<abstract xmlns="http://www.w3.org/1999/xhtml"><p>An accurate weather forecast is the basis for the valuation of weather derivatives, securities that partially compensate for financial losses to holders in case of, from their perspective, adverse outside temperature. The paper analyses precision of two forecast models of average daily temperature, the Ornstein-Uhlenbeck process (O-U process) and the generalized autoregressive conditional heteroskedastic model (GARCH model) and presumes for the GARCH model to be the more accurate one. Temperature data for the period 2000-2017 were taken from the DHMZ database for the Maksimir station and used as the basis for the 2018 forecast. Forecasted values were compared to the available actual data for 2018 using MAPE and RMSE methods. The GARCH model provides more accurate forecasts than the O-U process by both methods. RMSE stands at 3.75 °C versus 4.53 °C for the O-U process and MAPE is 140.66 % versus 144.55 %. Artificial intelligence and supercomputers can be used for possible improvements in forecasting accuracy to allow for additional data to be included in the forecasting process, such as up-to-date temperatures and more complex calculations.</p></abstract>

An accurate weather forecast is the basis for the valuation of weather derivatives, securities that partially compensate for financial losses to holders in case of, from their perspective, adverse outside temperature. The paper analyses precision of two forecast models of average daily temperature, the Ornstein-Uhlenbeck process (O-U process) and the generalized autoregressive conditional heteroskedastic model (GARCH model) and presumes for the GARCH model to be the more accurate one. Temperature data for the period 2000-2017 were taken from the DHMZ database for the Maksimir station and used as the basis for the 2018 forecast. Forecasted values were compared to the available actual data for 2018 using MAPE and RMSE methods. The GARCH model provides more accurate forecasts than the O-U process by both methods. RMSE stands at 3.75 °C versus 4.53 °C for the O-U process and MAPE is 140.66 % versus 144.55 %. Artificial intelligence and supercomputers can be used for possible improvements in forecasting accuracy to allow for additional data to be included in the forecasting process, such as up-to-date temperatures and more complex calculations.

Ornstein-Uhlenbeck process and GARCH model for temperature forecasting in weather derivatives valuation

Faculty of Economics and Business, Department of Statistics

Croatian Review of Economic, Business and Social Statistics

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

An accurate weather forecast is the basis for the valuation of weather derivatives, securities that partially compensate for financial losses to holders in...

Ornstein-Uhlenbeck process and GARCH model for temperature forecasting in weather derivatives valuation

Publié en ligne: 08 juin 2020

Pages: 27 - 42

Reçu: 03 avr. 2020

Accepté: 10 mai 2020

DOI: https://doi.org/10.2478/crebss-2020-0003

Mots clés
GARCH model, MAPE, Ornstein-Uhlenbeck process, RMSE, temperature forecasting, weather derivatives

© 2020 Berislav Žmuk et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Ornstein-Uhlenbeck process and GARCH model for temperature forecasting in weather derivatives valuation

Publié en ligne: 08 juin 2020

Pages: 27 - 42

Reçu: 03 avr. 2020

Accepté: 10 mai 2020

DOI: https://doi.org/10.2478/crebss-2020-0003

Mots clésGARCH model, MAPE, Ornstein-Uhlenbeck process, RMSE, temperature forecasting, weather derivatives

© 2020 Berislav Žmuk et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Mots clés
GARCH model, MAPE, Ornstein-Uhlenbeck process, RMSE, temperature forecasting, weather derivatives