Working Paper Series: Special Edition of 2016 to 2018 Interns

The panel Granger Causality advanced by Dumitrescu-Hurlin was employed. The method allows all coefficients to be different across cross-sections such that 0, ≠ 0, , 1, ≠ 1, , . . . , 1, ≠ 1, , ∀ , . 3 1, ≠ 1, , … , 1, ≠ 1, ∀ , . 4 Diagnostics Along with the expected diagnostics, cross-sectional dependence is checked. In the context of panel studies, it may be likely that substantial cross-sectional dependence in errors may arise due to common shocks, unobserved components and spatial dependence. The general hypothesis of no cross-section dependence is considered in terms of correlations between the disturbances in different cross-section units given by: 0 : = � , � = 0 ≠ . 5 In this study the Breusch-Pagan Chi-square test for cross-section independence was employed. In the context of dynamic panel estimators, the presence of cross-sectional dependence has severe consequences. Refer to De Hoyos & Sarafidis (2006) for a deeper discussion on the subject. Note, the Bayesian (Schwarz) information criterion (BIC)/(SIC) was used throughout the research. All discussion on diagnostics are found in the appendix. The time series data span the years 1995 to 2013 gathered on an annual basis where all variables are expressed in Eastern Caribbean dollars with 2010 as base year. Data utilized in this study were mainly gathered from the World Bank and the Eastern Caribbean Central Bank (ECCB). The results concluded that each variable is integrated of order one or I (1) while their linear combination (that is, the error term) is I (0) at 5% level of significance; see Table 2. Note, I (n) speaks to the number of times a variable is differenced to achieve stationarity and the fact that all variables are I (1) suggests there is a balanced order of integration. The implications of these results are that the model can avoid estimating inaccurate results in the presence of unit root variables. In addition, the balance order of integration allows the research to employ a robust cointegration test to enhance the results of the study. 5.0 Empirical Results and Analysis 5.1 Unit Root test

102