“Audit studies” have been popular in labor economics research for 10 years. The researcher sends resumés of artificial job applicants in response to job openings. Typically there is a crucial difference in some characteristic of the person that indicates a particular racial/gender/ethnic or other group to which one person within a pair of resumés belongs while the other does not. The differential response of employers to the difference in the characteristic implied by the resumés is taken as a measure of discrimination in hiring.
Is this ethical? Employers are deliberately lied to; and unbeknownst to them they spend time in an activity that cannot lead to a hire. If they look at 2000 resumés, spending only 30 seconds on each, and spend 2 minutes each calling back 10 percent of them (typical in such studies), employers are being asked to spend about 23 hours of their time in this activity. Valuing their time at $50/hour, this part of the time cost of the research, essentially a tax borne by participants who must be kept ignorant for the research to succeed, is over $1000.
I find this line of research to be unethical—it certainly violates the “informed consent” that is always required in biomedical studies. Nor is it even necessary: For example, we learned a huge amount about the impact of racial characteristics, in their case, names, from the excellent study by our own Steve Levitt and his coauthor Roland Fryer.