Table 5.
Panel data estimation results for Google Trend metric on dynamic conditional correlation between VIX & Stock changes.
| Dependent Variable |
abs (correlation between VIX & Stock changes) |
||||||
|---|---|---|---|---|---|---|---|
| Independent Variables | Exp. Sign | Pooled with PCSE and AR(1) | Fixed Effects with PCSE and AR(1) | ||||
| Model |
j = All |
j = EUR |
j = Asia |
j = All |
j = EUR |
j = Asia |
|
| C | +/− | 0.03349 | 0.0345 | 0.0348 | 0.097461 | 0.1008 | 0.0978 |
| (0.00)*** | (0.00)*** | (0.00)*** | (0.00)*** | (0.01)** | (0.01)** | ||
| Dj x GTR changes(t-1) | + | 0.026026 | 0.0296 | 0.0462 | 0.019484 | 0.0229 | 0.0210 |
| (0.02)** | (0.00)*** | (0.00)*** | (0.02)** | (0.01)** | (0.20) | ||
| (1-Dj) x GTR changes(t-1) | + | 0.0204 | 0.0222 | 0.0011 | 0.0191 | ||
| (0.23) | (0.04)** | (0.94) | (0.04)** | ||||
| ρj [AR(1) coeff.] | +/− | 0.955131 | 0.9550 | 0.9488 | 0.861368 | 0.9139 | 0.7969 |
| (0.00)*** |
(0.00)*** |
(0.00)*** |
(0.00)*** |
(0.00)*** |
(0.00)*** |
||
| Country Effects | no | no | no | yes | yes | yes | |
| R2 | 92.45% | 92.59% | 92.58% | 93.80% | 94.80% | 93.90% | |
| F Test | 4681.83*** | 2394.23*** | 2392.78*** | 818.78*** | 881.25*** | 739.27*** | |
| Durbin-Watson stat. | 2.13 | 2.14 | 2.14 | 2.03 | 2.04 | 2.04 | |
| N =(ixT) |
767 |
767 |
767 |
767 |
767 |
767 |
|
| Specification tests | |||||||
| Cross section F-test (pooled OLS vs. FEM) p-value | (0.00)*** | (0.00)*** | (0.00)*** | ||||
| Hausman test (FEM vs REM) | (0.00)*** | (0.00)*** | (0.00)*** | ||||
| Test of cross-sectional independence by Frees | (0.00)*** | (0.00)*** | (0.00)*** | ||||
| Modified Wald test for group wise heteroskedasticity | (0.00)*** | (0.00)*** | (0.00)*** | ||||
Notes: This table presents the results for the relationship between Google Trend metric and the dynamic conditional correlation between VIX and stock changes for three country groups (All, Europe, Asia). The “abs” refers to the absolute values of the dynamic conditional correlation estimates from the estimated DCC-GARCH models. The independent variable regressed on cross terms of the dummy variables Dj and (1-Dj) with lagged google trends indicator (GTRt-1). By using dummy variables Dj, (1-Dj) we decompose the sample between j geographic area and the rest. The j takes the value of either EUR for European markets studied, Asia for Asian markets or All for the markets of the whole sample. The pooled and fixed effects models with panel corrected standard errors (PCSE) for the period from 02 January 2020 to 09 April 2020. P-values are in parentheses. *,** and ***indicate statistical significance at the 10%, 5%, and 1% level respectively. The column Exp. Sign refers to expected signs of the coefficient of the relevant explanatory variable. Coefficient ρ is estimated assuming a common across countries autoregressive model [(AR(1)] for the residuals. Next panel of the table reports the adjusted R2 coefficient, the F-test, the Durbin–Watson value and the number of observations [market units (i) x time observations (t)]. The lower part of the table presents specification tests for the models estimated. Initially pooled versus fixed effects are tested, and then a Hausman test is applied between fixed versus random effects models for cases where previously fixed effects models are selected versus pooled models. Finally, p-values of tests for cross sectional independence and group wise heteroscedasticity are reported.