Page 73 - 臺大管理論叢第33卷第1期

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

IPR protection countries (β = -0.08, p-value < 0.01) while increasing the likelihood of
outsourcing innovation activities to the weak IPR protection countries (β = 0.12, p-value <
0.01). Overall, these robustness checks provide significant support for Hypothesis 1.
Following Aiken and West (1991), we further draw the interaction plot and conduct

the Hausman test to check upon the estimation of Hypothesis 2. For this, we first define
the regression of high task specificity by using the value of task specificity above the
mean plus one standard deviation, and the regression of low task specificity by using

the value of task specificity below the mean minus one standard deviation. Given the
estimation of low-cost talent on the likelihood of location choices (in terms of odds
ratios), we draw the interaction graph to illustrate two models: high-task-specificity and
low-task-specificity regressions (shown in Figure 1). According to Figure 1, the slope of
the high-task-specificity regression is steeper than the slope of the low-task-specificity

regression. We then calculate the simple slope of the high-task-specificity regression that
is significantly positive (β = 2.56, p-value < 0.01) and the slope of the low-task-specificity
regression that is also significantly positive (β = 1.64, p-value < 0.01). Although both

slopes are positive, we further conduct the Hausman test that shows that both regressions
2
are significantly different (χ = 4.9, p-value < 0.05). Moreover, we follow Hoetker (2007)
to evaluate the joint impact of low-cost talent and task specificity for the three groups
of very strong IPR protection countries, strong IPR protection countries, and weak IPR
protection countries from our data (see Figure 2). Consistent with Hypothesis 2, Figure

2 shows that the marginal impact of adding task specificity on the availability of low-
cost talent is positive when outsourcing an innovation activity to weak IPR protection
countries. Specifically, adding task specificity on low-cost talent is associated with a 4.60%

drop in the odds of outsourcing to very strong IPR protection countries, whereas there is
about a 5.77% increase in the odds of outsourcing to the weak IPR protection countries.
According to these robustness checks, the previous estimations and results of Hypothesis 2
are confirmed and supported.
We then plot the interaction graph, conduct simple slope analysis, run the Hausman

test, and evaluate the joint impact of the independent variable and moderating variable for
the robustness checks of Hypothesis 3. Firstly, we define the regression of high project
modularity by using the value of project modularity above the mean plus one standard

deviation and the regression of low project modularity by using the value of project

65