I plan to do a conjoint survey experiment in which I randomly change attributes of hypothetical job candidates and show two types of potential candidates with different attributes. Then, I ask respondents to select which candidate they prefer to hire. My aim is to analyze what attribute(s) affects the likelihood for being hired.
Let's say I am mainly interested in four attributes of candidates, 1. gender (male, female), 2. age (30, 40, 50, 55), 3. working experience in the same sector (no, yes). I understand that I should randomly assign these different attributes with different levels to respondents in conjoint experiment.
However, I wonder what I should do with other less important attributes for my research such as education level (BA, MA, Ph.D.), exam score (low, middle, high) in conjoint survey question. Should I also randomly assign these attributes to respondents? Or should I just show a fixed attribute (eg. BA, middle level exam score) as one of the attributes together with the above attributes? Or should I just explain that we assume that all job candidates have BA, middle level exam score, no political connection in survey instruction?
It depends on the hypothesis that you are testing, as well as how much time and effort do you have to account for or counter biases that you've offered a conduit for.
I would recommend to control as many variables as you realistically can to achieve an observation that gives perspective for your conclusion.
Remember, people don't answer accurately for their future in hypothetical self, so a baseline of their historical choices would be very helpful in countering those inaccuracies.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey