A growing body of research confirms the long-held belief of parents, school administrators, and policy makers that teachers are the key component to a good education and that there is substantial variation in teacher quality. This research differs fundamentally from prior work on teachers by focusing directly on differences in student learning outcomes across classrooms rather than differences in teacher experience, salary, education, or other quantifiable characteristics. Moreover, it raises questions about studies that draw inferences about implications for teacher quality based entirely on observable characteristics. When looked at in terms of teacher effectiveness, many of the policies commonly discussed are suspect if not wrong. The authors investigate teacher quality differences by transition status and school characteristics for teachers in a large urban district in Texas using estimates of teacher contributions to student learning as their measure quality. Further investigation confirms the finding that teachers who exit the Texas public schools are less effective on average than their colleagues in the same grade regardless of school average achievement or proportion black. More interestingly, the average value-added of school leavers in the sorted sample (described previously) tend to be more negative, particularly in the low proportion black and high achievement samples. This suggests that a difficult classroom placement may both downward bias estimated quality and precipitate a school change, though these results are far from definitive. (Contains 4 notes.)