The "Shukla-Gutzler" Method for Estimating Potential Seasonal Predictability

This paper reexamines a procedure proposed by Shukla and Gutzler for estimating potential seasonal predictability. Certain subtle and unverified assumptions required for the method to work are clarified, and Monte Carlo experiments are used to demonstrate that these assumptions are adequate even for auto-correlated processes in typical applications, provided the effective time scale T-0 of the stochastic process is known. This paper highlights the fact that the time scale T-0 is difficult to estimate reliably (as noted in other papers) and can be biased by an order of magnitude. This bias can seriously compromise the reliability of the Shukla-Gutzler method.