Michael Giberson
Learning from randomized controlled experiments: apparently free legal representation provided by the Harvard Legal Aide Bureau tended to delay (by about two weeks on average) receipt of unemployment benefits among a group of claimants appealing an initial disallowance of their unemployment claims. Representation had no effect on the likelihood of a claimant succeeding on their appeal.
We should be careful about leaping to broader conclusions about free legal aid from a single study, but (1) note that a well-conducted randomized controlled experiment is a powerful tool for improving understanding of the world, and (2) additional studies of legal aid are called for in order to discover what helps, what fails to help, and especially when trying to help ends up hurting.
Ian Ayres has additional commentary at the Freakonomics blog.