Underrepresentation of minority and poor households in scientific studies undermines policy decisions and public health. We study data from a serological study that randomized participation incentives. Participation is low (6% at $0, 17% at $100, 29% at $500) and unequal: minority and poor households are underrepresented at low incentive levels. We develop a framework for disentangling non-contact and “participation hesitancy” in explaining non-participation. We find that underrepresentation occurs because poor and minority households are more hesitant, not because they are harder to contact. The $500 incentive appears to overcome differences in hesitancy and restore representativeness along observable dimensions.