Purpose: The purpose of this paper is to determine whether teachers with high value-added scores (as a measure of teacher quality) stay or left test grades and subjects in a medium-sized school district. Design/methodology/approach: Panel data for this paper encompass teachers providing math and reading instruction and link to individual students in grades 3-10 from a single Florida school district (2000-2001 to 2004-2005). Value-added modeling is used to estimate a measure of teacher quality, which is entered into binomial logistic regression models. Findings: This paper finds a negative relationship between reading teachers' value-added scores and attrition (p less than 0.05): a finding consistent with the few that have examined the relationship between value added and teacher attrition. A significant relationship is not found between math value added and attrition. There is also no significant relationship between value added and transferring. Secondary and alternatively certified teachers are more likely to exit tested grades/subjects. Classroom percentages of students enrolled in the free/reduced lunch program (a proxy for poverty) are associated with leaving among math and reading teachers. Practical implications: Not all turnover is negative. Evidence from this paper suggests that schools are not losing the best teachers from tested subjects and grades: those in which schools and school leaders are held accountable. While there are costs associated with turnover, it can serve as an important matching function between workers and employers. Originality/value: Only, a few published studies have utilized value-added scores as the measure of teacher quality and tested their relationship with teacher attrition. (Contains 7 tables, 1 figure and 8 notes.)