Table 3.

Statistical tests of COVID-period differences in Russian individual investors’ trading behavior.

	Baseline model		Eq. (4)	Nonlinear model — Eq. (5)			Eq. (6)		Eq. (6) with bubble dummy
	Coefficient	t-stat.	Coefficient	t-stat.	Coefficient	t-stat.	Coefficient	t-stat.	Coefficient	t-stat.
$\omega$	0.000057	(3.41)⁎⁎	0.000033	(2.49)⁎	0.000078	(1.66)	0.000037	(1.32)	0.000044	(1.42)
$\omega$’			0.000093	(1.63)			0.000244	(2.39)⁎	0.000411	(2.45)⁎
B									0.000376	(1.58)
$\psi$	0.091	(2.05)⁎	0.078	(1.47)	0.116	(2.84)⁎⁎	0.061	(1.08)	0.042	(0.66)
$\psi$’			−0.025	(−0.41)			−0.013	(−0.20)	−0.098	(−0.99)
$\lambda$	−0.0062	(−3.56)⁎⁎	−0.0055	(−10.27)⁎⁎	0.0119	(−4.81)⁎⁎	−0.0062	(−5.84)⁎⁎	−0.0062	(−5.86)⁎⁎
$\lambda$’			−0.0010	(−0.35)			−0.0104	(−2.63)⁎⁎	−0.0112	(−3.22)⁎⁎
$\lambda$”					0.0093	(3.30)⁎⁎	0.0013	(0.77)	0.0013	(0.78)
$\lambda$”’							0.0140	(3.19)⁎⁎	0.0159	(3.75)⁎⁎
$\theta$	0.0006	(1.52)	0.0004	(0.92)	0.0018	(2.10)⁎	0.0005	(0.81)	0.0003	(0.42)
$\theta$’			0.0000	(−0.02)			0.0002	(0.16)	−0.0029	(−0.94)
$\theta$”					−0.0023	(−2.19)⁎	−0.0004	(−0.39)	−0.0001	(−0.12)
$\theta$”’							−0.0009	(−0.44)	0.0037	(0.89)

Each vertical block consisting of two columns (one for coefficients and one for Newey–West​ t-statistics in parentheses) represents one regression. The first block reports the results from $F_t=\omega +\psi {\sum }_{i=1}^p{F}_{t-i}+\lambda R_t+\theta {\sum }_{i=1}^p{R}_{t-i}+{\varepsilon }_{1,t}$ . The second block reports the results from Eq. (4), i.e., $F_t=\omega +{\omega }^{\prime }{D}_t+\psi {\sum }_{i=1}^p{F}_{t-i}+{\psi }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{F}_{t-i}+\lambda R_t+{\lambda }^{\prime }{R}_t{D}_t+\theta {\sum }_{i=1}^p{R}_{t-i}+{\theta }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{R}_{t-i}+{\varepsilon }_{1,t}$ which allows for COVID-period differentials in parameters.

The third block reports results from $F_t=\omega +\psi {\sum }_{i=1}^p{F}_{t-i}+\lambda R_t+{\lambda }^{\prime \prime }{R}_t{N}_t+\theta {\sum }_{i=1}^p{R}_{t-i}+{\theta }^{\prime \prime }{\sum }_{i=1}^p{N}_{t-i}{R}_{t-i}+{\varepsilon }_{1,t}$ which is the baseline model that allows for differential responses to positive and negative returns.

The fourth block reports results from Eq. (5), i.e., $F_t=\omega +{\omega }^{\prime }{D}_t+\psi {\sum }_{i=1}^p{F}_{t-i}+{\psi }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{F}_{t-i}+\lambda R_t+{\lambda }^{\prime }{R}_t{D}_t+{\lambda }^{\prime \prime }{R}_t{N}_t+{\mathit{\lambda }}^{\prime \prime \prime }{R}_t{D}_t{N}_t+\theta {\sum }_{i=1}^p{R}_{t-i}+{\theta }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{R}_{t-i}+{\theta }^{\prime \prime }{\sum }_{i=1}^p{R}_{t-i}{N}_{t-i}+{\theta }^{\prime \prime \prime }{\sum }_{i=1}^p{D}_{t-i}{R}_{t-i}{N}_{t-i}+{\varepsilon }_{1,t}$ which allows for COVID-period differentials in parameters in the model that allows for differential responses to positive and negative returns.

The last block reports results from $F_t=\omega +{\omega }^{\prime }{D}_t+{\omega }^{\prime \prime }{B}_t+\psi {\sum }_{i=1}^p{F}_{t-i}+{\psi }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{F}_{t-i}+\lambda R_t+{\lambda }^{\prime }{R}_t{D}_t+{\lambda }^{\prime \prime }{R}_t{N}_t+{\mathit{\lambda }}^{\prime \prime \prime }{R}_t{D}_t{N}_t+\theta {\sum }_{i=1}^p{R}_{t-i}+{\theta }^{\prime }{\sum }_{i=1}^p{D}_{t-i}{R}_{t-i}+{\theta }^{\prime \prime }{\sum }_{i=1}^p{R}_{t-i}{N}_{t-i}+{\theta }^{\prime \prime \prime }{\sum }_{i=1}^p{D}_{t-i}{R}_{t-i}{N}_{t-i}+{\varepsilon }_{1,t}$ which adds the negative bubble dummy to the previous model in the fourth block.

^*Denotes statistical significance at the 5% level.

^**Denotes statistical significance at the 1% level.