The Effect of Model Error on the Valuation and Hedging of Natural Gas Storage

How should price models be chosen for real option valuation and how should real option cash flows be hedged in the presence of model error? We consider a specific type of model error in the context of the natural gas storage real option: The assumed number of factors in a multifactor futures price model differs from the true number of factors. In terms of valuation, model error does not seem to be significant. Specifically, the valuations obtained with a family of empirically calibrated seasonal multifactor models of the futures curve are comparable as long as the most important three to six statistical factors are included. In terms of hedging, there are many ways to delta hedge using different combinations of futures contracts. Theoretically, these different hedging approaches are all equivalent without model error. However, they are not equivalent in the presence of model error. In particular, bucket hedging outperforms a näıve implementation of factor hedging, the performance of which can be disastrous. Enhanced factor hedging, in which the particular futures contracts traded are optimized, also performs well provided that a sufficient number of factors are used. Bucket hedging is remarkably robust to model error irrespective of the number of factors assumed, and is easier to implement than enhanced factor hedging. Our findings demonstrate the differential impact of model error on the valuation and hedging of the real option to store natural gas, and have potential relevance beyond this specific application.