An Empirical Test for Temperature Forecasting Models

• 발행기관: 한국자료분석학회
• 저자명: Zhixia Zou Kwangbong Lee
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.14 No.1, 91~100쪽, 전체 10쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2012.02.28

영문 초록

Modeling and forecasting temperatures is the most important process in pricing weather derivatives. Usually, dynamic modeling of daily temperature is considered to be more appropriate than modeling the temperature index. Studies in the literature have consistently pointed out that the observed temperature series exhibit seasonality in all of the mean, variance, and autocorrelation. Therefore, small specification error in the temperature models leads to large pricing error in the weather contracts. Applying the average daily temperature data for 32 years in Guangzhou, China, we estimate two different stochastic temperature behavior models: Cao and Wei (2004) and Campbell and Diebold (2002). Based on the comparison of the statistical properties and derivatives prices derived from the two competing models, the Campbell & Diebold’s time series approach better fits the data and whether it is in the sample or out of the sample, performs closer to the actual HDD/CDD values.

목차

1. Introduction

2. Description of the data

3. Modeling of temperature dynamics

4. Empirical estimations

5. Conclusions

Reference

키워드

Weather Derivatives Mean Reversion Risk Management HDD CDD