How to Recover Individual Preferences From Conjoint Experiments

🧭 What This Paper Does

Conjoint experiments are increasingly popular for studying multidimensional political preferences, yet common practice—estimating average marginal component effects (AMCEs) across subgroups—does not capture the full variation of preferences within a population. The paper proposes a procedure to estimate respondent-specific marginal component effects (individual MCEs), filling a methodological gap where no accepted technique previously existed.

🛠️ How Individual Estimates Are Constructed

A new estimation strategy is introduced that retrieves respondent-level component effects from conjoint responses. Key features:

• Estimates respondent-specific marginal component effects (individual MCEs).
• Does not require additional assumptions beyond those used in standard conjoint analysis, though some recommended changes to task design are discussed.
• Presents approaches to quantify and account for uncertainty in the resulting individual-level estimates.

📊 Empirical Illustration: Immigrant Admission Conjoint

A partial replication of an existing conjoint experiment on immigrant admission is conducted with the recommended design adjustments. The replication serves to show how the procedure performs in practice and to illustrate the kinds of individual-level insights the approach yields.

🔍 What Individual MCEs Make Possible

Using the respondent-specific estimates, the paper demonstrates several substantive applications:

• Mapping the distribution of preferences across respondents rather than relying on subgroup averages
• Examining intercorrelations between different preference dimensions
• Linking individual preference profiles to other respondent characteristics and outcomes

📌 Why It Matters

Recovering individual MCEs gives researchers direct access to within-population heterogeneity that AMCEs obscure. This enables richer descriptive and inferential work on how multidimensional preferences co-vary and relate to other variables of interest, expanding the analytical leverage of conjoint experiments.