IDB WORKING PAPER SERIES Nº IDB-WP-962

Transparency and Trust in Government:

Evidence from a Survey Experiment

Martin Alessandro
Bruno Cardinale Lagomarsino
Carlos Scartascini
Jerónimo Torrealday

Inter-American Development Bank

Department of Research and Chief Economist

February 2019
 Transparency and Trust in Government:

Evidence from a Survey Experiment

Martin Alessandro*
Bruno Cardinale Lagomarsino**
Carlos Scartascini**
Jerónimo Torrealday*

* Ciudad Autónoma de Buenos Aires

** Inter-American Development Bank

February 2019
Cataloging-in-Publication data provided by the
Inter-American Development Bank
Felipe Herrera Library

Transparency and trust in government: Evidence from a survey experiment / Martin

Alessandro, Carlos Scartascini, Bruno Cardinale Lagomarsino, Jerónimo Torrealday.
p. cm. — (IDB Working Paper Series ; 962)
Includes bibliographic references.
1. Transparency in government-Argentina-Surveys. 2. Information behavior-
Argentina-Surveys. 3. Government information-Argentina-Surveys. I. Alessandro,
Martín. II. Scartascini, Carlos G., 1971- III. Cardinale Lagomarsino, Bruno. IV.
Torrealday, Jerónimo. V. Inter-American Development Bank. Department of Research
and Chief Economist. VI. Series.
IDB-WP-962

http://www.iadb.org

Copyright © 2019 Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 Attribution-
NonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/
legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No
derivative work is allowed.

Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to
the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be
subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license.

Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised
version of this work may also be reproduced in any academic journal, including those indexed by the American Economic
Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore,
the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such
restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license
and these statements, the latter shall prevail.

Note that link provided above includes additional terms and conditions of the license.

The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American
Development Bank, its Board of Directors, or the countries they represent.
 Abstract
Does providing information improve citizens’ perception about government
transparency? Does all information matter the same for shaping perceptions
about the government? This paper addresses these questions in the context
of an online randomized survey experiment conducted in Argentina. Results
show that providing information to citizens matters for shaping perceptions
about transparency, and the content of the information matters for affecting
the evaluation people make about the government. Those who received a
“positive” treatment (showing that the government was over-performing on
its promises) increased their trust in the government more than those who
received a “negative” treatment (showing that the government was under-
performing). The evidence highlights that the channel between transparency
and trust may be mediated by the performance of the government.1

JEL classifications: C38, C83, C99, D83, D90, H11, H40

Keywords: Survey experiments, Information, Beliefs, Trust

1
We thank Martin Ardanaz, Cecilia Guemes, Philip Keefer, Jorge Streb, Mariano Tommasi,
Razvan Vlaicu, discussants and participants at the LASA and AAEP annual meetings, and an
anonymous reviewers for their helpful comments and discussions.

1
1 Introduction
In the last few decades, trust in governments has been in decline (OECD, 2018).
This is particularly true in Latin America, where corruption scandals and poor per-
formance of democracies have eroded public confidence in institutions for years.2
Interpersonal trust is also very low.3 Lack of trust has clear negative welfare conse-
quences (Levi and Stoker, 2000; Cahuc and Algan, 2013). Trust is necessary for key
economic activities, most notably finance, and for the success of government policies
that require and depend on cooperation and compliance of citizens (OECD, 2017).
Moreover, if citizens do not believe the government will fulfill its promises, or that
their fellow citizens will join them in voting against governments that default on
their promises, they will not demand policies that expand the provision of public
goods. If citizens do not trust that the government will follow through with those
policies that require upfront costs to reap large future benefits, they will not demand
long-term investments (Keefer et al., 2018).
Because part of the erosion of trust has been correlated with economic and
financial mismanagement, corruption scandals, and growing inequality, increasing
transparency in government has usually been touted as part of the solution. Trans-
parency, or “the disclosure of information by an organization that enables external
actors to monitor and assess the internal workings and performance” (Grimmelikhui-
jsen et al., 2013, p.3) has been pushed by international organizations, government,
and donors as standard practice.4 As such, increasing transparency and building
trust could go hand in hand (OECD, 2017). Trust is a psychological state involv-
ing positive confident expectations about the competence, benevolence, honesty and
predictablity of another person or organization, and the willingness to act on the
basis of these expectations (Welch et al., 2004; Adams, 2005; Ciborra, 2005; Tolbert
and Mossberger, 2006; Pina et al., 2007, 2010; Bertot et al., 2010; Grimmelikhuijsen,
2012b; Grimmelikhuijsen et al., 2013; Andrews and de Walle, 2013). More informa-
tion could provide citizens with evidence on government competence. It could also
provide a window from which to evaluate government values. The delivery and the
2
For example, historical data collected by Latinobarometer show that average confidence in
Congress has stagnated at around 30 percent and confidence in government around 40 percent for
the last 20 years. Data from the Americas Barometer Survey conducted by LAPOP in 2016/2017
shows that more than 50 percent of the population had very low trust in the president in Paraguay,
Mexico, Venezuela, and Brazil. Keefer et al. (2018) summarizes this evidence.
3
According to evidence from Latinobarometer, interpersonal trust has been hovering around 20
percent for the last 20 years.
4
Transparency plays a central role in the Paris Agreement signed in 2016. Organizations such as
Open Government Partnership or Transparency International are good examples: they share goals
and aim to secure concrete commitments from governments to promote transparency, empower
citizens, fight corruption, and harness new technologies to strengthen governance. There are
multiple definitions of transparency.

2
content of that information may matter too.
There is ample evidence that most people have limited time and do not follow
government actions closely because acquiring accurate information or processing
publicly available information is costly, and the probability of decisiveness in an
election or public matters tends to be negligible. Therefore, most people tend to
remain ignorant and do not pursue information, but they react rationally once they
receive information about the workings of government (Downs, 1957; Brennan and
Lomasky, 1997; Congleton, 2001; Martinelli, 2007; de Leon et al., 2014).5 As such,
while information matters in shaping public perceptions and voting choices, that
information cannot sit idle in a website but rather be delivered to citizens (Ferraz
and Finan, 2008; Mullainathan et al., 2010; Keefer and Khemani, 2012; Chong et al.,
2014; Cruz et al., 2016).
Because trust in institutions depends on competence and values, the informa-
tion provided to the public should allow citizens to update their beliefs about how
reliable, responsive to needs, and efficient the government is. It should also allow cit-
izens to evaluate the government’s integrity, openness and fairness (OECD, 2017).
Providing factual knowledge about government performance outcomes and about
the intentions of the government could be an important way of increasing citizen
trust in government (Grimmelikhuijsen, 2012b). Therefore, an important question
is whether any type of information matters, if information about results matters,
and whether subconscious and affective cues are also important (Grimmelikhuijsen,
2010, 2012b; Grimmelikhuijsen et al., 2013). Evidence is still scant of the strength
of the link between the content of the information (performance) and the framing
of the information, and the particular conditions under which transparency fosters
trust, particularly in the context of developing countries and young democracies.
The use of an experimental setting could help to yield more light into the existence
of the link and its determinants and generate usable knowledge for practitioners
(Levi and Stoker, 2000; Van de Walle and Bouckaert, 2014; Bouwman and Grimme-
likhuijsen, 2016; Cucciniello et al., 2017). Performing the experiment in a country
where overall levels of trust are very low expands the existing evidence and the
understanding of the phenomenon (Cucciniello et al., 2017).
In this paper, we test two hypotheses and two channels through which trans-
parency can affect citizens’ perception of the government. First, we evaluate whether
the provision/disclosure of information changes individuals perceptions about gov-
5
Gingerich and Scartascini (2018) show that because rationally ignorant individuals do not
validate the accuracy of any new information they (passively) receive they are not necessarily able
to separate the signal from the noise: their policy preferences are as likely to be driven by bad
information as by good. This state of affairs leaves them highly vulnerable to the machinations of
political entrepreneurs and other elites. In this perspective, voters act upon their beliefs in more
or less a rational way, but those beliefs are formed based on incomplete or biased information.

3
ernment’s transparency.6 Second, we evaluate different characteristics of that dis-
closure that could lead to more trust: i) whether the framing of the information
matters,7 and ii) whether the content of the information matters in terms of perfor-
mance. For testing these channels, we implemented a survey experiment on a sample
of 1,999 individuals living in the City of Buenos Aires, Argentina.8 First, partici-
pants were allocated at random with equal probability to receive information about
a series of commitments the government had made to the citizens of Buenos Aires.
The commitments are tracked, and results are publicly available on the webpage of
the city government. Second, participants were randomly allocated to a vignette
that highlighted the government pledge to efficiency and good management or to
a vignette that highlighted an empathetic message, where the government pledged
making life better for the inhabitants of the city. Additionally, individuals received
information showing either that the government was over-performing on its commit-
ments or that the government was under-performing. For evaluating perceptions of
transparency and trust we use a multidimensional approach that includes the many
components of trust: competence, benevolence, honesty, and confidence (Grimme-
likhuijsen, 2012b).
Results show that providing/disclosing information increases the perception about
how transparent the government is by about 8 percentage points, which implies an
increase of more than 10 percent. Does the framing or the content matter relatively
more for changing citizens perceptions about the government? Results show that
there are no relative differences in trust between an “efficiency-based” framing com-
pared to an “empathy-based” framing. However, differences in performance seem
to matter. The group that received information where the government was over-
performing on its goals increased trust significantly more than those who received
information that the government was under-performing on its goals. The difference
is equivalent to about 0.10 of a standard deviation.
These results have important implications. First, they highlight once more the
importance of actively providing information to citizens as a way to enhance trans-
parency. Second, they highlight that individuals may not care that much about
the framing of the message–they trust the government the same if the government
6
This could be seen as a tautology. However, the existence of more information and positive citi-
zen perceptions of government transparency do not necessarily go hand in hand (Grimmelikhuijsen,
2012b,a).
7
As will be clear later, we take a broad definition of “framing.”
8
A survey experiment is an intervention in which vignettes randomly vary by subject. Because
of random assignment, any variation on measures can be attributed to the vignettes (Nock and
Guterbock, 2010). The methodology has a long tradition in research but has gained popularity
recently because of decreasing costs (see Bouwman and Grimmelikhuijsen (2016) for some recent
examples.) In our case, we have tried to use a simple design that evaluated an actual policy setting,
but we have diversifed in terms of the context to provide evidence from a developing country setting
following the recommendations in Bouwman and Grimmelikhuijsen (2016).

4
makes explicit its desire to pursue efficiency or fairness–but they do care about the
content of the message, particularly about whether the government complies with
its promises or not. This result has important implications. On the one hand,
it should provide incentives to governments to exert extra effort to comply with
goals, given that it is going to be rewarded by the citizens. On the other, it could
generate incentives for politicians to report only good news. This behavior could
backfire in the long run as citizens learn about this deception and decide not to
trust government-provided information any longer.
The article proceeds as follows. Section 2 provides some background on Buenos
Aires and presents the survey experiment design. In Section 3 we describe our
sample, discuss our identification strategy, and show results. Section 4 concludes.

2 The Survey Experiment

2.1 Background
The Autonomous City of Buenos Aires (henceforth CABA), is the capital city of
Argentina. CABA is the most populous city in the country with almost 3 million
inhabitants in the urban area.9 Although its first foundation dates to the sixteenth
century, only in 1996 were citizens able to elect a chief of government for the first
time. Before the constitutional amendment that granted autonomy to the city in
1994, the mayor was directly appointed by the President. Since 2011, the administra-
tive division of the city has consisted of 48 neighborhoods grouped in 15 communes.
Each commune works as a territorial administrative and political entity. Electoral
sections and sanitary regions, for instance, are aligned with communes’ borders. In
addition to increasing decentralization and greater citizen participation, the city has
also been engaged in a process of increasing transparency.10 As part of the process,
the current Mayor of Buenos Aires issued a series of government commitments with
the citizens of Buenos Aires upon taking office. These commitments were clear and
measurable goals across many government areas and based on citizens priorities and
the Sustainable Development Goals of the United Nations (UN). The city has re-
ceived several awards because of the initiative. Today, there are more than 50 goals
that are explained and tracked on the city government webpage.11
9
Data from 2010 National Census.
10
A steady process of administrative capacity–building over the last decade or so has provided
the foundations for increasing transparency. The evidence is well-aligned with the framework
presented by Grimmelikhuijsen and Welch (2012).
11
The government commitments can be accessed at http://www.buenosaires.gob.ar/compromisos.
One example of a commitment is making 100 percent of streetcorners handicapped accessible.
Another commitment is installing 10 thousand security cameras around the city and public
transportation.

5
2.2 About the Survey
We designed an online questionnaire to elicit information about transparency, and
about perceptions of the competence, benevolence, and honesty of the government,
along with citizen confidence in the government.12 The questionnaire additionally
included questions for classification purposes. Data collection took place between
December 2017 and January 2018. The sample was stratified with quotas by gender,
age (18 years or older), socioeconomic status (NSE), and commune. Within strata,
individuals were selected at random until the desired sample sizes were achieved. A
total of 1,999 complete interviews were carried out by a company that specializes in
collecting online survey data.13
Participants were assigned to randomized treatments twice during the survey.
First, they were randomized between treatment and control. Those in the control
group were asked about their perceptions regarding transparency of the city gov-
ernment before they were shown the informational treatments about government
commitments. Those in the treatment group also received those questions, but only
after they were shown the informational treatments about the government com-
mitments. Then, all participants were assigned at random to three different infor-
mational treatments (vignettes) with the same structure. Each vignette describes
the policy commitments made by the government, gives specific examples of the
commitments, and shows how well the government has been doing in the implemen-
tation of the policy according to the benchmarks and objectives it had established
at the beginning of the mandate. Vignettes vary in two respects. First, the fram-
ing of the message is different. There is a framing that emphasizes the role of the
public commitments for improving efficiency and public management (“efficiency”
treatment) while the other emphasizes their role for improving lives of citizens in
the city (“empathy” treatment). Second, the government’s performance is different.
The government could be either under-performing or over-performing relative to the
benchmarks the mayor set for the administration. The vignette presents a short and
simple message to keep the attention of the participant but at the same time pro-
vides enough information for understanding the initiative. Because it would make
no sense to combine a message about efficiency with under-performance, instead of
working with a 2x2 design we ended up with three different treatments as described
in Table 4.14 The three treatments were the following: T1: empathy and positive
results, T2: efficiency and positive results, and T3: empathy and negative results.
In the empirical analysis, we evaluate the effect across comparable pairs: T1 and
12
We follow the framework of questions and concepts introduced by Grimmelikhuijsen (2012b).
Appendix B presents the survey module. The full questionnaire in Spanish is available upon
request.
13
For details about the company, see Appendix C.
14
See examples of vignettes in Figures 4, 5, and 6 in the Appendix.

6
T2 (performance is constant), and T1 and T3 (framing is constant).
In the survey, immediately after the module describing the commitments and
the assigned vignette, each participant was asked about their perceptions about the
government. This module consists of 11 questions designed to elicit information
about the components of trust and confidence in the government. Responses to
these questions are part of the set of dependent variables (questions are available in
Appendix B). The full description of the survey experiment -timeline- is presented
in Figure 1.

3 Data and Empirical Analysis

3.1 Data Description
Table 1 (and Table 5 in the Appendix) present summary statistics of the main
variables for the sample of respondents to our survey instrument. The average re-
spondent in our sample is 43 years old, female, employed, has completed high school,
and has not heard about the government commitments before. Interestingly, in spite
of the existence a dedicated website and public announcement of commitments pub-
lic, most people remain uninformed of the the government commitments. A third
of our sample is between 18 and 35 years old, and almost 35 percent of the partic-
ipants have a college degree. The lower panel in Table 5 in the Appendix includes
the multiple dependent variables we are working with.
We have two sets of dependent variables. First, we evaluate citizens’ perceptions
about transparency by asking individuals whether they believe the government in-
forms the citizenry about its plans and results, and how much it does. Second, we
construct a set of indexes from 11 questions to try to proxy the complexity and
multidimensionality of trust. This questions attempt to evaluate the perceptions
people have about the competence, benevolence, and honesty of the government,
plus the degree of citizens’ confidence in the government it. All the questions are
included in Appendix B.
All the variables capture different but presumably correlated attributes of the
government. Variables capturing competence, benevolence, and honesty range from
a minimum of 1 to a maximum of 7. Variables capturing confidence range from a
minimum of 1 to a maximum of 5. Table 1 shows evidence of randomization balance
on observable characteristics between treatment arms.

3.2 Empirical Strategy

The empirical strategy is relatively simple and the general model we estimate is
presented in the following equation:

7
 Y = α + T β + Xλ +  (1)
where T is the vector of treatments (the pooled treatments in one case–information–
or the individual treatments in the other), and X a vector of control variables. The
set of control variables includes all observable characteristics available from the
survey, such as age, gender, education, labor status, socio-economic level, and com-
mune.
Regarding the dependent variables, we work with the individual responses to
the questions and summary indices constructed to aggregate the information across
related outcomes. The aggregation improves statistical power to detect effects that
are consistent across specific outcomes when these specific outcomes also have id-
iosyncratic variation. Focusing on the indices helps us to reduce the number of
statistical tests performed so as to reduce the chance of false positives, although
interpretation could be cumbersome. We report evidence from principal compo-
nent analysis (PCA henceforth) because the first component explains more than 80
percent of the variance regardless of the index, but results remains the same with
alternative index definitions (see Table 6 in the Appendix).15 In addition, we have
conducted exercises where we correct p-values with three different methods which
rely on different assumptions. Conclusions on statistical inference remains the same
with our preferred correction.16 We construct four intermediate indexes (Compe-
tence, Benevolence, Honesty, and Confidence) and one global index that summarizes
all questions on perception (Trust in Government).

3.3 Empirical Results

The aim of the survey experiment is to study the relationship between the provision
of information and citizens perceptions about transparency and trust in the govern-
ment. First, we evaluate the role of information on perceptions about government
transparency, and then we evaluate the relative effect of the different vignettes on
indices that approximate the components of trust we try to explain.
Figure 2 provides the first approximation to the results. The Figure shows the
distribution of answers for the control and treatment groups. As can be observed,
15
Robustness exercises include running the regressions with the individual questions, standard-
ized indices, residuals from a regression, or indices as those proposed by Kling et al. (2007) where
specific outcomes are normalized in standardized units to study mean effect sizes of the indices
relative to the standard deviation of the control group. In all cases, results remain the same.
16
We correct p-values with Bonferroni-Holm, Šidák-Holm, and Westfall and Young procedures.
See Hochberg (1988); Šidák (1967); Westfall et al. (1993) respectively. We implement Westfall et al.
(1993) permutation algorithm to control for the FWER. The algorithm requires no assumptions
on the data-generating process to obtain multiple testing procedures. It is a step-down procedure
with adjusted p-values directly estimated by permutation. Improvements could be obtained with
assumptions; see Westfall and Tobias (2007) for a broad discussion. Results available upon request.

8
there is a movement of mass to the right for the treatment group. More people in
the treatment group than in the control group believe that the government provides
information on plans and results. We evaluate distributional differences with a χ2
statistic, and we strongly reject the null hypothesis of equality in distributions (p-
value smaller than 1%).
Table 2 shows the first set of regression results, which suggests that information
increases the perception about government transparency. The dependent variable
is a binary variable indicating whether the participant thinks that the government
reports what it plans to do. Columns 1 and 2 present the results for the pooled
informative treatments without and with controls, respectively. Basically, an indi-
vidual that is presented with the information about the government commitments
is about 8 to 9 percentage points more likely to indicate that he believes that the
government reports what it plans to do than an individual in the control group.
This is an increase of more than 10 percent. Results are statistically different from
zero.
In columns 3 and 4 we look at the effect of each individual informative treatment
to evaluate whether the content of the information affects this perception, with and
without controls. Results in those columns show that the magnitude and the signif-
icance is the same for every informative treatment (we cannot reject the hypotheses
that the coefficients are the same). That is, citizens’ perceptions about transparency
increase regardless of the framing and content of the message.
Does the content of information matter for affecting trust in the government?
To answer this question we evaluate the differential effect of the treatments on the
standardized version of the indices of government perceptions. Figure 3 provides the
first approximation to this question. The Figure shows the average answers for each
treatment for the overall index, which has been standardized to have zero mean
and unitary standard deviation. The first and last bar are statistically different
to each other (we don’t compare the second and third because both components
are changing). The difference between the “positive” and “negative” treatments is
equivalent to 0.10 of a standard deviation. The statistical analysis for the composite
index (presented in the Figure) and the component indices are presented in Table 7
in the Appendix.
Table 3 shows the main regression results. Table 3 has 10 columns (2 for each
index) and 2 panels. Odds columns present regressions without covariates, and
even columns present regressions with covariates. Panel A compares the effect of
the vignettes that show differences in performance while holding constant the fram-
ing (empathy). Panel B compares the effect of the vignettes holding constant the
performance (positive) but changing the framing of the message (empathy versus
efficiency).
Results show some very consistent patterns across different attributes that eval-
uate trust in the government. First, those who were exposed to the treatment where

9
the government was over-performing on its goals think that the government is more
competent, more benevolent, and more honest. Individuals also have more con-
fidence in the government. Second, those who were presented with the empathy
framing seem to evaluate the government slightly better than those who received
the efficiency one, but the differences are not statistically significant. This is true
for every index. Again, the vignette showing a positive performance increases trust
in government by about 0.10 of a standard deviation more than than the vignette
showing negative performance.
Table 8 in the Appendix shows similar results using as dependent variable two
count variables indicating what we define as “good” and “bad” perception. In
columns 1 and 2, the dependent variable is number of times (out of 11) the partici-
pant chooses a score greater than 5 in perception questions.17 In columns 3 and 4,
the dependent variable is number of times (out of 11) the participant chooses a score
lower than 3 in perception questions. Columns 1 and 3 show results from Poisson
model estimates. Columns 2 and 4 show results from Zero Inflated Poisson (ZIP)
model estimates. Voung tests suggest ZIP is a better fit than Poisson estimates.
The interpretation of Table 8 leads us to the same conclusions as before: differences
in performance seem to matter to evaluate the government; this is not the case for
differences in the way the government frames the communication.

4 Conclusion
Results in this paper from a survey experiment conducted in the City of Buenos
Aires show that it may not be enough to have information available on a website.
More than 40 percent of respondents had never seen the website or heard about the
commitments. In order to improve citizens perceptions about transparency, gov-
ernments should bring the information to the citizens. Once citizens are shown the
information, there is a significant increase in how they evaluate the government’s
transparency. Results also indicate that citizens’ trust in the government is affected
the same by a message that frames the commitments in terms of efficiency or em-
pathy. However, the content of the information in terms of performance seems to
matter more for how much citizens update their priors: showing good performance
has a relatively larger effect than showing relative poor performance.
This evidence has important implications for the literature, as it highlights that
increasing transparency perceptions is more than providing information on a website
and that the channel between transparency and trust is mediated by several factors,
including the performance of the government. Future studies may wish to explore
this issue. In particular, to evaluate what is the optimal disclosure of information,
whether rules would work better than endogenous decisions about disclosure, and
17
For Confidence1 and Confidence2, we use 3 as threshold.

10
the contexts under which performance does and does not matter. The interaction
between disclosure and media reporting could be a fruitful area for research.
The evidence also has important implications for policymakers. Again, the re-
sults show that in order to increase trust it is not only necessary to provide infor-
mation, but also to show good performance. Of course, this does not mean that
manipulating the information presented to the public is a good idea. In the context
of verifiable reports, manipulating information would eventually reduce trust in the
government.

11
References
Adams, B. D. (2005). Trust vs. confidence. Technical report, Humansystems Inc
Guelph (Ontario, Canada).

Andrews, R. and S. de Walle (2013). New public management and citizens’ percep-
tions of local service efficiency, responsiveness, equity and effectiveness. Public
Management Review 15 (5), 762–783.

Bertot, J. C., P. T. Jaeger, and J. M. Grimes (2010). Using ICTs to create a culture
of transparency: E-government and social media as openness and anti-corruption
tools for societies. Government information quarterly 27 (3), 264–271.

Bouwman, R. and S. Grimmelikhuijsen (2016). Experimental public administration

from 1992 to 2014: A systematic literature review and ways forward. International
Journal of Public Sector Management 29 (2), 110–131.

Brennan, G. and L. Lomasky (1997). Democracy and decision: The pure theory of
electoral preference. Cambridge University Press.

Cahuc, P. and Y. Algan (2013). Trust , Growth and Well-being : New Evidence
and Policy Implications.

Chong, A., A. L. De La O, D. Karlan, and L. Wantchekon (2014). Does corruption

information inspire the fight or quash the hope? A field experiment in Mexico
on voter turnout, choice, and party identification. The Journal of Politics 77 (1),
55–71.

Ciborra, C. (2005). Interpreting e-government and development: Efficiency, trans-

parency or governance at a distance? Information Technology & People 18 (3),
260–279.

Congleton, R. D. (2001). Rational ignorance, rational voter expectations, and

public policy: A discrete informational foundation for fiscal illusion. Public
Choice 107 (1-2), 35–64.

Cruz, C., P. Keefer, and J. Labonne (2016). Incumbent advantage, voter information
and vote buying. Technical report, IDB Working Paper Series.

Cucciniello, M., G. A. Porumbescu, and S. Grimmelikhuijsen (2017). 25 years of

transparency research: Evidence and future directions. Public Administration
Review 77 (1), 32–44.

12
de Leon, L., F. Leite, and R. Rizzi (2014). A test for the rational ignorance hypoth-
esis: Evidence from a natural experiment in brazil. American Economic Journal:
Economic Policy 6 (4), 380–98.

Downs, A. (1957). An economic theory of political action in a democracy. Journal

of political economy 65 (2), 135–150.

Ferraz, C. and F. Finan (2008). Exposing corrupt politicians: the effects of brazil’s
publicly released audits on electoral outcomes. The Quarterly Journal of Eco-
nomics 123 (2), 703–745.

Gingerich, D. W. and C. Scartascini (2018). A Heavy Hand or a Helping Hand?

Information Provision and Citizen Preferences for Anti-Crime Policies.

Grimmelikhuijsen, S. (2012a). A good man but a bad wizard. about the limits and
future of transparency of democratic governments. Information Polity 17 (3, 4),
293–302.

Grimmelikhuijsen, S. (2012b). Linking transparency, knowledge and citizen trust

in government: an experiment. International Review of Administrative Sci-
ences 78 (1), 50–73.

Grimmelikhuijsen, S., G. Porumbescu, B. Hong, and T. Im (2013). The effect of

transparency on trust in government: A cross-national comparative experiment.
Public Administration Review 73 (4), 575–586.

Grimmelikhuijsen, S. G. (2010). Transparency of public decision-making: Towards

trust in local government? Policy & Internet 2 (1), 5–35.

Grimmelikhuijsen, S. G. and E. W. Welch (2012). Developing and testing a theoret-

ical framework for computer-mediated transparency of local governments. Public
administration review 72 (4), 562–571.

Hochberg, Y. (1988). A sharper Bonferroni procedure for multiple tests of signifi-

cance. Biometrika 75 (4), 800–802.

Keefer, P. and S. Khemani (2012). Do informed citizens receive more... or pay more?
The impact of radio on the government distribution of public health benefits.
Technical report, World Bank Policy Research Working Paper.

Keefer, P., C. Scartascini, and R. Vlaicu (2018). Shortchanging the Future: The
Short-Term Bias of Politics. In A. Izquierdo and C. Pessino (Eds.), Better Spend-
ing for Better Lives (1 ed.)., Chapter 10. Washington, DC, USA: Inter-American
Development Bank.

13
Kling, J. R., J. B. Liebman, and L. F. Katz (2007). Experimental analysis of
neighborhood effects. Econometrica 75 (1), 83–119.

Levi, M. and L. Stoker (2000). Political Trust and Trustworthiness. American

Review of Political Science 3, 475–507.

Martinelli, C. (2007). Rational ignorance and voting behavior. International Journal

of Game Theory 35 (3), 315–335.

Mullainathan, S., E. Washington, and J. R. Azari (2010). The impact of electoral de-
bate on public opinions: an experimental investigation of the 2005 New York City
mayoral election,. Political representation, Ian Shapiro, Susan C. Stokes, Elis-
abeth Jean Wood, and Alexander S. Kirshner, eds.(Cambridge University Press,
2010).

Nock, S. L. and T. M. Guterbock (2010). Survey experiments. Handbook of survey

research 2, 837–865.

OECD (2017). Trust and Public Policy How Better Governance Can Help Rebuild
Public Trust (1 ed.). Paris: OECD Publishing.

OECD (2018). Trust and its determinants : Evidence from the Trustlab experiment.

Pina, V., L. Torres, and S. Royo (2007). Are ICTs improving transparency and
accountability in the EU regional and local governments? An empirical study.
Public administration 85 (2), 449–472.

Pina, V., L. Torres, and S. Royo (2010). Is e-government leading to more accountable
and transparent local governments? An overall view. Financial Accountability &
Management 26 (1), 3–20.

Šidák, Z. (1967). Rectangular confidence regions for the means of multivariate

normal distributions. Journal of the American Statistical Association 62 (318),
626–633.

Tolbert, C. J. and K. Mossberger (2006). The effects of e-government on trust and

confidence in government. Public administration review 66 (3), 354–369.

Van de Walle, S. and G. Bouckaert (2014). Public service performance and trust in
government : the problem of causality. International Journal of Public Adminis-
tration 29 (8&9), 891–913.

Welch, E. W., C. C. Hinnant, and M. J. Moon (2004). Linking citizen satisfaction

with e-government and trust in government. Journal of public administration
research and theory 15 (3), 371–391.

14
Westfall, P. H. and R. D. Tobias (2007). Multiple testing of general contrasts: Trun-
cated closure and the extended Shaffer–Royen method. Journal of the American
Statistical Association 102 (478), 487–494.

Westfall, P. H., S. S. Young, and S. P. Wright (1993). On adjusting P-values for

multiplicity. Biometrics 49 (3), 941–945.

15
Figure 1: Experiment ‘Timeline’

16
 Figure 2: Information and Perceptions about Transparency

50

40

30
Percent

20

10

0
Doesn't inform Informs plans, no results Informs plans and results
No informa planes Informa planes, no resultados Informa planes y resultados

Control Treatment

Notes: The figure shows the distribution of answers for the control and treatment groups.
As can be observed in the Figure, there is a movement of mass to the right. More people in the
treatment believe that the government provides information on plans and results. The difference
in the distribution is statistically significant.

17
 Figure 3: Information and Government Perceptions

0.04 Positive and Empathy

0.03 Positive and Efficiency

0.02

0.01

0.00

-0.01

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07 Negative and Empathy

Notes: The figure shows the average answers for each treatment for the overall index, which
has been standarized to zero mean and unitary standard deviation. The first and last bar are
statistically different from each other. The difference between the “positive” and “negative” treat-
ments is equivalent to 0.10 of a standard deviation. Regressions results are presented in Table 7,
first row.

18
 Table 1: Descriptive and Balance

Difference in means (p-value)

Mean Empathy Positive
Information
Efficiency Negative
(1) (2) (3) (4)
Female 0.500 0.070 0.993 0.701
(0.011)
Age 42.90 0.394 0.520 0.985
(0.261)
Employed 0.822 0.805 0.516 0.606
(0.009)
Unemployed 0.078 0.575 0.861 0.244
(0.006)
College degree 0.344 0.541 0.592 0.785
(0.011)
First exposure 0.429 0.013 0.872 0.863
(0.011)
ABC1 0.250 0.969 0.384 0.744
(0.010)

Observations 1,999 1,999 1,365 1,329

Notes: Column 1 shows mean and standard deviation in parentheses. In

columns 2 we show p-values from OLS regressions with the covariates
as dependent variables and the informative treatment as independent
variable, using the full sample. In columns 3 and 4 we show p-values
from OLS regressions with the covariates as dependent variables and the
“empathy vs efficiency” or “positive vs negative” treatments as indepen-
dent variables. In column 3, we keep positive performance constant and
compare between empathy and efficiency frames. In column 4, we keep
empathy framing constant and compare between positive and negative
performance.

19
 Table 2: Information and Perceptions about Transparency

Government Transparency
(1) (2) (3) (4)

Informative Treatment 0.087*** 0.080***

(0.0223) (0.0223)
[T 1] Empathy and Positive 0.085*** 0.079***
(0.0229) (0.0226)
[T 2] Efficiency and Positive 0.089*** 0.080***
(0.0233) (0.0231)
[T 3] Empathy and Negative 0.086*** 0.081***
(0.0239) (0.0239)
Constant 0.772*** 0.800*** 0.772*** 0.800***
(0.0132) (0.0231) (0.0132) (0.0231)

Observations 1,999 1,999 1,999 1,999

Controls No Yes No Yes
[1] = [2] 0.871 0.965
[1] = [3] 0.975 0.951

Notes: Dependent variable is a binary variable that indicates that

the participant thinks that the government reports what it plans
to do. Controls include age, gender, education, labor status, socio-
economic level, commune, and a factor variable that indicates if
the participant knew about the policy before the survey. Robust
standard errors in parentheses. For columns 3 and 4, rows labeled
[1] = [2] and [1] = [3] show p-values from tests of equality of coef-
ficients.*** p<0.01, ** p<0.05, * p<0.1

20
Table 3: Effect of Performance and Framing on Governance Perception – Main results

Government Competence Benevolence Honesty Confidence

Perception
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Panel A: Effect of performance
Positive vs Negative (framing constant)

Positive graphic 0.101* 0.107** 0.096* 0.099* 0.105* 0.112** 0.101* 0.110** 0.101* 0.110**
(0.0551) (0.0542) (0.0551) (0.0542) (0.0552) (0.0545) (0.0553) (0.0541) (0.0553) (0.0541)
Constant -0.063 -0.25*** -0.062 -0.233** -0.065 -0.245** -0.066 -0.281*** -0.066 -0.281***
(0.0404) (0.0980) (0.0402) (0.0985) (0.0407) (0.0986) (0.0403) (0.0983) (0.0403) (0.0983)

Observations 1,329 1,329 1,329 1,329 1,329 1,329 1,329 1,329 1,329 1,329
Controls No Yes No Yes No Yes No Yes No Yes
Panel B: Effect of framing
Empathy vs Efficiency (performance constant)

Empathy message 0.016 0.038 0.012 0.032 0.019 0.040 0.009 0.029 0.009 0.029
(0.0537) (0.0535) (0.0538) (0.0536) (0.0534) (0.0532) (0.0537) (0.0536) (0.0537) (0.0536)
Constant 0.021 -0.149 0.023 -0.112 0.021 -0.146 0.026 -0.172* 0.026 -0.172*
(0.0384) (0.0939) (0.0384) (0.0948) (0.0383) (0.0936) (0.0381) (0.0948) (0.0381) (0.0948)

Observations 1,365 1,365 1,365 1,365 1,365 1,365 1,365 1,365 1,365 1,365
Controls No Yes No Yes No Yes No Yes No Yes

Notes: Controls include age, gender, education, labor status, socio-economic level,
commune, and a factor variable that indicates if the participant knew about the
policy before the survey. The overall sample of 1,999 has zero mean and unitary
standard deviation for each index. Robust standard errors in parentheses. ***
p<0.01, ** p<0.05, * p<0.1

21
A Appendix: Tables and Figures

Table 4: Vignette composition

Information
Total
Control Treatment

1,017 982 1,999

Performance
Framing Total
Positive Negative

Empathy 695 634 1,329

Efficiency 670 0 670
Total 1,365 634 1,999

Notes: Numbers correspond to observations.

First, we show the split between control and
treatment group for the information experi-
ment. Second, we show treatment composi-
tion between features regarding performance
(either positive for over-perfomance or neg-
ative for under-performance), and framing
(empathy or efficiency).

22
 Table 5: Summary statistics

Obs Mean Std. Dev. Min Max P25 P50 P75

Variable
(1) (2) (3) (4) (5) (6) (7) (8)
Characteristics
Female 1999 0.500 0.500 0 1 0 1 1
Age 1999 42 11 18 65 33 42 53
Young 1999 0.300 0.458 0 1 0 0 1
Employed 1999 0.822 0.383 0 1 1 1 1
Unemployed 1999 0.078 0.268 0 1 0 0 0
College degree 1999 0.344 0.475 0 1 0 0 1
First exposure 1999 0.429 0.495 0 1 0 0 1
ABC1 1999 0.25 0.433 0 1 0 0 1
Transparency
Transparency 1999 0.814 0.389 0 1 1 1 1
Perceptions
Competence1 1999 4.426 1.837 1 7 3 5 6
Competence2 1999 4.336 1.837 1 7 3 5 6
Competence3 1999 4.498 1.885 1 7 3 5 6
Competence4 1999 3.765 1.91 1 7 2 4 5
Competence5 1999 4.492 1.984 1 7 3 5 6
Benevolence1 1999 4.189 1.875 1 7 3 4 6
Benevolence2 1999 3.996 1.886 1 7 3 4 5
Honesty1 1999 3.838 1.950 1 7 2 4 5
Honesty2 1999 3.956 1.846 1 7 3 4 5
Confidence1 1999 3.005 1.096 1 5 2 3 4
Confidence2 1999 2.958 1.105 1 5 2 3 4
Good (count) 1999 3.199 3.968 0 11 0 1 6
Bad (count) 1999 2.631 3.936 0 11 0 0 4

Notes: Column 1 present the number of observations. In columns 2 and 3,

we show mean and standard deviation. Columns 4 and 5 show minimum
and maximum values respectively. Values in columns 6 to 8 represent inter-
quartile range, and the median. Female is an binary variable for gender.
Age is a variable recording the age of each participant. Young is an binary
variable for those who are between 18 and 35 years old. Employed is an
binary variable for those who have a full time employment and those who
work in their house. Unemployed is an binary variable for those who are
looking for a job at the time of the survey. College degree is an binary
variable for those who have a college degree. First exposure is an binary
variable for those who have not heard about the policy before the survey.
ABC1 is an binary variable for those with the highest category for socio–
economic level. Transparency is an binary variable and takes the value one
if the participant thinks that the government informs what they plan to
do (see Appendix B and Figure 2). Perception variables are categorical
variables (see Appendix B).

23
 Table 6: PCA

Eigenvalue Proportion Std. Err. Cumulative Std. Err. Bias

(1) (2) (3) (4) (5) (6)
Trust in
Government
Comp1 8.99 0.82 0.005 0.818 0.005 0.00104
Comp2 0.55 0.05 0.002 0.867 0.004 0.00104
Comp3 0.36 0.03 0.001 0.900 0.003 0.00055
Comp4 0.27 0.02 0.001 0.924 0.002 -0.00018
Comp5 0.18 0.02 0.001 0.940 0.002 0.00044
Comp6 0.14 0.01 0.001 0.953 0.001 0.00090
Comp7 0.13 0.01 0.001 0.965 0.001 -0.00090
Comp8 0.11 0.01 0.000 0.975 0.001 0.00007
Comp9 0.10 0.01 0.000 0.984 0.001 0.00121
Comp10 0.10 0.01 0.000 0.993 0.000 -0.00329
Comp11 0.08 0.01 0.000 1.000 0.000 -0.00087
Competence
Comp1 3.39 0.85 0.005 0.847 0.005 0.00033
Comp2 0.28 0.07 0.003 0.917 0.003 0.00046
Comp3 0.22 0.05 0.002 0.972 0.001 -0.00052
Comp4 0.11 0.03 0.001 1.000 0.000 -0.00027
Benevolence
Comp1 1.88 0.94 0.003 0.940 0.003 0.00006
Comp2 0.12 0.06 0.003 1.000 0.000 -0.00006
Honesty
Comp1 1.82 0.91 0.004 0.908 0.004 0.00010
Comp2 0.18 0.09 0.004 1.000 0.000 -0.00010
Confidence
Comp1 1.92 0.96 0.002 0.960 0.002 0.00004
Comp2 0.08 0.04 0.002 1.000 0.000 -0.00004

Notes: Table shows eigenvalues from the principal component analysis (PCA) eigen de-
composition (column 1). The underlying eigenvectors are orthonormal (uncorrelated and
normalized). First eigenvalue is our index for each category because first component
explain more than 80% of the variance in each index (columns 2 and 4). Under PCA
assumptions, the first principal component is the best synthetic indicator (in the least
square sense) of the range of variability of variables considered. The index can be con-
sidered a sort of synthetic index that combines or condenses, in a single variable, the
consistent information originally dispersed over different measurements. Heteroskedastic
robust bootstrap confidence intervals are computed (columns 3 and 5).

24
Table 7: Effect of Performance and Framing on Perceptions about the Government –
Alternative

Policy outcome Positive Positive Negative

Meesage wording Empathy Efficiency Empathy [1]=[2] [1]=[3]
(1) (2) (3) (4) (5)
Trust in Government 0.037 0.021 -0.063 0.762 0.068
(0.990) (0.993) (1.017)
Competence 0.034 0.023 -0.062 0.830 0.081
(0.992) (0.995) (1.013)
Benevolence 0.040 0.021 -0.065 0.727 0.057
(0.983) (0.991) (1.026)
Honesty 0.035 0.026 -0.066 0.863 0.067
(0.997) (0.987) (1.014)
Confidence 0.035 0.026 -0.066 0.863 0.067
(0.997) (0.987) (1.014)

Observations 695 670 634 1,999 1,999

Notes: Columns (1), (2), and (3) show mean and standard deviation for
each sub-sample. The overall sample has zero mean and unitary stan-
dard deviation for each index. Columns (4) and (5) show p-values from
tests of equality of coefficients computed after a regression with each
index as dependent variable and a dummy variable for each vignette.

25
 Table 8: Count variables

Good Bad
perception perception
(1) (2) (3) (4)

Efficiency and Positive

-0.0539 -0.0392 0.000558 0.0381
(0.0653) (0.0303) (0.0828) (0.0346)
Empathy and Negative -0.133** -0.0481 0.144* 0.0670**
(0.0671) (0.0314) (0.0799) (0.0338)
Constant 1.384*** 1.806*** 0.684*** 1.684***
(0.0635) (0.0296) (0.0856) (0.0351)

Observations 1,999 1,999 1,999 1,999

Controls Yes Yes Yes Yes
Model Poisson ZIP Poisson ZIP
Vuong test 26.86 30.44

Notes: In columns (1) and (2), the dependent variable is number of times (out
of 11) the participant chooses a score greater than 5 in perception questions.
In columns (3) and (4), the dependent variable is number of times (out of 11)
the participant chooses a score lower than 3 in perception questions. Controls
include age, gender, education, labor status, socio-economic level, commune,
and a factor variable that indicates if the participant knew about the policy
before the survey. Columns (1) and (3) show results from Poisson model
estimates. Columns (2) and (4) show results from Zero Inflated Poisson
(ZIP) model estimates. The zero value is inflated by treatment arms and
a variable scoring Government Management. Voung tests suggest ZIP is a
better fit than Poisson estimates. Robust standard errors in parentheses.
Base category is Empathy and Positive. *** p<0.01, ** p<0.05, * p<0.1

26
Figure 4: Treatment 1 – Empathy message and positive outcome

27
Figure 5: Treatment 2 – Efficiency message and positive outcome

28
Figure 6: Treatment 3 – Empathy message and negative outcome

29
B Appendix: Survey Module
Transparency
Which of the following statements better represents the city government?

The government of the city of Buenos Aires...

...does not provide information about its plans.
...provides information about its plans but does not provide information about implementation and
...provides information about its plans, implementation and accomplishments.

Perceptions
Items adapted from Grimmelikhuijsen (2012b).
Using a scale from 1 to 7, where 1 is “Completely disagree,” and 7 is “Completely
agree,” please show your level of agreement with the following statements about
the Government of the city of Buenos Aires.

The government of the city of Buenos Aires. . .

Competence:
...is capable. (Competence1)
... is professional. (Competence2)
... is innovative. (Competence3)
... spends the available budget appropriately. (Competence4)
... thinks in the long term. (Competence5)
Benevolence:
... keeps the interest of the residents in mind. (Benevolence1)
... does everything in its power to help those in need. (Benevolence2)
Honesty:
... is sincere. (Honesty1)
... fulfills its promises. (Honesty2)
Confidence: Do you trust that the city government...
... to do their best for the city? (Confidence1)
... to do their best for the residents and their families? (Confidence2)

30
C Appendix: About the Company

Figure 7: Webpage

Source: See Quiddity official webpage (April 19, 2018)

31
 Figure 8: Webpage

Source: See Quiddity official webpage (April 19, 2018)

32