Specifications
Spatial coefficient
No model estimated.
Main finding
Estimation pending
Results will appear once the comparable-price sample reaches the minimum estimation threshold.
Spatial effect—per +1 SD of CI
95% CI · Conley 1 km
Diagnostics
What the model controls for
Preferred controlsdelivery platform, digital differentiation, local demand and district fixed effects
Local demand indexvillage youth density, observed MRT passenger flows and neighbourhood activity
Extended sensitivitypopulation density and MRT distance; broader controls remain documented rather than overfitted
Inferencedistrict fixed effects, HC3, district-clustered, Conley and exact wild-cluster tests
Robustness
Alternative competition measures
Estimates pending.
Identification design
From Salop's circle to the real city
1Population
Deduplicated union of OSM and drink vendors, restricted to Taipei City's 12 districts.
2Sample
Independent price setters only. Verified chain outlets are excluded from the outcome but retained in the competitive field.
3Product
Plain green tea with no milk, fruit, topping or flavour modifier. Size, discount, channel and date remain separate.
4Distance
Valhalla walking distances; CI = Σ exp(−distance/h), nearest neighbours and radius counts.
5Estimation
Pre-specified parsimonious log-price model; no control is selected by sign or p-value.
6Caution
Associative cross-section. Selection balance and trimmed IPW address observed menu availability only.
Reproducibility
Data and results
Analysis sampleCSV · one shop per row
Full research populationCSV · 927 candidates and all controls
Price auditproduct, size, channel and source
Data dictionarydefinition, unit and provenance
Model coefficientsHC3, clustered and Conley errors
Variable coverageobserved counts and missingness
Selection balancepriced versus unpriced independents
Exploratory heterogeneitydemand and differentiation interactions
Research workbookdata, audit, models and protocol