Skip to main content

Table 9 Robustness analysis to mitigate the influence of outliers

From: Store of value or speculative investment? Market reaction to corporate announcements of cryptocurrency acquisition

Window

OLS excluding p1 and p99

Median linear regression (MLR)

[0]

[− 1,1]

[− 2,2]

[− 5,5]

[0,1]

[0,3]

[0]

[− 1,1]

[− 2,2]

[− 5,5]

[0,1]

[0,3]

Panel A. High Exposure to Cryptocurrency

High Exposure Dummy

4.2529**

4.8908**

3.0610*

2.9734

3.8717*

4.1334*

2.3242*

3.5163

2.3471

1.4863

4.5306**

4.6807**

 

(1.50)

(1.69)

(1.57)

(4.62)

(1.82)

(2.20)

(1.11)

(2.27)

(2.80)

(3.79)

(2.08)

(1.65)

Control variables

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Obs

31

31

31

31

31

31

33

33

33

33

33

33

Adj. R-Sq

0.324

0.57

0.473

− 0.254

0.133

0.117

Panel B. Low Exposure to Cryptocurrency

Low Exposure Dummy

− 2.7781

− 4.8753**

− 1.7896

− 7.0344

− 3.7929

− 1.9693

− 1.5908

− 3.4063

− 3.3819

− 3.9626

− 4.1328**

− 2.2714

 

(1.84)

(1.91)

(2.33)

(5.92)

(2.29)

(2.10)

(1.41)

(2.80)

(2.90)

(3.60)

(1.71)

(1.53)

Control variables

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Obs

31

31

31

31

31

31

33

33

33

33

33

33

Adj. R-Sq

0.0277

0.495

0.352

− 0.105

0.0614

− 0.167

-

-

-

-

-

-

  1. This table shows the results of the robustness analysis that deals with the influence of outliers. OLS regressions are reported on a subsample after dropping extreme events (p1 and p99 of each CAR). Alternatively, on the right-hand side of the Table, we estimate each equation using Median Linear Regression (MLR). The dependent variables are the CARs associated with each event window—[− 1,1], [− 2,2], [− 5,5], [0,1], and [0,3]. High Exposure Dum (Low Exposure Dum) is a dummy that equals one if the event is classified as High Exposure (Low Exposure) to Cryptocurrency and zero otherwise. All regressions include country-fixed effects and firm, industry, and market-level controls, as reported in Table 4. Robust standard errors are reported in parentheses. ***, **, and * represent statistical significance at the 10%, 5% and 1% levels, respectively