*3.2. Data Diagnostics*

This investigation uses annual data (time series) for the Botswana economy for the period 1980–2015. The data sources are the International Financial Statistics (IFS), World Development Indicators (WDI) and the Federal Reserve Economic Data (FRED). The study adopts the ARDL bounds testing approach, which is applicable whether all the regressors are *I*(0) or *I*(1) or a mixture of both. However, the procedure is not valid if any of the variables under examination is *<sup>I</sup>*(2). This investigation used the generalised least squares transformed Dickey–Fuller (DF-GLS) test and the augmented Dickey–Fuller (ADF) unit root test to examine stationarity. An advantage of the DF-GLS procedure is that it has high power even when the root of the times series is closer to unity (Elliot et al. 1996). The results of the tests showed that the sample is a mixture of *I*(0) and *I*(1) series. Multicollinearity was assessed using the variance inflation factor (VIF) in this study. As a rule of thumb, VIFs greater than 10 signal a serious multicollinearity problem while VIFs between 5 and 10 signal a less serious problem of multicollinearity (Pedace 2013). The VIFs for the variables in this investigation are less than 6, which is acceptable for further analysis.
