Computational thinking (CT) is gaining increasing attention from researchers and practitioners all over the world to empower children in the digital era. However, there is no consensus on which components of CT to teach beginning coders in early childhood education (ECE). To address this issue, we conducted a systematic review of 42 empirical studies focused on teaching and assessing CT in ECE. We analyzed the included studies with the three-dimensional CT framework proposed by Brennan and Resnick (2012) and demonstrated how this framework could be modified to fit the context of ECE. Based on this systematic review, we sorted out the CT components that were proven suitable for young children to learn by incorporating emerging components and removing components inappropriate for young children. We thus proposed a CT curriculum framework for ECE that covers CT concepts (i.e., control flow/structures, representation, and hardware/software), CT practices (i.e., algorithmic design, pattern recognition, abstraction, debugging, decomposition, iteration, and generalizing), and CT perspectives (i.e., expressing and creating, connecting, perseverance, and choices of conduct). This systematic review and its associated CT curriculum framework provide important theoretical contributions and practical implications for early childhood CT education. (author abstract)