As a novel distributed and collaborative learning paradigm, federated learning is widely used in many scenarios with success to train better models via sharing multiple clients’ private data without leaking privacy. In this respect, evaluating clients’ contribution plays a key role in designing incentive mechanisms to motivate higher-quality data owners and also screen malicious clients, and thus is a vital component in maintaining the overall effectiveness and efficiency for federated learning. As a fair evaluation method originated from game theory, Shapley value has been intensively studied and used in evaluating contribution from clients in federated learning, and there is still no survey work on this topic. This motivates our work. In this paper, we presented a comprehensive survey for the current research progresses on Shapley-value-based contribution evaluation for federated learning, and also proposed several potential future research directions. To our knowledge, this is the first survey effort on Shapley-value-based contribution evaluation for federated learning, and we expect that our work can offer useful guidance and reference for future researchers.