Skip to content
Method

Common Method Bias (CMB / CMV)

Common method bias (CMB), or common method variance, is artificial covariance that arises when the same respondents answer all your questions with the same method at the same time. It can inflate or deflate the correlations between constructs. Harman's single-factor test flags a problem when one factor explains 50% or more of the variance, but the test is weak and should not be your only evidence. Stronger checks model the method variance directly: a common latent factor (ULMC) or, preferably, a measured marker variable.

common method bias

What it is

Common method bias (CMB), or common method variance, is the artificial covariance that can arise when the same respondents answer all your questions with the same method at the same time. It can inflate or deflate correlations between constructs and threaten the validity of your conclusions. You guard against it first by design — different scales, anchors, sources and timing — and then you screen for it statistically. Harman's single-factor test is the most cited screen but the weakest: it rarely detects bias and should not be treated as proof. Stronger options model the method variance directly — an unmeasured latent method construct (the common latent factor / ULMC approach) or, best of all, a measured marker variable theoretically unrelated to your constructs (Podsakoff et al., 2003; 2024).

When to use it

  • All your data come from the same respondents through one self-report survey.
  • Your predictor and outcome are measured with the same method at the same time.
  • A reviewer asked you to address common method bias.
  • You want statistical evidence to complement procedural remedies.

When not to use it

  • Your predictor and outcome come from different sources or times — the risk is already low by design.
  • You only have objective or behavioral measures rather than self-report.

Assumptions

  • A self-report dataset where method variance is plausible.
  • Enough indicators to fit a measurement model (for the CLF approach).
  • Procedural remedies were considered first — statistical screens are a complement, not a cure.

Thresholds & cutoffs

CriterionRule of thumbSource
Harman's single factorA single factor explaining < 50% of variance is reassuring, but the test is weak — do not rely on it alonePodsakoff et al. (2003; 2024)
Common latent factor (CLF / ULMC)Compare loadings with vs without an unmeasured latent method factor; large shifts signal biasPodsakoff et al. (2003; https://doi.org/10.1037/0021-9010.88.5.879)
Marker-variable techniqueUse a theoretically unrelated marker to partial out method variance (the preferred statistical control)Lindell & Whitney (2001); Podsakoff et al. (2024)

How to report it in APA

APA template — copy & adapt

Common method bias was assessed first with Harman's single-factor test as a preliminary check; the first factor accounted for XX% of the variance, below the 50% threshold. Because that test is not conclusive, a common latent factor was then added to the measurement model, and standardized loadings shifted by less than .XX, indicating that common method bias was unlikely to threaten the results (Podsakoff et al., 2003).

How Explonia runs it

Explonia screens for common method bias in the browser on the data you collect or upload. It runs Harman's single-factor test and reports the variance explained by the first factor, and it can add a common latent factor to your measurement model to compare loadings with and without it. The results draft an APA 'Article description' you can review and edit. Because CFA, reliability and HTMT live alongside it, you can present validity and method-bias evidence together. Export to SPSS (.sav) and R is available on paid tiers.

Results verified against R (lavaan)

Frequently asked questions

Is Harman's single-factor test enough for common method bias?

No. Harman's single-factor test is widely regarded as insufficient — it has low power to detect bias and is at best a minimum, not proof. Pair it with procedural remedies and a stronger statistical control such as the common latent factor (ULMC) or, preferably, a measured marker variable (Podsakoff et al., 2003; 2024).

What does the common latent factor (CLF) approach do?

It adds a latent factor connected to every indicator and compares the loadings with and without it. Small changes suggest method bias is limited; large changes suggest it is a concern.

How do I prevent common method bias before analysis?

Design helps most: separate predictor and outcome by source or time, vary scale anchors, protect anonymity, and word items clearly. Statistical screens come after these procedural remedies.

How much variance from one factor is too much?

In Harman's test, a single factor explaining 50% or more of the variance is the common red flag (Podsakoff et al., 2003). Below that is reassuring but not conclusive on its own.

Last updated: September 2026

Run CMB on your own data

Collect or upload your data, run the analysis in your browser, and get a draft APA paragraph you can edit. Free to start. No credit card.

Start free