Robust Discriminant Validity
Robust Discriminant Validity Assessment · Also known as: HTMT criterion, heterotrait-monotrait ratio, discriminant validity testing, RDV
Robust discriminant validity assessment determines whether distinct latent constructs in a measurement model are sufficiently different from one another. Unlike traditional AVE-based approaches, robust methods such as the Heterotrait-Monotrait (HTMT) ratio use the pattern of inter-indicator correlations to provide a more sensitive and simulation-validated criterion for judging discriminant validity in structural equation modeling contexts.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use robust discriminant validity assessment whenever you are validating a multi-construct measurement model, particularly in PLS-SEM or CB-SEM studies where two or more reflective scales may measure overlapping content. It is especially important when constructs are theoretically related (e.g., anxiety and depression, satisfaction and loyalty) because traditional Fornell-Larcker criteria are known to be too lenient in these cases. Do not rely solely on AVE comparisons or correlation matrices; supplement them with HTMT and bootstrap intervals. Avoid applying HTMT mechanically to formative constructs, as the logic of within-construct indicator correlations (monotrait correlations) does not apply the same way to formative measurement.
Strengths & limitations
- More sensitive than the Fornell-Larcker criterion — simulation studies show HTMT detects discriminant validity failures that AVE comparisons miss.
- Provides an inferential test via bootstrap confidence intervals, not just a descriptive point estimate.
- Grounded in the multitrait-multimethod logic of Campbell and Fiske, giving it strong theoretical foundations.
- Applicable across CB-SEM and PLS-SEM frameworks, making it versatile for diverse research designs.
- Transparent calculation from the indicator correlation matrix — can be computed by hand or in standard SEM software.
- HTMT was developed and validated primarily for reflective constructs; its interpretation for formative or composite constructs is less straightforward.
- Requires a sufficiently large sample for bootstrap confidence intervals to be stable; small samples may yield wide intervals that are uninformative.
- The choice of threshold (0.85 vs. 0.90) is a heuristic that depends on the theoretical relatedness of constructs, introducing some subjectivity.
Frequently asked
What is the difference between HTMT and the Fornell-Larcker criterion?
The Fornell-Larcker criterion compares the square root of each construct's AVE to its correlations with other constructs and requires the former to exceed the latter. Simulation studies by Henseler et al. (2015) showed this criterion has low sensitivity and frequently fails to detect genuine discriminant validity problems. HTMT uses the full pattern of inter-indicator correlations and is statistically more sensitive, making it the preferred modern criterion.
What threshold should I use for HTMT?
A threshold of 0.85 is recommended when the two constructs being compared are conceptually distinct and should not share much variance. A threshold of 0.90 is acceptable when constructs are theoretically related but still expected to be distinguishable. For formal inference, combine the threshold check with a bootstrap confidence interval: if the upper bound of the 95% CI falls below the threshold, discriminant validity is supported.
Does HTMT apply to formative constructs?
HTMT was designed for reflective constructs, where indicators are assumed to be caused by the latent variable and should correlate with one another. For formative constructs, where indicators are assumed to cause the construct and need not correlate, the monotrait correlation logic underpinning HTMT does not apply directly. Different validity checks — such as VIF for multicollinearity among indicators and external criterion validity — are recommended for formative constructs.
What should I do if HTMT is too high?
First inspect the cross-loadings and indicator correlations to identify which items are blurring the boundary between constructs. Options include removing overlapping items, revisiting the conceptual definition of the constructs, collapsing two theoretically similar constructs into one, or collecting new data with better-targeted items. A high HTMT may also signal a genuine theoretical problem — the constructs may not be as distinct as assumed.
Can HTMT be used in CB-SEM as well as PLS-SEM?
Yes. Although HTMT was popularized in the PLS-SEM literature, it is computed from the indicator correlation matrix and is therefore applicable in CB-SEM contexts as well. The same thresholds and bootstrap interval procedure apply regardless of the estimation method used.
Sources
- Henseler, J., Ringle, C. M. & Sarstedt, M. (2015). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. DOI: 10.1007/s11747-014-0403-8 ↗
- Campbell, D. T. & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56(2), 81–105. DOI: 10.1037/h0046016 ↗
How to cite this page
ScholarGate. (2026, June 3). Robust Discriminant Validity Assessment. ScholarGate. https://scholargate.app/en/psychometrics/robust-discriminant-validity
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Confirmatory factor analysisPsychometrics↔ compare
- Construct ValidityPsychometrics↔ compare
- Convergent ValidityPsychometrics↔ compare
- Structural Equation ModelingResearch Statistics↔ compare