The purpose of this report is to develop a strategy for improving both national and state forecasts of future attrition rates. A theory of teacher attrition is developed that attempts to account for the disparate reasons for attrition, and to explain the patterns of attrition unique to each life cycle and career stage. Existing literature on teacher attrition and present attrition patterns from several states are selectively reviewed for the purpose of testing hypotheses derived from the theory of teacher attrition. Existing data available to support improved attrition models are reviewed, and recommendations are made for better use of these data. Finally, sampling and data collection strategies are identified that will improve the value of data collected in a future national survey. (JD)