19. SPSS AMOS | Common Method Bias (Part 2) | Latent Common Method Factor

Common method bias can distort results by inflating (or, less commonly, deflating) the true correlations among measured variables—often because respondents answer items for independent and dependent constructs in the same survey session. That distortion can cascade into biased parameter estimates in confirmatory factor analysis (CFA) and subsequent structural modeling. A practical way to detect and control for this problem in IBM SPSS AMOS is to add a latent common method factor that links directly to every indicator in the measurement model.

The approach starts with a baseline CFA run using the original measurement structure. Then AMOS is used to build a second CFA that keeps the original constructs but introduces an additional latent variable—labeled as a “common method” factor—with no indicators of its own. Instead, the common method factor receives single-headed paths to every observed indicator in the model, capturing variance shared across indicators that may come from the measurement method rather than the constructs themselves.

To make the test interpretable, the model constrains the regression weights from the common method factor to all indicators to be equal. In AMOS terms, this means setting the variance/weight parameter for one of the common-method paths and then copying that parameter constraint across the remaining paths using parameter constraints. After estimating the constrained common-method CFA, the key diagnostic is a chi-square difference test between the original CFA and the CFA that includes the latent common method factor.

In the worked example, the original CFA shows χ² = 226.11 with 87 degrees of freedom. The CFA with the latent common method factor shows χ² = 23.769 with 86 degrees of freedom. The difference in chi-square values corresponds to 12.34 with 1 degree of freedom (because the constrained common-method model should add only one effective degree of freedom). The threshold used is 3.84 for 1 df at the conventional significance level. Since 12.34 is far above 3.84, the chi-square difference is treated as significant evidence of common method bias.

When the common method bias test is significant, the remedy is to include the common method factor in the structural analysis as well—so structural relationships among constructs are estimated while statistically controlling for method-driven covariance. To pinpoint where the bias is coming from, the standardized regression weights from the common method factor to indicators can be inspected. Even if the constrained setup keeps unstandardized weights equal, the standardized loadings can still reveal which indicators are most associated with the method factor.

If the chi-square difference test is insignificant, the guidance is to report that common method bias is not a substantial concern at the detected level. In that case, there is no need to carry the latent common method factor into the structural model; the analysis can proceed without it, while documenting the non-significant chi-square difference result.