An interesting case in point is New Zealand aid. Before a major reform, its largest program was giving college scholarships to poor Pacific Islanders. I don’t have any decisive evidence on how good they were at doing this, so this example is only suggestive, but a priori a scholarship sounds like a relatively effective way to help somebody help themselves — the ideal formula in aid. Then the “SHOULD” nannies took over, and now the tiny New Zealand budget is divided among ALL the fashionable causes in development. (The picture below shows the breakdown between 37 possible sectors in foreign aid.)I know basically nothing about New Zealand's foreign aid budget. The 2008 picture looks pretty fragmented. In 1999, about 8 areas got three quarters of total funding; it looks like that percentage is now split among about a dozen.
As the New Zealand example shows (and it is characteristic of most aid agencies), the SHOULD criteria defeats the whole idea of specialization, and even tiny agencies wind up giving 5 percent of the budget each to 20 different causes. Since there are fixed overhead costs of operating in a sector (like employing sector specialists), this means that a lot of the aid budget is going to be wasted on overhead costs.
The usual argument for individuals is that you should only give to one charity: there has to be some charity that does the most good as you see things, and your dollars are unlikely to have anything but marginal effects. Since you're not pushing your target charity downwards along its marginal benefit curve, just stick with one. In country aid, that'll be different, and especially for a country like New Zealand that takes on a regional mission in a bunch of tiny countries. New Zealand's spending could easily become inframarginal on all kinds of projects. If the foreign aid budget increased from 1999 to 2008, it's very plausible that the marginal student for a scholarship just wasn't worth the investment relative to other projects our foreign aid folks could be pursuing. Again, though, I know nothing about the actual make-up of the budget.