Page 36 - 34-3

In the second stage, we assess the effect of divesting from China on market performance. To
account for sample selection bias, we categorize observations into two groups, namely the
In the second stage, we assess the effect of divesting from China on market performance. To
divestiture group and the non-divestiture one. We also construct separate models for each group,
account for sample selection bias, we categorize observations into two groups, namely the
with sample selection correction factors being incorporated to correct for bias and to ensure that
divestiture group and the non-divestiture one. We also construct separate models for each group,
regression coefficient estimates are unbiased and approaching a normal distribution.
with sample selection correction factors being incorporated to correct for bias and to ensure that
regression coefficient estimates are unbiased and approaching a normal distribution.
We apply Heckman’s two-stage method to obtain the firm’s performance function, which is
expressed as:
We apply Heckman’s two-stage method to obtain the firm’s performance function, which is
expressed as:
, (1)
�
, (1)
where represents the market performance of a firm, is a vector of explanatory variables,
is the regression coefficient vector, and is the error term.
where represents the market performance of a firm, is a vector of explanatory variables,
is the regression coefficient vector, and is the error term.
In consideration of sample selection bias, equations (2) and (3) represent the firm’s divestiture
decision function and performance function, respectively, and are expressed as:
In consideration of sample selection bias, equations (2) and (3) represent the firm’s divestiture
decision function and performance function, respectively, and are expressed as:
≥ 5% ( ),
∗
�
∗ ≥ 5% ( ),
∗
< 5% ( ), (2)

�
� < 5% ( ), (2)
∗
, (3)
� � � � � � ∗ ∗ ∗ ∗ � �
�
, (3)
where is an unobserved variable, is a set of firm characteristics in equation (2) and
�
∗
�
� �
�
instrumental variables that affect the decision to divest, and is the error term.
where is an unobserved variable, is a set of firm characteristics in equation (2) and
Divestiture from China by Taiwanese Listed Electronic Information Firms: Effects of Host Country
∗
Performance and Selection Bias
instrumental variables that affect the decision to divest, and is the error term.
According to equation (3), when is greater than or equal to 5%, only the market
∗
performance of firms that have divested from China are being observed. Therefore, to accurately
According to equation (3), when is greater than or equal to 5%, only the market
∗
accurately estimate the performance function of a firm that has divested from China, we
estimate the performance function of a firm that has divested from China, we need to adjust the
performance of firms that have divested from China are being observed. Therefore, to accurately
need to adjust the model to obtain equation (4), which is expressed as:
model to obtain equation (4), which is expressed as:
estimate the performance function of a firm that has divested from China, we need to adjust the
model to obtain equation (4), which is expressed as:
( | ) ( | )
�
� ) ( | )
( | )
( | � �
� �
� ( | )
�
� � � � � �
�∅( ( )�
� � �� �
� � �
(4)
�
� �∅( ( )� (4)
��
, �
� �
� �
(4)
,
where is the inverse Mill’s ratio or Heckman’s lambda (i.e., the correction for self-selection).
�
where λ is the Inverse Mill’s Ratio or Heckman’s lambda (i.e., the correction for self-
� �
selection).
where is the inverse Mill’s ratio or Heckman’s lambda (i.e., the correction for self-selection).
In the second stage, the performance function of a firm that divested from China, that is, the
In the second stage, the performance function of a firm that divested from China,
inverse Mill’s ratio , is included in the subsequent analysis. Therefore, the performance function of
In the second stage, the performance function of a firm that divested from China, that is, the
that is, the Inverse Mill’s Ratio λ, is included in the subsequent analysis. Therefore, the
a firm that disinvested from China can be redefined as:
inverse Mill’s ratio , is included in the subsequent analysis. Therefore, the performance function of
performance function of a firm that disinvested from China can be redefined as:
a firm that disinvested from China can be redefined as:
. (5)
�
�
� �
. (5) (5)
23
�
� � �
23
To address and correct for sample selection bias and potential endogeneity resulting

from the use of self-selection factors, the present study adopts two measures. First, this
study employs the probit regression model established in the first stage to estimate the
Inverse Mill’s Ratio for correcting sample selection bias. Second, to mitigate self-selection
bias, the present study employs a modified version of the method developed by Leiblein,
Reuer, and Dalsace (2002). This is a quasi-experimental method that involves modeling
two sets of observations, with one model in this study being for firms with a divestment
strategy and the other being for firms with a non-divestment strategy.
We apply grouping modeling, a quasi-experimental method to identify observations
in the divestiture group where firms that divest should not have done so (i.e., decision
error) and observations in the non-divestiture group where firms that should have divested
did not do so (i.e., decision error). The estimation of the probability of decision errors is
based on the probit regression model established in the first stage of Heckman’s two-stage
method. Taking the divestiture group as an example, by subtracting the probability of a
firm divesting from China from 1, the probability of a firm that should have divested from
China but did not (i.e., decision error) is obtained; this probability is referred to as the
decision error variable in the present study.
26