You are on page 1of 14

Available online at www.sciencedirect.

com

ScienceDirect

Journal of Policy Modeling 40 (2018) 1126–1139

Public corruption: A study across regions in Italy夽


Germana Corrado a,∗,1 , Fiammetta Rossetti b
a Department of Management and Law, University of Rome Tor Vergata, Via Columbia 2, 00133 Rome, Italy
b Institute for Prospective Technological Studies, JRC-European Commission, Seville, Spain

Received 23 May 2017; received in revised form 16 November 2017; accepted 10 December 2017
Available online 31 January 2018

Abstract
This work explores the determinants of public corruption using a regional panel dataset on crimes perpe-
trated in Italy by public officials against the public administration in combination with a set of demographic
and socio-economic variables. The results suggest that both the size and the composition of public spend-
ing at the local level explain corruption. We also find that regions where social capital is higher are more
likely to face a lower incidence of corruption crimes. Moreover, regions which have historically placed less
importance on rooting out corruption may be stuck in a vicious circle of higher levels of corruption.
© 2018 The Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.

JEL classifications: C2; D7; H5; R1

Keywords: Corruption; Regions; Public sector; Italy; Panel data analysis

1. Introduction

A large political economy literature has highlighted that the more sizeable is the public sector
the higher is the likelihood of detecting illicit activities carried out by (mis-)using public money.
Corruption arises from the illicit behaviour of state-appointed bureaucrats who appropriate public
funds by misleading the government about the cost and quality of public goods provision. However,

夽 The authors wish to thank Professor Dominick Salvatore and four anonymous reviewers.
∗ Corresponding author.
E-mail address: corrado@uniroma2.it (G. Corrado).
1 She gratefully acknowledges the “Consolidate the Foundations Grant” awarded by the University of Rome Tor Vergata.

https://doi.org/10.1016/j.jpolmod.2018.01.001
0161-8938/© 2018 The Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.
G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1127

from an empirical perspective the evidence of the effect of government size (i.e., the quantity and
quality of public spending) on corruption is quite mixed (see, in particular, Acemoglu & Verdier,
2000; Blackburn, Bose, & Haque, 2011; Persson & Tabellini, 2003; Mauro, 1998; Treisman,
2007).
Since Italy has a long tradition of government interventions to foster growth, to stimulate
investments and to reduce unemployment, it is crucial to study the relationship between public
spending and corruption. In fact, an important issue to address when dealing with grafts within
public administrations relates to the relative share of public sector (with respect to the private
sector) within a country. Most of all, it can be insightful to observe how public expenditures
(in specific sectors) and corrupt practices within public administration jointly evolve across geo-
graphical areas (regions) of the same country. It is indubitable that the presence of corruption
hinders talented and clean activities; the best individuals – be they firms or citizens – tend to leave
corrupt systems since their ethical and moral values prevent them from conforming to dishonesty,
while their intellectual skills are frustrated by backward and inequitable conditions. Therefore, the
threat of corruption is to create a society made of its worst individuals and entrepreneurs which is
doomed to become more and more peripheral to economic development. In fact, corrupt systems
tend to support public projects that are less labour intensive and more capital intensive, being
the latter more prone to rent-seeking (Gupta, Davoodi, & Tiongson, 2000; Gupta, Davoodi, &
Alonso-Terme, 2002). As a consequence we observe a shrinking of resources aimed to maintain
social equality and inclusion (e.g., expenditure for public schools, hospitals, welfare, etc.). A
big bulk of studies documents how criminal conduct within the public administrations reduces
the wherewithal to public education system (Ades & Di Tella, 1999; Mauro, 1998) and to the
healthcare system (Baicker & Staiger, 2005; Factor & Kang, 2015; Golden & Picci, 2005). Whilst
Treisman (2000) focuses on the causes of corruption between-countries and finds that economic
and institutional variables explain corruption less than do socio-political variables (see also Ades
& Di Tella, 1996; Lambsdorff, 2005, 2006; Goel & Nelson, 2010). In a similar way we aim at
detecting the main factors that help to explain the incidence of corruption crimes conducting the
analysis at regional level within a country: Italy. A key feature of a country like Italy relates to the
strong heterogeneity of economic traits – and of the mechanisms of social fairness connected to it
– from North to South. An insightful perspective may be gained by looking at how the structure of
public expenditures varies across the regions of the same country. Also, observing how regional
heterogeneity of socio-economic and demographic traits relates to the incidence of illicit activ-
ities within the public administration might help for a better understanding of the geographical
determinants of corruption. The key questions we want to address are:

• What are the effects of the size of local public sector expenditure on the incidence of corruption
across regions?
• How important are geographic, cultural and historic influences in determining corruption?

We therefore believe that this analysis can contribute to the ongoing debate on regional differ-
ences brought at European level on quality of government, broadly defined, such as corruption,
impartiality, and quality of public services (Charron, Dijkstra, & Lapuente, 2014, 2015). The con-
cept of the quality of the local governance, which heavily depends on the control of corruption, is
central since it is related with economic and social development, better environmental conditions,
better quality of life and also more equitable society.
We employ a panel dataset for Italian regions on crimes perpetrated by public officials against
the public administration for more than a decade (from 2000 to 2011) in combination with a set
1128 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

of economic and socio-demographic variables identified by the prevailing literature to be causes


and effects of corrupt systems. We therefore use a measure of corruption that has the advantage to
be directly observable and therefore is ‘objective’ although it suffers of several significant well-
known problems. For example, while data are reported annually by the official statistics, there
is an unknown, and most likely variable, time lag between crimes and convictions; in addition,
data give little to no indication as to the seriousness of the incidence of corruption, but they just
reveal the tip of an iceberg of a vast word of corruption and other illicit practices. In fact, our
data refer to illicit activities that involve public officials and persons in charge of public offices
who have been prosecuted and convicted for such crimes. The database employed in the present
analysis offers a variety of indicators at the regional level that are likely to reveal the latent
existence of corruption, and mirror the effects of the statistics on crimes. Regional characteristics
included in the empirical analysis are the extent of public sector, complemented by the breakdown
of government spending in: general public services, social welfare, public safety, education and
the healthcare system. Finally, the analysis encompasses the degree of economic development
reflected by regional GDP per capita, the level of education and youth unemployment2 conditions
of local communities, and the attitude towards volunteering.3
The main findings of our work confirm that the dimension of the Italian public sector matters in
explaining the incidence of corruption crimes along with the socio-economic and “environmental”
(cultural) conditions that may greatly vary across Italian regions. In particular, the results suggest
that individuals who reside in regions where corruption is higher and persistent are less likely to
be satisfied with public services and to benefit from higher levels of social capital.
The paper proceeds as follows. Section 2 gives a picture of the socio-economic and histori-
cal conditions that lie behind the different levels of corruption across Italian regions. Section 3
describes the data employed. Sections 4 and 5 deepen the empirical methodology and discuss
results. Section 6 concludes.

2. Mapping corruption across Italian regions

The battle against corruption is now explicitly declared in the policy agenda of many national
and international institutions. The Italian case is definitely a blatant example of how corruption
frustrates hopes for a balanced growth and an equitable society. Countries like Italy are particu-
larly under scrutiny since they belong to the group of developed countries, therefore they should
have been responsible of an accountable management of their internal public resources. Recent
literature has pointed out that in Italy one of the main factor that explains regional path dependen-
cies is the presence of consolidated “clientelistic networks” existing in regions with historically
unconstrained rulers (Charron & Lapuente, 2013). As highlighted by Del Monte and Papagni
(2007) “. . .For politicians the probability of being elected was much more linked to the number
of favori (favours) that they could offer their ‘clients’ than to the efficiency of public expendi-
ture and the probability of being apprehended” (p. 387). This preferential treatment enjoyed by
individuals and groups of citizens keen to show compliance with the dominant political classes
was initially fed by the possibilities offered by periods of economic boost, but then ended with
pauperising the economic resources – and hindering growth – until present days. Favourable eco-
nomic conditions and available resources could have been used to create more equitable growth
paths, instead corrupt practices along with economic crises depleted resources.

2 See Mocan (2008).


3 See Del Monte and Papagni (2007).
G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1129

The very heterogeneous histories in the North and South of Italy generated diverse socio-
economic paths, and the cultural background was always considered the heart of clear-cut
economic and political divergences between the two macro areas of the country. Italy, as a unified
republic, enjoyed an economic boom after the Second World War, along with an increasing decen-
tralization of its public system since 1948. It is within this peculiar portrait that the political and
economic dynamics of Italy take place, and in particular the floods of corruption-related crimes
between the 70s and the 90s (La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1999).
Fig. 1 maps the regional incidence of crimes4 against the public administration committed by
public officials who have been prosecuted and convicted by the Italian judiciary system. These
data have been released from the Italian National Institute of Statistics (ISTAT) for the period
2000–2011. The episodes of misconduct are classified by differentiating between the broader
class of unspecified misbehaviour and illegal activities committed by public officials who abused
of their discretionary power. We should recall that statistics about unlawful behaviours may
underestimate the true phenomenon, since they refer only to crimes reported to the police and
prosecuted (and convicted) by the Italian judiciary system. In Fig. 1, a darker shade denotes regions
with a higher average incidence of recorded crimes against the public administration during the
period under consideration: Italian Central and Southern regions along with the northern region of
Friuli Venezia Giulia report on average a higher number of crimes. An increase in the number of
recorded crimes may not necessarily be due to an increase in the number of actual crimes; rather
it might be caused by an improvement in factors such as the willingness of individuals to report
crimes, the adequacy of police to detect illicit activities, or the efficiency of judicial institutions
to prosecute crimes.
In the following section we conduct a panel data analysis across Italy’s regions. Regional
data allow to better identify how structural changes in institutions, social norms and population
demographic characteristics over time may impact on corruption within the public sector; the only
drawback relates to the fact that regional data do not allow the use of variables that are the same
for all entities (here regions) such as the legal system or the administrative rules and law.

3. Data and statistics

This study employs a dataset which provides information for as many as twenty Italian regions
over a span of eleven years 2000–2011. Data come predominantly from the Italian National Insti-
tute of Statistics (ISTAT) and Eurostat. We use the information gathered across Italian regions on
the number of crimes perpetrated by public officials against public administration relative to the
population which have been reported, prosecuted and convicted with an irreversible provision of
sentence by the judicial authority and this constitutes our measure of corruption. These data5 refer
to a broader group of illicit activities that ISTAT classifies as crimes against public administration
and thus constitute a satisfactory proxy for the diffusion of corruption within the Public Admin-
istration; in the following we refer to this variable as “Crimes against the PA”. We assume the
existence of a significant correlation between recorded crimes and the actual misbehaviour within

4 These crimes embrace to a broader group of illicit activities that involve public officials and persons in charge of

public offices who have been prosecuted and convicted for such crimes.
5 The following specifications of crimes committed by public officials (i.e. individuals in charge of public offices) have

been considered in our analysis: 1. corruption; 2. embezzlement; 3. extortion; 4. bribery; 5. abuse, misuse or breach of
public office duties. Data source: data are gathered from the Italian Judicial Register (Casellario Giudiziario) and reported
in the crime statistics by from Italian National Institute of Statistics (ISTAT).
1130 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

Fig. 1. Crimes against the Public Administration. Years 2000–2011 (averages).


Source: Crime Statistics released by the Italian National Institute of Statistics (ISTAT).

public offices, even if a high number of denounces can reflect both a large number of corruption
cases and a high propensity to report, detect, and prosecute them. Also we know that this measure
of corruption certainly represents a lower bound as the level of actual corrupt practices may be
higher than the measured one but at least we can rely on an ‘objective’ measure of corruption not
biased by subjective perceptions.
The majority of variables are rescaled by the regional population, and turned in logarithmic
terms for ease of interpretation. Table 1 reports a description of the variables included in this
Table 1
Summary statistics and description of the variables.
Variable description Obs Mean Std. Dev. Min Max Source Unit

G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139


Crimes against the PA (crimes committed by 264 −10.5 0.6 −12.4 −8.6 ISTAT Per capita; logarithmic
public officials against the public
administration and prosecuted).
Share of public sector (government 240 9.1 0.2 8.6 9.8 ISTAT Per capita; logarithmic
expenditure per capita).
General public services 264 6.6 0.3 6.0 7.7 ISTAT Per capita; logarithmic
Education expenditure 264 7.0 0.2 6.5 7.5 ISTAT Per capita; logarithmic
Public healthcare care system expenditure 264 7.4 0.2 7.0 7.8 ISTAT Per capita; logarithmic
Social security and welfare 264 5.5 0.5 4.5 6.8 ISTAT Per capita; logarithmic
Public safety expenditure 264 6.2 0.1 6.0 6.7 ISTAT Per capita; logarithmic
Volunteering (number volunteers during the 242 −2.3 0.4 −3.3 −1.4 ISTAT Per capita; logarithmic
last 12 months).
GDP per capita 264 10.1 0.3 9.5 10.5 ISTAT Per capita; logarithmic
Gini index 198 −1.3 0.1 −1.5 −1.1 ISTAT logarithmic
Tertiary education (number of students 245 −3.6 0.6 −7.1 −2.9 EUROSTAT Per capita; logarithmic
enrolled in the first stage of tertiary
education).
NEETs (young people aged 15–24 neither in 241 2.6 0.5 1.1 3.5 EUROSTAT Logarithmic
employment nor in education and
training).
Roads (number of families declaring that 242 2.8 0.4 1.5 3.6 ISTAT Logarithmic
streets in their area of residence are in
very bad conditions, per 100 families with
the same characteristics).
Trade (used as instrumental variable) 240 −17.1 1.1 −19.5 −14.4 ISTAT Per capita; logarithmic
Polluted air (used as instrumental variable) 242 2.3 0.4 1.1 3.2 ISTAT Logarithmic

1131
1132 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

study along with summary statistics and data sources. We aim at detecting whether Italian regions
witness the association between illicit practices within the public administration and different types
of government expenditure. In particular, the way public spending is framed gives indication on
whether more public resources are allocated to labour-intensive or capital-intensive public goods.
In fact, strong differences in terms of rent-seeking activities are associated to investment in public
infrastructures rather than in health and education (Mungiu-Pippidi, 2015; Rose-Ackerman, 1997).
Traditional information such as GDP per capita is also included, more precisely the GDP per
capita is included with a lag of one year to deal with potential endogeneity. Whilst among socio-
demographic and cultural characteristics, we consider the fraction of people engaged in tertiary
education, the percentage of young people who are neither employed nor in education or training
(the variable NEETs), and the intensity of volunteering. We finally include data on the quality of
roads as declared by respondents and reported in the statistics on households’ living conditions
released by ISTAT. We expect a positive correlation between the level of corruption within the
public administration and the deterioration of some infrastructures such as roads.

4. Estimation strategy and empirical evidence

We sketch the relationship of crimes against the public administration committed by public
officials with the size of the public sector and the type of government expenditure by means of
pooled OLS regressions.
Crimes against PAi,t = α + βxi,t + ui,t (1)
where the variable “Crimes against PA” denotes crimes against the public administration (PA)
committed by public officials in region i at time t, xi,t is the vector of covariates we control
for, and ui,t is the idiosyncratic error term. In the variable “Crimes against PA” the following
specifications of crimes committed by public officials are comprised: corruption, embezzlement,
extortion, bribery, and abuse, misuse or breach of public office duties.
The first column of Table 2 shows a baseline regression performed on a minimum number of
socio-economic and cultural characteristics of the region (model (a)). Other regional controls are
added in the second regression reported in the second column of Table 2 (model (b)) where we
consider distinct items of public expenditure (social security and welfare, public safety, education,
general public services, healthcare system) plus other socio-economic controls. In particular, we
add two controls (i.e., the volunteering and the quality of the roads) in order to account both for
the level of local social capital and the quality of the public service offered in the area of residence.
Overall, evidence at this stage confirms the strong effect exerted by size of public sector as
a background for illicit activities (Acemoglu & Verdier, 2000). In fact, higher levels of public
spending definitely create a background of opportunities to raise irregular profits from businesses
with the public administration thus causing corruption (see column one of Table 2).
Whilst the negative relationship with public spending in education and with individuals’ edu-
cation attainment supports the argument that higher public investment in the formation of human
capital could lead to less crime (Ben-Porath, 1967; Eicher, García-Peñalosa, & van Ypersele,
2009). This finding is also in line with main literature6 originated by Lipset (1960) according to
which education represents a way to keep citizens politically involved, and civic engagement is
a means to monitor them more closely (Putnam, Leonardi, & Nanetti, 1994; Glaeser, La Porta,

6 See also Ades and Di Tella, (1999), Golden and Picci, (2005) and Mauro (1998).
G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1133

Table 2
Crimes against the PA. Pooled OLS and Panel estimates.
Controls Pooled OLS Panel

Model (a) Model (b) Model (a) Model (b)

FE RE FE RE
(1) (2) (3) (4) (5) (6)

GDP per capita (−1) −0.648** −0.669 2.221 −1.025** 0.729 −0.578
(0.307) (0.505) (2.158) (0.466) (2.344) (0.568)
Gini Index 0.893 0.578 −0.256 0.146 −1.036 −0.607
(0.607) (0.654) (0.646) (0.597) (0.668) (0.620)
Tertiary Education −0.118 0.0089 −0.951** −0.211 −0.287 −0.149
(0.111) (0.114) (0.460) (0.205) (0.422) (0.169)
NEETs 0.368 0.419* 0.060 0.153 −0.059 0.108
(0.233) (0.240) (0.267) (0.240) (0.254) (0.225)
Share of public sector 0.703** – 0.657 0.612 – –
(0.270) (0.768) (0.453)
General public services 1.355*** 0.974 1.474***
(0.388) (1.071) (0.529)
Social security/welfare 0.147 −0.003 0.0388
(0.244) (0.309) (0.263)
Public safety expenditure 0.0176 −6.095 −1.473
(1.372) (3.921) (1.501)
Education expenditure −0.769** −0.163 −0.369
(0.368) (0.769) (0.512)
Public health care system −1.602** −1.234 −1.167
(0.714) (0.912) (0.757)
Volunteering −0.268 −0.411 −0.486*
(0.255) (0.287) (0.250)
Roads −0.0279 0.525** 0.438**
(0.175) (0.212) (0.188)
√ √ √ √ √ √
Time dummies
Observations 163 161 163 163 161 161
R-squared within 0.449 0.428 0.474 0.458
R-squared between 0.066 0.535 0.066 0.526
R-squared overall 0.515 0.506 0.034 0.492 0.037 0.494
Hausman test (Prob > χ2 ) 6.26 (0.9362) 25.15 (0.1208)

Notes: Standard errors in parentheses; significance level *** p < 0.01, ** p < 0.05, * p < 0.1; FE stands for fixed effects;
RE stands for random effects; Constants are included.

Lopez-de-Silanes, & Shleifer, 2004). Therefore, under the so-called Lipset’s hypothesis voters
with higher education levels (and income) are expected to be more willing and also more capable
to monitor public employees and to take action (punish) when the latter violate the law (Fiorino &
Galli, 2010). The negative relationship between income per capita (GDP) and corruption across
Italian regions seems to confirm again the Lipset’s hypothesis: poor regions experience more
corruption than rich ones.
As shown in Table 2 (column two), the augmented model displays a positive association
between illicit practices within public administration and the extent of unemployment of untrained
young people (the NEETs). When work opportunities are scarcer more people might engage in
informal or even illegal businesses, this may be especially the case for untrained individuals
1134 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

with lower educational attainments. Therefore individuals with a precarious income and a poorer
cultural level may more easily decide to engage in illicit activities with the public sector in order
to carry their needs out. This might lead eventually to a vicious self-sustained circle of corruption
with youth unemployment and inactivity (NEETs). “These young people, excluded from the labor
market for years, are referred to as ‘Generation Jobless’. . .Youth have little faith in their skills
or qualifications as they attempt to navigate a system they perceive to be rife with corruption”
(World Bank, 2014, p. 26, chap. 3).
These results appear to be in line with the relevant literature (Schneider & Enste, 2000), and are
further followed up by exploiting the potential of the panel database. Longitudinal data have the
advantage to follow the dynamic of the same cross sectional units (i.e. Italian regions in our case),
and to enable the inclusion of individual heterogeneity. In this study the relevant heterogeneity
relates to unobservable socio-economic and cultural factors that change over time but not across
entities. The panel data model we estimate can be specified as follows:

Crimes against PAi,t = α + βxi,t + γt + εi + ui,t (2)

where again the variable “Crimes against PA” denotes crimes against the public administration
(PA) committed by public officials, xi,t is the vector of covariates we control for, γ t is a set of year
dummies, and εi and ui,t are respectively the region specific and the idiosyncratic error terms.
The term εi represents regional traits (i.e. socio-economic traditions and cultural values) assumed
to be time-invariant (i.e. fixed effects) but heterogeneously distributed within the country. The
two models identified with pooled OLS are now tested under random effects and fixed effects
hypothesis. The former assumes that the variation across regions is random and uncorrelated with
the independent variables — in other words εi is orthogonal to xi,t . The latter relaxes the strong
orthogonality assumption by time-demeaning the data in order to purge observations from their
time invariant components which the model does not control for, but that may intertwine with
relevant – however difficult to frame – regional aspects.
Table 2 (columns 3 and 4) presents the results of the baseline model where corruption is
regressed against a set of socio-economic variables using both panel random (RE) and fixed (FE)
effects. For choosing the best fitting model we performed the Hausman test: the random effects
estimation is preferred due to higher efficiency, in fact the null hypothesis7 of those two estimators
being different is not rejected. Consequently, we are going to analyse the RE model in Table 2
(column 4). Results are overall similar with respect to pooled OLS; in particular, we can note that
the coefficient of (lagged) GDP per capita remains negative and highly significant.
Finally, we consider the augmented RE model in column six of Table 2. Results highlight that
the poor quality of local infrastructure might positively impact on the incidence of corruption
crimes. For example, ordinary maintenance and operations of physical infrastructures, such as
roads, might be intentionally neglected so that they have to be rebuilt. This might create oppor-
tunities for (corrupt) officials to extract bribes and kickbacks from new public investment in
large infrastructure projects. As highlighted by Tanzi and Davoodi (1998) in some phases of a
large civil engineering project contract, such as for road infrastructures, “a strategically placed
high-level official can manipulate the process to select a particular project. He can also tailor the
specifications of the design to favour a given enterprise.” (p.4, Tanzi & Davoodi, 1998).

7 The RE model is preferred under the null hypothesis due to higher efficiency; while under the alternative FE model

is at least consistent and thus preferred.


G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1135

A negative relationship between misbehaviours or misconduct within public administration and


people’s participation in voluntary organizations comes to be highlighted. Average participation
in volunteering can be used as a measure of local social capital which captures the existence
at regional level of differences in people’s general attitude towards corruption. In mainstream
literature social capital is thought to measure the ability to cooperate and the level of trust and
honesty in society (Putnam et al., 1994); it is then intuitively obvious that regions in which people
appear to be more honest and trustworthy are more likely to experience less corruption (Del Monte
& Papagni, 2007).

5. Controlling for endogeneity: a GMM approach

By tracking the same cross-sectional units over time panel data offer the opportunity to exploit
the temporal dimension in order to gain insights about possible causality issues among variables.
With reference to corruption and other types of illicit conduct within public offices a reasonable
question is whether it is more the composition of government expenditure that causes the mis-
behaviour by creating the opportunities for it, or it is rather the unlawfulness which alters the
composition of government expenditure by creating inefficiencies within the economic system.
To answer this question, we should account for the potential endogeneity bias (reverse causal-
ity) between corruption and public spending (and possibly other covariates). In fact, government
expenditures are liable of endogeneity issues since they can be influenced by (corrupt) public
officials who are interested in the creation of opportunities for bribery thus affecting government
spending (as shown by Mauro, 1998). In addition we include the lagged value of the variable
“Crimes against PA” as a predictor.
We employ the system generalized method-of-moments (system GMM)8 particularly suitable
for panel analysis (Arellano & Bond, 1991). This estimation method can account for (i) the fact
that the process (i.e. illicitness within public sector) may be dynamic, with current realizations of
the dependent variable influenced by past; and (ii) that some controls may not be exogenous. In
the estimated model lags of the dependent variable are supposed to alleviate the misspecification
caused by the possible omission of relevant variables. These dynamic panel data models and their
estimation overcome the severe bias caused by adding a lagged dependent variable to a panel data
model and then estimating it via the within regression or the GLS estimator.
Therefore Eq. (2) is complemented with the inclusion of the lagged values9 of Crimes within
PAi,t −1 and estimated by the system GMM10 :
Crimes against PAi,t = δ1 Crimes against PAi,t−1 + βxi,t + γt + εi + ui,t (3)
The GMM estimates a system of two equations: one differenced and one in levels. By adding
the second equation additional instruments can be obtained. Variables in levels from the second
equation are instrumented with their own first-differenced values in order to increase efficiency.
Table 3 shows that today’s level of misbehaviour within the public sector is significantly
determined by its previous realizations. The persistency across time of illicit activities within

8 In the Arellano–Bond estimator (i.e. command xtabond2 of the software Stata 13) we estimate two equations (one
differenced and one in levels) which are instrumented by first-differences only (system GMM).
9 In the Arellano–Bond system GMM estimator the first-differenced lagged dependent variable is also instrumented

with its past levels to address for the autocorrelation problem.


10 The Hansen test of over-identifying restrictions is reported in Table 3. We also perform the Arellano–Bond test for

autocorrelation that presents no evidence of model misspecification: a significant AR(1) serial correlation and lack of
significant AR(2) serial correlation.
1136 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

Table 3
Crimes against PA. GMM estimates.
CONTROLS Model (a) Model (b)
(1) (2)

Crimes against PA (−1) 0.595*** 0.577***


(0.116) (0.114)
GDP per capita (−1) 0.028 0.191
(0.299) (0.786)
Gini Index −0.443 −0.726
(0.886) (0.929)
NEETs 0.485* 0.515
(0.193) (0.397)
Tertiary education 0.030 −0.171
(0.086) (0.134)
Share of public sector −0.238 –
(0.330)
Roads 0.424*
(0.231)
General public services 1.196*
(0.617)
Social security/welfare −0.298
(0.305)
Public safety expenditure 0.885
(1.497)
Education expenditure −1.379
(1.016)
Public health care system −1.565
(0.964)
Volunteering −0.478**
(0.216)
√ √
Time dummies
Observations 145 145
Arellano–Bond test for AR(1) −3.33 (0.001) −2.83 (0.005)
Arellano–Bond test for AR(2) 1.27 (0.206) 1.31 (0.191)
Sargan test of overid. restrictions 24.35 (0.711) 42.83 (0.170)
Hansen of overid. restrictions 3.32 (1.000) 0.00 (1.000)

Notes: Standard errors in parentheses; significance level at *** p < 0.01, ** p < 0.05, * p < 0.1. Instruments: exogenous
variables (trade, polluted air) and tertiary education, volunteering, GDPt−1 , year dummies.

Italian regions might imply the existence of a “culture” of corruption that is not easy to eradicate.
As a consequence some regions are more likely “to be stuck into a vicious circle” characterized
by pervasive corruption. We then can argue that those regions that have historically placed less
importance in combating corruption, and thus have weak anti-corruption social and cultural norms
(more tolerance for corruption) might experience higher and stronger levels of unlawfulness as
time elapses. Therefore, as stressed by Aidt (2003), the political and economic institutions, and
possibly history, might be among the main determinants of local corruption. In other words, a
historical inertia of institutions that induce corruption might be persistent (Goel & Nelson, 2010).
The positive relationship between youth (and untrained) unemployment and corruption,
NEETs, is still confirmed (as shown in column one of Table 3). The positive coefficient of the
variable NEETs highlights that existing inequalities in employment and education conditions of
G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1137

the young adults (age class 15–24) are likely to strengthen local unfairness and corrupt practices
across regions.
The augmented model (in column two of Table 3) shows that there is evidence that corruption
is negatively (but weakly) associated with higher (per-capita) levels of expenditure on general
public services that represents a very large share of total public spending (as they also include
interest payments on debt). Again a negative relationship between misbehaviours or misconduct
within public administration and people’s participation in voluntary organizations is found. The
dynamic panel confirms the existence of a strong causal effect from (high) social capital to (less)
corruption. In fact, social capital might be thought to measure the level of trust and honesty in
society (Uslaner, 2001), therefore a negative correlation between corruption-related crimes and
local social capital is expected to be found: regions in which people appear to be more honest
and trustworthy ought to experience less levels of corruption.11 Finally, the estimates confirm
the presence of a positive correlation between the quality of public infrastructures, Roads, and
the incidence of corruption (see column two of Table 3) thus suggesting that a high percentage
of paved roads in poor condition may be the result as well as a cause of corrupt practices: the
deterioration of roads caused by insufficient maintenance is likely to increase the quota of public
funds spent on investments for new infrastructure projects (repairing damaged roads or building
new ones) which create a fertile ground for illicit and corrupt practices to flourish.

6. Conclusions

This study investigates misbehaviours within public administration in Italy throughout a panel
analysis on regional statistics of crimes recorded by the judiciary authority. Episodes of crimes
are examined in connection with the dimension of the public sector, government expenditures by
type, economic and social characteristics of regions. We observed that the number of corruption-
related crimes within the public administration may be associated with the size of the public sector,
poorer quality of local infrastructure (quality of roads) and unequal socio-economic conditions
(high rates of young people not in employment, education or training and lower levels of social
capital). The estimated model also suggests that the persistence of unlawfulness for those regions
that delay their firm intervention against the misbehaviours of public officials increases the risk
to remain stuck in a harmful “vicious circle” which may lead to tolerate public crimes instead of
counteracting it. Our work contributes to the ongoing policy debate which is being developed at
the European level with the objective to disentangle the determinants of local corruption. This
analysis highlights the need to gather more accurate and complete information in order to enhance
the understanding of such a subtle phenomenon. This might help to better understand the deeper
and more hidden mechanisms that may start local and that – if not unveiled – may foster illegal
practices within the national public sector.

References

Acemoglu, D., & Verdier, T. (2000). The choice between market failures and corruption. American Economic Review,
90(1), 194–211.
Ades, A., & Di Tella, R. (1999). Rents, competition, and corruption. American Economic Review, 89(4), 982–993.

11 Our results confirm the findings of Paldam and Svendsen (2002) and Bjørnskov and Paldam (2004) who report that

higher level of social capital is a significant cause of less corruption, although the direction of causality between social
capital and corruption is not clear-cut.
1138 G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139

Ades, A., & Di Tella, R. (1996). The causes and consequences of corruption: A review of recent empirical contributions.
In B. Harris-White, & G. White (Eds.), Liberalization and the new corruption (vol. 27) (pp. 6–12). Brighton: Institute
of Development Studies Bulletin.
Aidt, T. (2003). Economic analysis of corruption: A survey. The Economic Journal, 113, F633–F652.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to
employment equations. Review of Economic Studies, 58, 277–297.
Baicker, K., & Staiger, D. (2005). Fiscal shenanigans, targeted federal health care funds, and patient mortality. Quarterly
Journal of Economics, 120(1), 345–386.
Ben-Porath, Y. (1967). The production of human capital and the life-cycle of earnings. Journal of Political Economy, 75,
352–365.
Bjørnskov, C., & Paldam, M. (2004). Corruption trends. In J. Graf Lambsdorff, M. Schramm, & M. Taube (Eds.), The
new institutional economics of corruption: Norms, trust and reciprocity (pp. 59–75). London: Routledge.
Blackburn, K., Bose, N., & Haque, M. E. (2011). Expenditures bureaucratic corruption and economic development. The
Manchester School, 79(3), 405–428.
Charron, N., Dijkstra, L., & Lapuente, V. (2015). Mapping the regional divide in Europe: A measure for assessing quality
of government in 206 European regions. Social Indicators Research, 122(2), 315–346.
Charron, N., Dijkstra, L., & Lapuente, V. (2014). Regional governance matters: Quality of government within European
Union member states. Regional Studies, 48(1), 68–90.
Charron, N., & Lapuente, V. (2013). Why do some regions in Europe have a higher quality of government? The Journal
of Politics, 75(03), 567–582.
Del Monte, A., & Papagni, E. (2007). The determinants of corruption in Italy: Regional panel data analysis. European
Journal of Political Economy, 23(2), 379–396.
Eicher, T., García-Peñalosa, C., & van Ypersele, T. (2009). Education, corruption, and the distribution of income. Journal
of Economic Growth, 14(3), 205–231.
Factor, R., & Kang, M. (2015). Corruption and population health outcomes: An analysis of data from 133 countries using
structural equation modeling. International Journal of Public Health, 60(6), 633–641.
Fiorino, N., & Galli, E. (2010). An analysis of the determinants of corruption: Evidence from the Italian regions. POLIS
Working Papers No. 171. Department of Public Policy and Public Choice, Università Piemonte Orientale.
Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2004). Do institutions cause growth? Journal of Economic
Growth, 9(3), 271–303.
Goel, R. K., & Nelson, M. A. (2010). Causes of corruption: History, geography and government. Journal of Policy
Modeling, 32(4), 433–447.
Golden, M. A., & Picci, L. (2005). Proposal for a new measure of corruption, illustrated with Italian data. Economics and
Politics, 17(1), 37–75.
Gupta, S., Davoodi, H. R., & Alonso-Terme, R. (2002). Does corruption affect income inequality and poverty? Economics
of Governance, 3(1), 23–45.
Gupta, S., Davoodi, H. R., & Tiongson, E. (2000). Corruption and the provision of health care and education services. In
A. K. Jain (Ed.), The political economy of corruption (pp. 111–141). London: Routledge.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1999). The quality of government. Journal of Law,
Economics, and organization, 15(1), 222–279.
Lambsdorff, J. Graf. (2005). Between two evils-investors prefer grand corruption!. Univ. Passau, Wirtschaftswiss. Fak.
Discussion Paper V-31-05.
Lambsdorff, J. Graf. (2006). Causes and consequences of corruption: What do we know from a cross-section of countries.
In S. Rose-Ackerman (Ed.), International handbook on the economics of corruption (pp. 3–51). Cheltenham: Edward
Elgar.
Lipset, S. M. (1960). Political man: The social bases of politics. Garden City, NY: Doubleday.
Mauro, P. (1998). Corruption and the composition of government expenditure. Journal of Public Economics, 69(2),
263–279.
Mocan, N. H. (2008). What determines corruption? International evidence from microdata. Economic Inquiry, 46(49),
493–510.
Mungiu-Pippidi, A. (2015). Corruption: Good governance powers innovation. Nature, 518, 295–297.
Paldam, M., & Svendsen, G. T. (2002). Missing social capital and the transition in Eastern Europe. Journal for Institutional
Innovation, Development and Transition, 5, 21–34.
Persson, T., & Tabellini, G. (2003). The economic effects of constitution. Cambridge, MA: MIT Press.
Putnam, R. D., Leonardi, R., & Nanetti, R. Y. (1994). Making democracy work: Civic traditions in modern Italy. Princeton
University Press.
G. Corrado, F. Rossetti / Journal of Policy Modeling 40 (2018) 1126–1139 1139

Rose-Ackerman, S. (1997). The political economy of corruption. In K. A. Elliott (Ed.), Corruption and the global economy
(pp. 31–60). Washington DC: Institute for International Economics.
Schneider, F., & Enste, D. H. (2000). Shadow economies: Size, causes, and consequences. Journal of Economic Literature,
38, 77–114.
Tanzi, V., & Davoodi, H. (1998). Corruption, public investment, and growth. Tokyo, Japan: Springer.
Treisman, D. (2000). The causes of corruption: A cross-national study. Journal of Public Economics, 76, 399–457.
Treisman, D. (2007). What have we learned about the causes of corruption from ten years of cross-national empirical
research? Annual Review of Political Science, 10(1), 211–244.
Uslaner, E. M. (2001). Volunteering and social capital: How trust and religion shape civic participation in the United
States. In P. Dekker, & E. M. Uslaner (Eds.), Social capital and participation in everyday life (pp. 104–117). London:
Routledge.
World Bank. (2014). Youth inactivity and unemployment. In Breaking the barriers to youth inclusion. pp. 23–35.
Washington D.C: The World Bank Group.

You might also like