Large-scale urban village redevelopment projects (UVRPs) have been implemented in different ways in China in recent years. This study develops an analytical framework for understanding the heterogeneity of institutional arrangements for UVRPs in China. The analytical framework is established from the perspective of neo-institutional economics. Decision-making and funding sources are used as two dimensions to classify the institutional arrangements into four categories: top-down with government funding (TGF), top-down with villager funding (TVF), bottom-up with villager funding (BVF) and bottom-up with private developer funding (BPDF). The comparative analysis reveals that the bottom-up institutional arrangement is generally more effective in political persuasion than the top-down one, especially when villagers’ committees (VCs) that are responsible for the project implementation are elected freely and fairly. When incorporating the dimension of funding sources into the analysis, a TVF institutional arrangement is likely to incur higher transaction costs than a TGF institutional arrangement, leading to the project having a longer duration. However, when private developers are allowed to participate and to develop UVRPs, the transaction costs will increase due to opportunism and information asymmetry. Policymakers should play an active role in arbitration as it would effectively mitigate the size of transaction costs used to deal with clashes between VCs, ordinary villagers and private developers. Based on information for 394 UVRPs collected from five coastal cities in Zhejiang province, the research findings reveal that a TVF institutional arrangement, which has the highest transaction costs, is still dominant in China. This may be the result of policy continuity and path dependence. Institutional change is contingent on the relative costs, which are the sum of the transformation costs and the transaction costs. Incremental institutional change, rather than radical institutional change, is more likely to occur in China.