ESTIMATING AND ANALYZING COTTON SPOT AND FUTURES PRICES VOLATILITY WITH ASYMMETRIC-ERROR GARCH MODELS by Mohamadou L. Fadiga 1 , Sukant Misra 1 , and Octavio Ramirez 2 Abstract This paper empirically confirms and illustrates recent theoretical results about the effects of error term misspecification on GARCH/ARCH models, showing that erroneously assuming normality can result in an incorrect determination of the order of the time-series autoregression, of the explanatory variables affecting the conditional mean and variance, and of the ARCH and GARCH parameter values; which produces unreliable estimates of the volatility series and impulse response functions often used for risk analyses and the development of risk management strategies. The recently developed A-GARCH model is proposed as an alternative to alleviate these problems. Key words: asymmetric error, mean-reversion, non-normality, policy changes, volatility estimation. 1 1 Graduate Research Assistant and Professor at the Department of Agricultural and Applied Economics of Texas Tech University, Box 42132, Lubbock, TX 79409. 2 Professor and Head of the Department of Agricultural Economics and Agricultural Business at New Mexico State University, Box 30003, MSC 3169, Las Cruces, NM 88003-8003.