Simulation-based valuation of project finance : does model complexity really matter?

This paper analyzes the impact of model complexity on the net present value distribution and the expected default probability of equity investments in project finance. Model complexity is analyzed along two dimensions: simulation complexity and forecast complexity. We aim to identify model elements which are crucial for the valuation of project finance in practice. First, we present a simulation-based project finance valuation model. Second, we vary several model aspects in order to analyze their impact on the valuation result. For forecast complexity, we apply different volatility and correlation forecasting techniques, e.g. correlation forecasts based on historical values and on a dynamic conditional correlation (DCC) model. Regarding simulation complexity, the number of Monte Carlo iterations, the equity valuation method, and the time resolution are varied. We find that the applied volatility forecasting models have a strong influence on the expected net present value distribution and on the probability of default. In contrast, correlation forecasting models play a minor role. Time resolution and equity valuation are both crucial when specifying a valuation model for project finance. For the number of Monte Carlo iterations, we demonstrate that 100,000 iterations are sufficient to obtain reliable results. -- Project Finance ; Investment Valuation ; Stochastic Modeling ; Monte Carlo Simulation ; Forecasting ; Model Complexity