Forecasting initial margin requirements: A model evaluation

The introduction of mandatory margining for non-cleared derivatives portfolios has major implications for the pricing and risk measurement of over-the-counter (OTC) derivatives. In particular, a model for estimating future initial margin requirements is necessary to enable the calculation of relevant pricing adjustments, net counterparty credit exposures and credit capital requirements. Existing literature on the topic suggests a model that makes use of regression techniques, but little detail is available on the predictive quality of these models within a Monte Carlo simulation framework. We review these regression-based initial margin models in detail and compare their output against the actual margin requirements measured by the International Swaps and Derivatives Association standard initial margin model (ISDA SIMM) methodology. We observe that the models generally perform well for single trades but show some degradation for larger diversified portfolios. We investigate potential extensions and improvements to the model, along with examining some additional ‘conservatism’ features that may have application in the context of credit exposure measurement. The initial margin modelling approaches discussed here are similarly applicable to centrally cleared or exchange-traded portfolios.