Construct Validity

Definition

Construct validity is the degree to which a measurement instrument accurately assesses the theoretical concept it is intended to measure. It encompasses two sub-types: convergent validity (the instrument correlates with other measures of the same construct) and discriminant validity (the instrument does not correlate strongly with measures of different constructs).

Why It Matters

In the social sciences, many important concepts — such as intelligence, motivation, or brand loyalty — cannot be observed directly. Researchers must infer them from questionnaires, behavioural indicators, or experimental tasks. Construct validity ensures that these inferences are sound and that the instrument is actually capturing the intended concept rather than something else. Without it, research findings rest on a shaky methodological foundation.

Example

A psychologist creates a new scale to measure "grit," defined as perseverance and passion for long-term goals. To establish construct validity, she shows that the scale correlates strongly with an existing perseverance measure (convergent validity, r = 0.72) but only weakly with a scale measuring momentary mood (discriminant validity, r = 0.12). She also conducts a confirmatory factor analysis showing that grit items load onto a single factor distinct from conscientiousness items. Together, these results support the construct validity of her new scale.

Related Terms

• Content Validity
• Factor Analysis
• Structural Equation Modelling
• Reliability
• Cronbach's Alpha

Software Notes

• SPSS: Convergent and discriminant validity are assessed through correlation matrices (Analyze > Correlate > Bivariate) and factor analysis (Analyze > Dimension Reduction > Factor). For advanced construct validation, use AMOS for confirmatory factor analysis.
• R: psych::cor.ci(data) for correlation matrices with confidence intervals. lavaan::cfa() for confirmatory factor analysis. psych::discriminant() can help assess discriminant validity via the Fornell-Larcker criterion.
• Stata: pwcorr scale1 scale2 scale3, sig for bivariate correlations. sem for confirmatory factor analysis. estat eqgof reports model fit indices that support construct validity claims.