Insights into tourism demand and tourism behavior: Four papers using multiple perspectives and structural equation modeling

Abstract

Goal and objectives of the dissertationThis work explores a significant issue in tourism - namely, tourism demand at a destination - as far as both academic research and tourism practice are concerned. The complexity of this issue demands multiple-perspective research based on scientific research methodologies. This study seeks to address this need by employing the structural equation modelling (SEM) approach to further examine tourism demand from both tourism development and tourist behaviour perspectives.Chapter 1 aims to identify the best practices when applying SEM in order to develop appropriate guidelines that are subsequently used to test the various proposed tourism demand models. Chapter 2 aims to advance the field's knowledge regarding the relationship between supply side variables used in tourism-based studies by extending existing tourism demand models to include additional factors - namely, environment and society variables. The chapter examines whether and how economic, infrastructural, social, and environmental variables impact tourists' inflow into a specific destination at country aggregate levels. Chapter 3 extends the analysis from Chapter 2, examining whether the causal relationships defined in Chapter 2 hold true across both developed and less developed countries equally. Finally, Chapter 4 further advances the literature on tourists behaviour (as opposed to aggregate demand level) by examining which customer-level factors (and to what degree) affect tourists'' immediate and future intentions to return to the destination. Thus, this study makes significant contributions to existing tourism literature.MethodologyThe present work applies different types of existing SEM models (e.g., traditional SEM with reflective constructs, multi-group constrained models, and latent growth curve models) to operationalize a variety of variables used in tourism demand forecasting. These models can also depict causal relationships among these variables to generate further results in tourism studies and advance theoretical development in tourism demand modeling at both the aggregate and individual (tourist behavior) levels.Chapter 2 uses AMOS 16.0 to test causal hypotheses between supply-side factors and tourism construct by means of traditional SEM model with reflective constructs using a cross-sectional data sample collected from a Euro monitor for 162 countries, for the year period 2004. Data from 2004 are used, as this dataset has fewer missing values compared to data available for 2005 onward. Chapter 3 uses advanced multigroup analysis sampling in AMOS 16.0 to test the same model across two groups of countries (developed vs. less-developed countries) with the same dataset collected previously for the 162 countries for 2004.Finally, Chapter 4 uses advanced Latent-Growth sampling in AMOS 16.0 to validate the proposed model on the effects of novelty-seeking, destination image, and overall satisfaction levels across intent to revisit trajectories, using data collected online in February 2009, among French, English, and German travelers, who flew for more than two hours to visit a sun-and-sand destination during the past six months.ResultsResearch results and findings from this work are specific to each chapter, and they help enhance both researchers' and practitioners' comprehension of the phenomena of tourism demand.In particular, results from Chapter 1 indicate that the use of SEM is still limited among tourism demand modeling, as only 21 studies were found to have used SEM in the past 12 years. Results also indicated that SEM techniques were applied inappropriately. Issues included the following: (1) the majority of authors failed to specify how they handled the issue of missing data; (2) few papers discussed data distribution (normality), although all used traditional estimation methods to fit the model, which assumes dataset normality; (3) all papers could use alternative fit indices to further support overall model fit; (4) only half of the studies tested for reliability and validity of individual constructs, even though not doing so could lead to specification errors; (5) most of the authors re-specified their model without providing a theoretical justification; and (6) few studies provided enough detail to permit the work to be replicated. …