Table 1.

Out-of-sample point forecast model evaluation in terms of MAFE relative to a random walk model. We also show the corrected version of Diebold and Mariano (1995) test statistics proposed by Harvey et al. (1997) for equal predictive accuracy (under absolute loss function). Here,^⁎⁎⁎,^⁎⁎, and^⁎ denote a Diebold–Mariano significance at 1%, 5%, and 10%, respectively. The evaluation period is from 2011 Q4 to 2019 Q4 (top) and 2011 Q4 to 2021 Q4 (bottom).

	France	Germany	Italy	Spain
Pre-Covid-19
AR(1)	0.87⁎⁎⁎	0.70⁎⁎⁎	1.00	1.00
BMA - 1st month	0.84	0.72⁎⁎⁎	1.29	1.67
BMA - 2nd month	0.29⁎⁎⁎	0.19⁎⁎⁎	0.56⁎⁎	0.61
BMA - 3rd month	0.29⁎⁎⁎	0.14⁎⁎⁎	0.48⁎⁎⁎	0.56
Covid-19 Included
AR(1)	0.99	0.73⁎⁎	0.89	1.07
BMA - 1st month	0.44	0.54⁎⁎	0.36	0.36
BMA - 2nd month	0.17⁎	0.20⁎⁎⁎	0.26⁎	0.19⁎
BMA - 3rd month	0.10⁎	0.18⁎⁎⁎	0.34⁎	0.19⁎

^***1% significance.

^**5% significance.

^*10% significance.