**4. Results and discussions**

*Linear and Non-Linear Financial Econometrics - Theory and Practice*

measures the constant of the model;

an autoregressive panel is eligible [52, 57].

sections and gives results of the full panel [57, 60].

any input variable omissions.

Tanzania), the study used panel annual data collected from the World Bank. The study period 1980–2019 and countries are selected due to obtainability of data, the chosen variables are based on the Keynesian theory of investment and some reviewed empirical literature [25, 30, 31]. The selected three monetary variables are money supply, lending rates, and exchange rate and investment is measured by

where GFCF measures gross fixed capital formation (investment); MS measures

policy variables, and μ the error term to make the model more accurate and cater for

This study employs a panel analysis that is more time-series than cross-sectional. The first step is to check for stationarity of variables as it is the common characteristics in time series dominated analysis [52, 53]. To test for stationarity, three tests were used to ensure the inexistence of unit root in the study data namely Levin-Lin-Chu (LLC) test, the Im-Pesaran-Shin (IPS) test and the Fisher-ADF. The LLC test allows for heterogeneity in the intercept terms, the IPS and the Fischer are less restrictive as they allow coefficients to be heterogeneous [54, 55]. The Fischer outperforms the IPS when it comes to the size-adjusted power [56]. Therefore, all the tests are used to reinforce each other and allow us to make robust decisions about which panel type to use for the analysis. If there are different orders of integration,

Panel cointegration is useful to determine if there are long term effects between

investment and the monetary variables. Additionally, panel cointegration can address issues of heterogeneity in the panel by looking at the parameters, how many cointegrating relationships across countries and if there is cointegration in different countries [57, 58]. For the cointegration exercise, the Pedroni, Kao and Johansen-Fisher tests are employed [52, 57]. The Pedroni consider four-panel statistics and three group panel statistics to test the presence of cointegration [59]. The advantage for the within-dimension-based four panels is to identify a first-order autoregressive process which is assumed to be the same in all countries in the series, and the three group panels are between-dimension-based and allow for parameters to vary across countries [59]. The Kao test reinforces the Pedroni as it uses the same approach but differs by specifying country-specific intercepts and homogeneous estimates on the first stage regressors. The Fischer combines individual cross-

After the realization that there is cointegration (long-run relationship) and variables are integrated at different orders, a panel autoregressive distributed lag (ARDL) model is employed. The ARDL regression is necessary to find the nature of coefficients, whether the negative or positive relationship and significant or not. If variables show different orders of integration in the unit root analysis and cointegration exist, then the ARDL is the best estimator to find short-run, long-run and error correction estimates in a single model [59, 60]. In this model, the error correction term can be determined by integration of short-run adjustments with long-run equilibrium maintaining the long-run information. The advantage of having a large panel ARDL starting from 1980 to 2019 is to address the bias problem caused by correlating error terms with the mean-differenced regressors. The cointegrating form of the ARDL model called the pooled mean group estimator permits estimates

money supply; LR measures lending rates and ER measures exchange rates; α

β

GFCF MS LR ER it =α+β +β +β +µ 1 it 2 it 3 it it (1)

1 3 <sup>−</sup> measures the estimates of monetary

gross fixed capital formation as stipulated in the following equation:

**240**

to differ across sections [53].

Brooks [61] emphasizes that variables should be free from a unit root to avoid spurious regression, therefore this study needed to difference the variables to attain stationary variables. **Table 1** provides panel stationarity tests as estimated using three stationarity test, LLC; IPS; and Fisher ADF.


**Table 1.**

*Summary of panel unit root test results.*

In **Table 1** gross fixed capital formation (GFCF) and money supply (MS) are generally shown to be integrated at levels I(0), while exchange rates (ER) and lending rates (LR) are integrated of order one I(1). Therefore, the variables used in the study are a mixture of I(0) and I(1) and none of them is I(2) which paves a way to run the panel ARDL [52, 60]. It is stated in Nkoro and Uko [60] that variables that show different orders of integration can be estimated best with ARDL. Moreover, cointegration results indicate the existence of a long-run relationship but do not give estimates, hence in addition to the cointegration analysis, there is a need for a robust estimation technique like ARDL.

**Tables 2**–**4** provide results of panel cointegration tests as estimated for the model specified in Eq. 1 under the Pedroni, Kao and Fisher-ADF tests for cointegration, respectively.

The Pedroni test results presented in **Table 2** confirm cointegration in three out of seven statistics. One out of four within dimensions accept the alternative hypothesis of cointegration at 10% significance levels (Panel v-Statistics) whereas two out of three between dimensions accept the alternative hypothesis of cointegration at 1% significance level (Group PP- statistics and Group ADF statistics). The Kao panel cointegration tests results, as shown in **Table 3** also confirm cointegration by rejecting a null hypothesis of no cointegration at 1% level of significance. **Table 4** illustrates a strong cointegration between the variables in the Fisher-ADF test. This is displayed by both the trace and the max Eigenvalues which both detect at least two cointegrated relationships between investment and the selected independent variables. All three cointegration tests reveal that a long-run relationship exists between the variables for the selected panel. This implies that investment has a long-run relationship with the selected monetary variables in the chosen panel of five Sub-Saharan countries. **Table 5** provides estimates of the model specified in Eq. 1, where investments are regressed against monetary variables such as lending rates, money supply and exchange rate.


*\* and \*\*\* indicate that the p-values are significant at 10 and 1% level of significance, respectively.*

#### **Table 2.**

*Summary of Pedroni cointegration test results.*


**243**

**Table 5.**

*Effects of Some Monetary Variables on Fixed Investment in Selected Sub-Saharan African...*

**Table 5** shows the summary of panel ARDL long-run and short-run results. As depicted in **Table 5**, lending rates, money supply and exchange rates all have a strong long-run relationship significant at 1% level with investment. Lending rates, as economic theory suggests, have been found to have a negative relationship with investment in this study [25, 44, 51]. The results are found to be in line with those of Malawi and Bader [44] and Ashraf et al. [50] where an increase in the real interest rate by 1% reduces the investment. It has been found that interest rate plays an

It turns out that the money supply is positively related to investment for our selected panel (**Table 5**). According to the results, when the money supply is increased, a relative increase in investment follows. Many scholars established that money supply has a positive long-run relationship with investment [38, 42, 62]. On the contrary, it has been discovered that there may exist a negative relationship between money supply and investment [31, 40, 43]. Li and Yang [40] further add that money supply is a weak instrument to be used to influence real estate invest-

The exchange rate also shows a significant and positive long-run relationship with investment in **Table 5**. It has been argued in the literature review section that a country's investment level can benefit from the exchange rate, provided exchange rate is stable [25, 30, 34, 36]. The argument is based on the fact that a depreciating exchange rate is associated with a stable environment and strong market power [36].

None 75.81\*\*\* 0.0000 47.55\*\*\* 0.0000 At most 1 15.57\*\* 0.0490 15.96\*\* 0.0429 At most 2 7.841 0.4492 6.332 0.6101 At most 3 9.222 0.3239 9.222 0.3239

**Variables Coefficient Std. Error t-Statistic P-value**

Lending rates −3.523144 0.677454 −5.200565\*\*\* 0.0000 Money supply 18.87173 2.946935 6.403849\*\*\* 0.0000 Exchange rates 0.012514 0.001282 9.763669\*\*\* 0.0000

Error correction term −0.834634 0.371897 −2.244262\*\* 0.0274 D(Investment) 0.133988 0.370028 0.362103 0.7182 D(Lending rates) 10.51841 3.009819 3.494700\*\*\* 0.0008 D(Money supply) 14.16886 24.00636 0.590213 0.5566 D(Exchange rates) 0.019939 0.111209 0.179291 0.8581

**P-value Fisher stat. (from** 

**max-Eigen test)**

**P-value**

**Fisher stat. (from trace test)**

*\*\* and \*\*\* indicate that the p-values are significant at 5% and 1% level of significance, respectively.*

*\*\*, and \*\*\* indicate that the p-values are significant at 5% and 1% level of significance, respectively.*

*DOI: http://dx.doi.org/10.5772/intechopen.93656*

important role in investment decision making.

ment in an inflation targeting environment.

*Summary of Johansen-Fisher panel cointegration test results.*

*Summary of long-run and short-run panel ARDL estimates.*

**Hypothesized no. of** 

**Long run estimates**

**Short-run estimates**

**CE(s)**

**Table 4.**

**Table 3.**

*Summary of Kao panel cointegration test results.*
