CHAPTER 4 DATA ANALYSIS 4.1. Data Hypothesis The hypothesis for each independent variable to express our expectations about the characteristic of each independent variable and the pay back performance i.e. whether the payment is on time or not, are explained below respectively: Hypothesis 1 : Probability of credit delinquency increases when interest rate increases due to increases in payback amount. Hypothesis 2 : female have less credit delinquency risk due to their precautionary motives. And males have more credit delinquency risk. Hypothesis 3 : older people have less credit delinquency risk due to their precautionary motives. Hypothesis 4 : married clients tend to pay back their loans faster. So we expect that married clients pay their installments on time. Hypothesis 5 : probability of credit delinquency increases when the loan size increases. So, we expect to find a positive relationship between loan size and clients payback performance, assuming there is no inflation. Hypothesis 6 : According to Özlem Özdemir s study, probability of credit delinquency increases when maturity increases because the longer term, the more probability of being missed installment because of forgetfulness and carelessness etc. 17
Hypothesis 7 : people working in private sector tend to pay back their loans on time more than others do since people working in private sectors are fixed income earners. Table 4.1 the Hypotheses Table Demography Description Relationship Interest Sex Age Marital status Principal Debt Payment period Occupation Increase interest Decrease in interest Male Female Younger Older Married Unmarried High Principal Low Principal Longer time of payment period Shorter time of debt payment period Fixed income (government officer) Entrepreneurship and professional More probability to get high collectability Less probability to get high collectability More probability to get high collectability Less probability to get high collectability More probability to get high collectability Less probability to get high collectability Less probability to get high collectability More probability to get high collectability More probability to get high collectability Less probability to get high collectability More probability to get high collectability Less probability to get high collectability Less probability to get high collectability More probability to get high collectability 4.2. Data Collection Data used to explain the relationship between credit loan payments and some financial and demographic variables were obtained through multipurpose consumer credit records of one of the biggest Indonesia s local government banks which is Bank Jabar. The data is taken from the main branch of Bank Jabar in Bandung. It is because the credit activities in Bank Jabar concentrate in this main branch and the data from the main branch is the most complete for all variables needed. 18
The data consist of the consumer s data that has been approved by Bank Jabar and monitored until now. Bank Jabar already create the collectability on each customer according to the customer performance. This final project analyze the data that has been approved by the bank qualification and collectability rated by the bank. Data collected during the data collection payment period to test the characteristic on non performing loan in Bank Jabar are collectability, gender, status, age, interest, principal, payment period of payment and job type. In order to preserve confidentiality, customer s personal information is not given in this research. The dataset consists of 125 individuals from West Java who were granted a loan between the years 2001 and 2007. When monitored on 2007, the loans were either still paying regular installments and interest or had been amortized completely. The data contains credits, which are repaid monthly with installments that are constant along the payback period. The wages is not included because it can be represented by the payment period and principle. Because of Bank Jabar has policy that the installment should not be more than 60 % of the wages (60% for PNS Pemda, 50% for PNS non Pemda and 40% for entrepreneur and professional). 4.2.1. Descriptive Statistic: Collectability Collectability varies on 5 types. There are collectability pass (1), special mention (2), sub standard (3), doubtful (4), and loss (5). More high the number of the collectability means on more bad of the quality of the credit. In this research, the data for collectability is equal to support the research result. Because of that, the portion of collectability will be 25 people for each of it. 19
COLLECTABILITY Level 1 20% 20% Level 2 Level 3 20% 20% Level 4 20% Level 5 Figure 4.1.Collectability 4.2.2. Descriptive Statistic: Gender Gender divided into 2 (two) types, female and male. It infers that male hold the biggest portion of the sample for 78.40% (98 people) and in the other hand female for 21.6% (27 people). GENDER Female 22% Male 78% Figure 4.2.Gender 4.2.3. Descriptive Statistic: Marital Status In this research the marital status are divided into 2 (two) kind, that are not married and married. The biggest portion in the sample is the married portions that have 77.6% and the other hand unmarried people have 22.4% for the sample. 20
MARITAL STATUS Unmarried 22% Married 78% Figure 4.3.Marital Status 4.2.4. Descriptive Statistic: Age Age are varies from 35 until 50. The biggest portion comes from the 50 years old that portion for 16% of the sample. Then the smallest portion comes from the 36 years old that portion for 1.6% for the sample. AGE 19% 36-40 41-45 52% 45-50 29% Figure 4.4 Ages The order from the highest of the sample are age 45-50 (52%), 41-45 (29%), and 36-40 (19%). 4.2.5. Descriptive Statistic: Interest Interest varies from 11%, 12.5%, 14%, 14.5% and 15%. Interests for 15% are arising from the policy of the Bank at 2001. Then at the beginning of 2003 it decreased to 14.5% and at the end of 2003 it decreased again to 14%. At the beginning of 2004, they decreased its interest to 12.5% and at the end of the year 2004 it decrease to 12%. It is 21
depend of the policy of the bank to decide the level of interest given to the customer. The adjustment will be comes from the BI rate, inflation and benchmark to another bank. INTEREST 10% 26% 16% 27% 21% 11% 12.50% 14% 14.50% 15% Figure 4.5.Interest The interest level at 11% is the biggest portion in the sample for 27.2% and then followed by 15% for 25.6% of the sample, 12.5% for 20.8% of the sample, 14% for 16% of the sample and the last 14.5% for 10.4% of the sample. 4.2.6. Descriptive Statistic: Principal Principal is the first initial principal of the credit borrowed by customer. The principal in the sample varies from 6 million until 50 million. The biggest portion of the principal is the principal for 26 million until 35 million (42%) then followed by 16 million until 25 million (38%), 6 million until 15 million (12%) and the last is for more than 35 million (8%). PRINCIPAL > 35 million 8% 6-15 million 12% 26-35 million 42% 16-25 million 38% Figure 4.6.Principal 22
The principal created from the customers has passed the criteria and regulation given by the Bank. Because of Bank Jabar has policy that the installment should not be more than 60 % of the wages (60% for PNS Pemda, 50% for PNS non Pemda and 40% for entrepreneur and professional) and it creates qualification on the principal. 4.2.7. Descriptive Statistic: Payment Period Payment period of time that the customers have a loan of in the sample varies from 2 years until 8 years. The biggest portion of payment period of time comes from 5 years payment period of time then followed by payment period for 3 years (12.8%), 4 years (4.8%), 6 years (2.4%), 8 years (2.4%), 2 years (0.8%) and 7 years (0.08%). PAYMENT PERIOD 2%1% 2% 1% 13% 5% 76% 2 years 3 years 4 years 5 years 6 years 7 years 8 years Figure 4.7.Payment period 4.2.8. Descriptive Statistic: Job type There are 3 types of job in the sample that are PNS Pemda, PNS Non Pemda and Entrepreneur and professional. PNS Pemda is the government employee that works in the local government and the PNS Non Pemda is the government employee that did not work in the local government but for another. Entrepreneur and professional is the entity, organization or corporation that make contract with the Bank to hold an agreement for the credit. Usually, Entrepreneur and professional did not have an active account in the bank and it result on more risk in the quality of the credit. 23
JOB TYPE Entrepreneur and Professional 27% Pemda 23% PNS Non Pemda 50% Figure 4.8.Job Type As the figure shows, the PNS Non PEMDA holds the biggest portion in the sample for 49.6%, the second is Entrepreneur and professional for 27.2% and the last is PNS Pemda for 23.2%. 4.3. Dummy Variables Several of the variables observed during the research are dummy variable. It means that the variable shows in the qualitative values, not the quantitative value. There is treatment used to convert the qualitative variable into the quantitative value. When there are two types of variable, will be type for the quantitative as 0 and 1. In this research there are two variables that have two type categorical data that are gender and status. For gender, female convert to 0 and male convert to 1. For status, unmarried convert to 0 and married convert to 1. Table 4.2.Dummy Variable Input for Gender and Status Gender Status Category Score Female 0 Male 1 Unmarried 0 Married 1 24
Then, where the variable is has 3 types in one dummy variable, need to make two variables there. In this research job type has three categories; those are Pemda, PNS Non Pemda and Professional. In order to convert it in the quantitative values, there need two variables. While the PEMDA explained by 1 at Z 11 and 0 at Z 12, PNS Non Pemda explained by 0 at Z 11 and 1 at Z 12 and Entrepreneur and professional explained by 0 at Z 11 and 0 at Z 2. Table 4.3.Dummy Variable Input for Job Type Z 11 Z 12 PEMDA 1 0 PNS NON PEMDA 0 1 Entrepreneur and professional 0 0 4.4. Data Analysis Data processing will be used the SPSS 15. There will be analysis of the linearity of the variable. Because of this research using sample to predict the behavior of the population, the level of confidence in this research is 95%. And this research uses 125 data of customer as the representative of the whole population, because of that the degree of freedom in this research is 125-1 = 124.Because of the research use the level of confidence at 95% and degree of freedom at 124 it get the t value for 1.998. The data that have T value around 1.998 and + 1.998 will be eliminated because it has no significant value for the model. 4.4.1 The First Attempt The first attempt done by insert all the demography and financial condition variable to see the characteristic to the colllectability. In this research, to discover the characteristic, the research use the multiple regression. 25
The equation of the result in the first attempt will be : y = β 0 + β 1 x 1 + β 2 x 2 + β 3 x 3 + β 4 x 4 + γ 1 z 1 + γ 2 z 2 + γ 3 z 3 (Eq. 4.1) β 0 = the intercept given by the model β 1, β 2, β 3 = the slope of the scale variable in each variable γ 1, γ 2, γ 3 = the slope of the dummy variable in each variable x 1= variable for the scale variable age x 2 = variable for the scale variable interest x 3 = variable for the scale variable payment period x 4 = variable for the scale variable principal z 1 = gender variable (dummy variable) z 2 = status variable (dummy variable) z 3 = job type variable (dummy variable) Then all the variables collected in the data collection (collectability, gender, status, age, interest, payment period, principal, and job type) are tested to the SPSS 15. The Result Table 4.4 Model Summary for the First Attempt R R Square Adjusted R Square Std. Error of the Estimate.859(a).738.720.749 a. Predictors: (Constant), Status, Plafond, Gender, Age, Interest, Period, Job R, the multiple correlation coefficients, is the correlation between the observed and predicted values of dependent values. The R for this model is 0.859 that means the stronger relationship. From table 4.4, The R Square is 0.738 It is means that the R Square are near 1 than 0, then the model is fit the data well with the population. 26
Table 4.5.ANOVA for the First Attempt Sum of Squares df Mean Square F Sig. Regression 181.534 8 22.692 40.500.000(a) Residual 64.434 116.560 Total 245.968 124 a. Predictors: (Constant), Status, Principal, Gender, Age, Interest, Payment period, job b. Dependent Variable: Collectability Table 4.5 summarizes the result of an analysis of variance. The output regression displays the variation result from the sample and the output residual displays the variation result that is not accounted by the model (sample). A model with a large regression sum of squares in comparison to the residual sum of squares indicates that the model accounts for most of variation in the dependent variable. The significance value of the statistic is 0.000(a) that means the independent variables do a good job explaining the variation in the dependent variable. In this model, the output in regression is larger than the output of residual. Then the model accounts for most of variation in the dependent variable. Table 4.6.Coefficients for the First Attempt Unstandardized Coefficients B Std. Error T Sig. (Constant).642 1.116.575.566 Gender (D) -.577.177-3.263.001 Status (D) -.100.173 -.578.564 Age.073.017 4.188.000 Interest.030.047.646.519 Payment -.192.086-2.229.028 period Principal.048.010 4.889.000 Z11_job -2.673.201-13.327.000 (D) Z12_job (D) -1.044.170-6.141.000 From table 4.6, as the coefficients table result, there are B results from the model. The model will be: Collectability = 0.642 0.577 Gender 0.1 Status + 0.073 Age + 0.03 Interest 0.192 Payment period + 0.048 Principal 2.673 z11_job 1.044 z12_job. 27
Some of the variables give no significance level, such as status and interest. Status has the significant value level for 0.564 and interest has the significant value level for 0.519. It means that the variables are not significance to explain the dependent variable. The significance value of the statistic is 0.000(a) that means the independent variables do a good job explaining the variation in the dependent variable. The significant level can be came from the significant variables or maybe came from the insignificant variables.because of that, the second attempt must be done to recheck again the significant of all the variables by eliminating the variables that have less significant for the model. 4.4.2 Second Attempt The second attempt in trying to find the best fit model is by eliminate the independent variable that have no significant with the dependent variable. From the first attempt, status and interest have no quite significant with the collectability. Status has the T value for -0.578 with the significant level at 0.564 and interest have the T value for 0.646 with the significant level at 0.519. Then the second attempt done by eliminates the status and interest variable. y = β 0 + β 1 x 1 + β 2 x 2 + β 3 x 3 + γ 1 z 1 + γ 2 z 2 (Eq. 4.2) β 0 = the intercept given by the model β 1, β 2, β 3 = the slope of the scale variable in each variable γ 1, γ 2, γ 3 = the slope of the dummy variable in each variable x 1 = variable for the scale variable age x 2 = variable for the scale variable payment period x 3 = variable for the scale variable principal z 1 = gender variable (dummy variable) z 2 = job type variable (dummy variable) 28
Table 4.7.Model Summary for the Second Attempt R R Square Adjusted R Square Std. Error of the Estimate.858(a).737.723.744 a. Predictors: (Constant), Principal, Age, Gender, Payment period, job Compared from table 4.4 and 4.7, the R square is decreasing from 0.738 to 0.737 it is because of the eliminating of the variable status and interest that caused the decreasing of the value. From table 4.7 the R square is high enough to explain the data fit well with the population. It means that the status and interest variable are not contribute most in the R square. Because of that, the model 1 can be enough to explain the characteristic of the collectability. The standard error of the estimate is 0.744. It means that the deviation arise for the model is.744. Table 4.8 ANOVA for the Second Attempt Sum of Squares Df Mean Square Regression 181.186 6 30.198 54.538.000(a) 64.782 118.554 Residual 245.968 124 Total a. Predictors: (Constant), Principal, Age, Gender, Payment period, job b. Dependent Variable: Collectability F Sig. From table 4.8, the sum of squares for the regression is 181.186 and the sum of squares for residual is 64.782. It means that the output in regression is larger than the output of residual. A model with a large regression sum of squares in comparison to the residual sum of squares indicates that the model accounts for most of variation in the dependent variable. The significance value of the F statistic is 0.000(a) that means the independent variables do a good job explaining the variation in the dependent variable. 29
Table 4.9.Coefficients Result for the Second Attempt Unstandardized Coefficients T Sig. B Std. Error (Constant) 1.098.896 1.225.223 -.593.170-3.482.001 Gender (D).071.017 4.233.000 Age -.188.085-2.211.029 Payment period.047.009 4.935.000 Principal (D) Job1 (D) -2.693.198-13.624.000-1.052.169-6.239.000 Job2 (D) a. Dependent Variable: Collectability There in the table of coefficient (table 4.9), the independent variables have great significant in explain the dependent variable. Gender has the significant level of 0.001, age have 0.000 payment period have 0.029 and principal and job have 0.000. Even though the constant intercept get the significant level for 0.223, from R square and significant level in the ANOVA are explain the intercept can be use in the model. From table 4.9, as the coefficients table result, there are B results from the model. The second model will be: Collectability = 1.098 0.593 Gender + 0.071 Age 0.188 Payment period + 0.047 Principal 2.693 z11 job 1.052 z12_job. For example, if there is an applicant that is male, age for 45, ask for 40 million consumer credit for 10 years period and he is work in PNS non Pemda. What will be the probability of its credit if credit officer use the empirical approach. Collectability = 1.098 0.593(1) + 0.071 (45) 0.188 (10) + 0.047 (1.88) 2.693 (0) 1.052 (1) = 2.648 30
From the model equation, the probability of the applicant is 2.648. It means that either the credit performance is in the 2 and 3 area. There is probability in the future the applicant credit is in around the special mention and sub standard. The bank can give more attention to this applicant if the bank wants to accept the loan.also, there should be consideration in the standard deviation of the model. When the prediction model result, the user must consider the standard deviation of the model in the decision making. 4.5 Test Hypotheses After the result arise, then compare the result to the hypotheses create before. The equation after the entire variable shows the variable indicates the significant result is: Collectability = 1.098 0.593 Gender + 0.071 Age 0.188 Payment period + 0.047 Principal 2.693 z11_job 1.052 z12_job. Interest is expected to get more probability to loss when there is increasing in the interest loan given to the customer. In fact, the model result shows that interest have no significant result in the relationship of the collectability (dependent variable). Gender is expected to get more probability to loss when the customer is male. In fact, the model result shows that female that has more probability to loss compare the male. For the customer is male there will be decreasing value for 0.593 to the collectability, where there is no decreasing value for the female customer. Age is expected to get more probability to loss when there is increasing of the age of the customer. That means that more age of the customer give more probability to loss. In fact, the model result shows that getting older of the age of the customer give the more probability of the bad debt. In every increasing of the age of the customer it will add an increasing value for 0.071. 31
Marital status is expected to get more probability to loss when the customer is unmarried. It means that unmarried customer give more probability to get bad debt rather than married customer. In fact, the model result shows that marital status has no significant result in the relationship of the collectability (dependent variable). Payment period of debt is expected to get more probability to loss when the customer has the longer time of debt payment period. It means that the customer that have longer time of debt payment period have more probability of getting bad debt in their credit. In fact, the model result shows that the shorter payment period of debt of the customer will give more probability of getting bad debt. In every increasing of value in payment period of the customer will decrease the value of collectability for 0.188. Principal of debt is expected to get more probability to loss when the customer has higher principal or principal. It means that the customer that have higher principal will give more probability of getting bad debt. In fact, the model result shows that the higher principal will give more probability of getting bad debt. In every increasing of value in principal of customer (in million) will increase the value of collectability for 0.047. Occupation is expected to get more probability to loss when the customer is having no fixed income compare to the fixed income customer. In fact, the no fixed income gives more probability to get bad debt. For the PNS Pemda there will be decreasing value for -2.693, for the PNS non Pemda there will be decreasing value for -1.052 and for the entrepreneur and professional there are no decreasing value. 32