A series is said to be stationary if the mean and autocovariances of the series do non depend on clip. A stationary clip series is of import because nonstationarity implies that its behaviour can merely be studied for the clip period under consideration. As a effect, it is non possible to generalise it to other clip periods. Therefore, for the intent of calculating, such ( nonstationary ) clip series may be of small practical value.

Hence, in this survey the first measure is to prove for stationarity of the series. The Augmented Dickey Fuller unit root trial is used for this intent. The ADF trial corrects for higher order consecutive correlativity by adding lagged differenced footings on the right-hand side.

The standards to find stationarity is as follows:

H0: Variables are non-stationary

H1: Variables are stationary

The trial is conducted at 5 % significance degree and the p-value bespeaking the exact point of significance, should be less than 0.1.

Standards for stationarity:

P-value & lt ; 0.1 and T-statistic must lie outside the critical value.

5.2.1 Result for Stationarity: Model 1

Table 5: Stationarity Consequences 1

Variables

t-value

1 %

5 %

10 %

p-value

GDPCAP

0.118

-3.709

-2.983

-2.623

0.9673

DGDPCAP

-4.865

-3.709

-2.983

-2.623

0.0000

GFCF

-1.162

-3.723

-2.989

-2.625

0.6899

DGFCF

-5.051

-3.709

-2.983

-2.623

0.0000

INF

-2.441

-3.709

-2.983

-2.623

0.1304

DINF

-6.123

-3.709

-2.983

-2.623

0.0000

HC

-1.634

-3.709

-2.983

-2.623

0.4654

DHC

-6.547

-3.723

-2.989

-2.625

0.0000

Florida

0.649

-3.723

-2.989

-2.625

0.9888

DFL

-3.103

-3.723

-2.989

-2.625

0.0263

Beginning: Computed

5.2.2 Result for Stationarity: Model 2

Table 6: Stationarity Consequences 2

Variables

t-value

1 %

5 %

10 %

p-value

FLI

1.551

-3.730

-2.992

-2.626

0.5081

DFLI

-8.000

-3.702

-2.980

-2.622

0.0000

Beginning: Computed

The Stationarity trial for the other variables are the same as theoretical account 1.

From the above tabular arraies, it can be seen that at 5 % of significance, the variables are ab initio non stationary hence they need to be differenced of order one-I ( 1 ) to go stationary at 5 % critical value.

In the instance of GDPCAP, the trial statistic lies inside the critical part at 5 % significance degree, hence, the void hypothesis is non rejected. Besides, the p-value is more than 0.1. Hence, the GDPCAP series is non stationary at order 0. It needs to be differenced of order one to be stationary: p-value is less than 0.1. Similarly, the void hypothesis is non rejected for GFCF. Hence, GFCF follows a non-stationary flow. Even the p-value is undistinguished. When differenced, the p-value becomes less than 0.1 and the t-statistic prevarications outside the critical part. Same applies to INF, SER, FL and FLI. They need to be differenced of order one to go stationary.

Since all the variables in this survey are of the same order of integrating I ( 1 ) , the Engle- Granger attack will be used over other techniques. This is because the Engle-Granger attack does non let for different grade of stationarity.

5.3 Arrested development

The variables in standard signifier have been regressed and the undermentioned consequences have been obtained.

5.3.1 OLS Appraisal: Model 1

The original theoretical account

GDPCAPt = a + b1 GDFCt + b2 HCt + b3 INFt + b4 FLt

is run utilizing OLS and the undermentioned consequences are obtained:

Table 7: Arrested development Consequences 1

Regressor

Coefficient

Std Error

t-ratio

p-value

Florida

0.5899757

0.075617

7.80

0.000

GFCF

0.2988969

0.0241362

12.38

0.000

INF

-0.005787

0.0014132

-4.09

0.00

HC

0.4283772

0.0522184

8.20

0.004

CONSTANT ( a )

-1.37193

0.7029715

-1.95

0.061

Beginning: Computed

DW Statistics=1.971615 F ( 4, 28 ) = 1335.24 ( 0.000 )

R2 =0.9948 R2-Bar = 0.9940

Mean VIF = 5.27 Tolerance ratio=0.1897

The theoretical account can be rewritten as:

GDPCAPt = -1.37193 + 0.5899757 FLt + 0.2988969 GFCFt -0.005787 INFt

+0.4283772 HCt

5.3.2 OLS Appraisal: Model 2

The original theoretical account

GDPCAPt = a + b1 GDFCt + b2 HCt + b3 INFt + b4 FLIt

is run utilizing OLS and the undermentioned consequences are obtained:

Table 8: Arrested development Consequences 2

Regressor

Coefficient

Std Error

t-ratio

p-value

Florida

0.0235042

0.0022604

10.40

0.000

GFCF

0.2483125

0.0224526

11.06

0.000

INF

-0.0038688

0.0011761

-3.29

0.003

HC

0.1857113

0.0525935

3.53

0.001

CONSTANT ( a )

2.979982

.904668

3.29

0.003

Beginning: Computed

DW Statistics= 2.340596 F ( 4, 28 ) = 2048.91 ( 0.000 )

R2 =0.9966 R2-Bar = 0.9961

Mean VIF = 9.34 Tolerance ratio=0.107

GDPCAPt = 2.979982 + 0.0235042 FLt + 0.2483125 GFCFt -0.0038688 INFt

+0.1857113 HCt

5.3.3 Significance

Standards for look intoing whether a variable is statistically important:

t-statistic & gt ; 2 and,

p-value & lt ; 0.05.

From the above tabular arraies, it can be concluded that all the variables are statistically important since the t-ratios are greater than two and the p-values are less than 0.5. Infact, the exact significance degree is normally known as the p-value.

5.3.4 Specious Arrested development

In regressing a clip series variable on another clip series variable ( s ) , one frequently obtains a really high R2 ( in surplus of 0.9 ) even though there is no meaningful relationship between the two variables. This state of affairs exemplifies the job of specious, or bunk, arrested development. Granger and Newbold ( 1974 ) argued that R2 & gt ; vitamin D ( Durbin-Watson ) is a good regulation of pollex to anticipate that the estimated arrested development is specious. From the consequences above, R2 is less than vitamin D ( 0.9940 & lt ; 1.971615 ) in the first theoretical account every bit good as in the 2nd theoretical account ( 0.9961 & lt ; 2.140596 ) demoing a non specious arrested development.

5.3.5 Goodness of tantrum

R2 measures the goodness of tantrum of a arrested development equation. This term steps by how much the fitted equation explain fluctuations in Y. R2 must lie between 0 and 1. If it is one so the fitted arrested development line explains 100 % of the fluctuation of Y. Hence the higher the R2, the better is the step of fittingness. is adjusted R2 which takes into history the loss of grades of freedom associated with adding excess variables. Hence it is a better measuring. In the consequences, the`R2 is 0.9940 and 0.9961in theoretical account 2 demoing a really good step of fittingness.

R2 and F

There exists a relationship between R2 and the F-test. The larger the value of R2 the greater the value of the F-test. The F – trial which is a step of the overall significance of the estimated arrested development, is besides a trial of the significance of R2. A important F-test indicates that the ascertained R-squared is dependable, and is non a specious consequence of unfamiliarity in the information set. Thus, the F-test determines whether the proposed relationship between the response variable and the set of forecasters is statistically dependable. In our surveies, the F-test is 1335.24 in the first theoretical account and 2048.91 which means that the theoretical accounts are rather important overall.

Muticollinearity

Multicollinearity is a state of affairs whereby two or more explanators move in perfect lock measure with one another so that one is a multiple of the other. A “ authoritative ” symptom of multicollinearity is “ High R2 but few important T ratios ” . From the above tabular arraies, the R squares are rather high but the t ratios are all important screening non multicollinearity. However to corroborate, another trial is carried out- the VIF and Tolerance trial. As a regulation of pollex, multicollinearity is present if the mean VIF value is greater than 10s and the tolerance ratio is less than 0.1. However the VIF ratio above is 5.27 ( & lt ; 10 ) and the tolerance ratio is 0.1897 ( & gt ; 0.1 ) bespeaking low multicollinearity. In the 2nd theoretical account, the VIF ratio is 9.34 ( & lt ; 10 ) and the tolerance ratio is 0.107 ( & gt ; 0.1 ) . It shows a higher grade of multicollinearity as compared to the first theoretical account.

Autocorrelation

Autocorrelation occurs in time-series surveies when the mistakes associated with a given clip period carry over into future clip periods. Autocorrelation may happen because of concern rhythms, wrong functional signifier, Cobweb phenomenon. There are several ways to observe autocorrelation but the most used 1 is the Durbin Watson 500 trial.

Durbin-Watson has derived a lower edge deciliter and an upper edge dU such that if the computed vitamin D lies outside these critical values, a determination can be made sing the presence of positive or negative consecutive correlativity. If the statistic prevarications near the value 2, there is no consecutive correlativity. But if the statistic prevarications in the locality of 0, there is positive consecutive correlativity. If it lies in the locality of 4, there is grounds of negative consecutive correlativity. From the above consequences, the DW statistics lie near 2 i.e, 1.971615 ( first theoretical account ) and 2.340596 ( 2nd theoretical account ) demoing no grade of consecutive correlativity.

Long tally Analysis of Coefficients

In simple or multiple additive arrested development, the size of the coefficient for each independent variable gives you the size of the consequence that variable is holding on your dependant variable, and the mark on the coefficient ( positive or negative ) gives you the way of the consequence. In arrested development with multiple independent variables, the coefficient tells you how much the dependent variable is expected to alter when that independent variable additions by one, A keeping all the other independent variables constant.

5.4.1 Fiscal Liberalization

The capable affair is the impact of Financial liberalization on economic growing in Mauritius and in Chapter two and three we have seen that there is a direct relationship between fiscal liberalization. This is backed by both theories such as Mac Kinnon and Shaw and by empirics such as Lanyi and Saracoglu ( 1983 ) , King and Levine ( 1993 ) and Hallwood and MacDonald ( 1994 ) . In many developed and developing states, private sector recognition has played a critical function and is considered as the engine of economic growing and development by supplying resources for investing to the private sector ( Barth and Calari, 2006 ; Levine, 1997, Levine and Renelt, 1992, King and Levine, 1993a and 1993b ) . The coefficient of fiscal liberalization for the first theoretical account is 0.58997 implicating that for every one per centum addition of fiscal liberalization, economic growing additions by 59 per centum. Equally far as the 2nd theoretical account is concerned, for every enterprise taken towards fiscal liberalization, economic growing additions by 2.35. Hence, the tabular arraies support the theories since the t-ratio is important and consequently there is a direct relationship between economic growing and fiscal liberalization. Furthermore harmonizing to Jankee ( 2006 ) , “ the major policy deduction is that the chase of fiscal liberalization and banking sector development is no uncertainty a right scheme to accomplish higher economic growing. ”

Table 9: Percentage distribution of GDP by fiscal sector, 1980-2009

1980

1985

1990

1995

2000

2005

2009

Fiscal Intermediation

5.0

4.7

4.9

6.5

9.7

10.3

11.5

Insurance

3.3

2.9

1.5

2.1

2.3

2.9

2.8

Banks

1.7

1.8

2.0

4.4

6.6

6.2

7.4

Other

0.8

1.2

1.3

Real Estate and Business Services

12.7

11.1

8.9

8.5

8.9

10.2

11.9

Beginning: Computed

The tabular array represents fiscal sector part to GDP every bit shortly as liberalization started. It shows that following fiscal liberalization, the fiscal sector has so contributed towards the economic development of the state which supports the findings of the arrested development.

5.4.2 GCFC-Gross Fixed Capital formation

Theoretically, the gross capital formation affects the economic growing positively either by increasing the physical capital stock in domestic economic system straight, Plossner ( 1992 ) or by advancing the engineering indirectly, Levine and Renelt ( 1992 ) . Hence investing placeholder by GFCF, has a positive relationship with economic growing. Both the theoretical accounts show a positive relationship as expected. However from theoretical account 1 for every 1 % addition of investing, economic growing additions by 39.50 % . From theoretical account 2, every 1 % addition of investing, economic growing additions by 24.83 % .

In add-on to the restructuring and modernisation of the fabric and sugar sectors in Mauritius, much more accent was put on the development of the ICT sector and the publicity of Mauritius as a seafood hub in the part, utilizing bing logistics and distribution installations at the Freeport ( free trade zone at the port and airdrome ) . To farther diversify the economic base and generate sustainable growing, the authorities engaed in following economic activities: the land-based pelagic industry, ) cordial reception and belongings development, health care and biomedical industry, ) agro-processing and biotechnology, ) the cognition industry, and renewable energy. In this manner, investing bossts cardinal sectors of the economic system and finally contributes to economic growing.

INF-Inflation

From the tabular arraies, rising prices as expected has an opposite relationship with economic growing. That is, as rising prices rate additions, economic growing lessenings. This is why rising prices acts a deflator to GDP. The IMF study 2010 states that the mean one-year rising prices rate between 2000 and 2009 in Mauritius was 6.0 % . However, the coefficient shows that for any 1 per centum addition in rising prices rate, economic growing lessenings by 0.66 per centum. From the 2nd theoretical account, as the rate of rising prices rises by 1 per centum, economic growing lessenings by 0.38 per centum.

Hence rising prices does non deflate economic growing by much in Mauritius. Fisher ( 1993 ) found negative associations between rising prices and growing in pooled cross-section, clip series arrested developments for a big set of states.

HC-Human Capital

The estimation of the human capital variable, proxied by secondary school and third school registration ratio, bears a positive mark ( 0.4283772 ) and is important. This figure shows that as HC additions by one per centum, economic growing additions by 42.83 per centum

Furthermore, the 2nd theoretical account shows that for every per centum addition in HC, economic growing rises by 18.57 per centum. This confirms the anticipations of the endogenous growing theory on the importance of human capital for economic growing. The inadequacy of skilled work force could decelerate down the development of the fiscal sector. Following this, the Mauritanian authorities has, justly so, opened up the island to foreign endowment. Furthermore, third establishments are supplying a scope of classs to run into the necessities of the fiscal sector in footings of labor demands. Harmonizing to ( Blin, 2004 ) , Mauritius has a big human capital base which is straight related with economic growing.

Cointegration

Economically talking, two variables are cointegrated if they have a long-run, or equilibrium relationship between them. Granger ( 1986 ) notes that a trial for cointegration can be thought as a pre-test to avoid specious arrested development. For the variables ( both dependant and independent ) to be cointegrated, we require that the mistake term, be I ( 0 ) . Testing for cointegration implies verifying if the remainders in the arrested development theoretical account are non-stationary or stationary. We can utilize the DF / ADF trial on the error term.

Engle and Granger ( 1987 ) have tabulated a new set of critical values and therefore the trial is known as the Engle Granger trial.

H0: remainders from cointegrating arrested development are non stationary

H1: remainders from cointegrating arrested development are stationary

Following this, an ADF trial was carried on the error term and the consequence is as follows:

Table 10: Stationarity of Remainders 1

Dickey-Fuller trial for unit root Number of obs = 32

Interpolated Dickey-Fuller

Test 1 % Critical 5 % Critical 10 % Critical

Statistic Value Value Value

Z ( T ) -6.368 -3.702 -2.980 -2.622

MacKinnon approximative p-value for Z ( T ) = 0.0000

Beginning: Computed

Table 11: Stationarity of Remainders 2

Dickey-Fuller trial for unit root Number of obs = 32

Interpolated Dickey-Fuller

Test 1 % Critical 5 % Critical 10 % Critical

Statistic Value Value Value

— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

Z ( T ) -8.000 -3.702 -2.980 -2.622

— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

MacKinnon approximative p-value for Z ( T ) = 0.0000

Beginning: Computed

Hence, since the p-values are less than 0.5 and the t ratios are important, so it can be said that the dependant and independent variable can be cointegrated and the arrested development is non specious.

5.6 Error Correction Model

After holding tested for the error term to be I ( 0 ) , one can continue to the mistake rectification theoretical account. When the alteration of one of the series is explained in footings of the slowdown of the difference between the series, perchance after scaling, and slowdowns of the differences of each series, it is said to be an mistake rectification theoretical account. The mistake rectification term ( R ) indicates the velocity of accommodation to reconstruct equilibrium in the dynamic theoretical account. The Engle Granger theorem provinces that after holding carried out the above trials, use the residuary ( error term ) as one variable in the mistake rectification theoretical account, i.e. make the following short run equation.

The consequences are as follows:

Table 12: Electronic countermeasures Consequences

Coef. Std. Err. t-statistic P-value

Model 1

rlag -0.541818 0.0767447 -7.06 0.000

Model 2

rlag -0.860033 0.1654613 -5.19 0.000

Beginning: Computed

ECM equation for theoretical account 1:

DGDPCAP= 0.2314352 DFL+ 0.2282853 DGFCF+ 0.3286344 DHC – 0.0028116 DINF – 0.541818 rlag

ECM equation for theoretical account 2:

DGDPCAP= 0.02233 DFLI+ 0.2286837 D GFCF+ 0.164454 DHC-0.0011113 DINF-0.860033 rlag

Here the electronic countermeasures is represented by the rlag. The t-ratio prevarications outside the critical value and the p-value is less than 0.5 and the coefficient of R is negative. This implies that in the following period economic growing will get down falling to rectify the equilibrium mistake. The size of the coefficient of the mistake rectification term ( -0.5418 ) suggests a comparatively high velocity of accommodation from the short tally divergence to the long tally equilibrium. More exactly, it indicates that around 54 per cent of the divergence from long tally growing is corrected every twelvemonth. It is to be noted that all the other variables in both the theoretical accounts following ECM are important.