Using Large Language Models to Predict Neighbourhood Change

This paper uses a large-language model (LLM) to read user-contributed text data – in this case property listings on Airbnb – and assign a score to indicate the degree to which the property’s neighbourhood is potentially undergoing gentrification. Preliminary results for a case study in Bristol are evaluated. These scores will ultimately be used as an input in a model of dynamic spatio-temporal neighbourhood change with the aims of (i) quantifying the impact that neighbourhood reputation has on historical house prices and (ii) predicting the emergence of new house price bubbles.