Self-selection into social intervention programs may bias the estimates of treatment impact. Data from an Individual Development Account (IDA) program (N = 758) are used to examine the self-selection process. Persons who applied but did not enroll are assumed to have had “second thoughts” about program participation. Multivariate logistic regression predicted second thoughts and showed that having children in the household and negative net worth, along with not owning a vehicle, were positively related to having second thoughts. Those saving for an education were more likely than those saving for a home or business to have second thoughts. Implications for social service administration and impact evaluation of IDAs are shared.