Home / Personal

Testing the Model's Predictive Accuracy


At the same time, the Nagelkerke R Square coefficient = 0.738 shows that 73.8% of the change in the dependent variable is explained by 7 independent variables, the rest is due to other factors.

4.4.2.3. Testing the model's predictive accuracy

The results of the model's prediction level are presented in Table 4.13.

Table 4.13: Forecast level



Observe

Forecast

RRTD

% accurate forecast

No credit risk

There is credit risk



No risk

credit

333

7

97.9

Step 1

RRTD

There is credit risk.

use


11


31


73.8



Total



95.3

Maybe you are interested!

Testing the Models Predictive Accuracy

(Source: Author's survey results, 2019)

Based on the classification table, it shows that: The number of customers with no real credit risk is 340 (333 + 7) customers, of which the prediction result is 333 customers, showing that the ability to predict customers with no credit risk is 97.9% accurate. For the subjects with real credit risk: 42 customers (31 + 11), the prediction result is: 31 customers, accounting for: 73.8%. From this, it shows that the model has an average prediction level of: 95.3%.

4.4.2.4. Logit regression model results

The results of the multivariate logit analysis are presented in Table 1.

4.14. Effect coefficient assesses the level of influence of factors on the possibility of personal credit risk at Vietinkbank Ba Ria Vung Tau branch.


Table 4.14: Logit regression model results


Independent variable

B

SE

Wald

df

Sig.

Exp(B)

Step

1 a

Sex (X1)

-1,541

0.601

6,584

1

0.010

0.214

Age (X2)

1,102

0.363

9,230

1

0.002

3,010


Honnhan (X3)

-3,022

0.994

9,237

1

0.002

0.049


Tvphuthuoc (X4)

1,044

0.415

6,327

1

0.012

2,840


Technology (X5)

-0.151

0.707

0.046

1

0.831

0.860


TGoDCHT (X6)

0.364

0.227

2,579

1

0.108

1,439


Vitrivieclam (X7)

1,142

0.447

6,519

1

0.011

3,134


Thuhap (X8)

-3,180

0.583

29,727

1

0.000

0.042


Tinhtrangnhao (X9)

1,355

0.374

13,109

1

0.000

3,877


Transfer (X10)

0.130

0.377

0.119

1

0.730

1,139


Constant

-2,947

2,380

1,534

1

0.215

0.052

(Source: Author's survey results, 2019)


Looking at Table 4.14, we see that the sig. of the independent variables occupation (sig. = 0.831), time at current address (sig. = 0.108), and professional competence (sig. = 0.730) are all greater than 5%, so the relationship between changes in occupation, professional competence, time at current address, and credit risk of individual customers is not statistically significant.

The Sig. values ​​of the variables gender (sig. = 0.010); age (sig. = 0.002); marriage (sig. = 0.002); number of dependents (sig. = 0.012); job position (sig. = 0.011); income (sig. = 0.000), housing status (sig. = 0.000) < 0.05 (5%). Therefore, the relationship between the remaining independent variables and the individual customer credit risk variable is statistically significant with a general confidence level of 95%.

4.4.2.5. Discussion of regression results

With a significance level of 5%, the regression coefficients of the variables gender (X1); age (X2); marriage (X3); number of dependents (X4); job position (X7); income (X8),


Housing status (X9) is statistically significant, in other words, the above factors affect the credit risk of individual customers at Vietinbank Ba Ria Vung Tau branch.

The model is defined as:

Ln(Pi/1-Pi) = -2.947 -1.541*X1 + 1.102*X2 -3.022*X3 + 1.044*X4 + 1.142*X7 –3.180*X8 + 1.355*X9 (1)

In table 4.14, it is shown that, using the results of the regression coefficient column (B) and the Exp(B)=eB column, the thesis will form a probability change scenario when the initial probability is 10%, 20%, 30%, 40% and 50% respectively.

Let P 0 : Initial probability

P 1 : Probability of change. P 1 is calculated by the following formula:

𝑃0∗ 𝑒

P 1 =

1−𝑃0 (1− 𝑒 𝛽 )

Table 4.14: Simulation results of individual customer credit risk probability



Dependent variable Credit risk (Y=1)


Regression coefficient


Exp(B) coefficient

Simulate the probability of Credit Risk when the independent variable changes by 1 unit and the initial probability is:

10%

20%

30%

40%

50%

Independent variable








sex

-1.541

6,584

42.2%

62.2%

73.8%

81.4%

86.8%

age

1.102

9,230

50.6%

69.8%

79.8%

86.0%

90.2%

kiss

-3.022

9,237

50.7%

69.8%

79.8%

86.0%

90.2%

TVphuthuoc

1,044

6,327

41.3%

61.3%

73.1%

80.8%

86.4%

vitrivieclam

1,142

6,519

42.0%

62.0%

73.6%

81.3%

86.7%

elastic

-3.180

29,727

76.8%

88.1%

92.7%

95.2%

96.7%

Tinhtrangnhao

1,355

13,109

59.3%

76.6%

84.9%

89.7%

92.9%

(Source: Author's survey results, 2019)


Explain the impact of factors:

The variable “Customer gender (X1)” has a coefficient of X1 = -1.541, statistically significant at the 5% level, negatively correlated with individual customer credit risk. This variable has the fifth strongest impact on individual credit risk in the regression model. Suppose the initial probability of individual customer credit risk is 10%. When other factors remain unchanged, if the customer is female, the credit risk of this customer is 32.2%. The results show that the Female group, coded as 0, has a higher level of credit risk than the Male group, coded as 1. This result is consistent with the study of Marjo Hörkkö (2010) when testing the impact of individual customers on credit risk at commercial banks in Finland.

The variable “Customer age (X2)” has a coefficient of X2 = 1.102, statistically significant at the 5% level, and is positively correlated with individual customer credit risk. This variable has the fourth strongest impact on individual credit risk in the regression model. Suppose the initial probability of individual customer credit risk is 10%. When other factors remain unchanged, if the customer’s age increases by 1 unit, the credit risk of this customer is 40.6%. This result shows that as the customer’s age increases, the customer’s credit risk also increases.

The variable “Marital status (X3)” has a coefficient of X3 = -3.022, statistically significant at the 5% level, negatively correlated with individual customer credit risk. This variable has the third strongest impact on individual credit risk in the regression model. Suppose the initial probability of individual customer credit risk is 10%. When other factors remain unchanged, if the customer is married, the credit risk of this customer is 40.7%. The results show that the single group, coded as 0, has a higher level of credit risk than the married group, coded as 1. This result is consistent with the research of Marjo Hörkkö (2010), Li Shuai et al (2013).

The variable “Number of dependent members (X4)” has a coefficient of X4 = 1.044, statistically significant at the 5% level, in the same direction as individual customer credit risk. This variable has the lowest impact on individual credit risk in the regression model.


Assuming the initial credit risk probability of an individual customer is 10%. When other factors remain unchanged, if the number of dependents in the family increases, the credit risk of this customer is 31.3%. The results show that the more dependents an individual has, the higher their personal credit risk will be. In addition to personal expenses, expenses for dependents are also an amount that must be taken into account when evaluating a borrower. The number of dependents directly affects the customer's income to repay the loan; the more dependents there are, the greater the expenses for dependents, leading to a decrease in the customer's income, affecting the customer's ability to repay the loan. The author's research results are also consistent with the research results of Marjo Hörkkö (2010), John M. C (1940), Truong D. Loc & Nguyen T. Tuyet (2011).

The variable “Job position (X7)” has a coefficient of X7 = 1.142, statistically significant at the 5% level, and is positively correlated with individual customer credit risk. This variable has the sixth strongest impact on individual credit risk in the regression model. Assuming the initial probability of individual customer credit risk is 10%. When other factors remain unchanged, if the customer is a business owner and manager, the credit risk of this customer is 32.0%. The results show that the position group of Business Owners and Managers has a lower credit risk than the groups of trained workers and other groups. It can be seen that, with job positions that tend to be manual labor, the level of risk will increase. This influence has been demonstrated in the study of Marjo Hörkkö (2010) conducted at commercial banks in Finland.

The variable “Financial capacity (X8)” has a coefficient X8 = –3.180, statistically significant at the 5% level, negatively correlated with individual customer credit risk. This variable has the strongest impact on individual credit risk in the regression model. Suppose the initial probability of personal customer credit risk is 10%. When other factors remain unchanged, if the customer's financial capacity can increase by 1 unit, the credit risk of this customer is 66.8%. Financial capacity is shown through


As the customer's income increases, the level of credit risk will decrease accordingly. The customer's income is the main source of debt. The customer's income is influenced by their working time and occupation. The higher the customer's income, the better the loan security, the higher the ability to repay the loan. Thus, the test results in the author's model coincide with the previously published research results of the authors (Roszbach, 2004); Agarwal et al. (2009); Hörkkö (2010).

The variable “Housing status (X9)” has a coefficient of X3 = 1.355, statistically significant at the 5% level, and is positively correlated with individual customer credit risk. This variable has the second strongest impact on individual credit risk in the regression model. Assuming the initial probability of individual customer credit risk is 10%. When other factors remain unchanged, if a customer has his own house, his credit risk is 49.3%. The group of people who own their own house has the highest level of credit safety, while the group who rents a house shows the highest level of risk. Banks often rely on collateral to ensure the safety of their loans, so the customer's home ownership is also an important factor in the bank's loan approval process. A customer who owns a house is considered more likely to repay the loan than one who does not own a house. This has been mentioned in the study on individual customer characteristics by Agarwal et al. (2009).


CHAPTER 4 SUMMARY

Chapter 4 presents the results of the analysis of the current situation of personal credit at the Branch in the period from 2015 to 2018, in which, it reflects the problem of bad debt status among individual customers at the Branch in this period. Chapter 4 also presents the results of the analysis of survey data for 382 customers using the personal credit services of the Branch at the present time, in which the survey content on customer characteristics is the key issue that the study exploits, in order to analyze the factors that affect the credit risk of customers. The research results show that gender (X1); age (X2); marriage (X3); number of dependents (X4); job position (X7); income (X8), housing status (X9) affect the credit risk of individual customers at Vietinbank Ba Ria Vung Tau branch.


CHAPTER 5: CONCLUSION AND RECOMMENDATIONS


5.1. Conclusion

Personal credit contributes to the circulation of capital in society, moving capital from low-efficiency areas to high-efficiency areas, from surplus areas to shortage areas to meet the capital needs for business or consumption of individuals/households. Commercial banks are increasingly showing interest in developing personal credit services, including developing the number of services, the number of customers, as well as the value of loans. However, the issue of personal credit risk has also increased recently. Therefore, determining which factors affect the risk level of these loans is necessary for banks to be proactive in minimizing credit risks for individual customers, which are very diverse in terms of subjects and have many potential risks when the issue of managing individual customers is still complicated, from information management to legal issues related to measures to resolve bad debts.

Vietnam Joint Stock Commercial Bank for Industry and Trade - Ba Ria Vung Tau Branch, as well as other banks in the NTHM system, have focused on developing personal credit services in recent years, and have also made important strides in providing services. Vietinbank Ba Ria Vung Tau Branch has seized the opportunity, seen the opportunity and fully utilized the above advantages to achieve encouraging results. Personal lending activities have always played a leading role in the Bank's credit activities and have grown with outstanding loans increasing continuously over the years and attracting a large number of customers to transact. At the same time, the development of personal lending activities has contributed to the overall development of the branch as well as the success of the entire VietinBank Vietnam system.

The personal credit products of VietinBank Ba Ria Vung Tau branch have satisfied most of the needs of customers in the area, especially the loan products for "buying houses", "production and business", "repairing houses", "consumer loans".

Comment


Agree Privacy Policy *