Permanent shocks and forecasting with moving averages

ABSTRACT

Moving averages are a common method of forecasting futures basis. We argue that the optimal lengths of moving averages depend on the frequency of structural breaks. A new stochastic time-series process including structural breaks is modelled by discrete probability distributions that capture the frequency and size of structural breaks. A permanent shock (means structural breaks in this article) is captured by a Poisson-jump or a Bernoulli-jump process, and a temporary shock is represented by a white noise process. Futures basis data are used to estimate the frequency of permanent shocks as well as the size of both shocks. Most shocks are permanent shocks. Since most shocks are permanent, the most recent year provides the best forecast and the optimal length of the moving average is one.

KEYWORDS:

• Poisson-jump process
• Bernoulli-jump process
• GMM
• calibration test
• autoregressive integrated moving average with outliers

JEL CLASSIFICATION:

• C53
• G13
• Q11

Acknowledgements

Brorsen receives financial support from the A.J. and Susan Jacques Chair, Oklahoma Agricultural Experiment Station and USDA National Institute of Food and Agriculture, Hatch Project number OKL02939.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the A.J. and Susan Jacques Chair; Oklahoma Agricultural Experiment Station; National Institute of Food and Agriculture [OKL02939];