Table 2.

Log marginal likelihood for the large BVAR.

	Weak	SZ	Strong SZ			Weak	SZ	Strong SZ
19Q4	−6480.35	−3756.80	−3918.06		19Q4	−5915.25	−3485.19	−3695.15
20Q1	−6615.97	−3864.18	−4021.19		20Q1	−5982.10	−3539.20	−3751.03
20Q2	−6756.28	−3978.87	−4151.69		20Q2	−6062.13	−3608.26	−3810.56
20Q3	−6889.06	−4067.78	−4233.83		20Q3	−6152.90	−3717.54	−3905.92
20Q4	−6964.35	−4114.55	−4298.21		20Q4	−6190.53	−3766.41	−3957.56
21Q1	−7049.20	−4151.13	−4334.66		21Q1	−6299.66	−3807.15	−4000.44
21Q2	−7134.59	−4199.77	−4388.15		21Q2	−6323.87	−3852.67	−4045.27
	(a) BVAR with Gaussian errors					(b) BVAR with fat-tailed errors

Note: The bold figure indicates the maximum log marginal likelihood for each model in a selected estimation window. ‘Weak’ stands for weakly informative, and ‘SZ’ stands for Sims and Zha.