Causal Inference With Regression: Leveraging Graphical Models in Tourism and Hospitality Context

Abstract

Regression models have long been a cornerstone of social science research, including in the tourism and hospitality field, providing flexible and versatile tools for analyzing relationships between constructs and testing theories. However, growing concerns have emerged over the misuse of regression models for causal inference, particularly when attempting to identify causal relationships using observational data. While randomized controlled experiments are considered the gold standard for causal inference, practical and ethical constraints often limit their feasibility, making regression models an important, albeit challenging, alternative in many contexts. In this article, we review a design-based approach to rethink regression models for causal inference and introduce the graphical causal model as an identification tool to support causal interpretations of regression coefficients. By addressing common pitfalls in regression-based causal inference, this study offers guidance for enhancing the credibility of regression analysis in social sciences, including tourism and hospitality.