Common Methods Bias: Does Common Methods Variance Really Bias Results?
通过结构方程模型和元分析，评估了社会科学期刊中共同方法变异对测量结果的影响，发现其导致约26%的偏差，但并未否定多数研究结论。
Methods variance and its effects are at the center of a debate in organizational science. Most of the debate, however, is focused on the prevalence of common methods variance and ignores common methods bias, or the divergence between observed and true relationships among constructs. This article assesses the level of common methods bias in all multitrait-multimethod correlation matrices published over a 12-year period in a set of six social science journals using a combination of structural equation modeling and meta-analysis. The results indicate that only 46% of the variation in measures is attributable to the constructs, that 32% of the observed variation in measures is attributable to common methods variance, and that common methods variance results in a 26% bias in the observed relationships among constructs. This level of bias is cause for concern but does not invalidate many research findings.