Estimation of Natural Rubber Export Model in International Markets
38 that can be used to calculate estimates for the FEM. In this study, the FEM
estimators are calculated by the Least Square Dummy Variable LSDV approach. Furthermore, the output from the STATA software is processed to estimate the
model, as can be seen in Figure 15.
Figure 15 The Results of The Gravity Model Testing for the basic assumption should be done to ensure that the
obtained parameters are unbiased, consistent and efficient. A test of basic assumptions
includes checking
for heteroskedasticity,
normality and
multicollinearity Appendix 1. The normality test can be done by looking at the distribution of the residual data using the kernel density plot application on
STATA Appendix 1. Meanwhile, the presence or absence of multicollinearity can be checked by looking at the correlation between the independent variables in
the model. As shown in Appendix 1, the correlation coefficient of each independent variable is lower than the value of the coefficient of determination
R
2
0.7133, thus, it can be concluded that the model does not have multicollinearity problems. It is also reinforced by the R-squared value is high and
the number of significant variables. Heteroskedasticity occurs frequently in panel data due to the fact that data
on different aggregation levels are combined and different sections have different numbers of observations. The variance is also dependent on specific country
characteristics. In order to formally test for heteroskedasticity, the Breusch-Pagan test was employed. The null hypothesis is constant variance, i.e.,
homoskedasticity. The test indicated a p-value of 0.0000, which means that we can clearly reject the H
: heteroskedasticity is present. In order to solve this problem, the robust and cluster country options are used in STATA. When
applying the Breusch-Pagan test for the explanatory variables, the same outcome is reached p-value 0.0000 Appendix 1.
Based on the estimation model, as shown in Table 9, it can be determined that the Fstat probability value is below a 1 significance level 0.00 0.01. This
_cons -81.32534 15.56643 -5.22 0.000 -111.953 -50.69764 dj_bel -.4636549 .3319673 -1.40 0.163 -1.116816 .1895066
dj_ind -4.626895 1.504841 -3.07 0.002 -7.587742 -1.666048 dj_br -2.147857 1.147711 -1.87 0.062 -4.406033 .1103185
dj_can -1.824441 .8397827 -2.17 0.031 -3.476753 -.1721285 dj_ger -2.223786 1.323001 -1.68 0.094 -4.826853 .3792812
dj_kor -1.207747 .8324896 -1.45 0.148 -2.84571 .4302162 dj_sin 0 omitted
dj_chn -2.414935 1.861297 -1.30 0.195 -6.077127 1.247256 dj_jpn -2.926822 1.434761 -2.04 0.042 -5.749784 -.1038611
dj_us -3.69344 2.095365 -1.76 0.079 -7.81617 .4292897 di_tha -.3759587 .510211 -0.74 0.462 -1.379823 .6279058
di_may 0 omitted di_ina .8905043 .4334498 2.05 0.041 .0376711 1.743337
erijt -.1383721 .0201704 -6.86 0.000 -.1780583 -.0986859 rjt 1.215068 1.649992 0.74 0.462 -2.031371 4.461507
gdpjt 1.573838 .529565 2.97 0.003 .5318935 2.615782 prodit 1.813713 .4403327 4.12 0.000 .9473373 2.680088
yijt Coef. Std. Err. t P|t| [95 Conf. Interval] Robust
Root MSE = .86373 R-squared = 0.7133
Prob F = 0.0000 F 15, 314 = 69.01
Linear regression Number of obs = 330
39 indicates that the model is a good fit for this data and that there is at least one
significant variable in the model. The obtained R
2
value is 0.7133, which means that the model is able to explain 71.33 of the diversity of natural rubber exports,
while the remaining 28.67 is explained by other factors outside of the model.