Empirical Results and Analyses of China

The table 4-1 shows the primitive regression model of China. It shows that unemployment rate (UR) is under 10% significance and consumer price index (CPI), industrial production (IP) and interest rate (IR) are under 1% level. In this primitive model, the dummy variable, the Olympics are not significant. However, this model is not revised through the modification process. There might be some

multicollinearity, heteroskedasticity or autocorrelation problems which do not fit the requirement of BLUE. Therefore, this study is required to make a number of modifications.

4.1.1 Primitive Regression Model of China Table 4-1 Primitive Regression Model of China

Dependent Variable: Number of Observation: 39 Shanghai Stock Exchange Composite Index (SSE) Frequency: monthly Regression Model:  !!"_! = ! + !_! !"_{! !}+ !_! !"#_{! !}+ !_! !"_{! !}+ !_! !"_{! !}

+!_! !"_{! !}+ !_! !"_{! !}+ !_!(!_!)_!+ !_!

Symbol Variable Coefficient Std. Error t-statistics Probability

! Constant term 21491.11 12543.00 1.713394 0.0966*

!"_! Unemployment

Rate -4130.299 2335.770 -1.768282 0.0869*

!"#_! Consumer Price

Index 405.4441 91.82625 4.415340 0.0001***

!"_! Industrial

Production 141.8304 47.55777 2.982277 0.0055***

!"_! Exchange Rate -57.42305 951.5984 -0.060344 0.9523

-617.1159 392.8636 -1.570815 0.1264

R-square 0.867925

Adjusted R-square 0.838101

F-statistics 29.10206

Durbin-Watson Statistics 1.328252

Note: Significant level 10%, 5% and 1% are marked by *, **and***

Source: This Study

4.1.2 Multicollinearity Test

We recall the Klein’s rule, whereas if R^!_! of independent variable is larger than !_!^! of dependent variable, there is multicollinearity. The test results are shown in Table 4-2, indicating the multicollinearity problems in consumer price index (CPI), exchange rate (ER) and money supply (MS).

Then, this study takes the significance and multicollinaerity into consideration and eliminates exchange rate (ER) and money supply (MS). The research sustains the consumer price index (CPI) because it could lose substantial amount of explanatory power after canceling consumer price index (CPI).

Table 4-2 Multicollinearity Test of China

Dependent Variable: Shanghai Stock Exchange Composite Index (SSE)

!_!^!: 0.867925

Symbol Variable !_!^! Multicollinearity

!"_! Unemployment Rate 0.793995 No

!"#_! Consumer Price Index 0.884052 Yes

!"_! Industrial Production 0.738861 No

!"_! Exchange Rate 0.961229 Yes

!"_! Interest Rate 0.734132 No

!"_! Money Supply 0.942091 Yes

Source: This Study

4.1.3 Revised Regression Model of China

After eliminating exchange rate (ER) and money supply (MS), this study uses the remaining five independent variables to establish the revised model, followed by conducting the heteroskedasticity and autocorrelation test to produce the final model.

Table 4-3 Revised Regression Model of China

Dependent Variable: Number of Observation: 39 Shanghai Stock Exchange Composite Index (SSE) Frequency: monthly Regression Model:

 !!"_! = ! + !_! !"_{! !}+ !_! !"#_{! !}+ !_! !"_{! !}+ !_! !"_{! !}+ !_!(!_!)_! + !_!

Symbol Variable Coefficient Std. Error t-statistics Probability

! Constant term 22798.54 9957.678 2.289544 0.0286**

!"_! Unemployment

Rate -3725.073 2482.238 -1.500691 0.1429

!"#_! Consumer Price Index 483.7010 88.12349 5.488900 0.0000***

!"_! Industrial Production 63.53341 34.96430 1.817094 0.0783*

!"_! Interest Rate -2131.158 413.8172 -5.150000 0.0000***

!_! =1, 2008/5~2008/11

=0, otherwise -358.8713 374.6051 -0.957999 0.3450

R-square 0.839321

Adjusted R-square 0.814975

F-statistics 34.47559

Durbin-Watson Statistics 0.945485

Note: Significant level 10%, 5% and 1% are marked by *, **and***

Source: This Study

4.1.4 Heteroskedasticity Test

This research use White Heteroskedasticity Test with Eview 5.0 to diagnose whether if there is heteroskedasticity problem associated with the revised regression model. This study assumes a null hypothesis without heteroskedasticity in the 5%

significant level and the result is displayed in Table 4-4. The null hypothesis is therefore not rejected due to the likelihood of larger than 5%. That is, there is no heteroskedasticity problem in this revised model.

Table 4-4 Heteroskedasticity Test for Revised Regression Model of China White Heteroskedasticity Test

F-statistics 3.107224 Probability 0.111700 Obs*R-square 36.35782 Probability 0.233179 Source: This Study

4.1.5 Autocorrelation Test

This study uses the Durbin-Watson Test to diagnose autocorrelation problem in the revised regression model. Note the criterion of without autocorrelation is

!_! < ! < 4 − !_!. In this case, !_! = 1.218  and 4 − !_! = 2.782. Table 4-5 shows that the D-W statistics in the revised regression model is 0.945485 and there is autocorrelation in this model. To solve this problem, this study adds a correction term, AR(1). After adding AR(1), the D-W statistics is 1.989166 and within the interval. Therefore, there is no long autocorrelation.

Table 4-5 Autocorrelation Test for Revised Regression Model of China

Criterion: 1.218 < ! < 2.782 Durbin-Watson Statistics Autocorrelation

Revised Regression Model 0.945485 Yes

Add AR(1) 1.989166 No

Source: This Study

4.1.6 Final Regression Model of China Table 4-6 Final Regression Model of China

Dependent Variable: Number of Observation: 38 (adjusted) Shanghai Stock Exchange Composite Index (SSE) Frequency: monthly Regression Model:  !!"_! = ! + !_! !"_{! !}+ !_! !"#_{! !}+ !_! !"_{! !}

+!_! !"_{! !}+ !_!(!_!)_! + !_!!"(1) + !_!

Symbol Variable Coefficient Std. Error t-statistics Probability

! Constant term 7416.442 6993.163 1.060528 0.2971

!"_! Unemployment

Rate -705.0147 1710.370 -0.412200 0.6830

!"#_! Consumer Price

Index 181.0842 90.01192 2.011780 0.0530*

!"_! Industrial

Production 55.47580 31.40433 1.766502 0.0872*

!"_! Interest Rate -794.6149 323.6378 -2.455260 0.0199**

!_!

=1,

2008/5~2008/11

=0, otherwise

-80.80179 289.8978 -0.278725 0.7823

AR(1) Correction Term 0.891030 0.079783 11.16811 0.0000***

R-square 0.928117

Adjusted R-square 0.914204

F-statistics 66.70890

Durbin-Watson Statistics 1.989166

Note: Significant level 10%, 5% and 1% are marked by *, **and***

Source: This Study

After the modification process, this study eliminated the independent variables, exchange rate (ER) and money supply (MS). This research also conducted heteroskedasticity test and autocorrelation test with addition of a correction term, AR(1). The final regression model is shown in Table 4-6. It shows that R-square is 0.928117 and adjusted R-square is 0.914204, which means 93% and 91%

explanatory power of this final regression model to the stock market of China.

Then, this study will analyze the impacts of the independent variables of this model. This study determines whether one variable significant or not through the t test. The analyses are shown in the follow:

1. Unemployment Rate:

The coefficient of unemployment rate is -705.0147 and it is negatively related to SSE index. The coefficient meets our prior expectation, however the probability is 0.683 and is not significant.

2. Consumer Price Index:

The coefficient of consumer price index is 181.0842 and it is positively related to SSE index. The coefficient meets our prior expectation.

Furthermore, the probability of consumer price index is 0.053 and under the 10% significance level.

3. Industrial Production:

Table 4-6 shows the coefficient of industrial production is 55.4758. There is a positive relationship between industrial production and SSE index.

The coefficient meets our prior expectation. Besides, the probability is

0.0872 and under the 10% significance.

4. Interest Rate:

Table 4-6 shows that the coefficient of interest rate is -794.6149. There is a negative relationship between interest rate and SSE index, which meets our prior expectation. The probability is 0.0119 and under the 5%

significance.

5. The Olympics:

In previous discussion, this study supposes that the Olympics is negatively related to stock market. Table 4-6 shows the coefficient is -80.80179 and it is the same as our discussion. However, the probability is 0.7823 and it cannot reject the null hypothesis of the coefficient is zero. Hence, the Beijing Olympics 2008 is insignificant to SSE index when the dummy variable D_! is setting at !_! = 1 from 2008/5 to 2008/11.