Should government transfers be provided in cash or in-kind? Simple economics clearly states cash is better for welfare. Just think about how cash raises the budget constraint leading to higher indifference curves. With an in-kind transfer, however, you (potentially) impose a constraint on the basket of goods, thus (potentially) reducing utility compared to he cash case. This argument is reminiscent of the deadweight loss of Christmas, discussed a year ago on this blog.
Janet Currie and Firouz Gahvani have a nice survey of reasons to provide transfers in-kind, and the circumstances where they apply. The parternalistic argument is the most obvious and applies when the government wants the recipient to consume specific goods. This would be called for when one wants to make sure children are adequately fed or are followed by health professionals. This violates consumer sovereignty, but it the whole point in this case.
But there are other reasons beyond paternalism. Social programs are often targeted, and in the presence of imperfect information for recipient identification, offering in-kind transfers leads to self-selection. for example, only needy people will show up at a soup kitchen or health clinic providing free services. Also, offering social housing in small apartments offers better selection among recipients than cash allowances. Of course, some needy people may be erroneously screened out. Note, however, that targeting does not necessarily imply overprovision of in-kind goods like paternalism does.
Cash transfers do not encourage poor people to do something about their future, as future cash payment depend on them being poor. Providing in-kind transfers takes care of this problem by forcing them into consuming goods that will help them out (education, health screening, etc.).
Finally, another circumstance where in-kind transfers make sense is when the target recipient is a member of the household that has no decision taking power. Children are such an example in most societies, and women in many as well.