Determining technical default factors for credit rating models, 2021

MINISTRY OF EDUCATION AND TRAINING STATE BANK OF VIETNAM

BANKING UNIVERSITY OF HO CHI MINH CITY

NGUYEN TRAN MAI VY

DETERMINING TECHNICAL DEFAULT FACTORS

FOR CREDIT RATING MODELS

BACHELOR THESIS

MAJOR: BANKING AND FINANCE

CODE: 7340201

Ho Chi Minh City, 2021

MINISTRY OF EDUCATION AND TRAINING STATE BANK OF VIETNAM

BANKING UNIVERSITY OF HO CHI MINH CITY

NGUYEN TRAN MAI VY

DETERMINING TECHNICAL DEFAULT FACTORS

FOR CREDIT RATING MODELS

BACHELOR THESIS

MAJOR: BANKING AND FINANCE

CODE: 7340201

INSTRUCTOR: Ph.D. NGUYEN MINH NHAT

Ho Chi Minh City, 2021

i

ABSTRACT

In the process of integration to the world economy, every bank in every

country has to face with new opportunities, as well as new challenges. The fierce

survival competition between commercial banks becomes not only a major problem,

but also regular one. To be survived and develop, every administrator needs to have

directions, and specific strategies to compete with others, and gain profit for the

bank in this current context.

Credit activities are traditional activities and bring the highest profit for

banks. But of course, with high returns comes great risks. This risk not only affects

credit lending banks but also can adversely affect the entire economy, especially the

developing economy in Vietnam.

The credit rating system always plays an important role at commercial banks

in assessing customers' credit risk and assisting the bank in making credit decisions

as well as in management activities and risk treatment at the bank. In fact, it is a

prerequisite for advanced credit risk management and each credit institution wishes

to establish an internal credit rating system for its own. Moreover, a legal

framework for credit rating has been establishing by the Government to improve

information transparency and support for banks to control credit risk from the very

beginning as well as support the bond market and the stock market to not only

promote capital mobilization through the stock market, but also protect the rights

and interests of investors.

Because of that, the research and selection of appropriate rating models will

significantly contribute to the development of credit rating activities in Vietnam.

Besides, at present, the world economy in general and Vietnam's economy in

particular are facing a lot of fluctuations, so the role of the bank becomes

particularly important in reviving and bringing the economy to development.

ii

However, through the research process, the current models revealed some

limitations and inconsistency about its reliability that leads to difficulty in choosing

the suitable models. Determining which factors affect the ranking result is an

inevitable and strategic issue to more complete the credit rating system. Up to now,

there are still not many studies published in Vietnam on finding out technical

default factors that affect credit rating models.

Because of those reasons above, the researcher choose the topic

"Determining technical default factors for credit rating models” to research on.

In this research, the author will conduct to find out the technical default

factors that affect credit rating models to systematically provide commercial banks

with theoretical basis and empirical evidence related to the selection of an

appropriate corporate bankruptcy prediction model to contribute to efficiency

improvement of the bank‟s credit risk management in the future.

By consulting with people who are knowledgeable about this field, have

long-term work experience and have a good grasp of reality to be able to give more

accurate technical default factors that can affect credit rating models.

Beside that, there are 04 stages that need to be implemented as below:

First stage: Collecting and processing data;

Second stage: Selecting the input variables of the model;

Third stage: Running the regression on selected credit rating models, which

are: the Logit model, the Probit model, and the Complementary Log-Log model;

Last stage: Using the Confusion matrix and F1 - Score for evaluating each

model's regression results. From that, selecting the appropriate credit rating model

that has the ability to accurately predict the default probability of customers.

This research was conducted based on the data taken from the annual

financial statements of about 400 enterprises from 09 business fields from 2017 to

2019. In order to ensure the quality of the information source, these financial

statements have been audited.

iii

COMMITMENT AND THANKS

This thesis is the researcher's own work, the research result is honest, in

which no previously published content or content of other researchers are presented,

except for citations are fully cited in the thesis.

After a period of time studying, the researcher was helped by all the teachers

and friends in support of implementing the knowledge more and more abundant.

With deepest gratitude, the researcher would like to sincerely thank all the

teachers of the Banking and Finance Department who use their knowledge and

enthusiasm to convey the precious knowledge to students during the study period at

Banking University HCMC. In particular, the researcher would like to give a special

thank to Ph.D. Nguyen Minh Nhat for spending time in guidance and support for

the researcher‟s bachelor thesis.

Due to the researcher‟s limited knowledge and many more, the researcher

will not be able to avoid the shortcomings. Consequently, the researcher would like

to receive valuable comments from teachers and classmates so that the knowledge

in this field will be enhanced and improved.

The researcher

NGUYEN TRAN MAI VY

iv

TABLE OF CONTENTS

CHAPTER 1. INTRODUCTION ............................................1

1.1. The urgency of the research........................................................................1

1.2. Objectives of the research...........................................................................6

1.3. Questions of the research............................................................................6

1.4. Object and scope of the research.................................................................6

1.4.1. Research object....................................................................................6

1.4.2. Research scope ....................................................................................7

1.5. Research methods.......................................................................................7

1.6. Determination of the study sample .............................................................8

1.7. Expected Contributions ..............................................................................8

1.8. Structure of the research.............................................................................9

CHAPTER 2. LITERATURE REVIEW...............................11

2.1. Overview of Technical Default.................................................................11

2.1.1. Definition of Technical Default .........................................................11

2.1.2. Regulations for Technical Default......................................................12

2.1.3. Types of Technical Default................................................................12

2.2. Probability of Default (PD) ......................................................................13

2.2.1. Definition of Probability of Default ...................................................13

2.2.2. Characteristics Probability of Default ................................................14

2.2.3. Application of Probability of Default.................................................14

2.3. Overview of credit rating models..............................................................15

2.3.1. Statistical models commonly used in credit rating..............................15

2.3.2. The difference between Logit model, Probit model and Complementary

Log-Log model ...............................................................................................25

2.4. Related studies .........................................................................................27

2.4.1. Related studies in the world ...............................................................27

2.4.2. Related studies in Vietnam.................................................................30

v

CHAPTER 3. DATA AND METHODOLOGY OF

RESEARCH ..........................................................................33

3.1. Theoretical framework .............................................................................33

3.2. Data collection and processing .................................................................34

3.3. Selection of input variables in the default prediction model......................37

3.4. Models for predicting the probability of default........................................49

3.4.1. Logit model .......................................................................................49

3.4.2. Probit Model......................................................................................50

3.4.3. Complementary Log-Log Model........................................................51

3.5. The evaluation criteria of default prediction models.................................51

3.5.1. Confusion Matrix...............................................................................51

3.5.2. F1 - Score ..........................................................................................54

CHAPTER 4. EMPIRICAL RESULTS................................55

4.1. Descriptive statistics results......................................................................55

4.2. Regression results of parametric models...................................................58

4.2.1. The Logit model ................................................................................58

4.2.2. The Probit model ...............................................................................60

4.2.3. The Complementary Log – Log model ..............................................62

4.2.4. Overall conclusion about the regression results of parametric models 64

CHAPTER 5. CONCLUSION AND RECOMMENDATION

OF THE RESEARCH..................................................................72

5.1. Conclusion for the research results...........................................................72

5.1.1. Achieved results of the study .............................................................72

5.1.2. Limitations of the study: ....................................................................74

5.2. Recommendations drawn from the results of the study .............................75

5.2.1. Recommendations for optimizing the most effective technical default

factors for credit rating models .......................................................................75

5.2.2. Recommend using the results of the model to predict default

probability of customers at commercial banks in Vietnam ..............................77

vi

5.2.3. Recommend using the results of the model to predict default

probability of customers at credit rating agencies in Vietnam .........................80

5.3. Future research direction ..........................................................................86

vii

LIST OF ACRONYMS

ACB A Chau Commercial Bank

ANN Artificial Neural Network

BIDV Bank for Investment and Development of Vietnam

CIC Credit Information Center

DA Discriminant Analysis

EBIT Earnings Before Interest and Taxes

E&Y Ernst and Young Corporation

Etc. Et Cetera

FICO Fair Isaac Corporation

FN False Negative

FP False Positive

HNX Hanoi Stock Exchange

HOSE Ho Chi Minh City Stock Exchange

IATA International Air Transport Association

KNN K - Nearest Neighbor

PBT Profit Before Tax

PD Probability of Default

ROS Return on Sales ratio

ROA Return on Assets

ROE Return on Equity

SME Small and Medium Enterprise

TA Total Assets

TD Total Debts

TN True Negative

TP True Positive

UK United Kingdom

US United States

viii

LIST OF FIGURES AND TABLES

 FIGURES:

Figure 1.1. Forecast Bankruptcy Rate in 2021 compared to 2019 Page 03

Figure 1.2. Euler Hermes‟s global insolvency index and regional indices (yearly

change in %) Page 03

Figure 5.1. General process of credit rating Page 83

 TABLES:

Table 2.1. Statistical models commonly used in credit rating Page 15

Table 2.2. The difference between Logit model, Probit model and Complementary

Log-Log model Page 25

Table 3.1. Synthesize about Business fields and Number of businesses Page 35

Table 3.2. Statistics of Bankruptcy and Non-bankrupt companies Page 36

Table 3.3. Some financial analysis criteria in corporate credit rating Page 40

Table 3.4. Independent variables in probability default prediction model Page 42

Table 3.5. The structure of variables data in Logit model Page 50

Table 3.6. Confusion Matrix Page 52

Table 4.1. Descriptive statistics of the independent variables Page 55

Table 4.2. Correlation Matrix Page 57

Table 4.3. Regression results of the Logit model Page 58

Table 4.4. Confusion matrix of the Logit model Page 59

Table 4.5. Regression results of the Probit model Page 60

Table 4.6. Confusion matrix of the Probit model Page 61

Table 4.7. Regression results of the Complementary Log – Log model Page 62

Table 4.8. Confusion matrix of the Complementary Log - Log model Page 63