Different Explanations for Correlated Disturbance Terms in MTMM Studies

Researchers often introduce correlated disturbance terms into their covariance structure models. This is only a substitute for unexplained residuals. One situation where this happens is in multitrait-multimethod (MTMM) research. Results from MTMM studies show that the correlated uniqueness model-a factor model with correlated disturbance terms-leads to proper solutions in almost all cases. One possible explanation for the observed correlated disturbance terms is the similarity of methods for the different traits. This idea has been explored many times in the past, with methods effects introduced in the model as extra factors. However, there may be quite different explanations for these correlated disturbance terms. This article concentrates on models to explain correlated disturbance terms in survey research. Seven alternative models are presented that can all give an explanation for these correlated disturbance terms. We try to explain the correlated disturbance terms by method effects, relative answers, the acquiescence bias, and variation in response functions. All these models are tested on 7 data sets. The results show that the model that assumes unequal method effects offers the best explanation for the correlated disturbance terms, although other explanations cannot be ruled out entirely.