Diverging patterns of interaction around news on social media: insularity and partisanship during the 2018 Italian election campaign

Summary

This paper investigates how partisan communities in Italy engaged with political news during the 2018 general election campaign, combining Twitter-based measures of news source partisanship with Facebook engagement data on 84,815 political news stories. The authors develop a Multi-Party Media Partisanship Attention Score (MP-MPAS) and an original “insularity” index, then show that sources favored by populist party supporters (M5S, League) are markedly more insular than those favored by mainstream parties. They further demonstrate that insularity systematically alters engagement dynamics on Facebook — insular sources are amplified through shares, while cross-partisan sources draw comment-based contestation — and that the Five Star Movement uniquely manages to bend the shares/comments ratio in its favor by amplifying positive coverage and contesting negative coverage.

Key Contributions

• Adapts the US-developed Media Partisanship Attention Score to multiparty systems via a replicable MP-MPAS metric.
• Introduces a novel insularity index combining maximum partisan attention with a Gini-based dispersion measure across parties.
• Triangulates Twitter (for partisan attention) and Facebook (for engagement) under contemporary data-access constraints.
• Provides non-US, polarized-pluralist empirical evidence on partisan news engagement, addressing the US-centric bias in the echo-chamber literature.
• Refines the echo chamber thesis by distinguishing amplification (shares) on insular sources from contestation (comments) on cross-partisan ones.
• Identifies a distinctive “cyber party” engagement signature for M5S, linking populist ideology to digitally orchestrated attention tactics.

Methods

The authors collected ~216,765 retweets of official Italian party/leader accounts in January 2018 to estimate user partisanship, then tracked the top 5,000 partisan users via DMI-TCAT (Feb–Mar 2018), yielding 4.4M tweets and ~130M resolved URLs used to compute MP-MPAS for 634 news sources. Each source was assigned an insularity score (max MP-MPAS × Gini of its partisan attention distribution) and classified into four tiers. In parallel, 84,815 political news stories from 4,113 domains were harvested via Huginn (Google News, GDELT, Twitter) and tracked on Facebook through the Graph API for one week each. Three coders manually annotated the top decile (3,731 stories) for sentiment toward M5S, PD, League, and Forza Italia (Krippendorff’s alpha 0.72–0.84). Analyses used Kruskal-Wallis, Dunn’s post-hoc with Bonferroni correction, and Spearman correlations.

Findings

• Median insularity: M5S = 0.73, League = 0.67, LeU = 0.63, PD = 0.63 (system mean 0.65); M5S significantly higher than all other groups.
• Of 634 adjudicated sources: 33.9% League, 27.9% M5S, 19.1% cross-partisan, 9.5% PD, 7.7% LeU; almost none for Forza Italia/FdI.
• Insularity is negatively correlated with the comments/shares ratio (Spearman ρ = −.31, p < .001): more insular → more shares per comment.
• For M5S, positive stories received more shares per comment and negative stories more comments per share (p < .001).
• For PD, Forza Italia, and the League, the inverse held — negative stories were shared, positive stories commented.
• Cross-partisan, low-insularity sources function as arenas of contestation, not echo-chamber amplification.

Connections

This paper sits at the intersection of work on hyper-partisan ecosystems and populist mobilization in Italy, and is foundational for the authors’ subsequent program on coordinated link sharing and problematic information — see Giglietto2020-9d8acdd7, Giglietto2023-fa71a001, Marino2024-2fbc690f, Giglietto2025-1765bb4f, Giglietto2025-1e9a0917, and Giglietto2026-632ef967. It engages directly with debates over selective exposure and cross-cutting exposure exemplified by Bakshy2015-rn, and connects to broader studies of partisan asymmetries and hyperpartisan media such as Rossini2026-jn and Ghezzi2023-8bebc91f.