In many areas of education, it is interesting to observe how easily wrong conclusions can be drawn from data. In a recent piece in the New Yorker, Malcolm Gladwell implies that because there is no difference in student performance between certified and non-certified teachers that we should eliminate the certification process altogether. You may label this the "Let them all in and sort them out later" strategy.
It is true that if the bar is set so low that it does not improve quality, you should eliminate it altogether. The bar is useless. But you should only do that if you are getting the quality necessary for great schools. This is obviously not the case.
We need to raise the bar, not eliminate it. The McKinsey study says that selectivity is the key to great schools. Our current teacher certification processes are anything BUT selective. Our Ed schools are not all that rigorous to get through and most state teacher tests have pass rates over 95%.
This is not highly selective. Therefore you would not expect to see much difference between a certified and uncertified teacher.
Choose the right inputs to measure in potential teachers - the ones that lead to greater student achievement (we do). Set the bar high enough to ensure people have the skills and knowledge to perform and you will see better results. Don’t end certification. Make it relevant and make it rigorous.