Finding Out the Hard Way: Uncovering the Structural Foundations of Political Dissatisfaction in Italy, 1973–2013

Abstract

This article analyses the long-term dissatisfaction of citizens with democracy in Italy in light of the economic and institutional transformations that have occurred over the last four decades. Using multilevel analysis, the results challenge previous research showing how poor economic performance in the form of unemployment and inflation has increased dissatisfaction. Most importantly, the article demonstrates that the widespread political dissatisfaction can be connected to fragmentation of the party system. In this regard, the impact of the electoral reforms that have been introduced are assessed. These favoured a recovery of satisfaction between the ‘First’ and the ‘Second’ Republic due to the emergence of bi-polar political competition. However, the effect has been contained by an inability to substantially reduce the number of parties. By drawing on compelling data, this study provides systematic evidence of the macro-foundations of Italian political discontent.

Acknowledgements

A previous version of this article was presented at the panel ‘The agenda of the Second Republic. A first look at the priorities of public opinion, media and institutions in the political cycle of alternation’ of the XXVII conference of the Italian Political Science Association. We would like to thank all the panel participants for their useful comments. We also want to express our gratitude to the anonymous reviewers and the Editors for their suggestions which have contributed to the improvement of the article. The authors are equally responsible for the content of this work.

Notes

1. We used the Eurobarometer Trend File; EB 58.1; EB 60.1; EB 61.0; EB 63.4; EB 65.2; EB 68.1; EB 71.3; 73.4; EB 76.3; EB 77.3; 79.3.

2. This choice is due to the well-known geographical differences in the Italian context (Diamanti 2003). North-west includes: Piemonte, Valle d’Aosta, Liguria, Lombardia; north-east includes: Trentino-Alto Adige, Veneto, Friuli-Venezia Giulia, Emilia Romagna; centre includes: Toscana, Umbria, Marche, Lazio; south includes: Abruzzo, Molise, Campania, Puglia, Basilicata, Calabria; islands includes: Sicilia and Sardegna. We follow the NUTS1 classification.

3. The measure is computed using seat shares.

4. To create the index, we used parties before 1994 and coalitions after 1994. See Chiaramonte (2007).

5. We used the election results for the Camera dei deputati (the Lower chamber). For the period between 1994 and 2001, we included in the models the scores computed using the proportional part of the electoral rule. However, we also ran the models using the majoritarian part of the electoral rule and the average scores using both parts of the electoral rule. No substantial differences between the models were found. See Bardi (2006) on the topic.

6. Employment status is missing in 1973, education in 1995, satisfaction with life in 1980, 1981 and 2004, political discussion in 2009 and the left–right scale is missing in 2011, 2012 and 2013.

7. This model is also known as the hierarchical age–period–cohort model (HAPC) (Yang and Land 2006), and is specifically suited to modelling outcomes using repeated surveys in one country. Published works using the HAPC model do not discuss the issue of potential autocorrelation among the random effects. However, in order to exclude potential bias, we run time-series diagnostics tests on the models. The tests for autocorrelation and partial autocorrelation show that the assumption of independence of random effects is not violated. Other diagnostic tests show that the normality assumption is not violated either. All the models are checked for multicollinearity. The correlation matrix of the aggregate variables and the variance inflation factors of the models can be found in the online supplementary material.

8. We also run the same models without including the cohort level. These models show worse goodness-of-fit statistics compared to those including the cohort effect.

9. The predicted change in probability is the difference between the probability of being ‘not very satisfied’ or ‘not at all satisfied’ computed at the maximum of the covariate of interest and the probability of being ‘not very satisfied’ or ‘not at all satisfied’ computed at the minimum of the covariate of interest. All the other covariates are held at their means.

10. The Intraclass Correlation Coefficient (ICC) is ‘the fraction of total variation in the data that is accounted for by between-group variation’ (Gelman and Hill 2006: 448).