E-journal Article
Yüklə 0,55 Mb. Pdf görüntüsü
|
- Bu səhifədəki naviqasiya:
- _________________________
| Statistical Method
This study suggests that more expansive immigration policy directly attracts FDI due to the positive effects of immigrant networks. In addition to this direct effect, this paper argues that more expansive immigration policies indirectly lead to increased FDI inflows by putting downward pressure on labor costs. To assess both the direct and indirect effects, I use path
4 All control variables are from the World Bank’s World Development Indicator 2017 unless otherwise noted. Economics: The Open-Access, Open-Assessment E-Journal 14 (2020–15) www.economics-ejournal.org
12
analysis of the causal mechanisms among immigration, labor costs, and FDI. Spurious regression problems, including the presence of unit root, the existence of co-integration, and endogeneity, can be expected in a traditional regression analysis of time series data. Therefore, I conduct a set of diagnostic tests. First, running an augmented Dickey–Fuller (ADF) test, I detect that all variables are non- stationary, except for three: refugee ceiling per capita, economic growth, and market size. The results of the ADF are given in Table 2. Second, I perform the Johnsen co-integration test with co-integrating ranks (
) to see if there is any co-integration between the explanatory and the outcome variables. The results are presented in Table 3. The adjusted trace statistics indicate that the null hypothesis of no co-integration relationship between variables is rejected, and there exist at least four co-integration relationships between variables at 5%. Therefore, we believe a long-term relationship exists between the variables in the model. The results are reported as follows. Overall, the results of the diagnostic tests suggest that single equation error-correction models (ECM) are the most appropriate for dealing with the presence of non-stationarity and co- integration in time series data (Beck and Katz, 1996; De Boef and Keele, 2008). The ECM allows us to observe the immediate short-term impact of independent variables using differenced values and their long-term impact using the lagged levels. Naturally, the ECM models remove trends for all variables, including the FDI measure. Intuitively, it would be unrealistic to expect that change in immigration policy would only cause a change in FDI inflows in the following year. Immigrants usually take time to make the decision to migrate when the immigration laws change. Thus, it would take a while to see some substantial effects of immigration reforms, and foreign firms would want to take time to see the results of a policy. Traditional linear regression methods, therefore, cannot capture the possible long-run impacts of a change in immigration policies. Following the traditional literature on mediation and the merits of the ECM, both direct and indirect effects of immigration policies are estimated, using the following set of linear equations, which model single-level, simple mediation analysis. It is also important to empirically specify the models that examine how the increased labor costs induced by more restrictive immigration policy affect FDI using two estimations. Structural equation models are used to investigate the causal mechanisms for the key independent (immigration policy), mediating (labor costs), and dependent (FDI inflows) variables.
∆𝐹𝐹𝐹
𝑡 = 𝛽 0 + 𝛽
1 𝐹𝐹𝐹
𝑡−1 + 𝛽
2 ∆𝐹𝐼𝐼𝐹𝐼𝐼𝐼𝐼𝐹𝐼𝐼 𝑡 + 𝛽
3 𝐹𝐼𝐼𝐹𝐼𝐼𝐼𝐼𝐹𝐼𝐼 𝑡−1 + 𝛽
4 ∆𝐿𝐼𝐿𝐼𝐼
𝑡 + 𝛽 4 𝐿𝐼𝐿𝐼𝐼
𝑡−1 + 𝛽
𝑘 ∆𝑋 𝑡 + 𝛽 𝑘 𝑋 𝑡−1 + 𝜀
𝑡 ……………………………… (1) ∆𝐿𝐼𝐿𝐼𝐼 𝑡
0 + 𝛽
1 𝐿𝐼𝐿𝐼𝐼
𝑡−1 + 𝛽
2 ∆𝐹𝐼𝐼𝐹𝐼𝐼𝐼𝐼𝐹𝐼𝐼 𝑡 + 𝛽
3 𝐹𝐼𝐼𝐹𝐼𝐼𝐼𝐼𝐹𝐼𝐼 𝑡−1 +
𝑘 ∆𝑋 𝑡 + 𝛽 𝑘 𝑋 𝑡−1 + 𝜀
𝑡 ………………………………………………………………… (2)
where subscript t denotes time, and ɛ denotes a stochastic error term, and ∆ denotes the first difference operator. I model each equation by operating such that the change in the dependent variable from the year t– 1 to year t is a function of both the differenced and lagged major
|