| Panel A. Monthly Data Observations | Panel B. Quarterly Data Observations |
---|
Lags | 2 | 4 | 6 | 2 | 4 | 6 |
---|
PMG /→ PML | 0.390 | 0.851 | 0.622 | 0.594 | 0.566 | 0.436 |
PML /→ PMG | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
P MGF /→ P MLF | 0.967 | 0.952 | 0.908 | 0.852 | 0.792 | 0.390 |
P MLF /→ P MGF | 0.000 | 0.000 | 0.000 | 0.002 | 0.004 | 0.014 |
- Note. X /→ Y means the null hypothesis that X does not Granger-causes Y. P MGF and P MLF mean filtered PMG and PML respectively. This table reports the p-values of the F -statistics. When performing Granger causality test, we set m = n in Equation (5) for the sake of being convenient. Di erent lags are used for being robust since the Granger causality tests are sensitive to the lag selection. Panels A and B report respectively the results for monthly and quarterly data observations. Decomposing stock returns with high price extremes are presented as
-
\( {\displaystyle \begin{array}{l}{r}_t=\left[\mathit{\log}\left({\mathrm{C}}_t\right)-\mathit{\log}\left({\mathrm{L}}_t\right)\right]-\left[\mathit{\log}\left({\mathrm{L}}_t\right)-\mathit{\log}\left({\mathrm{O}}_t\right)\right]\\ {}\kern1.5em =P\ {MG}_t-P\ {ML}_t\end{array}} \)