Conjoint analysis
Estimate preferences across combinations of attributes
Conjoint analysis presents profiles with systematically varied attributes to study multidimensional preferences. Traditional rating or ranking designs estimate part-worth utilities; choice-based designs model selections. Randomised conjoint experiments can estimate average marginal component effects under explicit assumptions. Attribute levels, profile realism and repeated choices determine what the estimates mean and where they generalise.
Use when people trade off several product, policy or service attributes. Ensure feasible profile construction, ethical presentation and a respondent population relevant to the intended preference or causal estimand.
Strengths
- Reveals trade-offs across several attributes
- Randomised attributes can identify well-defined component effects
Limitations
- Hypothetical choices may differ from real behaviour
- Attribute distributions and implausible combinations affect interpretation
Know the boundary
An attribute effect averages over the profile distribution studied; it is not an intrinsic universal preference.