Corruption and EU Institutions: The Italians' Opinion

Abstract

Corruption is unanimously recognized as an endemic pathology of the Italian political system. Even after the ‘Tangentopoli’ scandals that broke down the Prima Repubblica and rapidly changed the partisan configuration, corruption remains one of the most relevant problems that affects the public sector. The different successive governments do not seem to have achieved significant results in constraining corruption. Italian citizens perceive institutions as corrupt and are increasingly disenchanted about politics. But what is the opinion of the Italians about the role of the European Union (EU) institutions in facing the problem of corruption? The purpose of this paper is to analyse the perceptions of Italian citizens about the spread of corruption within EU institutions and their potential role in preventing and fighting corruption in their country. In line with the scholarly literature on this topic, we expect that citizens' attitudes toward the EU in relation to the problem of corruption are mainly driven by their perceptions of the domestic national context. Taking advantage of data gathered from the Eurobarometer and comparing the Italian and the European contexts, the paper shows that citizens' opinions about corruption within EU institutions are drawn by their perceptions at the national level. However, at the same time, citizens who express more negative evaluations of the performance at the national level tend to be more confident about the role played by the EU in restraining corruption.

Key Words:

• Corruption
• European Union
• Italy
• political support

Notes

¹ For information and data on CPI, see http://www.transparency.org/research/cpi/ (accessed 20 June 2013).

² In statistical terms, the subjects of the analysis are arrays of contingency, whose elements indicate the number of times that the characteristics of two different sizes have been detected jointly. In the arrays, rows and columns play similar roles as they represent a breakdown of the whole data according to two magnitudes of the qualitative type, each in turn divided into a group of characteristics or modalities. The analysis aims to explain why the data matrix deviates from a position of homogeneity, which occurs when the rows (or columns) are proportional. In this work, the active variables – which contribute to the formation of the two axes factors – included in our MCA are 71. We did not consider the variables in which the modalities presented a frequency percentage of less than 2.

³ Since eigenvalues obtained by MCA give a pessimistic evaluation of the variability explained by factorial axes, we have computed the variance considering only eigenvalues higher than 1/k, where k is the number of variables used in the analysis (see Benzécri, 1965, 1976).

⁴ This is why Figure 2 does not report the position of those modalities and countries that collapse in the centre of the two axes.

⁵ The exact wording of the question is thus: ‘Could you please tell me whether you totally agree, tend to agree, tend to disagree, totally disagree with the statement: “There is corruption within the national institutions of your country/the institutions of the EU”.’

⁶ Given the low number of respondents for each Italian region, it is not possible to analyse the inter- and intra-regional differences in the opinions on corruption and the role played by the EU. This is why we have preferred to aggregate respondents coming from neighbouring regions and to analyse the differences in the three main Italian sub-areas. 

⁷ For a more complete review of the theoretical approaches in studying public support towards the EU, see, among others, Gabel (1998a, 1998b) and Serricchio (2010).

⁸ It is worth noting that the term ‘objective’ referring to experienced-based measures is used in the literature to distinguish these measures from the more ‘subjective’ perceptions-based indicators. However, as pointed out by Clausen, Kraay & Nyiri (2011), objective measures are also biased and present some methodological problems.

⁹ More precisely, Models 2 and 3 are estimated using generalized linear latent mixed models (gllamm). As a further robustness check, we have dichotomized the dependent variable and we have also estimated these two models using random-effects logistic regressions. The categories ‘totally agree’ and ‘tend to agree’ are dichotomized into the category ‘agree’ (1), while the categories ‘tend to disagree’ and ‘disagree’ are dichotomized into the category ‘disagree’ (0). The results of these models are very similar to the main ones and present the same levels of significance.

¹⁰ We have run a Hausman test and the results obtained suggested that we apply a random-effect regression model, rather than a fixed-effects one.