A big new study (PDF) from the U.S. Department of Education's Institute of Education Sciences performed by Mathematica Policy Research confirms that Teach for America teachers outperform peer teachers in hard-to-staff schools and also indicates that teachers recruiting through the New Teacher Project's Teaching Fellows program are about the same as peer teachers recruited through traditional certification methods.
Since empirical evidence doesn't change anyone's mind about this issue, if you disagree with those conclusions I urge you to click through to the report and find methodological points to quibble with. I believe Page xxi is where you'll find the most relevant quibble-able details.
But I would say the most important takeaways are these. For starters, over the years it's become clear that Teach for America's most important impact on American education probably isn't through the teaching efforts of TFA teachers. Rather, many alumni of the TFA program have continuing careers in the education field. That's sometimes as classroom teachers in traditional public schools, but often as leaders in charter schools or charter school networks or else in policy advocacy organizations. Since many people do not like charter schools or disagree with the policies being advocated by TFA aligned policy advocacy organizations, there's now a TFA backlash. It would be convenient for the leaders of that backlash if TFA's core claim that it's a way of improving the average level of teaching in low-income schools was just some kind of ruse. But we can see here that while it is—in part—a ruse through which new cohorts of education policy reformers are recruited, it's certainly not just a ruse. The existence of Teach for America raises the student's school learning, completely apart from any downstream policy-level consequences.
The Teaching Fellows finding is less of a political hot button, but it has broader and more important policy implications. The basic upshot is that the current certification process has no meaningful validity. On the other hand, we also saw that TNTP's effort to devise a better ex ante screening process failed. They came up with one that's totally different, but about as good. The implication is that we don't really have a great idea about how to ex ante screen for effective teachers. The reasonable response is to say that lowering the bar for becoming a teacher in the first place probably wouldn't do much harm, and raising the bar for staying as a classroom teacher once you've been on the job for two or three years and can be evaluated ex post could do quite a bit of good.