: Null hypothesis is that each variable contains unit root. The asterisks *** and ** denote the significance level
: The asterisks *** and ** denote the significance level at 1% and 5%, respectively.
Economics: The Open-Access, Open-Assessment E-Journal 14 (2020–15)
www.economics-ejournal.org
14
independent variables with the lagged dependent variable.
∆𝑋
𝑡
denotes the set of the differenced
control variables, and
∆𝑋
𝑡−1
represents their lagged values. Equation (1) estimates
the direct
effects of immigration policies on FDI inflows with the mediating variable—labor costs—
included as the predictor in the model. Equation (2) estimates the indirect effects of immigration
policies on FDI inflows, using the mediating variable from equation (1) as the dependent
variable. These equations describe labor costs and FDI inflows as endogenous factors
determined by an exogenous variable, immigration policy.
5
Ordinarily, structural equation
models are limited to situations where linear regression models are appropriate for both
mediator and outcome (Keele, 2015). Thus, I include control variables to explain both mediating
variable (unit labor cost) and outcome variable (FDI) in the models. While variables for
differenced FDI and labor cost should ameliorate the endogeneity problem, it is important to
ensure that at least a weak exogeneity exists in the models. To detect the presence of
endogeneity, I perform Granger causality tests. The results show that the null hypotheses that
FDI does not Granger-cause immigration policy, and that FDI does not Granger-cause labor
costs cannot be rejected.
Dostları ilə paylaş: