It really depends on the kind of problem. For example, recursive solutions to linked-list/tree/graph problems are typically substantially easier to understand than their iterative counterparts, in large part because the data structures themselves are recursively defined.