Skip to content

felix-reichel/woombikes-price-contagion

Repository files navigation

woombikes-price-contagion

Repository for VU Online Marketplaces

Observational Unit: "Woom Bikes" (Marketplace → Sport / Sports Equipment → Bicycles / Cycling → Bicycles)

Number: Approximately 3,000+ (Two snapshots / Web Scraper Executions)

Sample Size: All, but only approximately 900 due to duplicates and missing values, especially for variables involved in further encodings.

Correlation Heatmap:

Correlation Heatmap

Spatial_distribution of one binary variable

Spatial weight matrice construction using knn=1

Analysis of Willhaben Ads (Supply Side)

Descriptive Statistics

Histograms

Model Specification (Draft)

Model 1: Base MLR Model

$$ logPrice_i = \beta_0 + \beta * X + u_i $$

Description:

  • $logPrice_i$: Observed ad listing price of product i on the marketplace, reflecting a proxy variable of the "realized" price.
  • $X$: Matrix of explanatory variables for product i.

statInference.log

Model 2: Spatial Regression Model

Model choice based on research aim / Summary of Cook et al. (2020)

Cook et al. (2020) emphasize guiding model selection based on research aims. For unbiased estimates of non-spatial parameters, use SDM -> includes both observable spillovers (Wy and Wx). For testing spatial theories, use SAC or SDEM -> distinguishes spillovers in observables (SAC: ρWy, SDEM: WXγ) from unobservables -> prevents erroneous conclusions about diffusion/spillovers by controlling for non-observable clustering sources (p. 738).

spatial.log

Data Sources

Example of scraping code using 'scrapy': GitHub - scrapy-tutorials