[Air-L] Gender in surveys/tests

Hi all, I have always followed the rule of thumb to not ask the demographic questions - gender, race/ethnicity - in the beginning of the survey or a test because it affects how people answer the questions. Can somebody please share some research that supports this? I remember reading research about how test scores are affected when students are reminded about their gender and gender related stereotype. I also remember reading a thread on air listserv where somebody pointed out not to ask gender questions in the beginning of the survey. But cannot find any of the papers or that email thread. If Can anybody share literature that offers reason for not asking gender question in the beginning of the survey? Thanks, Devayani