Policy changes from the past decade have resulted in a growing interest in identifying effective teachers and their characteristics. This study is the third study to use data from a randomized experiment to test the validity of measures of teacher effectiveness. The authors collected effectiveness measures across three school years from three broad areas: value-added, classroom observation, and student surveys. In the first two years, they collected non-experimental estimates of these measures and, in the third year, they designed a randomized experiment to test the validity of these estimates. Using these data, they answer two questions: (1) Does a combination of these three distinct non-experimental measures identify teachers who, on average, produce higher student achievement gains following random assignment?; and (2) Does the magnitude of the gains correspond with what we would have predicted based on their non-experimental estimates of effectiveness? The analysis sample consisted of 66 fourth- and fifth-grade teachers from four large East coast school districts in the 2010-2011 through 2012-2013 school years. To answer the research questions, the authors first constructed the best linear combination of non-experimental student test score, survey, and classroom observation data from the first two years of the study (2010-11 and 2011-12) to predict teachers' average contribution to student growth on state standardized math tests another year. They used these predicted outcomes as their non-experimental estimates of teachers' contributions to student test score growth in 2012-13. Then, they examined actual student growth in 2012-13 (the third year of the study in which re randomly assigned students to teachers) and compared their non-experimental prediction of growth to actual growth. Tables are appended.