Skip to main content
. 2021 Aug 6;56(4):2199–2214. doi: 10.1007/s11135-021-01214-7

Table 7.

Forecast evaluation using Clark and West test

In-sample Out-of-sample
h = 30 h = 60
Palladium
Full sample 0.3383*** (0.0481) [7.0396] 0.3364*** (0.0478) [7.0332] 0.3384*** (0.0476) [7.1040]
Pre-COVID 0.3927*** (0.0569) [6.9010] 0.3931*** (0.0567) [6.9386] 0.3957*** (0.0564) [7.0182]
Post-COVID 2.9112* (1.5511) [1.8769] 2.8054** (1.3859) [2.0243] 2.6173** (1.2514) [2.0916]
Platinum
Full sample 0.0736** (0.0301) [2.4477] 0.0742** (0.0299) [2.4785] 0.0734** (0.0298) [2.4609]
Pre-COVID 0.0599* (0.0331) [1.8085] 0.0604* (0.0330) [1.8326] 0.0595* (0.0328) [1.8135]
Post-COVID 2.0805** (1.0209) [2.0379] 0.0742** (0.0299) [2.4785] 2.1824** (0.8643) [2.5251]
Silver
Full sample 0.8496*** (0.0971) [0.0971] 0.8508*** (0.0966) [8.8043] 0.8457*** (0.0962) [8.7933]
Pre-COVID 0.7921*** (0.0965) [8.2063] 0.7921*** (0.0962) [8.2335] 0.7878*** (0.0957) [8.2293]
Post-COVID 0.8739* (2.8652) [0.3050] 1.1383* (2.5710) [0.4427] 1.2966* (2.3221) [0.5584]

The rejection of the null hypothesis implies the preference for the uncertainty-based model for gold returns, otherwise, it is not preferred to the benchmark model. The null hypothesis of a zero coefficient is rejected if this statistic is greater than + 1.282 (for a one sided 0.10 test), + 1.645 (for a one sided 0.05 test) and + 2.00 for 0.01 test (for a one sided 0.01 test) (see Clark and West 2007). Values in parentheses—() denote standard errors while those reported in square brackets—[] are for t-statistics

***, **, *Imply the rejection of the null hypothesis of equal forecast accuracy at 1%, 5%, 10% levels of significance, respectively