Structural equation modelling and regression analysis in tourism research

Robin Nunkoo, Haywantee Ramkissoon

Research output: Contribution to journalArticlepeer-review

164 Citations (Scopus)


This paper explores the concept of structural equation modelling (SEM) and discusses the steps which researchers should follow when using this technique. This involves the development of a theoretical model, testing of a confirmatory measurement model, and evaluating the structural model with hypothesised path relations. For the benefit of readers, the discussion is supported by an illustration of a theoretical model predicting residents' support for tourism, developed on the premise of the social exchange theory. The paper emphasises that the proper application of SEM depends largely on theory, where every step in the analysis is based on theoretical reasoning. The advantages of SEM over regression analysis are discussed and these are grouped in four categories: (1) modelling of measurement errors and unexplained variances, (2) simultaneous testing of relationships, (3) ability to link micro- and macroperspectives, and (4) best-fitting model and theory development. The limitations of SEM over regression analysis are: (1) difficulty in choosing and using SEM software packages; (2) complexity and ambiguity; (3) limited use in exploratory research; and (4) inability to model 'truly' categorical variables. The paper concludes that although SEM has considerable advantages over regression analysis, it does not replace it.

Original languageEnglish
Pages (from-to)777-802
Number of pages26
JournalCurrent Issues in Tourism
Issue number8
Publication statusPublished - 2012
Externally publishedYes


  • Fit indices
  • Regression analysis
  • Structural equation modelling
  • Theory

ASJC Scopus subject areas

  • Geography, Planning and Development
  • Tourism, Leisure and Hospitality Management


Dive into the research topics of 'Structural equation modelling and regression analysis in tourism research'. Together they form a unique fingerprint.

Cite this