A political cartography of news sharing: Capturing story, outlet and content level of news circulation on Twitter

TL;DR

This paper tackles the limited, one-dimensional view of news sharing by proposing a multi-dimensional pipeline that jointly analyzes news sources and content on Twitter. It constructs a CHES-validated, two-dimensional political space from MPs’ follower networks and maps news sharing across stories and topics by linking 26 outlets and analyzing full-text articles with a Structural Topic Model to derive metatopics. The results demonstrate that a two-dimensional space reveals directional patterns and fault lines invisible in one-dimensional analyses, and that considering individual stories and content-level topics uncovers substantial heterogeneity in sharing across outlets and topics. The approach offers a robust framework for understanding online information circulation with potential applicability beyond Germany and Twitter, advancing research on media influence, polarization, and the information environment online.

Abstract

News sharing on digital platforms shapes the digital spaces millions of users navigate. Trace data from these platforms also enables researchers to study online news circulation. In this context, research on the types of news shared by users of differential political leaning has received considerable attention. We argue that most existing approaches (i) rely on an overly simplified measurement of political leaning, (ii) consider only the outlet level in their analyses, and/or (iii) study news circulation among partisans by making ex-ante distinctions between partisan and non-partisan news. In this methodological contribution, we introduce a research pipeline that allows a systematic mapping of news sharing both with respect to source and content. As a proof of concept, we demonstrate insights that otherwise remain unnoticed: Diversification of news sharing along the second political dimension; topic-dependent sharing of outlets; some outlets catering different items to different audiences.

Figures (11)

• Figure 1: The process of identifying interpretable political axes from a two-dimensional space derived from Correspondence Analysis (CA). Left panel: Initial two-dimensional space that includes MPs represented by colored dots. Their colors are corresponding to their political party (i.e., AfD blue, Die Linke pink, Greens green, CDU/CSU black, FDP yellow, SPD red). The green density plot shows user distribution across this space, dimension-wise distributions are given on the margins. Middle: The average positions of each party's MPs are calculated (top). Subsequently, the space is rotated and the dimension-wise correlation with every othogonal combination of the left-right plus any other CHES dimension (examples displayed at bottom) is calculated, until we arrive at an angle for which two CHES dimensions maximize combined correlation. (CDU/CSU are treated as separate parties to calculate the correlation, hence the two grey dots in the figure of the average party positions -- they represent their respective positions.) Then, the whole space (that is, including the non-MP users) is assigned the rotation and used as the basis of analysis in the following (right panel). (In fact, four CHES dimensions rank nearly equally high in orthogonal combination with the left- right axis with similar rotation angles. Hence we average over their four angles and combine them in the second dimension, which we call 'Elite-/EU-skeptical/Protectionist'.)
• Figure 2: Circulation of individual news stories (A., B.), outlet-level distributions of two outlets (C., E.) and distribution of story means (D., F.). Two articles from focus.de are shared in different regions of the political space (individual shares are displayed as blue crosses, the mean of all shares of the respective story by a red 'X'). If we investigate the distribution of users sharing focus.de articles (regions of high density are darker, and since we deal with sharing and not with the baseline distribution now, we color the distribution blue), we see that the outlet is both shared by users placed close to AfD as well as by center-right to right, less elite-/EU-skeptical/protectionist users (C.). Still, if the distributions of individual story means is considered, we find that articles by the outlet rarely bridge the latter divide. They are either shared on one side or the other of the cleavage (D.). An outlet that succeeds in publishing articles being shared on both sides is nachdenkseiten.de: Most mean story positions are between AfD and Die Linke (F.).
• Figure 3: Comparison of the sharing distributions of welt.de and focus.de for different metatopics. While articles on environmental issues and climate change by the latter outlet are shared comparatively more often by users left of center (B.), its coverage on national politics is shared by right-leaning users both high and low on the EESP axis (A.). welt.de articles on national politics (C.) are very similarly distributed to focus.de's articles on the same topic. News on climate change and environmental issues, on the other hand, are shared to a disproportionate extent by right-wing EESP users (D.).
• Figure 4: Four topics and their corresponding metatopic (environment and climate change). News on environment and climate change are shared by users across the political spectrum. Yet, stories on extreme weather events and Fridays For Future are shared nearly only left of center. News on the more extreme climate movement Last Generation and on the science and institutions dealing with climate change are shared by the political counterparts (right of center and by right-wing elite-/EU-skeptical/protectionist users, respectively).
• Figure 5: CA percent of variance per dimension.