THE ANALYSIS OF LIBYA USING SOLOW GROWTH MODEL AND FDI. A CASE OF STUDY: LIBYA Participants 1.Haji, Yusuph 2.Hamisi, Medison 3.Mhenga, Denis.G Supervised by Michael.R.Baha
THE ANALYSIS OF LIBYA USING SOLOW
GROWTH MODEL
AND FDI.
A CASE OF STUDY: LIBYA
Participants
1.Haji, Yusuph
2.Hamisi, Medison
3.Mhenga, Denis.G
Supervised by Michael.R.Baha
Introduction
Solow growth model is a model of economic growth originally developed by MIT’s Robert
Solow in the 1950s. Solow’s purpose in developing the model was to deliberately ignore some
important aspects of macroeconomics, such as short-run fluctuations in employment and savings
rates, in order to develop a model that attempted to describe the long-run evolution of the
economy. The Solow growth model shows how saving, population growth and the technological
progress affect the level of a country’s economy output and its growth at a particular time.
(Mankiw, 2001)
From the framework of Mankiw, the Solow model is presented as here under;
),( LtKtFYt
And under the Cobb-Douglas function with constant returns to scale can be written as;
1)(ALKY , 10
Where Y is output, K is capital, L is labor, A is labor-augmenting technological progress, and
is the share of capital in total output.
A. Empirical Model
Foreign direct investment (FDI) is frequently seen as an important catalyst for economic growth
in the developing countries. It affects the economic growth by stimulating domestic investment,
increasing human capital formation and by facilitating the technology transfer in the host
countries. (Falki, 2009). The main purpose of the study is to investigate the impact of FDI on
economic growth in Libya, for the period 1980-2011. The role of the foreign direct investment
(FDI) has been widely recognized as a growth-enhancing factor in the developing countries
(Khan, 2007).
According to Levin and Raut (1997) and Zhang (2003), FDI can be applied into growth model in
two ways depending on different assumptions. FDI can be postulated to cause growth directly or
indirectly through the spillover effects. First, we assume that FDI would directly cause growth,
and then the capital stock in Solow production function is assumed to consist of two components.
I.e. domestic and foreign owned capital stock ftdtt KKK . at this point Equation 1 is
obtained:
321 b
fit
b
dit
b
ititit KKLAY
where Y is denoted as output, dtK and ftK as the domestic and foreign owned capital stocks,
Lit as labor, A it as total factor productivity, which explains the output growth that is not
accounted by the growth in factors of production specified. The subscript i = 1.N indicates
sample country i to N.Subscript t = 1, T represents time period t, starting from 1 to T. After
taking logarithm to Equation 1, the production function is as follows:
ftditititit KbKbLbALogY logloglog 321
The theoretical model that is used to investigate the interaction of additional variable which is
FDI and economic growth based on the following production function
1FDItLtAtKtYt
So to test the hypothesis empirically the effect of FDI on economic growth, the model used can
be specified as follows;
tttt FDIbLbKbbYt 3210
For the purpose of estimation the above equation to be tested was obtained by taking the Log on
both sides of the equation, the equation that could be written as here under;
ttttt FDIbLbKbbLogY logloglog 3210
Where the variable on the left side is dependent variable and the variables on the right side are
exogenous variables.
B. Preliminary Data Analysis
The statistical method that is concerned with the presentation of numerical data, in form of
tables, graphs and charts to simplify data analysis is known as descriptive statistics. Thus,
descriptive statistics is useful in summarizing a set of data clearly and easy for analysis. This
includes; Mean Median, Standard Deviation, Skewness and Kurtosis.
TABLE 1: The descriptive statistics for the variables used in the model
N Range Mean Std.
Deviation
Skewness Kurtosis
FDI 32 16022.1 2766.93 4810.414 2.178718 3.44038
LABOR 32 1641.78 1534.35 543.6186 0.170103 -1.389
GDP 32 73.0302 -0.5543 12.8701 -3.35871 14.5353
CAPITAL 31 8326.85 5412.14 1941.432 1.230933 1.44342
Valid N (listwise) 31
SOURCE: UNCTAD DATA, STATA (2013)
FIGURE 1: GDP GROWTH RATE FOR THE YEAR (1980-2010) IN LIBYA
-70
-60
-50
-40
-30
-20
-10
0
10
20
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Total
Figure 1 shows that the GDP growth rate in Libya keeps on changing over time and that it
experience negative and positive GPD growth. From the graph in the year
1980, 1981,1983,1984,1986 and other periods showed from the graph and this is due negative
effect of labor and capital. In 2011 Libya experienced political revolution which led to lower the
GDP.
FIGURE 2: LABOR FORCE OF ALL SECTORS FOR LIBYA (1980-2011)
Time (years)
Figure 4 shows that labor force participation to promote growth in Libya, has an upward trend
from year to year from 1980 and 2011.
0
500
1000
1500
2000
2500
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Total
FIGURE 3: GROSS CAPITAL STOCK FOR LIBYA (1980-2011)
Time (years)
The trend of gross capital formation in Libya was constant over time , year to year from 1980 to
2011as shown in the figure 3 above.
FIGURE4: FDI IN LIBYA FOR THE YEAR (1980-2011)
Time (years)
0
0.2
0.4
0.6
0.8
1
1.21980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Total
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Total
Figure 5 shows that the trend of FDI inflow in Libya was in a horizontal movement year to year
from 1980 to 2004, but there was a very rapid increasing trend of FDI inflows to Libya from
2005 towards 2011
C. REGRESSION ANALYSIS
loggdp Coef. Std. Err. t P>t [95% Conf. Interval]
loglabor -3.77456 1.056859 -3.57 0.003 -6.015 -1.53412
logcapit -4.17455 1.172956 -3.56 0.003 -6.6611 -1.68799
logfdi 1.271049 0.330741 3.84 0.001 0.569911 1.972188
_cons 24.19987 6.394651 3.78 0.002 10.64381 37.75592
SOURCE: UNCTAD DATA, STATA (2013)
tttt FdiLabCapY 271.1775.3175.420.24
SE (6.395) (1.173) (1.057) (0.331)
The analysis was made through Analysis of Variance (ANOVA) to see the influence of those
factors affecting the economic growth of Libya say GDP. The study revealed that those factors
has significance influence on economic growth since the F calculated (Fc) is greater than critical
value at 5% which is 3.24 that means those factors are clearly significance difference and that
95% (C.I) probability that the conclusion is correct or 5% probability that conclusion is wrong.
ANOVA TABLE
Source SS df MS
Number of
obs = 20
F( 3, 16) = 5.12
Model 1.628005 3 0.542668183 Prob > F = 0.0113
Residual 1.694655 16 0.105915957 R-squared = 0.49
Adj R-
squared = 0.3943
Total 3.32266 19 0.174876834 Root MSE = 0.32545
a: Predictors:(constant);LogCAPITAL,LogLABOR,LogFDI.
b: Dependent variable: LogGDP
D. The Main Problems in Conducting Regression Analysis
Multicollinearity
The term multicollinearity originally it meant the existence of a “perfect”, or exact, linear
relationship among some or all explanatory variables of a regression model. (Gujarat, 2004).
Multicollinearity is a nature of a sample problem and state of nature that results in relatively
large standard errors for the estimated regression coefficient, but not biased. (Thomas, 2007)
Measuring the degree of multicollinearity.
There are three ways to measure the degree of multicollinearity in the regression model; these
are such as the use of a correlation matrix, the Variance Inflation Factor (VIF), and the Tolerance
measure. in our case multicollinearity was measured by using VIF with the help of STATA and
it was found to be less than ten. As the rule of thumb, if VIF is greater than 10 then there is no
exact linear relationship among the independent variables.
Variable VIF 1/VIF
Loglabor 4.65 0.214845
Logfdi 4.17 0.240086
Logcapit 3.78 0.264422
Mean VIF 4.2
AUTOCORRELATION
Autocorrelation or serial correlation often appears when working with time series data. On
should understand that in order for autocorrelation to appear it is necessary that observations are
correlated over a sequential order. In statistics terms it could be expressed as here under;
0, ji UUCov j
Hence autocorrelation is a problem that frequently appears when working with data that has time
dimension. This means that it is meaningless to look for autocorrelation when working with
cross sectional data which usually are based on random samples from a population, at a given
point in time.
Autocorrelation can easily be detected by using graphical method and Durbin–Watson Test.
Graphical Method involves the plotting of the residuals against time, the time sequence plot.
Durbin Watson d Test is the most celebrated test for autocorrelation by many statisticians. It is
defined as
T
T
N
T
U
UUd
T
1
2
2
2
1
d Test is the ratio of the total sum of squared differences in residuals over the residual sum of
squares. d Statistic is very useful as it is concerned with estimating residuals that has been
computed in regression analysis.
It is possible to see that the DW test statistic only takes values between 0 and 4 since
autocorrelation coefficient only takes values between -1 and 1. Hence when the autocorrelation
coefficient equal 0, the DW test statistic equal 2. If DW>2 we have an indication of a negative
autocorrelation and if DW<2 we have an indication of a positive autocorrelation
Unfortunately there exist no simple distribution function for this test function since it depends on
the number of observations used as well as the values of the explanatory variables used in the
regression. For that reason it is not possible to establish a precise critical value for the DW test
statistic. However, Durbin and Watson made some simulations so that you, based on the number
of observations used, and the number of parameters included in the model, can find a lower value
(L) and an upper value (U) to compare the DW test value with.
Possible outcomes from the Durbin-Watson test for autocorrelation
Positive AC Inconclusive No AC Inconclusive Negative AC
0<DW<L L<DW<U U<DW<4-U 4-U<DW<4-L 4-L<DW<4
Table above show five different regions where the DW-test value potentially could end up, if you
receive a test value that is located in the interval between the lower value (L) and the upper value
(U) your test is inconclusive and you have no use of the DW-test. However, if the DW-value is
between 0 and the lower value (L) you can draw the conclusion of having positive
autocorrelation. In case of negative autocorrelation you have to form the upper and lower value
for yourself using L and U as is done in table above.
Detection of autocorrelation
Under this the Durbin Weston test was used to see whether there is autocorrelation on those time
series data or not.
The result from the STATA we received that a DW-test value is equal to 2.242.from the
statistical table the lower value(L) was 1.160 and the upper value was 1.735.
UDWU 4
Where DW is 2.242.then the value substituted on the above equation which represent absent of
autocorrelation
735.14242.2735.1
265.2242.2735.1 , which means the model doesn’t suffer from the problem of
autocorrelation.
Heteroscedasticity and Diagnosis.
The classical assumption require for the ordinary least square estimator to be efficient states that
the variance of the error term has to be constant and same for all the observation, this is referred
to homoscedasticity error term. When this assumption is violated and the variance is different
from different observations then it is referred as heteroscedasticity. This assumption is important
and cannot be ignored since for the ordinary least square to be efficient then this assumption
must hold, otherwise the estimator will be inefficient and therefore we cannot claim our
estimator is the best among unbiased estimator.
Heteroscedasticity can be detected in two ways, by graphical method and statistical test. The
Graphical method is where the dependent is plotted against its explanatory variables, the
graphical method are useful but sometimes is difficult to say whether heteroscedasticity is
present and found harmful, it is therefore necessary to use statistical test.
Statistical test involve three tests which are Gold-feld Quant test, Breusch pagan test and white’s
test.
Ways to detect heteroskedasiticty problem
Graphical Method
A natural starting point in detecting possible deviation from homoskedasticity is to plot he data,
since we are interested on behavior of error term and its variation
Goldfeld-Quandt Test.
This work under the assumption that the error variance is equal for all observation, which is to
say the error term is homoskedasicity when this is true the variance of one part if the sample
must be the same as the variance of another part of other sample if this is not the case we
conclude that there is heteroskedasticty problem
Breusch–Pagan–Godfrey Test.
It is more general than GQ since it allow more than variable to be tested It is The success of the
Goldfeld–QuandttesIt depends not only on the value of c (the number of central observations to
be omitted) but also on identifying the correct X variable with which to order the observations.
The study used the Breusch-Pagan-Godfrey Test to determine whether the model is suffering
from the problem of heteroscedasticty and the below results obtained;
The Breusch-Pagan-Godfrey Test Results
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of loggdp
chi2(1) = 0.85
Prob > chi2 = 0.3569
SOURCE: UNCTAD DATA, STATA (2013)
The results above shows that a large chi-square led to rejection of the null hypothesis
(homoskedasticity null), and not reject the alternative hypothesis. So the result indicates the
presence of heterosckedasticty in the model.
E. Nonstationarity
Nonstationarity refers to values of mean, variance and covariance of time series data are
changing over time. The typical case is an example of the random walk model where
econometricians do several analyses and came to conclude that; random walk model is no
stationary because as t increases, its variance increases.
Testing for Stationary
There are several tests of stationary in time series data, for the case of space, two tests will be
briefly discussed:
Graphical analysis and
Autocorrelation Function (ACF) and Correlogram
Graphical Analysis
For better analysis of any time series data, it is advisable firstly to present data graphically in
order to determine the likely nature of the data. These visual plots provide impression that all
data used in regression process trends upward with time suggesting that the mean and variance of
the variables are changing and hence, a non-stationary in character.
Autocorrelation Function (ACF) and Correlogram
This is the simple test of stationary in time series data and it is defined as below
iance
klagatarianceACF
var
cov
F. Correction of problems identified in D
Correction for Heteroscedatsicity by using robust standard error
Number of
observation 20
F( 3, 16) 4.28
Prob > F 0.0212
R-squared 0.49
Root MSE 0.32545
Robust
loggdp Coef. Std. Err. t P>t [95% Conf. Interval]
loglabor -3.77456 1.184802 -3.19 0.006 -6.28623 -1.26289
logcapit -4.17455 1.168147 -3.57 0.003 -6.65091 -1.69819
logfdi 1.271049 0.380017 3.34 0.004 0.46545 2.076648
_cons 24.19987 6.803688 3.56 0.003 9.776694 38.62304
SOURCE: UNCTAD DATA, STATA (2013)
The above is the new model corrected from the problem of heterosckedasticty using the robust
standard error. Here under is the new model resulted from the above model
tttt FdiLabCapY 271.1775.3175.420.24
SE (6.804) (1.168) (1.185) (0.380)
REFERENCE
Gujarati, D.N. (2007) “Basic Econometrics”, Tata McGraw Hill Education Private Limited, New
Khan Arshad (2007) “Foreign Direct Investment and Economic Growth; The role of Domestic Financial
Sector”, PIDE working paper.
Levin, A. and L.K. Raut (1997) “Complementary between Export and Human capital in Economic
Growth; Evidence from the semi-industrialized countries” Economic Development and Cultural
Change.
Nuzhat Falki (2009) “International Review of Business Research Paper”, vol.5 No. 5 September
2009.Pp.110-120
Sadia Shaikh (2012), “Impact of FDI, Capital Formation and International Trade on Economic
Growth of Pakistan”: An Empirical Analysis
Thomas Andren.(2007), “econometrics” Ventus publishers
UNCTAD (2002) Investment Report, United Nations, New York
Zhang, K. H. (2003) “Foreign Direct Investment in China”; Asian Economic and Political Issues.
Index1: Libya data of Output (GDP), FDI, Capital and Labor for all sectors
YEAR FDI STOCK LABOR CAPITAL GDP FDI FLOW
1980 1855.425 739.036 9279.96502
-
0.284523744 -1089.34
1981 1111.295 783.912 11194.7776 -19.1901808 -744.13
1982 719.475 828.015 8641.13444 2.806056774 -391.82
1983 392.845 864.052 7610.95782
-
2.472227282 -326.63
1984 375.955 900.471 7478.80056
-
5.041781428 -16.89
1985 495.195 938.975 5769.59877 8.326778426 119.24
1986 318.275 982.409 5131.34535
-
11.35250631 -176.92
1987 311.862 1023.6 3897.0133
-
14.70205316 -97.63
1988 393.962 1059.467 4772.12023 7.579023356 97.98
1989 519.132 1103.917 4613.49386 7.199417141 125.17
1990 678.043 1165.705 5787.33248 3.720131634 158.911
1991 769.947 1212.919 4511.78927 13.34321975 91.9038
1992 869.04 1263.211 4431.33432
-
1.217100258 99.0931
1993 927.193 1314.536 5367.75704 3.172647467 58.1529
1994 854.569 1369.475 5026.66325 0.507538436 -72.6243
1995 766.082 1435.694 3439.82378
-
2.220406953 -88.4861
1996 654.36 1498.863 4848.71231 2.134123184 -111.722
1997 586.46 1570.788 4101.62246 5.212671027 -67.9
1998 458.566 1648.291 3575.5142
-
3.552244711 -127.894
1999 330.497 1726.342 3763.42296 0.749659243 -128.069
2000 471.497 1801.142 5015.41419 2.338436921 141
2001 338.497 1874.88 4205.47894 0.511727078 -133
2002 483.497 1943.685 3153.6324
-
1.250254005 145
2003 626.497 2006.347 2867.92879 12.96465682 143
2004 983.497 2068.739 3333.99358 4.435780624 357
2005 2021.497 2129.007 4063.72512 10.2900164 1038
2006 4085.497 2187.509 5486.79501 6.715576478 2064
2007 7935 2246.753 6040.58535 5.086807211 3850
2008 11115 2303.012 7742.10861 2.721379199 3180
2009 14425 2349.633 5702.86375
-
0.745390583 3310