Abstract
Since the end of the Cold War, Italy has radically transformed its foreign and security policy, participating in several Military Operations Abroad (MOA) across the world. A few qualitative studies have already analysed how Italian parties debated and voted on this issue, underlining a bipartisan consensus between centre-left and centre-right parties, based on a common humanitarian narrative. This article provides a substantial methodological contribution to this research agenda, explaining party support in Italy for the six most relevant MOAs during the so-called ‘Second Republic’ (1994–2013), through the employment of automated text analysis and linear regression models. In line with existing literature on the party politics of military interventions, the findings indicate a curvilinear distribution of support across the left–right axis, the strong impact of government–opposition dynamics and the interaction between international legitimacy of the specific operation and ideological leaning.
Notes
Acknowledgements
The author would like to thank Fabrizio Coticchia, Andrea Pedrazzani and the two anonymous reviewers for their useful comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 For the full report of the debate, see the digital archive of the Chamber of Deputies (http://documenti.camera.it/_dati/leg14/lavori/stenografici/sed298/pdfel.htm).
3 For an overview of the state of the art and the latest developments in the sub-field of FPA, see Smith et al. ().
5 For a complete list of all the expired and ongoing MOAs, see the website of the Italian Ministry of Defence: https://www.difesa.it/OperazioniMilitari/Pagine/OperazioniMilitari.aspx.
7 On 21 February 2007, the second Prodi government was defeated at the Senate in a vote on the renewal of Italian commitment to NATO Operation ISAF in Afghanistan and the expansion of a U.S. military base in Italy due to the defection of a couple of MPs from the extreme left. For a more detailed account of this event, see Carbone (Citation2007).
8 For applications of Wordfish in the field of political science, see Slapin and Proksch (), Klüver (), Proksch and Slapin (Citation2010), Proksch et al. (), Ceron (, ) and Frid-Nielsen (Citation2018).
9 For an application of Wordscores, see Laver et al. (Citation2003). For an extensive overview of unsupervised and supervised scaling methods, see Grimmer and Stewart (Citation2013).
10 For applications of Wordfish on party manifestoes, see Slapin and Proksch (), Proksch et al. (Citation2011). For applications on parliamentary speeches, see Proksch and Slapin (Citation2010); Frid-Nielsen (Citation2018).
11 All the speeches were collected from official reports of the debates. These reports are available at the digital archive of the Chamber of Deputies (http://storia.camera.it).
12 The names of the operations refer to both the international denomination and the Italian contributions to them. The years in the parentheses correspond to the ones in which the debates took place and not the beginning and the end of the missions. Contributions to related operations Allied Force and Allied Harbour in Kosovo were discussed in the same debates.
13 For a more extensive discussion of the approval process of MOAs in Italy, see Di Camillo and Tessari ().
14 The mission Iraqi Freedom is another example of a mission formally approved through a law decree, even though government referred it to Parliament before. Other ones are UNMIBH (Bosnia), Allied Harmony (Macedonia), and UNSMIS (Syria).
15 The conversion of law decrees was also an opportunity to debate other smaller MOAs not taken into account in this article.
16 According to the Italian law, the minimum number of seats to form a parliamentary group is 20. In order to give more homogeneity to the data, I continued to take into account a pivotal junior coalition partner moving to the opposition after election, even though it did not hold enough seats.
17 For more details on the debates, missions and parties, see the Online appendix.
18 This procedure admittedly neglects potential intra-party disagreements.
19 As a robustness check, party positions were also extracted without including the texts containing the speeches made by government representatives. However, positions extracted including these texts were slightly more in line with those underlined by previous qualitative studies (Calossi and Coticchia Citation2009; Coticchia Citation2011). Correlation between the two measures is statistically significant (p < 0.01). See the Online appendix.
20 The author can provide a complete list of these words on request. As a robustness check, the analysis was replicated on the positions extracted not including the speeches made by government representatives. Discriminating and non-discriminating words were very similar. For an example of the qualitative validation process, see the Online appendix.
21 The words are translated from Italian by the author.
22 As a robustness check, voting patterns were also correlated with the position extracted not including the speeches made by government representatives. Correlation is almost the same (r = 0.5117) and significant (p < 0.01). For tables and scatterplots, see the Online appendix.
23 The author can provide the complete dataset on request.
25 Party positions were attributed on the basis of the closest survey available in terms of timing.
26 Various robustness checks on these two models were conducted. First, I replicated Model 1 and 2 using the positions extracted not including the texts containing speeches made by government representatives as dependent variable (Support_Nogov). In this case, the statistical significance of Rile squared drops considerably, while the effect of being in government is equally strong and significant. This difference is due to the increased accuracy in the measures obtained including the speeches made by the government representatives. Moreover, I replicated Model 1 and 2 using the positions extracted including the speeches made by government representatives as dependent variable (Support) with fixed effects for debates and clustered standard errors for parties. These two replications do not present any significant difference with respect to the one included in the main text. Finally, I replicated Model 2 with measures extracted including speeches made by government representatives (Support) as dependent variable and an interaction term between presence at the government (Gov) and position on the left–right axis (Rile). There is no impact on the significance of Rile squared. For all these replications see the Online appendix.
Additional information
Notes on contributors
Valerio Vignoli
Valerio Vignoli is a PhD Candidate at the University of Milan. His research interests are comparative politics, international relations, foreign policy analysis, Italian politics and foreign policy. [[email protected]]