Replication Data and Code for "Words That Price: How Listing Language Encodes Audiences and Shapes Market Value"
Description
This repository contains the data and analysis code supporting "Words That Price: How Listing Language Encodes Audiences and Shapes Market Value" (submitted to the Journal of Business Research). The dataset comprises 12,682 anonymized rental listings collected from Meqasa, a major Ghanaian online real estate marketplace, covering Greater Accra between 2013 and 2026. Each record includes structured property attributes (rent, bedrooms, bathrooms, floor area, property type, amenities, locality, listing date) alongside an anonymized free-text description. The repository also includes the corpus-driven term-discovery output and thematic-coding materials used to classify listing language into seven registers (including a Western-Standard Signaling register and a Domestic-Reference register), along with inter-rater reliability data from two independent professional coders (Cohen's kappa = 0.875). Python code reproduces all reported pricing models and robustness checks, including coarsened exact matching, within-agent and cell fixed-effects comparisons, repeat-listing analysis, Oster (2019) and Cinelli-Hazlett (2020) sensitivity bounds, quantile regression, and the seasonal audience-tracking test, plus all figures. A step-by-step analysis protocol and full variable-level documentation are included. All platform-identifying information (URLs, agent names, internal IDs) has been removed from the data; agent identity is preserved only as an anonymized integer code, to support agent fixed-effects estimation.
Files
Institutions
- University of RichmondVirginia, Richmond