Policies that require the use of information about student achievement to evaluate teacher performance are becoming increasingly common across the United States, but there is some question as to how or whether to use student test-based teacher evaluations when student assessments change. We bring empirical evidence to bear on this issue. Specifically, we examine how estimates of teacher value-added are influenced by assessment changes across 12 test transitions in two subjects and five states. In all of the math transitions we study, value-added measures from test change years and stable regime years are broadly similar in terms of their statistical properties and informational content. This is also true for "some" of the reading transitions; we do find, however, some cases in which an assessment change in reading meaningfully alters value-added measures. Our study directly informs contemporary policy debates about how to evaluate teachers when new assessments are introduced and provides a general analytic framework for examining employee evaluation policies in the face of changing evaluation metrics. [This paper was published in "Economics of Education Review" v62 p48-65 2018.]