Convergent vs. Discriminant Validity: Which Is More Important?

Convergent vs. Discriminant Validity in SEM

If I had to prioritize one type of validity in structural equation modeling, I would consider convergent validity to be slightly more important. Convergent validity ensures that items intended to measure the same construct are strongly related, which is fundamental for establishing a reliable measurement model.

In my study on refusal strategies in money-related requests, for example, items measuring indirect refusal strategies must show high loadings on the same factor. If these items do not converge, it would indicate that the construct itself is not clearly defined, making further analysis unreliable. According to Hair et al. (2019), adequate convergent validity (e.g., AVE ≥ 0.5, CR ≥ 0.7) is essential before evaluating other aspects of the model.

However, weak convergent validity can lead to serious issues, such as inconsistent measurement and invalid conclusions. Even if discriminant validity is achieved, the model would still be problematic if the constructs themselves are not measured accurately.

That said, discriminant validity is also crucial, as it ensures that constructs are distinct. Without it, overlapping constructs could lead to redundancy and misinterpretation. Therefore, both should be considered together, but convergent validity forms the foundation.