Recent research shows that school principals exert a significant influence over student achievement.  Yet, there is relatively little empirical evidence showing which characteristics and actions of principals that are so effective in improving student academic performance.  In this paper, we first estimate a principal’s value added to student achievement with rich administrative data on elementary school students in North Carolina over 11 years.  Our main specification regresses student test score gains in grades 3-8 on student and school characteristics, and a set of school and principal fixed effects. The principal fixed effects form our estimate of principal value added. We show that the standard deviation of the principal effects is roughly 0.1 in both math and reading, which is similar to teacher value added estimates.  We then relate our measure of principal value added to various characteristics of the principal to determine what characteristics determine principal quality. We show that principal value added is significantly related to both education and experience. Finally, we use the many principal changes within schools across time in our sample to evaluate the effect on various school outcomes of receiving a new principal of high or low quality. We show that principal turnover has a negative effect on student achievement, but that this negative effect is smaller for high-quality principals.  Our results shed light on which principals are high quality, and some of the ways that they are able to improve student performance, which can be used by districts in selecting the appropriate individual to lead a school.