Predicting the probability of default for small and medium enterprises based on financial indications: Bachelor thesis of Banking and Finance / Nguyen Dieu Linh ; supervisor Nguyen Minh Nhat

BANKING UNIVERSITY OF HO CHI MINH CITY

-----oOo-----

NGUYEN DIEU LINH

PREDICTING THE PROBABILITY OF DEFAULT FOR

SMALL AND MEDIUM ENTERPRISES BASED ON

FINANCIAL INDICATORS

GRADUATION THESIS

MAJOR: FINANCE & BANKING

CODE: 7340201

HO CHI MINH CITY, 2021

MINISTRY OF EDUCATION AND TRAINING STATE BANK OF VIETNAM

BANKING UNIVERSITY OF HO CHI MINH CITY

NGUYEN DIEU LINH

PREDICTING THE PROBABILITY OF DEFAULT FOR

SMALL AND MEDIUM ENTERPRISES BASED ON

FINANCIAL INDICATORS

GRADUATION THESIS

MAJOR: FINANCE & BANKING

CODE: 7340201

SCIENCE INSTRUCTOR

Ph.D. NGUYEN MINH NHAT

HO CHI MINH CITY, 2021

MINISTRY OF EDUCATION AND TRAINING STATE BANK OF VIETNAM

i

ABSTRACT

The internal 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, risk treatment at the bank. At the same time, the

Government has been building a legal framework for the credit rating to improve

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

beginning as well as support the stock market, the bond market to promote capital

mobilization through the stock market, protect the rights and interests of investors.

Researching and selecting suitable rating models will significantly contribute to the

development of credit rating activities in Vietnam. However, the current models for

predicting default probability have certain limitations and are being debated,

inconsistency about these models' reliability, which leads to difficulty in choosing the

model is suitable to predict the probability of default of the business. Besides,

determining which financial ratios affect the ranking results is always the goal, which

needs to be studied in default prediction research. Up to now, there are still not many

studies published in Vietnam on selecting models to forecast the probability of default

of enterprises based on financial indicators.

Therefore, the thesis focuses on the issue of "Predicting the probability of default

for Small and Medium Enterprise based on financial indicators" to provide

commercial banks systematically a theoretical basis and empirical evidence related

to the selection of an appropriate business bankruptcy prediction model to contribute

to improving the efficiency in credit risk management of the bank in the future.

Based on the importance and necessity, the objective of this study is to: (i) determine

the criteria of an appropriate forecasting model; (ii) how to choose a model capable

of predicting the default probability of Small and Medium Enterprises (SMEs) at

Vietnamese commercial banks based on financial indicators. The results obtained

from this study aim to provide additional quantitative scientific evidence to answer

which predictive model gives the best results in predicting the probability of default

of medium firms and small in Vietnamese commercial banks; (iii) The most important

ii

contribution of this study is to develop a basic idea in the use of financial indicators

to forecast the default probability of SMEs, thereby contributing to improving

efficiency results in the credit risk control of commercial banks in Vietnam in the

coming time.

SMEs play a major role in most economies, particularly in developing countries.

SMEs account for the majority of businesses worldwide and are important

contributors to job creation and global economic development. Micro, small and

medium enterprises, commonly known as small and medium enterprises, are small￾sized enterprises in terms of capital, labor or turnover. Small and medium enterprises

can be divided into three categories based on their size: micro enterprises, small

enterprises and medium enterprises. According to the World Bank Group's criteria, a

micro enterprise is an enterprise with a number of employees less than 10 people; a

small enterprise with a number of employees from 10 to less than 200 people and a

capital of 20 billion or less; medium enterprises have from 200 to 300 employees with

capital of 20 to 100 billion.

Probability of default is an important component applied in many credit risk analysis

and risk management activities. According to Basel II, it is a key parameter used in

calculating the level of economic capital capable of absorbing risks at credit

institutions. PD is one of the most useful ratios for classifying borrowers. All banks,

whether using standard or other advanced methods must provide supervisors with an

internal estimate of the PD relative to the borrower to the extent of the score. The

ranking result based on PD is considered relatively accurate as it is calculated on the

firm's actual financial ratios and can practically reflect the business's state. PD can

effectively reduce credit risk if fully considered.

Through a review of domestic and foreign studies shows that financial institutions

can apply many different credit rating models to predict the default probability of

enterprises. These predictive models can be polynomial models, logit models, probit

models, artificial neural network models. Besides, these ranking models use inputs or

different financial indicators to forecast the bankruptcy of a business. Financial ratios

are commonly used as short-term solvency, rate of return/total assets, total

iii

liabilities/total assets. However, with data sets built in different periods, the

conclusions about choosing the appropriate credit rating model and financial

indicators affecting the probability of default in the researchers are different, as well

as the application in Research to predict the possibility of default of SMEs customers

in Vietnam according to the author which is a new point. Through the analysis,

comparison and synthesis of the above studies and related issues, the author has

pointed out some research gaps, proposing the proposed research model and expected

method for the topic.

To accomplish the research objectives, the author implemented through 04 stages

according to the following steps: Stage one is collect and process data; The second

phase select the input variables of the model; The third stage run the regression on

selected credit rating models (the logit model, the probit model, the complementary

log-log model); The last stage use the Confusion matrix and F1 - Score to evaluate

each model's regression results. On that basis, select an appropriate credit rating

model and has the ability to predict well the probability of default of customers.

The study was conducted based on the data, which are taken from the annual financial

statements of approximately 400 companies from 2017 to 2019. These financial

statements have been audited to ensure the quality of the information source. Out of

400 businesses, there are 31 businesses in the field of consumer goods trading; 35

enterprises in the petroleum business sector; 39 businesses in the automotive

business; 40 enterprises in the construction and installation industry; 43 enterprises

in the pharmaceutical industry and medical equipment; 45 enterprises in the textile

and garment industry; 47 enterprises in the fisheries sector (fish, shrimp, clam,...); 54

businesses in the iron and steel industry and 66 businesses in the agricultural sector

(rice, coffee, pepper,...). Based on the studies, the author selected 14 financial

indicators as independent variables for the credit rating models in the research paper.

Through analyzing the regression results from parametric models, and based on

criteria calculated from the confusion matrix (Accuracy, Sensitivity, Specificity,

Precision, F1 - Score) to compare and evaluate the ability to predict default

iv

probability of each model. Thereby finding a suitable model to predict the default

probability of enterprises.

The final result of the research shows that there are 5/14 variables play an important

role in predicting the default probability of customers, these are Income before

tax/Total assets, Total liabilities/Total assets, Earnings before tax, interest and

amortization/Long-term debt, Average cost of goods sold/Inventory and Total

revenue/Total assets. Through the research results, commercial banks can evaluate

and select customers in practice to minimize the risk that customers cannot repay their

loans.

From the research results, the author proposes some suggestions for commercial

banks on the development of the internal credit rating system in the coming time. The

thesis has found a model to predict the solvency (default probability) of SMEs

customers at commercial banks in Vietnam. The model can help stabilize credit

quality, minimize arising bad debts. Customers with a qualified credit rating (rated A

or higher) combined with the results of measuring the good repayment capacity

according to the model will have a low probability of incurring bad debt, according

to which credit risk for this group of customers is small.

The model can be seen as a supporting tool for commercial banks in credit granting,

assuring credit quality, and facilitating an efficient, safe, and sustainable expansion

and growth. From there, it can help banks select and maintain a good customer

structure, promote marketing strategies towards low-risk customers and develop a

network of reputable customers, ensuring debt repayment.

The model results are the basis for commercial banks to orient credit shrinking to

weak customers (high probability of default) and effective credit growth for well￾performing customers (low probability of bankruptcy). Simultaneously, building a

credit policy suitable for each type of customer in terms of credit terms, interest rates,

fees, requirements for security measures…to ensure safety in operation.

On the other hand, information to measure the solvency and the results of the model

also reflects many problems related to the business performance of the business and

v

the field - production and business sector. As a result, the model becomes a source of

information for future credit policy analysis, assessment, forecast and administration.