This paper analyzes attrition patterns among teachers in New York City public elementary and middle schools and explores whether teachers who transfer among schools, or leave teaching entirely, are more or less effective than those who remain. We find that the first-year teachers who are less effective in improving student math scores have higher attrition rates than do more effective teachers. The first-year differences are meaningful in size; however, the pattern is not consistent for teachers in their second and third years. Attrition patterns differ between schools having disproportionate numbers of low- vs. high-scoring students. A relatively high percentage of the ineffective first-year teachers in low-scoring schools leave teaching altogether; whereas inefficient first-year teachers in higher-scoring schools disproportionately transfer within NYC. In general, first-year teachers who transfer, on average, are less effective than their peers in their new schools, as was the case in their original schools. Furthermore, the more effective first-year teachers who transfer differentially move to schools with fewer low-scoring, poor, Black and Hispanic students, possibly contributing to achievement gaps. As discussed in the paper, these findings raise important questions about policies and policy proposals intended to reduce teacher attrition. Value-added analysis and estimation of attrition model are appended. (Contains 12 figures, 8 tables and 19 footnotes.) [This research was supported by the National Center for the Analysis of Longitudinal Data in Education Research (CALDER).].