Can anyone point me to some think pieces on what to do with census type data where at least some of the people surveyed do not self-identify a race/ethnicity, gender, or other demographic data?
In studies (and datasets) examining diversity in various communities, I see a wide range of strategies. Some simply remove records where a person declines to specify a race/ethnicity or gender. Some include them and create an "unknown" category. If looking at race/ethnicity, for example, some studies just state they assume all non-responders are white.
None of these solutions seem perfect. Are there any others? And if you remove responses from people who, to take one example, decline to specify a gender, aren't you potentially just erasing part of the population that does not identify as male or female?
A number of surveys now allow participants the option to "decline to state" a gender or "decline to state" a race/ethnicity or some similar option, or to just not answer the question at all, is there any developing consensus on what to do after that data is collected? Does it depend on the population being studied?