Three Cs in measurement models: causal indicators, composite indicators, and covariates

Area of Science:

• Measurement in social sciences and statistics.
• Methodology in latent variable modeling.

Background:

• Growing attention to causal and formative indicators in the last two decades.
• Existing belief in a simple dichotomy of effect (reflective) versus causal (formative) indicators.

Purpose of the Study:

• To challenge the simplistic dichotomous view of indicators.
• To introduce and differentiate three types of variables influencing latent variables: causal indicators, composite (formative) indicators, and covariates.
• To resolve confusion regarding indicator types, error terms, and coefficient stability.

Main Methods:

• Conceptual analysis and theoretical argumentation.
• Distinguishing between causal, composite, and covariate variables.
• Developing new guidelines for model identification, scaling, estimation, and validity assessment.
• Illustrating points with an empirical example on self-perceived health.

Main Results:

• Proposes a more nuanced classification of indicators beyond a simple causal/effect dichotomy.
• Defines causal indicators (conceptual unity, structural effects), composite indicators (linear combinations, weights), and covariates (control variables).
• Demonstrates how distinguishing these "Three Cs" resolves common methodological questions.

Conclusions:

• The dichotomous view of indicators is inadequate.
• Clear differentiation of causal indicators, composite indicators, and covariates is crucial for accurate measurement and analysis.
• Subject matter expertise is vital for correctly identifying and utilizing these variable types.