Personal Characteristics Factor Group:

Table 4.1 – Statistical table describing data of variables in the model Source: Author uses SPSS software for calculation

In 180 observed samples, the lowest age is 18 and the highest is 67 years old, the average age is 40 years old, which is quite suitable because this is the stage when borrowers have stability in work, career, capital and experience, so the ability to repay debt on time is much higher.

The minimum loan size is 30 million because this is an unsecured loan, the credit limit is based on the borrower's salary income. Along with the loan size, the borrower's income is respectively the lowest at 5 million and the highest at 182 million. The reason for the lowest income is 5 million is because there are unsecured loans for officers and employees of socio-political organizations, officers and employees working at Vietinbank, ... so the credit limit depends on the actual salary received by the customer after deducting living expenses. The loan term ranges from 12 months to 240 months because the dependent variable is considered in terms of monthly principal and interest payments. In addition, the observed sample accounts for 65.6%, mainly consumer loans, so the loan term is quite long.

The average loan interest rate of 9.9%/year is quite suitable for medium and long-term loans. Vietinbank's medium and long-term loan interest rate up to now is 10%/year, in which the lowest loan interest rate is 6.1%/year for customers who are eligible for preferential promotion programs and the highest loan interest rate is 12.5%/year because this is an unsecured loan.

4.1.2 Sample structure by independent variables :

4.1.2.1 Group of personal characteristics factors :

The research data includes 180 samples, of which 105 are male borrowers, accounting for 58.3%, and 75 are female borrowers, accounting for 41.7%. The data are quite similar for the gender variable.

Sex

Frequency

Percent

(%)

Part value

hundred (%)

Plus Percent

cumulative (%)

Male

105

58.3

Maybe you are interested!

Female

75	41.7	41.7	100
Total	180	100	100

Table 4.2 - Gender characteristics

Source: author's calculation based on SPSS software

The research results show that the majority of borrowers are married, accounting for 67.8% of the total observation, while the unmarried status accounts for 32.2%. This is quite appropriate because when considering the average age of borrowers is around 40 years old. This is the age when borrowers are usually married and have a higher need to borrow to serve their living and consumption needs.

Kissing status

core

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
Married	122	67.8	67.8	67.8
Unmarried	78	32.2	32.2	100
Total	180	100	100

Table 4.3 – Marital status

Source: author's calculation based on SPSS software

The statistical results show that the highest number of dependents in a family is 3 people and the lowest number is no dependents in the family. Of which, the highest number of dependents is 2 people, accounting for 37.2% and the lowest number is 3 people, accounting for 1.1%.

Number of members

dependent member

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
0	61	33.9	33.9	33.9
1	50	27.8	27.8	61.7
2	67	37.2	37.2	98.9
3	3	1.1	1.1	100
Total	180	100	100

Table 4.4– Number of dependent family members Source: author's calculation based on SPSS software

4.1.2.2 Borrower capacity group :

The bank also collected educational qualifications during the appraisal process. The survey showed that 36.1% of borrowers had university degrees or higher, 33.9% of borrowers had secondary to college degrees, and 30% had high school degrees or lower. These figures are quite similar. This shows that high educational qualifications are not an important criterion for deciding to grant credit limits to customers at Vietinbank. In addition, in the bank's loan application, educational qualifications are not mentioned as requiring confirmation.

Education level

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
Intermediate/Advanced class	61	33.9	33.9	33.9
Undergraduate/Graduate learn	65	36.1	36.1	70
Other	54	30	30	100
Total	180	100	100

Table 4.5 – Education level

Source: author's calculation based on SPSS software

In the research sample, the status of office work or work requiring high intelligence only accounts for 31.1% while other jobs account for 68.9%. This shows that the status of office work or work requiring high intelligence is not a decisive factor for lending at Vietinbank. The results are also similar to the sample selection of the educational level variable.

Public status

job

Frequency

Percent

(%)

Part value

hundred (%)

Plus Percent

cumulative (%)

Office work

room

31.1

Other

124	68.9	68.9	100
Total	180	100	100

Table 4.6– Job status

Source: author's calculation based on SPSS software

4.1.2.3 Loan characteristics :

With the goal of safe, effective and sustainable development set out in the bank's annual report, mortgaged assets are still the most prioritized form of lending by the bank with a mortgage rate of 81.7% in the observed sample and 18.3% including unsecured loans and unsecured loans, mainly lending to Vietinbank officers and employees.

Collateral

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
No assets warrant	33	18.3	18.3	18.3
Secured	147	81.7	81.7	100
Total	180	100	100

Table 4.7– Collateral

Source: author's calculation based on SPSS software

The dependent variable is the customer's ability to repay debt on time, considered under the condition that the customer pays principal and interest in part, so the loan purpose is mainly consumer loan, with a rate of 65.6% in the total observation, and 34.4% is business loan.

Loan purpose

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
Consumption	118	65.6	65.6	65.6
Business	147	34.4	34.4	100
Total	180	100	100

Table 4.8 – Loan purpose

Source: author's calculation based on SPSS software

4.1.2.4 Moral hazard group of borrowers :

The customer's overdue debt history is provided through the credit information center (CIC). Currently, banks mainly grant credit to customers who have no credit relationship or have a credit relationship history with group 1 - Standard debt group. In some cases, banks still support granting credit to group 2. Therefore, the overdue debt history variable of customers at Vietinbank with the group of no overdue debt accounts for 90% of the total observation.

History of overdue debt

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
No/Not yet	162	90	90	90
Have had	18	10	10	100
Total	180	100	100

Table 4.9 – Customer overdue debt history Source: author's calculation based on SPSS software

4.1.2.5 Operational risk group from the bank:

In terms of experience, loan appraisal staff with an observation sample of employees with 2 years or more accounted for 61.7%, equivalent to 111/180 samples, while employees with 2 years or less accounted for 38.3% of the total observation. The results show that Vietinbank, as well as other banks, focus on appraisal staff with many years of experience, so in recruitment rounds, priority is often given to experienced staff.

Experience

CBTD

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
< 2 years	69	38.3	38.3	38.3
> 2 years	111	61.7	61.7	100
Total	180	100	100

Table 4.10 – Experience and qualifications of loan appraisal officers Source: author's calculation based on SPSS software

4.1.2.6 Ability to repay debt on time :

Ability to pay debt

Frequency	Percent (%)	Part value hundred (%)	Plus Percent cumulative (%)
Late	70	38.9	38.9	38.9
On time	110	61.1	61.1	100
Total	180	100	100

Table 4.11 – Ability to repay debt on time Source: author's calculation based on SPSS software

According to the sample selection results of Table 4.11, there are 110/180 customers who are able to repay loans on time, accounting for 61.1%, and 38.9% of customers who repay loans late. The number of samples selected is quite appropriate because the group of customers with overdue debt history accounts for 90% of the total observation, which is customers who do not have overdue debt at the Bank.

4.2 Optimization model building process:

Through the collected data processed through Excel software and using SPSSS software to run binary logistic regression, the author conducted regression analysis through the following steps:

Step 1 : Enter all independent variables into the model. After running the binary logistic model data through SPSS software, we have the following results:

B	SE	Wald	df	Sig.	Exp(B)
Step 1 a	SEX	9,198	3,834	5,756	1	.016	9.875E3
	AGE	.162	.097	2,817	1	.093	1,176
	MAR	-1.723	1,662	1,076	1	.300	.178
	HOS	3,705	2,095	3.130	1	.077	40,667
	EDU_1	1,844	2,220	.690	1	.406	6.323
	EDU_2	-1.695	2.016	.707	1	.401	.184

WORK	5.360	2,585	4,300	1	.038	212,760
SOL	-.027	.010	7.121	1	.008	.973
TIME	.189	.079	5,723	1	.017	1,207
INC	.959	.355	7,271	1	.007	2,608
INT	476,934	212,952	5.016	1	.025	1.348E2 07
SEC	-42.368	16,152	6,880	1	.009	.000
TOL	30,720	11,495	7,143	1	.008	2.196E1 3
CIC	21,583	7.096E3	.000	1	.998	2.362E9
IOS	-6.745	2,880	5,484	1	.019	.001
Constant	-65.098	7.096E3	.000	1	.993	.000

Table 4.12 – Model results after running step 1 Source: the author used SPSS software for research

Based on table 4.12, we have the following research model:

Table 4.12 shows that this model is not selected because the variables AGE, MAR, HOS, EDU_1, EDU_2 and CIC have sig levels greater than 0.05, so they are not statistically significant. Therefore, the above variables will be eliminated from the model (see appendix for detailed results).

Step 2 : Remove the variables AGE, MAR, HOS, EDU_1, EDU_2 and CIC from the model, continue running the model with the remaining variables. We get the results of the second model run as follows:

B	SE	Wald	df	Sig.	Exp(B)
Step 1 a	SEX	3,374	1,268	7,077	1	.008	29,195
	WORK	.997	1,049	.904	1	.342	2,711
	SOL	-.015	.004	10,842	1	.001	.985
	TIME	.061	.024	6,518	1	.011	1,063
	INC	.542	.171	10,077	1	.002	1,719
	INT	342,812	112,533	9,280	1	.002	7.609E148
	SEC	-24.493	7,241	11,441	1	.001	.000
	TOL	16,420	4,478	13,447	1	.000	1.353E7
	IOS	-2.766	1,208	5.245	1	.022	.063
	Constant	-22.265	8,385	7,051	1	.008	.000

Table 4.13 – Model results after running step 2 Source: the author used SPSS software for research

Based on table 4.13, we have the following research model:

Table 4.13 shows that this model is not selected because the variable WORK has a sig level greater than 0.05, so it is not statistically significant. Therefore, the variable WORK will be removed from the model (see the appendix for detailed results).

Step 3 : Remove the WORK variable from the model, continue running the model with the remaining variables. We get the results of the third model run as follows:

Comment