LLM기반 정책지수를 이용한 부동산 정책의 주택시장 파급효과 분석

초록

This study examines the effectiveness of Korean real estate policies by combining a large language model (LLM)-based policy variable construction with a local projection methodology. Drawing on policy documents issued between 2006 and 2023 and monthly panel data from seven major cities, we show that identical policy measures generate markedly different effects in the Seoul Metropolitan Area (SMA) and in non-metropolitan regions. The results further demonstrate that house prices respond significantly, and in directions consistent with policy intentions, only when policies constitute genuine “surprises,” that is, when new measures are substantially stronger than those previously in place. Conversely, policies that are weaker than preceding measures often produce statistically insignificant effects or outcomes that run counter ,to stated policy objectives. These findings underscore the importance of accounting for regional characteristics and for the relative intensity of new interventions compared with recent measures when designing real estate policy.

키워드