Conjoint Survey Design and Ordered Information

In this paper, we design a conjoint experiment that brings order effects back into the experimental design. While conjoint experiments do away with order effects through randomization of trait ordering, we argue that there are some circumstances in which ordered information can be useful in conjoint experimental survey design. This includes specifically real world circumstances in which prior information affects the way that people make decisions about their preferences. By bringing order effects back into the design, we can more accurately represent real world decision making. We develop a survey design based on information gained from a previous conjoint survey we conducted to analyze how this might look in an updated conjoint design with ordered information. This design can help us understand how the order that respondents see information affects respondent’s preferences and allows for a more accurate estimation of how information learned first and information contained within the conjoint work together to predict preferences. In this paper, we use an example conjoint experiment of voter preferences for the gender and race of a potential U.S. vice presidential candidate based on prior, ordered information about the gender and race of their presidential running mate.