She? The Role of Perceived Agent Gender in Social Media Customer Service

Abstract

This work investigated the role of perceived agent gender in customer behavior using a unique dataset from Southwest Airlines’ Twitter account. We inferred agent gender based on the first names provided by agents when responding to customers. We measured customer behavior using three outcomes: whether a customer decided to continue the service conversation upon receiving an agent’s initial response as well as the valence and arousal levels in their second tweet if the customer chose to continue the interaction. Our identification strategy relied on the Backdoor Criterion and hinged on the assumption that customer service requests are assigned to the next available agent, independent of agent gender. The findings revealed that customers were more likely to continue interactions with female agents than male agents and they were more negative in valence but less intense in arousal with the former group than with the latter.