One important issue in measuring teacher success is peer effects–how does the presence of particular subgroups of students impact the classroom? There is quite a body of evidence to suggest that peer effects are significant and are not accounted for in various statistical models that purport to measure teacher effectiveness. Reading several of the VAM critiques listed earlier here will bring up several articles that raise this issue.
After Hurricane Katrina, thousands of New Orleans students relocated to Houston and made Houston schools a perfect laboratory to look at peer effects. In brief, the poor students coming to Houston brought down the success of the schools they entered, and the better-off students raised the achievement of their schools. And, these effects with both academic and behavioral–absenteeism went up among both New Orleans and Houston native students in poor schools, for example. For a full study of these effects, see this study: http://econweb.tamu.edu/common/files/workshops/PERC%20Applied%20Microeconomics/2009_5_6_Scott_Imberman.pdf