Singer promotes giving to Oxfam in public speeches because it's easier for most people to understand the direct benefits of their work, but in private conversation he agrees that it is far better to donate to meta-charities. For instance, you can donate directly to the Poverty Action Lab, which conducts rigorous controlled, randomized studies to evaluate the effectiveness of interventions, often finding that billions of dollars are being wasted at low cost... Another strong candidate meta-charity would be the start-up nonprofit GiveWell, which seeks out the most demonstrably effective charities at improving human welfare in a transparent published research process to guide individual donors.
It's an intriguing idea: rather than trying to directly help the less fortunate, we would do better to serve as a 'catalyst' that boosts the effectiveness of others' giving. (Of course, it would defeat the purpose if everyone went the 'meta' route, as Brandon notes. But there's little risk of that in practice. And as things stand, it appears that we would do best to shift the balance at least slightly more in the 'meta' direction.)
Though it raises the question: are there meta-meta-charities to help us decide which of the various meta-charities is most worth donating to?