I have run across an interesting strategy for survey design -- that may interest others as well. A common criticism of surveys (especially those on potentially sensitive subject -- or anything with a social affirmation bias) is that people may not answer honestly and there is no way to ensure that they do. One strategy to overcome this is randomized response.
In randomized response surveys, you ask people about some behavior (lets say that a response indicating that a respondent has exhibited some behavior is a "1", "0" otherwise). The survey tells the respondent to flip a coin. If the coin comes up heads, always say "1." If not, answer honestly. It is impossible to reconstruct whether an individual "1" was the product of a coin flip or actual behavior. However, at a group level one can say that the 50% of "1"s are the product of coin flips, factor this out, and get a population average of the behavior.
I am not sure whether I trust people to actually flip a coin and follow directions (in part because I don't know if they will trust the instructions -- or just say "0" regardless).
Interesting idea, though.
No comments:
Post a Comment