Twitter as a mirror of political space(s): An analysis of multi-level party competition in Italy

ABSTRACT

The literature on the nationalization of party systems has found that decentralization of authority leads to increasing territorial heterogeneity of voting behaviour. In this article, we intend to further extant knowledge on this topic by exploring whether different levels of government imply distinct representations of party competition at the citizens’ level. We do so by analysing Twitter data collected during the 2015 regional elections in five Italian regions (Liguria, Veneto, Tuscany, Campania and Apulia) to estimate national and regional party configurations by means of multi-dimensional scaling. Analyses show that national party competition is generally perceived in the same way across regions, but representations of regional political spaces tend to stand out from it and also to differ from each other. These results suggest that, in line with the process of political decentralization of authority, voters perceive party competition in different ways according to the level of territorial governance at stake.

KEYWORDS:

• Political spaces
• multi-level party competition
• regional elections
• Twitter data
• Italy

Acknowledgements

We would like to thank the organizers and participants of the panel ‘Strumenti e tecniche di analisi dei social media’ (Conference of the Italian Political Science Association – Milano, 15/17 September 2016) and of the ‘Values in European and Japanese Politics’ Conference (Brussels, 8/9 March 2019) for their valuable comments and feedback on the early draft of this paper.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 On the other hand, there are geographical-based specificities that have characterised the electoral support in the different Italian regions. For instance, Veneto was traditionally (before the 1990s) the stronghold of the Christian Democratic Party (DC), while Tuscany was the one of the Italian Communist Party (PCI).

2 Please note that some regions modified aspects of the original electoral law defined at the national level, including the number of possible preference votes. Specifically, this is the case of Tuscany, where up to three preference votes can be cast, and Campania, where citizens are allowed to choose up to two candidates of different gender.

3 For the record, Apulia was the region with the highest incidence of preference votes (71% of the total), while Veneto was the least affected (31%). Campania, Liguria and Tuscany held instead an intermediate position. Please refer to Rombi (2015) for further details.

4 In our data, the two-dimensional solutions fit better than the one-dimensional across all regions (S-Stress index < 0.15 according to a well established rule of thumb, see Dugard, Todman, and Staines 2010). We also find that three-dimensional models would not improve substantially the goodness of fit, in line with previous studies (De Sio and Schadee 2013).

5 We have excluded from analyses the coalitions/lists that had an extremely small number of followers (N < 100).

6 For a broader discussion on the meaning of such vertical axis when space dimensions are not determined a priori, i.e. on the basis of pre-defined sets of issue and/or ideological preferences, but are rather ‘inductively’ derived from patterns of preferential data such as likes/dislikes and propensity scores, please refer to Schadee, Segatti, and Vegetti (2019).

7 This was a mixed mode survey relying on CATI, CAMI and CAWI interviewing, performed in June 2015 by an Italian survey agency, i.e. Demetra Srl. The survey agency provided all involved subjects with appropriate survey description, as well as with the opportunity to express their informed consent at the beginning of the interview. As regards the online part, participants were informed via an email and after clicking on the link they were given the opportunity to express (or deny) their consent to participate. In the case of CATI and CAMI interviews, this whole process took place verbally. Since procedures for an ex-ante formal ethics assessment have been put in place at the relevant institution (i.e. the Ethics Committee of the University of Milan) only in early 2019 (Rector’s Decree of 18th January 2019), i.e. four years after this research was fielded, a specific approval number is not available for the electoral survey at issue. Nevertheless, all involved data have been elaborated and treated in compliance with the national and international regulations in force at that time and were made available to our research group in a completely anonymised form.