Sure, but it isn't necessary to withhold aid in order to do a comparative study. "All" that's required is that agencies maintain useful records, and conduct post-event cost/benefit studies on what was done, then interrogate the data - "for Event Y, which had conditions A, B, and C, we used approaches a, b, and the results were Z. For Event X, which had conditions ..." etc. Then matricize that whole bad boy, and sift out common themes using Ragin's Qualitative Comparative Analysis.
Dr Chris Paul (RAND) used that to figure out which counter-insurgency approaches work, and which don't. The same methodology would, I think, work in figuring out what aid approaches work.
But, as you mention in another post, many agencies shy away from even modest attention to book keeping, which would make such a study somewhat tricky.