Formative vs. reflective measures: Facets of variation

摘要

Research that treats observed measures of a business construct as a function of a latent true score plus random error is making two strong assumptions. The first, the assumed direction of causality, has generated the burgeoning formative measurement literature to which Cadogan and Lee (2013) contribute. The second is overlooked. Classical test theory assumes there is only one legitimate source of variance. Academics validating measures of business constructs invariably assume that this source is their respondents, who can be managers, employees, or customers, depending on the context (i.e. business discipline). The invoked causality accounts for covariance at the respondent level, ignoring whether it also applies to other sources of variance-such facets as brands, companies, departments, locations, service providers or work groups. Researchers need to be clear about a business construct's conceptual domain, about the sources of variance that are focal to the theoretical relationships being investigated, and whether the construct's relationship with its indicators is formative or reflective for each facet.