**5. Methodology**

This section discusses the methodology used in the study which was inspired by the FDI theory and the interest rates and investment theory. The study adopts an econometric model by using the panel ARDL procedure to examine occurrence of foreign direct divestment in Sub-Saharan countries [32].

#### **5.1 Data**

The study used panel annual data of six countries in the Sub-Saharan economies, emerging and developing economies from 1998 to 2018 due to availability of data and these countries were randomly selected to avoid any biases. The selected countries are South Africa, India, Nigeria, Italy, Egypt and Botswana. The secondary data for the following variables: foreign direct investment, trade openness, lending rates, real gross domestic product per capita and urbanisation was obtained from the World Bank.

$$FDI\_{\rm \tiny \rm \}} = \infty + \beta\_1 LOPENNESS\_{\rm \tiny \rm \}} + \beta\_2 LR\_{\rm \tiny \rm \}} + \beta\_3 LRGDP\_{\rm \tiny \rm \}} + \beta\_4 URB\_{\rm \tiny \rm \}} + \varepsilon\_{\rm \tiny \rm \}} \tag{1}$$

where ∝ = represents the constant parameter, LOPENNESS = log of trade openness measured by the sum of imports and exports. Trade openness is expected to have a positive influence on foreign direct investment according to the FDI theory. LR = lending rates which represents the cost of borrowing. Lending rates are expected to have a negative impact on foreign direct investment according to the inverse relationship between interest rates and investment theory. LRGDP = log of real gross domestic product per capita proxy for market size. URB = urbanisation as a percentage of total population.

#### **5.2 Empirical analysis**

The following econometric measures are undertaken to investigate the existence of foreign direct divestment in the Sub-Saharan, developing and emerging economies.

#### *5.2.1 The panel unit root test*

Before testing for long run cointegration among variables, the study used several tests for stationarity. The panel unit root test is conducted to determine the order of integration among variables which helps in identifying the best suitable model for the data used in the study. The several approaches to unit root testing used in the study for the panel data were Levin, Lin and Chu (2002) (LLC) test; Im, Pesaran, Shin (2003) (IPS) test and the Fisher- Augmented Dickey-Fuller (ADF) test as supported by Maddala and Wu [33].

#### *5.2.2 The panel cointegration test*

The panel cointegration test is useful when examining the existence of long run relationships between the regressors and the regressed variables. The Kao panel cointegration test which follows the same basic approach as the Pedroni test extends the Engle-Granger [34] framework to panel cointegration test. The distinct feature of the Kao test from the Pedroni test is that it typically stipulates the cross sections exact intercept and similar coefficients of regressors on the early stage. Also, the Kao and the Pedroni panel cointegration tests are generally used to examine the long run relationship between variables used in a study [35]. The Johansen-Fisher cointegration test uses the findings of the individual independent tests [36]. The Johansen-Fisher panel cointegration pioneered by Maddala and Wu [33] to examine the cointegration in panel data by incorporating the test from individual cross

*Foreign Direct Divestment Phenomenon in Selected Sub-Saharan African Countries DOI: http://dx.doi.org/10.5772/intechopen.100304*

sections to get a test statistic for the whole panel. Say the π\_i is the p-value from the distinct cointegration test for cross section i, under the null hypothesis of the panel.

$$-2\sum\_{i=1}^{N} \log\left(\boldsymbol{\varkappa}^{2}\boldsymbol{\pi}\_{i}\right) \to \boldsymbol{\varkappa}^{2}2N \tag{2}$$

Therefore, the value of <sup>2</sup> *x* is built upon the MacKinnon-Haug-Michelis p-values for Johansen's cointegration trace test and maximum eigenvalue test.

#### *5.2.3 The panel autoregressive distributed lag procedure*

The autoregressive distributed lag (ARDL) procedure supported by Pesaran et al. [32] which combines lags of both explained and explanatory variables as regressors is used in the study. The ARDL model is used owing to its ability to join small sample size data and yet generating useful findings [34, 37]. Johansen and Juselius [37] point that the traditional cointegration technique have fewer advantages compared to ARDL that has several advantages. First, it requires small sample size, with variables that are pure I(1), purely I(0) or integrated at different orders of integration but not I(2) [37]. Secondly, it does not require variables to be integrated in the same order compared to the Johansen cointegration approach. Thirdly, ARDL approach caters for any structural breaks in a time series. And lastly, this approach carries a method of measuring the long run and short run findings of one variable on the other and as well distinct both once an appropriate selection of order of the ARDL model is made [38]. Regardless of these advantages, the study employed this model due to its small sample sized panel data and the variables used are integrated at different orders of integration.

### **6. Results and discussion**

Levin, Lin & Chu, Im, Pesaran and Chin W-stat and Augmented Dickey Fuller – Fisher Chi-square tests were employed to perform the panel unit root test and it was discovered that the variables are integrated at different orders of integration [ 0 *I*( ) and *I*(1 ]) and none of which are *I*(2) . This gave justification to use panel ARDL. For instance, foreign direct investment was stationary at level, *I*(0) for all tests. Trade openness was stationary at *I*(1) for IPS and Fisher-ADF. Further, lending rates became stationary at 1%, 5% and 10% level of significance for all tests after first differencing. Gross domestic product was stationary at *I*(1) for LLC at 10% level of significance and IPS and Fisher-ADF at 5% respectively. Finally, urbanisation was stationary at *I*(1) for all tests (**Table 1**).

Having established the order of integration for the panel series, the next step is to examine the probability of long-run association between variables. The study will begin with the Kao panel cointegration test. The p-value of 0.004 in the ADF test is less than 0.05 and thus the null hypothesis of no cointegration is rejected and fail to reject the alternative hypothesis of cointegration between the variables. Therefore, the variables have a long run relationship according to the Kao panel cointegration test (**Table 2**).

The test results of the Johansen Fisher panel cointegration with linear deterministic trend test are shown in **Table 3**. Johansen Fisher panel cointegration test results indicate that the trace statistic has five cointegrating equations and the Fisher


**Table 1.**

*Summary of panel unit root test results.*

maximum-eigen test also shows five cointegrating equations at a 10%, 5% and 1% significance level. The first four equations shows that all p-values are statistically significant at 10%, 5% and 1% level of significance respectively (only one equation at 10%) thus rejecting the null hypothesis of no cointegration. This indicates that there is long run relationship between the variables.

In **Table 4**, the test results of Johansen Fisher panel cointegration with no deterministic trend test indicates that all p-values are significant at 10%, 5% and 1% level of significance respectively (only the last equation at 10%). Therefore, the *Foreign Direct Divestment Phenomenon in Selected Sub-Saharan African Countries DOI: http://dx.doi.org/10.5772/intechopen.100304*


#### **Table 2.**

*Kao panel cointegration test results.*


*Note: \*, \*\*, and \*\*\* indicate that the p-values are significant at 10%, 5% and 1% level of significance respectively. The Fisher's test applies regardless of the dependent variable.*

#### **Table 3.**

*Johansen Fisher panel cointegration with linear deterministic trend test.*


*Note: \*, \*\*, and \*\*\* indicate that the p-values are significant at 10%, 5% and 1% level of significance respectively. The Fisher's test applies regardless of the dependent variable.*

#### **Table 4.**

*Johansen Fisher panel cointegration with no deterministic trend test.*

null hypothesis of no cointegration is rejected, indicating that there is a long run relationship between the variables.

**Table 5** shows the test results of the Johansen Fisher panel cointegration with Quadratic deterministic trend test. The results indicate that all p-values are significant at 10%, 5% and 1% level of significance, meaning that the null hypothesis of no cointegration is rejected. This indicates that the variables are cointegrated in the long run.

**Table 6a** and **b** show the Autoregressive Distributive Lag Short Run and Long Run Results. The long run results indicated that there is an insignificant long run relationship between gross domestic product and FDI (dependent variable) and cannot be used for policy making purposes in this study. These findings contradicts with the findings of Pegkas [39] with FDI as an independent variable. Further, the results showed that lending rates coefficient had a negative significant long run impact on FDI at 5% level of significance. This indicates that if lending rates were to increase by one percent, FDI for the panel six Sub-Saharan and developing economies would

#### *Macroeconomic Analysis for Economic Growth*


#### **Table 5.**

*Johansen Fisher panel cointegration with quadratic deterministic trend test.*


*Notes: D-denotes differenced results for short run.*

#### **Table 6.**

*(a and b) Autoregressive distributed lag short run and long run results.*

decrease by 16.7845%. The findings are in line with Musyoka and Ocharo [31] who discovered that real interest rates have a negative significant impact on foreign direct investment inflows. This further implies that increasing the cost of borrowing will lead to foreign direct divestment in the countries, thus resulting in a decline in inflows. This also indicates that FDI is particularly sensitive to increase in cost of borrowing as suggested by theory on interest rates and investment.

Trade openness had a positive significant long run relationship with FDI at 10%, 5% and 1% level. This concur with the findings of Kumari and Sharma [2] which revealed that trade openness was significant at 10% level. Furthermore, these findings supports the FDI theory that states that for investment purposes, degree of openness indicates the ease with which a host country is accessible in the world market. Finally, urbanisation showed a negative significant long run relationship with FDI at 10%, 5% and 1% level of significance.

The relationship between the coefficients in **Table 6** can also be represented in an equation to further understand their meaning and their influence on FDI. Lending rates and urbanisation both had a significant negative influence on foreign *Foreign Direct Divestment Phenomenon in Selected Sub-Saharan African Countries DOI: http://dx.doi.org/10.5772/intechopen.100304*

direct investment indicating that an increase in these variables would lead to what we refer to as foreign direct divestment for these selected Sub-Saharan and developing economies. Trade openness showed a positive significant influence on FDI, implying increasing trade openness increases FDI and a decrease leads to foreign direct divestment.

$$\begin{array}{l} \text{FDI}\_{\textit{u}} = \text{\textit{\textdegree}} + \text{20.6322LOPEINES}\_{\textit{u}} + \text{20.6322LR}\_{\textit{u}} + \text{-1.4117LRGDP}\_{\textit{u}}\\ + \text{-2.2362URB}\_{\textit{u}} + \text{\textit{\textdegree}} \end{array} \tag{3}$$

In the short run, the critical part of the analysis is the error correction term (ECT), which must always be negative according to theory otherwise the model will be explosive and may never return to equilibrium if it is positive. ECT is also referred to the speed of adjustment which shows whether the estimated economic models will be able to return to equilibrium or not and at what speed.

The estimated speed of adjustment, which is at −0.925834, has a negative sign and it is significant at 1% level of significance, as expected by theory. A highly significant speed of adjustment also confirms the existence of cointegration among the variables and a stable long run relationship. This indicates that there is a longrun causality running from the independent variables to the dependent variable and that approximately 93 percent of disequilibrium is corrected each year. It will take 93 percent each year for foreign direct investment activity to return to equilibrium, which is not a slow movement back to equilibrium.

#### **7. Conclusion and recommendations**

The aim of the study was to investigate the presence of any noticeable foreign direct divestment, that is any perceptible rapid drop of FDI inflows in the Sub-Saharan African countries using annual panel data from 1998 to 2018. The FDI inflows trends established that there was a rapid decline of FDI inflows for all selected countries after the 2008 global financial crisis and the spillover effects US-China trade war. The study used panel autoregressive distributed lag to determine long run and short run equation for variables that are likely to influence foreign direct investment. The study began by testing for unit root using Levin, Lin & Chu, Im, Pesaran and Chin W-stat and Augmented Dickey Fuller – Fisher Chi-square tests and it was discovered that the variables are integrated at different orders of integration [I(0) and I(1)] and none of which are I(2). The Kao and the Johansen Fisher panel cointegration tests confirmed the long run cointegration among the variables.

The findings of the long run equation revealed that lending rates and urbanisation have a negative and significant influence on foreign direct investment. Further, the findings revealed an insignificant influence of real gross domestic product per capita on FDI. Finally, trade openness showed a positive significant impact on foreign direct investment. This was in line with the FDI theory that states that for investment purposes, degree of openness indicates the ease with which a host country is accessible in the world market. The error correction model also divulged that approximately 93% of disequilibrium will converge towards equilibrium annually.

Like any quantitative or econometrics research, this study also had some limitations. For instance, the period of study ended in 2018 due to unavailability of data for some countries. Lack of data and literature on foreign direct divestment and the absence of data on key variables such as corruption, political risks, labour costs, natural resources and exchange rates may be perceived as limitations. Also,

controlling variables such as corruption, exchange rate, political risk and labor cost could make significant improvements to this study. The study also have practical and significant implications for researchers, scholars, governments, policy makers, managers and notably foreign investors.

We recommend policies that increase FDI through cost of borrowing since increasing interest rates result in foreign direct divestment. Real gross domestic product per capita (market size) cannot be used for policy making purposes in the study. Trade openness makes a country accessible on the world market, therefore, policies that promote foreign trade such as exporting complex products, shifting production capabilities from raw materials to more sophisticated products and services, reduce dependence on the primary sector, trade liberalisation, free trade agreements and open trade systems could help reduce the presence of foreign direct divestment in the selected countries. Finally, urbanisation deter foreign direct investment, therefore countries should invest on infrastructure and reduce poverty in rural areas to transform them into urban areas to reduce urbanisation.
