On the Importance of Conducting Construct-Level Analysis for Multidimensional Constructs in Theory Development and Testing

Abstract

On top of not defining explicitly the relationship between multidimensional constructs and their dimensions, many management researchers have conducted their analyses of multidimensional constructs at the dimension level only and assumed that these analyses would be valid at the construct level. This article discusses the potential problems of these practices. Data from two reported management studies are reanalyzed to show that different conclusions might be arrived when analyses are conducted at the construct rather than the dimension level. This finding is especially true when the multidimensional construct is defined under the latent model. Implications are discussed.