An efficient method for modelling tourists' length of stay

Modelling tourists' length of stay imposes relevant challenges to researchers. The log-linear ordinary least squares (OLS) regression is a simple but limited statistical technique due to its log-normal distributional assumption. Duration models allow more flexible distributional assumptions, but they create unnecessary statistical complexity. Considering these limitations, a third and most efficient alternative that joins statistical simplicity and distributional flexibility is examined in this article. Generalized linear models (GLMs) are used to explain tourists' length of stay at each of 100 Brazilian destinations. The log-gamma distribution is shown generally to fit data better than the log-normal alternative, indicating that GLMs outperform the log-linear OLS without requiring the complexity of duration models.