Research Note: The Determinants of Tourists' Length of Stay: Some Further Modelling Issues

Abstract

Very complicated so-called ‘survival’ (or duration) models have featured strongly in research aimed at explaining variation in tourists' length of stay at destinations. In a constructive critique of this research, Thrane (2012) has shown that use of these models lacks sound footing on conceptual as well as statistical grounds. In recent studies, length of stay has been thought of as a count variable, and the variation in this variable has accordingly been modelled with count data regression models. The purpose of the present study is to provide a constructive critique of this research. There are two conclusions. First, count data regression models should be used when analysing ‘how-many-times-something-happened’ data. Consequently, these models are not ideal when the non-discrete dependent variable is length of stay measured in days. Second, since an OLS regression analysis on the natural log of length of stay yields the same results as a far more complicated count data regression model, the former is preferable on the principle of parsimony and with regard to reaching out to an audience not well acquainted with statistics.