Working Paper Series: Special Edition of 2016 to 2018 Interns

to this variable may be permanent versus temporary on economic performance. Other benefits of the panel unit root test are noted in the power of such test compared to individual unit root tests of times series data. Panel unit root tests Im, Pesaran & Shin (IPS) and Hadri panel unit root tests were used. The IPS tests individual unit root processes assuming that varies across the cross-sections. Furthermore, the test combines individual unit root tests to derive a panel specific result. On the other hand, the Hadri builds on the premise that there is a common unit root process. The test is based on residuals obtained from individual OLS regressions; one may specify a constant or a constant and trend. While the IPS tests the null as unit root, Hadri tests the null as no unit root. Panel Cointegration tests Cointegration has long been a statistical consideration in empirical studies. The concept concerns itself with the long-run relationship between non-stationary variables. The algebra of cointegration states that if two variables are integrated of the same order (say, I(1) ) with their linear combination of a lower order (say, I(0) ) of integration, then the variables are said to be cointegrated. However, the need to difference a non-stationary series to obtain stationarity potentially throws away this important long-rung relationship Maddala and Lahiri (2009). Therefore, running this test was important to determine if the long-run relationship has been eradicated via differencing or not. Further, if there is no cointegration amongst the variables it suggests the pure Vector Autoregressive model (VAR) is sufficient but if there is cointegration then a Vector Error Correction Model (VECM) is appropriate because it corrects the presence of cointegration with an error correcting term according Karanfil and LI (2015). One of the key drawbacks of panel cointegration which panel unit root is also subjected to is short time series such that more robust methods may have to be employed. The Fisher combined Johansen panel cointegration test was employed, examining the nature of cointegration between the variables. The non-residual based test provides a number of benefits over its residual based counterparts. These include (i) it is not affected by the randomness of the residuals (ii) it provides valid test statistic and (iii) it has the ability to speak to the precise number of cointegrating relationships (iv) not affected by sample size and (v) can handle imbalance order of integration relationships.

100