The study that I was looking for in my last email about only keeping the top 20% of teachers is at: http://www4.gsb.columbia.edu/rt/null?&exclusive=filemgr.download&file_id=733883&rtcontentdisposition=filename%3DStaiger (slides). The paper is at http://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.24.3.97 and concludes:
The implications of our analysis are strikingly different from current practice.
Schools and school districts attempt to screen at the point of hiring and require
significant investment in education-specific coursework but then grant tenure
status to teachers as a matter of course after two to three years on the job. Performance
evaluation is typically a perfunctory exercise and, at least officially, very
few teachers are considered ineffective (Weisberg, Sexton, Mulhern, and Keeling,
2009). Rather than screening at the time of hire, the evidence on heterogeneity of
teacher performance suggests a better strategy would be identifying large differences
between teachers by observing the first few years of teaching performance
and retaining only the highest-performing teachers.