Analyzing tourists’ length of stay at destinations with survival models: A constructive critique based on a case study

Abstract

The statistical modeling of tourists’ length of stay at destinations is a topic that recently has received much attention from tourism scholars. In this regard, so-called “survival models” have been introduced as a means of studying how a set of independent variables explain variation in the number of days tourists spend at destinations. This paper provides a critical look at these studies. There are two main findings. (1) The various justifications offered for favoring the survival models over the traditional OLS regression do not hold up under closer scrutiny. (2) An empirical study shows that the OLS regression model describes the association between a set of independent variables and length of stay at least as effectively as a battery of survival models. In line with the principle of parsimony it is concluded that future studies on tourists’ length of stay should abandon survival models if they are conducted along similar lines as the ones to date.

Introduction

Analyzing variation in tourists’ length of stay at destinations with so-called “survival models” (or duration models) has become very popular of late.¹ For this reason, we presently know a great deal about what causes tourists to lengthen or shorten their stays at destinations. Yet a general feature of these survival models is that they are much more statistically complex than more familiar alternatives, such as the OLS regression model. A related issue concerns the coefficients produced by survival models because they are much less transparent than their intuitive OLS counterparts. In other words, the results from survival models are often in need of a transformation into a more accessible metric if they are to be successfully communicated to readers – and especially those not very familiar with statistical reasoning. The purpose of this paper is to show that using survival models in order to analyze tourists’ length of stay at destinations is to make matters more complicated than strictly necessary. Even more so, it will be shown that the reasons for justifying survival models in the social sciences and economics more generally do not apply in the context of tourists’ length of stay. The study’s main conclusion might thus be thought of as an application of the principle of parsimony: Why use complex and demanding statistical models when simpler and less demanding ones do the job just as well?

As indicated above, studies based on survival analysis tend to be loaded with statistical jargon and are thus probably not so intuitive for those tourism scholars who are not very well acquainted with the field of statistics. For this reason a central aim of the present study is also to facilitate these researchers’ knowledge to the point that they can carry out the most basic analyses on their own. To achieve this goal, statistical equations and derivations will be eschewed for the benefit of a more accessible, and hopefully more clarifying, approach. The paper is structured as follows. The next section introduces the basics of the survival model in a non-technical fashion (Section 2), whereas Section 3 briefly reviews the tourism applications of this model that focuses on tourists’ length of stay. Section 4 offers some critical remarks on the practice adopted in the tourism literature. Section 5 is a case study that aims to rectify some of the shortcomings noted in Section 4. Finally, Section 6 concludes and offers some implications for future research.