Using a dataset covering over 10,000 Australian school teachers and over 90,000 pupils, I estimate how effective teachers are in raising students' test scores. Since the exams are biennial, it is necessary to take account of the teacher's work in the intervening year. Even adjusting for measurement error, the teacher fixed effects are widely dispersed, and there is a strong positive correlation between a teacher's gains in literacy and numeracy. Teacher fixed effects show a significant association with some, though not all, observable teacher characteristics. Experience has the strongest impact, particularly in the early years of a teacher's career. Female teachers do better at teaching literacy. Teachers with a master's degree or some other form of further qualification do not appear to achieve significantly larger test score gains. Overall, teacher characteristics found in the departmental payroll database explain only a small fraction of the variance in teacher performance. (Contains 3 figures and 6 tables.)