Publicizing malfeasance: When media facilitates …...Our point estimates imply that each additional...
Transcript of Publicizing malfeasance: When media facilitates …...Our point estimates imply that each additional...
Publicizing malfeasance: When media facilitates
electoral accountability in Mexico ∗
Horacio A. Larreguy †
John Marshall ‡
James M. Snyder Jr. §
April 2015
We estimate the effect of broadcast media outlets revealing information about incum-bent performance on electoral accountability in Mexico. Focusing on malfeasance bymunicipal mayors, we study federal grants earmarked for infrastructure projects tar-geting the poor, and leverage two sources of plausibly exogenous variation. First, weexploit variation in the timing of the release of municipal audit reports around elec-tions. Second, and moving beyond existing studies, we compare neighboring electoralprecincts on the boundaries of media stations’ coverage areas to isolate the effects of anadditional broadcast media station. We find that voters punish the party of malfeasantmayors, but only in precincts covered by local media stations, which emit from withinthe precinct’s municipality. An additional local radio or television station reduces thevote share of an incumbent political party revealed to be either corrupt or neglectfulof the poor by around 1 percentage point. The effect of a local media station is largerwhen local media stations principally serve the market inside their municipality. Incontrast, while we find no evidence that non-local media stations contribute to theelectoral sanctioning of malfeasant mayors, our results indicate that non-local mediacrowds out the sanctioning effects of local media. Our findings thus indicate that elec-toral accountability requires that the media market structure provides media stationswith incentives to supply politically-relevant information to their audiences.
JEL: D72, D78, H41, O17.Key words: corruption, elections, malfeasance, media, voter behavior.
∗We thank Daron Acemoglu, Rakeen Mabud, Ben Olken, Jesse Shapiro, and David Stromberg, as wellas participants at the Harvard Comparative Politics Workshop and MIT Political Economy Lunch for usefulcomments. Thanks to Andrea Ortiz and Daniel Silberwasser for excellent research assistance, and to ASFofficials for providing information about the auditing process. Horacio Larreguy acknowledges financialsupport from the IQSS Undergraduate Research Scholars Program. All errors are our own.†Department of Government, Harvard University. [email protected].‡Department of Government, Harvard University. [email protected].§Corresponding author. Department of Government, Harvard University. [email protected].
1
The functionaries of every government have propensities to command at will the
liberty and property of their constituents. There is no safe deposit for these but
with the people themselves, nor can they be safe with them without information.
Where the press is free, and every man able to read, all is safe. (Thomas Jefferson
to Charles Yancey, 1816. ME 14:384).
1 Introduction
A large body of scholarship in political economy asserts that in large democracies: (i) elec-
tions are one of the key institutions for producing political accountability; (ii) in order for
elections to function well, voters must be adequately informed; and (iii) the mass media play
an essential role in informing voters. One important application of this trio is the electoral
sanctioning of malfeasant behavior such as corruption and diverting funds away from the
projects for which they are earmarked. This is particularly relevant in developing democra-
cies, where corruption is prevalent (e.g. Mauro 1995) and voters are relatively uninformed
(Greene 2011; Lawson and McCann 2005; McCann and Lawson 2003). This article seeks to
identify when broadcast media effectively hold local government to account by facilitating
the election of high-quality representatives.
Although the media often cover corruption scandals and other cases of malfeasance (e.g.
Puglisi and Snyder 2011), there is little solid evidence establishing (a) whether such infor-
mation causes voters to punish politicians at the polling booth, and (b) what kinds of media
station induce such punishment. Identifying such effects is challenging because identifica-
tion requires exogenous variation in both malfeasance revelations and voter access to media
coverage. Ferraz and Finan (2008) find that incumbent mayors in Brazil who are randomly
revealed to be corrupt just before an election suffer more at the polls in municipalities with
more AM radio stations. However, as the authors acknowledge, without exogenous variation
in media access the study cannot separate the effects of AM radio from its many correlates,
such as education and demand for political news.1
Several studies also show that randomizing access to information about incumbent rep-
1Table 2 in the Online Appendix shows that media access is significantly greater in moreurban, literate, and economically developed areas in Mexico. This is likely to induce bias,given that such variables themselves have important political consequences. For example,Klasnja (2011) finds evidence that voters with greater “political awareness” are more likelyto punish incumbents in corruption scandals in the U.S., while Weitz-Shapiro and Winters(2014) find similar results for voters with higher literacy in Brazil.
2
resentatives can induce voters to sanction poor performance in office (Banerjee et al. 2011;
Humphreys and Weinstein 2012).2 However, such studies cannot separate the effects of the
information provided from their mode of transmissions since they lack exogenous variation
in malfeasance revelations. We combine these literatures and overcome such concerns by
exploiting plausibly exogenous variation in both the release of audit reports and access to
media.
Similarly, there is limited evidence identifying how the structure of the media market
affects the electoral sanctioning of malfeasant politicians in developing contexts (Larreguy
and Monteiro 2014). First, not all voters receive politically-relevant news from the media
stations that that cover them, given that media stations vary in their incentives to sup-
ply politically-relevant news to segments of their audience. While Snyder and Stromberg
(2010) examine the implications of media market and electoral boundary overlap for behav-
ior in office in the U.S., how such media supply incentives affect the electoral punishment
of malfeasant behavior is an outstanding question. Second, surprisingly little empirical at-
tention has been devoted to the role of market concentration. Since media it is easier to
capture media in concentrated markets (Besley and Prat 2006), voters may require multiple
signals to substantially update their beliefs (Gentzkow, Shapiro and Sinkinson 2014), and
new media stations may engage new segments of the market (Prat and Stromberg 2005),
additional media stations providing similar news content could considerably increase voter
sanctioning. Despite the potential importance of market concentration, existing research
has largely focused on identifying the electoral implications of access to a new media station
or media market offering distinct news content (DellaVigna and Kaplan 2007; Enikolopov,
Petrova and Zhuravskaya 2011; Snyder and Stromberg 2010).
In this article, we identify large electoral effects of local broadcast media stations—
which emit within an electoral precinct’s municipality—publicizing revelations of mayoral
malfeasance in Mexico, particularly when media stations primarily serve local audiences. To
do so, we utilize detailed geographic data and exploit two sources of plausibly exogenous
variation. First, as in Ferraz and Finan (2008), we leverage variation in the timing of the
release of municipal audit reports around elections. In particular, we use a difference-in-
differences design to compare mayors who engage in malfeasant behavior—either corruption
2Observational studies find that corrupt politicians are more likely to be punished elec-torally when their corruption is covered in the news, or when political corruption is moresalient (Chang, Golden and Hill 2010; Costas, Sole-Olle and Sorribas-Navarro 2011; Eggersand Fisher 2011). However, such studies are vulnerable to the concern that the presence ofmedia coverage is correlated with the severity of malfeasant behavior.
3
or diverting funds to other projects that do not benefit the intended poor recipients—that
is revealed in audit reports published before an election to comparable mayors whose audit
reports are not published until after the election. Second, and moving beyond existing
studies, we also leverage within-neighbor variation in commercial quality radio and television
signals that differ across urban electoral precincts from within the same municipality due
to plausibly exogenous factor such as antenna power and geographic features lying between
the antenna and particular precincts.3 Mexican voters rely largely on such local media,
particularly television, to find out about malfeasance in the use of public funds (Castaneda
Sabido 2011).
In Mexico’s federal system, a significant proportion of government spending is admin-
istered by municipal mayors. Amidst widespread concerns about corruption, the Mexican
Congress passed a law institutionalizing independent audits of the use of federal funds in
1999. We focus on audit reports pertaining to the Municipal Fund for Social Infrastructure
(FISM), a major social program—representing about 25% of mayors’ annual budgets—that
provides mayors with funds for infrastructural projects required to benefit impoverished cit-
izens. Audits are announced the year after the funds have been allocated, and reports reveal
the share of FISM money spent “in an unauthorized manner,” as well as the share spent on
projects “not benefiting the poor.” The first figure clearly represents malfeasance, and usu-
ally actual corruption. The second figure indicates malfeasance of a different sort—diverting
funds from their intended targets. By law, 100% of FISM projects must benefit the poor, so
any money not spent on the poor represents illegal misallocation. Provided that voters care
about malfeasance, we are thus able to also address the important question of what types of
malfeasance matter most.
Our results first demonstrate that each additional local media station substantially in-
creases voter punishment of the party of mayors revealed to be either corrupt or neglectful
of the poor. Although mayors cannot seek re-election (due to term limits), voters primarily
choose between parties since mayoral elections are very low-information races, and inter-
nal selection procedures ensure that candidates types within parties are highly correlated
(Langston 2003). Our point estimates imply that each additional local radio or television
station reduces the vote share of the incumbent political party whose mayor was revealed
3Specifically, we provide estimates of the “intention-to-treat” voters with access to media,because commercial quality coverage boundaries reduce the likelihood that voters receive amedia signal but cannot preclude coverage entirely. Such issue are discussed below. Sincewe lack the data required to compute a first stage and the exclusion restriction might nothold, we focus on providing unbiased reduced form estimates.
4
to be corrupt by more than 0.7 percentage points. The effects of failing to spend funds on
the poor are even larger: our point estimates imply that if the incumbent party’s mayor
was revealed to have misallocated funds away from the poor, each additional local radio or
television station covering a given precinct reduces the party’s vote share by up to 1.2 per-
centage points, depending on the severity of the misallocation. However, we also find that
voters reward good performance in office: when the incumbent party’s mayor did not engage
in corruption or correctly spent the money on the poor, an audit report released before an
election increases the party’s vote share by almost 0.7 percentage point for each additional
local media station. Differentiating media formats, we find that exposure to an additional
local television station—the most prevalent source of political information in Mexico—has
substantially larger effects on electoral sanctioning than an additional FM radio station,
which has similar signal range.4
While there are large effects for local media stations, we find no evidence that non-local
media stations contribute to the electoral sanctioning of the party of malfeasant mayors.
Although this could simply be because non-local media stations fail to capture audiences
outside their municipality, we find that the presence of an additional non-local media station
in fact crowds out the sanctioning effect of local media. Furthermore, consistent with the
extent of news coverage of mayoral performance increasing with audience demand for such
information, we find—akin to Snyder and Stromberg (2010)—that the effect of a local media
station is larger when the media market of such media is principally from inside the munici-
pality. These findings suggest that electoral accountability requires that the media market is
structured to ensure that stations have incentives to supply politically-relevant information
to their audiences.
We demonstrate that these findings are robust to a series of sensitivity checks. First,
although the timing of audit report releases and the number of media stations are well
balanced across a wide variety of political, socioeconomic and demographic characteristics,
we show that our findings are robust to the inclusion of flexible interactions between audit
report outcomes and potential confounds of local media’s effect. Second, focusing only on
reports released before the election sharpens our estimates. Finally, we experiment with our
definitions of malfeasance and show that corruption’s effect are primarily driven by the most
egregious cases, while not spending on the poor is punished more commensurately to the
4We are unable to estimate the effect of an additional AM radio station since we lack asufficiently large sample of neighboring precincts within the same municipality that differ inAM radio coverage. This reflects the high power of AM radio antennae, which often coverall urban areas in their municipality.
5
magnitude of the misallocation.
Our findings contribute to the literature in a variety of ways. First, we demonstrate the
particular importance of local media for local electoral accountability, rather than media in
general.5 This is an important consideration because local radio and television are often the
only way in which isolated voters can learn about the performance of their incumbent politi-
cians, but only 45%, 50% and 56% of Mexican electoral precincts are respectively not covered
by a single AM radio, FM radio or television station emitting from inside their municipal-
ity.Moreover, understanding the role of local media in supporting electoral accountability is
particularly salient given that local media markets are shrinking in many countries.6 This
trend is particularly worrying in light of the fact that our findings provide a clear rationale
for malfeasant politicians to exploit the weakening economic position of local media and to
seek its control (Besley and Prat 2006), by purchasing radio stations (Boas and Hidalgo
2011) or preventing “defamation” (Stanig forthcoming).
Second, we demonstrate that media market structure plays a key role in explaining the
extent to which local media matter for electoral accountability. Our results, in a developing
context, thus buttress the finding that the electoral composition of media markets has impor-
tant implications for the types of political news that media stations report (Ansolabehere,
Snowberg and Snyder 2006; Snyder and Stromberg 2010; Stromberg 2004), which in turn
affects electoral accountability (Snyder and Stromberg 2010). In addition to comparing
the relatively efficacy of more and less locally-oriented broadcast media stations, we also
show that the presence of non-local media stations detracts from electoral accountability by
crowding out politically-relevant news.
Third, we show that voters respond to different types of malfeasance. With the exceptions
of Banerjee et al. (2011) and Humphreys and Weinstein (2012), who respectively analyze the
discretionary allocation of funds to slums rather than urban areas and the relative parlia-
mentary performance of legislators, the bulk of the literature on political accountability in
developing democracies has focused on corruption. Our results are most similar to Banerjee
et al. (2011), who also find that voters punish politicians for diverting funds away from the
5In contrast, Larreguy and Monteiro (2014) show the importance of regional and nationalmedia networks for national electoral accountability.
6Over the last 15 years, Mexico has experienced a 40% decline in the share of individualsclaiming to read political news in newspapers. 60% of Latinobarometer respondents claimedreading political news in newspapers 1996, compared to 36% in 2009. In the U.S., dailynewspaper circulation dropped from just over 1.0 newspapers per household in 1950 to about0.3 per household in 2010.
6
poor. However, the relatively large effects that we document —which are slightly greater for
revelations that politicians did not spend on the poor than for corruption revelations—may
reflect the fact that in Mexico such diversion is a direct violation of FISM program rules.7
Finally, given that audits are announced after FISM funds have been allocated, the data
suggest that the positive likelihood of being audited is insufficient to prevent municipal may-
ors from engaging in malfeasance. Our study thus complements previous research suggesting
that audits can be effective at reducing corruption only if politicians know prior to spending
that the reports could result in criminal prosecution (Olken 2007) or will be released before
an election (Bobonis, Fuertes and Schwabe 2013). Rather, the corruption levels we observe
in Mexico are broadly similar to those found in Brazil (Ferraz and Finan 2008), where the
municipal audit scheme was only announced after spending had occurred. This is consistent
with the dynamic optimizing behavior observed in India (Niehaus and Sukhtankar 2013), but
partially contrasts with recent findings from Brazil suggesting that increasing the probabil-
ity of audit reduces corruption but does not affect spending patterns (Zamboni and Litschig
2014). Our findings ultimately suggest that media-induced electoral accountability can mit-
igate adverse selection problems by reducing the probability that the parties of malfeasant
mayors are re-elected.
The article proceeds as follows. Section 3 provides a brief overview of local governments
in Mexico, the FISM funds that we study, the audit of such funds, and local media in Mexico.
Section 5 details our data and identification strategy. Section 6 presents our main results
and robustness checks. Section 7 concludes.
2 Why the intensity of media coverage matters
An influential literature has emphasized the political importance of access to a new media
station or media market offering distinct content, both in terms of the focus of their news
(e.g. Ansolabehere, Snowberg and Snyder 2006; Enikolopov, Petrova and Zhuravskaya 2011;
Snyder and Stromberg 2010) and their ideological stance (e.g. DellaVigna and Kaplan 2007;
Enikolopov, Petrova and Zhuravskaya 2011). However, despite significant theoretical interest
in the role of market concentration (e.g. Besley and Prat 2006; Gentzkow and Shapiro 2006),
little is known about the marginal effect of an additional media station providing similar
content for electoral accountability. Combining existing arguments, we suggest that access
7Conversely, legislators in India are free to allocate their discretionary project fundsanywhere in their districts.
7
to more media stations providing similar content may also play an essential role in informing
voters about the behavior of their politicians in office. This is of particularly relevance in
light of the mixed evidence regarding whether voters punish malfeasant behavior in office
(e.g. Chong et al. forthcoming, Ferraz and Finan 2008).
In theory, the existence of a single news-reporting media outlet could be sufficient to
support electoral accountability. Consider a situation where the incumbent is corrupt, but
voters may only learn this through the news. Abstracting from demand-side biases, for
one news-reporting media station to be sufficient to generate electoral sanctioning, many
voters would have to receive credible information from such an outlet, sufficiently update
their beliefs about their incumbent (party), and ultimately vote according to such beliefs. In
practice, many voters may not consume information, or the information consumed may not
cause voters to substantially update their beliefs. We argue that, once these conditions are
violated, an additional news-reporting media station may substantially affect voter behavior.
In their model of media capture, Besley and Prat (2006) show that political capture of
media stations decreases with the number of stations in the market, and in turn increases
turn over of low-quality politicians by ameliorating the adverse selection concern. Intuitively,
this is because high market concentration makes it costlier for politicians to buy off many
media stations that must be paid monopoly prices to resist the commercial profits associated
with revealing information. Consistent with this model, Djankov et al. (2003) provide sug-
gestive evidence that state media ownership reduces government performance in developing
countries, while Boas and Hidalgo (2011) provide evidence that Brazil politicians actively
seek to control local media stations once in office.
Without political capture, media firms and journalists may have incentives not to re-
veal politically-relevant information (see e.g. Anderson and McLaren 2012; Baron 2006;
Gentzkow and Shapiro 2006; Mullainathan and Shleifer 2005). For example, Gentzkow and
Shapiro (2006) suggest that the media may bias its news coverage toward the priors of their
consumers in order to increase their reputation (as a high-quality news outlet). However,
greater competition and diversity in the media market increases the probability that competi-
tors will expose the factual inaccuracies of a report pandering to audience priors, and thus
ensures that voters receive truthful signals about political performance.8 Although newspa-
per biases are correlated with the biases of their readers (Gentzkow and Shapiro 2010), there
is some evidence that media market competition indeed reduces biased reporting (Gentzkow
8Anderson and McLaren (2012) make a similar argument in a model where media ownersare politically biased.
8
and Shapiro 2006; Puglisi and Snyder 2011), although it may come at the cost of journalistic
quality (Cage 2014). Moreover, receiving the same information from multiple sources should
increase voter certainty regarding incumbent performance, especially when the information
is consistent across outlets with different political biases (Gentzkow, Shapiro and Sinkinson
2014).
Furthermore, the entry of a new media station may also affect media consumption, and
consequently increase consumption of politically-relevant news. Even in relatively large mar-
kets, new media can expose new listeners. Supporting this claim, Prat and Stromberg (2005)
show that the introduction of commercial television disproportionately attracts relatively un-
informed voters, causing them to become more politically knowledgeable and increase their
political participation. In particular, media stations with specific qualitiesmay be particu-
larly effective at gaining new audiences, given that voters are willing to change their media
consumption patterns in response to changes in the types of available media (Durante and
Knight 2012). Similarly, the introduction of media stations with different partisan biases
can attract sufficiently large audiences to alter electoral behavior (DellaVigna and Kaplan
2007; Enikolopov, Petrova and Zhuravskaya 2011). Below, we offer suggestive evidence that
the availability of additional local media station increases news consumption in our context
of Mexico.
Finally, even if voters consume unbiased news coverage, such coverage must still be
politically relevant in order to affect electoral accountability. Profit-maximizing media firms
face strong incentives to tailor their programming to consumer demands, and may thus focus
on catering to particular tastes (Stromberg 2004). In diverse media markets, small consumer
groups may simply not be served. For example, television stations predominantly cover in-
state politicians even when the media market includes voters outside the state (Ansolabehere,
Snowberg and Snyder 2006; Snyder and Stromberg 2010). This phenomenon may detract
from electoral accountability, not only by failing to provide information about incumbent
performance, but also by crowding-out media stations that do provide such content.
There are thus good reasons to believe that reducing media market concentration may
enhance electoral accountability, by increasing voter exposure to credible and informative
signals. However, the effect of an additional media station has yet to be identified. While
Ferraz and Finan (2008) provide suggestive evidence of the importance of AM radio stations,
this article aims to causally identify the impact of an additional broadcast media station
providing news about incumbent party malfeasance. Furthermore, to capture the role of
market structure, we differentiate local and non-local media stations and examine how the
9
effects of media coverage change with media market composition.
3 Political accountability in Mexico
Following 70 years of Institutional Revolutionary Party (PRI) hegemony, national and local
politics have become relatively competitive. Elections to the Chamber of Deputies, the
lower house of Mexico’s national legislature, are held every three years, while the President
and Senate are concurrently elected to six-year tenures.9 State and municipal elections are
instead staggered across the electoral cycle and held every two or three years. In the period
of our sample, three main political parties competed for political control: the left-wing Party
of the Democratic Revolution (PRD), the populist PRI, and the right-wing National Action
Party (PAN).10 Competition in most parts of the country was generally between only two of
these parties, with the PRI performing best in rural areas and the PAN and PRD performing
best in urban areas (Larreguy, Marshall and Querubın forthcoming).
3.1 Municipality audits
In Mexico’s federal system, states and municipalities exercise significant control over local
policy. In 2010, Mexico’s 31 states contained 2,435 municipalities, varying in population from
102 to 1.8 million with a mean of 46,134. Following major fiscal decentralization reforms
in the 1990s, the average municipality annual budget has been around nine million U.S.
dollars, which constitutes 20% of total government spending.11 Municipal governments are
led by mayors, who are responsible for delivering basic public services and managing local
infrastructure. Mayors are normally elected every three years, although they serve four-year
terms in some states, and could not stand for re-election.12
9In the Chamber of Deputies, 300 members are elected via plurality rule from single-member districts and 200 members are elected via proportional representation. The Senatecomprises 128 Senators, where three are allocated from each state (including the FederalDistrict) where the largest party receives two Senators and the second largest receives oneSenator, and a further 32 allocated according to the national vote share.
10The PRD’s candidate in the 2006 and 2012 presidential elections—Andres ManuelLopez Obrador—left the PRD to form a new party, the National Regeneration Movement(MORENA).
11Education and health were decentralized between 1992 and 1996 and the decentralizationof infrastructure projects followed in 1997 (Wellenstein, Nunez and Andres 2006).
12Re-election will become possible for those running starting in 2015.
10
An important component of a mayor’s budget is the Municipal Fund for Social Infras-
tructure (FISM). On average, this represents 24% of a municipality’s total resources. FISM
funds, which are allocated to municipalities according to the Fiscal Coordination Law (LCF)
passed in 1997, are direct federal transfers provided exclusively for the funding of public
works, basic social actions, and investments that directly benefit the socially disadvantaged
population living in extreme poverty. Spending may be allocated in any of the following cat-
egories: potable water, sewage, drainage and latrines, municipal urbanization, electrification
or rural and poor suburban areas, basic health infrastructure, basic education infrastructure,
improvement of housing, rural roads, and rural productive infrastructure.
The use of FISM funds is subject to independent audits by Mexico’s Federal Audi-
tor’s Office (ASF). The ASF, which was established in 1999 in response to widespread
concerns regarding the mismanagement of public resources, is an independent body with
constitutionally-enshrined powers to audit the use of federal funds by the federal, state and
municipal governments. In each year since 2000, the ASF audits FISM spending in multiple
municipalities per state. Audits focus on the spending and management of FISM resources in
the prior fiscal year, and the list of municipalities to be audited in a given year is announced
the year after the spending occurred. Between 2007 and 2012, 14% of Mexican municipalities
were audited at least once, with around 120 municipalities being audited each year.
Although municipalities are not randomly chosen to be audited, the timing of an audit
is essentially random. The appendix to the ASF’s summary report makes clear that only
the following criteria are taken into account: the financial importance of FISM funds to the
municipality, relative to the municipal budget; historical performance indicators and insti-
tutional weaknesses that raise the likelihood of misallocation; whether FISM spending has
recently been audited; whether other federal audits are occurring simultaneously; and where
a specific mandate exists to examine a particular municipality.13 Direct communication with
the ASF confirmed these criteria, and clarified that an audit should not have occurred within
the last two years, and that—for logistical reasons—they often simultaneously audit neigh-
boring municipalities.14 Crucially, given that our identification strategy exploits the timing
13This information is formally stated on page 240 of their 2014 summary report, “In-forme del Resultado de la Fiscalizacion Superior de la Cuenta Publica 2012”, available here.Federal auditors may also audit Funds for the Strengthening of Municipalities and FederalDemarcations of the Federal District (FORTAMUNDF) funds allocated in proportion to thenumber of citizens and intended to strengthen municipal social spending, or Subsidy forSecurity in Municipalities (SUBSEMUN) funds allocated to support public security.
14Based on a personal interview with the Licentiate Jaime Alvarez Hernandez, GeneralDirector of Research and Evaluation of the Special Audit of Federal Spending, in July 2012,
11
of audits, our correspondence with the ASF also confirmed that the selection of municipali-
ties for audit does not depend upon the electoral cycle or the government’s political identity.
Ensuring that this claim applies in practice, our balance tests below confirm that the timing
and content of audit reports are uncorrelated with a wide variety of political, demographic,
and socioeconomic indicators.
Independent ASF auditors check that officials abide by the rules established for the
management of FISM resources (e.g., procurement rules, accounting procedures), that the
status of the funded projects is in accordance with the books, and that funds are given the
use they were intended for. Reports break down the use of FISM funds across multiple
dimensions; an example report is provided in the Online Appendix. Most importantly,
reports state the percentage of FISM funds spent on projects not benefiting the poor and the
percentage of funds used for unauthorized spending. Spending that does not benefit the poor
ranges from the diversion of resources to support agricultural production in areas without
poverty to paving the streets of relatively rich urban areas. We interpret unauthorized
spending, which includes the diversion of resources for personal expenses of the mayor or
electoral campaigning and funds that are unaccounted for, as corruption.15 Audit reports are
publicly released two years after the spending actually occurred, when they are presented in
Congress by the last working day of February each year and made publicly available online
at the ASF’s website.16
Relative to previous social programs, FISM funding has been comparatively successful at
targeting resources at the poor (Wellenstein, Nunez and Andres 2006). However, funds are
often misallocated. Unlike previous studies focusing on corruption in more general programs
(e.g. Ferraz and Finan 2008; Bobonis, Fuertes and Schwabe 2013), the specific targeting of
FISM funds allows us to examine the electoral response by voters to both corruption and
the misuse of funds intended to serve a disadvantaged population.
The ASF can impose a variety of punishments on malfeasant public officials. In par-
ticular, the ASF can inflict fines on the municipality to recover FISM funds, recommend
that the Ministry of Public Function removes, suspends or imposes economic sanctions on
officials, or file (or recommend) a criminal case against culpable individuals. In practice,
and formal email correspondence with the ASF.15This definition resembles Ferraz and Finan (2008) in that we focus on violations that
include procurement fraud, diversion and over-invoicing, but differs in that we quantifythe relative importance of such corruption. Rather than the percentage of unauthorizedspending, Ferraz and Finan (2008) count the number of corruption violations.
16Guıa para el ciudadano. Que es y que hace la Auditorıa Superior de la Federacion?
12
these punishments have not been used regularly: between December 2006 and July 2012,
the Ministry of Public Function only recovered two million U.S. dollars, sanctioned 9,000
public employees for serious misdemeanors, and incarcerated one hundred officials.17
The largest punishment may be electoral. Since Mexican mayors cannot stand for re-
election, any electoral penalty hits the party of a malfeasant mayor. This feature differen-
tiates our study from many preceding studies (e.g. Banerjee et al. 2011; Ferraz and Finan
2008; Humphreys and Weinstein 2012). However, there are good reasons to believe that a
mayor’s political party may be punished by voters at the next election. First, although some
voters are aware of particular candidates, Mexico’s strong party system ensures that party
labels play a key role in voting decisions. Previous studies show that voters are poorly in-
formed about local politics, and that voters are willing to punish the party’s of local mayors
for their actions in office (e.g. Chong et al. forthcoming). Second, the top-down internal
structure of Mexican parties at the state level ensures that within-party candidate choice is
highly correlated at the local level (Langston 2003).
3.2 Local broadcast media
As in many developing countries, radio and television stations are the principal source of
news in Mexico. Conditional on providing news, both radio and television stations provide
around 2 hours of news coverage a day.18 While a few television stations are devoted to
national news, the focus of radio stations and most television stations is predominantly
local. In fact, while the majority of entertainment content is common across media stations
within Mexico’s large media networks, news programming is typically local. Such content
sharing among major radio networks such as Radiorama, ACIR, Radiocima, Multimedios
Radio and MVS Radio, and Mexico’s two main television networks, Televisa and TV Azteca,
allows Mexico’s many small media stations to support themselves through local advertising
revenues including government advertising expenditures.
Voters are generally unaware of mayoral responsibilities and the use of public funds
(Chong et al. forthcoming), as well as the institutions that are responsible for auditing the
use of public resources (Castaneda Sabido 2011). Most public spending is invisible and
inaccessible to most voters. A study by Castaneda Sabido (2011) indicates that only 33.6
% of surveyed individuals think that municipal governments are transparent about the use
17El Universal, “A la carcel, solamente 100 ex servidores”, 29th May 2014, link.18These figures are based on IFE monitoring of a non-random sample of 200 radio and
television stations providing news coverage during the 2012 Mexican Federal election.
13
of public resources.19 Moreover, only 25% of surveyed individuals can mention a public
institution in charge of auditing the use of public funds, and only 1.4% of those individuals
mention the ASF as the main institution responsible for that task.
Voters learn about public spending primarily through media coverage. Figures from the
2009 Latinobarometer indicate that 83% of respondents gather political information from
television, 41% gather political information from radio, 30% gather political information
from newspapers, and 41% gather political information from family, friends and colleagues
(many of whom, of course, gather their information from television, radio and newspapers).20
Internet is not widespread: according to the 2010 Census, only 24% of households in the
average electoral precinct have internet access. Additionally, according to Castaneda Sabido
(2011), 83% of individuals report that they receive information about malfeasance in the
management of public resources through media, and 61% regard such information as reliable.
We thus expect that television may be the most the important media source for electoral
accountability.
Furthermore, the likelihood that a voter follows the news increases with the availability
of local media. Using detailed data on media consumption from the 2009 CIDE-CSES Sur-
vey, the Online Appendix demonstrates a strong positive correlation between access to local
media stations (defined in detail below) and consumption in a predominantly urban sam-
ple. In particular, an additional local radio and television station respectively increases the
probability that an individual regularly watches a news program by 0.4 and 1.4 percentages
points, as well as increasing the total number of news programs regularly watched by 0.005
and 0.021 respectively. In total, 80% of respondents listed at least one television program,
and 30% listed one radio program. Conversely, we find no evidence that the number of
non-local media stations, which are less likely to provide local news coverage, are associated
with news consumption.
The release of municipal audit report results each February is a major annual media
event, especially for television stations which extensively cover audit report outcomes. As in
many other countries (see Pande 2011), revelations about political malfeasance are widely
19In principle, local governments are required to inform the public about the arrival ofFISM funds. However, only about 50% comply with this requirement. Moreover, amongthose that do comply, the main communication channels used are newspapers and theinternet—i.e., two types of mass media. Furthermore, media relies extensively on the ASFaudit reports since governments are extremely reluctant to release information about theirexpenses to the public (Lavielle, Pirker and Serdan, 2006).
20These are the only four responses to an open-ended question that received a non-negligible number of mentions.
14
covered, and—as indicated by the many news stories we found from later in the year—
continue to remain salient news for many months. News reports generally cover mayors
within the local vicinity, and typically focus on cases of corruption and mayors not spending
FISM funds on projects targeting the poor.21 Most reports accurately cite the proportions
of unauthorized spending and spending on projects not targeted at the poor, and many dig
deeper to describe the nature of the malfeasance. There are many such examples available
online.22 For example, in 2013 BBM radio station reported that Oaxaca de Juarez’s mayor
had created a fake union to collect payments, presided over many public works contracts
without offering open tender, diverted payments for advertising and consulting fees, and
failed to provide details of considerable quantities of spending.23 While this particular case
represents one of the most corrupt mayors, such behavior was not uncommon: many media
reports pointed to mayors diverting payments, using FISM funds for personal and family
expenses and manipulating tender processes. Failures to spend FISM funds on the poor were
just as common in media reports. In many cases, public works projects were undertaken in
urban and affluent parts of the city. In others, the alleged project never materialized despite
being paid for, or was diverted for alternative uses such as supporting local candidates from
the incumbent’s party. To more systematically demonstrate the substantial coverage of the
audit report releases, Figures 1a and 1b show spikes in Google searches for the ASF and
FISM around February and March of each year.
We now turn to our data, and to the empirical strategy we use to identify the effects of
the release of audit reports in electoral precincts that differ in the number of local media
outlets they are covered by.
4 Data
This section describes our main sources of data: incumbent electoral performance at the
electoral precinct level; municipal audit reports released just before and just after an election;
and precinct-level radio and television coverage.
21Little mention was made of other features of the reports such as the the degree ofparticipation of the community in the allocation of funds or the share of FISM funds thatwere spent.
22For example, see: BBM Noticias, “ASF: desvio Ugartchechea 370.9 mdp”, October21st 2013, here; El Informador, “Hallan irregularidades en gasto tapatıo contra pobreza”,February 28th 2013, here; Revolucion Tres Punto Cero, “En 2012, se desviaron a campanas29 millones de pesos para combate a la pobreza en Tabasco”, March 6th 2014, here.
23BBM Noticias, “ASF: desvio Ugartchechea 370.9 mdp”, October 21st 2013, here.
15
4.1 Mayoral election outcomes
Mexico’s municipalities are divided into approximately 67,000 electoral precincts. Using
data from the Federal Electoral Institute (IFE) and State Electoral Institutes, we collected
electoral returns for every available precinct in each municipal election between 2004 and
2012. We thus accumulated up to four election results per electoral precinct, which enabled
us to identify the municipal incumbent and incumbent’s past vote share in all the elections
in our period of analysis, 2007-2012.24
We focus on two main outcomes: the change in incumbent party’s vote share at the
precinct level, and whether the incumbent party was re-elected at the municipal level. The
former measure quantifies the extent of precinct-level voter sanctioning, while the latter
captures the municipality-level implications for the identity of the office-holder. When ex-
ploiting fine-grained variation in media coverage across precincts, our analysis solely focuses
on changes in the incumbent vote share. We define the vote share by the proportion of
voters that turned out.25 The average incumbent party in our sample received 48% of votes
in the average electoral precinct, which represents a 4.7 percentage point decline in their
vote share.
Since Mexican mayors cannot stand for re-election in our sample, we focus on the party
of the incumbent mayor. However, municipal politics often entail the formation of local
coalitions between political parties, and this can change across elections. For example, in
2009 the incumbent mayor of the municipality of Colima represented a two-party coalition
containing the PRI and the Green Party (PVEM). However, the 2009 election saw six groups
stand for election: the PT, PVEM and PC all stood separately against three coalitions, PAN-
ADC, PRI-PANAL and PRD-PSD.To classify such cases where the incumbent coalition split
at the next election, we determined the party affiliation of the mayor by researching their
identity and party ties.
4.2 Audit reports
Since audit reports are released with a two year lag, reports released in the February of a
municipal election year generally refer to the first year the incumbent mayor was in office.26
24By ending our sample in 2012, our sample does not feature any mayors that can seekre-election.
25We obtain similar results when measuring vote share as a proportion of registered voters.26In Coahuila, where mayors are elected to four-year terms, the report refers to the second
year of their term.
16
Since municipal elections take place later in the calendar year, we define a pre-election audit
report release by whether an audit was released in February of an election year.27 Typically,
the report is released 4 months before the election. Our control group will be mayors in
municipalities where the audit was released in February the year following the election. In
such cases, the audit report generally pertains to their second year in office.
The results of audit reports, which quantify the use of FISM funds, are publicly available
on the ASF’s website. We extracted the proportion of funds spent in an unauthorized
manner and the proportion of funds not spent on projects benefiting the poor from every
available report between 2007 and 2012.This yielded a total of 742 municipal audits, which
were relatively evenly spread across years and covered 351 unique municipalities. Of these,
470 reports from 321 different municipalities were released in an election year or the year
after. We henceforth restrict attention to this subsample of audits, which are shaded by
their levels of malfeasance in Figure 2.
[Figure 2 about here.]
We operationalize malfeasance using indicators to capture severity. For corruption, we
define indicators for precincts with mayors in the third and fourth quartiles of the distribution
of unauthorized FISM spending. Unauthorized spending in the third quartile ranges from
0.6% to 11.2% of available FISM funds with a mean of 5.1%, while spending in the fourth
quartile exceeds 11.2% with a mean of 30.7%. For neglectful spending, we similarly define
indicators for mayors in the third and fourth quartiles with respect to FISM funds not
allocated to spending on the poor. Not spending on the poor in the third quartile contains
any positive value between up to 12.9% of available FISM funds with a mean of 5.7%, while
spending in the fourth quartile ranges exceed 12.9% with a mean of 38.0%. Since around 50%
of precincts did not experience corruption or not spending on the poor, the 25th percentile of
each distribution is 0; hence we do not use an indicator for the second quartile. We employ
binary performance metrics—which are also more flexible than linear measures—identifying
more egregious cases of bad performance since standard theoretical models suggest that
voter sanctioning involves cut-off rules (e.g. Barro 1973; Ferejohn 1986). Furthermore, our
examination of media reports indicates that only relatively serious cases are widely reported.
Nevertheless, our robustness checks show similar results when using a continuous measure
27Although states differ in the month in which they hold elections, only the state ofBaja California Sur holds elections before mid-February. We adjust for Baja California Suraccordingly.
17
where we instead allow sanctioning to be a linear function of revealed performance, and also
when examining cutoffs for different levels of malfeasance.
4.3 Radio and television coverage
In addition to our fine-grained electoral data, a key feature of this study is the detail of our
media coverage data. Following a major media reform in 2007 (see Serra 2012), the IFE
required that every AM and FM radio station and every television station in the country
provide signal coverage data.28 Specifically, for each media station we are able to define
the municipality from where it broadcasts, as well as the commercial quality coverage range
of its signal.29 Inside a station’s coverage area the signal is of high quality, so precincts
inside the area have good access to the station’s broadcasts. Precincts outside the coverage
area experience increasingly poor coverage as the distance from the boundary increases. See
Larreguy, Marshall and Snyder Jr. (2014) for further details of the coverage data.
Figures 3-5 map the location and coverage of each of the 852 AM, 1097 FM and 1255
television stations. Although media coverage is extensive, with most precincts receiving
at least one media signal and most municipalities containing at least one media station,
there is considerable variation in the number of media stations covering each precinct that
emit from within the precinct’s own municipality.30 The figures also clearly indicate that
the commercial quality coverage range of AM radio is substantially greater than for FM
and television. However, as discussed in detail below, we restrict attention to more urban
precincts that primarily differ in terms of FM and television signals.
[Figures 3, 4 and 5 about here.]
Our principal measure of local media coverage is the total number of AM, FM or tele-
vision stations covering a given electoral precinct that broadcast from within the precinct’s
municipality. To avoid counting signals that barely cover a given precinct, we use (urban)
28For only a small number of FM and television stations did the same station broadcastfrom multiple municipalities. No electoral precincts received the same signal from multipleantennae.
29The IFE defines the boundary of the coverage area using a 60 dBµ threshold for signalstrength. This is the threshold commonly used to determine a radio station’s audience andsell advertising space commercially in the U.S., where it “is recognized as the area in whicha reliable signal can be received using an ordinary radio receiver and antenna” (NTIA link).
30Since the number of radio and television stations has remained constant between 2003and 2010, we cannot exploit temporal variation in media coverage.
18
block and (rural) locality-level population data from the 2010 Census to define a precinct as
covered by a given media station only if at least 20% of its population lies in side the com-
mercial quality coverage boundary. The average precinct is covered by 4.4, 5.8 and 2.5 local
AM, FM and television stations respectively, while the total number of local media stations
covering a precinct ranges from 0 to 44. Given these precinct totals are highly correlated
across media types, simply adding the total together yields similar results to examining each
type of media separately.31
To compare the effects of local media to non-local media, we also consider the number
of media stations that cover a precinct but transmit from outside their municipality. The
average precinct receives roughly as many FM and television signals from inside their mu-
nicipality as outside, although the greater signal range of AM stations means that precincts
are typically covered by three times as many AM stations emitting from outside their mu-
nicipality.
5 Empirical strategy
Our goal is to identify the effect of local media coverage of municipal audits on the incumbent
party’s electoral performance. To achieve this, we exploit exogenous variation in both the
timing of audit report releases around elections and access to local media. In particular,
we combine the difference-in-differences design of Ferraz and Finan (2008) with plausibly
exogenous variation in the number of local media stations covering neighboring electoral
precincts.
5.1 Identifying the effects of audit reports
The difference-in-differences (DD) component of our design rests upon exogenous variation
in the timing of audit report releases. To first identify the effects of audits we compare
municipalities where an audit report was released just before a municipal election to a control
group of municipalities where the audit was released after a municipal election. We then
move beyond this first difference by also comparing municipalities where the mayor is corrupt
or neglectful. The DD sample contains 47,938 precinct-election observations.
Conditional on our sample of municipalities that have been audited at least once, the
identifying assumption required to estimate the effects of audits released just before an
31This procedure yielded a Cronbach’s alpha of 0.80. The minimum pairwise correlationbetween the variables is 0.57.
19
election is that the timing of the releases is effectively random. As discussed in detail above,
the independent ASF’s procedure for determining which municipalities will be audited in a
given year is apolitical: the timing of elections does not feature in their selection rule.
The summary statistics in Table 1 empirically support this identifying assumption. Con-
sistent with the ASF’s claim that selection is independent of electoral considerations, difference-
in-means tests confirm that differences between electoral precincts in municipalities where an
audit was released in the year before an election and precincts where an audit was released
the following year across 49 political, demographic, media, and economic characteristics
are consistent with chance. Specifically, we find only 1 statistically significant difference
(which is only statistically significant at the 10% level). The final 26 variables are from the
precinct-level Census data from 2010, and are described in detail in the Online Appendix.
[Table 1 about here.]
A second potential concern is that FISM spending decisions—the content of audit reports—
might differ across reports released before and after elections. This could reflect differences
in the behavior of either auditors or mayors. First, even though the timing of audits is
effectively random, auditors could still be more lenient or more meticulous in the knowledge
that a report will be released in an election year. Second, mayors anticipating the release
of an audit report in an election year may spend more appropriately (Bobonis, Fuertes and
Schwabe 2013). Alternatively, the relative inexperience of first-year mayors—about whom
audit reports are released before an election—may induce them to allocate their funds dif-
ferently from second-year mayors (Chattopadhyay and Duflo 2004).
[Figure 6 about here.]
However, there is no empirical support for such concerns. Table 1 shows that there is
no systematic correlation between pre-election audits release and either the proportion of
unauthorized spending or the proportion not spent on the poor. This holds regardless of
whether we measure such spending on average or in terms of being in the third or fourth
quartile of either distribution. Furthermore, Figure 6 compares the report outcome distribu-
tions across audit reports released just before and just after an election.32 The distributions
of unauthorized spending and spending not on the poor are very similar, and in neither
32The graphs are very similar if we compare audits from election years to all non-electionyears (i.e. not just including audits released in the year after a municipal election).
20
case does a Kolmogorov-Smirnov test reject equality of distribution.33 Combined with our
randomization check, this strongly suggests that the audits results released in election years
are typical of “normal” FISM spending.
Given that the timing of audit report releases, and their content, is effectively random,
we estimate the following simple DD specification to identify the effect of revealing a mayor
to be corrupt or neglectful before an election:
Yp,m,t = β1auditm,t + β2outcome Q3m,t + β3outcome Q4m,t + β4
(auditm,t × outcome Q3m,t
)+β5
(auditm,t × outcome Q4m,t
)+ εp,m,t, (1)
where Yp,m,t is the incumbent party’s vote share in precinct p in municipality m in year t
(or whether the incumbent party won the election in municipality m), auditm,t is an in-
dicator for an audit being released in the year before the election, and outcome Q3m,t and
outcome Q4m,t are indicators for municipalities in the third and fourth quartiles of the distri-
butions of corrupt or neglectful mayors (regardless of whether the audit was released before
or after the election). Since precincts differ in size, we weight each observation by the number
of registered voters.34 Throughout, we cluster by municipal election to account for spatial
correlation between precincts receiving the same audit report.
Our main coefficients of interest are β4 and β5, which identify the effect of an audit
conditional upon it revealing corruption, or that the mayor did not—as legally required—
spend FISM money on the poor. We are also interested in β1, which identifies the effect of
an audit conditional upon it revealing no malfeasance.
5.2 Identifying the effects of local media stations revealing audit
reports
To examine how media affects whether voters punish malfeasant behavior, Ferraz and Finan
(2008) further interact auditm,t × outcomem,t with the number of AM stations located in a
municipality. If the number of AM stations were effectively randomly assigned, then this
would estimate the average effect of an audit report being released for each additional local
33Collapsing to the municipality-audit level (with 470 observations) to avoid duplicationacross precincts, the p values of the corruption and not spending on poor distributions arerespectively 0.39 and 0.99.
34Given that electoral precincts were all designed to contain up to 1,500 voters, precinctnumbers remain relatively similar, and the unweighted results yield very similar estimates.
21
media station.35
However, in general, media stations are not randomly assigned across municipalities. As
we show in the Online Appendix, the number of local media stations received by an elec-
toral precinct in our Mexican sample is significantly correlated with almost every Census
characteristic in Table 1. Local media is significantly more prevalent in more highly de-
veloped, urban and politically uncompetitive precincts. These correlations may upwardly
bias our estimates of local media’s effects if, for example, the better educated and informed
citizens in such precincts are more willing or able to sanction incumbent mayors revealed
to be malfeasant (e.g. Alt, Lassen and Marshall 2014; Weitz-Shapiro and Winters 2014).36
Controlling for the available observables does not fully address the major concern that local
media is actually picking up the effect of correlated local characteristics.
To identify the effects of local media coverage, we compare neighboring electoral precincts
from within a given municipality that differ in the total number of local AM radio, FM radio
and television stations that they are covered by (for similar designs, see Ansolabehere, Snow-
berg and Snyder 2006; Enikolopov, Petrova and Zhuravskaya 2011; Fergusson 2014; Snyder
and Stromberg 2010). A media station is local to a precinct if it emits from within the same
municipality. As explained above, we define a precinct as covered by a given media station
if at least 20% of its population lie inside the commercial quality boundary. Since broadcast
signals decay gradually rather than abruptly, and whether any given household receives a
signal may depend upon the quality of their receiver, discrete differences in commercial qual-
ity signal coverage do not imply that neighboring precincts differ strictly between receiving
or not receiving a station’s signal. Rather, our design estimates the “intent to treat” effect
of differences in the proportion of an electoral precinct that can receive commercial quality
local media.37 Our estimates will thus understate the effect of an terminal drop in coverage.
35Furthermore, unlike our precinct-level data, this strategy rests upon between-municipality differences in media coverage.
36In theory, these correlations could also downwardly bias our estimates if such precinctscontain voters with stronger prior beliefs about their incumbent’s quality (Zaller 1992).
37Ideally, we could also identify the electoral effect of receiving or consuming an additionalmedia station using instrumental variable techniques. To estimate the relevant first stage,we would need to measure either the proportion of voters in each precinct that can access allmedia stations or the proportion of voters that actually listen to each radio stations or watcheach television station. Unfortunately, such detailed individual-level data is not available.Survey datasets typically cover only 1-2% of all electoral precincts and never ask specificallyabout which radio or television stations voters have access to or actually consume. Althoughthe Comparative Study of Electoral Systems and Mexican Panel surveys did ask whetherrespondents listen to the radio or television, the surveys are predominantly urban and cover
22
Similarly, any spillovers (e.g. arising from driving to work across radio coverage boundaries)
would further attenuate our estimates toward zero.
Our identifying assumption is that neighboring precincts differ only in their local media
coverage. Restricting attention to within-neighbor variation removes a wide variety of po-
tential confounds. Ideally, otherwise similar precincts near the commercial quality coverage
boundary only vary due to exogenous and fixed signal impediments or facilitators such as
large physical objects, terrain, and salt water that affect ground conductivity (in the case
of AM long waves) and line of sight (in the case of FM and television waves). As noted
above, the coverage maps provided by the IFE account for such obstacles and facilitators,
as well as the frequency and power of transmitters signals. While the conductivity of AM
signals is sensitive to variable weather conditions and the night-time ionosphere, FM radio
and television coverage is relatively constant because such waves travel by line of sight. Fur-
thermore, focusing on within-municipality neighbors removes municipal political differences
and ensures that neighboring precincts vote on the same incumbent.
Nevertheless, strategic sorting represents an important concern. Our estimates would be
biased if certain types of voters choose to live in areas with better local media coverage or
media stations strategically choose the strength of their signal to exclude certain types of
voters. However, such sorting is unlikely. First, if voters were migrating according to media
availability, they would likely choose a location close to the antennae, rather than near the
commercial quality coverage boundary, in order to guarantee high-quality signal coverage.
Second, media stations lack the technology to precisely target certain types of voters: beyond
the fact that excluding voters is challenging when signals are not discontinuous, the antennae
strengths that media stations purchase are highly discrete.38
To maximize the plausibility of our identification strategy, we focus on relatively urban
precincts. Specifically, our sample only contain precincts with at most an area of 10 km2.
Between such neighbors, strategic sorting is particularly unlikely because media stations
only 1-2% of electoral precincts. The Latinobarometer, which also asks basic questionsabout media consumption, does not provide precinct-level identifiers for its respondents.Even if such surveys had greater coverage, none of the surveys could identify the numberor identity of the media available to voters—such measures would be necessary to computethe relevant first stage. Furthermore, since voters are likely to discuss the news that theyreceive with their friends and family, the exclusion restriction requiring that a commercialquality coverage signal only affects electoral outcomes through either access or especiallyconsumption is hard to sustain.
38The power output in watts for the AM, FM and television stations in our sample arealmost exclusively round thousands and divisible by 5.
23
cannot plausibly choose technologies to separate markets. Given the substantial reach of AM
stations, a second key advantage of this design is that more urban areas almost exclusively
differ with respect to FM radio and television coverage. Removing the most volatile media
signal maximizes the accuracy of our coverage measures. Moreover, this area restriction
prevents rural-urban comparisons, which could be problematic because distance from urban
areas may be correlated with other politically-relevant characteristics and media signals decay
at different rates in rural areas. Supporting the validity of this approach, our balance tests
show that only 2 out of our 36 precinct-level balancing variables are significantly correlated
with local media (see below).39
Figure 7 illustrates our empirical strategy graphically. Electoral precincts 1571 and 1583
in the municipality of Villa de Tututepec de Melchor Ocampo are neighbors, but differ
because only precinct 1583 is covered by a television station emitting from within the mu-
nicipality. Since some neighbors differ by more than one local media station,40 we include
neighbor group fixed effects to ensure that our estimates are not confounded by differences
in the types of places where neighbors differ by one as opposed to two (or more) local media
stations. Operationally, we define a “treated” precinct as one which differs from at least
one neighboring precinct in terms of the number of local media stations that it receives.
For each such precinct, we then collect all possible neighboring “control” precincts that re-
ceive a different number of local media stations. This yielded a sample containing 17,312
observations, where neighbors typically differ by one or two local media stations. Given AM
radio typically travels substantial distances, focusing on relatively urban neighbors typically
identifies off differences in access to FM radio and television stations.
[Figure 7 about here.]
To combine variation in the timing of audit and the number of media stations covering a
given electoral precinct, we estimate the following triple-difference (DDD) specification using
39Relative to Table 1, we removed municipality level variables and, naturally, local media.4054% of neighbor pairs differ by more than one local media station. Therefore, we cannot
simply compare treated and control units because the difference in the number of mediastations between the two groups (or the treatment intensity) is not constant.
24
our neighboring precincts sample:
Yp,k,m,t = β1auditm,t + β2outcomem,t + β3mediap,m
+β4
(auditm,t × outcome Q3m,t
)+ β5
(auditm,t × outcome Q4m,t
)+β6
(auditm,t ×mediap,m
)+ β7
(auditm,t × outcome Q3m,t ×mediap,m
)+β8
(auditm,t × outcome Q4m,t ×mediap,m
)+ ξk + εp,k,m,t, (2)
where mediap,m is the total number of local media stations. Including a neighbor group
fixed effect ξk ensures that our DDD estimates identify only off within-neighbor variation in
media coverage (given we restricted attention to neighbors within a given municipality). In
addition to weighting by the number of registered voters in the precinct, we further divide
this weight by the number of control units per comparison to give each neighbor comparison
equal weight. Using a similar design, we also examine the effect of the total number of
non-local media stations, i.e. those broadcasting from outside the municipality.
The coefficients β6, β7 and β8 respectively identify the electoral effect of an additional
media station for a precinct in a municipality where the audit report reveals the incumbent
party not to be malfeasant, to be in the third quartile of malfeasance, and to be in the fourth
quartile of malfeasance. The coefficients β4 and β5 estimate the effect of an audit report in
the absence of media. These coefficients thus capture how media coverage supports electoral
accountability—both in terms of punishing malfeasant incumbent parties and rewarding
incumbents that correctly allocate federal transfers. By exploiting within-neighbor variation,
we can plausibly identify the average effect of an additional media station. Unfortunately,
however, the demanding structure of our identification strategy means that we lack the power
to non-parametrically estimate media’s effect.41
41The ideal non-parametric approach would be to allow media’s effect to vary for numberof media stations. However, by requiring 4 coefficients for each of the 40 levels in our data,we quickly lose power and rely on cells with very little support. A second best approachcould involve using multiple categories (e.g. at least ten, twenty, thirty media stations etc.)or quantiles. However, since 90% of neighbors only differ by one or two local media stations,using broad categories eliminates most of this variation because most neighbor comparisonwill not cross the threshold defining media intensity categories. Rather than taking theaverage marginal effect, this approach would estimate the marginal effect around a giventhreshold, e.g. among neighbors with 9 and 10 and 9 to 11 media stations. By ignoringwithin-neighbor variation away from the threshold, this simply yields an under-poweredaverage effect for an additional media station for a somewhat arbitrary threshold.
25
[Table 2 about here.]
To assess the plausibility of the design, we test whether local media predicts pre-treatment
covariates using specifications akin to equation (2). Despite our restriction to relatively
urban areas and our arguments above, a key concern is that precincts receiving more media
are likely to be closer to the municipal center, where antennas are typically located, and
thus differ in politically salient ways. To examine this possibility, Table 2 tests for balance
over distance to the centroid of the municipal head (from both a precinct’s border and its
centroid), area, the total population and number of registered voters, and population density
(all as natural logarithms). The results are consistent with local media’s random assignment
between neighbors: for none of these variables we register a statistically significant difference,
while the variation in coefficients directions is consistent with chance rather than access to
local media correlating with distance from relatively large urban developments. Crucially for
distinguishing the effects of local and non-local media, we also find no relationship between
the number of local and non-local media stations. Furthermore, we find no evidence that the
number of local media stations covering a precinct predicts standard development measures
such as education, basic housing needs, or owning more luxurious items like computers,
washing machines or cell phones. Similarly, we find no correlation between the presence
of local media and prior political behavior: an additional media station is insignificantly
and negligibly associated with incumbent electoral performance, local political competition
(proxied by the effective number of political parties) and electoral turnout. Together, our
wide range of balance tests thus offer substantial support for our identification strategy.
6 Results
We first briefly examine whether audits revealing the incumbent party’s mayor to be corrupt
or neglectful before an election reduces the party’s vote share and probability of retaining
office. However, our main contribution is to identify the interactive effect of local media
and the pre-election release of incumbent performance information in holding the incumbent
party to account. Our results demonstrate that an additional local media station significantly
increases the electoral punishment that the party of a corrupt or neglectful mayor faces.
Such differences in access to local media may explain why the average effects of audit report
releases are noisily estimated. Reinforcing the importance of local media, we also show that
non-local media do not enhance electoral accountability and crowd out the effects of local
26
media. Finally, we show that local media is most effective when the station’s market is
predominantly based within the municipality.
6.1 Audits and electoral accountability
Table 3 presents our DD estimates of the average effect of an audit revealing a mayor to
be corrupt or neglectful of the poor on the mayor’s electoral prospects. The outcome in
columns (1) and (2) is the change in the incumbent party’s vote share at the precinct level,
while the outcome in columns (3) and (4) is whether the incumbent party was re-elected in
the municipality.
[Table 3 about here.]
The results suggest that an audit report released may have important electoral impli-
cations. Although the point estimates are relatively large, they are imprecise and never
statistically significant. Column (1) suggests that revealing the incumbent party’s mayor
to be in the most corrupt quartile before the election, on average, reduces the party’s vote
share by 4.5 percentage points. Although the effect is not statistically significant, the mag-
nitude represents 9% of the average incumbent’s initial vote share. Column (2) suggests
that revealing the incumbent party’s mayor to have been neglectful similarly reduces their
vote share: the vote share of mayors in the third quartile declines by 2.6 percentage points,
while mayors in the fourth quartile lose a further 4.2 percentage points. In both cases, the
audit coefficient in the first row—which captures the baseline category of essentially zero
or negligible malfeasance—suggests that the parties of mayors whose reputations are not
negatively affected by the audits may increase slightly their vote share, especially when they
actually spent the FISM funds on its intended poor recipients.
Looking at the probability of re-election similarly suggests that voters may severely punish
mayoral malfeasance. Measured at the municipal level, the change in incumbent vote share
maps to large but very imprecise reductions in the probability of being re-elected. Column (3)
finds that revealing a mayor as one of the most corrupt reduces their re-election probability
by 29 percentage points, although this is not quite statistically significant.42 Column (4)
shows that the publication of an audit report showing that a mayor did not spend FISM
federal transfers on the poor is 17 percentage points less likely to be re-elected. For both
42The particular relative lack of precision for the incumbent win probability reflects thefact that we have 481 audited municipalities, of which only 51 had mayors that were revealedto be corrupt before the election.
27
audit outcomes, the effect is much larger for mayors in the fourth quartile relative to the
third, where the estimate is surprisingly positive.
The electoral sanctioning suggested by these results is broadly similar in magnitude to
that found by Ferraz and Finan (2008) in Brazil, although we measure corruption in terms of
stolen funds rather than the number of corrupt spending violations. However, our results also
suggest that incorrectly spending money earmarked for the poor could also evoke sanctioning
of similar magnitude to corruption.43 However, our DD estimates are very noisy. A plausible
explanation for the lack of precision is that audit reports only affect voter behavior when
the information is effectively conveyed by local media stations.
6.2 Broadcast media audit report coverage
We now address the central question of this article: is the party of a malfeasant mayor more
likely to be sanctioned by voters who live in areas covered by media stations that publicize
audit reports revealing the mayor’s behavior in office? Combining our DD and the within-
neighbor design, we first estimate the effects of local media stations –those emitting from
the municipality that the electoral precinct belongs to–before turning to non-local media
stations. Since mayoral corruption and neglect are primarily important local issues, we
expect to find that local media is more effective at facilitating electoral accountability.
6.2.1 Importance of local media
Table 4 provides our estimates for the sanctioning effect of an additional media station
emitting from the precinct’s own municipality. Since we now focus on precinct-level variation
in media coverage, our analysis focuses on the change in the incumbent’s vote share at the
precinct level. Columns (1) and (2) provide our preferred estimates, while columns (3) and
(4) show that the results are robust to controlling for the presence of non-local media stations.
[Table 4 about here.]
We find that local media play a key role in supporting electoral accountability. Voters
only punish the incumbent party when an audit report released before an election reveals
the incumbent mayor to be corrupt only in the presence of sufficient local media stations.
43The Online Appendix reports quantitatively similar, but far noisier, estimates for theneighbors sample, which includes only a selection of precincts from 181 municipality electionsand only identifies out of within-neighbor variation.
28
Column (1) shows that, on average, mayors in the third and fourth quartiles of the corruption
distribution experience a significant loss in their vote share—almost 1 percentage point for
each additional local media station. A standard deviation increase in the number of media
stations, which entails 10.6 more media stations, thus reduces the vote share of an incumbent
revealed to be corrupt by around 7.5 percentage points. This represents half of the average
municipal victory margin (in the sample) of 15 percentage points. However, the insignificant
positive interactions between our audit dummy and a mayor’s corruption quartile indicate
that revealing a mayor to be corrupt is not punished electorally in precincts covered by no
(or few) media stations. On the other hand, clean incumbent parties are only meaningfully
rewarded for their good performance in the presence of local media stations. The statistically
significant coefficient on the interaction between being audited before an election and the
number of local media stations shows that an additional local media station publicizing
that the incumbent mayor did not engage in unauthorized spending increases the incumbent
party’s vote share by 0.7 percentage points.44
In precincts where local media reveals a mayor to have neglected the poor, we observe
slightly larger effects, and punishment is increasing in the severity of a mayor’s neglect.
Column (2) shows that an additional local media station reduces a neglectful mayor’s vote
share by 0.8 percentage points for mayors in the third quartile and 1.2 percentage points
for mayors in the most neglectful quartile. A standard deviation increase in the number
of local media stations thus entails an 13 percentage point decrease in the vote share of
the most neglectful mayors if their behavior is revealed before an election. This represents
a 25% reduction in their precinct vote share. Again, the insignificant interaction between
the pre-election audit release and not spending on the poor shows that in locations with
zero local media stations the party of the mayor is not meaningfully electorally sanctioned.
Furthermore, the significant positive interaction between revealing an audit and the total
number of local media stations shows that parties that were revealed to have correctly spent
all their FISM funds on projects benefiting the poor are boosted at polls, although not
as much as malfeasant mayors are punished. It is possible that voters are more likely to
believe that their municipal representatives are corrupt than neglectful, and thus positive
information that contradicts this prior is rewarded more.
44By contrast, there is no significant effect associated with the (unreported) interactionsbetween local media and audit report outcomes released after the election, for either corrup-tion or not spending on the poor. These coefficients are omitted to save space.
29
It is important to remember that the average proportion of funds not spent on the poor in
the third and fourth quartiles are slightly higher than for corruption. Nevertheless, the point
estimates suggest that the electoral impact of an additional local media station reporting a
fixed proportion of misallocated funds is, on average, around 25% higher for not spending
on the poor than corruption. The findings thus suggest that local-media revelations about
misallocated spending are at least as as important to voters as revelations about unauthorized
spending.
[Table 5 about here.]
Which types of media platform drive such electoral accountability? To better understand
local media’s effects, we implemented the same identification strategy separately for local
radio and television stations. Table 5 reports the main results for FM radio and television
when we only identify out of neighbors that respectively differ in the number of such media
stations that they receive.45 Consistent with television’s status as the main source of news
consumption, the effect of local media appears to be driven primarily by additional local
television stations. For both corruption and neglect of the poor, television has very large
effects on incumbent electoral performance: while each additional local television reduces a
malfeasant incumbent party’s vote share by 3-5 percentage points, a high-performing incum-
bent party benefits by a similar margin. Although television stations are far less prevalent
than radio stations, a standard deviation increase in local television stations nevertheless
translates into reducing the incumbent party’s vote share by almost 25% in the most egre-
gious cases. With a similar sample size, there is no clear sanctioning effect associated with
FM radio.
6.2.2 No role for non-local media
We now consider the role of non-local media. Columns (3) and (4) of Table 4 add interactions
with non-local media to our main estimates for the interaction between audit reports released
before an election and local media. First, and unsurprisingly given that local media is well
balanced across the number of non-local stations, we find that the local media point estimates
are unaffected. If anything, the estimates slightly increase and are more precisely estimated.
Second, the non-local media interactions suggest that non-local media does not support
45Given the extensive reach of AM radio stations and our restriction to urban areas, theAM sample cannot provide informative estimates and is thus omitted. The available samplesize drops to 1,038.
30
either rewarding or sanctioning the incumbent party. This suggests that local media is the
key driver of electoral accountability, and thus any media coverage is not sufficient to punish
the parties of malfeasant mayors. However, because our design does not isolate exogenous
variation in non-local media, this finding does not prove that non-local media serves to
accountability function.
[Table 6 about here.]
To better test whether non-local media also facilitates electoral sanctioning, we again
implement the same identification strategy to isolate plausibly exogenous variation in non-
local media coverage. The results are presented in Table 6, and due to the larger sample size
offer a more powerful test than for non-local media. However, the triple interaction between
revealing an audit before the election, the audit’s outcome and non-local media provides no
evidence that the number of media stations broadcasting from outside the municipality affect
the incumbent party’s vote share. Comparing these estimates to those for local media in
Table 4, the effect of such non-local media is indistinguishable from zero and always smaller
than for local media.46 This evidence reinforces our claim that the presence of local media
is essential for electoral accountability.
6.2.3 Robustness checks
We now demonstrate that our main findings for local media are robust to a variety of
robustness tests. Table 7 presents the results of these specification and variable definition
checks, focusing on the triple interactions identifying the effects of local media revealing
mayoral malfeasance.
First, although the number of local media is very well balanced across pre-treatment vari-
ables, we nevertheless ensure that our results are not driven by the most plausible potential
confounds. Columns (3) and (4) of Table 4 showed that our results are robust to including
an interaction with the number of non-local media. As a more general test, columns (1)
and (2) of panel A in Table 7 simultaneously include the interaction of audit results with
second-order polynomials for four important background indicators: non-local media, dis-
tance to the municipal head (from the precinct centroid), average years of education, and
the proportion of households with a car. For both corruption and neglectful incumbents, if
anything we find larger and precisely estimated effects.
46We also find the same results when using the neighbors sample used to estimate theeffect of local media.
31
[Table 7 about here.]
A second potential concern is that our results are driven by restricting attention to
relatively small neighboring precincts. Accordingly, columns (3)-(6) in panel A of Table 7
show similar, if not larger, estimates when our 10 km2 restriction is relaxed to 25km2 and
50km2.47
Third, to address the possibility that media and election dynamics in municipalities
receiving audit reports after the election are different, we drop the post-election comparison.
We thus focus only on precincts in municipalities where audit reports were released before
an election, and identify entirely off variation in local media before the election. The results
in columns (7) and (8) of panel A show that we obtain similar results. Given the noise
introduced by the DD design, this approach increases the precision of our estimates despite
dropping 40% of the sample.
Fourth, we consider alternative operationalizations of malfeasance. Columns (1) and
(2) of panel B in Table 7 first report linear measures of corruption and not spending on the
poor. In both cases the effects are negative and fairly large in magnitude: for each additional
10% of the FISM budget not spent correctly, a standard deviation increase in local media
stations reduces the incumbent vote share by 2.2 percentage points in the case of corruption
and 3.9 percentages points in the case of not spending on the poor. However, the estimate for
corruption is not statistically significant. Together with our previous results, this suggests
that corruption’s effects are non-linear. To further examine when the impact of malfeasance
kicks in, columns (3)-(8) sequentially operationalize malfeasant behavior using an indicator
for corruption or funds not spent on the poor exceeding 5, 10 and 20% of FISM funds. At
each level, not spending on the poor is punished, although the magnitude of punishment is
greater for the 20% cutoff. For corruption, the evidence suggests that only the highest levels
of corruption—greater than 20%—are severely punished.
Fifth, at the cost of losing randomization in local media, we estimate equation (1) in the
larger DD sample to check the external validity of the neighbor sample estimates. The results,
provided in the Online Appendix, are broadly similar to our neighbor sample estimates. For
both corruption and spending not on the poor, a mayor revealed to be in the most malfeasant
quartile experiences a significant decrease in their vote share for each additional local media
station. However, malfeasance in the third quartile is not punished. Although the samples
differ in terms of both composition and quality of identification, the similarity of the results
is encouraging.
47Similar results are obtained when using all possible neighbors.
32
6.3 Media market structure
Our main results show that only local media stations facilitate electoral accountability in
terms of informing voters about the performance of their incumbent party. To better un-
derstand when media can support this important social function, we further investigate the
media market structure. In particular, we focus on two important features: crowd-out by
non-local media and the geographic composition of a station’s potential audience.
6.3.1 Non-local media crowd out local media
Since non-local media are less likely to cover the relevant political actors, and the preced-
ing analysis demonstrated that non-local media stations do not affect incumbent electoral
performance, an additional non-local media station could crowd-out the effects of local me-
dia. This possibility rests upon the likelihood that a new media station attracts listeners or
viewers away from local media stations, plausibly because voters consume media for reasons
other than acquiring politically-relevant information and are unable to increase consumption
of all stations commensurately.
[Table 8 about here.]
To examine this possibility empirically, we interact the number of local media stations
with the number of non-local stations in Table 8. The results are consistent with crowd-
out: for both and low performance incumbent mayors, non-local media stations weakens the
impact of a local media station. Particularly for not spending on the poor, the significant
quadruple interactions show that each additional non-local media station reduces the effect
of an additional local media station by 0.1-0.16 percentage points. In the average precinct,
covered by nearly 15 non-local media stations, the impact of an additional local media
station thus falls by around 2 percentage points. These positive interactions are are not
statistically significant in the case of corruption. This evidence thus implies that the presence
of uninformative non-local media stations drowns out the electoral accountability supported
by local media, specially when it comes to investments not benefiting the poor. Implicitly,
the fact that an additional non-local media station reduces the effect of local media suggests
that voters indeed use local media to follow the news—otherwise, an additional media station
should not affect voter behavior.48
48This could arise because an additional non-local media station causes pre-existing firmsto alter their content, or because voters shift toward the new non-local station. Withoutdetailed media consumption data, we cannot differentiate these explanations.
33
6.3.2 Media markets determine news content
To further explore the mechanism driving local media’s effect, we consider the composition
of the media market. As Snyder and Stromberg (2010) show, the congruence of political
boundaries and media markets plays a key role in the content that broadcasters choose to
provide. In our context, local media stations predominantly cover consumers within their
municipality have the strongest incentives to cover audit reports relating to their municipal-
ity. However, despite emitting from a given municipality, some media stations may primarily
serve audiences in other municipalities, and adjust their news coverage accordingly.
To test this argument, we employ a similar approach to the congruence measure in-
troduced by Snyder and Stromberg (2010). Specifically, for each local media station, we
calculate the proportion of the population that receives a commercial quality signal that
reside inside the municipal of the emitter. We then computed the average such share across
all AM radio, FM radio and television stations covering a given electoral precinct. In our
sample, 63% of the average local media station’s audience is from within the municipality.
Higher values imply that a given media station has a strong incentive to cover audit report
outcomes in depth. As with our crowd-out analysis, we interact this measure of the local
market with local media to examine how the accountability-enhancing effects of local media
depend upon the media market.
[Table 9 about here.]
The results in Table 9 suggest that a media’s station’s market has important effects on
electoral accountability. Consistent with news coverage depending upon the extent to which
a station’s market is local, we find that a larger local market share increases the reward and
punishment of incumbent parties. This is particularly true in the case of not spending on the
poor, where the large and statistically significant negative quadruple interaction implies that
each additional media station with an entirely local market share would reduce the incumbent
party’s vote share by 2.9 percentage points for mayors in the third quartile and 3.7 percentage
points for mayors in the fourth quartile. Similarly, mayors that correctly spend FISM funds
on projects benefiting the poor gain 2.7 percentage points for an additional media station
exclusively serving their market. Although the analogous coefficients for corruption are not
statistically significant, they paint a similar picture where a mayor from the third quartile
is punished by 1.9 percentage points and a mayor from the fourth quartile is punished by
1 percentage point. The party of a non-corrupt mayor receives a significant boost of 1.7
34
percentage points for each additional local media station serving only the local municipal
market.
7 Conclusion
Many scholars call media “the fourth estate,” due to its potential to inform voters about the
behavior of politicians in office. Both national and local media are needed: while national
media outlets cover national level actors, local media are necessary to inform voters about
the performance of local politicians. While the influence of national media has received con-
siderable attention, this article demonstrates the importance of local media for local electoral
accountability in a federal setting where local governments play an increasingly important
role in service delivery. Furthermore, we show that effective electoral accountability requires
more than just a single local media outlet: we find that additional local media stations have
large effects, even in relatively congested media markets.
Using detailed local data, and an identification strategy that exploits both the timing
of the release of audit reports with respect to elections and differences in signal coverage
across neighboring electoral precincts, we identify the impact of the media environment on
electoral accountability. We show that voters punish the party of malfeasant mayors, but
only in electoral precincts covered by local radio or television stations. In particular, we find
that each additional local media station reduces the vote share of an incumbent political
party revealed to be corrupt by nearly 0.7 percentage points. Local media similarly reduces
the vote share of an incumbent political party revealed to have diverted funds away from
the poor by around 1.2 percentage points. However, we find no effect of non-local media
stations that are based in other municipalities. Delving further into the structure of media
markets, we show that non-local media crowd out the effects of local media—plausibly by
attracting away consumers of local news—and that even local media may not as effectively
support electoral accountability when their audience contains many consumers from outside
the municipality. Our findings thus demonstrate the importance of media, especially local
media, in supporting local electoral accountability by sanctioning malfeasant behavior.
35
References
Alt, James E., David D. Lassen and John Marshall. 2014. “Information sources, belief
updating, and the politics of economic expectations: Evidence from a Danish survey
experiment.” Working paper.
Anderson, Simon P. and John McLaren. 2012. “Media mergers and media bias with rational
consumers.” Journal of the European Economic Association 10(4):831–859.
Ansolabehere, Stephen, Erik C. Snowberg and James M. Snyder. 2006. “Television and the
incumbency advantage in US elections.” Legislative Studies Quarterly 31(4):469–490.
Banerjee, Abhijit V., Selvan Kumar, Rohini Pande and Felix Su. 2011. “Do Informed Voters
Make Better Choices? Experimental Evidence from Urban India.” Working paper.
Baron, David P. 2006. “Persistent media bias.” Journal of Public Economics 90(1):1–36.
Barro, Robert J. 1973. “The control of politicians: an economic model.” Public choice
14(1):19–42.
Besley, Timothy and Andrea Prat. 2006. “Handcuffs for the grabbing hand? Media capture
and government accountability.” American Economic Review 96(3):720–736.
Boas, Taylor C. and F. Daniel Hidalgo. 2011. “Controlling the airwaves: Incumbency advan-
tage and community radio in Brazil.” American Journal of Political Science 55(4):869–885.
Bobonis, Gustavo J., Luis R. Camara Fuertes and Rainer Schwabe. 2013. “Monitoring
Corruptible Politicians.” Working paper.
Cage, Julia. 2014. “Media competition, information provision and political participation.”
Working Paper.
Castaneda Sabido, Fernando. 2011. “Investigacion sobre la percepcion ciudadana acerca de
la transparencia, rendicion de cuentas y fiscalizacin en el uso de los recursos publicos en
Mexico.” Reporte para la Auditorıa Superior de la Federacion .
Chang, Eric C.C., Miriam A. Golden and Seth J. Hill. 2010. “Legislative malfeasance and
political accountability.” World Politics 62(2):177–220.
36
Chattopadhyay, Raghabendra and Esther Duflo. 2004. “Women as Policy Makers: Evidence
from a Randomized Policy Experiment in India.” Econometrica (72):1409–1443.
Chong, Alberto, Ana De La O, Dean Karlan and Leonard Wantchekon. forthcoming. “Does
Corruption Information Inspire the Fight or Quash the Hope? A Field Experiment in
Mexico on Voter Turnout, Choice and Party Identification.” Journal of Politics .
Costas, Elena, Albert Sole-Olle and Pilar Sorribas-Navarro. 2011. “Do voters really tolerate
corruption? Evidence from Spanish Mayors.” Working paper.
DellaVigna, Stefano and Ethan Kaplan. 2007. “The Fox News effect: Media bias and voting.”
Quarterly Journal of Economics 122(3):1187–1234.
DellaVigna, Stefano and Matthew Gentzkow. 2010. “Persuasion: Empirical Evidence.” An-
nual Review of Economics 2(1):643–669.
Djankov, Simeon, Caralee McLiesh, Tatiana Nenova and Andrei Shleifer. 2003. “Who Owns
the Media?” Journal of Law and Economics 46(2):341–382.
Durante, Ruben and Brian Knight. 2012. “Partisan control, media bias, and viewer re-
sponses: Evidence from Berlusconi’s Italy.” Journal of the European Economic Association
10(3):451–481.
Eggers, Andrew and Alexander C. Fisher. 2011. “Electoral Accountability and the UK
Parliamentary Expenses Scandal: Did Voters Punish Corrupt MPs?” SSRN working
paper.
Enikolopov, Ruben, Maria Petrova and Ekaterina Zhuravskaya. 2011. “Media and political
persuasion: Evidence from Russia.” American Economic Review 101(7):3253–3285.
Ferejohn, John. 1986. “Incumbent performance and electoral control.” Public choice 50(1):5–
25.
Fergusson, Leopoldo. 2014. “Media markets, special interests, and voters.” Journal of Public
Economics 109(C):13–26.
Ferraz, Claudio and Frederico Finan. 2008. “Exposing Corrupt Politicians: The Effects of
Brazil’s Publicly Released Audits on Electoral Outcomes.” Quarterly Journal of Economics
123(2):703–745.
37
Gentzkow, Matthew and Jesse M. Shapiro. 2006. “Media bias and reputation.” Journal of
Political Economy 114(2):280–316.
Gentzkow, Matthew and Jesse M. Shapiro. 2010. “What drives media slant? Evidence from
US daily newspapers.” Econometrica 78(1):35–71.
Gentzkow, Matthew, Jesse M. Shapiro and Michael Sinkinson. 2014. “Competition and ide-
ological diversity: Historical evidence from U.S. newspapers.” American Economic Review
104(10):3073–3114.
Greene, Kenneth F. 2011. “Campaign Persuasion and Nascent Partisanship in Mexico’s New
Democracy.” American Journal of Political Science 55(2):398–416.
Humphreys, Macartan and Jeremy Weinstein. 2012. “Policing Politicians: Citizen Empow-
erment and Political Accountability in Uganda Preliminary Analysis.” Working paper.
Klasnja, Marko. 2011. “Why Do Malfeasant Politicians Maintain Political Support? Testing
the “Uninformed Voter” Argument.”.
Langston, Joy. 2003. “Rising from the ashes? Reorganizing and unifying the PRI’s state
party organizations after electoral defeat.” Comparative Political Studies 36(3):293–318.
Larreguy, Horacio A. and Joana C. M. Monteiro. 2014. “Media Networks and Political
Accountability: Evidence from Radio Networks in Brazil.” Working paper.
Larreguy, Horacio, John Marshall and James M. Snyder Jr. 2014. “Leveling the playing
field: How equalizing access to political advertising helps locally non-dominant parties in
consolidating democracies.” Working paper.
Larreguy, Horacio, John Marshall and Pablo Querubın. forthcoming. “Parties, Brokers and
Voter Mobilization: How Turnout Buying Depends Upon the Party’s Capacity to Monitor
Brokers.” American Political Science Review .
Lavielle, Briseida, Kristina Pirker and Alberto Serdan. 2006. “Monitoreo de Programas
Sociales en Contextos Electorales.” Fundar Working paper.
Lawson, Chappell and James A. McCann. 2005. “Television News, Mexico’s 2000 Elections
and Media Effects in Emerging Democracies.” British Journal of Political Science pp. 1–30.
38
Mauro, Paolo. 1995. “Corruption and Growth.” Quarterly Journal of Economics 110(3):681–
712.
McCann, James A. and Chappell H. Lawson. 2003. “An Electorate Adrift?: Public Opinion
and the Quality of Democracy in Mexico.” Latin American Research Review 38(3):60–81.
Mullainathan, Sendhil and Andrei Shleifer. 2005. “The market for news.” American Eco-
nomic Review pp. 1031–1053.
Niehaus, Paul and Sandip Sukhtankar. 2013. “Corruption Dynamics: The Golden Goose
Effect.” American Economic Journal: Economic Policy 5(4):230–269.
Olken, Benjamin A. 2007. “Monitoring Corruption: Evidence from a Field Experiment in
Indonesia.” Journal of Political Economy 115(2).
Pande, Rohini. 2011. “Can informed voters enforce better governance? Experiments in
low-income democracies.” Annual Review of Economics 3(1):215–237.
Prat, Andrea and David Stromberg. 2005. “Commercial Television and Voter Information.”.
CEPR Discussion Paper No. 4989.
Puglisi, Riccardo and James M. Snyder. 2011. “Newspaper coverage of political scandals.”
Journal of Politics 73(3):931–950.
Serra, Gilles. 2012. “The Risk of Partyarchy and Democratic Backsliding Mexico.” Taiwan
Journal of Democracy 8(1):93–118.
Snyder, James M. and David Stromberg. 2010. “Press Coverage and Political Accountabil-
ity.” Journal of Political Economy 118(2):355–408.
Stanig, Piero. forthcoming. “Regulation of Speech and Media Coverage of Corruption: An
Empirical Analysis of the Mexican Press.” American Journal of Political Science .
Stromberg, David. 2004. “Radio’s impact on public spending.” Quarterly Journal of Eco-
nomics 119(1):189–221.
Stromberg, David. 2015. “Media and Politics.” Working paper.
Weitz-Shapiro, Rebecca and Matthew S. Winters. 2014. “Discerning corruption: Credible
Accusations and the Punishment of Politicians in Brazil.” Working paper.
39
Wellenstein, Anna, Angelica Nunez and Luis Andres. 2006. Social Infrastructure: Fondo de
Aportaciones para la Infraestructura Social (FAIS). In Decentralized service delivery for
the poor, Volume II: Background papers, ed. The World Bank. Mexico City: The World
Bank pp. 167–222.
Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge University Press.
Zamboni, Yves and Stephan Litschig. 2014. “Audit Risk and Rent Extraction: Evidence
from a Randomized Evaluation in Brazil.” Working paper.
40
-2-1
01
23
Sta
ndar
dize
d se
arch
act
ivity
(si
gma)
Feb 06 Feb 07 Feb 08 Feb 09 Feb 10 Feb 11 Feb 12
(a) Searches for ASF
-10
12
3S
tand
ardi
zed
sear
ch a
ctiv
ity (
sigm
a)
Feb 06Feb 06 Feb 07Feb 06 Feb 07 Feb 08Feb 06 Feb 07 Feb 08 Feb 09Feb 06 Feb 07 Feb 08 Feb 09 Feb 10Feb 06 Feb 07 Feb 08 Feb 09 Feb 10 Feb 11Feb 06 Feb 07 Feb 08 Feb 09 Feb 10 Feb 11 Feb 12
(b) Searches for FISM
Figure 1: Google searches related to audit reports by month, 2006-2012
Notes: Extracted using Google Correlate (http://correlate.googlelabs.com) on 15th July 2014.
The data cover Google searches in Mexico for the period used in our sample.
41
Figure 2: Distribution of audit report outcomes by municipality.
Notes: Only the 268 municipalities in our final sample are included. Where more than one
audit occurs, we take the average audit outcome.
42
Figure 3: AM radio signal coverage areas (source: IFE).
Figure 4: FM radio signal coverage (source: IFE).
43
Figure 5: TV signal coverage (source: IFE).
44
05
1015
2025
Den
sity
0 .2 .4 .6 .8 1
Proportion of spending unauthorized
After election Before election
(a) Unauthorized spending
05
1015
20
Den
sity
0 .2 .4 .6 .8 1
Proportion of spending not on the poor
After election Before election
(b) Spending not on the poor
Figure 6: Distribution of audit report results, by release around election
Notes: The audit outcomes kernel density distributions are based on electoral precincts in
municipalities that received an audit just before and just after an election. As with the main
analysis, precincts are weighted by the number of registered voters. We use an Epanechnikov
kernel with a bandwidth of 0.025.
45
Figure 7: Identification strategy example
Note: Both precincts are from the municipality of Villa de Tututepec de Melchor Ocampo in
the state of Oaxaca. While precinct 1583 is covered by the television emitting from within the
municipality, 1571 is not.
46
Table 1: Summary statistics by audit status (DD sample)
Control (no audit) mean Audit difference
Unauthorized spending 0.106 -0.028 (0.030)Corrupt Q3 0.186 0.062 (0.065)Corrupt Q4 0.322 -0.110 (0.080)Spending not on the poor 0.097 0.028 (0.033)Not poor Q3 0.229 0.026 (0.075)Not poor Q4 0.263 -0.012 (0.079)PAN incumbent 0.354 0.022 (0.085)PRI incumbent 0.464 0.073 (0.086)PRD incumbent 0.171 -0.083 (0.058)Coalition partners 1.751 0.134 (0.228)Municipal incumbent victory margin (lag) 0.151 0.010 (0.019)Municipal effective number of political parties (lag) 2.615 -0.038 (0.077)Incumbent party vote share (lag) 0.484 -0.003 (0.012)Effective number of political parties (lag) 2.485 -0.031 (0.063)Registered voters (log) 7.450 -0.026 (0.060)Turnout (lag) 0.472 0.030 (0.018)Distance to municipal head from precinct border (log) 8.160 -0.015 (0.067)Distance to municipal head from precinct centroid (log) 8.460 -0.045 (0.054)Area (log) 1.039 -0.077 (0.115)Population (log) 7.688 0.026 (0.055)Population density (log) 7.932 0.159 (0.210)Local media 10.446 3.743* (2.182)Non-local media 27.626 -3.349 (4.566)Average children per woman 2.298 -0.012 (0.034)Share households with male head 0.749 -0.002 (0.005)Share indigenous speakers 0.041 -0.001 (0.008)Average years of schooling 8.947 0.209 (0.154)Share illiterate 0.052 -0.001 (0.005)Share no schooling 0.058 -0.001 (0.004)Share incomplete primary schooling 0.942 0.001 (0.004)Share complete primary schooling 0.828 0.006 (0.010)Share incomplete secondary schooling 0.674 0.011 (0.013)Share complete secondary schooling 0.623 0.011 (0.013)Share higher education 0.385 0.018 (0.015)Share economically active 0.408 0.007 (0.006)Share without health care 0.321 -0.015 (0.014)Share state workers health care 0.052 0.007 (0.005)Average occupants per dwelling 3.925 0.041 (0.038)Average occupants per room 1.068 -0.008 (0.024)Share non-dirt floor 0.944 0.003 (0.005)Share toilet at home 0.961 0.003 (0.005)Share running water 0.898 0.014 (0.016)Share drainage 0.923 0.011 (0.011)Share electricity 0.981 0.001 (0.002)Share washing machine 0.711 0.020 (0.018)Share fridge 0.857 0.015 (0.013)Share cell phone 0.686 0.015 (0.016)Share car 0.468 0.014 (0.023)Share computer 0.325 0.018 (0.019)
Notes: The audit difference results are from regressions of the outcome variables on the left-hand-side of the table on an
indicator for an audit being released the year before an election, where standard errors clustered by municipality election
are in parentheses. There are 47,938 observations for each variable. * denotes p < 0.1, ** denotes p < 0.05, *** denotes
p < 0.01.
47
Tab
le2:
Lin
ear
bal
ance
over
loca
lm
edia
(nei
ghb
orsa
mple
)
Incu
mb
ent
Eff
ecti
veR
egis
tere
dT
urn
out
Dis
tan
ceto
Dis
tan
ceto
Are
aP
opu
lati
onP
opu
lati
onp
arty
vot
enu
mb
erof
vote
rs(l
og)
(lag
)m
un
icip
alm
un
icip
al(l
og)
(log
)d
ensi
tysh
are
(lag)
pol
itic
alh
ead
from
hea
dfr
om(l
og)
par
ties
(lag
)b
ord
er(l
og)
centr
oid
(log
)
Loca
lm
edia
0.00
04
0.0
005
-0.0
040
0.00
030.
0022
-0.0
024
-0.0
077
-0.0
027
0.00
68(0
.0004
)(0
.001
8)(0
.005
6)(0
.000
4)(0
.006
9)(0
.004
0)(0
.006
8)(0
.004
8)(0
.007
1)
Non
-loca
lA
vera
geS
har
eS
har
eA
vera
geS
har
eS
har
en
oS
har
eSh
are
med
iach
ild
ren
hou
seh
old
sin
dig
enou
syea
rsof
illi
tera
tesc
hool
ing
inco
mp
lete
com
ple
tep
erw
om
anw
ith
mal
esp
eaker
ssc
hool
ing
pri
mar
yp
rim
ary
hea
dsc
hool
ing
sch
ool
ing
Loca
lm
edia
-0.1
172
-0.0
021
-0.0
006*
**-0
.000
10.
0344
-0.0
002*
-0.0
002
0.00
020.
0005
(0.2
114
)(0
.002
3)(0
.000
2)(0
.000
2)(0
.027
9)(0
.000
1)(0
.000
1)(0
.000
1)(0
.000
5)
Sh
are
Shar
eS
har
eS
har
eS
har
eS
har
eA
vera
geA
vera
geS
har
ein
com
ple
teco
mp
lete
hig
her
econ
omic
ally
wit
hou
tst
ate
occ
up
ants
occ
up
ants
non
-dir
tse
con
dar
yse
con
dar
yed
uca
tion
acti
veh
ealt
hw
orker
sp
erp
erfloor
sch
ool
ing
sch
ooli
ng
care
hea
lth
care
dw
elli
ng
room
Loca
lm
edia
0.00
08
0.0
009
0.00
120.
0000
-0.0
003
0.00
03-0
.002
8-0
.001
90.
0002
(0.0
009
)(0
.001
)(0
.001
4)(0
.000
2)(0
.000
5)(0
.000
2)(0
.003
2)(0
.001
5)(0
.000
3)
Sh
are
Shar
eS
har
eS
har
eS
har
eS
har
eSh
are
Sh
are
Sh
are
toil
etru
nn
ing
dra
inag
eel
ectr
icit
yw
ash
ing
frid
gece
llca
rco
mp
ute
rat
hom
ew
ater
mac
hin
ep
hon
e
Loca
lm
edia
0.00
00
0.0
007
0.00
050.
0001
0.00
080.
0002
0.00
060.
0006
0.00
16(0
.0003
)(0
.000
6)(0
.000
3)(0
.000
1)(0
.000
6)(0
.000
4)(0
.000
7)(0
.001
1)(0
.001
3)
Notes:
Each
coeffi
cien
tis
from
ase
par
ate
OL
Sre
gres
sion
incl
ud
ing
nei
ghb
orfi
xed
effec
ts,
are
esti
mat
edu
sin
gO
LS
,an
dw
eigh
tby
elec
tora
tesi
zed
ivid
edby
the
nu
mb
erof
mat
ches
per
nei
ghb
orse
t.E
ach
regr
essi
onco
nta
ins
17,3
12ob
serv
atio
ns.
Sta
nd
ard
erro
rsar
ecl
ust
ered
by
mu
nic
ipal
elec
tion
.*
den
otes
p<
0.1,
**d
enot
esp<
0.05
,**
*d
enot
esp<
0.0
1.
48
Table 3: The effects of audits revealing malfeasance before an election (DD sample)
Change in incumbent Incumbent partyparty vote share re-elected(1) (2) (3) (4)
Constant -0.057*** -0.066*** 0.483*** 0.621***(0.015) (0.015) (0.098) (0.090)
Audit 0.005 0.023 0.031 -0.025(0.019) (0.021) (0.122) (0.113)
Corrupt Q3 -0.008 0.068(0.027) (0.159)
Audit × Corrupt Q3 0.022 0.136(0.037) (0.191)
Corrupt Q4 0.014 0.152(0.036) (0.153)
Audit × Corrupt Q4 -0.045 -0.291(0.045) (0.208)
Not poor Q3 0.014 -0.200(0.022) (0.156)
Audit × Not poor Q3 -0.026 0.242(0.035) (0.196)
Not poor Q4 0.034 -0.117(0.041) (0.162)
Audit × Not poor Q4 -0.068 -0.171(0.047) (0.203)
Observations 47,938 47,938 47,938 47,938Outcome mean -0.05 -0.05 0.55 0.55Outcome std. dev. 0.15 0.15 0.50 0.50
Notes: All specifications weight by the number of registered voters, and are estimated using
OLS. Similar estimates for the neighbor sample are provided in the Online Appendix. The
omitted category for corruption and not spending on the poor is Q1 and Q2. Standard errors
are clustered by municipal election. * denotes p < 0.1, ** denotes p < 0.05, *** denotes
p < 0.01.
49
Table 4: Effects of local media and audits reports revealing malfeasance before an election(neighbor sample)
Change in incumbent party vote share(1) (2) (3) (4)
Audit -0.029 0.043 -0.041 0.044(0.054) (0.050) (0.069) (0.080)
Audit × Local media 0.007*** 0.006* 0.008*** 0.005(0.002) (0.003) (0.002) (0.003)
Audit × Corrupt Q3 × Local media -0.007* -0.006**(0.004) (0.003)
Audit × Corrupt Q4 0.104 0.199**(0.074) (0.095)
Audit × Corrupt Q4 × Local media -0.007* -0.008**(0.004) (0.004)
Audit × Not poor Q3 0.123 0.031(0.098) (0.139)
Audit × Not poor Q3 × Local media -0.010* -0.008(0.006) (0.006)
Audit × Not poor Q4 -0.042 0.020(0.076) (0.101)
Audit × Not poor Q4 × Local media -0.011*** -0.012***(0.004) (0.004)
Audit × Non-local media 0.000 0.001(0.003) (0.001)
Audit × Corrupt Q3 × Non-local media 0.002(0.003)
Audit × Corrupt Q4 × Non-local media -0.003(0.004)
Audit × Not poor Q3 × Non-local media 0.002(0.004)
Audit × Not poor Q4 × Non-local media -0.002(0.002)
Observations 17,312 17,312 17,312 17,312Outcome mean -0.04 -0.04 -0.04 -0.04Outcome std. dev. 0.13 0.13 0.13 0.13Local mean 20.60 20.60 20.60 20.60Local std. dev. 10.57 10.57 10.57 10.57
Notes: All specifications include neighbor fixed effects, are estimated using OLS, and weight by electorate
size divided by the number of matches per neighbor set. The omitted category for corruption and not
spending on the poor is Q1 and Q2. Media, corruption and non poor spending lower order terms are
omitted. Standard errors are clustered by municipal election. * denotes p < 0.1, ** denotes p < 0.05,
*** denotes p < 0.01.
50
Table 5: Effects of local FM radio and television stations and audits reports revealing malfea-sance before an election (neighbor sample)
Change in incumbent party vote share(1) (2) (3) (4)
Audit × Local FM media 0.005 -0.003(0.004) (0.004)
Audit × Corrupt Q3 × Local FM media 0.003(0.006)
Audit × Corrupt Q4 × Local FM media -0.009(0.007)
Audit × Not poor Q3 × Local FM media 0.013**(0.007)
Audit × Not poor Q4 × Local FM media -0.001(0.005)
Audit × Local TV media 0.037*** 0.037***(0.005) (0.008)
Audit × Corrupt Q3 × Local TV media -0.031***(0.008)
Audit × Corrupt Q4 × Local TV media -0.047***(0.011)
Audit × Not poor Q3 × Local TV media -0.037**(0.015)
Audit × Not poor Q4 × Local TV media -0.042**(0.018)
Observations 9,194 9,194 9,686 9,686Outcome mean -0.03 -0.03 -0.05 -0.05Outcome std. dev. 0.12 0.12 0.14 0.14FM/television Media mean 11.39 11.39 4.09 4.09FM/television Media std. dev. 5.22 5.22 2.37 2.37
Notes: Both the FM radio and television neighbor samples were computed in the same way
as for local media (see text for details). All specifications include neighbor fixed effects, are
estimated using OLS, and weight by electorate size divided by the number of matches per
neighbor set. The omitted category for corruption and not spending on the poor is Q1 and Q2.
Media, corruption and non poor spending lower order terms are omitted. Standard errors are
clustered by municipal election. * denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
51
Table 6: Effects of non-local media and audits reports revealing malfeasance before anelection (non-local media neighbor sample)
Change in incumbent party vote share(1) (2)
Audit -0.111 -0.011(0.070) (0.051)
Audit × Non-local media 0.004* 0.000(0.002) (0.001)
Audit × Corrupt Q3 0.083(0.086)
Audit × Corrupt Q3 × Non-local media -0.003(0.003)
Audit × Corrupt Q4 -0.010(0.103)
Audit × Corrupt Q4 × Non-local media -0.003(0.003)
Audit × Not poor Q3 0.021(0.084)
Audit × Not poor Q3 × Non-local media 0.003(0.002)
Audit × Not poor Q4 -0.153(0.111)
Audit × Not poor Q4 × Non-local media 0.003(0.002)
Observations 40,007 40,007Outcome mean -0.06 -0.06Outcome std. dev. 0.14 0.14Non-local media mean 32.97 32.97Non-local media std. dev. 22.73 22.73
Notes: This neighbor sample was computed in the same way as for local media (see text for
details). All specifications include neighbor fixed effects, are estimated using OLS, and weight
by electorate size divided by the number of matches per neighbor set. The omitted category
for corruption and not spending on the poor is Q1 and Q2. Media, corruption and non poor
spending lower order terms are omitted. Standard errors are clustered by municipal election. *
denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
52
Tab
le7:
Rob
ust
nes
sch
ecks
(nei
ghb
orsa
mple
)
Pan
el
A:
Sp
ecifi
cati
on
test
sC
han
ge
inin
cum
ben
tp
art
yvo
tesh
are
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Au
dit×
Cor
rup
tQ
3×
Loca
lm
edia
-0.0
06*
-0.0
09**
-0.0
09**
-0.0
04***
(0.0
03)
(0.0
04)
(0.0
04)
(0.0
01)
Au
dit×
Cor
rup
tQ
4×
Loca
lm
edia
-0.0
17***
-0.0
07
-0.0
08**
-0.0
08***
(0.0
03)
(0.0
04)
(0.0
04)
(0.0
01)
Au
dit×
Not
poor
Q3×
Loca
lm
edia
-0.0
11*
-0.0
10**
-0.0
10**
-0.0
08**
(0.0
06)
(0.0
05)
(0.0
04)
(0.0
03)
Au
dit×
Not
poor
Q4×
Loca
lm
edia
-0.0
13***
-0.0
15***
-0.0
17***
-0.0
08**
(0.0
04)
(0.0
05)
(0.0
05)
(0.0
03)
Are
are
stri
ctio
n<
10km
2<
10km
2<
25km
2<
25km
2<
50km
2<
50km
2<
10km
2<
10km
2
Qu
adra
tic
inte
ract
ive
contr
ols
XX
Pre
-ele
ctio
nau
dit
son
lyX
XO
bse
rvat
ion
s17,3
12
17,3
12
22,4
89
22,4
89
28,1
47
28,1
47
11,2
85
11,2
85
Pan
el
B:
Malf
easa
nce
defi
nit
ion
sC
han
ge
inin
cum
ben
tp
art
yvo
tesh
are
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Au
dit×
Loca
lm
edia×
Cor
rup
tion
-0.0
22
(0.0
23)
Au
dit×
Loca
lm
edia×
Not
poor
-0.0
39**
(0.0
17)
Au
dit×
Loca
lm
edia×
Cor
rup
t≥
5%0.0
01
(0.0
06)
Au
dit×
Loca
lm
edia×
Not
poor≥
5%-0
.009**
(0.0
04)
Au
dit×
Loca
lm
edia×
Cor
rup
t≥
10%
-0.0
06*
(0.0
03)
Au
dit×
Loca
lm
edia×
Not
poor≥
10%
-0.0
07**
(0.0
03)
Au
dit×
Loca
lm
edia×
Cor
rup
t≥
20%
-0.0
28**
(0.0
13)
Au
dit×
Loca
lm
edia×
Not
poor≥
20%
-0.0
16***
(0.0
04)
Ob
serv
atio
ns
17,3
12
17,3
12
17,3
12
17,3
12
17,3
12
17,3
12
17,3
12
17,3
12
Notes:
All
spec
ifica
tion
sin
clu
de
nei
ghb
orfi
xed
effec
ts,
are
esti
mate
du
sin
gO
LS
,an
dw
eight
by
elec
tora
tesi
zed
ivid
edby
the
nu
mb
erof
mat
ches
per
nei
ghb
orse
t.A
llot
her
inte
ract
ion
sare
om
itte
d.
InP
an
elA
,th
eco
ntr
ols
are
the
nu
mb
erof
non
-loca
lm
edia
stati
on
s,(l
og)
dis
tan
ceto
the
mu
nic
ipal
hea
dfr
omth
ep
reci
nct
’sce
ntr
oid
,av
erage
years
of
sch
ooli
ng,
an
dth
esh
are
that
own
aca
r.S
tan
dard
erro
rsare
clu
ster
edby
mu
nic
ipal
elec
tion
.*
den
otes
p<
0.1,
**
den
ote
sp<
0.05,
***
den
ote
sp<
0.01.
53
Table 8: Effects of local media and audits reports revealing malfeasance before an election,conditional on the number of non-local media stations (neighbor sample)
Change in incumbent party vote share(1) (2)
Audit × Local media 0.014*** 0.020***(0.005) (0.005)
Audit × Non-local media × Local media -0.0003 -0.0005***(0.0002) (0.0001)
Audit × Corrupt Q3 × Local media -0.011(0.008)
Audit × Corrupt Q3 × Non-local media × Local media 0.0002(0.0002)
Audit × Corrupt Q4 × Local media -0.015*(0.009)
Audit × Corrupt Q4 × Non-local media × Local media 0.0004(0.0003)
Audit × Not poor Q3 × Local media -0.031***(0.008)
Audit × Not poor Q3 × Non-local media × Local media 0.0016***(0.0004)
Audit × Not poor Q4 × Local media -0.039***(0.008)
Audit × Not poor Q4 × Non-local media × Local media 0.0010***(0.0002)
Observations 17,312 17,312Non-local media mean 14.85 14.85Non-local media std. dev. 16.33 16.33
Notes: All specifications include neighbor fixed effects, are estimated using OLS, and weight
by electorate size divided by the number of matches per neighbor set. The omitted category
for corruption and not spending on the poor is Q1 and Q2. Media, corruption and non poor
spending lower order terms are omitted. Standard errors are clustered by municipal election. *
denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
54
Table 9: Effects of local media and audits reports revealing malfeasance before an election,conditional on local market media coverage share (neighbor sample)
Change in incumbent party vote share(1) (2)
Audit × Local media 0.002 -0.006**(0.005) (0.003)
Audit × Corrupt Q3 × Local media -0.005(0.006)
Audit × Local media × Local market 0.017** 0.027***(0.008) (0.009)
Audit × Corrupt Q3 × Local media × Local market -0.014(0.015)
Audit × Corrupt Q4 × Local media -0.000(0.006)
Audit × Corrupt Q4 × Local media × Local market -0.010(0.013)
Audit × Not poor Q3 × Local media 0.007(0.007)
Audit × Not poor Q3 × Local media × Local market -0.029**(0.012)
Audit × Not poor Q4 × Local media 0.001(0.004)
Audit × Not poor Q4 × Local media × Local market -0.037***(0.013)
Observations 17,312 17,312Local market mean 0.63 0.63Local market std. dev. 0.28 0.28
Notes: All specifications include neighbor fixed effects, are estimated using OLS, and weight
by electorate size divided by the number of matches per neighbor set. The omitted category
for corruption and not spending on the poor is Q1 and Q2. Media, corruption and non poor
spending lower order terms are omitted. Standard errors are clustered by municipal election. *
denotes p < 0.1, ** denotes p < 0.05, *** denotes p < 0.01.
55