"Technical Analysis in Financial Markets"

1: Contract specifications of January 26, 1998.
2: We thank the cocoa-trading firm Unicom International B.V. and ADP Financial Information Services for providing the data.
3: The pasting date is equal to the roll over date.
4: H₀: σ_r(csce)²=σ_r(liffe)² vs H₁: σ_r(csce)² ≠ σ_r(liffe)²; F=S_r(csce)²/S_r(liffe)²;
5: Because sample autocorrelation may be spurious in the presence of heteroskedasticity we also tested for significance by computing Diebold (1986) heteroskedasticity-consistent estimates of the standard errors, se(k)=1/n (1+γ(r², k)/σ⁴), where n is the number of observations, γ(r², k) is the k-th order sample autocovariance of the squared returns, and σ is the standard error of the returns. ***, **, * in table 2.2 then indicates whether the corresponding autocorrelation is significantly different from zero.
6: Positions are unchanged until the moving averages really cross.
7: In practice traders can hold a margin of 10% of the underlying value. The broker issues frequently a margin call, that is to add money to the margin, if the trader is in a losing position. However, to keep things as simple as possible we assume a fully protected trading position by setting the required margin to 100% of the underlying value.
8: Nelson (1991) replaces the normal distribution used here with a generalized error distribution.
9: We checked for significance of the estimated coefficients. We did diagnostic checking on the standardized residuals, to check whether there was still dependence. We used the (partial) autocorrelation function, Ljung-Box (1978) Q-statistics and the Breusch-Godfrey LM-test. The Schwartz Bayesian criterion was used for model selection.
10: This model is found to fit the data the best, see page ??.
11: We would like to thank Guido Veenstra, employed at the Dutch cocoa firm Unicom, for pointing this out to us.