https://hal-emse.ccsd.cnrs.fr/emse-00744939Roustant, OlivierOlivierRoustant3MI-ENSMSE - Département Méthodes et Modèles Mathématiques pour l'Industrie - Mines Saint-Étienne MSE - École des Mines de Saint-Étienne - IMT - Institut Mines-Télécom [Paris] - Centre G2ILaurent, Jean-PaulJean-PaulLaurentSAF - Laboratoire de Sciences Actuarielle et Financière - UCBL - Université Claude Bernard Lyon 1 - Université de LyonAn Empirical Study of the Prices Uncertainties of Some Weather DerivativesHAL CCSD2001weather derivativeuncertaintybootstrap[INFO.INFO-MO] Computer Science [cs]/Modeling and SimulationBreuil, Florent2012-10-24 11:14:502022-09-30 11:10:082012-10-24 11:14:50enConference papers1Weather derivatives provide protection from losses caused by non-catastrophic climatic events. The underlying climatic risks are measured by means of indexes built from available meteorological data. Weather derivatives are then built on one of these indexes, which can therefore be considered as underlyings. In this paper, we consider temperature-based weather derivatives in an actuarial framework and quantify Futures contracts net premia uncertainties. The precision of the computed prices depends on the precision of the parameters estimators of the temperature process and on the importance of that process misspecifications. The aim of the present article is to quantify the effects of those two kinds of uncertainty - estimation errors and model errors - on these weather derivatives prices. Price uncertainty issues are current in other fields of finance. For instance, in option pricing framework, it is well known that one can quantify the uncertainties around the Black-Scholes option price resulting from the estimation error of volatility. Here, two kinds of errors are quantified independently. First, assuming that the model is well specified, the net premia can be computed analytically as a function of the parameters. Thus, approximate confident intervals for the net premia can be deduced from the asymptotic properties of the parameters maximum likelihood estimator. This is roughly the same methodology that was used to quantify the Black-Scholes option price estimators. Secondly, assuming that the estimated parameters are the "true" ones, we evaluate the effects of some model misspecifications using two corresponding bootstrap techniques.