areas, such as parks, shopping malls, and busy commercial streets, as detailed below.
In designing these experiments, we sought to preserve the nature of the websites as much as possible, thereby prizing experimental realism. Our videos showed multiple screenshots of the websites, preserving both the content and layout of the information. Our treatments are faithful to the sources and nature of the information found on the websites. In our experiment on sources, we use the actual sources named on a transparency website even though different sources may have been better known or produced interesting results. In our experiment on framing, we used the framing available on the transparency website, even though other information frames would be possible.
Experiment 1: Transparency Portal Overview
In this first experiment, we tested whether a direct relationship between transparency and legitimacy might occur. That is, would awareness of the government’s transparency efforts and their scope alter political attitudes? We emphasized the existence and extent of the Lima’s transparency information, but did not discuss the sources or content of the information.
We use two treatments and one control. The first treatment provided only a limited amount of information to the subject, with a video of 1 minute and 40 seconds in length. This video showed Lima’s main transparency portal and the types of information available there, including government investment projects, the mayor’s schedule, personal information about members of the municipal council, and government contracts and acquisitions. No specific information about these categories was displayed. The second treatment, with a video of 3 minutes and 40 seconds, provided a glimpse of detailed information available on the portal in each category. This video showed budget information and location of several government investment projects, details of the mayor’s daily schedule, the email addresses and résumé (curriculum vitae) of a member of the municipal council, and details on a government contract, including the name of the contractor and the amount paid. While this treatment included some substantive information, it had no particular frame but rather illustrated the type of details available on the website. The control group received the survey without viewing any information.
Experiment 2: Source of Information
In this experiment, we wished to understand whether the purported source of the information might influence subjects’ political attitudes. The experiment was composed of three treatments and one control. Each treatment, as well as the control, provided the same information on the municipality of Lima, including Lima’s budget per capita, illiteracy rate, infant mortality rate, the percentage of households with access to water and sewage, a rating of the government’s capacity to execute plans for investments, information on whether the government published required details on the participatory budget process, the rate of political participation of women, and information on the local government’s accountability to the national government. Where applicable and available on the website, the statistics for Lima were compared with the national average. Lima generally, but not always, compared favorably to other areas of the country on these indicators. This overall positive frame made it more likely that people would increase their positive feelings about the government. We did not test how the sources might affect negative information provided by the government, in large part because we valued realism, and government performance in Lima was relatively positive by measures reported on the transparency websites.
We accessed this information on the jointly sponsored website discussed above, a cooperative venture between the Peruvian Controller, USAID and an NGO. The videos for the treatment and control groups are identical except for the logo that is displayed as the source of the information and the narration for the video that calls attention to that logo. The first treatment attributes the information to the Controller, an office within the government of Peru that serves (in theory) as an independent auditor. The second treatment attributes the information to USAID, the foreign aid agency of the U.S. government. The third treatment attributes the information to Peru ProDescentralización, a little-known Peruvian NGO. The control does not provide a source of the information.
As Peruvian trust in government is so low, it follows that Peruvians would be skeptical of any data provided by their government via transparency initiatives even though (or because) the information was positive. It also seems unlikely that an unknown Peruvian NGO would be perceived as credible. However, when an independent actor, such as USAID, provides information to Peruvian citizens, it seems plausible that Peruvians would view it as more credible and legitimate. Therefore, we predict that the information attributed to USAID will increase support for Peru’s government.
Experiment 3: Framing Effects of Comparative Well-Being
In this experiment, with two treatment groups and one control group, we varied the framing of information about government performance to be more positive or more negative. We focused on the district of San Juan de Lurigancho (SJL), one of forty-three municipal districts in the province of Lima. All individuals were shown information on socioeconomic indicators for SJL in comparison with ten other districts in Lima. These indicators included the percentage of children under the age of 5 that are malnourished, the percentage of households with electricity, the percentage of households with access to water and sewage, the investment budget spent per capita, and the Human Development Index. We retrieved the information from the same website employed in Experiment 2. We emphasized and slightly altered the graphs on the website to emphasize SJL’s relative placement compared to other districts in Lima.
The first treatment group was shown information in which SJL performed better than ten other districts, providing a frame in which socioeconomic indicators related to government programs were relatively good. The second treatment group was shown the same information, but SJL was contrasted with other districts that performed better. In this condition, the framing suggests that socioeconomic indicators related to government programs are not performing well. The control group was shown the same information on the socioeconomic performance of the respondents’ district without providing a comparison to any other districts. While the information in these treatments reflects government performance, it cannot simply be reduced to that performance as the information is framed in particular ways and those frames should influence opinions.
The difference in content between Experiments 2 and 3 is important to note. The information in Experiment 2 was generally, though not always, positive, and it pertained to the entire Lima metropolitan area, which has a population of around 8.5 million with marked differences between districts. In Experiment 3, SJL has a population about one-tenth that size and its socioeconomic indicators fall in the middle of Lima’s inter-district range. Moreover, the information focused exclusively on basic, everyday concerns such as child malnutrition, electricity and water. In the previous experiment, the information also included indicators less relevant to daily needs such as whether the municipal government was complying with nationally defined bureaucratic processes.
We selected respondents randomly from those walking down a major street in the district, several blocks from the offices of the municipality. The street has a variety of small shops, stores and street vendors and was relatively busy. However, it was not a shopping or business destination for other residents of Lima. A large shopping mall and major bus routes that may have attracted non-residents of SJL lay several minutes by taxi from this particular street. It is worth noting that SJL is comparatively geographically isolated within Lima by the Rímac River on the south and by hills to the east and west. It is most easily accessible by public transportation from downtown Lima rather than from other areas of the city. It thus seems unlikely that non-residents of SJL – or those without vested economic interests in the area by way of their business – were roaming the street in which we administered the survey, though we cannot rule it out as we did not ask for the district of residence of our respondents.
Survey Design and Administration Method
Our dependent variable, system support, was measured by a survey consisting of fourteen questions about a variety of political attitudes. These survey items were mostly taken from AmericasBarometer, a survey conducted every other year through the Latin American Public Opinion Project (LAPOP) at Vanderbilt University. For each control and treatment group within each experiment, participants viewed the corresponding video in its entirety before responding to the survey questions. In efforts to curb framing effects caused by item ordering, we randomized the order in which the survey items appeared. We also included a battery of demographic questions at the end of the survey.
The surveys administered to the treatment groups for Experiments 2 and 3 included an additional manipulation check question designed to decipher which respondents had absorbed the treatment. In Experiment 2, respondents were asked to identify the source of the information they were shown (the Peruvian government, USAID, or the Peruvian NGO); and in Experiment 3, they were required to report whether San Juan de Lurigancho had been compared favorably or unfavorably to other areas in the presentation they saw. Though there was some variance among treatments and experiments in the rates at which respondents answered the manipulation checks correctly, well over half of respondents were able to identify correctly the treatment they had received (see Appendix Table 10).
We conducted these experiments in Peru’s capital city of Lima in June and July 2013. We hired and trained a total of twenty Peruvian nationals to conduct the surveys, mostly students aged 20+ studying at a variety of technical institutes. Each enumerator was given a tablet computer and assigned a randomized treatment and a randomized selection number. The enumerator then approached the nth person that passed by according to the selection number. If the enumerator was rejected (the person they approached chose not to participate), he or she would then approach the next nth person. Enumerators repeated this process until they were successful in administering a survey. At the end of the treatment video and survey, the subjects were offered three Peruvian soles (the rough equivalent of US$1). The enumerators were then assigned a new random treatment and random selection number, repeating the process.
With an average of ten to twelve enumerators per day working over the course of roughly two and a half weeks, we conducted 1,431 surveys. The average survey lasted about fourteen minutes, though there was some variation across experiments. These surveys took place in four different locations within Lima: a public park in the district of Miraflores, a busy commercial street in the city’s center, a middle-class neighborhood in the district of San Juan de Lurigancho, and outside of a popular mall, ‘Plaza Norte’, in the northern region of Lima. We utilized the SJL location only for Experiment 3, which compared SJL to other districts. The conditions were not optimal for viewing and listening to videos, but most subjects seemed to be able to see them and understand them, as demonstrated by our manipulation check. Subjects faced distractions such as loud background noise, street performers, constant movement around them, and – on one memorable afternoon – tear gas from a nearby demonstration.
Some considerations with regards to both external validity and internal validity bear mentioning. First, our experiments were carried out exclusively in Lima and thus we make no claims about generalizing to the population of Peru as a whole or to other countries. Even though we utilized a ‘convenience’ sample of passers-by in public spaces, our sample is not much different from the nationally representative AmericasBarometer survey (see Appendix Table 1). We include demographic controls in some models of our analyses to help correct for random imbalances between treatment and control groups within a given experiment (see Appendix Table 2–4 for demographic details on each experiment). As mentioned above, the experimental setting was not ideal; clearly, a busy city street is not a prime survey location. It is possible that differing levels of street activity from day to day could have affected respondents’ attentiveness. However, we control for fixed effects associated with the day on which each survey was conducted and find that our results remain unchanged. We also controlled for interviewer effects, also without changing the findings substantially. Finally, we were conservative in our estimates of the treatment effects by selecting a relatively large control group in comparison to each treatment group (Appendix Table 5).
Dependent Variables: Constructing Indices
Previous research suggests that system support is a multi-dimensional concept.Footnote 80 As mentioned, we identified four dimensions of interest: National Political Community, Regime Performance, Confidence in Regime Institutions, and Local Government. Exploratory factor analyses suggested the presence of these dimensions in our survey data.Footnote 81 We constructed indices for each dimension comprised of three to six questions each. To facilitate index construction, we rescaled individual questions from 0–100 (the same scale used by AmericasBarometer), where 100 represents higher levels of approval or trust. Confirmatory factor analysis of these indices in MPlus provided strong evidence that each index measures a single dimension and that the included indicators loaded well onto the index. See the Appendix (Tables 6–9) for the specific questions used in these indexes.
Statistical Analysis
We conducted three types of analysis in order to determine the extent to which treatments affected dependent variable outcomes: (1) difference of means tests, (2) standard OLS regression (with variations), and (3) calculation of the complier average causal effect (CACE). In this article, we report on our preferred model of the OLS regressions that include both treatments and demographic controls (gender, age, race, education level, and income level). For space reasons, we only report the coefficients for the treatment effects; the full set of results for the OLS tests as well as for the other tests may be found in Appendix Tables 11–22.
The CACE is a measure that represents the average effect of a treatment on an individual who absorbed the experimental treatment.Footnote 82 The CACE was calculated by dividing the difference between the control and treatment means for a given dependent variable by the proportion of the treatment group respondents who were ‘compliers’, that is, those who correctly answered the manipulation check questions (see Appendix Table 10). Calculation of the CACE allows for isolated analysis of treatment effects when information in the treatment is absorbed and understood.
FINDINGS
Experiment 1: Transparency Alone Is Insufficient
In Experiment 1, subjects watched brief videos that provided an overview of the official transparency portals for the municipality of Lima. The results suggest that, in the Peruvian context, transparency has no direct effect on individuals’ evaluations of their government or the extent to which they believe the government is trustworthy. Table 1 presents the results. As shown, neither a brief overview of the portals nor a somewhat longer, more detailed overview had a statistically significant effect on the indices, with the exception of National Political Community in the case of the detailed overview. This result is robust to all model specifications. It is possible that respondents felt some pride in the nation-state as they viewed detailed information about their country. Generally, however, if the Peruvian government seeks to improve its standing with its constituents, it appears that transparency efforts alone are insufficient. At the same time, it is important to note that for the detailed overview treatment, the size of the coefficients are much larger (compared to the brief overview) and that the difference of means tests (Appendix Table 11) show that system support increases about 2–5 points on a 100-point scale when compared to the control group. It is possible that transparency is having a small direct effect but that our sample size is too small to pick it up.
Experiment 1, Regression Results
Notes: Controlling for gender, age, education, income, and race; robust standard errors in parentheses.
***p<0.01, **p<0.05, *p<0.1.
Experiment 2: Source Credibility Matters
Experiment 2 indicates large differences in public response to information depending on the source to which the information is attributed. Interestingly, when presented with socioeconomic indicators attributed to USAID, Lima residents reported attitudes and evaluations of the Peruvian political system that were significantly more positive than the control group responses. By contrast, when the Office of the Peruvian Controller or an unknown NGO was identified as the source, respondents’ attitudes did not change with regard to any of the dependent variables. This contrast seems to indicate that (1) Peruvians attribute differing levels of credibility to different sources; (2) Peruvians do not consider their government to be a credible source, even when a semi-independent office such as that of the Comptroller is mentioned; and (3) transparency information does, in fact, appear capable of improving respondents’ perceptions of their government, but only when the information is attributed to a source previously considered credible.
Importantly, then, a government considered less-than-credible seemingly cannot foster credibility for itself on its own; we suspect that if the public does not consider the government a credible source, even positive information will likely do little to foster improved appraisals of governmental performance. In general, based on these results, we believe that a credible third-party endorser is required to alter respondents’ political attitudes. Table 2 presents the regression results. Results from other models (reported in the Appendix) are qualitatively similar, though in the fixed effects model the USAID treatment sometimes loses its statistical significance as the coefficient shrinks slightly.
Experiment 2, Regression Results
Notes: Controlling for gender, age, education, income, and race; robust standard errors in parentheses.
***p<0.01, **p<0.05, *p<0.1.
The effect size for the USAID treatment is relatively large. Control group means for National Political Community, Regime Performance, and Local Government range between 40 and 52 on our 100-point scale. The USAID treatment raises the index levels by approximately 6 points in every case. If one calculates the CACE, which accounts for those who actually received the treatment, levels go up by 7–8 points. When compared with the standard deviations of each corresponding index in the control group, these increases represent about 38–47 per cent of those values. We believe increases of nearly half of a standard deviation are substantively large.
Experiment 3: Information Frames Matter
Experiment 3’s findings suggest that the framing of substantive information matters. In particular, we found that individuals’ attitudes toward their government improved when they were shown that their community was benefitting relative to others. As Table 3 depicts, residents responded with improved perceptions of their government on the dimensions of National Political Community and Confidence in Regime Institutions. Results from other models (reported in the appendix) are qualitatively similar, though in the fixed effects model Confidence in Regime Institutions loses significance as the coefficient decreases slightly and Local Government gains significance as the coefficient increases slightly. However, those who were shown information about their community not performing well showed no difference from the control group.
Experiment 3, Regression Results
Notes: Controlling for gender, age, education, income, and race; robust standard errors in parentheses.
***p<0.01, **p<0.05, *p<0.1.
Substantively, these increases are even larger than those in the previous experiment. The control group means for National Political Community and Confidence in Regime Institutions are about 50 and 45 on our 100-point scale. The treatment of ‘neighborhood compares well’ increases those averages by about 7 points for each index. The CACE calculates the treatment increase as about 10 and 11 points respectively. For each index, this increase is about 55 per cent of the standard deviation of the control group, a substantively large change.
Generally, this experiment suggests that information frames matter in shaping public opinion. While this general result is hardly news, it is more interesting to note that a frame demonstrating the relative well-being of one’s community can improve some dimensions of system support. No previous scholarship on frames or system support has examined such a linkage. In the control group for this experiment, the interpretation of the data was less clear and the frame thus less strong. We simply presented five socioeconomic indicators and briefly showed a figure placing SJL in the context of all of the forty-two other districts in Lima. Even someone paying close attention would have been hard-pressed to surmise much more than the fact that there was plentiful data available by district. The first experiment in this article also lacked strong frames in both treatment and control conditions, again taking the approach of providing a lot of raw information in a relatively short period of time. Hence, we cannot say which frames work the best, but we can say that this particular frame alters system support. As previous scholarship on system support has focused mostly on individual and national levels of analysis, it seems possible that scholars have overlooked an important level of analysis – one’s own community – in thinking about the bases of system support.Footnote 83
DISCUSSION
Generally, we found no direct relationship between transparency (in itself) and political attitudes. We did find that information provided by transparency affected legitimacy, depending on its source and content. Why do residents of Lima trust their government more when the information is endorsed by USAID or when the information is framed in a way that associates the government with community-specific comparatively good socioeconomic outcomes? While it is difficult to say with certainty, we can bring some evidence to bear on these questions.
It seems unlikely that most Peruvians know USAID by name. Rather, they are likely to rely instead on their views of the United States generally when making judgments about USAID-endorsed information. USAID has made strong efforts to improve branding in recent years and has had some success at increasing public knowledge of the agency and its mission.Footnote 84 Still, it seems unlikely that most Peruvians possess any knowledge of or views about the agency. On the other hand, almost all Peruvians would have some views about the United States. The USAID logo clearly utilizes the letters ‘US’ and ‘USA’, which are well-known acronyms of the United States in Latin America. Our video presentation stated clearly both at the beginning and at the end that USAID is an agency of the United States government. Hence, it would have been easy for respondents to associated USAID with the United States. For Peruvians who do have an understanding of USAID, we expect their views of USAID to be correlated with their overall views of the United States in any case.
How, then, do Peruvians feel about the United States? Baker and Cupery report that about 80 percent of Peruvians hold favorable attitudes toward the United States, higher than any other country in South America.Footnote 85 Moreover, they find that aid is a significant predictor of pro-American attitudes in Latin America.Footnote 86 Pew research surveys put the US favorability percentage among Peruvians at 65 percent, the global median.Footnote 87 Perhaps an even better measure is an AmericasBarometer question about the trustworthiness of the United States. In 2012, Peruvian trust in the US government stood at 53.6 on a 100-point scale.Footnote 88 That level of trust was higher than Peruvian citizens’ trust in their own president (52.9), armed forces (52.0), municipal governments (41.2), Supreme Court (40.3), national police (40.1), justice system (39.4), Congress (35.5) or political parties (32.0). It was lower than trust in the Catholic Church (60.5), the media (58.7) or ‘elections’ (55.2).Footnote 89 The survey did not ask about trust in the government generally or in the particular arm of the government utilized in our survey, but one can infer that trust in the Peruvian government generally would be lower than trust in the United States government because most elements of the Peruvian government (Congress, justice system) do not fare well.
It is also possible that respondents were influenced not by good feelings toward or trust in the United States, but rather in its credibility as an outside party with conflicting interests. Most of the information shown in our treatment reflected positively on the city of Lima, including low illiteracy and infant mortality rates, and high access to water and sewage, among other indicators. In 2013 when we conducted our survey, both the mayor of Lima (Susana Villarán) and the president of Peru (Ollanta Humala) were on the left of the political spectrum and expected to implement policies and programs unfriendly to US preferences. Hence, apparent US endorsement of socioeconomic achievements in Lima could have sent a credible signal to respondents that Peruvian government institutions and individuals could be trusted.
It is also unlikely that Peruvians knew much about the other two specific entities in our treatments, the Controller, the primary auditor of government expenditures, or the NGO, PeruProDescentralización. As with USAID, we suspect that our respondents would have relied on attitudes about associated entities as they encountered the names of these organizations. In the case of the Comptroller, we noted in our treatment that it was part of the Peruvian government and in the case of PeruProDescentralización, we simply noted that that it is an NGO. Hence, we expect that we are examining the results of respondents’ general views of the Peruvian government and of NGOs. We interpret the absence of change in respondents’ attitudes as a general unwillingness to trust either the Peruvian government or unknown NGOs. It is also possible that Peruvians understood the NGO to have a particular political agenda – namely, decentralization, which is the meaning of its name. In this case, some may have found it more credible and some less credible, depending on their views of decentralization.
We chose these organizations not for their visibility but to maintain the realism of the experiment. The transparency website we relied on most heavily was in fact produced by these three organizations. The control group viewed unsourced information. It is possible that control group respondents assumed that the information source was the government, thus explaining the lack of change in results when we specifically referenced the government as a source. With respect to the NGO condition, it is perhaps too broad a category to be of any assistance to respondents in interpreting the credibility of the information. Further research might investigate what would happen if the data source were a more well-known organization.
Turning to the issue of framing, we utilized socioeconomic indicators that would be meaningful and important to most Peruvians and that they are likely to associate with government institutions and programs. Thus, we expect that as they evaluate how their municipality compares with others, they are also judging the quality of their government, both at the local and national level, as the two levels work together to implement the relevant programs in Peru.
The first indicator in our treatment videos concerned the percentage of malnourished children under 5 years old. Since the 1980s, the national government has sought to combat malnourishment through well-known programs like ‘Glass of Milk’, which serves children under 6.Footnote 90 In the mid-2000s, these programs dramatically increased in scope and visibility. The national government adopted a conditional cash transfer system, JUNTOS, aimed at the poorest Peruvians, with one of its chief goals a decrease in child malnutrition.Footnote 91 Around the same time, it adopted a National Strategy for Poverty Reduction that broadened the approach to malnutrition and that involved decentralization of important programs to local municipalities. A new public–private partnership, the Child Malnutrition Initiative, was established and won public commitments from all presidential candidates in the 2006 election to work on malnutrition.Footnote 92 The winning candidate, Alan Garcia, made it a centerpiece of his first hundred days in office. Problems arose as municipalities either lacked the capacity or the political interest to implement national programs to combat malnutrition, and the success of the programs clearly depends on co-operation between local and national governments.Footnote 93 When viewing indicators of childhood malnutrition in their municipality, Peruvians would naturally associate those with government performance and services.
We make a similar conclusion about the other indicators used in our treatments. Water is a particularly important problem in Lima, which exists in a very dry desert. Access to water was included in the five socioeconomic indicators in our treatment video. The Peruvian government tried to privatize water services in the mid-1990s, but only partly succeeded. Access to water in Lima is thus determined by a set of overlapping interests, including private companies and municipal and national governments.Footnote 94 Residents of Lima tend to see access to water as a political game in which the well-off benefit and the poor lose.Footnote 95 While private companies provide water, government officials determine the process by which those companies bid for contracts, which areas will be served, and the building of infrastructure like water treatment plants. High-profile corruption scandals involving cabinet ministers and utility officials receiving bribes and kickbacks for water services feed public perceptions that the provision of household water services is fundamentally a political decision.Footnote 96
In brief, we interpret the increase of trust in Experiment 3 to be the result of respondents associating the relatively good socioeconomic outcomes they see in the information frame with relatively good government performance. Why does system support not decrease when respondents see information showing relatively poor socioeconomic outcomes? One possibility is that they expect poor outcomes and so when they observe what they expect to observe, they do not change their opinions of their government. In mid-2013 when we conducted our survey, expectations about Pres. Humala’s government were quite low in SJL. In the 2011 election during the two-candidate runoff phase, only 49 percent of the municipality’s residents voted for Humala, though this was higher than the 42 percent who voted for him in Lima generally. From April to July 2013, a series of problems drove down his approval rating by 20 points to the mid-30s.Footnote 97 In this context, specific good news about the community’s socioeconomic indicators increased system support, though not support for Pres. Humala.
CONCLUSIONS
A difficult problem in many developing countries is a corrosive lack of confidence in government institutions. While skepticism of government motives and activities undoubtedly plays a healthy role in any democratic system, hard-core cynical attitudes go beyond reasonable skepticism because they are disconnected from actual government performance.Footnote 98 Latin American and Eastern European countries have been particularly susceptible to these ‘cycles of despair,’Footnote 99 in which states manage resources relatively well but citizens do not recognize or believe that to be the case. Citizen belief in hapless and corrupt governments can be a self-fulfilling prophecy in which government actions cannot make any progress as no one believes they will do any good. If everyone believes government officials are corrupt, it is easier for them to be corrupt. Likewise, governments might have to allocate resources away from development issues and social problems in order to simply maintain stability and power.Footnote 100
Transparency is an oft-recommended remedy for many government ills. We wanted to know whether transparency and the associated information provided by that transparency might help break this cycle of cynicism. The Peruvian government has undertaken extraordinary transparency efforts on the internet, producing a flood of information about its activities. Peruvian citizens are also among the most cynical in the world; they predict their government will perform 30 percent worse than it actually does.Footnote 101 Of the seventy-two states in Gilley’s study, only seven have a more cynical citizenry.
Given this level of cynicism, it is perhaps not unexpected that transparency information on its own did little to build system support in Peru. At the same time, we found that information provided by transparency can increase system support, depending on its source and content. Generally positive information about government performance in the Lima region increased system support if the information was endorsed by a credible third party. Clearly positive information about government performance in a specific local community, on issues relevant to daily well-being such as childhood nutrition, water and electricity, also increased system support. The changes in system support were quite dramatic. When USAID was reported as the source of transparency data, system support increased from 6–8 points (on a 100-point scale) in all the indexes, or nearly half of one standard deviation of those indexes for the control group. Evaluations of system support increased by 7–11 points for two of the indexes, or a little more than half of a standard deviation, when citizens received information showing that they benefitted relative to other Lima communities. The results here are asymmetrical, however, contrary to our hypothesis. If citizens received information that their government was performing poorly in their community, the information did not decrease system support compared to the control group. We suspect that this is due to preexisting low levels of system support. When the public’s expectations of government performance are already so low, evidence of poor performance does not decrease system support.
These findings are interesting but do not necessarily bode well for governments trapped in cycles of cynicism. Our results suggest multiple paths but relatively high bars for governments seeking to increase system support. They not only need transparency and good performance; they also need a credible endorser or good community-specific information on important socioeconomic indicators. It may be quite difficult to identify credible third-party endorsers without yielding (or appearing to yield) some control over important government data and programs. It may also be psychologically difficult for government officials to ask for or accept outside endorsements. Moreover, if citizens are only satisfied by information suggesting that they are benefitting relative to nearby communities, then system support can only increase if some citizens improve their well-being while others do not, of if they improve more rapidly than others are improving. Such scenarios may be undesirable for a number of reasons. We did not test whether third-party endorsements or positive socioeconomic news would increase system support on their own – for example, if this information were encountered through the news media – without government transparency websites. It is possible that the information on its own would have a positive effect even without a government transparency program, but it also seems less likely that citizens would ever encounter such information.
Our findings have important implications for research on transparency, information and system support. To the best of our knowledge, this is the first study to examine the effects of government transparency websites and some associated information on political attitudes in a developing country. While transparency is widely applauded and supported in principle, we know relatively little about its effects and need to learn more. Governments do not (solely) undertake transparency to boost system support, but the link between these factors is plausible and deserves further exploration. Peru is not alone in its efforts to increase transparency, nor is it alone in the problems it faces with credibility and weak support for democratic institutions. Future experiments should identify the effects of information on its own (in the absence of government transparency efforts), of different types of information, and of the interaction between source credibility and framing (as we tested each dimension on its own).
With respect to source credibility and framing, scholars have repeatedly found that source credibility facilitates persuasion, but they have not fully specified the characteristics that make a source credible. This study raises the intriguing possibility – not previously explored in this context – that one government could serve as a credible source for another with respect to the trust of its own citizens. In international relations, credible third parties are frequently essential for conflict resolution among two warring parties, including civil wars. Perhaps credible third parties could also ease problems created by corrosive mistrust in everyday political life or, alternatively, call into question the credibility of governments whose citizens are too trusting. Finally, this study makes an important contribution to scholarly knowledge about framing government performance. Most of the literature examines government performance in different issue areas (e.g., political vs. economic) or asks whether citizen attitudes are influenced by pocketbook or sociotropic measures. Our study, by contrast, suggests the importance of local community well-being in increasing system support.
