HR: 15:25h
AN: NG12A-08 [PDF]
TI: Coping with climate noise: Long-range dependence and weather derivatives
AU: Jewson, S
EM: stephen.jewson@rms.com
AF: RMS, 10 Eastcheap, London, IL EC3M 1AJ
United Kingdom
AU: * Caballero, R
EM: rca@geosci.uchicago.edu
AF: Department of the Geophysical Sciences, University of Chicago, 5734 S Ellis Av, Chicago, IL 60637 United States
AB:
Random day-to-day changes in weather lead to random year-to-year
fluctuations in monthly and seasonal means, a feature known as
"climate noise". Such climate noise has direct economic impact on a
wide variety of businesses. A typical example is energy vendors, whose
annual revenues are closely correlated with seasonal mean
temperatures. To deal with this risk, a form of insurance known as
weather derivatives has been developed in recent years. We discuss a
Monte Carlo approach to the pricing of weather derivatives based on
stochastic modeling of daily temperature. It will be shown that this
approach can only be succesful if the time-series model correctly
captures the autocorrelation structure of the data even at very high
lags. Evidence will be presented that observed daily temperatures
exhibit long-range dependence, i.e. power-law decay of the
autocorrelation. This means that classical Box-Jenkins ARMA models are
unequal to the task, since their autocorrelations decay
exponentially. ARFIMA, a generalisation of ARMA explicitly
incorporating long-range dependence, does however prove to be suitable. We also briefly discuss the physical mechanisms which
give rise to the power-law scaling found in the data.
DE: 3300 METEOROLOGY AND ATMOSPHERIC DYNAMICS
DE: 6300 POLICY SCIENCES
SC: Nonlinear Geophysics [NG]
MN: 2003 Fall Meeting