Principal evaluation has become a key component of national policy debates on developing effective leaders. To contribute to these debates, this study draws on survey data to explore how principals in Michigan and metropolitan Beijing behaved differently in enacting leadership related to teaching and learning, and how they were evaluated differently by districts in these two regions in 2007-08 (Michigan) and 2008-09 (Beijing). We further combine these two independent samples and build two-level Multivariate Hierarchical Linear Models (HMLM) to estimate the extent to which district evaluation features were associated with principal leadership practice. Descriptive results indicated that Beijing principals were more likely to engage in core activities of instructional leadership (such as direct interactions with students about their learning, observing classrooms, and providing feedback to teachers) than their Michigan peers. Moreover, HMLM results suggest that district principal evaluation can serve as a powerful policy instrument to promote instructional leadership and should emphasize principals' organizational impact on instructional and learning outcomes in determining evaluation purposes, contents, and sources of evidence.