Data and Codes for "Analyzing Price Efficiency Using Machine Learning Generated Price Indices: the Case of the Chilean Used Car Market "

This dataset contains all the necessary data and codes to replicate the main findings of the article, which examines how new car import prices affect the valuation of used vehicles in Chile's secondary car market. The study uses event study and difference-in-differences (DiD) methodologies to evaluate price efficiency.

The replication package includes:

- Used car price indexes dataset: Filtered and pre-processed to include key vehicle attributes (model, year, mileage, transmission, fuel type, seller type, region, etc.).

A new car import dataset built from official customs records covering shipment dates, CIF prices, and units imported per model and version.

R and Python scripts for estimation and analysis:

"Unit Root Test.R": Tests for stationarity using Im, Pesaran, and Shin (IPS) panel tests; and "Event Study CARS.py":

"Event Study CARS.py": Performs event study estimation using cumulative abnormal returns (CAARs), including subsample analyses by vintage and vehicle segment.

"DiD Event Studies.R": Estimates difference-in-difference (DiD) regressions using staggered treatment timing and fixed effects and calculates cumulative abnormal returns (CAARs) from fitted values.

Each code script includes comments and references to the corresponding tables and figures in the paper.

Key findings that can be replicated using these files include:

- Evidence of prompt and statistically significant price responses in the used car market following increases in new import prices.

- Stronger responses among newer and high-end used cars.

- These responses occur before the public release of import data, suggesting high informational efficiency.

- The results are robust across different methodological approaches and sample partitions.

This replication package allows for the independent verification of results. All datasets are anonymized and formatted for reproducibility. The codes are compatible with R 4.2+ and Python 3.8+ environments.