Home Team Effect and Opinion Network after the Sewol Ferry Disaster: A mixed-method study of the influence of symbol and feedback on liberal versus conservative newspapers’ negative opinions

The paper investigates how symbols (emotional and cognitive) and political system feedback (positive and negative) shaped media negativity toward the conservative Park Geun-hye administration following the April 16, 2014 Sewol Ferry Disaster. Recognizing disaster politicization and the centrality of negative media opinions to policy change, the study asks: How did symbols and feedback influence newspapers’ negative opinions, and did responses differ by media ideology (liberal vs. conservative) under a conservative administration? The Sewol event—highly politicized due to many student victims and proximity to elections—offers a unique case to examine media responses and agenda dynamics. The study’s purpose is to evaluate theoretically and empirically, via mixed methods, how emotional and cognitive symbols and feedback affected negative opinions, and to compare liberal and conservative newspapers to test the Home Team Effect (HTE). The work is important for understanding media-policy dynamics, disaster politicization, and for introducing opinion networks as a tool to measure and visualize negative opinion as policy attention.

The study integrates three frameworks: (1) Multiple Streams Approach (MSA), where policy entrepreneurs deploy symbols to couple problems, policies, and politics; symbols have emotional (affect-laden, tension-reducing) and cognitive (information-providing) functions. (2) Punctuated Equilibrium Theory (PET), where feedback processes (positive amplifying change; negative stabilizing) shape attention and policy shifts, though distinguishing their effects can be difficult and time-varying. (3) Home Team Effect (HTE) and motivated reasoning in media: outlets tend to align with partisan preferences; Korean newspapers exhibit discernible ideological orientations (Hankyoreh liberal; Chosun conservative), with editorials strongly reflecting outlet stance. Media shape public agendas and emphasize negative opinions post-disaster, affecting policy responses. Prior work focused on budgets/attention as punctuation indicators; this paper adds negative opinion as another attention metric and addresses gaps: limited empirical study of symbols, incomplete application of PET feedback constructs, and scarce cross-ideology comparisons of symbol/feedback effects. The paper also situates opinion networks within social science as a means to capture direction and intensity of inter-actor sentiments in texts.

Design: Mixed-method study combining qualitative case study and quantitative content analysis, opinion network construction, and multiple regression analyses. Data: Archival sources from South Korea, April–December 2014. Newspaper editorials: 424 total (Chosun Ilbo [conservative] 192; Hankyoreh [liberal] 232), retrieved via “Sewol Ferry” search. Additional government/economic data (e.g., KOSPI) used as controls. Content analysis and opinion network: Hand-coded opinion mining at sentence level to identify positive/negative opinions, senders and receivers (nodes), and magnitude (±1, with some elements weighted ±2 for editorial headlines/pictures). Edge lists were created per sentence (sender→receiver with polarity and weight). Intercoder reliability reached 82.99% (Holsti); coders reconciled differences to achieve final agreed dataset. Total coded ties: 8,107 (Hankyoreh 4,796; Chosun 3,311). Network measures: Weekly aggregation over 38 weeks post-disaster. Degree centrality (in-degree, sum of weights; standardized) computed with Netminer 4. Dependent variable: magnitude of negative opinion toward Park administration (sum of centrality across President Park, Park administration, Blue House, ruling party). Independent variables: Symbol measures—Emotional symbol (weekly sum of positive centrality toward victims, their families, high school victims, survivors); Cognitive symbol (log weekly count of retrieved bodies). Feedback measures—dummy-coded weekly periods for: resignation of PM Chung (initially intended negative feedback, weeks 3–11 coded 1), chasing Mr. Yoo (weeks 5–14 coded 1), announcing dismantling KCG (weeks 6–38 coded 1), passing three Sewol Acts (weeks 31–38 coded 1), ending body retrieval mission (weeks 31–38 coded 1), establishing MPSS (weeks 32–38 coded 1). Controls: negative opinion toward “government” and “KCG” (separately), economic factor (KOSPI; logged), partisan shift indicator around June 4 local elections. Unit of analysis: week (1–38). Analytic strategy: Separate OLS regressions with robust standard errors for each newspaper (liberal vs. conservative), modeling weekly negative opinion on Park administration as a function of symbols, feedback dummies, and controls. Complementary qualitative case narrative aligned events (symbols and feedback) with observed opinion dynamics and attention/budget punctuations (editorial counts, Google Trends; safety budget increase of 17.9%). Opinion network visualizations illustrated changes in negative opinion under key events (e.g., after bills passed and mission ended).

- Emotional symbols vs. cognitive symbols (MSA): Emotional symbols significantly increased negative opinion toward the Park administration in both newspapers, with stronger effects in the liberal Hankyoreh than in the conservative Chosun. Cognitive symbols (retrieved bodies) showed no significant effect in either outlet. - Feedback effects (PET): Effects varied by event and newspaper ideology; feedback did not operate uniformly as purely amplifying or dampening. The resignation of PM Chung (initially negative feedback, functionally positive via failure/controversy) increased negative opinion in both newspapers. Chasing Mr. Yoo did not significantly affect negative opinion overall; in some specifications it was marginal. Announcing the dismantling of KCG increased liberal negative opinion but not conservative. Passing the bills and ending the retrieval mission corresponded with decreased liberal negative opinion (mixed feedback cancelation), with no effect in the conservative paper. Establishing MPSS did not reduce negative opinion as a negative feedback would predict; liberal negativity sometimes increased or remained sensitive. - Home Team Effect (HTE): Under a conservative administration, the conservative newspaper was more lenient (lower magnitude and sensitivity of negative opinion) than the liberal newspaper. Content counts show stronger liberal criticism: liberal paper recorded 791 negative opinions targeting Park administration entities vs. 303 in conservative; victims received more positive/emotional mentions in the liberal paper (409 vs. 188). Descriptives of dependent variable: mean weekly negative opinion magnitude was larger in Hankyoreh (-0.019) than in Chosun (-0.007). Regression fit: R^2 ≈ 0.755 (liberal) and 0.606 (conservative). Controls indicated conservatives deflected blame more toward KCG while liberals blamed “government.” - Punctuation indicators: Safety-relevant budget increased by 17.9% post-disaster; editorial attention and Google Trends showed sharp spikes, evidencing punctuated attention alongside the introduced measure of negative opinion as attention. Overall, emotional symbols outweighed cognitive ones in driving negative opinion; feedback effects were context-dependent; and ideological stance shaped sensitivity and leniency, supporting HTE.

Findings empirically substantiate that symbols, especially emotional ones, are pivotal tools for coupling streams (MSA) and elevating negative opinion toward incumbents after focusing events. Cognitive information (retrieved body counts) did not mobilize blame as effectively as affective appeals. Feedback processes (PET) did not yield uniform amplification or dampening; failed or controversial attempts at blame avoidance (e.g., PM resignation saga) transformed into positive feedback that amplified negative opinion. Mixed feedback (passing bills and ending mission) could cancel out or reduce negativity in the liberal paper, suggesting temporal and contextual contingencies and supporting claims that distinguishing positive/negative feedback effects is difficult. Ideological alignment mattered: the liberal paper, opposed to the conservative administration, reacted more strongly to symbols and feedback, whereas the conservative paper was more lenient, consistent with HTE and motivated reasoning. Introducing negative opinion as an attention metric, operationalized via opinion networks, provides a complementary lens to budgets and coverage counts, linking media discourse structure directly to policy-relevant negativity.

The study contributes by (1) demonstrating that emotional symbols, more than cognitive ones, heighten negative opinions toward incumbents in disaster politicization; (2) showing that political system feedback effects are situational and can invert (negative to positive) or cancel, challenging assumptions of uniform feedback dynamics in PET; (3) providing evidence for the Home Team Effect in media—ideology conditions sensitivity and leniency; and (4) advancing the opinion network method to measure and visualize directed, weighted negative opinion as a form of attention. Practically, policymakers and policy entrepreneurs should account for media stance and the potency of emotional symbolism when managing crises and communications. Future research should extend opinion network analysis to other issues, outlets, and platforms (social media, online forums), explore additional symbol types, and examine when and why feedback effects cancel or invert across contexts, languages, and political systems.

- Media scope: Analysis is limited to two newspapers’ editorials (Hankyoreh and Chosun) as ideological proxies; results may not generalize to all media or to non-editorial content. - Temporal scope: April–December 2014 window may miss longer-term dynamics and delayed effects of feedback or policy implementation. - Measurement constraints: Hand-coding, while carefully reconciled, remains subject to interpretive judgment; mapping complex sentences to single-opinion edges may simplify nuance. - Feedback identification: Coding feedback periods and their intended valence (positive/negative) is theory-informed but may overlap and interact in ways that complicate causal attribution. - External validity: Findings are situated in South Korea’s political-media context during a conservative administration; HTE and feedback dynamics may differ in other systems or administrations. - Cognitive symbol proxy: Using retrieved-body counts as the cognitive symbol may not capture all forms of cognition-oriented messaging that could influence opinion.