An efficient method for modelling tourists' length of stay. (December 2016)

• Record Type: Journal Article
• Authors: Santos, Glauber Eduardo de Oliveira
• Other Names: Li ShiNa guest-editor.; Pratt Stephen guest-editor.; Song Haiyan guest-editor.
• Is Part Of: Tourism economics. Volume 22:Number 6(2016)
• Page Start: 1367
• Page End: 1379
• Publication Date: 2016-12
• Subjects: duration model -- duration of the trip -- generalized linear model -- length of stay -- regression -- tourist behaviour; Tourism -- Periodicals; 338.4791
• DOI: 10.5367/te.2015.0490
• Languages: English
• ISSNs: 1354-8166

Abstract

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.