Despite the availability of various probabilistic and non-probabilistic methods for determining budget contingency, still many large capital projects are suffering from cost overrun. Furthermore, most of the effort has been concentrated on calculation of contingency for a single project rather than a group of projects. This paper introduces a new Bayesian-based model for allocating contingency budget to a portfolio of correlated construction projects. The proposed model enables an owner agency to define the individual project confidence level for contingency calculation taking into account the portfolio budget and the desired portfolio confidence level. The model recognizes the correlation between each pair of projects in the portfolio and calculates the required increase in budget in such a way to ensure adequate budgets with respect to individual projects and the portfolio. Using the information from newly completed projects, the Bayesian technique can be used to update the model parameters periodically so that more accurate contingency budget can be established for the portfolio. A numerical example is presented to show the application of the model on a portfolio of transit projects. The proposed model can be employed as an effective tool for the owner agencies in charge of funding a group of projects every year.