(In)effectiveness of Frame-Inducing Rating Instructions in State-Trait Research.

Abstract

Negative affectivity (Trait-NA) is often measured at baseline and treated as a control variable in models of workplace stress and well-being. However, recent research suggests that routine procedures for the measurement of Trait-NA (i.e., using trait-inducing instructions) may not adequately distinguish Trait-NA from State-NA, leading to probable model misspecification with substantive theoretical implications. We employ Trait-State-Occasion modelling of intensive longitudinal data to examine the degree to which trait-inducing instructions are actually reflective of Trait-NA, as intended. Results indicated that (a) NA, like other often-used measures (e.g., positive affect and workload) exhibit substantial portions of both State and Trait variance, and (b) that rating instructions were generally ineffective in achieving shifts in rating perspectives (5%-15% or so on the average). We discuss implications of using latent variable modelling to distinguish Traits and States in 'shortitudinal research.'