Common factors can always be found but can they also be rejected?

Factor models are often used to reduce the complexity of data. It is simple to find common factors and to interpret them. However, it is doubtful whether factor models are always appropriate. Especially when the common factor is considered to be an attitude, it must be stable through time and explain a considerable part of the variances. A procedure is suggested to test the usefulness of a common factor. In an experimental pilot study in which an attitude towards immigrant workers was investigated, the instability and superfluousness of a common factor is shown. An alternative model is presented which suggests that changing opinions can be explained by specific variables. Finally it is shown that a more stable common factor is found if a so-called MIMIC-design is used.