Convergent Validity

The conventional threshold is AVE ≥ 0.50, established by Fornell and Larcker (1981). Some reviewers accept values as low as 0.40 when composite reliability exceeds 0.60, but this tolerance is not universal. Values below 0.40 almost always indicate a problematic scale requiring item revision.

Reliability (e.g., Cronbach's alpha, McDonald's omega) measures internal consistency — the degree to which items produce stable, repeatable scores. Convergent validity is a construct validity concept: it asks whether the items actually measure the intended construct. A scale can be internally consistent yet still measure the wrong thing, so both must be evaluated.

Yes, but with less precision. In purely exploratory contexts, high inter-item correlations, a clear single-factor EFA solution, and strong item–total correlations all provide informal convergence evidence. However, the AVE statistic is only meaningful from CFA because it requires constrained, construct-specific loadings.

First inspect individual item loadings: drop or revise items with loadings below 0.50, check for cross-loadings, and consider whether the construct definition is too broad. Recollecting data with improved items is preferable to lowering the threshold in the paper.

No. The multitrait–multimethod matrix approach of Campbell and Fiske (1959) examines correlations across both traits and methods and remains a rigorous gold standard. Correlation with well-established criterion measures (criterion validity) also provides convergence evidence. AVE is simply the most compact and commonly reported index in SEM-based research.