Common Method Bias

When one respondent reports both the independent and dependent variables in one sitting, part of any observed relationship may reflect the shared method (mood, social desirability, consistency motives) rather than the constructs.

Defences are procedural (separating measurement, assuring anonymity, mixing item formats) and statistical (Harman's single-factor check, marker variables, common-latent-factor tests). Journals in management now expect an explicit CMB section — address it before reviewers raise it.

Harman's single-factor test is the most widely reported check and also the weakest. You load every item into an unrotated exploratory factor analysis; if one factor explains less than 50% of the variance, method bias is considered unlikely to dominate. Reviewers increasingly treat it as necessary but not sufficient. Stronger options include a marker variable — a construct theoretically unrelated to your model whose correlations estimate the method effect — and, in PLS-SEM, the full collinearity assessment, where variance inflation factors above 3.3 flag possible contamination.

The best defence is built in before data collection, not tested for afterwards. Collect predictor and outcome variables at different time points or from different sources where the design allows — supervisor-rated performance rather than self-rated, for instance. Assure respondents that there are no right or wrong answers, keep the survey anonymous, avoid ambiguous item wording, and separate related scales within the questionnaire. Describe these procedural remedies in your methodology chapter and the statistical check in your results chapter.