Review Text Predicts Rating Shifts in U.S. Rental Markets

A recent study published on arXiv (2504.14053) examines the potential of review text from accommodation platforms to forecast future shifts in displayed ratings. Analyzing over 200,000 listings from 34 short-term rental markets in the U.S., the research develops a predefined sentiment index based on comprehensive review histories. It considers text and ratings as parallel channels reflecting guest experiences at varying speeds, investigating whether text can predict upcoming rating changes. The methodology involved locking the model and conducting falsification checks beforehand, with half of the markets set aside for confirmatory analysis. Results indicate that review text can act as a leading indicator of displayed reputation, addressing the issue of consistently high ratings. This study is significant for platform design and consumer choices, offering a data-driven perspective on reputation dynamics.