Page 65 - 臺大管理論叢第33卷第1期

NTU Management Review Vol. 33 No. 1 Apr. 2023

offshore in the first-stage estimation. In the first-stage estimation, specifically, we adopt the
following equation:
Z = W α + S γ + ϑ , (1)
ij
i
ij
ij
where
Z is the dependent variable with a binary value of 1 if firm i adopted the offshore
ij
outsourcing mode for the outsourced project j; 0 otherwise.
W includes independent and control variables that influence firm i’s choice to adopt
ij
the offshore outsourcing mode for the outsourced activity j.
S includes the instrumental variables.
ij
ϑ is the error term.
i
The second-stage estimation is our main model, in which we introduce the inverse
Mills ratio (a selection correction term constructed by the results from the first-stage

estimation) to control for the potential bias of sample selection in estimating the effects of
independent variables and moderating variables on the dependent variable. In the second-
stage estimation, we then estimate the following equation:

Y = X β + C σ + ϵ , (2)
ij
ij
i
i
where
Y represents the dependent variable with a binary value of 1 if firm i chose the weak
ij
IPR protection country to outsource activity j; 0 otherwise.
X represents independent, moderating, and control variables affecting the dependent
ij
variable (the location choice of firm i to outsource activity j).
C is the inverse Mills ratio, a selection correction term, denoting the probability
i
density function over the cumulative distribution function of a distribution.

ϵ is the error term.
i
Our dataset also includes firm-activity observations. A given firm might have
multiple associated observations, meaning that the analyzed data may result in a clustered
structure. To overcome this potential problem, we chose a maximum likelihood estimator
by controlling fixed factors and constant parameters with clustered robust standard errors

(Greene, 2000). This two-stage model can also correct for the endogeneity caused by
unobserved and fixed factors (Wooldridge, 2012).

57