An assessment of the waste effects of corruption on infrastructure provision

The estimates of the cost of corruption are rather elusive ranging from 3 % of GDP (Kaufmann et al. 2006) to more than 5 % of global GDP (World Economic Forum 2012). These estimate, however, do not take into account the negative long-term costs deriving by the decreasing trust for institutions, the lower competitiveness and the reduction of attractiveness for foreign investments, or the impact on income inequality. As reported by OECD (2013), the World Bank estimates that each year 20–40 % of official development assistance is misplaced from public budgets.

Among the three airport institutional arrangements—government-branch; port authority and airport authority—corruption has the strongest impact on the productivity of airport authorities.

For more details on parametric and nonparametric approaches to frontier efficiency estimation, see among the others: Cooper et al. (2007), Fried et al. (2008).

In a cross-sectional setting, Jondrow et al. (1982) derived an estimation of one-sided residuals, interpreted as inefficiency scores, which permitted the estimation of inefficiency for individual units. More recently, the assumption of time-invariant efficiency has been relaxed by developing appropriate panel data models. For an extensive survey of parametric methods, see Kumbhakar and Lovell (2000), Greene (2008).

An alternative approach would be to include as many environmental variables as inputs when estimating the efficiency frontier (Banker and Morey 1986).

Moreover, Simar and Wilson (2007) show that estimating [2] with Tobit leads to the violation of the assumption of the independence between \(\varepsilon _{i}\) and \(z_{i}\).

The Italian procurement system, during the last 20 years, has been subjected to continuous changes ‘to improve the efficiency, the effectiveness, the transparency and the quality of public works’, e.g. the so-called legge Merloni (Law 11/2/94, no. 109, art. 1). More recently, new rules have been devised, and a law was passed (Codice dei contratti pubblici di lavori, servizi, forniture) which transposes the EU Directive no. 2004/18/CE on the coordination of procedures for the award of public works, public supply and public service contracts into the Italian legislation.

Italy is at the 69th position in the world ranking, with a score of 43 out 100.

Engineering estimated costs are used as reserve price in tendering procedures.

The database has been built on data obtained by the authors upon official request to Osservatorio per i lavori Pubblici, AVCP. For a detailed description of AVCP data, see De Carolis and Palumbo (2011).

Differently by Guccio et al. (2012a) we perform our DEA estimates on the whole sample without distinguishing between the new and maintenance public works and between different levels of reserve price. The results do not largely differ from those of Guccio et al. (2012a).

Their method consists in smoothing the probability distribution of the efficiency estimates under the CRS and VRS formulations of the DEA model. A bootstrap re-sampling method (\(B\) = 500) is implemented to develop a robust \(p\) value, which enables us to test whether public work contracts operate under CRS or VRS.

The results of DEA VRS estimates are not reported here but are available from the authors upon request.

To compute the bootstrapped DEA estimates, we use FEAR package for R software (Wilson 2008).

The bootstrap bias-correction procedure slightly affects the estimates (92.58 %). This is clearly shown by Fig. 1 that jointly scatters DEA efficiency scores and bias-corrected ones.

All estimates are available from authors upon request.

As proposed in Del Monte and Papagni (2001), we use as a proxy of environmental corruption the number of reported crimes against the public administrations. Data at provincial level have been taken from the Annals of Criminal Statistics (Statistiche giudiziarie), National Institute of Statistics (ISTAT) from year 2000 to 2004.

Other studies that focus on public procurement procedures have detected corruption looking at the prices paid for goods and services provided by the public sector. For instance, Olken (2009) measures corruption by comparing the official prices of road-building project in Indonesian villages with an independent estimate of the cost of project realisation provided by a team of engineers. Di Tella and Schargrodsky (2003) analyse the prices of some inputs purchased by the hospitals of Buenos Aires before and after an investigation on corruption has been run. Bandiera et al. (2009) use detailed data on the prices of goods obtained by Italian public administrations from an approved supplier, CONSIP. They distinguish between the corruption (called ‘active waste’) and inefficiency in managing purchases (called ‘passive waste’). Their results show that the weight of passive waste is four times stronger than the one of active waste.

However, in the paper, there is no attempt to examine corruption in public contracts, because the main aim is to investigate what affects the performance of public work execution. Therefore, environmental corruption is used in the analysis as factor affecting the efficiency of public contracts for infrastructure provision.

The WCI is constructed taking into account the sub-categories involved in each project as well as their relevance. Complexity may be assumed to be decreasing in the concentration of works in one or few subcategories. More formally, the Weighted Composition Index (WCI) is defined as a follows: If \(W_{[i]j}\) is the amount of money to be spent, within the \(j\)th project, with (\(j; 1,\ldots ,n\)), for works of the \(i\)th sub-category (\(i; 1,\ldots ,G\)), and \(W_{[i]j} \ge W_{\left[ {i+1} \right] j} \forall i\), then \(WCI_j =\sum \limits _i {i} \frac{W_{[i]j} }{\sum {W_{\left[ j \right] }}}\in {\left[ 1,\frac{G+1}{2} \right] }\).

According to De Carolis (2009), the Italian procurement rules to exclude anomalous bids turn out to be not to work properly and to cause significant efficiency losses in bidder selection.

We computed three classes with reference to public works with reserve prices in the ranges of 150,000–500,000; 500,000–1,500,000; and 1,500,000–5,000,000 of Euros.

Following a major reform giving greater autonomy to local government launched in 1990 and still under way, regions can adopt their own specific regulations in the public procurement field. Moreover, the new Article 117 of the Italian Constitution gives regional governments the legislative power parallel to that of the central government in specific fields such as town planning, health and public works. This legislative power must be exercised within the limits laid down by central government legislation and must not conflict with the national interest or the interests of other regions. In the public works sector, the national legislation applicable to all contract procedures lays down that the regions and the bodies financed by them are only required to comply with the EU Directives. They may adopt implementing legislation that is different from national legislation. Moreover, regions with special statuses have adopted their own autonomous regulations for public procurement procedures.

We also perform an F-test for joint significance of these variables (YEARS and REGIONS) for our baseline model employing the Banker and Natarajan (2008) estimator. The null hypothesis of no YEARS and REGIONS effects can be rejected at any conventional levels of significance. All estimates are available from the authors upon request.

Similar results are obtained by Guccio et al. (2014) showing that local governments do not seem to be under sufficient and effective pressure to behave efficiently in the execution of public works.

Table 12, in the Appendix, reports the descriptive statistics of the variables employed in this section.

To make possible the comparison of the regional and provincial measures of environmental corruption, we have normalized the former by the median according to what has been done by (Golden and Picci (2005), footnote 24). Moreover, since the regional index is time invariant in these estimates, we do not use regional fixed effects.

When social capital is lower, politicians face weaker incentives to pursue social welfare, and corruption is more likely to occur.

Table 13 shows that only SOCIAL_K and ACCOUNTABILITY turn out to be significant and with the expected signs. Overall, the results are in line with those expected. Finally, it has to be noted that the number of observations reported in Table 13 is lower than that in the previous Tables because the indexes of social capital and accountability are not available for all the provinces included in the analysis.

The whole model is available from the authors upon request. Beside the Cobb-Douglas production function with half-normal distribution, we have also estimated a Cobb-Douglas production function with exponential distribution and a truncated-normal distribution with results similar to those reported. Also these further estimations are available from the authors upon request.