Analyzing Housing Pension Policy Discourse Using Structural Topic Models: Topic Dynamics and Media-Type Framing Differences in News Coverage, 2007–2026

This study analyzes the long-term trajectory and differentiation of Korean housing pension (reverse mortgage) policy discourse using Structural Topic Modeling (STM). From 12,799 housing pension–related news articles collected from BIGKinds (July 2007–March 2026), a 12-topic solution was adopted after SimHash near-duplicate removal. The discourse centered on policy and household debt (13.4%), asset management and retirement planning (11.9%), and enrollment statistics (10.1%). Period-specific (time-smoothed) comparisons showed enrollment statistics predominating under the Lee Myung-bak administration (18.0%), institutional design under Moon Jae-in (13.2%), and elderly-household and bequest-intention surveys under Lee Jae-myung (10.5%). Media-type differentiation was pronounced: regional dailies emphasized enrollment figures, national dailies intergenerational narratives, and the business press retirement planning. Following the March 2026 guarantee-fee reduction, the elderly-household and bequest-intention topic surged by +25.65 percentage points, though this is provisional given a post-event sample of 31 articles. The KNU sentiment lexicon and KoELECTRA agreed on only 49.0% of classifications, confirming a domain gap in policy-news sentiment measurement. Government-period estimates represent descriptive statistics conditional on time-smoothing rather than causal effects. This study contributes a replicable STM-based framework for policy-discourse monitoring in housing finance.