**3. Methodology**

To analyze the effects of monetary variables on investment in the selected Sub-Saharan African countries (Kenya, Mozambique, Nigeria, South Africa and

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

The effects of monetary policy variables on investment are based on the Keynesian theory of investment. The theory was developed by Keynes [24] who state that investment decisions are determined by a conducive environment for the investor and a long run survival behavior of an investor. For this to happen the investor need to consider the accumulation of capital which is influenced by lending rates [25–27]. The longer the investor survive in business, the more the economy can

Keynes theory of investment further compares the marginal efficiency of capital (MEC) with interest rates [26, 28, 29]. If the MEC exceeds the rates, the investment will be increased. But because the production process demands the use of more and more capital, the MEC will suddenly fall. Once the MEC equals to the level of interest rate, there will not be any additional investments on income-earning assets. Additionally, Duesenberry [28] developed the financial theory of investment which assumes that there is a relationship between the cost of capital and interest rates. The Keynesian theory of investment can be extended to include the effects of all the selected monetary variables on investment. For instance, according to Nucci and Pozzolo [30], investment is a function of the cost of capital and exchange rate. Also in Amiti and Weinstein [31] investment can be determined by money supply

It is vital to investigate the influence of monetary variables on investment activities as an investment is an important economic resource needed for economic growth. Literature suggests that monetary variables such as exchange rate do affect investment levels of a country in several setups. For instance, Osemene and Arotiba [32] advocate for a stable exchange rate environment to have positive effects of volatile exchange rate on foreign portfolio investment. Therefore, it can be argued that monetary authorities should formulate policies that result in a stable exchange rate as a way of boosting investors' confidence. These findings are enforced in Teddy [33] that a high volatile (highly unstable) exchange rate in Zambia harmed

There are several conditions found in the literature on how the exchange rate can affect investment. These conditions vary depending on the developing state of the country. In Harchaoui, Harchaoui et al. [34], the exchange rate can influence investment through three channels: domestic and foreign demand, prices of variable inputs and the investment price. When a domestic currency depreciates, sales of goods and services yield higher revenues and profits. At the same time, the variable cost and imported capital increase to counterbalance the positive effects of higher revenues [34, 35]. This is because revenue from both domestic and foreign sales is increased. Nucci and Pozzolo [30] supported this argument when they investigated the exchange rate- investment nexus for some selected Italian manufacturing firms. The authors discovered that exchange rate depreciation impacts investment positively through revenue channel and negatively through the cost channel, and added

that businesses need monopoly power to achieve this relationship.

The most important factors deliberated in the literature about what can cause positive effects of exchange rate on investment are stable exchange rate, monopoly power, the openness of trade, amount of imported inputs and developing level of a country [32, 34, 36, 37]. For instance, Atella et al. [36] emphasized that for a country's

**2. Literature review**

through bank supplies.

**2.2 Empirical literature**

private capital inflows.

grow [26].

**2.1 Theoretical literature**

**238**

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 gross fixed capital formation as stipulated in the following equation:

$$\text{GFCFF}\_{\text{it}} = \alpha + \beta\_{\text{i}} \text{MS}\_{\text{it}} + \beta\_{\text{z}} \text{LR}\_{\text{it}} + \beta\_{\text{y}} \text{ER}\_{\text{it}} + \mu\_{\text{ir}} \tag{1}$$

where GFCF measures gross fixed capital formation (investment); MS measures money supply; LR measures lending rates and ER measures exchange rates; α measures the constant of the model; β1 3 <sup>−</sup> measures the estimates of monetary policy variables, and μ the error term to make the model more accurate and cater for any input variable omissions.

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, an autoregressive panel is eligible [52, 57].

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

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 to differ across sections [53].

**241**

**Table 1.**

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

**Variable Test Test equation Level**

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

*GFCF* LLC I&I 0.0011 —

*MS* LLC I&I 0.0138 —

ER LLC I&I 0.9968 0.0000

LR LLC I&I 0.4142 0.0000

*I&I: individual and intercept; II&T: individual, intercept, and trend.*

*Summary of panel unit root test results.*

**p-value**

II&T 0.0350 — None 0.0155 —

II&T 0.0491 —

II&T 0.0304 — None 0.0342 —

II&T 0.0000 — None 0.7522 0.0000

II&T 0.0000 —

II&T 0.0000 — None 0.0000 —

II&T 0.1617 0.0000 None 0.9999 0.0000

II&T 0.3255 0.0000

II&T 0.3980 0.0000 None 1.0000 0.0000

II&T 0.0376 — None 0.3539 0.0000

II&T 0.0967 0.0000

II&T 0.0567 0.0000 None 0.8643 0.0000

IPS I&I 0.0116 —

IPS I&I 0.0000 —

IPS I&I 1.0000 0.0000

IPS I&I 0.3366 0.0000

Fisher-ADF I&I 0.2363 0.0000

Fisher-ADF I&I 0.9999 0.0000

Fisher-ADF I&I 0.0000 —

Fisher-ADF I&I 0.0124 —

**1st order p-value**

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

three stationarity test, LLC; IPS; and Fisher ADF.

**4. Results and discussions**

*Effects of Some Monetary Variables on Fixed Investment in Selected Sub-Saharan African... DOI: http://dx.doi.org/10.5772/intechopen.93656*
