Kindness capital and rating bias. A sentiment analysis on Airbnb reviews

We study whether the personal interactions between buyers and sellers may bias the rating scores left by reviewers on digital platforms and affect consumer demand. Using data and text reviews from Airbnb in Barcelona in 2019, we perform semantic and regression analyses to measure the host’s kindness and to quantify the impact of the kindness-related bias on listing ratings and demand. To identify the bias, we exploit the listing’s location, an attribute that can be objectively measured through GPS coordinates and cannot be strategically manipulated by the host. We find that kinder hosts receive significantly higher location ratings, given the distance, and higher listing demand, given the overall rating. Furthermore, kindness mitigates the negative impact of a bad location both on location rating and on listing demand. We take advantage of Covid-19 lockdown in early 2020 as a natural experiment to address endogeneity concerns.