Quantitative Measurement, Psychometrics, and Scale Construction
Measuring the Unobserved
Operationalization
Many central concepts in social science, such as ideology, class consciousness, trust, and efficacy, are latent constructs that cannot be directly observed. Measurement therefore requires theory-driven operationalization, careful item design, and evidence that scores behave as intended.
Measurement properties
Common scale types
| Type | Best for | Caution |
|---|---|---|
| Likert scale | Attitudes and agreement | Treating ordinal data as interval without justification |
| Index | Combining multiple indicators | Unequal weighting can distort interpretation |
| Latent variable model | Complex constructs | Model fit and identification must be checked |
| Behavioral count | Observable actions | May miss meanings or motivations |
Avoiding construct drift
A measure can become weaker over time if the items no longer capture the same theoretical concept in changing social contexts.
Which statistic is most commonly used as a preliminary indicator of internal consistency for a multi-item scale?
Cronbach's alpha is a common summary of internal consistency, though it should not be treated as a complete validity assessment.
Correct answer: Cronbach's alpha
Why is measurement invariance important in comparative social research?
Comparable scores are essential before interpreting cross-group differences substantively.
Correct answer: Because without invariance, observed group differences may reflect different measurement properties rather than true differences in the construct.
Latent variable thinking
Latent variable models treat observed indicators as imperfect reflections of an underlying construct. In advanced applications, researchers test whether factors load similarly across groups, whether error terms are defensible, and whether the model preserves theoretical meaning.
A measure is most directly concerned with which issue?
Measurement quality is about how well an indicator or scale captures the construct it is meant to represent.
Correct answer: Capturing the intended construct accurately
What is one reason researchers prefer latent variable models over simple summed scores?
Latent models often provide a more theory-consistent representation of complex constructs.
Correct answer: They can model measurement error and the relationship between multiple indicators and an underlying construct.