Skip to main content

Table 7 Robustness tests using the CH(1995) model

From: The witching week of herding on bitcoin exchanges

 

Extreme 1% lower market returns

Extreme 1% larger market returns

DLDexp

p-value

DL(1 − Dexp)

p-value

DUDexp

p-value

DL(1 − Dexp)

p-value

Aver. D0pre − D150pre

0.0004

0.10

0.0017

0.00

0.0004

0.11

0.0011

0.00

Aver. D1post − D150post

0.0007

0.12

0.0017

0.00

0.0005

0.18

0.0010

0.00

Aver. D4pre − D8pre

− 0.0003

0.00

0.0016

0.00

0.0003

0.21

0.0010

0.00

Aver. D8post − D11post

0.0001

0.59

0.0016

0.00

− 2.9E−05

0.63

0.0010

0.00

  1. The table shows the estimates of the CH(1995) equations including five lags for the dispersion of returns \(CSSD_{t} = \alpha _{0} + \beta _{{11}} D_{{\exp }} D_{t}^{L} + \beta _{{12}} (1 - D_{{\exp }} )D_{t}^{L} + \beta _{{21}} D_{{\exp }} D_{t}^{U} + \beta _{{22}} (1 - D_{{\exp }} )D_{t}^{U} + \varepsilon _{t}\)
  2. Dexp is the dummy variable, defined differently, that takes value 1 in specific times around expiration and 0 otherwise. DL = 1 if the market return at time t lies in the 1% extreme lower tail of the return distribution and 0 otherwise. DU = 1 if the market return at time t lies in the 1% extreme upper tail of the return distribution and 0 otherwise. In the table: Aver. D4pre − D8pre is the variable representing the average of Dexp estimates (from D4pre to D8pre) and their average significance in parentheses (from D4pre to D8pre); Aver. D0pre − D150pre is the variable representing the average of Dexp estimates (from D0pre to D150pre) and their average significance in parentheses (from D0pre to D150pre); Aver. D8post − D11post is the variable representing the average of Dexp estimates (from D8post to D11post) and their average significance in parentheses (from D8post to D11post) and Aver. D1post − D150post is the variable representing the average of Dexp estimates (from D1post to D150post) and their average significance in parentheses (from D1post to D150post)