EXAMINING THE RELATIONSHIP BETWEEN GDP AND NATIONAL INCOME MODEL COMPONENTS OF 185 COUNTRIES IN 2013

19B1008 FARIDA HAMIMI BINTI MOHD SALLEH

Assumption of National Income Model: 2 sector model help to explain the national income model and the changes Increase & Decrease in component C, I, G, X,M will change the national income (GDP). Similarly, changes in GDP will cause a change in national income (GDP).

Regression 1

Coefficient Standard Error t-ratio constant 4.71e+08 1.49e+09 0.32 Government -.1634183 .0336677 -4.85 Investment 1.256981 .011899 105.64 Consumption .9803206 .0097932 100.10 Import -1.150317 .0505533 -22.75 Export 1.134428 .0431802 26.27

n=185 R-squared= 1.000 F-statistic=99999.0 Prob> F = 0.0000

Coefficient Standard Error t-ratio constant 4.71e+08 1.49e+09 0.32 Government -.1634183 .0336677 -4.85 Investment 1.256981 .011899 105.64 Consumption .9803206 .0097932 100.10 Import -1.150317 .0505533 -22.75 Export 1.134428 .0431802 26.27

A $1 increase in government spending leads to a $0.16 decrease in GDP (constant USD2010) A $1 increase in Investment leads to a $1. 26 increase in GDP (constant USD2010) A $1 increase in Consumption leads to a $0.98 increase in GDP (constant USD2010) A $1 increase in Import leads to a $1.15 decrease in GDP (constant USD2010) A $1 increase in Export leads to a $1.13 increase in GDP (constant USD2010)

_cons = P>t = 0.753 The P>f result shows the regression is 0.000 which is significant. However, looking at the _cons, the p>t is greater than 0.05 which is insignificant. Therefore, there is an error in the regression.

Significant or Not ?

[Audio] In order to identify heteroskedasticity we will perform the GOLDFELD QUANT test or GQ test. This test is used in regression analysis to test for homoscedasticity. First, the data should be in ascending order, then we will find the new no of observations and d.o.f. By regressing gdp to 5 components in order to find the RSS (residual sum of squares) of the lower bracket and upper bracket of the data, higher residual value means there is high in variance/ varialibility. If the variance increasing thus heteroskedasticity exist.

RSS Sample 1 (Lower Bracket) N= 70 countries d.o.f= 70-6= 64 By regressing the GDP into GOV CONS INV IM EX RSS= 1.91E+19

Heteroskedastic Test (Goldfeld Quandt Test/GQ Test)

RSS Sample 2 (Upper Bracket) N= 70 countries d.o.f= 70-6= 64 By regressing the GDP into GOV CONS INV IM EX RSS= 6.53E+22

[Audio] There fore it can also be said that we reject the null hypothesis. This means heteroscedasticity is present in this case.

RSS= 1.91E+19 RSS= 6.53E+22 (lower bracket) (upper bracket) ‘’ The variance increasing. Thus, heteroskedasticity exist‘’

Regression 2 Treat Heteroskedasticity Using Logged

Coefficient Standard Error t-ratio constant .6607206 .0922733 7.16 InGovernment .0005441 .0180529 0.03 InInvestment .2559683 .0187215 13.67 InConsumption .7264814 .0210383 34.53 InImport -.3035008 .0260145 -11.67 InExport .3190214 .01768 18.04

n=185 R-squared= 0.9989 F-statistic= 33578.19 Prob> F = 0.0000

[Audio] Such relationships where both X and y are log transformed, are commonly referred to as elastic in econometrics, and the coefficient of logx is referred to as an elasticity,

Coefficient Standard Error t-ratio constant .6607206 .0922733 7.16 InGovernment .0005441 .0180529 0.03 InInvestment .2559683 .0187215 13.67 InConsumption .7264814 .0210383 34.53 InImport -.3035008 .0260145 -11.67 InExport .3190214 .01768 18.04

A 1% increase in government spending leads to a little or no changes in GDP (coefficient of govt spending) A 1% increase in Investment leads to a increase 25.6% in GDP (coefficient of investment) A 1% increase in Consumption leads to a 72.6% increase in GDP (coefficient of consumption) A 1% increase in Import leads to a 30.4% decrease in GDP (coefficient of import) A 1% increase in Export leads to a 31.9% increase in GDP (coefficient of export)

InGov = P>t = 0.976 The P>f result shows the regression is 0.000 which is significant. However, by looking at the InGov, the p>t is greater than 0.05 which is insignificant.

Significant or Not ?

RSS Sample 1 (Lower Bracket) N= 70 countries d.o.f= 70-6= 64 By regressing the Logged GDP into to Logged GOV,CONS, INV, IM & EX Logged RSS=0.658722233

Treat Heteroskedasticity Using Logged

RSS Sample 2 (Upper Bracket) N= 70 countries d.o.f= 70-6= 64 By regressing the Logged GDP into Logged GOV,CONS, INV, IM & EX Logged RSS= 0.098200743

[Audio] The variance show a decrease which indicates an improvement, Hence, it can be said that the heteroshedasticity is cured.

Logged RSS= 0.658722233 Logged RSS= 0.098200743 (lower bracket) (upper bracket) ‘’ The variance decreasing. Thus, the heteroscedasticity is cured‘’

In summary, The first regression output shows heteroskedasticity after the GQ test. To fix heteroskedasticity, use logged regression. The second regression improve the variances during GQ test. Hence, the heteroskedasticity is cured.