Variance models for project financial risk analysis with applications to greenfield BOT highway projects

Assessment of BOT project financial risk is generally performed by combining Monte Carlo simulation with discounted cash flow analysis. The outcomes of this risk assessment depend, to a significant extent, upon the total project uncertainty, which aggregates aleatory and epistemic uncertainties of key risk variables. Unlike aleatory uncertainty, modelling epistemic uncertainty is a rather difficult endeavour. In fact, BOT epistemic uncertainty may vary according to the significant information disclosed during the concession period. Two properties generally characterize the stochastic behaviour of the uncertainty of BOT epistemic variables: (1) the learning property and (2) the increasing uncertainty property. A new family of Markovian processes, the Martingale variance model and the general variance model, are proposed as an alternative modelling tool for BOT risk variables. Unlike current stochastic models, the proposed models can be adapted to incorporate a risk analyst's view of properties (1) and (2). A case study, a hypothetical BOT transportation project, illustrates that failing to properly model a project's epistemic uncertainty may lead to a biased estimate of the project's financial risk. The variance models may support, guide and extend the thinking process of risk analysts who face the challenging task of representing subjective assessments of key risk factors.