ISSN: 2038-7296 POLIS Working Papers [Online]

Dipartimento di Politiche Pubbliche e Scelte Collettive – POLIS Department of Public Policy and Public Choice – POLIS

Working paper n. 171

September 2010

An analysis of the determinants of : Evidence from the Italian regions

Nadia Fiorino and Emma Galli

UNIVERSITA’ DEL PIEMONTE ORIENTALE “Amedeo Avogadro” ALESSANDRIA

Periodico mensile on-line "POLIS Working Papers" - Iscrizione n.591 del 12/05/2006 - Tribunale di Alessandria An Analysis of the Determinants of Corruption: Evidence from the Italian Regions

Nadia Fiorino∗ Emma Galli Università degli Studi L’Aquila and Università degli Studi di Roma “La CREI Sapienza” and CREI

Abstract

This paper investigates the causes of corruption in the Italian regions for the period 1980 to 2002 by selecting a number of hypotheses assessed in the literature. Corruption turns out to be driven by the level of per capita income and of education. While, as expected, income is negatively related with corruption, education is not; its positive impact on corruption can be explained as if corruption in was typically a ‘white collars’ phenomenon during the 1980s and 1990s. Moreover the size of public investments in both the economic and politico- institutional dimensions seems to be a crucial factor in the explanation of the phenomenon.

Keywords: Corruption; Determinants; Institutions; Italian regions JEL classification: O1; H0

1. Introduction

 Corresponding Author, Nadia Fiorino, Dipartimento di Sistemi e Istituzioni per l’Economia, Università degli Studi dell’Aquila, Via , 25 - 67100 Coppito-L’Aquila, ITALY. E-mail: [email protected]. We would like to thank Luisanna Onnis, Fabio Padovano, Simona Scabrosetti, and the participants to the 2008 Società Italiana di Economia Pubblica Conference and to the 2009 European Public Choice Society Meeting. The usual caveat applies. The literature on the historical, cultural, economic and political determinants of corruption is quite developed: countries characterized by common law systems that embody greater protection of property against the state (La Porta et al., 1999), less hierarchical religions like Protestantism (La Porta et al., 1997), higher levels of income and education (Lipset, 1960), less government expenditures and regulations (Glaeser and Shleifer, 2003), lower diversity along ethnic and income lines (Mauro, 1995; Alesina et al., 2002) and higher degrees of civicness of the population (Putnam, 1993) are expected to be less corrupt. Mixed predictions have been provided by the literature about the impact on corruption of federalism (Weingast,

1995; Breton, 1996; Tanzi, 1995) and of electoral systems (Persson and Tabellini, 1999;

Lijphart, 1999).

The empirical research however faces major difficulties in measuring corruption rates.

Recently economists and political scientists have used indexes of ‘perceived’ corruption constructed by business risk analysts and polling organizations based on survey responses of businessmen and local residents. Among others, Mauro (1995), La Porta et al. (1997, 1999),

Ades and Di Tella (1999) and Persson et al. (2003) used those indexes to investigate the causes of corruption in cross-national analyses.

The Corruption Perceptions Index produced by Transparency International places Italy in the same group of several Eastern European, Asian and South-American countries with a score of about 5.2 calculated in a range between 10 (highly clean) and 0 (highly corrupt). Indeed since its unification in 1861, the Italian political system has been characterized by a strong interaction between politicians, bureaucrats and citizens aimed to reap high benefits through special laws or through political appointments (the so-called clientelismo). In the 1980s yet the increasing practice of bribing politicians and bureaucrats associated with the ongoing fiscal crisis of the public finances began to modify the general attitude towards corruption and to spread greater intolerance towards it in the public opinion. Del Monte and Papagni (2007) show that corruption crimes increased steadily between the mid-1970s and the first half of the 1990s, especially in the Southern regions, and slightly decreased after 1993 as a consequence of the so- called (Clean Hands) campaign undertaken by the judiciary system. However, the distribution of corruption crimes is not homogeneous across the territorial areas of the country.

Our paper is an attempt to take account of this differentiation by selecting a number of hypotheses assessed in the literature about the causes of corruption and applying them to the

Italian regions. In order to comparatively the explanatory power of these hypotheses we built a set of measures of the level of corruption and of the economic and politico-institutional determinants on regional basis for the period 1980 to 2002.

The paper is organized as follows: in Section 2 we present the leading theories of what causes corruption. The empirical framework is analyzed in Section 3 where variables and data are described. Section 4 discusses the results. Section 5 offers some concluding remarks.

2. The determinants of corruption: a review of the literature

It is difficult to fully capture the economic and social significance of corruption as well as its causes and consequences. The economic literature tries to investigate why public officials misuse their office for private gain more frequently in some countries than in others by identifying economic as well as social, cultural and political variables that correlate with the incidence of corruption. If we define corruption as a crime made by public officials for personal gain (as in Rose-Ackerman, 1975) the economic theory of corruption follows closely the economic theory of crime (Becker, 1968). The potential criminal, in this case a public official, weighs the benefits of crime against its costs. The various economic theories on the determinants of corruption are thus driven by this cost-benefit calculus of the potential criminal.

National or local characteristics, however, will influence the level of corruption as they alter the benefits and costs of crime.

The most relevant cost come from the risk of being caught and punished that depends in turn on the effectiveness of the country’s legal system and on the degree of protection it provides to private individuals harmed by corrupt acts of public officials1.

The risk of being caught may be higher in more democratic systems where competitors for office have an incentive to disclose and publicize the incumbent’s misuse of office during the

1 On the impact of different legal systems – common law versus civil law – on corruption see, among others, La Porta et al., 1999 elections. This hypothesis is based on an extensive literature originated by Lipset (1960) that considers political attention a luxury good, education a way to lead individuals towards having a higher value of staying politically involved, civic engagement a channel towards closer monitoring (Putnam, 1993; Glaeser et al., 2004). Therefore, voters with higher income and education are expected to be both more willing and capable to monitor public employees and to take action when the latter violate the law.

The literature considers also that the effectiveness of a legal system is rooted not only in the formulations of the laws but also in the ‘legal culture’, that is the expectations and practices that inform the way they are enforced. Different conceptions of the social role of law may imply dissimilar perceptions of the gravity of corruption. Besides losing their jobs, corrupt officials face a social stigma that highly depends on the prevailing moral norms and cultural expectations. In other words, the more dissonance occurs between the law and those norms, the lower legitimacy the legal system will tend to have and the higher will be then the degree of corruption (Treisman, 2000).

The benefits of corruption come from public agents being able to allocate resources to private individuals, including the right to bypass certain regulations. Then the official uses his power to generate some gains for the private individual beyond those he would receive without state interventions. State measures such as regulations, taxation, public economic activities may be used to give the private agent advantages wit respect to others in the market and in return the private agent pays the official part of the rent. Therefore, the larger the public sector and the greater the number of regulations of the economy, the greater will be the opportunities for helping private actors either to benefit or to evade these regulations, increasing the possibilities for corrupt practices (Tanzi, 1994; Glaeser and Shleifer, 2003; Adsera et al., 2003). Countries characterized by more regulations will experience higher levels of corruption as public officials and private individuals will have more resources to steal and more rules to exploit.

Another set of theories on the determinants of corruption has focused on the effect of ethnic fragmentation on corruption and wasteful redistribution (Mauro, 1995; Fearon and Laitin,

1996; Alesina et al., 2002). If an area is worn out by ethnic divisions and leaders tend to allocate resources towards groups of their own ethnicity, then members of one ethnic group might continue to support a leader of their own group, even if he is known to be corrupt. Furthermore, other forms of division, such as income inequality, may also reduce voters’ desire to oppose corruption since they will focus on the redistribution rather than on the honesty of government officials. An increase in income inequality would positively affect the degree of corruption.

More recently, Persson and Tabellini (1999) and Persson et al. (2003) suggest the existence of a systematic link between corruption and electoral rules. Larger voting districts are associated with less corruption, whereas larger shares of candidates elected from party lists with more corruption. Majoritarian systems are more accountable because voters seek consensus among individuals rather than among parties, which should restrict rent extraction. Then majoritarian elections are more effective in deterring political rents since the outcome of an election is generally more sensitive to the incumbent’s performance. By developing a different line of reasoning Lijphart (1999) argues that proportional systems are more accountable than majoritarian because they lead to policy decisions closer to voters’ preferences, which discourages corrupt practices.

Differences among countries in the extent of corruption may also depend on the degree to which officials compete against each other to sell mutually substitutable benefits to private agents (Shleifer and Vishny, 1993). Rose-Ackerman (1978) suggested that the existence of competition at the level of the officials receiving bribes reduces corruption. When a bureaucrat provides a scarce service, the existence of competing officials to reapply in case of being asked a bribe will reduce the equilibrium amount of corruption. Put it differently, when officials compete to sell the same or substitutable benefits to private actors, the price they can charge is driven down zero. This logic motivates two alternative arguments about the relationship between federalism and corruption: 1) federal systems may induce more honest and efficient governments because of the competition between jurisdictions (Weingast, 1995; Breton, 1996);

2) Corruption may be higher at local level because the relationships between private agents and public officials are closer and more frequent and because the potential briber has to influence only a segment of the government (Tanzi, 1995). To our knowledge there is only one empirical contribution to the analysis of the determinants of corruption in the Italian Regions by Del Monte and Papagni (2006) for the period 1963-2001. Their estimates show that economic variables (government consumption and level of economic development) as well as political and cultural factors (party fragmentation, presence of voluntary organizations and absenteeism at national elections) affect corruption in

Italy. The major shortcoming of Del Monte and Papagni (2007) lies in the lack of a complete regional data set for all the variables they use as determinants of corruption for the overall period. To face this severe problem they resort to a mix of national and regional measures. In order to propose a more accurate interpretation of the phenomenon at local level we decided to restrict our sample period and provide a full set of regional variables and data that measure the determinants of corruption.

3. The empirical determinants of corruption

In this Section we select the hypotheses developed in the literature about the causes of cross-national variation in corruption that seem to be more suitable to explain the phenomenon in the Italian Regions2. We take into account that many factors are likely to be correlated, causing problems of causal relationship and endogeneity; at the same time some variables may capture more than only one underlying hypothesis causing difficulties in the interpretation of the results. All these problems need to be controlled for in the empirical analysis by including in the regression all factors simultaneously (given that our data set contain enough variation to distinguish clearly among them) and by using the Instrumental Variable methodology.

3.1 Variables and data description

The dependent variable. Real abuse of political and administrative office can show up both in corruption and, more generally, in misgovernance. As Tanzi (1998) observes, it is difficult to define corruption in abstract, also because corruption is illegal and consequently hidden.

Cultural and legal differences across countries make it hard to investigate corruption without

2 The Italian Constitution, promulgated in 1948, foresees the principle of decentralization of the government functions and the establishment of Regional Governments (Article 5 and Title V of the Constitution). Italy has thus been divided in 20 Regions. Five of them, the first to be established between 1948 and 1963, enjoy a special statute (Regioni a Statuto Speciale, or RSS), because of their multilingual status, borderline position or particularly low level of development. The remaining 15 Regions characterized by an ‘ordinary statute’ (Regioni a Statuto Ordinario, or RSO) were established in 1970, 22 years after the Constitutional provision. taking country-specific features into account. Good proxies for should thus offer reliable information on the unlawful abuse of political power, as well as a strong level of comparability across different countries.

Following Rose-Ackerman (1975) and Glaeser and Saks (2006), we define corruption as crimes by public officials for personal gain; then in a first step we measure corruption as the number of regional government officials prosecuted for corrupt practices relative to the population over the 1980-2002 period. These prosecution levels capture the extent to which prosecutors charge and accuse public officials for misconduct in each of the twenty regions. The crimes that we consider are based on the Libro II, titolo II (crimes against the Public

Administration) of the Italian Criminal Law as reported in the Annals of Judicial Statistics of the ISTAT (various issues). We believe that these data on prosecutions are more reliable than those from Casellario Giudiziale on convicted corruption crimes because they are able to capture part of the hidden corruption shelved by corrupt judges.

Generally, cross-countries studies on corruption rely on opinion surveys resulting in the

Corruption Perceptions Index (CPI) produced by Transparency International3 or similar measures4. This index is constructed from a number of individual surveys of businessmen or local populations of the relevant countries, as well as from several ratings compiled by staff of economic risk analysis firms on the basis of reports from country experts. While measures like the CPI contain valuable information, they suffer of a few shortcomings: first, the meaning of corruption is subjective and can vary greatly from one country to the other one; furthermore, the types of the corrupt activities could be substantially different in each country making comparative analyses even more difficult. In addition, they are computed on country’s basis and this a fortiori supports our choice of an alternative measure of corruption.

Using prosecution rates as a measure for corruption bumps against the circumstance that in corrupt countries the judicial system is itself corrupt and fewer people will be charged with corrupt practices. We consider this problem of the existence of a general corruption spread

3 This index is computed for 54 countries from 1980 to 2003 and is calculated as the simple average of a number of different surveys assessing each country’s performance in a given year. 4 For a general discussion on these indexes see Persson, Tabellini and Trebbi (2003); for a comparison of different corruption measures for Italy see Del Monte and Papagni (2007). through the whole system, even though in Italy the judicial system is centralized and then relatively more isolated from local corruption. Although the legal system is the same in all the

Italian regions, its degree of legitimacy is not. Even if he does not specifically deal with corruption, Putnam (1993) shows that the different degrees of trust, civicness and legal culture have an impact on the effectiveness of regional governments in Italy.

Moreover, such a measure reflects only the ‘revealed’ corruption, most likely by leaving part of the phenomenon hidden. A first look of table 1a highlights this problem. The table shows the average per capita prosecutions for 1980-2002 per 100,000 inhabitants. As expected the

Northern regions are less corrupt than the Center and Sourthern regions; however, the ranking is not completely in line with the ‘common opinion’ about the different extent of corruption in the

Italian regions. As matter of fact, prosecutions for corrupt practices in the Corte d’Appello district of Reggio Calabria5 in the last twenty years resulted in two convictions only.

Nevertheless, similar conditions characterize other districts of as well as districts of other ‘perceived’ corrupt regions, like , and (Davigo and Mannozzi,

2007). To take into account the hidden corruption and avoid potential bias between official statistics and ‘true’ data, we consider the existing link between corruption and associative crimes (crimes ex art. 416 and 416 bis of the Italian Criminal Law). This implies that, as the so- called Mani Pulite criminal trials confirmed, corruption emerges not only as corrupt practices but also as associative crimes in the most ‘perceived’ corrupt regions. Then we construct a composite index annually computed per each region as per capita prosecutions multiplied for per capita associative crimes. Table 1b ranks this index from the least to the most corrupt regions. This ranking lines up reasonably well with our pre-conceived notions about the real distribution of corruption in Italy.

5 Reggio Calabria is one of the major town of Calabria. Table 1a - Regions with most and least prosecutions per capita (1980-2002)

Average annual prosecutions per 100,000 Pop. Emilia Romagna 34 Lombardia 35 38 39 Piemonte 42 45 Puglia 48 49 Toscana 49 Campania 53 59 60 Sardegna 60 Sicilia 61 Calabria 64 76 89 96

Table 1b - Regions with most and least prosecutions and associative crimes per capita (1980-2002) Average annual prosecutions per 100,000 Pop. Emilia Romagna 34 Lombardia 35 Veneto 38 Marche 39 Piemonte 42 Umbria 45 Toscana 49 Abruzzo 59 Friuli Venezia Giulia 60 Sardegna 60 Molise 89 Lazio 96 Puglia 97 Basilicata 98 Liguria 153 Campania 159 Calabria 192 Sicilia 244 The explanatory variables. To test the Lipset hypothesis we measure the region’s economic development and level of education by using respectively regional GDP per capita and secondary school enrolment ratio for male and female population.

Social and cultural factors may be different in the various regions even though the institutions as well as the legal system are the same. Direct democracy institutions signal the degree of civicness of the population6 and can be also interpreted as a measure of people’s attitude against the practices of corruption. This allows us to control for the degree of corruption generally ‘accepted and tolerated’ by the system as a whole, since in corrupt regions the judicial system may be itself corrupt and fewer people may be charged with corrupt practices. As a measure of direct democracy we use voters’ participation to referenda and expect a negative sign of the coefficient of this variable.

Another set of theories on the determinants of corruption has focused on the presence of the state into the economy. More regulations and greater government size create a potential for corruption since more resources may be stolen and more rules exploited or subverted (Tanzi,

1994; Glaeser and Shleifer, 2003). We test this hypothesis by using the capital public expenditure, the number of regional laws (or, alternatively, the number of articles of regional laws) and the number of legislators7.

Since Italian regions are overall ethnically homogeneous8, but show relevant income differentials between the North and the South we also test the hypothesis that as voters become more diverse along the income line, they will focus on the redistribution rather than on the honesty of government officials (Mauro, 1995; Alesina et al., 2002). We expect that an increase in income inequality will positively affect the degree of corruption. We measure income inequality with the Gini index calculated using data on family income per regions.

6 Though he does not specifically deal with corruption, Putnam (1993) shows that the effectiveness of regional governments in Italy is lower where the degree of trust and civicness are lower.

7 On the basis of the law passed in 1968 (n. 108) the number of regional legislators was 30 in regions with less than one million inhabitants; 40 for regions with more than one million inhabitants; 50 for regions with more than three million inhabitants; 60 for regions with more than four million inhabitants; and, finally 80 for regions with more than six million inhabitants. 8 The only exceptions in this respect is represented by the German minority in Trentino-Alto-Adige and French minority in Valle d’Aosta that are both excluded from our estimates. Political fragmentation as a result of the electoral rules (Persson, Tabellini and Trebbi,

2003) can be measured by using the Herfindhal index for concentration. The index is built on the seats of the majority supporting the regional government with respect to the overall legislature and ranges from 0 (a legislature in which each legislator belongs to a different party) to 1 (when all members belong to the same party). We then expect a positive impact of this measure on corruption. The use of this variable is also suggested by a recent change of the regional electoral system occurred in 1995. The mechanism by which the members of the regional Council are elected switched from a pure proportional representation to a mixed one. A top-up number of seats for the winning coalition is also introduced, so that the absolute majority of the legislators will be held by the coalition linked to the regional list that has obtained the relative majority of the votes. Furthermore, the law reduced the tenure length of the Council from five to two years if the relationship of confidence between the Council and the Cabinet breaks down during the first two years. This reform was completed in 1999 when it was established that the President of the regional Cabinet is elected by universal and direct suffrage.

Decentralized systems have been considered alternatively either as a potential source of corruption or as a source of honesty and efficiency of sub-national governments. Since mid-

1990s and especially with the approbation of federalist reform in 2001 the Italian regions were characterized by a gradual, though slow, process of reduction of grants received by the central government associated to an increase of fiscal autonomy. We capture the evolution of the decentralization process by using two different variables: 1) a dummy that takes the value of 1 from 1995 on and 0 before; 2) the share of per capita local taxes over total revenues.

Data sources. Income per capita is the ratio of the regional gross domestic product in real terms

(1995 base = 100) over the regional population. Data on regional GDP and population are taken from Crenos (2004). Data on the regional secondary school enrollment for male and female population come from Crenos (2004). Data on the regional participation to referendums and on the total number of voters are taken from the Ministero dell’Interno (various years). Data on regional legislative output come from the website of the Camera dei Deputati della Repubblica italiana. Data on Regional election’s results, used to measure the Herfindhal index of political concentration, are from Istat (1990) and Ministero dell’Interno (various years). We build the

Gini index using micro-data on the households’ disposable income coming from the Survey of

Household Income and Wealth (SHIW), conducted by the (1980; 1986; 1989,

1991, 1993, 1995, 1998; 2000; 2002). Table 2 – Descriptive statistics

Share of Corruption Income Income per Share school Number of Public Direct voluntary rate inequality capita attainment Legislators Fragmentation laws investments Local taxes democracy organizations Mean 1.86E-08 0.330655 0.013845 0.047398 53.08556 0.670597 51.68182 0.000370 0.000142 0.610032 0.000149 Median 5.52E-09 0.329437 0.013724 0.047153 50.00000 0.686400 46.00000 0.000219 2.40E-05 0.632643 0.000114 Maximum 5.52E-07 0.479429 0.023014 0.065018 90.00000 0.880000 165.0000 0.001794 0.001435 1.303661 0.000759 Minimum 0.000000 0.236252 0.000800 0.019830 30.00000 0.128000 8.000000 3.34E-05 8.27E-10 0.200333 1.44E-05 Std. Dev. 6.27E-08 0.036212 0.004192 0.007202 17.70924 0.133744 25.61925 0.000321 0.000252 0.212849 0.000123 Skewness 6.648611 0.434022 -0.277167 -0.259954 0.503150 -1.226576 1.295616 1.504923 2.406279 0.091105 2.072315 Kurtosis 49.10177 3.809933 3.283196 3.961819 2.373921 5.691410 5.370545 4.608223 8.536013 3.045466 8.817946

Jarque-Bera 35875.79 21.96461 6.038315 18.62831 21.88859 206.6606 192.2042 181.4765 838.5104 0.549592 795.1622 Probability 0.000000 0.000017 0.048842 0.000090 0.000018 0.000000 0.000000 0.000000 0.000000 0.759727 0.000000

Observations 374 374 374 374 374 374 374 374 374 374 374 Cross 18 18 18 18 18 18 18 18 18 18 18 sections Table 2 reports the summary statistics for the above variables respectively.

3. Empirical results

In Section 3 we have presented several hypotheses about the causes of corruption, each deriving from a theoretical literature in political science or economics. Since income and education are likely to be correlated, however, it would risk omitted variable bias to test hypotheses individually without also controlling for alternative hypotheses. Education and income might be related to other unobservable factors that are the true explanation for variation in corruption across locations. Second, income and education are likely to be endogenous: whether or not they cause corruption, corruption may cause them. If a low level of economic development is conducive to corruption, corruption itself is known to impede development. Long standing corruption might induce capital to flee and reduce the quality of schools, which would produce a negative relationship between education and corruption. Thus we address these problems byusing the Instrumental Variables methods or two-stage least squares. This, however, requires the identification of suitable instruments. We then regress:

Corruption Rate = a Income + b Education + Other Controls

Another empirical concern is the position of the Lazio in the rank we present in table 1.

Lazio is the region where the Ministers, and more generally the highest bureaucratic power as well as the national legislative branch, are located. Hence in Lazio both the regional and the central bureaucracies coexist. Since the Italian criminal law identifies the locus of corruption crimes as the place where the exchange of money occurs, phenomena of ‘export’ of corrupt practices in Lazio may emerge. To take account of this possibility we perform our estimation with and without Lazio.

Our sample includes the Italian regions with the exception of Valle d’Aosta and

Trentino Alto Adige for which the data set is incomplete. Table 3 shows the OLS results.

Column (1) presents the results of a regression that includes per capita income. As expected, income is significantly negative. Furthermore, we also include in the regression direct democracy institutions, public capital expenditure, income inequality and degree of decentralization. Public investments are in line with theoretical predictions. Participation to referenda as well as income inequality does not seem to have an impact on corruption. The decentralization of spending functions and consequently of the power to issue licences, to select project to finance, to expand office increase the opportunities for corruption. Most likely for local policy makers the probability of being elected was related more to the number of favours they could offer to their ‘clients’ than to the efficiency in the provision of public policies.

In column (2) we replace the income variable with the share of school attainment in the region. The significance of education is strong and much more robust than those of the income variable. Yet, in contrast with theoretical predictions, the share of school attainment exhibits a positive sign. The rest of the results are similar to those presented in column (1). Column (3) includes both education and income and confirms the previous findings. In column (4) we estimate the model analyzed in column (3) by excluding Lazio to control for possible phenomena of ‘export’ of corruption previously underlined. As the results shows, the general picture does not change. All the regressions presented in table 3 display an adjusted R 2 that explains about the 42% of the variability of the corruption measure in each equation. The negative correlation between education levels and corruption is generally interpreted in the empirical work as an indication that the middle-class opposes corruption more because corrupt practices are more likely to benefit the lower class (Schlesinger and Meier, 2002 among others).

2 The positive sign of education variable in our regressions seems to suggest though that corruption in the Italian regions concerns more educated classes. This is clearly shown in the so- called Mani Pulite investigation that signed the passing from a system of political patronage to a system of corruption that involved legislators, bureaucrats and businessmen. In other words our results seem to suggests that corruption of the 1980s and 1990s was typically a ‘white collars’ phenomenon.

Table 3 – OLS results

Dependent variable: Corruption (1) (2) (3) (4) Constant 2.80E-08 -9.79E-08 -6.17E-08 -5.24E-08 (0.476) (0.011) (0.152) (0.249) Income per capita -1.95E-06 -2.74E-06 -3.50E-06 (0.092) (0.075) (0.047) Share school 2.16E-06 2.26E-06 2.28E-06 attainment (0.000) (0.000) (0.000) Dum95 2.84E-08 3.43E-08 3.87E-08 4.02E-08 (0.002) (0.000) (0.000) (0.000) Direct democracy -1.67E-08 9.46E-09 -3.78 -7.28E-09 (0.459) (0.651) (0.864) (0.755) Public investments 3.18E-05 2.35E-05 3.01E-05 3.21E-05 (0.019) (0.065) (0.023) (0.020) Income inequality 1.84E-08 -3.47E-08 -3.18 -2.86E-08 (0.825) (0.673) (0.699) (0.741) Obs. 414 414 414 391 Adj.R2 0.42 0.41 0.42 0.42 F 13.17 14.55 14.13 14.04 (0.000) (0.000) (0.000) (0.000) Notes: p-value in parenthesis. Column (4) reports the estimates of the regression in column (3) that excludes Lazio.

As the OLS estimates may be biased we apply the Instrumental Variables method to deal with the reverse causality between education, income and corruption. Table

4 replicates the estimations performed in table (3). Column (1) reports the results of the equation including per capita income. We use as instrument the investments in industry. In column (2) we show the result of the equation that excludes income and includes the share of educated population. The equation has been instrumented using the lagged values of

3 education. Finally, column (3) presents the results of the full equation by using both the instruments. Column (4) replicates the estimates of column

(3) excluding Lazio. The Sargan’s over-identification test maintains that the instruments used in the second stage equations are exogenous. Either in OLS or in IV estimates we include regional fixed effects to avoid bias from omitted regional specific variables (as for example the difference in the attitude towards corruption practices in each region) and correct them for heteroskedasticity. Taken as a whole, estimates via the IV method do not significantly differ from the OLS results.

Table 4 – TSLS results

Dependent variable: Corruption (1) (2) (3) (4) Constant 1.60E-06 -5.96E-08 3.42E-08 6.73E-08 (0.004) (0.414) (0.687) (0.468) Income per capita -6.95E-05 -7.63E-06 -9.71E-06 (0.002) (0.017) (0.007) Share school attainment 2.02E-06 2.42E-06 2.46E-06 (0.003) (0.000) (0.000) Dum95 -5.94E-08 3.68E-08 4.14E-08 4.02E-08 (0.765) (0.076) (0.017) (0.027) Direct democracy -9.30E-07 -8.35E-09 -6.93E-08 -9.51E-08 (0.052) (0.902) (0.213) (0.115) Public investments 0.000224 2.16E-05 3.40E-05 3.70E-05 (0.027) (0.282) (0.10) (0.090) Income inequality -3.62E-07 -1.03E-07 -3.00E-08 -7.13E-09 (0.701) (0.602) (0.873) (0.972) Obs. 414 324 342 374 Adj. R2 - 3.74 0.49 0.45 -0.28 Notes: p-value in parenthesis. The instruments used in column (1), (2), (3) and (4) are the investments in industry and the lagged values respectively for income per capita and school attainment. The over- identification test is Sargan’s statistic. Columns (4) reports the estimates of the full regression that excludes Lazio.

In Table 5 we turn to deal more specifically with the link between regulation and corruption. While in the specification of the model presented in Table 3 and 4 we proxy regulation by using public investments, in further specifications we include different measures of the size of government, specifically the output of the legislative activity and the size of

4 legislatures (McCormick and Tollison, 1981; Weingast et al., 1981), to capture respectively the procedural and the politico-institutional dimension of the regulation. The coefficient on the number of regional laws does not turn out significant whereas the number of legislators as expression of interest groups’ activity has a positive a significant impact on corruption.

A further step concerns the role played in the empirical analysis by the dummy variable that identifies in 1995 the beginning of a federalist process in Italy. As we mentioned above in that year it occurred a change in the electoral system of the regional legislatures from a proportional to a mixed one. In Section 2 we emphasize that both the literature on the nexus between federalism and corruption, on one side and the literature on the nexus between electoral rules and corruption, on the other, provide conflicting predictions about the direction of the relation. In our estimates the dummy shows a positive and significant coefficient, i.e. a higher degree of decentralization together with a change towards a majoritarian system generate more corruption. In order to separate more clearly these two institutional events and to take into account their time and spatial dimension we use the share of local taxes over total revenues as a measure of the degree of decentralization and the political fragmentation index as a proxy of the electoral rules. It turns out that the level of corruption is not affected by the degree of fiscal autonomy probably because the latter is quite uniform in the Italian regions during our sample period. It is rather the political fragmentation of the regional governments that increases corruption through its impact on spending decisions. Indeed public investments are multicollinear with political fragmentation. When we include in the regression the latter, as for legislators a positive relation with corruption is established (see Column 5 in Table 5 and column 3 in Table 6).

Overall these results (that are confirmed in the IV estimations presented in Table 6) suggest that corruption in the Italian regions is driven by the level of per capita income, i.e. poor regions are more corrupt than rich ones and by the level of education, showing that corruption

5 was typically a ‘white collars’ phenomenon. Generally the Lipset hypothesis suggests a combined interpretation of the negative impact of income and education on corruption. Our result on education is only apparently in contrast with the negative relationship between economic development and corruption because of two reasons: 1) in Italy education is largely provided by the public sector and therefore is basically unrelated to income; and 2) Southern regions, that are the most corrupt, are characterized by lower levels of income associated with higher levels of education because typically education represents a substitute for job opportunities in these economies that are structurally weak. Finally, the size of government corruption creates scope for corruption; the political channel affects corruption through its impact on public investments while the amount of regulation does not.

Table 5 – OLS results

Dependent variable: Corruption (1) (2) (3) (4) (5) (6) Income per capita -1.65E-06 -1.17E-06 -6.90E-08 (0.027) (0.01830) (0.09288) Income inequality -4.39E-08 -4.18E-08 1.88E-09 (0.586) (0.5376) (0.9750) Share school 1.88E-06 1.91E-06 1.34E-06 attainment (0.000) (0.0016) (0.0163) Legislators 6.78E-09 8.28E-09 6.88E-09 8.35E-09 1.00E-08 1.04E-08 (0.000) (0.000) (0.0179) (0.006) (0.0210) (0.0163) Number of laws -1.15E-10 -5.16E-11 -1.29E-10 -6.88E-11 -1.43E-10 -1.19E-10 (0.43) (0.725) (0.1701) (0.444) (0.1927) (0.2713) Fragmentation 4.74E-08 4.29E-08 (0.0122) (0.0166) Local taxes -0.011247 -0.00729 (0.5996) (0.7351) Public 2.87E-05 2.88E-05 investments (0.030) (0.029) Dum95 3.20E-08 2.06E-08 3.12E-08 2.05E-08 (0.000) (0.0147) (0.000) (0.0005) Direct democracy 8.60E-09 2.52E-09 7.02E-09 -1.91E-09 -1.50E-08 -1.62E-08 (0.699) (0.904) (0.7148) (0.913) (0.4353) (0.2736) Obs. 414 414 414 414 374 374 Adj.R2 0.44 0.42 0.44 0.42 0.46 0.45 F 50.59 82.55 57.681 107.44 48.537 82.170 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Notes: p-value in parenthesis.

6 Table 6 – TSLS results

Dependent variable: Corruption (1) (2) (3) C -1.71E-07 -3.47E-07 -2.09E-07 (0.1118) (0.2711) (0.0313) Income per capita -6.62E-06 -1.70E-05 -1.56E-06 (0.0257) (0.0635) (0.02410) Income inequality -3.97E-08 -2.50E-07 2.94E-08 (0.5161) (0.1174) (0.3103) Share school 1.33E-06 3.51E-06 4.95E-07 attainment (0.0025) (0.0035) (0.0284) Legislators 3.29E-09 9.17E-09 3.58E-09 (0.0864) (0.1001) (0.0602) Number of laws -1.46E-10 -4.39E-10 -7.25E-11 (0.1969) (0.0924) (0.1466) Fragmentation 5.25E-08 (0.0138) Local taxes 6.64E-06 (0.3974) Public 3.29E-05 investments (0.0672) Dum95 6.18E-08 1.15E-07 (0.0022) (0.0288) Direct democracy 4.85E-08 1.76E-08 -1.03E-08 (0.1684) (0.8356) (0.5662) Obs. 396 396 374 Adj. R2 -0.103182 0.104643 0.223172 Notes: p-value in parenthesis. The instruments used the investments in industry and the lagged values respectively for income per capita and school attainment. The over-identification test is Sargan’s statistic. The estimates of the full regression that excludes Lazio do not show substantial differences.

5. Concluding remarks

Our paper is an attempt to investigate the differentiation in the distribution of corruption crimes across the Italian territorial areas by selecting a number of hypotheses assessed in the literature about the causes of the phenomenon. In order to comparatively test the explanatory

7 power of these hypotheses we built a set of measures of corruption and of economic and politico-institutional determinants on regional basis for the period 1980 to 2002.

Corruption turns out to be driven by the level of per capita income and of education.

While, as expected, income is negatively related with corruption, education is not; its positive impact on corruption can be explained as if increases the ability of the private and public actors to bypass and evade regulations. The interpretation of this result corroborates what it was clearly revealed in the Mani Pulite investigation, namely that corruption in Italy was typically a ‘white collars’ phenomenon during the 1980s and 1990s.

Moreover the size of public investments in both the economic and politico-institutional dimensions seems to be a crucial factor in the explanation of the phenomenon suggesting that the design of anti-corruption policies should pass through a process of enhancement of the quality of the public expenditures as well as of the political decision making process.

8 References

Adsera, A., Boix, C., Payne, M. (2003). Are you being served? Political accountability and quality of government. Journal of Law, Economics and Organization, 19: 445-490. Alesina, A., Baqir, R., Easterly, W. (2002). Redistributive public employment. Journal of Urban Economics 48, 219–241. Becker, G. (1968). Crime and punishment: an economic approach. Journal of Political Economy 76. Breton, A. (1996). Competitive Governments: An Economic Theory of Politics and Public Finance. Cambridge University Press, Cambridge. Davigo, P., Mannozzi, G. (2007). La Corruzione in Italia. Percezione sociale e controllo penale. Bari, Editori Laterza. 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. Fearon, J.D., Laitin, D.D. (1996). Explaining interethnic cooperation. American Political Science Review 90, 4: 715-35. Geddes, B. (1997). The political uses of corruption and privatization. Paper prepared for presentation at the annual meeting of the APSA, Washington DC. Glaeser, E., Saks, R. E. (2006) Corruption in America. Journal of Public Economics. 90: 1053- 1072. Glaeser, E., Shleifer, A. (2003). The rise of the regulatory state. Journal of Economic Literature 41 (2), 401–425. Glaeser, E., La Porta, R., Lopes-de-Silanes, F., Shleifer, A., (2004). Do Institutions Cause Growth? Journal of Economic Growth 9 (3), 271– 303. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W. (1997). Trust in large organizations. American Economic Association Papers and Proceedings 87 (2): 333-8. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W. (1999). The quality of government. Journal of Economics, Law and Organization 15, 1: 222-279. Lipset, S. (1960). Political Man: The Social Bases of Politics. Doubleday, Garden City, NY. Lijphart, A. (1999). Patterns of democracy, government forms and performance in thirty six countries. Yale University Press. New Haven Mauro, P. (1995). Corruption and growth. Quarterly Journal of Economics 110, 681–712 (August).

9 McCormick, R.E., and Tollison, R.D. (1981). Politicians, Legislation, and the Economy. Boston: Martinus Nijhoff. Persson, T., Tabellini, G., Trebbi, F. (2003). Electoral rules and corruption. Journal of European Economic Association, 1: 958-989. Persson, T., Tabellini, G. (1999). The Size and Scope of Government: Comparative Politics with Rational Politicians, 1998 Alfred Marshall Lecture. European Economic Review, 43: 699–735. Putnam R. (1993). Making Democracy Work: Civic Traditions in Modern Italy, Princeton, N.J.: Princeton University Press. Rose-Ackerman, S. (1975). The economics of corruption. Journal of Public Economics 4, 187– 203 (February). Rosen, L. (1978). Law and social change in the new nations. Comparative Studies in Society and History, 20: 3-28. Shleifer, A., Vishny, R.W., (1993). Corruption. Quarterly Journal of Economics 108: 599-617. Schlesinger, T., Meier, K. J. (2002). Variations in corruption among the American states. In: Heidenheimer, Arnold J., Johnston, Michael (Eds.), Political Corruption: Concepts and

Contexts. Transaction Publishers, New Brunswick. Tanzi, V. (1994). Corruption, governmental activities, and markets. IMF Working Paper, IMF, Washington DC. Tanzi, V. (1995). Fiscal federalism and decentralization: a review of some efficiency and macroeconomic aspects. in: World Bank, Annual World Bank Conference on Development Economics. World Bank, Washington, DC. Tanzi, V. (1998). Corruption Around the World: Causes, Consequences, Scope, and Cures. IMF Staff Papers , IMF, Washington DC, 45 (4): 559–594. Treisman, D. (2000). The causes of corruption: A cross-national study. Journal of Public Economics.76: 399-457. Weingast, B.R., Shepsle, K.A., and Johansen, C. (1981). The political economy of benefits and

costs: a neoclassical approach to distributive politics. Journal of Political Economy 89:

642-664.

Weingast, B.R. (1995). The economic role of political institutions: market-preserving federalism and economic development. Journal of Law, Economics, and rganization 11 , 1:1-31.

10

Recent working papers

The complete list of working papers is can be found at http://polis.unipmn.it/pubbl

*Economics Series **Political Theory Series ε Al.Ex Series TTerritories Series tTransitions Series Q Quaderni CIVIS

2010 n.171* Nadia Fiorino and Emma Galli: An analysis of the determinants of corruption: Evidence from the Italian regions

2010 n.170* Jacopo Costa and Roberto Ricciuti: State capacity, manufacturing and civil conflict

2010 n.169* Giovanni B. Ramello: Copyright & endogenous market structure: A glimpse from the journal-publishing market

2010 n.168* Mario Ferrero: The cult of martyrs

2010 n.167* Cinzia Di Novi: The indirect effect of fine particulate matter on health through individuals' life-style

2010 n.166* Donatella Porrini and Giovanni B. Ramello: Class action and financial markets: Insights from law and economics

2010 n.165** Corrado Malandrino: Il pensiero di Roberto Michels sull'oligarchia, la classe politica e il capo carismatico. Dal Corso di sociologia politica (1927) ai Nuovi studi sulla classe politica (1936)

2010 n.164ε Matteo Migheli: Gender at work: Productivity and incentives

2010 n.163Q Gian-Luigi Bulsei and Noemi Podestà (Eds): Imprese differenti. Le organizzazioni cooperative tra crisi economica e nuovo welfare

2010 n.162* Claudia Cusinello and Franco Amisano: Analysis for the implementation of a sustainable transport model in the eastern county of Alessandria, Italy

2010 n.161* Roberto Ricciuti: Accumulazione del capitale e crescita economica tra Italia liberale e regime fascista

2010 n.160* Carla Marchese and Giovanni B. Ramello: In the beginning was the Word. Now is the Copyright

2010 n.159ε Peter Lewisch, Stefania Ottone and Ferruccio Ponzano: Free-riding on altruistic punishment? An experimental comparison of third-party-punishment in a stand- alone and in an in-group environment 2009 n.158* Rongili Biswas, Carla Marchese and Fabio Privileggi: Tax evasion in a principal-agent model with self-protection

2009 n.157* Alessandro Lanteri and Stefania Ottone: Economia ed etica negli esperimenti

2009 n.156* Cinzia Di Novi: Sample selection correction in panel data models when selectivity is due to two sources

2009 n.155* Michela Martinoia: European integration, labour market dynamics and migration flows

2009 n.154* Massimo Pasquariello and GianMarco Chiesi: Valore aggiunto e tipologia di spesa in Piemonte. Un confronto tra Alessandria e gli altri capoluoghi di provincia

2009 n.153* Massimo Pasquariello: Produttività, sistemi locali del lavoro, specializzazione produttiva e scenari futuri. Studio preliminare in preparazione del Piano strategico per il di Alessandria

2009 n.152* Massimo Pasquariello and GianMarco Chiesi: L’analisi dei determinanti di un’area territoriale. Studio preliminare in preparazione del Piano strategico per il comune di Alessandria

2009 n.151* Cristina Elisa Orso: Formal and informal sectors: Interactions between moneylenders and traditional banks in the rural Indian credit market

2009 n.150* Michele Giuranno: The logic of party coalitions with political activism and public financing

2009 n.149* Matteo Migheli: Sharing the pie: the Lutheran is neither opportunistic nor generous

2009 n.148* Amedeo Fossati and Marcello Montefiori: Migrants and mafia as global public goods

2009 n.147* Alberto Cassone: L'impatto economico dell'Universita' del Piemonte Orientale “Amedeo Avogadro”: un aggiornamento al 2008

2009 n.146* Daron Acemoglu, Davide Ticchi and Andrea Vindigni: Persistence of civil wars

2009 n.145* Daniele Bondonio: Impact identification strategies for evaluating business incentive programs

2009 n.144* Barry Eichengreen, Rachita Gullapalli and Ugo Panizza: Capital account liberalization, financial development and industry growth: a synthetic view