Corruption is widely accepted as a major problem plaguing Public-Private Partnership (PPP) projects and annihilating most of their potential benefits. It is believed that the corruption in post-tender phase of PPP is more severe than other phases in the life cycle of the projects. In this paper we study corruption in government-led renegotiation process in the post-tender phase of PPP. A highway serving heterogeneous residents is built by central government through PPP with a contractor in an uncertain environment. After the resolution of uncertainty, a change in design of the highway can be made through renegotiation process in post-tender phase to make all parties, i.e. central government, residents and contractor, to be better-off. The central or state governments cannot interact with each contractor or residents group within their jurisdiction on account of administration cost. Therefore, such decision is delegated to a local official, which leaves the official the opportunity to collude with the other two parties and seek for possible personal gain. We derive a general social optimal design of highway under user heterogeneity is derived with the presence of asymmetric information. Based on the results, outcomes of different renegotiation games with various types of local officials are provided. Our derivation and results provide several important policy implications and regulatory guidelines, including that the official will be more likely to renegotiate the contract if she can collude with both parties than only one party, restricting official’s colluding ability to only part of residents group will reduce social welfare loss, and corruption can be dected through whether renegotiation is declined by the residents group or contractor. Beyond its contribution to PPP literature, our paper enriches the mechanism design and collusion theories with a model containing both consumer and producer. Also, our model links existing economic and engineering models in PPP literature and provides a uniform and useful framework for future discussion.