Skip to main content

Table 7 ARMA-GARCH estimation of realised volatility for 5- and 15-min intervals

From: Impact of trading hours extensions on foreign exchange volatility: intraday evidence from the Moscow exchange

Variable

5-min intraday intervals

15-min intraday intervals

Regular return—rr (A)

Maxmin Return—rmm (B)

Regular return -rr (C)

Maxmin Return -rmm (D)

Coef

z-stat

Coef

z-stat

Coef

z-stat

Coef

z-stat

C

− 5.200

− 16.66***

− 4.190

− 13.35***

− 5.583

− 23.92***

− 4.192

− 14.42***

\({\varvec{DTHC}}\)

0.587

3.42***

0.555

3.44***

0.517

2.49**

0.599

3.62***

\({\varvec{DUR}}\)

0.014

3.43***

0.015

3.97***

0.015

5.01***

0.014

3.96***

AR(1)

0.255

6.27***

0.267

6.25***

0.351

9.45***

0.253

6.73***

AR(2)

0.142

3.60***

0.124

3.25***

0.235

5.87***

0.114

2.76***

AR(3)

0.640

8.13***

0.637

7.89***

–

–

0.076

1.94*

AR(4)

− 0.116

− 2.16**

− 0.091

− 1.59

0.116

3.04***

0.456

6.92***

MA(3)

− 0.574

− 6.44***

− 0.587

− 6.76***

0.133

3.08***

–

–

MA(4)

–

–

–

–

–

–

− 0.400

− 5.95***

Variance Equation

        

C

0.013

2.10**

0.018

2.22**

0.005

1.61

0.020

1.80*

RESID(-1)2

0.083

3.85***

0.109

3.97***

0.015

1.67*

0.088

3.17***

GARCH(-1)

0.905

42.13***

0.873

32.05***

0.976

104.56***

0.889

28.68***

T-DIST. DOF

–

–

–

–

10.653

2.39**

10.970

2.14**

Adj. R2

0.4141

0.4363

0.5238

0.4561

ARCH-LM Stat

0.1185

0.2368

0.1136

0.1001

  1. ***, ** and * indicate 1%, 5% and 10% levels of significance, respectively. Correlogram Q-statistics show that no autocorrelation is present in the ARMA-GARCH regression. Although the coefficient of DTHC and DUR have positive directions and the same significance level in OLS regression, the serial correlation LM tests and ARCH LM tests verify that OLS estimates indicate autocorrelation and heteroscedasticity problems. ARMA-GARCH estimations past the ARCH LM test state that there is no heteroscedasticity. According to the Jarque–Bera test, the residual of the ARMA-GARCH estimations is normally distributed