In this section, we use econometric analysis to examine the impact of financial liberalization of the financial sector. As a result of financial liberalization, I will be checking if the variables in my model, which are main variables that play a role in the process of financial liberalization has had any impact on GDP (Gross Domestic Product), the variables are foreign direct investment, financial deepening, gross national savings, and degree of openness. The data for the analysis was obtained from the statistical bulletin of the central bank of Nigeria, the World Bank and the IMF's international financial statistics. The data series covered a period ranging from 1975 to 2008 for both Ghana and Nigeria.
The primary reason for implementing policy reform is to influence the targeted economic variable; the corresponding change in this target variable would then serve as an indicator of policy impact. Liberalization is a policy mix designed to influence the target variables through the same other intermediate variables. Thus these policy instruments could impact the gross national savings, foreign direct investment, financial deepening and the general level of the economic activities. Several studies have assessed the impact of financial liberalization on savings, investment and growth in the economy. McKinnon (1973) and Shaw (1973) concludes that liberalization leads to higher level of investment and economic growth, their main argument was that higher interest rate which is as a result of liberalization would result in savings, which in turn will lead to a higher level of investment and economic growth. Some recent works have showed that though financial liberalization results in higher interest rates and financial deepening , it does not in all times lead to higher savings and investment. Ostry and levy (1995) in their work on France's economy concluded that financial development caused by liberalization reduced the rate of savings in the economy.
MEASUREMENT OF FINANCIAL LIBERALIZATION
The need for the movement from financial repression to financial liberalization gives markets a bigger role in development and allows empirical research to measure the effectiveness and impact of this role. There are a very large number of studies investigating the effect of financial openness and economic growth. Generally, the research can be divided into two different groups which employ two separate measures of financial liberalization. The first one I will be discussing is the de-facto measure of financial integration. It is a or price based financial liberalization The de-facto measure of financial liberalization can be used as an endogenous variable to measure the actual observed outcome of the enforcement of existing regulations on financial markets. de-jure measure of financial integration which is the next category, it's measures are quality based measures of financial liberalization which concentrate on events such as changing regulations and the response of the monetary authorities to financial flows Prasad, Rogoff, Wei and Kose (2003), as well as Aizenman and Noy (2003) question the reliability of de-jure measures to assess financial openness. Squalli and Wilson (2006), in their work based on the new approach for measuring openness, confirmed that de-jure financial openness measures are systematically impacted by economic and political economy factors which include commercial openness, political regime, corruption, and institutional developments, especially in developing countries, mainly due to structural problems, the de-jure measure may not be effective in finding the impact of financial liberalization. Therefore in this work we will be using the de-facto measure because it is a more realistic tool to assess financial liberalization.
Model Specification:
The Log linear model that is estimated in this paper is specified as follows:
A multiple regression will be used since the independent variables are more than one.
GDP= f (LogGNS, LogDOP, LogFD, LogFDI) ……………………………………. (1)
LogGDP=α0+α1LogGNSt+α2LogDOPt+α3LogFDt+LOGFDIt Ut
Where
LOGDOP: Degree of openness
LOGGNS: Gross national Savings
LOGGDP: Gross domestic product
LOGFD: Financial deepening
LOGFDI: Financial direct investment
NOTE:
For the sake of clarity, let us examine the variables in the model in turns.
The description of the variables in the models is contained in Table below
Description of Variables
Variables
Description of Variables
LOG(DOP)
Degree of openness of the economy measured by (X+M)/GDP
Openness has been measured in various ways by different researchers investigating this issue, but most of them all have similar features which is expressing DOP as trade in terms of its share of income for a given country. But the three most measures used includes; M/GDP, X/GDP and (X+M)/GDP(Wilson,2006). But whichever measure that is employed in a research, they all provide a method for determining how open an economy is to world trade.
LOG(GNS)
GROSS NATIONAL SAVINGS
Savings in layman's language is income not spent or deferred consumption. GNS is the sum of net foreign savings and domestic savings. Most countries have considered the act of increasing national savings as a way to reduce the dependence on foreign savings in other to protect the economy from external shocks (Serven& Schmidt)
LOG(GDP)
GROSS DOMESTIC PRODUCT
LOG(FDI)
FOREIGN DIRECT INVESTMENT
This is the most talked about form of capital flows by academicians, policy makers, e.t.c.
LOG(FD)
FINANCIAL DEPEENING (M/GDP)
It is measured as the ratio of money supply to GDP. This measures the financial depth of a country's economy.
Method of Analysis
I am going to use the Ordinary Least Square (OLS) technique in carrying out my analysis.. The technique has a unique property of being the best linearly unbiased estimator (BLUE) once its assumption have been satisfied. The criteria for decision making includes the Sign and Magnitude of Regression Coefficient, the Co-efficient of Determination (R2), the F-statistics, the Student T-test, and the D.W Statistic.
Before carrying out this analysis we have to check if the series are stationary or not in order to find out the stationarity of the series, the unit root test will be employed. Estimation and interpretation of a model depends on if the series are stationary or not like I mentioned above. Intuitively regression analysis involves using an independent variable to explain a dependent variable, if the dependent variable is y and the independent variable is x and x's properties differs from the properties of y, then it becomes difficult for x to explain y, in more sense, it means that it is hard for a stationarity series to explain the stochastic trend variation in a unit root series.
So before running any time series regression, it is important you examine the univariate properties of the variables (Koop, 2009 p169). If the result of the unit root test shows that they are stationary then you can go on to run your regression, but if they are non stationary, it means there is a problem with the data and if it is used for analysis, it may lead to spurious regression. The major problem in the data is that the test statistics do not follow the usual t-distribution under the null hypothesis.
Unit root analysis
I will be using the augmented Dickey Fuller test for each variable's in Eviews. The general formula for ADF test is as follows:
\Delta y_t = \alpha + \beta t + \gamma y_{t-1} + \delta_1 \Delta y_{t-1} + \cdots + \delta_p \Delta y_{t-p} + \varepsilon_t,
In the model above, α is a constant, and β is the coefficient on a time trend and p the lag order of the autoregressive process. Another test that was used to check for stationarity was the Phillips -Perron test, I used it because it has a more comprehensive theory of unit root non -stationarity. The test is quite similar to the Augmented Dickey Fuller test but they incorporate an automatic correction to the Augmented Dickey Fuller procedure to allow for autocorelated residuals (Brooks, 2008, pp330). Although the Phillips-perron test is more developed than the Augmented Dickey Fuller test, they still arrive at the same conclusion, and this was shown in my work. According to (Brooks 2008,p331), the results of these test will be compared with a stationarity test which is the KPSS test to see if the same conclusion is obtained, and if the results should fall under outcomes A and B below, then the conclusion is robust(Brooks,2008,pp331)
The null and alternative hypotheses under each testing approach are as follows: (Brooks, 2008, pp331)
KPSS ADF/PP
HO: yt ~ I(0) H0: yt ~ I(1)
H1: yt ~ I(1) H1: yt ~ I(0)
Reject H0 and Do not reject H0
Do not reject H0 and Reject H0
Reject H0 and Reject H0
Do not reject H0 and Do not reject H0
We checked for stationarity on all the variables for both Ghana and Nigeria, using the Phillips-perron test, ADF test and we came out with the same conclusion and when it was compared with the KPSS test, it met the first two outcomes above.
The test for the unit root showed that all the variables have unit roots and are non stationary, so it is not appropriate to examine the coefficient standard errors or their t - ratios in the test regression.
There are lots of ways to deal with these. First of all the we can difference the series to make it stationary. In most cases if the series are non- stationary, then a test will be carried out to check if the variables are co integrated. This was carried out in this work.
COINTEGRATION
If the result shows that the variables are not stationary, I will employ the co integration technique, to know if there is a long term relationship between the variables. A cointegrating relationship can be seen as an equilibrium phenomenon, since most times variables that are co integrated deviate from their relationship in the short run, but their association would return in the long run. (Brooks, 2008 Pp336). Economically speaking, two variables are will be co integrated if they have a long term or equilibrium relationship between them (Gujarati, 2004 Pp822), That is they tend to move in the same direction in the long run. Many time series are non stationary but they tend to move together in the future or as time goes on, which implies that the series are bound by some relationship in the long run, and when this happens it means that they are co integrated, but if it is the other way round, it means they are not co integrated.
If the variables are co integrated then the spurious regression problem does not apply, consequently we can run an OLS regression on the series and obtain valid results (Koop, 2008). Using the co integration technique the result showed that there is co integration, so as the above statement puts it, I can run an OLS regression on the time series.
PANEL DATA
I will exploit the cross-section and time-series dimension of our data by using panel data estimation techniques. I employed the fixed effects. Fixed effects estimates have been used to correct for the problems such as omitted variable bias that may arise from pure cross-section regressions (Islam, 1995; Caselli et al., 1997; Baltagi, 2001). The fixed effects model takes account of the unobservable country specific effects which are assumed to be fixed parameters to be estimated. The panel data consist of two countries, which are Ghana and Nigeria, from the period of 1975 to 2008.
