When do individuals actually improve their financial behavior in response to advice? Using survey data from current defined-contribution plan holders in the RAND American Life Panel (a probability sample of US households), the authors find little evidence of improved DC plan behaviors due to advice, although they cannot rule out problems of reverse causality and selection. To complement the analysis of survey data, they design and implement a hypothetical choice experiment in which ALP respondents are asked to perform a portfolio allocation task, with or without advice. Their results show that unsolicited advice has no effect on investment behavior, in terms of behavioral outcomes. However, individuals who actively solicit advice ultimately improve performance, in spite of negative selection on financial ability. One interesting implication for policymakers is that expanding access to advice can have positive effects (particularly for the less financially literate); however, more extensive compulsory programs of financial counseling may be ultimately ineffective.