.
contributo inviato da fabio1963 il 22 maggio 2012

Quello che segue è un articolo pubblicato sulla rivista internazionale SSRN:

 
dai 2 italiani: Lorenzo Casaburi & Ugo Troiano.
 
VEDERE IMPORTANTISSIME IMMAGINI E TABELLE IN FONDO ALL'ARTICOLO

Abstract (Traduzione di Fabio Marinelli):
 
Qual è la risposta elettorale per le politiche anti-evasione fiscale? 
Studiamo i risultati politici amministravi locali ottenuti da un programma nazionale in Italia. 
 
Nel 2007, un grande intervento contro l'evasione fiscale (Programma Edifici Fantasma) ha individuato immobili non correttamente elencati nel catasto, e quindi esclusi dalla base imponibile, mediante un confronto di mappe catastali con le immagini aeree di tutto il territorio italiano. 
In primo luogo, noi ripercorriamo il corso degli eventi per valutare a livello nazionale il livello di evasione delle tasse e gli studi ad essa correlati, utilizzando i dati amministrativi raccolti dall'Agenzia Italiana del Territorio.
In secondo luogo, stimiamo l'impatto del programma sulla probabilità di rielezione del sindaco. Usiamo dati incrociati eterogenei delle città in base alla densità degli edifici fantasma, e quindi del potenziale per il programma di aumentare la riscossione fiscale attuando una differenza di approccio. 
Stimiamo che ci sia una correlazione tra  il numero degli edifici fantasma trovati, destinatari del programma, e la probabilità di rielezione della carica di sindaco, che aumenta di circa 3,5 punti (se le promesse elettorali vengono mantenute dopo le elezioni).
In terzo luogo, troviamo che l'effetto è guidato dalle zone più sviluppate. Infine, utilizzando i dati analizzati, forniamo prove che suggeriscono che è proprio l'applicazione della tassa addizionale indotta dal programma che guida i risultati.
--------------------------------------- 1
 
Ghost-House Busters: The Electoral Response
 
to a Large Anti-Tax Evasion Program
Lorenzo Casaburi Ugo Troiano
Harvard University Harvard University
First version: October 2011
This Version: March 2012
Abstract
What is the electoral response to anti-tax evasion policies? We study the political
returns local administrators obtain from a nationwide intervention in Italy. Starting in
2007, the Ghost Buildings program identied buildings not properly listed in the land
registry, and thus illegally excluded from the tax base, by juxtaposing land registry
maps with aerial pictures of the whole country. A substantial registration wave then
targeted the detected buildings. First, we provide a novel nationwide measure of tax
evasion and study its correlates by using administrative data from the intervention.
Second, we estimate the impact of the program on incumbent mayor reelection likeli-
hood. We use cross-town heterogeneity in the intensity of ghost buildings, and thus in
the potential for the program to increase tax enforcement, to implement a di_erence-in-
di_erences approach. We nd that a one standard deviation increase in the intensity of
ghost buildings targeted by the program raises the likelihood of reelection of the incum-
bent by about 3.5 percentage points in post-program elections, relative to pre-program
ones. Third, we nd that the e_ect is driven by more developed areas. Finally, using
registration data, we provide evidence suggesting that it is indeed the additional tax
enforcement induced by the program that drives the results.
 
Lorenzo Casaburi, casaburi@fas.harvard.edu, Ugo Troiano, troiano@fas.harvard.edu .We are indebted to the Agenzia del
Territorio that provided the administrative data used in this paper. We wish to thank Alberto Alesina, Joseph Altonji, Josh
Angrist, Robert Barro, Raj Chetty, David Cutler, Ed Glaeser, Larry Katz, Asim Kwhaja, Michael Kremer, Massimo Maoret,
Sendhil Mullainathan,Manisha Padi, Rohini Pande, Dina Pomeranz, Tristan Reed, Giovanni Reina, L_aszl_o S_andor, Andrei
Shleifer, Monica Singhal, David Yanagizawa-Drott and seminar audiences at the Harvard Development, Labor, Macro, and
Political Economy lunch seminars for their comments. All errors are our own. The views expressed here re
ect the views of
the authors alone, and do not necessarily re
ect the views of any of the organizations that provided support for this work. In
particular, the views expressed here do not necessarily re
ect those of the Agenzia del Territorio.
Electronic copy available at: http://ssrn.com/abstract=1950473
 
 
 
--------------------------------------- 2
 
1 Introduction
How do voters react to anti-tax evasion policies? The electoral response to such policies
shapes the choice of the optimal level of tax enforcement for an o_ce-motivated politician.
Fighting tax evasion can a_ect the incumbent's probability of reelection both positively
and negatively. On the one hand, increasing enforcement hurts the targeted tax evaders.
On the other hand, voters can obtain monetary benets from increased tax revenues and
non-monetary benets from increased e_ciency in tax collection. In this paper, we study
the impact of a nationwide program implemented in Italy to provide, to the best of our
knowledge, the rst evidence on the electoral impact of anti-tax evasion policies.
The importance of tax evasion both in industrialized and in developing economies is well
known. In Italy, the country targeted in this research, it is estimated that the evasion rate
on personal income tax is 13.5% (Marino and Zizza, 2008). Estimates from other developed
countries deliver similar gures (Slemrod, 2007, for the United States). In developing coun-
tries, where the share of the informal economy is typically larger, the gures are much higher
(Gordon and Li, 2009). Thus, it is little surprise that voters often consider increasing tax
compliance one of the priorities on the political agenda. For instance, a recent survey ad-
ministered by Bank of Italy nds that around 76 percent of the sample considers tax evasion
among the rst priorities for Italian public policy (Cannari and D'Alessio, 2006). Yet, the
ght against tax evasion is costly from an economic standpoint, as it typically employs costly
auditing, and, potentially, from a political standpoint, as it hurts a portion of the voters.
This paper uses novel administrative data collected by the Italian Agenzia del Territorio
to provide evidence on the political returns to the ght to tax evasion. In 2007, the Italian
government started a program to identify unregistered buildings by comparing o_cial land
1
registry maps with high resolution aerial pictures of Italian municipalities.This four-year
endeavor delivered, for each of the land registry parcels, a measure of Ghost Buildings, build-
ings identied by the aerial photographs but that were not registered in the building registry
maps. The intervention detected more than two million ghost buildings. Non-registration is
a form of tax evasion since the values of the registered buildings enter the tax base for per-
sonal income tax, property tax, and other taxes. In our rst contribution, we use data from
1
The program excluded the autonomous region of Trentino Alto-Adige
1
Electronic copy available at: http://ssrn.com/abstract=1950473
 
 
 
--------------------------------------- 3
 
the program to provide a new disaggregated measure of tax evasion. The auditing program
undertaken by the Agenzia del Territorio provides a nationwide administrative measure of
tax evasion. The availability of such a measure is a rare opportunity for researchers, who
typically have to rely on the comparison between aggregated macro indicators (most often
income and expenditures) or on discrepancies between income levels reported to the tax
2
authority and the ones reported in anonymous surveys.For the purpose of our analysis,
we dene the \ghost buildings Intensity" as the ratio between the number of parcels with
unregistered buildings and the total number of parcels in each town. We combine this data
with additional information at the municipal level in order to identify the determinants of tax
evasion. We nd that geographical features of Italian municipalities are strongly correlated
with tax evasion, consistently with the results presented in Saiz (2010).
Following the completion of the aerial photographing exercise, a large registration program
targeted the identied ghost buildings. The registration occurred in two phases. First, until
April 30th 2011, owners of ghost buildings could proceed with the registration in the land
registry. Second, after the April deadline, the Agenzia del Territorio proceeded to assign
imputed values to the buildings not yet registered. The program led to a substantial wave of
registrations. According to the administrative data, around 40 percent of the ghost buildings
were registered as of 30th of December 2011, leading to an increase of approximately 2.4
percent in the tax base deriving from building ownership (Rendita Catastale'). Available
data show high local variance in the registration rate, suggesting that town-specic factors
play an important level. Ghost buildings intensity is a strong predictor of the intensity of
parcels with registered ghost buildings (relative to the total number of land registry parcels).
Yet, the ghost buildings registration rate -i.e. the percentage of ghost buildings that get
registered - is decreasing in the initial ghost buildings intensity.
While the central government began the program, local administrations played an impor-
2
For instance, Fisman and Wei (2004) quantify the e_ects of tax rates on tax evasion by examining the
relationship in China between the tari_ schedule and the \evasion gap" which they dene as the di_erence
between Hong Kong's reported exports to China at the product level and China's reported imports from Hong
Kong. Marino and Zizza (2008) estimate that Italian income tax evasion by comparing reported incomes
to the tax authority and to an anonymous survey administered by Bank of Italy. Saez (2010) observes that
\some sources of income are more responsive to taxation than others, or more easier to manipulate, or more
prone to tax avoidance or evasion", and that this could be consistent that in specic cases reported incomes
bunch at kinks, particularly for taxpayers whose income is not subject to third-party reporting (Kleven et
al., 2011).
2
Electronic copy available at: http://ssrn.com/abstract=1950473
 
 
 
--------------------------------------- 4
 
tant role. In particular, they: i) contributed to the di_usion of information about targeted
parcels; ii) collaborated to follow-up inspection to assign imputed values; iii) proceeded with
the collection of overdue local taxes up to ve years before the program; iv) veried con-
formity of ghost buildings to the city plan. Media reports highlight both the importance of
3
the role of local administration,and the heterogeneity in their actions in response to the
4
program.
The program can impact mayors' reelection likelihood through several channels. First, the
program shifts tax enforcement by reducing tax evasion. The potential e_ect of this channel
on reelection is a-priori ambiguous. On the one hand, evaders face costs from the increased
tax enforcement. On the other hand, non-evaders benets from the increased tax compliance,
which can lead to extra public goods or lower tax rates. Additionally, non-evaders could have
non-monetary benets by the increased e_ciency in tax collection. Even if evaders represent
a minority of the population (i.e. the median voter is a non-evader), the benets for the non-
evaders are potentially di_use among many voters, and they might be small in per capita.
On the contrary, evaders bear potentially high per capita costs. Thus, the net e_ect depends
on specic parameters such as preference for fairness of tax collection, and marginal utility
of private consumption from evaded tax payment. Second, the program can in principle
reveal information about the stock of tax evasion in each Italian municipality. However, as
we discuss later in the paper, it should be noted that most of the buildings targeted by
the program were likely to be built when the incumbent was in o_ce. This channel could
hurt the incumbent in areas characterized by higher political accountability and more tax
evasion, and could benet her in areas characterized by lower political accountability and less
tax evasion. Third, the program can in theory provide the incumbent with an opportunity to
5
extract political or monetary rents by selectively not enforcing the registration process.We
expect this e_ect to be more widespread in areas characterized by higher political shirking,
where voters are also less likely to punish politicians for misbehavior.
3
For example, Il Sole 24 Ore (2011b)
4
Among many others, La Nazione (2011) and L'Informazione (2008) discussed the particular way in which
respectively the city of Montecatini and some cities in the Reggio Emilia province implemented the program.
5
For example, the press agency of one of the mayor of a city in our sample, Capaccio Paestum, explicitly
criticized the \excessive media attention" to the program, indicating how the unregistered buildings in that
city were due to \citizens' needs" of his constituency (Comunicato Stampa n. 134/10 del 29.07.2010, Comune
di Capecci Paestum).
3
 
 
 
--------------------------------------- 5
 
Given that the central government initiated the program, it is important to understand
under which conditions the implementation should a_ect voters' preferences for local politi-
cians. In a standard Bayesian learning framework, voters would correctly anticipate on
average the distribution of responsiveness of each town to the program. Thus, according to
the martingale property, the signal generated by the program should lead, in expectation, to
no change in beliefs about the quality of the incumbent administration. However, the liter-
ature provides at least two classes of models that can predict an impact of the program on
voters' behavior. First, voters have imperfect information over the source of the additional
enforcement in their town. For instance, they could ascribe, on average, an excessive por-
tion of the additional enforcement to the local politician (Wolfers, 2007). In addition, Bubb
(2001) showed that even in a rational political agency model (Banks and Sundaram, 1998)
voter's rationality alone does not rule out a correlation between observable exogenous shocks
and support for the incumbent. The shock can change the signal extraction problem about
the type of incumbent. For instance, the politician's type could be more easily observed in
the presence of the program (higher \transparency"). This can occur, among many other
cases, in the presence of voters' risk aversion and skewness in the distribution of incumbents'
6
types.These models are consistent with a potential impact of the program on voters' be-
havior. Yet, as we argued, the fact that tax enforcement has an a-priori ambiguous e_ect on
incumbent support implies that the direction of such an impact is an empirical question.
To test whether the described anti-tax evasion policy a_ected incumbent reelection, we
exploit the cross-town heterogeneity in the intensity of ghost buildings targeted by the pro-
gram. This in turn shaped the program's potential to reduce evasion in each specic town.
This enables us to implement a di_erence-in-di_erences approach, based on the potential
intensity of the program, as opposed to the actual intensity, which is a_ected by endoge-
nous responsiveness at local level. In our preferred specication, we nd that, in elections
occurring after the beginning of the program, an increase in one standard deviation in the
intensity of ghost buildings targeted by the program raises the likelihood of re-election of
6
For example, risk-averse voters will on average prefer the candidate for which they have more information.
In this setting the incumbent is expected to benet from any \transparency" shock that reduces dispersion
of the beliefs, even without changing their mean. Alternatively, when the distribution of incumbents' types is
skewed, it is possible that whenever an administrative measure of tax evasion is readily available, ine_ective
actions by the government to reduce tax evasion good incumbents' e_ort might be more likely to separate
them from bad incumbents.
4
 
 
 
--------------------------------------- 6
 
the incumbent by approximately 3.5 percentage points, out of a mean of 45.4, relative to
pre-program elections. We provide evidence that the trends in reelection probability before
the program are homogeneous in the cities in our sample. In addition, our results survive a
large number of robustness checks.
We argue that that the additional tax enforcement induced by the program is the pre-
vailing force in driving the results. In addition to the \reduced form" approach used for our
main results, we thus provide other evidence consistent with this hypothesis. As we argue
later in the paper, these additional results are not consistent with the two alternative chan-
nels we discussed above (i.e. information revelation about the evasion baseline levels and
incumbent rent extraction). First, we show that the e_ect is driven by cities characterized by
higher social capital, where there is a higher degree of political accountability and political
misbehavior is usually punished (Nannicini, Stella, Tabellini and Troiano, 2011). Secondly,
we show that the program intensity has no e_ect in the South of Italy, an area where shirking
is more widespread compared to the rest of Italy (Ichino and Maggi, 2000). Finally, we show
that, for a given intensity of ghost buildings targeted by the program, a higher registration
rate of ghost buildings under the incumbent has a positive e_ect on reelection likelihood.
This strategy alleviates the concerns arising from potential correlation between registration
rates and incumbent type by providing an instrumental variable based on the time elapsed
between the program start date and the town election date.
The remainder of the paper is organized as it follows. Section 2 provides institutional
background on the Ghost Buildings program. Section 3 describes the dataset and provides
summary statistics. Section 4 lays out our empirical strategy and comments the main results
on the impact of the program on incumbent reelection. Section 5 provides additional evidence
on the impact of registration activities and rules our alternative explanations for the main
results. Section 6 concludes.
5
 
 
 
--------------------------------------- 7
 
2 Institutional Background
7
Italian legislationimposes that owners should register new buildings into the land registry
8
at the local o_ce of the Agenzia del Territorio within thirty days from their completion.
Failing to register is a form of tax evasion. Registered buildings enter the tax base for
9
\ICI", the local property tax,and for \IRPEF", the personal income tax which includes the
inferred opportunity cost of living in the house, with both a local and a national component.
Additional taxes associated with registered building ownership include registration fees and
local waste management tax.
In 2006, the government approved new anti-tax evasion legislation, the Ghost Buildings
10
Program,aimed at detecting buildings not registered in the land registry maps. The Agenzia
del Territorio, which, among other tasks, manages the land registry, coordinated the e_ort.
The Agenzia del Territorio rst juxtaposed the land registry maps and the buildings registry
maps to obtain the \O_cial Buildings Map". Then, it undertook an intense activity of
high-resolution (50cm) aerial photographing of the whole country. This activity provided
the basis for the identication of the ghost buildings. Figures 1a-1c summarize the steps
of the identication. First, the aerial photograph for a certain location is created (Figure
1a). Second, the pictures are matched with the o_cial buildings map for the corresponding
area (Figure 1b). After this comparison, unregistered buildings that do not qualify for being
11
dened as tax evasion are excluded.Finally, the ghost buildings are identied (Figure 1c).
These include stand-alone buildings but also substantial extensions of previously registered
buildings that should have been notied to the land registry.
As a result of this process, the Agenzia del Territorio identied around two millions land
7
Legge 9 Marzo 2006 n.80 - Art. 34-quinquies.
8
Before this law, the Reggio Decreto n. 652 of April 1939 had as threshold the 31st of January of the year
following to the end of the completion of the building. All buildings in Italy need a building permit before
their construction starts. Obtaining a building permit makes the building part of the City Plan. The process
of obtaining building permits is administered independently from the registration in the land registry maps.
Buildings not in the City Plan are required to be demolished.
9
Property tax on the rst owned house has been abolished starting from 2008, and replaced by a transfer
from the national government to the municipalities equivalent to the virtual revenues if rst owned houses
had been taxed.
10
Legge 24 novembre 2006, n. 286 subsequently modied by Legge 30 Luglio 2010, n. 122
11
According to the Law Decreto Ministero delle Finanze 2 gennaio 1998, n.28.Art. 3 the following buildings
do not increase the tax base of their owners: (i) buildings that are not completed (ii) buildings particularly
degraded (iii) solar collectors (iv) greenhouses (v) henhouses or other reserved for animals
6
 
 
 
--------------------------------------- 8
 
12
registry parcels with unregistered buildings.Beginning in August 2007, the Agenzia del
Territorio started to publish parcel-level data on unregistered properties in the Gazzetta Uf-
ciale, in order to induce registration of the ghost buildings. Within three years, it coded
detailed information on the number of ghost buildings in the universe of Italian municipal-
ities (with the exception of Trentino Alto-Adige). The order of publication relied on the
availability of land registry maps in digitalized form at the time when the program started.
The Agenzia del Territorio had 60% of the land registry maps of the Italian territory in
digitalized form before the Ghost Buildings program was approved. After 2006, the Agen-
zia del Territorio started the process of digitalization of the rest of the land registry maps,
proceeding by province (i.e. they coded simultaneously municipalities in the same province).
At completion of the land registry map updates, new waves of publishing of unauthorized
13
buildings occurred.
According to the initial legislation, owners could register the detected ghost building into
the land registry within ninety days from the publication. Widespread media campaign, both
on newspapers and broadcasting, and local administration follow-up activities supported the
program to achieve high registration rates. The deadline was rst extended to seven months
after the publication and then further extended to April 30, 2011. After that, the Agenzia
del Territorio proceeded with a follow-up inspection, with the support of local administra-
14
tionand local contractors, in order to impute the tax base for the remaining unregistered
15
buildings.
Owners of ghost buildings registered within the April 2011 deadline were subject to pay-
ment for overdue taxes going back up to 2007, or to the construction date if post 2007. In
16
addition, owners also had to pay penalties for the delayed payments.Additional penalties
and a fee for the extra inspection applied to the building for which the Agenzia del Territorio
imputed the tax base after April 2011.
12
The unit of the Italian land registry maps is the parcel (parcella), which is dened as a portion of land
belonging to a given physical or legal person. In the case where the land is shared across several owners, the
parcel is split into several sub-parcels (subalterni)
13
O_cial data were published by the Agenzia del Territorio in the journal Gazzetta U_ciale in the following
waves: August 2007, October 2007, December 2007, December 2008, December 2009, December 2010.
14
In order to further increase incentives for the local administrations to promote the registration activities,
an additional bonus was introduced in 2011 for each registered ghost building.
15
Decreto Legge 79/2010, art. 10, 11
16
Penalties were determined by Legge 29 Dicembre 1990, n. 408, subsequently modied by Decreto Leg-
islativo. 18 Dicembre 1997
7
 
 
 
--------------------------------------- 9
 
The Agenzia del Territorio published the detailed economic impact of the program for the
year 2011. According to the gures provided, the program increased yearly total tax revenues
17
by 472 millions euros.We estimate that around 65 percent of those revenues come from
local taxes. According to a simple back of the envelope calculation using information on the
number of land parcels with ghost buildings, on the registration rates and on total additional
tax revenues from the program, we estimate that a one standard deviation of ghost buildings
18
targeted by the program increased the tax revenues by around 120 thousands Euros. Using
the same information, we nd that, on average, the owner of a registered ghost building faces
an additional yearly tax burden of approximately 528 Euros.
3 Data and Summary Statistics
We obtain data from the Agenzia del Territorio on the number of land parcels (particelle)
19
for each town where the exercise was conducted. The database includes information on the
number of parcels containing Ghost Buildings in each town. The aerial photograph activity
detected more than two millions of such parcels. As discussed above, the exercise did not
cover one of the semi-autonomous regions, Trentino Alto-Adige, because in that region land
registry maps are autonomously administered. Thus, we target a population of 7,760 of the
8,092 Italian towns (Comuni). Figure 2 presents the intensity of ghost buildings across the
Italian provinces (normalized per thousand of total houses). As expected, tax evasion is
more prevalent in Southern Italy (including Sicily), and it is less widespread in the North.
In Figure 3, we summarize the fraction of ghost buildings that get registered by the 30th of
April 2011, the cuto_ date for which we obtained the registration outcomes we use for our
20
analysis.There is a strong territorial heterogeneity in the waves of registrations that occur
in response to the program, and this suggests that idiosyncratic factors might play a role in
the implementation of the program.
17
This gure does not include payment for overdue taxed from previous years.
18
To compute this gure, we divide the total increase in tax revenues, by the total number of parcels with
ghost buildings and then we multiply it by the standard deviation of those parcels
19
Only 1.4% of towns was found to have no ghost building
20
As we discussed above, the Agenzia del Territorio subsequently released a registration summary at the
end of 2011, which included results from several months of follow-up inspections to determine tax base for
buildings not yet registered by April 2011. However, given that the latest election round in our database
occurred in May 2011, we choose to focus on the April 2011 outcomes.
8
 
 
 
--------------------------------------- 10
 
We complement this information with data from the Italian Ministry of Internal A_airs
21
containing outcomes for the universe of municipal elections from 1993 until 2011on electoral
outcomes and local tax revenues. Additionally, we collect geographical, social and economic
data at the municipality level from the Italian National Statistical O_ce. In Figure 4 we plot
the number of elections by year, from 1993 to 2011. We distinguish between elections before
and after the beginning of the Ghost Buildings program. Given that the program started
in 2006, and the disclosure started in the end of 2007, it is not surprising to note that all
the elections from 1993 to 2007 are considered elections \pre-program". It should be noted
from the graph that the dates of municipal elections vary across towns. In addition, it is
important to discuss two institutional reforms that happened in our sample of interest. In
1993, the Italian municipal politics was overhauled: a law introduced a new electoral rule
(single round cities with lower than 15,000 inhabitants and a runo_ above) and a two term
limit for the mayors elected from 1993 onward. The other institutional reform that is relevant
for our sample is the one in 2000, when the length of the mayoral term was extended from
four to ve years. The fact that both municipality elections and disclosure of tax evasion
information do not happen at the same time across towns will constitute the basis for one
of our robustness checks. In Figure 5 we plot the ranking of the municipal elections with
respect to the beginning of the program in the town. As we can see from the graph, there are
almost 5,200 municipalities for which we have data on an election that is after the disclosure
of the tax evasion data (about 66% of the total number of towns targeted by the program).
Table 1- Panel A presents summary statistics on the unauthorized buildings share, and our
controls and variables of interest, for each Italian city in our sample. In Panel B we instead
22
summarize the election panel variables that we use to test our empirical specications.
We exploit the amount of evasion detected by the Ghost Buildings program to identify
the correlates of tax evasion. We estimate the following equation (where i is municipality)
GBi = _ + Gi + DSi
 + _i(1)
where GBi is the intensity of ghost buildings, Gi includes geographical controls (altitude, area
21
The Italian municipal government (Comune) is composed of a mayor (Sindaco), an executive committee
(Giunta) appointed by the mayor, and an elected city council (Consiglio Comunale).
22
It should be noted, given that our main outcome of interest is the probability of reelection, that we do
not summarize the variables of the elections in which the mayor has a term limit. The summaries for the
whole sample are qualitatively similar and are available upon request.
9
 
 
 
--------------------------------------- 11
 
size of the municipality, number of land registry parcels), and DSi includes demographic and
socio-economic (population, income per capita and social capital, measured as number of non
prot associations per capita, number of rms, urbanization rate). Table 2 includes three
specications. We rst study whether geographical controls are correlated with tax evasion,
then we add socio-economic controls and in the third column we show that our results are
23
una_ected by regions xed e_ects. Geographical factors are strongly associated with tax
evasion. It is particularly interesting to note that tax evasion is associated with a large area
size of the municipality. Plausibly, in cities with wide geographical extension, the oppor-
tunities for unregistered buildings are higher as the enforcement of building registration is
harder. However, we cannot decisively interpret this evidence as causal. Previous literature
has shown for example that borders are endogenously determined (see, among others, Alesina
and Spolaore, Alesina et al. 2004, Alesina et al. 2011b). Finally, as expected, tax evasion is
negatively associated with both social capital and income.
Finally, we show that the number of unregistered buildings detected by the program is
a good predictor of the number of registered ghost buildings because of the intervention. In
Table 3, we provide some correlations between the percentage of parcels with ghost buildings
eventually registered as of April 30, 2011 and the percentage of parcels that were detected
as containing ghost buildings. We study the raw correlation in column (1), and we add town
controls and region xed e_ects respectively in columns (2) and (3). According to these
results, an increase of one s.d. in the intensity of targeted ghost buildings raises the intensity
of registered ghost buildings at April 2011 by 0.75 s.d. Thus, the program potential intensity
strongly predicts actual intensity. This premise motivates the analysis of the impact of the
program on electoral outcomes that we present in the next section.
23
We notice that by including all the available town level controls we lose about 3.5% of the towns in our
sample. The results of our paper are una_ected by the inclusion of this subsample of towns
10
 
 
 
--------------------------------------- 12
 
4 The Electoral Response to the Ghost Buildings Pro-
gram
4.1 Baseline Results
In this section we investigate the electoral consequences of the Ghost Buildings program. Our
identication strategy relies on the fact that the program had di_erent potential intensity
24
across towns, depending on the town-level intensity of parcels containing ghost buildings.
We thus implement a di_erence-in-di_erences approach based on the intensity of detected
ghost buildings, which we dened above as the ratio between the number of land registry
parcels with ghost buildings and total number of land registry parcels in the town. By
focusing on the program potential intensity, this approach allows us to deal with the endoge-
nous level of actual ghost buildings registrations in each town, which is likely to re
ect town
25
idiosyncratic elements, such as citizens' and politicians' responsiveness.
Equation 2 reports our baseline estimation model:
Rie = _ + P ostP rogramie +
P ostP rogramie GBi + _i + _it(2)
The dependent variable Rit is a dummy that indicates whether the incumbent of municipality
i has been reelected in the elections e. Observations where the incumbent cannot legally be
26
reelected because of a term limit are excluded from the regression sample.The dummy
PostProgram assumes the value 1 when election e occurs after the beginning of the Ghost
Buildings program in the town and zero otherwise. GBi is the intensity of ghost buildings
detected in town i, and _i is a municipal xed e_ect. The coe_cient of interest is
, which
captures the di_erential impact of the program by ghost building intensity. As it is standard
in di_erence-in-di_erences estimation, identication of such a coe_cient requires the parallel
trend assumption to hold, a condition that we verify later in the paper. Similarly, the estimate
24
Importantly, the Agenzia del Territorio conducted the detection activities homogeneously throughout
the country. Thus, heterogeneity in the number of detected unregistered buildings is a proper measure of
actual levels of non-registration at the time of the aerial photographing.
25
Our approach can be a considered a reduced form one. Importantly, we believe that the ghost buildings
intensity cannot be used as an instrument for actual registration intensity. In principle, as we discussed in
Section 1, the program could a_ect incumbent election probability through other channels besides registra-
tion. This would make the standard exclusion restriction required for an instrumental variable approach
invalid. Later in the paper, we show evidence consistent with the hypothesis that actual registration plays
an important role in the results.
26
Our results are una_ected by this sample selection.
11
 
 
 
--------------------------------------- 13
 
of the coe_cient captures time-trend and thus, in this basic specication, cannot simply
be interpreted as the causal e_ect of implementing the program in a town with no detected
27
ghost building. We always cluster standard errors at provincial level.
One potential identication threat may arise from the timing of publications of the unau-
thorized building lists. For instance, if local administrators had in
uence over publication
date, unpopular mayors in cities with high evasion might lobby to delay the publication. In
such a case, our estimates of a positive interaction between baseline tax evasion levels and
post-program dummy might just capture a selection problem. However, as we discussed in
Section 2, the timing of the publication was determined by the availability of digital land
registry maps and was highly clustered at provincial and regional level. Any provincial unob-
served heterogeneity driving the initial availability of land registry maps will be absorbed by
the town xed e_ects in our model. In addition, we nd that about 7% of the post-program
elections have values for the post-program indicator di_erent than the one they would have
had based on the modal date of publication in the province. In order to deal with these id-
iosyncratic discrepancies we implement an instrumental variable approach. We code elections
based on whether they occur before or after the modal date of publication of the unautho-
rized building lists in the province. We then instrument the actual P ostP rogram dummy
with this indicator. We adopt this strategy for our main specications.
Table 4 presents our baseline results on the impact of the Ghost-Building program on
local incumbent reelection. Column (1) reports the basic OLS specication and Column (2)
reports the reduced form specication. In the latter, our independent variable of interest is
the interaction between ghost buildings intensity and a dummy variable that re
ects whether
the election in the municipality is after the modal date of publication of the unauthorized
buildings list in the province. The reported coe_cient on the ghost buildings intensity is
1.659, signicant at 1%. This magnitude implies that a one s.d. increase in the intensity of
ghost buildings targeted by the program raises the likelihood of reelection of the incumbent
by 3.5 percentage points in post-program elections, relative to pre-program ones (out of a
sample mean of 45.4).
Starting in Column (3), and for the rest of the table, we instrument the post-program
indicator with the provincial post-program dummy described above, so to rule out concerns
27
The towns targeted by the program belonged to 101 provinces.
12
 
 
 
--------------------------------------- 14
 
regarding the manipulation of the date of the beginning of the program. The coe_cient is
stable across the di_erent specications. The inclusion of year xed e_ects and town-xed
e_ects, respectively in Columns (4) and (5) does not change our results. Finally, in Column
(6), we add year trend interacted with ghost building intensity. The coe_cient slightly raises
and remains signicant at 1%. Under this specication, this magnitude implies that one s.d.
increase in the intensity of ghost buildings targeted by the program raises the likelihood of
reelection of the incumbent by 4 percentage points in the post-program.
4.2 Robustness Checks
In this section, we discuss some potential concerns related to our empirical specication and
provide robustness checks to our results. The results are presented in Table 5. Column (1)
shows our baseline specication, previously reported in column (5) of Table 4, which includes
both year and town xed e_ects. In column (2) we introduce the interaction between the
post-program indicator and a xed e_ect for each of the nineteen regions targeted by the
28
program.The coe_cient falls to 1.12 (though not signicantly di_erent than the one
estimated in our baseline specication) and remains signicant at 1%. In Column (3) we
add the interaction between the town controls reported in Column (3) of Table 2 and the
post-program indicator. The coe_cient is stable relative to column (2). In Column (4),
we introduce an interaction between each year xed e_ect and ghost-buildings intensity. In
this specication we thus identify the coe_cient
 by comparing electoral outcomes across
towns with the same intensity of ghost buildings but di_erent program status (i.e. started
vs. non-started) in a given electoral year. As Figure 4 shows, identication of the coe_cients
of interest in this specication comes mainly from one year, 2009, for which we have a sizable
number of both pre-program and post-program elections. Lack of power thus prevents us from
precisely estimating the coe_cient. Thus, the magnitude of the coe_cient, even if comparable
to the one estimated from the di_erence-in-di_erences identication, should be interpreted
with caution given the few cities that drive this estimate. Another possible concern is related
to the choice of the control group in our di_erence-in-di_erences strategy. Given that we
have a post election for a subset of Italian cities, in Column (5) we repeat our empirical
28
When we implement this specication our instrument becomes weak, because of the high correlation
between regions and the beginning of the program. Results are unchanged is we re-estimate our equation
through LIML rather than 2SLS.
13
 
 
 
--------------------------------------- 15
 
specication by restricting the sample only to cities for which we have at least one election
after the treatment. The results are unchanged: the size and statistical signicance of the
e_ect are unaltered. Additionally, one might be concerned that our results are driven by
cities with a large number of not residents who own a building. If this were the case, mayors
would not pay the electoral cost of enforcing tax evasion. We therefore exclude in Column
29
(6) cities that are classied as touristic by the Italian Association of Cities, Ancitel.One
could also be concerned that outliers are driving our results (i.e. cities with an abnormally
large fraction of unregistered buildings). In Column (7) we show that our results hold if we
trim our outcome at the top 1 percent. If anything, the coe_cient size grows. Finally, we
report an alternative normalization for our dependent variable. In our baseline specication,
since the data reported by the Agenzia del Territorio included the number of parcels with
unregistered buildings, we divided this outcome by the total number of particles in each
municipality. Alternatively, in Column (8), we estimate our equation using as a normalizing
factor the total number of buildings recorded in the town, rather than the total number of
land registry parcels. We get very similar results. The e_ect of one standard deviation in
this alternative variable (.084) is comparable in magnitude to the one obtained when using
our main ghost building intensity variable based on parcel number.
5 Channels: Impact Heterogeneity and Buildings Reg-
istration
At their starkest, the results presented in the previous section show that an increase in
the intensity of ghost buildings targeted by the program raises the incumbent reelection
likelihood. In this section, we present additional evidence consistent with the hypothesis
that it is the registration induced by the program that drives our results. According to
this interpretation, the program generates an increase in the level of tax enforcement, which
a_ects the electoral consensus for the incumbent. We provide support for this statement.
First, we report an additional set of results consistent with this hypothesis, testing for impact
heterogeneity and then using actual ghost buildings registration data. Second, we discuss
29
Ancitel denes as touristic those cities with a large percentage of touristic income over the total city
income. As noted by the Italian Association of Real Estate Agents (FIAIP) in their yearly report cities with
touristic activities are usually those with a large number of buildings not owned by residents
14
 
 
 
--------------------------------------- 16
 
several theoretical frameworks consistent with the empirical evidence. Third, we use our
results to rule out potential alternative explanations.
In order to shed further light on the interpretation of the results, we test for two dimen-
sions of heterogeneity in the electoral responsiveness for the program. The rst dimension
is the geographical one. Italy is ideally suited for nding geographic heterogeneous treat-
ment, because large di_erences in economic development, history, attitudes and many other
dimensions persist within the country. The second dimension is heterogeneity in the degree
of social capital across towns. Several political scientists and economists have shown that
social capital is important for understanding economic development and the functioning of
institutions (Baneld 1958; Putnam 1993, 2000; Fukuyama 1995; Guiso, Sapienza, and Zin-
gales 2008; Tabellini 2008, 2009; Algan and Cahuc 2010; Aghion et al. 2010). Following
Guiso, Sapienza and Zingales (2008), we use as a proxy for social capital the number of non
prot associations per capita, which is, to the best of our knowledge, the only proxy for social
capital available at the city level.
Table 6 assesses to which extent this heterogeneity matters. We add the triple interac-
tion between ghost buildings intensity, post program election dummy, and the heterogeneity
dimension of interest to Equation 2, as well as the interaction between post program election
dummy and heterogeneity. In Column (1), we notice that ghost buildings intensity does
not a_ect the reelection probability in the post program elections in Southern Italy. Ichino
and Maggi (2000) argue that misconduct and shirking episodes are much more widespread in
Southern Italy. In Column (2), we notice that the electoral response to the program is weaker
in areas with social capital below median. Nannicini, Stella, Tabellini and Troiano (2012)
argue that Italian voters in higher social capital areas are more likely to hold politicians ac-
countable, and are less tolerant of misbehaviors and shirking by their elected representatives.
These results provide suggestive evidence that the results are driven by voters rewarding
politicians for the increase in tax evasion enforcement generated by the program. If, on the
contrary, voters were rewarding politicians for having allowed tax evasion in the past -one of
the potential alternative interpretations discussed in Section 1 - we would expect the electoral
impact to be stronger, not weaker, in Southern regions and in areas with lower social capital.
In order to provide further support for our interpretation, we then move to the analysis
of the actual registration data. In Section 3 we emphasized several important limitations
15
 
 
 
--------------------------------------- 17
 
of these data. As we mentioned, we obtained data on the number of parcels that contain
a registered ghost building by April 2011. We rst normalize this outcome by the total
number of parcels in which the Agenzia del Territorio detected ghost buildings, thus creating
a \registration rate" variable. We then construct a measure of registration imputable to
the incumbent administration. Specically, we multiply the registration rate by the ratio
between the time elapsed between program start date and election date and the time elapsed
between program start date and April 2011. Measurement error should provide a downward
bias in estimating the e_ect of registration on reelection probability. However, measurement
error is not the only factor that makes us very cautious in using registration data. The e_ort
in registering ghost buildings is endogenous and could be correlated with many potential
confounders, such as mayors ability. Ideally, we would need an instrument that a_ects the
probability of reelection only though the registration program. We argue that the timing
of the program might allow us to implement a strategy that can alleviate the endogeneity
concerns related to the mayoral ability. Even if the program in most of our cities started in the
same year, we can exploit the variation generated by the fact that Italian municipalities hold
elections in di_erent years to alleviate the endogeneity concerns. More time to implement
the program naturally leads to more registration activities. This generates variation in the
registration rate achieved at election date across towns that is uncorrelated with mayoral
quality and re
ects only programs characteristics.
Table 7 presents the results. In Column (1) and (2), we present the correlation between the
registration rate and the incumbent reelection. We nd that, controlling for ghost building
intensity, a one s.d. increase in registration rate raises reelection likelihood by 1.3 percentage
points. The result holds both when using the April 2011 registration rate while controlling
for year xed e_ects and when using the registration rate in the election year computed as
30
we described above.In Column (3), we show that the motivation for our instrument is
supported by the data, and years elapsed since the program start at election time are a good
predictor of the registration rate at that time. In Column (4), we present the reduced form
results. In Column (5) we use the years elapsed since the program start as an instrument for
30
In our IV specication we do not control for year xed e_ects. Three quarters of the post-program
elections come from cities that started the program in 2007. Thus, we loose statistical signicance when
running this specication, even if it is reassuring that the coe_cients remains of similar size. Results are
available upon request.
16
 
 
 
--------------------------------------- 18
 
31
registration rate.Columns (3)-(5) show a consistent story. For a given potential intensity
of the program, as determined by the intensity of detected ghost buildings, towns with higher
registration rates are more likely to reelect the incumbent. In the IV specication, a one s.d.
increase in the registration rate (.079) raises the reelection likelihood by 4 percentage points
in post-program elections. We notice that the IV estimates are larger than the OLS ones.
This is relatively common (see, for example, the return to schooling literature, reviewed in
Card, 2001), and it can be explained by the fact that the use of IV may reduce the downward
attenuation bias due to measurement error, or by the fact that in the set of cities which
the IV estimates are identied o_ - that is, cities where the registration activity depends on
program duration - the political returns to regularization might be bigger than in the rest of
the cities (i.e. we are estimating a LATE).
The results presented so far in this section suggest that voters reward mayors for the reduc-
tion in tax evasion induced by the program. It is important to understand which theoretical
models can generate such a prediction. As we discussed in Section 2, the local administra-
tion did play an important role in the registration and the tax enforcement process. Yet,
in a standard model with rational voters and perfect information about the program, voters
should fully incorporate information about the nationally implemented program and discount
accordingly the increased enforcement they observe in their town when updating their beliefs
about incumbent's type. Thus,on average, the signal generated by the reduced tax evasion
should not change the beliefs about the current politician. The literature presents two classes
of models that relax some of the assumptions of the basic Bayesian voting model and deliver
predictions that are consistent with the empirical evidence we presented. In the rst, voters
systematically ascribe to the local administrations an excessive component of the observed
additional enforcement, without properly incorporating the program implementation in their
updating (Wolfers, 2009). In the second, even under the assumption of rational voters, the
program changes the signal extraction problem about the incumbent type in a way that could
systematically generate an incumbent advantage or disadvantage (Bubb, 2008).
Even if voters ascribe a portion of the additional enforcement to local politicians, the
impact on incumbent relection prospect is a-priori ambiguous. The motivating intuition is
31
In our IV specication we dont control for year xed e_ects. Given that the program started in three
quarters of the cities at the same time, this specication is too demanding and we loose statistical signicance,
even if it is reassuring that the coe_cients remains of similar size. Results are available upon request.
17
 
 
 
--------------------------------------- 19
 
that tax enforcement is not a \windfall" but rather a policy variable that induces gains
and losses for di_erent members of the constituency. First, cracking tax evasion obviously
hurts evaders. On the other hand, non evaders can gain utility by the increased tax base
or by a perceived increase in the e_ciency of the tax collection process, resulting from the
punishment of former evaders. The net e_ect on voters' average responsiveness depends on
model parameters, such as those characterizing the distribution of evasion, preference for
fairness, and marginal utility of private consumption from evasion. Thus, our results, which
show a positive incumbent e_ect of ghting tax evasion, crucially depend on the specic form
of tax evasion and on the magnitude of these forces in our specic case. We acknowledge
this as an important caveat in generalizing the results, an issue on which we come back in
Section 6.
Finally, we use the entire set of our results to argue that the impact on incumbent re-
election probability arising from the increase in tax enforcement more than o_sets several
alternative interpretations we mentioned in Section 1. According to the rst of these expla-
nations, the publication of the number of ghost buildings generates information about the
incumbent. We believe this to be both unlikely and inconsistent with our ndings. First, the
set of ghost buildings is a slow moving stock variable likely to be accumulated over decades,
rather than a re
ection of just the most recent years. Most of the buildings found by the
Agenzia del Territorio were not new constructions. In addition, the existence of a term limit,
paired with the fact that the time to complete a building in Italy is generally longer than
most of the other OECD countries, suggests that most of these buildings could not have
been built when the incumbent was in o_ce. Second, we notice that voters who could poten-
tially receive information from the publication are most likely the ones that were not evading
before, since evaders already knew they were evading. With this premise, we then use our
results to rule out this explanation. In one version of this alternative story, voters, after
learning about low level of evasions detected by the program, reward the current mayor for
having properly enforced tax payment in the past. This hypothesis predicts a negative impact
of the detected ghost buildings intensity on incumbent reelection in post-program elections,
and as such it is obviously inconsistent with our baseline results. In another version, voters
reward incumbent mayor for having allowed high levels of evasion in the past. First, this is
inconsistent with the intuition that non-evaders, rather than those previously evading, are
18
 
 
 
--------------------------------------- 20
 
the ones who are potentially acquiring new information. Second, this is unlikely since the
purpose of the program, and thus of the publication, is explicitly to shut down the evasion
opportunity. Third, it is at odd with the fact that the positive impact of program intensity
on incumbent reelection is lower in the South and in towns with low social capital. Fourth,
this is inconsistent with our results showing that, for a given intensity of ghost buildings,
towns with higher registration levels are more likely, rather than less likely, to reelect the
incumbent mayor.
In a second potential alternative explanation, the program gives the incumbent an elec-
toral rent by allowing her not to register the targeted ghost buildings, for instance by report-
ing errors in the results generated by the mapping process. As before, such an explanation is
inconsistent with the result showing that, for given baseline evasion intensity, a higher share
of registered ghost buildings by the time of the election increases the reelection likelihood.
To summarize, we believe the results of this section provide strong evidence that it is the ad-
ditional tax enforcement induced by the program the driving force which leads to an increase
in the reelection prospect of the incumbent.
6 Conclusion
A rapidly growing literature studies which factors could potentially increase tax compliance
(Kleven et al. 2011, Pomeranz 2011). In this paper, we build a bridge between this literature
and the one estimating the political return of scal reforms (Brender and Drazen, 2008;
Alesina et al., 2011a). The premise for our analysis is that politicians set tax enforcement
levels by weighing cost and benets. A marginal increase in enforcement hurts a portion
of the voters (the evaders) and benets others, through either direct utility from increased
e_ciency of tax collection or additional public good provision and lower tax rates. We
use a nationwide tax evasion intervention in Italy to estimate the political returns from an
exogenous shifter in tax enforcement. In our rst contribution, we use administrative data
from the Ghost Buildings program to provide a new disaggregated measure of tax evasion,
using information on the number of unauthorized buildings in each parcel of the national
land registry.
We then estimate the impact of the anti-tax evasion policy on local politicians' reelection.
19
 
 
 
--------------------------------------- 21
 
While the central government started the program, local administrators played an important
role in circulating the information, collaborating to the follow-up inspections and enforcing
payment of overdue taxes. We implement a di_erence-in-di_erences approach, relying on the
cross-town heterogeneity in the intensity of ghost-buildings, and thus in the potential impact
of the program on tax enforcement. Using estimates from our preferred specication, we nd
that, in post-program elections, an increase of one s.d. in baseline tax evasion raises the
likelihood of reelection by 3.5 percentage points (relative to pre-publication elections).
We then provide evidence consistent with the hypothesis that the registration of ghost
buildings induced by the program drives our results. We provide a set of further results
to support this claim. First, the results are driven by places that are more socially and
economically developed, i.e. where previous literature showed that political accountability
and compliance are generally higher. Second, we show that, for a given intensity of targeted
ghost buildings, towns with higher registration rates during the current mandate are more
likely to reelect the incumbent mayor. We also develop an instrumental variable approach
based on the timing of the elections to deal with the endogeneity in registration activities.
This set of results allows us to rule out alternative interpretations arguing, for instance,
that voters responded to the information about the baseline level of tax evasion detected
by the program or that mayors extracted electoral rents by limiting registration of targeted
buildings.
Our ndings provide the rst evidence that voters reward the ght to tax evasion. There
are at least two possible explanations that would be consistent with this nding. First, tax
compliers reward the incumbent for increasing tax revenues, and hence increasing public
goods provisions or reducing future tax rates. This e_ect starts with the beginning of the
program. Second, tax compliers get non monetary benet by the increased e_ciency of the
tax enforcement. The results are consistent with two families of political economy models,
either requiring voters' imperfect information or perfectly informed voters and an interac-
tion between politicians' e_ort and the implementation of the program. In the rst, voters
have imperfect information over the source of the additional enforcement in their town, and
ascribe, on average, an excessive portion of the additional enforcement to the local politi-
cian (Wolfers, 2007). In the second, voters have perfect information on the program, but its
implementation changes the voters' signal extraction problem about the administrator type
20
 
 
 
--------------------------------------- 22
 
(Banks and Sundaram, 1998; Bubb, 2011).
We thus nd that reforms aiming to achieve higher tax enforcement have positive political
returns for local administrators, consistent with the ndings of Brender and Drazen (2008),
who show that scal stabilization is associated with higher reelection, against conventional
wisdom. We are aware that the enhanced internal validity of studying the political returns
to reforms in a within country setting comes at the price of lower external validity. However,
we think that the internal heterogeneity that characterizes Italy allows us to draw some
suggestive conclusions on the external validity of our results. Italian cities di_er not only in
terms of economic development, but also in terms of history colonization (South: Spanish
domination; North: Austrian-German domination), crime rates, shirking, passive waste, and
political accountability. In particular, the fact that our results are driven by places with
higher economic development (Northern and Central Italy) and higher political accountability
(places with higher social capital) suggests that the ndings might be more relevant for more
developed and mature democracies, consistently with the heterogeneity found by Brender
and Drazen (2008).
21
 
 
 
--------------------------------------- 23
 
References
Agenzia del Territorio, (2011), \Comunicato Stampa: Regolarizzazione degli immobili fan-
tasma: dati al 30 Aprile 2011"
Aghion, P., Y. Algan, P. Cahuc, and A. Shleifer (2010), \Regulation and Distrust," Quarterly
Journal of Economics, 125, 1015{1049.
Alesina A. and Enrico Spolaore (1997) \On the number and size of nations". The Quarterly
Journal of Economics, 112(4):1027-1056
Alesina, A., R. Baqir and C. Hoxby (2004) \Political jurisdictions in heterogeneous commu-
nities". The Journal of Political Economy, 112(2):348-396
Alesina, A., D. Carloni and G. Lecce (2011a), \The Electoral Consequences of Large Fiscal
Adjustments", Harvard University working paper.
Alesina, A., W. Easterly and J. Matuszeski (2011b) \Articial States", Journal of the Euro-
pean Economic Association, 9(2): 246-277
Algan, Y., and P. Cahuc (2010), \Inherited Trust and Growth," American Economic Review,
100, 2060{92.
Andreoni, James, Brian Erard and Jonathan Feinstein (1998) \Tax Compliance" Journal of
Economic Literature. 36(2):818-860.
Bandiera, O., A. Prat, and T. Valletti (2009): \Active and Passive Waste in Government
Spending: Evidence from a Policy Experiment," American Economic Review, 99, 1278{
1308.
Baneld, E. (1958), The Moral Basis of a Backward Society, New York, NY: Free Press.
Banks J. and R. Sundaram(1998): \Optimal Retention in Agency Problems" Journal of
Economic Theory, 82(2), 293-323.
Berdejo C. and N. Yuchtman (2010) Crime, punishment, and politics: an analysis of political
cycles in criminal sentencing, working paper
22
 
 
 
--------------------------------------- 24
 
Brender, Adi and Allan Drazen (2008) \How Do Budget Decits and Economic Growth A_ect
Reelection Prospects? Evidence from a Large Panel of Countries", American Economic
Review, 98(5): 2203-20
Bubb (2011), \ Blame It On The Rain? Voter Rationality and Exogenous Economic Shocks?",
New York University working paper
Cannari, L., and G. D'Alessio (2006), \The opinion of Italians on tax evasion" , Bank of Italy
working paper.
Chetty, R., J. Friedman and E. Saez (2011), \Using Neighborhood E_ects to Uncover the
Impacts of Tax Policy: The E_ect of the EITC on Earnings", working paper
Comunicato Stampa n. 134/10 del 29.07.2010 Comune di Capaccio Paestum
Di Tella, R. and E. Schargrodsky (2003), \The Role of Wages and Auditing during a
Crackdown on Corruption in the Cityof Buenos Aires", Journal of Law and Economics
46(1):269-292, 2003
Ferraz, C. and F. Finan (2008): \Exposing Corrupt Politicians: The E_ects of Brazils Pub-
licly Released Audits on Electoral Outcomes," Quarterly Journal of Economics, 123, pp.
703-745
Fisman, Raymond and Shang-Jin Wei. 2004. \Tax Rates and Tax Evasion: Evidence from
Missing Imports in China" Journal of Political Economy 112 (471-500)
Gordon, Roger H. and Wei Li. (2009). \Tax Structure in Developing Countries: Many
Puzzles and a Possible Explanation." Journal of Public Economics. 93:7-8 (855-866).
Guiso, L., P. Sapienza, and L. Zingales (2008), \Long Term Persistence," mimeo.
Ichino, A., and G. Maggi (2000): \Work Environment and Individual Background: Ex-
plaining Regional Shirking Di_erentials in a Large Italian Firm," Quarterly Journal of
Economics, 15, 1057{90.
Il Sole 24 ore (2011), \Case Fantasma", Franco Guazzone 23rd Janyart 2011
23
 
 
 
--------------------------------------- 25
 
Il Sole 24 Ore (2011b), \Il Tesoro delle Case Fantasma", Dell'Osta C. and G. Trovati
Kleven, H., C. Kreiner, and E. Saez (2009). \Why Can Modern Governments Tax So Much?
An Agency Model of Firms as Fiscal Intermediaries." NBER Working Paper 15218,
August 2009.
Kleven H., M. Knudsen, C. Kreiner, S. Pedersen , and E. Saez. (2011) \Unwilling or Unable to
Cheat? Evidence from a Randomized Tax Audit Experiment in Denmark" Econometrica
79(3), 2011, 651-692
L'Informazione Provincia Montagna, (2008), \I primi cittadini della montagna commentano
i dati apparsi sul sito dell'Agenzia del territorio sugli immobili non denunciati", Matteo
Barca 21th March
La Nazione (2011) \Il Comune va a caccia delle case 'fantasma" , Daniele Bernardini 18th
March
Marino, M. and R. Zizza (2008), \The personal income tax evasion in Italy", Bank of Italy
working paper.
Nannicini, T., A. Stella, G. Tabellini, and U. Troiano (2010): \Social Capital and Political
Accountability," CEPR Discussion Paper 7782.
Peltzman, Sam. \Voters as Fiscal Conservatives." Quarterly Journal of Economics 107
(1992): 327-361.
Pomeranz D. (2011), \No Taxation without Information: Deterrence and Self-Enforcement
in the Value Added Tax ", working paper
Putnam, R. (1993), Making Democracy Work. Civic Traditions in Modern Italy, Princeton,
NJ: Princeton University Press.
Saez, Emmanuel (2010) \Do Taxpayers Bunch at Kink Points?" American Economic Journal:
Economic Policy, 2, pp. 180-212.
Saiz, A. (2010). \The Geographic Determinants of Housing Supply," The Quarterly Journal
of Economics, vol. 125(3), pages 1253-1296
24
 
 
 
--------------------------------------- 26
 
Tabellini, G. (2008), \The Scope of Cooperation: Values and Incentives," Quarterly Journal
of Economics, 123, 905{950.
Tabellini, G. (2009), \Culture and Institutions: Economic Development in the Regions of
Europe," Journal of the European Economic Association, 6, 255{294.
Wolfers, J. (2007). \Are Voters Rational? Evidence from Gubernatorial Elections", working
paper
25
 
 
 
--------------------------------------- 27 
 
Figure 1: The Ghost Buildings Identication
Source:Agenzia del Territorio
26
 
 
 
 
--------------------------------------- 28

 

Figure 2: Ghost Buildings Intensity (parcels with ghost buildings per 1,000 parcels)
Notes: The gure presents the number of land registry parcels with ghost buiings per 1,000 land registry
parcels. The color bins refer to the ve quintiles of the distribution. White areas identify towns for which27
we do not have data.
 
 
 
 
--------------------------------------- 29 

Figure 3: Ghost Buildings Registratoion Rate
Notes: The gure presents the registration rate of ghost buildings at April 30, 2011, dened as the ratio
between the number of land registry parcels with registered ghost buildings and the number of land registry
parcels with ghost buildings targeted by the program.
28
 
Ghost Buildings Registration Rates at April 2011
.15
.1
Fraction
.05
00.2.4.6.81
Ghost Buildings Registation Rate
 
 
 
--------------------------------------- 30 

Figure 4: Number of Elections by Year
Notes: The gure presents, for each calendar year, the number of elections elections occurring before or after
the beginning of the Ghost Buildings program. The gure includes elections from 1993 to 2011.
29
 
Number of Elections by Year
5,000
4,000
N 3,000
2,000
1,000
0
1993199419951996199719981999200020012002200320042005200620072008200920102011
Pre-Program ElectionPost-Program Election
 
 
 
--------------------------------------- 31

 

Figure 5: Number of Elections Pre/Post Publication
Notes: The gure presents the number of elections relative to the beginning of the Ghost Buildings program.
The gure includes elections from 1993 to 2011.
30
 
Number of Elections Pre/Post Program
8,000
6,000
4,000N
2,000
0-8-7-6-5-4-3-2-11
 
 
 
--------------------------------------- 32
 
Figure 6: Ghost Buildings Intensity Coe_cient by Election Pre/Post Treatment
Notes: The graph reports the coe_cients on the ghost building intensity for each election before and after
the beginning of the Ghost Buildings program. On the x-axis, elections are ranked based on their occurrence
relative to the program. The regression includes town and year xed e_ects. For each election rank, we
report the point estimate and the 95% condence interval. The last election before the program ( "-1") is
the omitted category.
31
 
3
2
1
0
Beta&95%C.I. on Election*Ghost Buildings
-1
-4-3-2-101
Election Pre/Post
 
 
 
--------------------------------------- 33
 
Table 1: Summary Statistics

 
 
 
 
--------------------------------------- 34

 

Table 2: The Determinants of Ghost Buildings Intensity

 
 
 
 
--------------------------------------- 35

 

Table 3: Registered Ghost Buildings Intensity 

 
 
 
 
--------------------------------------- 36

 

Table 4: Program Intensity and Incumbent Reelection: Baseline Results

 
 
 
 
--------------------------------------- 37

 

Table 5: Program Intensity and Incumbent Reelection: Robustness Checks

 
 
--------------------------------------- 38 

Table 6: Program Intensity and Reelection: Heterogeneity

 
 
 
--------------------------------------- 39

 

Table 7: Ghost Buildings Registration and Incumbent Relection

 
TAG:  CASE  FISCALE  FANTASMA  2 MILIONI  STUDIO SCIENTIFICO  NON ACCATASTATE  RESPONSO ELETTORALE  RECUPERO TASSE 

diffondi 

commenti a questo articolo 0
informazioni sull'autore
ISCRITTO A PDNETWORK DAL
25 gennaio 2010
attivita' nel PDnetwork