Recent Issues in Marketing Scale Testing

Abstract

Structural equation models are becoming increas-ingly popular in a growing number of areas of scientificinquiry, including marketing (Baumgartner and Hom-burg 1996). Such models permit researchers to quicklytest a wide range of hypotheses concerning the relation-ships among any combination of manifest and latentvariables. However, employing a structural equationmodel to test a measurement scale can in many cases beproblematic, especially in a marketing context wheremultidimensional scales may contain many variables.There is now evidence that testing a scale throughstructural equation modeling methods can easily leadresearchers to question the value of a new or existingscale. The reasons why tests of a measurement scalemight be expected to reject a scale or model are: (1)overly high statistical power (e .g ., MacCallum, Browne,and Sugawara 1996); and (2) the magnitude of thecorrelations ofitems between factors (Fornell andLarcker198l;Hu, Bentler, andKano 1992; McQuittyandZumbo1998A