Thanks for the comment, Angela. I think you hit the nail on the head. The way that economist's USED to think of this was by evaluating the welfare LOSSES from conditional cash, which in its purest reincarnation was "in-kind" subsidies. One excellent example of how this can be done in Hanan Jacoby's paper in Journal of Human Resources (http://econpapers.repec.org/article/uwpjhriss/v_3a32_3ay_3a1997_3ai_3a2_3ap_3a223-249.htm).
Since then, the research is primarily interested in using the design of cash versus conditional cash experiments to shed light on market failures that might make conditional cash a more attractive option. One example is that there could be an externality within families. Parents want children to provide for them in their old age. IF they could sign a contract with their kids, they would educate them to the efficient levels. The problem could be that they can't actually sign these contracts and educating them makes it more likely that they will leave the "farm". Then, because of the commitment problem between parents and children, conditional cash works like a "subsidy" to education, in the classic welfare economics sense. This, for instance, is the gist of Rob Jensen's very nice paper that you can find at http://www.econ.brown.edu/econ/events/jensen.pdf.
Similarly, researchers have used designs where money is either given to the mother or father to see whether there are problems due to bargaining within the household. That is, traditional economic theory modeled households as "unitary", so that parents pooled resources and spent them in the "best" possible way. But that's not the way the world works, and bargaining between spouses could change what is consumed. Interestingly, this research does not find large differences depending on the identity of the recipient on household outcomes--but does find that the gender of the person matters when they are running a business. Here is my colleague, David Mckenzie’s paper on this: http://poverty-action.org/sites/default/files/McKenzie-Paper.pdf. Of course, you can complicate this further: If you gave money to the mother and they spent more on health, but if you gave it to the father and they spent it more on education, which one is better? To deal with this vexing problem, economists' have figured out a way to deduce whether the spending is "efficient" and a very nice paper finds that even when consumption decisions vary according to the identity of the recipient, you cannot rule out efficiency. Here is the paper: http://homes.chass.utoronto.ca/~bobonis/GJBmarriage_nov08.pdf.
Using the design of CCT versus UCTs to uncover "deep" parameters or potential market failures is clearly am important task. There are very specific ways to do this and the framing of the question has to be entirely different from what The Economist article tried to do.