This project investigates the causal impact of the Tel Aviv Light Rail’s Red Line on property prices in its vicinity. The Red Line, running through key municipalities in Gush Dan, offers an opportunity to analyze the effect of new transportation infrastructure on housing values. We collected property transaction data from public real estate databases and enriched it with address information using web scraping techniques.
We obtained data on real estate transactions through public government sources. Given the structure of the data, we used a Selenium Python script to map each property by its "gush-helka" (plot ID) to its full address via the Israeli Centre for Mapping. To facilitate grouping by distance, we used Google API services to determine the proximity of each property to Red Line stations. This allowed us to categorize properties into three distance bands: 0-500m, 500-1000m, and 1000m+.
To estimate the causal effect, we relied on several analytical techniques:
- ATE Calculation using the Rubin-Neyman framework to measure average treatment effects across distance groups.
- Propensity Score Matching (PSM) and Mahalanobis Distance Matching for balancing treatment and control groups, enhancing comparability.
- Regression Discontinuity Design (RDD) to assess impact near predefined proximity thresholds.
These methods account for confounding variables like year of construction, property area, and number of bedrooms to control for baseline differences in property characteristics across distance bands.
Preliminary findings suggest a positive impact of the Red Line on property values within 500m of stations, with diminishing effects as distance increases. However, variability in results and non-significant findings in some cases indicate that further analysis with larger datasets is needed for more definitive conclusions.
Data limitations, including address mapping challenges and pre-existing differences among properties, contribute to some residual confounding. Additionally, market-wide trends during the study period may influence our estimates.
-
Data Folder:
raw_data.csv: The raw real estate data gathered from public sources.preprocessed_data.csv: The data after preprocessing, with properties divided into distance groups.
-
scraping.py: Python script for web scraping to collect property data, mapping each property by "gush-helka" to its full address.
-
google_api_distances.py: Script for categorizing properties into distance bands from Red Line stations using Google API.
-
Causal_Inference_Final.ipynb: Jupyter Notebook containing the data analysis, including the causal inference experiments, statistical tests, and final results.
While proximity to the metro shows potential for enhancing property values, this effect is context-sensitive and varies across distance bands. Our study underscores the importance of comprehensive causal inference designs for urban infrastructure projects.
