Design-build is proved to be a preferred project delivery method for infrastructure projects, yet how to design an appropriate and fair risk sharing mechanism within design-build coalition (DBC) is essential to achieve project success. Further, despite the well documented literature on risk allocation for infrastructure projects, prior research has largely neglected participants' social preference which is proved to be salience in affecting their attitudes, behavior and, in turn, decision-making process. Accordingly, participants' risk management behavior will be situational, which means they will response to the risk sharing mechanism based on their perception whether the allocation ratio is fair to them. To address the gaps, a quantitative approach is presented to analyze risk sharing arrangement in design-build project by considering DBC member's fairness preference. Fehr and Schmidt's inequity aversion (IA) model is integrated into the proposed risk allocation model. The objective of this paper is to derive results for DBC leader's optimal risk-sharing ratio and DBC members' optimal risk-management effort simultaneously. The derivation is based on solving a restrained optimization problem using the conception and methods from Stackelberg game theory. Analysis results show that:(1) DBC members are prone to different fairness preference (in terms of envy preference and sympathy preference) depending on the value of risk-sharing ratio; (2) DBC members' optimal risk investment decreases with the enhancement of IA level when DBC member is envy preference; (3) DBC members' optimal risk investment increases with the enhancement of IA level when DBC members are sympathy preference; (4) It is beneficial for DBC leader to allocate more risk-sharing ratio to DBC members as their levels of envious preference increase; (5) DBC leader can allocate less risk-sharing ratio to DBC members as their levels of sympathy preference increase.Practically, the article will benefit for those who write DBC negotiation with recommendations on risk allocation strategies. Theoretically, this research sheds a light on establishing optimal risk allocation via considering members' social preference, and fills the gaps where traditional risk allocation models are based on the hypothesis of completely rational person. Accordingly, literature is enriched by providing mathematical evidence on designing a fair risk allocation strategy and, in turn, future research directions are provided for scholars to explore empirical evidence to support notions proposed in this paper.