Test design under voluntary participation and conflicting preferences✩

At issue is the design of a test which can generate a signal about an agent’s quality when the agent is imperfectly informed and his participation in the test is voluntary. The designer strives for information generation, whereas the agent strives for a high quality perception and is averse to information generation. When quality is binary, the optimal test is binary and not subject to false positives. For an extended setting with non–binary quality types, a possibly non–binary adapted version of such a test is optimal. JEL classification: D82, D83