Priced at the promise, not the platform: when a planned railway enters house prices, and why value capture is mistargeted
Description
This dataset reproduces every table and figure in the article. The empirical setting is the Sinbundang Line Gwanggyo-Homaesil extension in Suwon, Republic of Korea. The estimation panel covers 258 apartment complexes over 2016Q1-2025Q4 (7,111 complex-quarter observations) and is built from publicly available transaction records.
Files
Steps to reproduce
DATA SOURCES Apartment sale transactions were obtained from the Real Transaction Price Disclosure System of the Korean Ministry of Land, Infrastructure and Transport (MOLIT, https://rt.molit.go.kr), covering all apartment sales in Suwon-si (Jangan-gu, Paldal-gu, Gwonseon-gu) from [2010Q1] to [2025Q4]. Complex-level attributes (construction year, number of units, floor area, parking spaces, heating type, brand) were compiled from the MOLIT/Korea Real Estate Board apartment complex registry. Station locations for the Shinbundang Line extension were digitised from the official route and station plans in the MOLIT approval notices. Complex coordinates were geocoded from road-name addresses using the [Juso.go.kr / Kakao Local] geocoding API and visually verified against the cadastral base map. PROCESSING Duplicate and cancelled contracts and records with missing price, floor or exclusive area were removed; [the top and bottom 1% of price per m2 were trimmed]. All coordinates were projected to EPSG:5186 (Korea 2000 Central Belt, metres), and straight-line distance from each complex centroid to the nearest new station was computed in metres, from which distance-ring dummies ([<=500 m, 500-1000 m, 1000-1500 m, >1500 m]) were generated. The data were aggregated into a complex-by-quarter panel with an integer quarter index (tq = year*4 + quarter) and a relative-time index centred on the [announcement / opening] quarter. ESTIMATION Effects were estimated with a difference-in-differences specification (treated = complexes within [500 m] of a new station) and a complementary event-study specification with relative-quarter dummies, both including complex and quarter fixed effects, transaction-level controls (exclusive area, floor, building age, units, parking per unit, heating type) and standard errors clustered at the complex level. SOFTWARE Data assembly, distance calculation and all figures: Python [3.11] with pandas, numpy, shapely and matplotlib. Estimation: Stata [18] with reghdfe, ftools and estout. No commercial GIS software(Qgis) is required. STEPS TO REPRODUCE 1. Unzip the archive (authors used D:\sinbundang_replication). 2. python -m pip install numpy pandas matplotlib shapely 3. In Stata: cd to the folder, then do 00_setup_paths.do do 01_main_analysis.do -> all estimation tables are written to output\tables\ 4. In Windows PowerShell: python -u "02_make_figures2.py" --root "D:\sinbundang_replication" -> Figures 1-7 are written to output\figures_manuscript\ (PDF and PNG) The archive includes the final analysis file 19_stata_analysis_data.csv (UTF-8) and 00_codebook.csv with full variable definitions, so every table and figure in the article can be regenerated without re-downloading the raw transaction data.
Institutions
- Sogang UniversitySeoul, Seoul