Classic Assumption Test Data Processing

73 Based on the multicollinearity test that provided in the table 4.12, it is known that the value of each variable GDP, INF, UNEMP, IMP, CAB toward the other variables is less than 0.8. Thus, it can concluded that there is no multicollinearity contained on GDP growth, inflation, unemployment, import, and current account balance variable. 4 Autocorrelation Test The aim of autocorrelation test is to detect the internal correlation among the groups of a series observation arrange in a series of place and time. According to Ariefianto 2012:30 the commonly uses testing method to test the autocorrelation Durbin- Watson test DW tests. If the result shows that 2 DW 4 – d u it can be concluded that there is no autocorrelation. The result of Durbin-Watson test can be seen in the following table: Table 4.13 The Result of Autocorrelation Test using Durbin-Watson Statistic DW count d u table value 4 – d u dL table value 4 – dL 1.747141 1.8025 2.1975 1.4326 2.5674 Source: data processed 74 Table 4.13 shows the result of autocorrelation test using Durbin- Watson statistic. As the result in the table 4.13 it is known that autocorrelation problem is detected. The calculation shows that 2 1.747141 2.1975. Based on the result, thus it can be concluded that the data is conceive of autocorrelation problem. Nevertheless, according to Gujarati 2004:475, if one research is using GLS Generalized Least Square model, thus, the output does not have an autocorrelation problem. The regression model that have used in this research is using GLS model, thus it can be concluded that the autocorrelation problem is solved.

c. Model Selection in Panel Data Regression

As mentioned in the chapter 3, there are three models in panel data regression, they are: Pooled Least Square PLS, Fixed Effect Model FEM, and Random Effect Model REM. Before determine which one is the best model for the research, there are some of test that have to be done. The first is chow test. Chow test used to determine whether PLS or FEM model that will be used to processing the data. If the result of chow test indicates that the FEM is better than PLS, the next stage will be the hausman test. Hausman test used to determine whether FEM or REM that will be used to processing the data. 75 1 Chow Test The aim of chow test is to determine whether PLS or FEM model that will be used to processing the data. The requirements to reject H is, if the F chow is F table or if the probability of F chow is α 0.05. Table 4.14 The Result of Chow Test Cross-section F F Chow F Table Probability F Probability α Decision 2.688961 2.36 0.0059 0.05 FEM Source: processed data Based on the chow test result that have processed in the table 4.14, it is known that the value of Cross-section F is higher than F table that is 2.68 2.36 and the probability of F chow in Cross- section F is lower than α tht is 0.0059 0.05, thus the H PLS model is rejected. Thus, the tentative conclusion is FEM model is used. After finished the chow test, the next stage will be the hausman test, to determine whether the FEM or REM model that will be used to process the data. 76 2 Hausman Test The aim of hausman test is to determine whether FEM or REM that will be used to processing the data. The base for reject the H REM is using statistic consideration of chi square. If the hausman test is significant probability of hausman α, thus the null hypothesis is rejected and the FEM model is used. The result of Hausman test will be provided in the following table: Table 4.15 The Result of Hausman Test Cross-section RE Probability α Decision prob Prob 0.3366 0.05 REM Source: processed data Based on the chow test result that have processed in the table 4.15, it is known that the probability of Random Effect is higher than the probability of α that is 0.3366 0.05. Thus, it can be concluded that the null hypothesis REM is accepted and the REM model is used for processing the data. 77

d. Hypothesis Testing

After testing the stationary test, classic assumption test, and model selection test, it can be concluded that all of the data that have used for conducting this research are passed from stationary test, free from normality, heteroscedasticity, multicollinearity, and autocorrelation problem. After completing all of the requirements needed, next step will be test the hypothesis. After selecting the model that will be used, thus the Random Effect Model in panel data regression is used to testing the hypothesis. 1 The Impact of Macroeconomic Variables toward Credit Default Swap Spreads Partially T-Test T-test aimed to observe how much every single dependent variable is affecting the dependent variable. The hypothesis of t-test is: H = the independent variables are not affecting the dependent variable H 1 = the independent variable are affecting the dependent variable H is rejected if the t statistic t table or if the probability of t statistic α. The significance level α that used in this research are 1 0.01, 78 5 0.05 and 10 0.10 The result of t-test can be seen in the following table: Table 4.16 The Result of t-Test Variable Coefficient Std. Error t-Statistic Prob. C 35.74421 61.98233 0.576684 0.5662 GDPGROWTH? -2.531825 5.504267 -0.459975 0.6471 INFLATION? 26.14017 7.410467 3.527466 0.0008 UNEMPLOYMENT? 14.37829 4.754548 3.024112 0.0036 IMPORT? -0.068843 0.551008 -0.124939 0.9010 CAB? -8.353070 4.278162 -1.952491 0.0553 Note: Significance level at 10 Significance level at 5 Significance level at 1 a. Dependent variable: Credit Default Swap spreads Source: processed data The result of the t-test can be interpreted as follows:

a GDP Growth

GDP growth variable has the probability of 0.5662. The result shows that the GDP growth variable is not significantly affect the Credit Default Swap spreads at significance level of 1, 5 or 10. The coefficient of GDP growth variable is negative, it means that the higher GDP growth leads to the lower Credit Default Swap spread.