148 / 274
經理人外部連結對新興產業內公司之表現影響
148
members’ employment history, this study used companies’ SEC filings and four industry
magazines, namely–
Telephony, Cellular Business, Communications (Englewood)
, and
Mobile Phone News
, that reported the movement of managers within the industry during the
study period. To supplement these sources, searches of
Dun & Brad street Corporate
Management, LexisNexis
, and
Hoover’s Online
were also performed.
5.4 Analysis
The hypotheses are tested using the fixed-effects model because it is considered a
conservative test that can control for all constant, unobserved heterogeneity among firms
(Greene, 2000). In addition, there is no justification for treating the individual effects as
uncorrelated with the other regressors, as is assumed in the random-effects model (Greene,
2000). Moreover, the results from Hausman tests on specified regression models support the
use of fixed effects, not random effects. The general form for estimation is as follows:
y
it
=
α
+
X
it
‧
β
+
v
i
+
ɛ
it
(1)
where
y
it
is the increase in cell-phone subscribers for firm
i
from year
t
-1 to
t
,
v
i
+
ɛ
it
is
the residual that includes the regular error term and firm-specific component
v
i
. Note that a
TMT’s external ties in year
t
-1 interacted with the addition in a TMT’s intra-industry
experience for firm
i
from year
t
-1 to
t
. The reasons for such specification are that (1) the
external ties at year
t
-1 further shape the “development of intra-industry experience from
year
t
-1 to
t
” and (2) the future experience “∆”, rather than current experience, can be
shaped.
6. RESULTS
6.1 Descriptive Statistics
The final data set comprises 41 firms and 272 firm-year observations. The
corresponding descriptive statistics and correlation matrix for key variables are summarized
in Tables 2 and 3. On average, the firms in the sample have 9.35 team members, 0.26 years
addition in a TMT’s cell-phone service industry experience, and 0.15 team members whose
background information is missing. Although some correlations are high in Table 2, high
correlations do not occur among the key explanatory variables. In addition, using panel data
analysis can reduce the problem of multi-collinearity among variables (Hsiao, 2003).