Following the analysis of the time series model to study the variables that are non-stationary at the level and non-stationary at the first difference, and the OLS regression analysis to demonstrate the sensitivity of the dependent variables to the changes in the independent variables and finally controlling for series of co-integrating relations, the final stage of the analysis is the investigation of the causal relationships between the independent variables and the dependent variables representing the macro-economic performance of the Saudi economy. In order to investigate these causal relationships, the Granger causality test is employed for all models.
The probability values generated from the Granger Causality Tests are depicted in the Table (Table 6.14, 6.15, 6,16, 6.17 and 6.18). The reported F-statistics are a set of
standard tests for the joint hypothesis that the independent variables have causal relationships with the dependent variables.
6.5.1. Model 1: GDP
Starting the causality analysis by investigating the relationship between the independent variables, Market Capitalization Ratio (LNMCR), Number of Shares Traded (LNNST), Value of Shares Traded (LNVSTR), General Index (LNGI), Number of Transaction (LNNT) and Government Spending (GS) with the GDP, the results can be summarised as in Table 6.14:
Table 6.14: Granger Causality Test
Null Hypothesis F-Statistic Probability
LNMCR LNMCR does not Granger Cause LNGDP 2.27141 0.1305 LNGDP does not Granger Cause LNMCR 2.01736 0.1605 LNNST LNNST does not Granger Cause LNGDP 1.45685 0.2578 LNGDP does not Granger Cause LNNST 0.73656 0.4919 LNVSTR LNVSTR does not Granger Cause LNGDP 2.09423 0.1507 LNGDP does not Granger Cause LNVSTR 0.90721 0.4204
LNGI LNGI does not Granger Cause LNGDP 2.30773 0.1267
LNGDP does not Granger Cause LNGI 1.93336 0.1721
LNNT LNNT does not Granger Cause LNGDP 1.29224 0.2977
LNGDP does not Granger Cause LNNT 4.06149 0.0340
LNGS LNGS does not Granger Cause LNGDP 5.54287 0.0127
LNGDP does not Granger Cause LNGS 0.13363 0.8757 As the results demonstrates, the probability of accepting the null hypothesis for Number of Shares Traded (LNNST) not causing a change in GDP (LNGDP) is 25.78%, while the probability of rejecting it is 74.22%. The causality from the other direction suggests a similar relationship. In the similar manner, the probability that LNGDP does not cause a change in LNNST is 49.19%, whereas the probability of rejecting such a relationship is 50.81%. Both of these results suggest a causal relationship between GDP and MCR, albeit in a weaker form, particularly from GDP to the Number of Shares Traded (LNNST).
The test for Value of Shares Traded (LNVSTR) also provides evidence for a similar relationship. While the probability of accepting the null hypothesis is 15.07%, the probability for rejecting it is 84.93%. On the other hand, the probability of rejecting a causal relationship from LNGDP to LNVSTR is 42.04%, whereas that of accepting it is 57.96%. In other words, 57.96% of the times, LNGDP causes a change in LNVSTR.
Continuing the analysis for the fourth independent variable, General Index (LNGI), the probability of accepting the null hypothesis is 12.67% and the probability of rejecting is 87.33%. In the other direction, the probability of accepting the null hypothesis that GDP does not Granger cause GS is 17.21% and rejecting it is 82.79%, providing evidence for a causal relationship between these two variables.
In addition, checking the causal relationship between LNGDP and LNNT, the probabilities for accepting the null hypothesis are 29.77% and 3.4% respectively.
These results indicate that at 70.33% of the times, a change in the Number of Transactions (LNNT) causes a change in the GDP, and 96.6% of the times a change in the GDP results in a change in LNNT. The strength of the results indicates that with higher per capita income, Saudi people have a higher tendency to invest in the stock market.
Finally, the relationship between Government Spending (LNGS) and the GDP gives rather interesting results. First, the probability of accepting the null hypothesis that Government Spending does not cause GDP is only 1.27% and rejecting it is 98.73%, implying a strong influence of Government Spending on the GDP. More interestingly, the probability of accepting the null hypothesis that GDP does not cause Government Spending is 87.57%, suggesting that in 85.57 out of 100 times, the GDP does not create Government Spending. This one-way causal relationship suggests that the Saudi government uses Government Spending as a stimulator for the general economy and has a tendency to intervene by injecting liquidity into the economy when the macro-economic performance is not as strong. In addition, these results also suggest that this tendency is not for the long run and when the performance reaches the targeted levels, the spending becomes more controlled.
6-5-2- Model 2: NOGDP
When the analysis is further detailed by investigating the relationships between the Non-oil GDP as the dependent variable and the independent variables, the results in table 6.15. also indicate some interesting findings. First, the probability of accepting the null hypothesis for Market Capitalization Ratio (LNMCR) does not cause a change in Non-oil GDP (LNNOGDP) is 14.5%, while the probability of rejecting it is 85.5%. The causality in the other direction suggests a similar relationship.
Table 6.15: Granger Causality Test
LNGI LNGI does not Granger Cause LNNOGDP 2.55079 0.1044 LNNOGDP does not Granger Cause LNGI 0.94684 0.4055
LNNT LNNT does not Granger Cause
The probability that LNNOGDP does not cause a change in LNMCR is 33.55%, whereas the probability of rejecting such a relationship is 66.45%. Both these results suggest a causal relationship between NOGDP and MCR.
It should also be noted that the probability of accepting the null hypothesis for Number of Shares Traded (LNNST) does not cause a change in Non-oil GDP (LNNOGDP) is 67.07%, while the probability of rejecting it is 32.93%. On the other hand, the causality from the other direction suggests a different relationship. The probability that LNNOGDP does not cause a change in LNNST is 30.16%, whereas the probability of rejecting such a relationship is 69.84%. These results suggest that while NOGDP causes a change in Number of Shares Traded (LNNST), NST does not cause a change in NONGDP.
The test for Value of Shares Traded (LNVSTR) provides evidence of a non-causal relationship from both ends. While the probability of accepting the null hypothesis is 63.75%, the probability for rejecting it is 36.25%. On the other hand, the probability of rejecting a causal relationship from LNNOGDP to LNVSTR is 61.74%, whereas accepting is 38.26%. In other words, LNNOGDP causes a change in LNVSTR only 38.26% of the times.
Continuing the analysis for the fourth independent variable, General Index (LNGI), the probability of accepting the null hypothesis is 10.44% and the probability of rejecting is 89.56%. In the other direction, the probability of accepting the null hypothesis that NOGDP does not Granger because GS is 40.55% and rejecting it is 59.45%, providing evidence for a causal relationship between these two variables.
In addition, checking the causal relationship between LNNOGDP and LNNT, the probabilities for accepting the null hypothesis are 91.73% and 19.18% respectively.
These results indicate that only 8.27% of the times a change in the Number of Transactions (LNNT) causes a change in the NOGDP, and 81.82% of the times a change in the NOGDP results in a change in LNNST. These results suggest a one-way causal relation from NOGDP to NT, and indicate that the macro-economic performance is a determinant for the Number of transactions in the Saudi capital markets.
Finally, the relationship between the Government Spending (LNGS) and the NOGDP suggest the probability of accepting the null hypothesis that Government Spending does not cause GDP is 28.97% and rejecting it is 71.03%, This implies a strong influence of Government Spending on the NOGDP. More interestingly, the probability of accepting the null hypothesis that NOGDP does not cause Government Spending is 66.31%, suggesting that 66.31 out of 100 times, the NOGDP does not create Government Spending.
These results suggest a less clear causal relation between the macro-economic performance of the Saudi economy and the Saudi financial markets once oil revenues are taken out of the equation. Thus, it can be suggested that oil revenues are not the only determining factor of the dynamism of the Saudi economy, as the bourgeoning non-oil sector through economic diversification has contributed to the expansion of the economy. Such diversification can also be noticed in the non-oil economic activity of the government sector as well.
6-5-3- Model 3: NOPSGDP
In order to capture the real dynamics of the macro-economic performance of the Saudi economy and the Saudi financial markets without the influence of oil revenues, the analysis is further detailed by examining the Granger causality relationships between the Non-oil Sector GDP (LNNOPSGDP) and the independent variables. As can be seen in the results depicted in table 6.16, the probability of accepting the null hypothesis that Market Capitalization Ratio (LNMCR) does not cause a change in Non-oil Sector GDP (LNNOPSGDP) is 21.19%, while the probability of rejecting it is 79.81%. The causality in the other direction suggests a similar relationship. The probability that LNNOPSGDP does not cause a change in LNMCR is 31.58%, whereas the probability of rejecting such a relationship is 69.42%. Both these results suggest a causal relationship between NOPSGDP and MCR.
Similarly, the probability of accepting the null hypothesis that Number of Shares Traded (LNNST) does not cause a change in NOPSGDP (LNNOPSGDP) is 40.55%, while the probability of rejecting it is 59.45%. The causality from the other direction suggests a similar relationship. The probability that LNNOPSGDP does not cause a change in LNNST is 43.78%, whereas the probability of rejecting such a relationship
is 56.22%. Both these results suggest a causal relationship between NOPSGDP and NST.
The test for Value of Shares Traded (LNVSTR) also provides evidence for a different kind of relationship. While the probability of accepting the null hypothesis is 16.62%, the probability for rejecting it is 83.38%. On the other hand, the probability of rejecting a causal relationship from LNNOPSGDP to LNVSTR is 29.49%, whereas accepting it is 70.51%. In other words, only at 29.49% of the times does LNNOPSGDP cause a change in LNVSTR, indicating a somewhat one-way relationship from VSTR to NOPSGDP.
LNGI LNGI does not Granger Cause LNNOPSGDP 2.11585 0.1480 LNNOPSGDP does not Granger Cause LNGI 0.98395 0.3921 LNNT LNNT does not Granger Cause LNNOPSGDP 0.07936 0.9240 LNNOPSGDP does not Granger Cause LNNT 1.84064 0.1859 LNGS LNGS does not Granger Cause LNNOPSGDP 0.28822 0.2897 LNNOPSGDP does not Granger Cause LNGS 0.35754 0.7040
Continuing the analysis for the fourth independent variable, General Index (LNGI), the probability of accepting the null hypothesis is 14.80% and the probability of rejecting it is 85.20%. In the other direction, the probability of accepting the null hypothesis that NOPSGDP does not Granger because GS is 39.21% and rejecting it is 60.79%, providing evidence for a causal relationship between these two variables.
In addition, checking the causal relationship between LNNOPSGDP and LNNT, the probabilities for accepting the null hypothesis are 92.40% and 18.59% respectively.
These results indicate that at only 7.6% of the times does a change in the Number of Transactions (LNNT) cause a change in the GDP, and 81.41% of the times a change in the NOPSGDP results in a change in LNNT. These results indicate another one-way causal relationship with NOPSGDP.
Finally, the relationship between Government Spending (LNGS) and NOPSGDP gives rather interesting results. First, the probability of accepting the null hypothesis that Government Spending does not cause GDP is 28.97% and rejecting it is 71.03%, implying a strong influence of Government Spending on NOPSGDP. More interestingly, the probability of accepting the null hypothesis that Non-oil Public Spending GDP does not cause Government Spending is 70.4%, suggesting that out of 100 times, in 70.4 the GDP does not create Government Spending.
6.5.4. Model 4: GFC
When the analysis shifts to investigating the Gross Fixed Capital (GFC) of the Saudi economy, the Granger Causality Test also suggests some interesting results, as the results presented in table 6.17, with stronger emphasis on one-way relationships.
First, investigating the probability of accepting the null hypothesis of Market Capitalization Ratio (LNMCR) not causing a change in GFC (LNGFC) is only 3.05%, while the probability of rejecting it is 96.95%. The causality from the other direction suggests a relationship of totally opposite nature. The probability that LNGFC does not cause a change in LNMCR is 93.56%, whereas the probability of rejecting such a relationship is 6.44%. Both these results suggest a one-way causal relationship from MCR to GFC.
Second, the probability of accepting the null hypothesis of Number of Shares Traded (LNNST) not causing a change in GFC (LNGFC) is 2.39%, while the probability of
rejecting it is 97.61%. The causality from the other way around suggests a relationship of a totally opposite nature once again. The probability that LNGDP does not cause a change in LNNST is 68.59%, whereas the probability of rejecting such a relationship is 31.41%. Both these results suggest another one-way causal relationship from the Number of Shares Traded (LNNST) to Gross Fixed Capital (LNGFC).
Testing for Value of Shares Traded (LNVSTR) also provides evidence of a similar relationship, as the probability of accepting the null hypothesis is 1.97%, the probability of rejecting it is 98.03%. On the other hand, the probability of rejecting a causal relationship from LNGFC to LNVSTR is 80.77%, whereas accepting is 19.23%. In other words, 19.23% of the times, LNGFC causes a change in LNVSTR.
Table 6.17: Granger Causality Test
Null Hypothesis
F-Statistic
Probability
LNMCR LNMCR does not Granger Cause LNGFC 4.31607 0.0305 LNGFC does not Granger Cause LNMCR 0.06678 0.9356 LNNST LNNST does not Granger Cause LNGFC 4.69148 0.0239 LNGFC does not Granger Cause LNNST 0.38553 0.6859 LNVSTR LNVSTR does not Granger Cause LNGFC 4.99459 0.0197 LNGFC does not Granger Cause LNVSTR 0.21632 0.8077 LNGI LNGI does not Granger Cause LNGFC 4.59634 0.0254 LNGFC does not Granger Cause LNGI 0.00204 0.9980 LNNT LNNT does not Granger Cause LNGFC 5.90993 0.0113 LNGFC does not Granger Cause LNNT 1.45342 0.2614 LNGS LNGS does not Granger Cause LNGFC 0.27133 0.7656 LNGFC does not Granger Cause LNGS 1.75711 0.2025 Continuing the analysis for the fourth independent variable, General Index (LNGI), the probability of accepting the null hypothesis is 2.54% and the probability of rejecting is 97.46%. In the other direction, the probability of accepting the null hypothesis that GFC does not Granger cause GS is 99.80% and rejecting it is 1.2%, providing evidence of another one-way causal relationship.
In addition, checking the causal relationship between LNGFC and LNNT, the probabilities for accepting the null hypothesis are 1.13% and 26.14% respectively.
These results indicate that at 98.87% of the times, a change in the Number of Transactions (LNNT) causes a change in the GFC, and 73.86% of the times a change in the GFC results in a change in LNNT. The strength of the results indicates that with higher investments in fixed capital Saudi people has a higher tendency to invest in the stock market.
Finally, the relationship between Government Spending (LNGS) and GFC give rather interesting results. First, the probability of accepting the null hypothesis that Government Spending does not cause GFC is 76.56% and rejecting it is 23.44%, implying that Government Spending does not have an influence on GFC. More interestingly, the probability of accepting the null hypothesis that GFC does not cause Government Spending is 20.25%, suggesting that 79.75% of the times, Gross Fixed Capital creates Government Spending.
These results presented table 6.17 suggest an interesting trend in the causal relationship between the Gross Fixed Capital of Saudi Arabia and the independent variables representing the stock market. The relationships seem to be one-way from Gross Fixed Capital to the financial markets, suggesting that additions to the fixed capital encourage activity and strength in the financial markets. Another interesting result comes from the nature of the causal relationship between the GFC and the Government Spending. While the results above in Tables 6.14, 6.15 and 6.16 indicate that Government Spending is an important factor in defining the level of GDP, the relation goes in the opposite direction, when it comes to the GFC, and GFC itself becomes an important cause explaining the level of Government Spending itself.
These interesting results will further be investigated by excluding the influence of the oil revenues by examining the relationships with Non-oil Gross Fixed Capital. .
6.5.5. Model 5: NOGFC
The results presented Table 6.18 suggest similar types of causal relationships between the independent variables representing the Saudi financial markets and the Non-oil Gross Fixed Capital to the results presented above in Table 6.16. First, investigating the probability of accepting the null hypothesis for Market Capitalization Ratio
(LNMCR) is not causing a change in NOGFC (LNNOGFC) is 24.29%, while the probability of rejecting it is 75.71%. The causality from the other way around suggests a relationship of totally opposite nature. The probability that LNGFC does not cause a change in LNMCR is 86.16%, whereas; the probability of rejecting such a relationship is 13.84%. Both these results suggest a one-way causal relationship from MCR to NOGFC.
Second, the probability of accepting that the null hypothesis for Number of Shares Traded (LNNST) does not causing a change in NOGFC (LNNOGFC) is 5.13%, while the probability of rejecting it is 94.87%. The causality from the other direction suggests a relationship of the totally opposite nature once again. The probability that LNGDP does not cause a change in LNNST is 44.88%, whereas the probability of rejecting such a relationship is 55.22%. Both these results show a causal relationship between the Number of Shares Traded (LNNST) and Non-oil Gross Fixed Capital (LNNOGFC), although the causality is much stronger from the former to the latter.
Table 6.18: Granger Causality Test
LNGI LNGI does not Granger Cause LNNOGFC 1.45596 0.2608 LNNOGFC does not Granger Cause LNGI 0.01855 0.9816
LNNT LNNT does not Granger Cause 2.79931 0.0889
LNNOGFC
LNNOGFC does not Granger Cause LNNT
2.28832 0.1318
LNGS LNGS does not Granger Cause
LNNOGFC
0.72918 0.4968
LNNOGFC does not Granger Cause LNGS
0.92865 0.4142
Testing for Value of Shares Traded (LNVSTR) also provides evidence of a similar relationship. While the probability of accepting the null hypothesis is 11.51%, the probability for rejecting it is 88.49%. On the other hand, the probability of rejecting a causal relationship from LNNOGFC to LNVSTR is 80.57%, whereas accepting it is 19.43%. In other words, 19.43% of the times LNNOGFC causes a change in LNVSTR.
Continuing the analysis for the fourth independent variable, General Index (LNGI), the probability of accepting the null hypothesis is 26.08% and the probability of rejecting is 73.92%. In the other direction, the probability of accepting the null hypothesis that NOGFC does not Granger cause GS is 98.16% and rejecting it is 1.84%, providing evidence for another one-way causal relationship.
In addition, checking the causal relationship between LNNOGFC and LNNT, the probabilities for accepting the null hypotheses are 8.89% and 13.18% respectively.
These results indicate that at 91.11% of the times a change in the Number of Transactions (LNNT) causes a change in the Non-oil GFC, and 86.82% of the times a change in the Non-oil GFC results in a change in LNNT. This strength of the results indicates that with higher investments in fixed capital, Saudi people have higher tendencies to invest in the stock market, similar to the analysis for the GFC presented above in table 6.17.
Finally, the relationship between the Government Spending (LNGS) and the NOGFC gives rather interesting results. First, the probability of accepting the null hypothesis that Government Spending does not cause NOGFC is 49.68% and rejecting it is 50.32%, implying that Government Spending does have a weak causal influence on the NOGFC. More interestingly, the probability of accepting the null hypothesis that
NOGFC does not cause Government Spending is 41.42%, suggesting that 58.58% of the times Non-oil Gross Fixed Capital creates Government Spending.
Overall, these results are consistent with the results for the Gross Fixed Capital (GFC) presented above in table 6.16, which are also consistent with the results of the other, table 14, table 15 and table 16. Thus, the level and the strength of the relationships seem to decrease when the influence of the oil revenues is taken out of the equation. These results suggest an expected significant influence of oil revenues on the Saudi economy and Saudi financial markets.