Organized crime and corruption: what are the consequences for Italian Cohesion Policy investments?

ABSTRACT The paper evaluates the effect of illegal activities on Cohesion Policy implementation. This issue is fundamental in Italy, where large delays in expenditure risk to undermine the growth-enhancing effects of funds. To explain delays, we focus on two criminal behaviours: corruption and organized crime. By exploiting a two-step approach, an empirical analysis is carried out on Italian provinces, with a focus on Southern ones, between 2007 and 2015. The findings show that both crimes, impacting the efficiency of funds, cause delays in the implementation of Cohesion Policy. These consequences are higher if linked to corruption rather than to organized crime.


INTRODUCTION
Cohesion Policy is the main EU investment policy, aiming at promoting regional imbalances reduction and long-term sustainable growth (Dall'Erba & Fang, 2017).Over the last 40 years, Cohesion Policy has allocated around €300 billion to all European regions, with its very bulk committed to the so-called convergence regions (i.e., those regions reporting gross domestic product (GDP) per capita levels below the 75% of EU GDP per capita average levels) (Crescenzi et al., 2021).The relevance of Cohesion Policy has further increased when it has been exploited as a countercyclical instrument (Great Recession, Sovereign Debt Crisis and COVID-19 pandemic shock; Boffardi et al., 2022).
During the 2007-13 programming cycle, Italian regions received almost €28 billion (Coppola et al., 2020).The expenditure of these funds was characterized by difficulties in meeting the schedule agreed with EU, in terms of projects implementation and payments (Camera dei Deputati, 2018).According to the European Commission (2016), in 2010 payments made by public administrations were still extremely low (12% of total available funds), highlighting long delays in projects implementation.In 2013, this value stood at 50%, while in 2015, more than the 80% of the resources available had been spent.By the very end of the programming cycle, total fund expenditure was almost achieved through significant efforts made by both programme authorities and ad hoc taskforces.Italian government adopted several legislative changes with the aim of speeding up payments and fostering financed projects implementation (Arbolino et al., 2020).Similarly, with reference to the 2014-20 programming cycle, in August 2022, the share of funds of the 2014-20 Cohesion Policy cycle expended (i.e., final payments made to funds recipients) stood at 55.7% of total funds allocated, while the share of committed funds was 73.2% (MEF, 2022).
Delayed expenditure or non-expenditure of funds generates two unintended consequences, among the main.On the one side, the missing-growth effect does not allow to reduce output differentials among European regions areas (Varga & Veld, 2011).On the other, by the end of each seven-year programming cycle, the European Commission de-commits the amount of non-expended funds: in this way, timely implementation of funded projects becomes central for achieving the targets stated at the beginning of each programming cycle.
The economic literature has long debated the causes of the inability to expend EU funds, identifying low managerial abilities of local administrations and socio-economic preconditions (human capital availability, production fabric structure, institutional framework, private investments, infrastructural endowment; Bachtrögler et al., 2020;Kersan Škabić & Tijanić, 2017) among the main.With reference to managerial ability, quality and efficiency of regional institutions play a key role (Arbolino & Boffardi, 2017).Indeed, local administrations enjoy a high degree of discretion in the expenditure decisions, as well as in the monitoring of funded activities.In contexts characterized by low institutional quality and, thus, managerial abilities, this is likely to ingenerate resources misuse, abuse and distortion of resources (Sergi, 2015).Among the main, scholars recognize in corruption and coercive mafia-related crimes two phenomena that might alter the progress of funded activities and initiatives (Aceto, 2013;Achim & Borlea, 2015;Capasso et al., 2022).Corruptive practices usually arise in any bureaucracy having the authority to allocate benefits and impose costs (Soreide & Rose-Ackerman, 2018).According to Haque and Kneller (2015), public expenditures are vulnerable to the influence of corruptive practices, due to the high discretion of government authorities.Since corruption distorts public spending, it generates inefficiencies in terms of both public goods and services provision and policymaking.Mauro (1995) argues that corruption reduces the efficiency of government expenditure, since resources to be spent on public services and infrastructures are partially embezzled for private gains, before reaching their destinations.According to Mungiu-Pippidi (2013), corruption poses significant barriers to the effective expenditure of EU Cohesion Funds and, subsequently, to the achievement of funds development objectives.However, there is no agreement on the potential impact of corruption on economic dynamics (Cooray & Schneider, 2018).A first hypothesis ('sand-the-wheel' effect) is that corruption reduces public policies efficiency and effectiveness by generating distortions on economic decisions and imposing additional cost on economic actors.Alternately, the 'grease-the-wheel' hypothesis claims that in contexts characterized by weak governance and ineffective policy mechanisms, corruption can speed up the implementation of policies by allowing the reduction of the timing of bureaucracy through bribery (Méon & Sekkat, 2008).
Besides corruption, a weak institutional contextcharacterized by favouritism and clientelismmight favour the infiltration of criminal organizations, thus allowing them to exert a detrimental influence of policy implementation.This kind of organization, often referred to as mafia-type associations, is a major problem, strangling not only Southern regions from a social, economic and political point of view.Gambetta (1996) maintained that mafia-type organizations are crime groups specialized in producing, promoting and selling private protection.According to Barone and Narciso (2015), organized crime infiltrations hinder public policy effectiveness by manipulating funds assignment towards, for example, specially created fictious firms.This results in the reduction of efficiency (funds are often allocated but not properly expended), long-term growth, investments and development.A vast array of studies proved that administrations colluded with criminal associations are more prone towards misuse of public funds (Sergi & South, 2016).Territories characterized by large infiltrations experience a loss of efficiency in fund expenditure, partially due to the collusion of politicians with organized crime (Sergi, 2015).According to Aceto (2013), criminal organizations put efforts into creating obstacles to the implementation of EU-funded activities, since they prefer to keep undeveloped areas in a state of underdevelopment which facilitates their control over territories and economic activities.
Given these premises, the following research studies how illegal activities (i.e., corruption and organized crime) might affect the scheduled implementation of EU Cohesion Policy by causing delays.It contributes to those strands of the literature studying economic implication of illegal activities (Becker & Klößner, 2017;Corrado & Rossetti, 2018;Capasso, 2005;Mauro, 1995) and the differentials in the effectiveness of EU Cohesion Policy (Bachtrögler et al., 2020;Crescenzi et al., 2016;Di Caro & Fratesi, 2022).However, it presents some relevant and innovative contributions.Our study differs from those dealing with the effect of EU Cohesion Policy on criminality (e.g., Banca d'Italia, 2015) or, vice versa, on the effect of criminality on funds allocation (Poprawe, 2015).First, it includes and separately analyses the influence of two illegal activities.Second, it exploits a novel set of metrics to study a particular aspect of fund governance, that is, the delays.In fact, the literature has studied Cohesion Policy, but has posed its focus on the determinants of funds allocation (Charron, 2016), funds absorption (i.e., Tosun, 2014) and policy effectiveness (Crescenzi et al., 2016), with a focus on the effect of corruption (Kersan Škabić & Tijanić, 2017).
The remainder of the paper is structured as follows.Section 2 discusses the literature on the main topics related to the research.Section 3 describes the methodology adopted.Descriptive evidence is presented in section 4. The econometric approach and the main results are presented in sections 5 and 6.A discussion and the following conclusions are presented in Section 7.

LITERATURE REVIEW
Cohesion Policy is an EU-wide policy aimed at supporting regional convergence by fostering the growth of the most disadvantaged areas and promoting long-term sustainable growth.Although there have been several attempts to simplify it, the complexity of its management structure as well as the numerosity of the programmes funded has increased over time (Crescenzi & Giua, 2020).Indeed, Cohesion Policy relies on a multilevel governance structure, with a strong role for different actors involved.On top, the European Commission is responsible for setting the overall budget, allocating funds through different programmes whose framework is negotiated with the member states.Then, national governments and regional administrations are responsible for selecting final recipients, expending funds by funding projects and monitoring their implementation and the respect of the target stated at the EU level (Mendez & Bachtler, 2022).
Given the risk of fund decommitment, the effective expenditure of fund is one of the main challenges for all stakeholders involved in the management of Cohesion Policy (Kersan Škabić & Tijanić, 2017).Overall, different expenditure capabilities have been constantly reported across EU member states (Terracciano & Graziano, 2016).The existence of these differentials has underlined the necessity of studying which factors determines them.Kersan Škabić and Tijanić (2017) found that, on a regional scale, the expenditure of Cohesion Policy-related funds is crucially determined by local macroeconomic preconditions, such as labour force employment, human capital endowment, institutional framework, investments and infrastructure development.Similarly, regional economic structure (Gagliardi & Percoco, 2017;Percoco, 2017), institutional quality (Arbolino et al., 2020), economic openness (Ederveen et al., 2006) and the typology of expenditure considered (Di Caro & Fratesi, 2022) are fundamental issues for explaining funds expenditure and effectiveness.
Starting from Coase (1960) and North (1990), the literature has widely recognized the relevance of institutions for all aspects of economic development.More in detail, it assigns a central role to institutions and governance mechanisms to achieve public programmes expenditure targets, as well as the highest gains in terms of policy effectiveness (Arbolino & Boffardi, 2017, 2021;Rodríguez-Pose & Garcilazo, 2015).
In fact, empirical evidence suggests a higher impact of Cohesion Policy in those areas characterized by higher institutional quality (Crescenzi et al., 2016).This result is underlined in contexts characterized by high differentials in terms of quality of subnational institutions, such as Italy (Albanese et al., 2021).With reference to Italy, Milio (2007) has provided evidence on the relevance of administrative capacity of local administrations in the realization of EU-funded projects.It found that the lack of fund expenditure is not only a matter of political and socio-economic factors, but mainly of administrative efficiency.Catalano et al. (2015) studies the effect of the structure of Italian bureaucracy on the implementation of Cohesion Policy by focusing on employment and social policy, finding that the high degree of specialization in the Italian bureaucratic structures negatively affects the coordination required to the different administrative stakeholders, thus hampering the realization of EU-funded initiatives.Similarly, Terracciano and Graziano (2016) identifies administrative capacity as a key factor to explain regional implementation of Cohesion Policy, defining it in terms of ability of programming, managing, monitoring and evaluating.
Within institutional quality, the control of corruption is considered a key issue for increasing the spending of funds and their absorption (Incaltarau et al., 2020).Despite the increased attention of political institutions towards the influences of criminal organizations on Structural and Cohesion Fund investments, at the date there is no large literature developing such issue.
According to Kersan Škabić and Tijanić (2017), corruption negatively affect the expenditure of EU funds on a local scale.To evaluate member states' ability of effectively investing and absorbing Structural Funds, Gruševaja and Pusch (2011) referred to three different institutional variables (corruption perception index, a regional NUTS-2 administration variable and an equalization/ growth objective policy variable).With reference to Romanian regions, Braşoveanu et al. (2011) found that corruptive behaviours negatively affect the absorption and the effective returns of such funds.The same case has been analysed by Mihailescu (2012), who claims that corruption is one of the factors inhibiting national and regional-level expenditure capability.
In Italy, the influence of organized crime in EU funds management is a major problem, where regions receiving the largest amount of funds are deeply influenced by historical criminal organizations: Mafia, Camorra, 'Ndrangheta and Sacra Corona Unita (SCU).Criminal organization infiltration in Structural Funds management is enhanced by administrators through corruption and absence of controls (Sergi & South, 2016).As demonstrated in the case of the Calabria region, clans, politicians and administrators cooperate in the use/abuse/misuse of Structural Funds (Sergi, 2015).Such convergence of interests (among clans and local political representatives) favours negligent governance of funds, which negatively influences both management and absorption of funds and, subsequently, hinders their effectiveness.The influence of organized criminality on public procurements and programmes has been pointed by EU as a risk (Barone & Narciso, 2015).
The literature surveyed can be summarized as follows: . Institutions' weaknesses facilitate corruption and criminal organization infiltration. .Criminal organization infiltrations generate loss of efficiency in the implementation of publicly funded projects. .Loss of efficiency results in delays in funds expenditure due to fund misallocation, embezzlement, lenient or absent monitoring practices. .Delays reduce the growth-enhancing effect of Cohesion Policy.
From the above discussion, the following hypothesis results: Hypothesis 0: The delay in fund expenditure is a function of illegal activities, that, in turn, are facilitated in low quality institutions.

EMPIRICAL ANALYSIS: METHODOLOGY AND VARIABLES
The literature has developed a vast array of empirical methodologies to evaluate the effectiveness of public administration in EU Structural Cohesion Funds expenditure: theory-based evaluations, macroeconomic modelling and econometric analyses, among others (Kersan Škabić & Tijanić, 2017).Economic analyses normally favour the econometric approach (Bachtler et al., 2000).Given the structure of data and our research question, we identified econometric panel data analysis as the most appropriate technique in measuring the influence of a set of independent variables on Italian provinces expenditure delays.
In order to disentangle the effects of the two phenomena, a two-step methodology is applied.Indeed, criminal behaviour in the administration of place-based investment policies is a very broad phenomenon, intertwining with a large set of local-scale features.Properly isolating the effects of these phenomena is not an easy task.For doing so, we first removed the effect of contextual factors (I step) and, subsequently, estimated the impact of the concerned illegal activities (II stage).In the second stage, that side of policy which is not affected by the determinants recognized by the literature (human capital endowment, infrastructures, etc.) is identified.With this aim, we assume that criminal phenomena find their place in this unexplained side, belonging to a hidden and 'fuzzy' logic.We applied two different and subsequent regressions (Ali & Zhang, 2015;Montagnoli & Napolitano, 2003): . The first allows one to understand ifand, in case, how macroeconomic factors explain delays.For doing so, the measure of delays is regressed on a set of variables proxying the main determinants of Cohesion Policy implementation. .The second explains the role of criminal phenomena by regressing the residuals predicted from the first equation (i.e., the new dependent variable) on the interest variables (index of corruption and index of organized crime).Following this procedure, the estimated residuals are adjusted for the effects of the most wellknown factors influencing the ability to implement projects funded via Cohesion Policy Funds.
This approach provides a further advantage.Potentially high correlation might exist among the two main regressors (index of corruption and index of organized crime), on the one side, and other independent variables, on the other.The two-step strategy contributes to solve the potential issues of multicollinearity.According to Tabachnick and Fidell (1989), the independent variables with a bivariate correlation of 0.70 should not be included in multiple regression analysis.As shown in the correlation matrix reported in Appendix C in the supplemental data online, the two indices show correlation coefficients lower than the selected thresholds.
Therefore, the first step of the analysis relies on the estimation of the following panel data models: where t ¼ 1, … , 9 refers to a year between 2007 and 2015; i is the provinces included in the sample; and DELAY describes the index evaluating progresses and delays in Cohesion Policy implementation (i.e., payments delay index -PDI) and its two sub-indices (private and public payments delay indices).
According to the literature, three groups of independent variables have been included: . Administration factors describe aspects related to both managerial abilities of local administrators and the institutional framework.These factors are widely recognized as key issues for explaining Cohesion Policy implementation.To assess the former, (1) a score assessing the educational level of the member of local governments and councils (Education) was added.Furthermore, the degree of regional autonomy was considered through a dummy variable (taking value 1 for regions with higher degrees of legislative and executive autonomy).Its effect cannot be stated a priori (Tosun, 2014): on the one side, local institutions might promote more efficient resource allocation; and, on the other, devolution of powers might be a major cause of inefficiencies due to clientelism, favouritism and corruption. .The second group includes five variables accounting for policy-specific factors, such as governance level, activation procedure and complexity.First, the complexity of projects realised is analysed by means of two variables: (1) the per capita amount of funds allocated to each province (Total Funds) to capture whether Italian local administrations are capable to cope with larger amount of funds; and (2) the average dimension of projects (Dimension).We also consider the impact of the Piano di Azione e Coesione (PAC) strategy, implemented by Italian government from 2011 onwards to accelerate fund expenditure and project implementation (Arbolino & Di Caro, 2021).Second, with reference to the activation and selection procedure, two main typologies were identified: activation through standard open calls (Calls) and negotiated procedures (Negotiated Procedures).While the former should be allowed to select the best recipient according to a set of predetermined and clearly verifiable criteria and stimulate competition among potential recipients, negotiated procedures aim at creating a direct partnership among the involved stakeholders and local administrations.The effectiveness of these activation procedures cannot be stated a priori and largely depend on local management capabilities (Crescenzi et al., 2021).Finally, to assess the impact of the governance level, two variables were included: (i) the share of projects related to regional operational programmes (POR share); and (ii) the share of projects related to funding programmes directly managed by municipalities (Municipal projects share).In the current framework of Cohesion Policy, funding programmesand the related projectscan be managed by central administrations (ministries), regions and municipalities. of industrial activities (Industry) (Giannola et al., 2016;Percoco, 2017).Second, by recognizing the negative influence of the Great Recession on the ability of implementing projects (IFEL, 2017), (2) a dummy variable (Crisis) taking value 1 for the years of the Great Recession was included.
(3) Social Capitalthat is, stronger civic engagement and pro-social behaviourwas then proxied by a variable assessing voter turnout during the last election.A positive impact is expected (Andini & Andini, 2019).Finally, (4) a variable accounting for the share of population holding at least a secondary school degree (Human Capital) was included to evaluate the impact of higher levels of education on local capacities to timely expend funds.A positive effect of stronger human capital is expected (Fratesi & Wishlade, 2017).
And 1 i,t describes the error term of the specification.A set of time-period annual effects was introduced to account for omitted variables and exogenous shocks that are likely to influence Cohesion Policy implementation (e.g., reforms of provincial powers, among others).Following these estimations, residuals (Res) were predicted and used as dependent variables of a second regression: The potential relationship between organized crime infiltration and corruption is analysed by including two variables, summarizing the results of the two indices: index of organized crime (Crime) and index of corruption (Corr).In addition, by recognizing that corruption is often exerted by organized crime together with, or instead of, violent actions and that corruption in public administrations often happens through violent channels, an interaction term was included.

Sample and timespan
Modelling was applied over nine years , which is over the programming cycle 2007-13.The choice of extending the timespan up to 2015 is justified by the regulatory framework underlying Cohesion Policy, according to whom, under the t + 2 rule, 2015 is the last year to receive payments.The analysis investigates the relation between Cohesion Policy implementation and illegal activities based on two samples: the whole set of Italian provinces and the 36 provinces of the Mezzogiorno area: Abruzzo, Molise, Campania, Puglia, Basilicata, Calabria, Sicilia and Sardegna.The peculiar focus on the Southern part of the country is required because of a set of socio-economic issues: regions and more subject to corruption (Nifo & Vecchione, 2014). .This area includes the less developed regions of the country, suffering from a historical socio-economic delay, in comparison with the rest of the country (Banca D'Italia, 2015).

DATA
Evaluating the effects of the two illegal activities on the delays in Cohesion Fund expenditure required the assessment of three different issues.In doing so, two indicators have been built, while an already existing one has been modified.First, the PDI was built to measure the occurrence of delays in the implementation of projects financed through Cohesion Policy funds.Second, based on the index of organized crime, the strength of criminal organizations at the provincial level was proxied.Finally, one of the pillars of the institutional quality index (Nifo & Vecchione, 2014) was modified to measure corruption levels.

Cohesion Policy implementation
The delay in funds expenditure was evaluated by building the PDI, while two sub-indices measure delays related to projects managed by public and private recipients, 1 respectively: public PDI and private PDI.These indices move beyond the frequently used measures of Cohesion Policy implementation, such as (1) the ratio of payments to total funding (Tosun, 2014), (2) the ratio of committed funds to total funding allocated in the region and (3) the ratio of payments to committed funds (Kersan Škabić & Tijanić, 2017).The construction of the PDI is based on the policy framework (REG.(CE)no.1083/2006) regulating the different phases of an EU-funded project.Projects are considered eligible for funding by the administration in charge based on a precise roadmap.The time required for its implementation is divided into phases.Once a precise phase is concluded, the administration makes the established payment.At the beginning of the project, 'Pre-payment accounts' anticipate between 5% and 10.5% of the total funds committed in the approval process.Another share is transferred through a final settlement.This procedure encourages administrative controls over the implementation of financed projects.A major share of payments is made following pre-established steps, only in case actuators/recipients meet the scheduled time frames.
Based on this regulatory framework, payments can be used as valuable proxies to assess project progress.Therefore, the index is conceived as a measure of the overall progress of projects financed within Cohesion Policy.From a methodological point of view, the index is built as a weighted average of payments (in euros) made by the administration.Indeed, in the reference year, the database OpenCoesione associates each project with information on (1) the financial progress of the project (indicating the current implementation status of the project) and ( 2) the total payments made by the administrations so far.Following the 'financial progress' criterion, projects might be classified into three categories: (1) 'dismissed', in case the 95% of payments have been made; (2) 'ongoing', when payments made are more than 0% and less than 95% of the resources committed; and (3) 'not started'.
To provide a consistent measure of the delays, each project is associated with a weight corresponding to the values of a standardized normal distribution: 0.000 to 'dismissed' projects, 0.475 to 'ongoing' and 0.975 to 'not started' ones.The index is constructed based on the following equation: where Paym.dismissed is payment (in euros) made for projects whose financial progress is indicated as 'dismissed'; Paym.ongoing measures payment made for projects whose financial progress is indicated as 'ongoing'; and Paym.NotStarted indicates money transfers made to notstarted projects.
The indices range between 0 and 0.975.The lowest value means that by the end of the programming cycle, all the projects have been closed and, consequently, all payments have been finalized.Higher values indicate that part of the projects had not been completed in 2015, highlighting a delay in Cohesion Policy implementation.Therefore, the index succeeds in capturing the implementation progress of each project in terms of overall progress at the end of the programming period.
The same procedure was applied to three kinds of projects, that is, (1) total sample of projects, (2) projects managed by public authorities and (3) projects managed by private entities.In doing so, we calculated three: PDI, public PDI and private PDI.
Index construction was based on data on almost 800,000 co-financed projects from the governmental database OpenCoesione.We limited the analysis to the funds related to the regional operative programmes (ROP), excluding the national operative programmes (NOP) and to those projects whose starting date was not reported on the OpenCoesione database.NOPs are the main funding channel for major infrastructural projects, whose realization times usually overlap a single programming cycle.Therefore, not accounting for NOP allows one to exclude projects whose 'ongoing-ness' would have affected the results of the index.

Organized crime
The index of organized crime is built based on the data provided by the Italian Justice Department (Ministero della Giustizia) on the number of proceedings against known and unknown subjects pending before the District Anti-Mafia Directorates (Direzione Distrettuale Antimafia -DDA).DDAs have an exclusive jurisdiction on proceedings relating to mafia-type crimes (Jamieson, 1998).In more detail, DDAs exert a leading role on investigations of both crimes listed in Article 51.3 bis of the Italian Code of Criminal Procedure and those others committed by members of criminal organizations (Lavorgna & Sergi, 2014).Therefore, since all cases allocated to DDAs must be previously recognized as mafiarelated ones (Jamieson, 1998), in our opinion, these data represent a good proxy for the concerned phenomenon.We acknowledge that due to 'Omertà' and fear of retaliations, data on the crimes reported to authorities might underestimate the real levels of infiltrations.Table 1 reports criminal offences for which proceedings have been registered at the DDAs ('Offences linked to organised crime').

Corruption
The 'corruption' pillar of the institutional quality index was chosen to assess the level of corruption in provincial public institutions (Nifo & Vecchione, 2014).It is constructed by aggregating three different indicators: the Golden-Picci index (Golden & Picci, 2005), the number of crimes against the public administration and the number of municipalities overruled by the national authorities.The original values range between 0 (the most corrupted area) and 1 (the less corrupted one), thus providing a 'control of corruption' measure.Instead of using the original variable, it was modified so that a higher value of the new variable means a higher level of corruption (hence, 1 now represents provinces with the most corrupted administrations).This is an innocuous transformation not affecting empirical analyses; it was done in order to both increase consistency with the other variables (whose higher values indicate 'more') and eliminate a possible source of confusion in the reading of the results of the estimations.

Descriptive evidence 4.4.1. Cohesion Policy implementation
Based on data provided by the governmental database OpenCoesione, the PDI and the two sub-indices (public and private PDI) were calculated.Table B1 in Appendix B in the supplemental data online reports the averages of PDI, public PDI, private PDI, index of corruption and index of organized crime.
Southern areas report the highest delays in all typologies of projects analysed, with an average score above 0.11, while the rest of the country is characterized by lower ones.Lombardy was the most virtuous region, together with the Liguria, Piemonte and Marche regions.On the opposite, major delays refer to projects implemented in Southern Italian regions, achieving the lowest overall scores (PDI): Calabria, Sicilia and Campania.
Similar results are provided by subsequent elaborations based on provincial (NUTS-3) data.Figure 1 summarizes evidence related to the three PDIs: that the most efficient provinces are the Northern ones and that delays increase in Southern provinces, with few exceptions (i.e., Marche provinces).For detailed results on all provinces, see Table B2 in Appendix B in the supplemental data online.
Indeed, the top five positions in all three PDI rankings are taken by Northern provinces.In more detail, five provinces of Lombardy report the most reduced delays, with almost all the scheduled payments made.By contrast, the lowest scores of all three rankings are almost all recorded in Campania and Sicilia provinces.

Organized crime and corruption
Data on the main macro-areas of the country (see Table B1 in Appendix B in the supplemental data online) show that the incidence of the selected crimes is clearly higher in the Southern Italy than in other areas.These data shed light on the geography of organized crime in Italy, with the main criminal organizations rooted in Southern Italy, namely in Calabria ('Ndrangheta), Apulia (SCU), Campania (Camorra) and Sicily (Mafia).
The 2007-15 averaged index of organized crime at the provincial level supports this evidence (see Table B2 in Appendix B in the supplemental data online).Figure 2 summarizes these results: darker shades indicate higher values of the index of organized crime.As can be seen, Southern Italian provinces are almost all characterized by the darkest shade, apart for a few exceptions.Reggio di Calabria province (Calabria) is the area of the country where the power of criminal organizations ('Ndrangheta clans) is exerted in the most violent and manifest way, as showed by the high score achieved (0.978).The following four provinces in the top-five positions are Siracusa, Enna, Caltanissetta (all Sicily) and Benevento (Campania).At the opposite, provinces with the lowest incidence of organized crime belong to a North-Western region: Piedmont (Verbano-Cusio-Ossola, Asti, Alessandria, Novara and Vercelli).
Finally, data on corruption confirm the gap among Centre-North and South, the latter being the most corrupted area of the country, followed by the two insular regions.This evidence is supported by the provincial ranking (see Appendix B in the supplemental data online and Figure 2), according to whom the lowest corruption levels are achieved by Central (Perugia and Siena) and Northern (Vicenza, Piacenza and Verona) provinces.

RESULTS
Table 2 reports estimates computed through equation ( 1) by using the feasible generalized least squares (GLS) estimator, which is considered an appropriate choice when regional and time-fixed effects are present and heteroskedasticity might be a concern (Hsiao, 2014).
We estimated the model on the whole set of Italian provinces (models A1-A3) and, subsequently, to the sample of 36 Southern provinces (models B1-B3).Models A1 and B1 adopt the PDI as a dependent variable; models A2 and B2 the private PDI; and models A3 and B3 the public PDI.
Signs associated with the covariates are mostly as expected.Higher educational levels of public administrators correspond to a significant reduction of delays in the public sector (model A3), while no significant impact is clear in the private sector (model A2).
The group of policy-specific variables provide fundamental evidence to explain implementation delays.Indeed, the complexity of local-level policy (proxied by the per capita amount of funds transferred to provinces, and the average dimension of funded activities) significantly increases delays in both sectors and in both areas analysed.At the opposite, the PAC strategy did not prove to be effective on a national scale, while it supported significant reductions of private project delays in Southern Italian regions.This result is in line with the targets stated by the policymakers when the policy was designed, that was to provide countercyclical instruments to support local economies through a reallocation of funds, the reduction of the bureaucratic burden and speeding-up payments (Arbolino & Di Caro, 2021).Indeed, most of the regions showing expenditure delays were in Southern Italy, thus these areas have been the main recipients of this policy interventions.Moreover, from the second phase of the PAC (2012 onwards), most of the actions have been targeted to enterprises and private activities (i.e., fund assignment for supporting digital transition, employment, human capital formation, firm competitiveness and innovation), thus allowing to explain the major significant effect in the private sector (Boffo & Gagliardi, 2014).
In addition, the activation procedure adopted played a relevant role.Resorting to open calls has allowed to significantly reduce delays in Cohesion Policy implementation, rather than adopting negotiated procedure.Therefore, the use of predetermined criteria for selecting beneficiaries has allowed one to select more capable funds recipients, thanks to the increase in competition among the applicants (Crescenzi et al., 2021).Finally, governance level was not significant for explaining delays.
As shown by the main literature on the topic, territorial macroeconomic and social dynamics are determinant in explaining funds expenditure capacities.In more detail, a stronger industrial structure allows one to reduce delays in projects implementation, as well as in the expenditure of funds, in the private sector.
Heterogeneous effects are associated with the variable describing the effects of the Great Recession on projects delays.On the one side, in the context of the crisis, projects managed by private beneficiaries in Southern Italy have reported significantly increased delays.This is consequence of the economic slowdown deriving from the crisis, whose impact is not significant when the whole set of Italian provinces is analysed.This underlines the weakness of the Southern Italian entrepreneurial environment (characterized by lower levels of productivity and efficiency) in comparison with the rest of the country (Rungi & Biancalani, 2019).In a similar context, private investments are more exposed to economic turmoil, as Great Recession was, and thus more likely to be delayed.
At the opposite side, significant reductions of delays can be identified in publicly managed projectsand associated payments.As already, during the Great Recession, the Italian government and regional administrations were active in accelerating the expenditure of funds to support countercyclical measures, among others  (Arbolino & Di Caro, 2021).These actions contribute to explain the acceleration of expenditure on behalf of public recipients.
Finally, in Southern Italy, where the lowest level of institutional quality is reported, together with the highest degree of organized crime infiltrations, we detect a (partially) significant role for social capital in reducing delays.Indeed, in similar context, a strong social capital provides support, know-how and connections required to face external challenges (such as the implementation of funded projects), as well as it is a signal of pro-civic engagement on behalf of population.Evidence on the relevance of social capital in the policy networks underlying Cohesion Policy implementation have already been provided by the literature (Jordana et al., 2012).
The second step of the analysis regresses the predicted residuals from the previous three estimates on the variables of interest: index of corruption, index of organized crime and an interaction term between the two.Table 3 shows the results for both Italy (models B1-B3) and Southern Italy (models B4-B6).
As expected, in all three models, both variables of interest exert an effect on the delays in Cohesion Policy projects.However, differently from the case of Southern Italy, when the whole set of Italian provinces is considered (model B1), no significant effect of the intimidatory and violent power of organized crime on the implementation of total projects results.
Indeed, over the last decades, Italian mafia-types organizations have expanded their activities beyond their traditional territories towards Northern Italy (Dagnes et al., 2020).However, when establishing in non-traditional areas, clans prefer to exploit political corruption rather than violence, which, by contrast, is majorly exerted in those areas where clans are traditionally rooted (Alfano et al., 2019).Our results on the whole set of Italian provinces might be interpreted in this sense.Delays in Cohesion Policy implementation on a national scale are significantly determined by inefficiencies deriving from the corruption channels, which is also preferred by criminal organization, rather than by the control on territories through violence.
By contrast, when Southern areas of the country are considered, both variables of interest exert a significant effect on the delays in Cohesion Policy projects.In Southern Italy, the magnitude of the impact of illegal activities is sensibly higher than those estimated in models C1-C3, referring to the whole sample of provinces.In the case of projects managed by private entities in Southern Italy (model C5), a strong second-order effect exists.This effect is proved by the significance of the coefficient associated with the interaction between the variables Crime and Corruption.Indeed, in a province with an average level of organized crime (0.12), the effect of Corruption on Cohesion Policy delay is 0.08 ¼ (0.010) + (0.01*0.12).This result proves that corruption constitutes a non-secondary channel for criminal organization.Although EUprojects procurement procedures are strictly controlled to avoid misuse or criminal use of funds, subsequent expenditures are not always transparent and easy to control.Indeed, as also stated by Aceto (2013), criminal organizations are interested in creating obstacles to the implementation of projects aimed at strengthening local entrepreneurship and, in general, local development, since this allows them to maintain their territorial control.Overall, private activities (usually small, medium and private enterprises) have less force to resist bribery requests from public agents or violent intimidations from criminal organization.Therefore, when private agents are the aim of illegal behaviours, they suffer greater damage and, thus, their ability to implement projects reduces to a greater extent.Organized crime and corruption: what are the consequences for Italian Cohesion Policy investments? 187 REGIONAL STUDIES

ROBUSTNESS CHECKS
Further analyses were performed to test the robustness of the results.In these models (Table 4), although correlation analysis shows that there is no issue related to multicollinearity (see Appendix C in the supplemental data online), we used a composite indicatorsummarizing information on the two phenomenaas an independent variable.It was built to overcome potential problems linked to interconnections between the two interdependent variables of interest (i.e., the index of corruption and index of organized crime).The model includes all the covariates listed in equation ( 1).Results of this model are summarized in Table 4: no relevant changes are observed in comparison with the results in Tables 2 and 3. Subsequently, by recognizing potential risks of endogeneity, we adopted an instrumental variables (IV) approach, 2 which is considered suitable when (1) institutions are characterized by high persistence across time, as in Italy (Arbolino et al., 2020), and (2) in presence of interaction terms in the model (Wooldridge, 2010).The economic theoretical and empirical literature recognizes that when dealing with the analysis of institutions, issues of endogeneity are likely to arise (Efendic et al., 2011).In our analysis, endogeneity might arise for two main causes: 188 Roberta Arbolino and Raffaele Boffardi

REGIONAL STUDIES
. Measurement errors, mainly related to the so called 'inspection-detection' problem: enforcers turn a blind eye, nothing happens and the number of detected corruption cases becomes insignificant.In this way, measures of crime resorting to objective data (official statistics, number of reported crimes, etc.) are likely to underestimate the concerned phenomena and those places with the greatest corruption might be at the bottom of the scale (Daniele & Marani, 2011).Furthermore, in the case of organized crime, the risk of underestimation is heightened by two problems: the 'hidden mafia' and omertà (Pinotti, 2015). .Reverse causality, since the relation between expenditure delays and illegal activities might be endogenously determined.First, organized crime could arise, for example, where funds arrive in large amount, as new opportunities for (il)licit business arise.Second, EU funds go to poor provinces, in general more plagued by corruption and criminal organization infiltrations.This can happen because of a need for funds or because of a strong connection/corruption of entrepreneurs and politicians.
We adopt an 'historical approach' for instrumenting both corruption and organised crime indices.We hold that historical dominations have left a permanent effect on the institutional setting (either formal or informal), which in turn affects the performance of regional economic actors (Arbolino et al., 2020;Di Liberto & Sideri, 2015).Therefore, the main foreign domination of Italian provinces between 1100 and 1800 by the Normans, Anjou, the Spanish and Austrians, and Savoy have been used as instruments for institutional issue, that is, corruption.Second, following Buonanno and Pazzona (2014) and Scognamiglio (2018), we used data on the number of convicted bosses who had been 'exiled' in a different province from that in which they usually live in the period between 1961 and 1972.The ratio of forcedly exiled to the total population (per 100,000 inhabitants) provides a source of exogenous variation since it is linked to the spread of mafia-type organizations outside their traditional regions (IPAC, 1976), but it has no conceptual link with Cohesion Policy implementation (which was first designed and enforced in 1986).
The IV estimation proceeds as follows.In the first stage, the two independent variables of interest are regressed on the described covariates.Table 5 reports the results of the first-stage regressions, second-stage coefficients for the variables of interest along with the main IV post-estimation diagnostics.The post-estimation diagnostics provide evidence about the goodness of our models. 3The results of the panel data model estimations (Table 3) are mostly supported when the analyses are carried on by exploiting IV approach.

DISCUSSION AND CONCLUSIONS
The economic literature has widely demonstrated the link existing between institutional systems and illegal activities, on the one hand, and the implementation of public policies, on the other (Barone & Narciso, 2015).
In the case of Cohesion Policy, much attention has been paid to the influence that the quality of institutions exerts on the level of fund expenditure (Kersan Škabić & Tijanić, 2017).This paper approached the study of regional policymaker spending capabilities from a different point of view, focusing on the delays in EU funds expenditure.The Cohesion Policy framework provides for the possibility that, at the end of each programming cycle, the amount of funds that has not been expended by recipients is de-committed.The Italian case is quite representative of this issue, considering that during past Cohesion programmes, a large amount of funds was risked being 'lost' due to delays in projects implementation.By the very end of the programme, funds were almost completely expended (IFEL, 2017), thanks to a large set of legislative interventions allowing a speed-up in payments.However, the presence of delays is recognized as a significant determinant of the loss of growth effects of EU-funded activities (Varga & Veld, 2011).Therefore, reducing delays stands as a policy priority.
Following the literature, we acknowledge that a poor institutional context promotes the emergence of illegal activities, which, in the Italian context, might be traced back to two main phenomena: infiltrations by organized crime and corruptive behaviour.The latter strongly characterizes the Italian context: in fact, the largest part of Italian citizens (97% of the total according to the 2013 Eurobarometer) thinks that corruption is a widespread phenomenon in Italy (the EU average is 76%).According to nine Italians out of 10, corruption is the easiest way to access specific public services (the EU average is seven citizens out of 10).
These illegal activities represent a direct cost for beneficiaries of EU funds putting efforts in the implementation of an EU-funded project.On the one side, a corrupt administrator might require bribes or other forms of payments to accept a project for funding or grant the required permissions for implementing it.On the other, violent actions against the funds recipient or their properties require the diversion of funds which might be used for 'productive' activities.Furthermore, there is a wide set of indirect impacts of illegal activities on publicly funded investment.For example, rent-seeking public servants are likely to be more prone to accept bribes and favour the access to funding of projects rather than others potentially more deserving for funding.Similarly, in the presence of corruption or under the threat of mafia retaliation, monitoring of the progresses in project implementation might be more lenient or absent, thus favouring inefficiencies.Moreover, mafia-type organizations create obstacles to the expenditure of funds in order to avoid the development of local economies.
The first step of the analysis involved the construction of a set of indices describing the delays in the implementation of projects funded through EU funds.They do not assess the level of public expenditure, but the timing of project implementation and the levels of payments made by the public administration to the recipient subjects.Descriptive evidence underlines the presence of a strong imbalance between the different areas of the country.Even though significant delays have characterized all Italian regions, Northern regions have been more able to meet the planned deadlines in comparison with the Mezzogiorno (which was the main target of EU development policy) (MEF, 2022).Similarly, the strength of criminal phenomena in the Southern and insular areas of the country is evident.
Based on a panel data model, our analysis unveils that both criminal behaviours are key determinants for EU funds expenditure delays across Italian provinces.In particular, in Southern areas, the implementation time of all typologies of EU-funded activities are negatively and significantly affected by the two phenomena.By contrast, Northern areas report a lower impact of corruptive practices on delays (reduced magnitude of coefficients) and the absence of significant organized crime-associated coefficients.This result follows the lead of the literature maintaining that Northern Italian regions are usually characterized by less corrupt and more civically minded administrations, which are less prone to organized crime and violence, rather than Southern ones (Del Monte & Papagni, 2007;Mocetti & Orlando, 2019;Putnam et al., 1994).
Our study underlines the pervasiveness of criminal behaviours in the public and private spheres of the Italian economic and institutional system, especially in the South where traditionally criminal organizations are rooted.Apart from suggesting a decisive action against corruption and criminal organization, the main policy implications deriving from these results underline that the monitoring systems imposed during the 2007-13 programming period have not been effective in the Italian context.
Other relevant policy implications might be drawn from our study.First, the adoption of the PAC strategy has been fundamental to reduce delays in Southern Italian provinces (i.e., those experiencing the largest delays).Therefore, this experience might provide useful suggestions to policymakers, in view of the new challenges posed by the expenditure of funds related to both the 2021-27 programming cycle and National Plan for Recovery and Resilience.In addition, even activation procedures might be reconsidered by increasing the share of projects activated through open calls, rather than through negotiated procedures.

Figure 1 .
Figure 1.Payments delay index (PDI), public and private payments delay indices.

Figure 2 .
Figure 2. Index of organized crime (a) and index of corruption (b).
Organized crime and corruption: what are the consequences for Italian Cohesion Policy investments? 185

Table 1 .
Crimes selected to describe organized crime and build the index.

Table 2 .
Impact of governance, policy-specific and macroeconomic factors on Cohesion Policy delays.

Table 3 .
Impact of the selected crimes on Cohesion Policy delays.
Note: Results include a constant term, regional and time annual effects.Errors are shown in parentheses).*Significance at 10%, **at 5% and ***at 1%.PDI, payments delay index.
Note: Results include a constant term, regional and time annual effects.Errors are shown in parentheses.*Significance at 10%, **at 5% and ***at 1%.PDI, payments delay index.

Table 5 .
Two-stage least squares.Organized crime and corruption: what are the consequences for Italian Cohesion Policy investments? 189 Note: Results include a constant term, regional and time annual effects.Errors are shown in parentheses.*Significance at 10%, **at 5% and ***at 1%.PDI, payments delay index.