In selecting a hotel room, guests evaluate specific room characteristics. After their stay, these characteristics are used as reference levels in deciding which hotel room to stay in the future. According to prospect theory, the gains (i.e., upgrade or improving) and losses (i.e., downgrade or worsening) with respect to the reference level are perceived differently by individuals. In particular, losses are weighted more than gains. This research investigates the asymmetric preference in hotel room choice by performing a stated choice experiment among the guests of a hotel in Hong Kong. A mixed logit model is estimated by deriving different coefficients for improving and worsening conditions, and attribute-specific cluster analyses are performed to identify those segments with similar preferences. The results confirm the validity of reference-dependent specification in hotel room choice and provide insights for revenue managers in relation to their aim to maximize revenue for repeat guests.