Master thesis Banking Finance: Impact of capital structure on performance of listed material manufacturing enterprises in Vietnam

MINISTRY OF EDUCATION AND TRAINING

STATE BANK OF VIETNAM

BANKING UNIVERSITY HO CHI MINH CITY

TRUONG THI KIM NGAN

IMPACT OF CAPITAL STRUCTURE ON PERFORMANCE

OF LISTED MATERIAL MANUFACTURING ENTERPRISES

IN VIETNAM

MASTER THESIS

Major: Banking - Finance

Code: 8 34 02 01

Ho Chi Minh City – 2023

MINISTRY OF EDUCATION AND TRAINING

STATE BANK OF VIETNAM

BANKING UNIVERSITY HO CHI MINH CITY

TRUONG THI KIM NGAN

IMPACT OF CAPITAL STRUCTURE ON PERFORMANCE

OF LISTED MATERIAL MANUFACTURING ENTERPRISES

IN VIETNAM

MASTER THESIS

Major: Banking - Finance

Code: 8 34 02 01

MENTOR: ASSOC. PROF. LE PHAN THI DIEU THAO

Ho Chi Minh City – 2023

i

DECLARATION OF AUTHENTICITY

This thesis is the author's own research, the research results are authentic, in which

there is no previously published content or the content done by others except the

citations cited in the thesis.

I declare that all statements and information cited in this thesis are true, accurate and

correct to the best of my knowledge and belief.

I am fully responsible for the thesis’s authenticity.

Ho Chi Minh City, February 25th, 2023

Student

TRUONG THI KIM NGAN

ii

GRATEFULNESS

Firstly, in order to complete this graduation thesis, I would like to express my deep

gratitude to mentor Associate Professor Le Phan Thi Dieu Thao who has guided

me from writing outline, enthusiastically giving feedback, encouraging and

following the plan to ensure that the thesis is completed in a best and timely manner.

Next, I would also like to express my sincerest thanks to all the lecturers of Banking

University of Ho Chi Minh City for imparting valuable knowledge to me during the

course so that I have a good foundation in order to complete the graduation thesis.

Finally, it is indispensable to the enthusiastic support from my friends and family

who always accompany and facilitate to perform the graduation thesis perfectly.

During implementing the thesis, due to some limitations in knowledge and

experience, it is impossible to avoid shortcomings. Therefore, I am very grateful to

receive advice and suggestions from teachers and friends to help me complete the

graduation thesis.

Once again thank you.

Ho Chi Minh City, February 25th, 2023

Student

TRUONG THI KIM NGAN

iii

ABSTRACT

Title: Impact of capital structure on performance of listed material manufacturing

enterprises in Vietnam.

Abstract: The thesis analyzes the impact of capital structure on performance of listed

material manufacturing enterprises in Vietnam. Secondary data is extracted from

annual reports, financial statements of listed material manufacturing enterprises

selected as a sample for the 10-year period from 2012 to 2021. FGLS estimation

method combined with matrix correlation is used to test the influence of the

explanatory variables on the dependent variable. The impact of capital structure

including the short-term debt ratio, the long-term debt ratio on performance is

measured by return on average assets (ROA), return on average equity (ROE), in

addition, the thesis also uses additional control variables such as enterprise size

(SIZE) and revenue growth (GROWTH). Regarding the level of influence, the

impact of short-term debt ratio on performance is a non-linear relationship. For long-

term debt, because most enterprises use a very small ratio of long-term debt, only

about 10%, the increase in long-term debt also has an impact on financial

performance but follows a linear relationship. For two control variables are firm size

and revenue growth. Both of these variables show a positive impact on financial

performance. Therefore, creating better profitability than other businesses.

Keywords: Capital structure, short-term debt ratio, long-term debt ratio.

iv

ABBREVIATIONS LIST

Abbreviations Meaning

FEM Fixed effect model

GLS Generalized least square

HoSE Ho Chi Minh Stock Exchange

Pooled OLS Pooled ordinary least square

REM Random effect model

ROA Return on asset

ROE Return on equity

WACC Weighted average cost of capital

v

TABLE OF CONTENTS

DECLARATION OF AUTHENTICITY .................................................................... i

GRATEFULNESS .................................................................................................... ii

ABSTRACT ............................................................................................................. iii

ABBREVIATIONS LIST .........................................................................................iv

LIST OF TABLES .................................................................................................. vii

LIST OF CHARTS ................................................................................................. viii

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

1.1. Introduction ................................................................................................. 1

1.2. The necessity of the research ...................................................................... 2

1.3. Research objectives ..................................................................................... 3

1.4. Research questions ...................................................................................... 3

1.5. Research subjects and scope ....................................................................... 4

1.6. Research method ......................................................................................... 4

1.7. Research structure ...................................................................................... 5

Chapter 1’s summary .................................................................................................. 6

CHAPTER 2 : THEORETICAL AND EMPIRICAL RESEARCHES ..................... 7

2.1. Theory of capital structure ......................................................................... 7

Traditional approach ........................................................................ 7 2.1.1.

M&M theory .................................................................................... 8 2.1.2.

Trade-off theory ............................................................................. 11 2.1.3.

Pecking order theory ...................................................................... 13 2.1.4.

2.2. Empirical researches ................................................................................. 15

Foreign researches .......................................................................... 15 2.2.1.

Domestic researches ....................................................................... 24 2.2.2.

Researches’ gaps ............................................................................ 28 2.2.3.

Chapter 2’s summary ................................................................................................ 29

CHAPTER 3 : RESEARCH METHOD ................................................................... 30

3.1. Research hypothesis .................................................................................. 30

3.2. Description of variables in the research model ...................................... 31

3.2.1. Dependent variable ......................................................................... 31

vi

3.2.2. Independent variables ..................................................................... 32

3.2.3. Control variables ............................................................................ 34

3.3. Theoretical research model ...................................................................... 36

3.4. Research data ............................................................................................. 38

3.5. Research method ....................................................................................... 40

3.6. Data analysis sequence .............................................................................. 42

3.6.1. Descriptive statistics ....................................................................... 42

3.6.2. Correlation analysis ........................................................................ 42

3.6.3. Regression Model Selection ........................................................... 43

3.6.4. Model’s deficiencies test ................................................................ 44

3.6.5. Model’s deficiencies remedy ......................................................... 45

Chapter 3’s summary ................................................................................................ 46

CHAPTER 4 : RESEARCH RESULTS AND DISCUSSIONS .............................. 47

4.1. Descriptive statistics .................................................................................. 47

4.2. Correlation analysis .................................................................................. 50

4.3. Regression analysis .................................................................................... 51

4.3.1. The impact of short-term debt on ROA ......................................... 52

4.3.2. The impact of long-term debt on ROA .......................................... 60

4.3.3. The impact of short-term debt on ROE .......................................... 67

4.3.4. The impact of long-term debt on ROE ........................................... 75

Chapter 4’s summary ................................................................................................ 84

CHAPTER 5 : CONCLUSION AND POLICY RECOMMENDATION ............... 85

5.1. Conclusion .................................................................................................. 85

5.2. Recommendations on capital structure for enterprises ......................... 87

5.3. Recommendations for relevant organizations ........................................ 88

5.4. Limitations of the topic ............................................................................. 89

Chapter 5’s summary ................................................................................................ 90

REFERENCES ............................................................................................................ i

APPENDIX 1 ......................................................................................................... viii

APPENDIX 2 ............................................................................................................. x

vii

LIST OF TABLES

Table 3.1: Description of variables in the research model ....................................... 36

Table 4.1: Descriptive statistics of variables ............................................................ 47

Table 4.2: Correlation between variables ................................................................. 50

Table 4.3: Correlation between variables (after adjustment) ................................... 51

Table 4.4: Model regression results of short-term debt for the ROA ...................... 52

Table 4.5: Model’s selection for short-term debt on the ROA ................................ 53

Table 4.6: VIF test result for short-term debt on the ROA ...................................... 54

Table 4.7: Heteroskedasticity test result for short-term debt on the ROA ............... 55

Table 4.8: Auto-correlation test result for short-term debt on the ROA .................. 56

Table 4.9: FGLS regression model for short-term debt on the ROA ....................... 57

Table 4.10: Regression results for long-term debt on the ROA ............................... 60

Table 4.11: Model’s selection for long-term debt on the ROA ............................... 61

Table 4.12: VIF test result for long-term debt on the ROA ..................................... 62

Table 4.13: Heteroskedasticity test result for long-term debt on the ROA .............. 63

Table 4.14: Auto-correlation test result for long-term debt on the ROA ................. 64

Table 4.15: FGLS regression model for long-term debt on the ROA ...................... 65

Table 4.16: Model regression results of short-term debt for the ROE ..................... 67

Table 4.17: Model’s selection for short-term debt on the ROE ............................... 69

Table 4.18: VIF test result for short-term debt on the ROE ..................................... 70

Table 4.19: Heteroskedasticity test result for short-term debt on the ROE ............. 70

Table 4.20: Auto-correlation test result for short-term debt on the ROE ................ 71

Table 4.21: FGLS regression model for short-term debt on the ROE ..................... 72

Table 4.22: Regression results for long-term debt on the ROE ............................... 75

Table 4.23: Model’s selection for long-term debt on the ROE ................................ 76

Table 4.24: VIF test result for long-term debt on the ROE ..................................... 77

Table 4.25: Heteroskedasticity test result for long-term debt on the ROE .............. 78

Table 4.26: Auto-correlation test result for long-term debt on the ROE ................. 79

Table 4.27: FGLS regression model for long-term debt on the ROE ...................... 80

Table 4.28: Overall comparison results .................................................................... 83

viii

LIST OF CHARTS

Chart 4.1: Average rate of return over 10 years ....................................................... 48

Chart 4.2: Average debt ratio over 10 years ............................................................. 49

1

CHAPTER 1: INTRODUCTION

1.1. Introduction

Capital structure decision is one of the essential decisions of every enterprise. One

of the key issues of capital structure is determining the optimal structure to achieve

good performance. Capital structure decisions must be made well before the

company is incorporated or when there is a capital requirement to meet the cost

needs. The CFO of a company must analyze the cost and profit factors of various

sources before choosing the best one, the optimal capital structure or the capital

structure that reduces the cost for the business. Therefore, capital structure decision

is an ongoing process and must be done whenever the company has capital

requirements for projects. The capital structure is said to be optimal once it

maximizes the market value of the firm (Chadha and Sharma, 2015).

Capital structure is the combination of debt and equity that a company uses to finance

its business (Damodaran, 2001). In the capital structure decision, the term financial

leverage is mentioned. Financial leverage is the ratio between debt and equity,

indicating the relationship between borrowed funds and owners’ equity in a firm’s

capital structure (Chadha and Sharma, 2015). The owner has a commitment to the

company in the belief that the company will grow in the near future. In contrast,

creditors do not have a solid and long-term commitment because they are more

interested in paying their debt on time. CFOs will want to invest cash in future

projects to generate better returns, while shareholders are more interested in paying

regular dividends (Chadha and Sharma, 2015). The impact of decisions on capital

structure will help enterprises to cope with the harsh competitive environment to

finance the company’s assets (Zuraidah et al., 2012). Thus, the basic objective of

capital structure optimization is to decide the proportions of debt and equity in order

to maximize corporate value, improve business performance, and at the same time

minimize the average cost of capital (Yu-Shu et al., 2010).

2

Capital structure has an influence on the performance of enterprises and has been

researched by domestic and foreign authors. Some of the foreign studies can be

mentioned are Mireku et al (2014); Ramachandran and Candasamy (2011); Olokoyo

(2013); Sheikh et al (2013); Pouraghajan et al (2012); Gill, Biger and Mathur (2011);

Margaritis and Psillaki (2010). Meanwhile in Vietnam, assessing the impact of

capital structure on the performance of enterprises is studied by many authors

Nguyen Thi Thanh Vinh (2021); Tran Thi Kim Oanh and Hoang Thi Phuong Anh

(2017); Doan Vinh Thang (2016); Quang et al (2014).

1.2. The necessity of the research

Theoretically, capital structure reflects the ratio of debt to equity that the business is

carrying. Prior studies focused on looking at factors affecting the capital structure of

enterprises, making comments on the ratio between debt capital and equity or the

ratio of short-term and long-term debt. Theoretically, researchers have shown that a

reasonable capital structure will contribute to increasing profits and increasing

enterprise value. However, what capital structure is reasonable in the economic

environment in Vietnam has not been clarified. Therefore, it is necessary to have an

overall picture of the common types of capital structures currently applied by

Vietnamese enterprises and specific to an industry.

In terms of economic practice, firstly, materials are inputs for all production

industries as well as having spillover effects, positive impacts promoting the

development of many industries, such as: manufacturing, information technology,

electronics, chemicals, high-tech industries, products for agricultural, forestry,

fishery, livestock production, etc. Secondly, the self-production for domestic

production has also contributed to reducing the import of raw materials from other

countries and optimizing investment costs in the production of some industries.

Thirdly, in general, the production capacity and quality of our country’s material

industry is still limited. The localization rate of production of materials for the

manufacturing industry is still low, such as cast iron materials (under 30%);

aluminum material, copper material (about 5%); chemicals for the plastic and rubber

3

industries still have to import up to 70%; raw materials for the textile industry have

to be imported nearly 90% of fabrics, 80% of yarns; ...

Based on the practices and recent researches, the thesis decided to choose the topic

“Impact of capital structure on performance of listed material manufacturing

enterprises in Vietnam”. This study provides both theoretical and practical

significance to know which factors in capital structure affect enterprises’

performance in order to make rational decisions about capital structure. With the

ultimate goal is to help businesses get the right strategies to maximize their value

with the minimal capital cost.

1.3. Research objectives

The thesis aims to determine the impact and degree of impact of capital structure on

the performance of listed material manufacturing enterprises in Vietnam from 2012

to 2021. In which, to achieve the above purpose, the thesis will accomplish 3 specific

objectives, (i) determine the factors in capital structure that affect the performance

of listed material manufacturing enterprises in Vietnam; (ii) examine the impact of

capital structure factors on the performance of these firms; (iii) then make

recommendations to improve the efficiency in the operation of Vietnamese

enterprises in general.

1.4. Research questions

Based on the purpose set out above, the study will make 3 corresponding questions

to solve 3 specific objectives, (i) What factors in capital structure affect performance

of listed material manufacturers in Vietnam? (ii) How do capital structure factors

affect the performance of these firms? (iii) Determine the optimal capital structure

based on the quadratic research model? (iv) What solutions will help improve

performance in business activities of Vietnamese enterprises.

4

1.5. Research subjects and scope

The object of the research are the capital structure’s factors that affect the

performance of the listed material manufacturing enterprises in Vietnam. The study

will carry out sample, based on secondary data of listed material manufacturing

enterprises on the Ho Chi Minh Stock Exchange (HoSE). Listed firms have a fairly

significant effect on economies in term of scale or business activities (Choi et al.,

2014). Using data from audited financial statements of listed companies will ensure

the reliability and transparency of the information disclosed by the company. Based

on data extracted from HoSE’s website, the total number of enterprises in the

material manufacturing industry is 67 enterprises, ranking second among the overall

of 406 listed ones, so the research sample will be guaranteed. In which, the research

period is 10 years from 2012 to 2021 that ensures the representativeness of the

selected sample to explain the research results (Chadha and Sharma, 2015; Choi et

al., 2014).

1.6. Research method

The thesis uses a combination of qualitative and quantitative research methods to

solve the initially set objectives.

Qualitative research method used in the thesis includes approaching and analyzing

the synthetic theory, combined with references and discussions of previous related

empirical studies both in domestic and foreign countries. Then, the thesis designs the

research model and interprets variables, giving hypothesis for each independent

variable. Finally, the study discusses the research results, gives relevant suggestions

and recommendations to the subjects.

Quantitative research method is to examine the extent of impact of capital structure’s

factors on performance of enterprises. It includes specific technical methods such as

descriptive statistics, correlation analysis and panel data regression. Then, the author

performs regression analysis in the order of OLS model, FEM model and REM

model. Hausman test is used to select the model that fits the data. Finally, when a

5

suitable model has been chosen, the thesis will test the defects of the model,

including (i) the multicollinearity, (ii) the heteroskedasticity, (iii) the auto-

correlation.

1.7. Research structure

The thesis will focus on the following 5 detailed chapters. Chapter 1 is the

introduction. In this chapter, the thesis will present the reasons for choosing the topic.

After that, the thesis sets out specific goals and objectives to be solved, meanwhile,

the research questions corresponding are set out. Next is about the research object

and scope and research methods. And finally the layout of the study. Chapter 2 is

about theoretical research and experimental research. Firstly, the study performs a

synthesis of theories of capital structure such as Modigliani Miller theory, pecking

order theory, trade-off theory and examines relevant empirical studies to find

research gaps and meanwhile gives out the factors in capital structure that will affect

the performance of the business. Chapter 3 presents the research methodology. The

thesis explains in detail the method used to solve the research questions. After that,

the thesis will mention the method used to collect the data. From the factors obtained

in Chapter 2, the thesis sets out research hypotheses and builds a theoretical model

to test the impact of the set factors. Chapter 4 is about the research results obtained

from model regression, which will include the descriptive statistics, correlation

analysis, regression analysis and the results’ explanations. And finally, Chapter 5 is

the conclusion and recommendations. The thesis will make general conclusions

about the achieved results and suggest appropriate recommendations and policies to

build capital structure to improve the financial efficiency of Vietnamese enterprises

in general.

6

Chapter 1’s summary

In the first chapter, the thesis stated the necessity of the topic as well as clearly

defined the objectives, objects and scope of the research. Besides, some contributions

of the thesis on capital structure of Vietnamese enterprises are also clearly shown in

this chapter.

7

CHAPTER 2: THEORETICAL AND EMPIRICAL

RESEARCHES

2.1. Theory of capital structure

Information about the financing for business operations, the enterprise’s capital source

is presented on the balance sheet and explained in detail in the financial statements’

notes, including two groups are liabilities and equity, sorted by increasing stability.

Liabilities reflect capital sources formed by enterprises borrowing, purchasing goods

and services on credit from suppliers, accumulated debts (unpaid taxes to the State,

salary and unpaid social insurance to employees). Liabilities are divided into two groups

based on the repayment period, including short-term liabilities - with a maturity date of

1 year or 1 business cycle, and long-term liabilities – with a maturity is over 1 year or 1

business cycle. Unlike debt, equity is formed by the contributions of owners or

accumulated from business results during the operation. Equity is a source of financing

characterized by dynamic, non-repayable and high stability. To reflect the combined

relationship between different sources of financing, company often emphasizes the

relationship between debt and equity for the entire invested assets, thereby finding out

how the company’s management has decided to sponsor, and whether there will be

positive or negative effects on the goals of the funding decision.

2.1.1. Traditional approach

The traditional approach also known as the theory of optimal capital structure, indicates

that exists an optimal capital structure. The optimal capital structure is the combination

of debt and equity at a certain ratio to minimize the average cost of capital (WACC),

hence maximizing the business’s value and the shares’ price.

Some assumptions from the traditional approach:

− Financial markets are not perfect, businesses and investors cannot borrow at the same

interest rate.

− Enterprises operate in an environment with corporate income tax.

− Enterprises are at risk of falling into financial distress due to the use of debt. However,

when utilization increases but debt ratio remains low, shareholders and creditors can

8

either ignore the risk and either not increase the required rate of return, or increase it

but not significantly.

The traditional approach explains that firms can reduce the average cost of capital by

using debt, because the cost of debt is lower than the cost of equity. The reason for

optimal capital structure theory indicates that the cost of debt is lower than the cost of

equity is due to the tax savings from interest. When firm uses debt, the interest will

reduce the income tax payable. Meanwhile, the profits that the business distributes to

the shareholders do not generate tax savings. Thus, in order to reduce the average cost

of capital, enterprises should use debt. However, the use of debt will increase risks for

creditors and shareholders, so the average cost of capital will only decrease when the

level of debt use is still within the limit so that creditors and shareholders do not increase

the required rate of return or insignificant increase. In summary, the average cost of

capital will change as the capital structure changes, the optimal capital structure must

include debt and equity in an appropriate ratio so that the average cost of capital is

lowest. .

Thus, the traditional approach admits and provides arguments to prove the existence of

an optimal capital structure for the firm, which is because (1) the cost of debt is lower

than the cost of equity, (2) both of these costs will increase with an increase in debt

utilization. Although the traditional approach has been supported by some famous

financial experts such as Ezra Solomon and Fred Westo, it has not been convinced by

other researchers, and has even been rejected. As in reality, there is not enough certain

bases to confirm that when increasing the level of debt use, the cost of debt and the cost

of equity will increase.

2.1.2. M&M theory

The Modigliani and Miller Theory (Modigliani Miller Theory) often referred as M&M

theory is the work of two researchers Franco Modigliani and Merton Miller, published

in 1958. The basic assumptions of M&M theory include (i) no tax; (ii) no transaction

costs; (iii) no financial distress costs; (iv) both investors and businesses can borrow at

the same interest rate; (v) the information is available to all investors and no charge for

9

the information; (vi) all investors have as much information as the company’s

management about the investment opportunities; (vi) companies with comparable

business risks, operating in similar environments; (vii) the company distributes 100% of

profits to shareholders, so the growth rate is zero. M&M theory states that the average

cost of capital and firm value are independent of capital structure. However, an

important and prominent point that increases the persuasiveness of M&M theory is that

the two authors provide behaviour and technical evidence to explain why the average

cost of capital and firm value do not change, when the degree of financial leverage is

changed. The reason is due to arbitrage activities.

Arbitrage activities

Arbitrage is the process of buying securities in a low-priced market and then selling

them in a higher-priced market, thereby restoring equilibrium to the market price of a

security. M&M theory illustrates that due to the existence of arbitrage, the value of firms

with different capital structure will the same. This is explained by M&M if there exist

two companies with different values due to different capital structures, then investors

will buy the shares of the low priced company and sell the higher ones. This action will

increase the stock price of the low value company and decrease the share price of the

high value firm. This buying and selling process is continued until the market prices of

the two stocks are equal.

The value reservation rules

Starting from the view that the value of the business does not depend on the capital

structure and how the earning before interest and tax is divided among creditors and

shareholders, M&M believes that changing capital structure only transfers value from

shareholders to creditors or vice versa without changing firm value. According to M&M,

the value of a business is governed by two factors (i) its operating income and (ii) the

level of business risk associated with the above operating income. In a perfect capital

market with no corporate taxes, no bankruptcy costs, a firm’s value is unaffected by its

capital structure. In other words, the value of the firm without debt and the value of the

firm using debt is the same. Thus, companies cannot increase its value depending on the

change of capital structure, so enterprises cannot find the optimal capital structure.

10

Financial leverage and the cost of capital

Since the expected annual operation return and the market value of the debt and non-

debt firm are the same, the weighted average cost of capital of the two firms will also

be the same. For equity, financial leverage increases expected earning per share, but also

increases the standard deviation and variation’s coefficient of earning per share. Thus,

financial leverage increases the risk of equity. As risk increases, the required rate of

return on equity must increase. In M&M’s view, in the absence of taxes, the required

rate of return on equity of the company using debt, is equal to the one using no debt,

plus risk premium. In other words, the required return on equity is positively related to

the level of financial leverage.

M&M theory in tax environment

The original M&M theory was built with the assumption that the business operates in a

tax-free environment and some other assumptions. Based on these assumptions M&M

concludes that firm value and average cost of capital are independent of capital structure

as discussed above. The assumption that businesses do not have to pay taxes is

unrealistic, because in reality most businesses have to pay income tax to the State. Thus,

the assumption of no taxes reduces the validity of the M&M theory. Therefore, in the

next section, the author will discuss M&M theory in an environment with corporate

income tax, the remaining assumptions are unchanged. In tax-free terms, the earning

before interest and tax belongs to its creditors and shareholders. If there are taxes,

earning before interest and tax will be divided into three groups: creditors, the State and

shareholders. Tax law allows interest to be included in a deductible expense when

calculating income tax. Therefore, interest helps businesses create a tax shield or create

an interest tax savings. Thanks to this savings, the use of debt will increase the return to

creditors and shareholders. M&M theory under tax conditions also has two postulates,

in which the postulate I deals with enterprise value and the postulate II deals with the

capital cost.

Postulate I – Enterprise’s value

In tax condition, the value of the leveraged firm is greater than the value of the

unleveraged firm the amount of the present value of the tax savings from the interest.

11

Without using debt, the operating profit will belong to the State and shareholders. If debt

is used, the operating profit will belong to creditors, the State and shareholders. When

there is the participation of the State in the distribution of profits, the profit brings to

both creditors and shareholders of the company that uses debt is greater than the one

that does not use debt is the amount of tax savings from interest.

Postulate II - Cost of capital

As the level of debt utilization increases as to tax and interest savings, the real cost of

debt decreases, so the average cost of capital decreases of debt-using firms will be lower

than the non-debt firms. As debt utilization increases, financial risk increases, so the

cost of equity increases. This is stated by poposition II that iIn an environment with no

taxes and no bankruptcy costs, the required rate of return or the cost of equity will

fluctuate with the degree of financial leverage but taking into account the impact of the

tax shield.

Thus, M&M theory has given a new perspective in analyzing and explaining the

relationship between capital structure and firm value and cost of capital. M&M theory

applies arbitrage as an evidence to explain investor behavior, thereby drawing the

following conclusions that in the absence of taxes, enterprise value and weighted cost

of capital is independent of capital structure. However, the next section M&M removes

the assumption of no taxes and concludes that firm value increases when using debt

thanks to tax shield from interest; the weighted cost of capital of firm that uses debt is

lower than the one that does not use the debt; the cost of equity increases with an increase

in the use of debt. Although M&M theory offers many different perspectives on the

impact of capital structure on firm value and cost of capital, it is built on many unrealistic

assumptions. Therefore, unrealistic assumptions need to be removed for more accurate

and convincing conclusions. In the next section, the author will consider capital structure

in terms of financial distress costs, agency costs and information asymmetric.

2.1.3. Trade-off theory

The trade-off theory of capital structure proposed by many researchers is to explain the

phenomenon in practice, that firms only use debt to a certain limit, while the M&M

12

theory suggests that the higher the firm value, the higher the level of debt usage. To

explain this phenomenon, researchers believe that the M&M theory is based on the

unrealistic assumption that there is no financial distress that generates financial distress

costs due to the use of debt. In 1973, two researchers Alan Kraus and Robert H.

Litzenberger concluded that firms with optimal financial leverage reflect the trade-off

between the benefit of the tax shield from interest and the costs of bankruptcy. In 1984,

Stewart C. Myers stated that firms following the trade-off approach would establish a

target debt ratio based on the balance between the benefits of the tax shield and the costs

of bankruptcy and gradually adjust the capital structure towards that goal. Thus,

according to the trade-off theory of capital structure, firms using debt will benefit from

the tax shield of interest, but using debt incurs additional costs, especially the cost of

financial distress. . Therefore, businesses need to use both debt and equity to balance

the benefits of tax shields and bankruptcy costs, the value of the business will be highest,

and that is the optimal capital structure.

Financial distress occurs when a business is unable to fulfill its promises to creditors, or

is able to do so but very difficult. Financial distress can be temporary, leading to a

number of problems for operations such as feasible projects are delayed or canceled,

labor productivity decreases, creditors do not continue to lend, suppliers tighten their

credit policy... But sometimes financial distress will lead to bankruptcy, in which case

enterprises have to spend large amount of money on the lawyers, courts, auditors,

business managers, etc. Thus, in both cases, financial distress causes serious

consequence to the business, and investors assume that a business uses debt can fall into

financial distress, which is a factor that reduces firm value. The cost of financial distress

depends on the likelihood of financial distress and the magnitude of the costs involved.

Financial distress costs fall into two categories are direct costs and indirect costs. Direct

costs include decrease in asset value due to liquidation, decrease in selling price,

increase in legal and administrative costs, some companies’ managers can reduce

investment in research and development, market research and some other investments.

Indirect costs include the company’s reputation and reputation decline, possible loss of

customers, suppliers of capital require higher returns leading to an increase in the cost

of capital, suppliers set out stricter credit standards and tightening terms of sale or

13

incurring losses due to competitive pressure. When the level of debt utilization is low,

the risk of financial distress is still negligible, the present value of the financial distress

costs are lower than the present value of the tax shield from interest, so the firm’s value

increases with the level of debt. However, when the level of debt utilization exceeds the

rapid increase of financial distress costs, when the present value of the marginal cost of

financial distress equals the present value of the tax savings, the firm value is highest

where the capital structure is optimal. If the level of debt usage continues to increase,

the value of the enterprise will decrease.

Thus, the trade-off theory of capital structure indicates that firms must consider the

benefits from the tax shield and the costs of financial distress to choose the one with the

highest firm value. However, this theory also has limitations that (1) it is difficult to

accurately determine the costs of financial distress, especially indirect costs such as loss

of customers, loss of suppliers, loss of reputation and reputation; (2) in practice many

large and successful firms still use debt much lower than the optimal debt ratio

determined by theory.

2.1.4. Pecking order theory

The pecking order theory of capital structure was first proposed by Gordon Donaldson

in 1961 and revised by Stewart C. Myers and Nicolas Majluf in 1984. The pecking order

theory does not address the existence of whether there is optimal capital structure as a

starting point in the approach, but this theory asserts that firms prefer to use internal

financing such as retained earnings or excess liquid assets, rather than external funding.

If internal capital is not enough to finance investment opportunities, firms can choose

external sources in the direction of minimizing increasing costs due to asymmetric

information. In order to minimize the cost of external sources, businesses prefer to use

debt, then to issue preferred shares, and finally to issue common shares. Thus, the basic

content of the pecking order theory provides managers with the priority order when

deciding to choose funding sources, specifically businesses will choose first are internal

funding sources (mainly retained earnings) to finance operations and investment in new

projects, if internal capital is not sufficient, external funding is mobilized in which debt

is preferred first of all and then direct owners funding. This prioritization reflects the

14

goals of financial managers who want to ensure control for existing owners, to reduce

agency costs of new equity, and to avoid negative feedback from market if new shares

are issued.

The two basic assumptions of pecking order theory include (i) corporate managers

acting in the interest of current owners; (ii) information asymmetry exists between

managers and investors. The pecking order theory states that managers know more

information about a firm’s prospects, risks, and values than outside investors, and this

information asymmetry affects decision between internal and external financing, debt

and equity financing. Internal funding is encouraged as the first choice, because the

company does not have to disclose information about potential investment opportunities

and the expected return from investment opportunities to shareholders or owners. debt,

except for the disclosure of dividend payments. On the other hand, choosing internal

capital sources also ensures control for existing shareholders, ownership rights to assets

are not dispersed.

The cost of the internal sources from retained earnings is less than the cost of newly

issued shares. Practical research shows that businesses in the service sector where

products and services need to be kept confidential to avoid copying, such as computer

software technology businesses prefer to retain profit to form internal capital, limiting

mobilization of external sources. For debt, this is the first preferred source of external

financing over equity financing, since the cost of debt is lower than the cost of preferred

equity and common stock. On the other hand, like retaining earnings, using debt does

not make the existing owners’ control dispersed, since creditors have no right to

participate in the management of the business. Although creditors need information

about the business’s credit risk and managers are responsible for providing complete

and reliable news to them. On the other hand, the existence of asymmetric information

also favors the use of debt rather than equity from owners. As the company’s managers

use debt instead of issuing new shares, considered by investors as a positive signal, show

that the company has many investment opportunities that are expected to bring high

returns and the company’s managers want to take advantage of debt to increase profits

for existing shareholders, contributing to ensuring the goal of extending corporate’s

15

value. Raising fund from owners through the issue of additional shares is the last option,

because this source has a higher cost than debt. In addition, the issuance of new shares

is considered by investors as a negative signal, indicating that the company’s prospects

are not good, and it is likely that earnings will be reduced in the future. Therefore,

issuing new shares will reduce the share price compared to the current level, negatively

affecting the asset value of existing shareholders, so this is the last choice.

Thus, the pecking order theory does not provide a model for determining the optimal

capital structure for businesses, but rather guides the priority order of choosing funding

sources for operation activities or future financing projects. This theory has clarified the

financing choices of managers, as well as explained the market’s reaction to the firm’s

decision to raise fund from outside. However, the pecking order theory does not consider

the effects of income taxes and financial distress, ignore the negative problems that can

arise if the managers keep too much cash… Thus, pecking order theory cannot replace

other theories but supplement and clarify financing decisions of corporate managers.

2.2. Empirical researches

Research on the impact of capital structure on the company’s performance is a topic

raising interest and attention of both domestic and foreign authors. The thesis will make

an overview of the relevant reseaches in order to provide research gaps and direction for

the topic.

2.2.1. Foreign researches

Sedeaq Nassar (2021) examined the impact of capital structure on performance of

companies listed on the Palestine Stock Exchange (PEX). The study uses data of 32

companies listed on the Palestine Exchange (PEX) for the period 2015 - 2019 with about

160 observations. The data is collected from the financial statements of the companies

published on the PEX’s website. The independent variables used to measure capital

structure include the ratio of short-term debt to total assets (STDTA), long-term debt to

total assets (LTDTA), total debt to total assets (TDTA). Performance is measured by

return on assets (ROA), return on equity (ROE), return on equity (ROI). In addition, the

study uses firm size (SIZE) as the control variable.

16

(2.1) ROA = 𝛽0 + 𝛽1CAPITAL + 𝛽2SIZE + 𝜀

(2.2) ROE = 𝛽0 + 𝛽1CAPITAL + 𝛽2SIZE + 𝜀

The results show that the capital structure of the company has a negative and statistically

significant relationship with performance measured by ROA, which means the use of

high debt will negatively affect the ROA of the company. On the other hand, the results

show that the capital structure of the company has a positive and statistically significant

relationship with the return on equity (ROE) and the return of investment (ROI).

Berzkalne (2015) studied the capital structure and enterprises’ profitability through

array data analysis. The author studies on the non-linear relationship between capital

structure and profitability by using threshold regression analysis with a sample of 58

listed companies in the Baltic, including 22 enterprises from the Baltic region, the rest

are businesses in the vicinity, the research period is from 2005 to 2013. The study uses

debt ratio as the ratio of total debt/total equity to represent capital structure and selects

stock prices to be the best indicator of a company’s profitability. The authors conclude

that there is a non-linear relationship between capital structure and profitability. For

Baltic listed companies with small market capitalization, an increase in leverage will

increase the company’s profitability and reach its peak value if the debt ratio reaches the

rate of 24.64%

Mireku et al (2014) established the relationship between capital structure measures and

performance in order to determine which capital structure measure has a stronger

relationship with performance. Two definitions of capital structure measures (book

value and market value) and six measures of performance were used in the study. The

sample in this study was 15 companies on the Ghana Stock Exchange (GSE) that had

been selected over a 6-year period (2002–2007). The results showed that the capital

structure of firms affects their performance. Many indicators showing the company's

performance were negatively correlated with financial leverage. That means, Ghana

companies with less debt would have high profit margins and good performance. The

(2.3) ROI = 𝛽0 + 𝛽1CAPITAL + 𝛽2SIZE + 𝜀

17

findings of the study showed that firms rely more on short-term debt than on long-term

debt. This was probably due to the absence of a well-developed bond market in Ghana,

where companies could raise long-term debt on demand. Performance measures,

especially profitability, have a negative relationship with financial leverage. Firms with

high profitability and good performance in Ghana had less debt and were more

dependent on internal sources of financing according to pecking order theory. This study

concluded that the market value of capital structure should be considered more when

assessing capital structure because it was more closely linked to financial performance

than book value. The author recommended future empirical research on other factors

affecting the performance of companies in Ghana besides debt policy and a deeper

investigation into capital structure trends and dependence on short-term liabilities that

could be taken.

Olokoyo’s study (2013) presented empirical findings on the impact of leverage (debt

ratio) on the performance of firms. The results were based on data from 2003 to 2007 of

101 listed companies in Nigeria along with pecking order theory and static trade-off

theory of capital structure. The study used panel data analysis using fixed-effect

estimation, random-effect estimation and regression model. Hausman's Chi-square and

conventional tests were used to check whether the fixed-effects model is a suitable

alternative to the random-effects model when calculating for each model. In the research

model, the author used the ratio of return on total assets (ROA), return on equity (ROE)

and the ratio between market value to book value (Tobin's Q) to measure performance.

In which, the variables used to explain the relationship between capital structure and

performance included the ratio of total debt to total assets (TD/TA), the ratio of long-

term debt to total assets (LD/ TA), the ratio of short-term debt to total assets (SD/TA)

and firm size (S). Financial leverage had a negative effect on a company’s performance

measure (ROA).

(2.4) ROAit = 𝛼i + 𝛽1.TD/TAit + 𝛽2.LD/TAit + 𝛽3.SD/TAit + 𝛽4.Sit + 𝜇it

(2.5) ROEit = 𝛼i + 𝛽1.TD/TAit + 𝛽2.LD/TAit + 𝛽3.SD/TAit + 𝛽4.Sit + 𝜇it

18

All measures of leverage had a positive and significant relationship with the market

performance measure (Tobin's Q). Nigerian companies were financed primarily with

equity or a combination of equity and short-term funding. A major difference between

the capital structure of Nigerian firms and those in developed economies was that

Nigerian firms tended to use shorter-term financing and significantly less long-term

debt. This showed that Nigerian companies mainly relied on short-term sources rather

than long-term sources. The difference in long-term versus short-term debt, to some

extent, limited the validity of capital structure theories in Nigeria. This suggested the

theoretical basis remained largely unresolved.

Sheikh et al. (2013) conducted a study on the effect of capital structure on the

performance of non-financial companies listed on the Karachi Stock Exchange in

Pakistan in the period 2004-2009. With three econometric tools, namely the pooled

ordinary least squares method, fixed effect estimation and random effect estimation,

were used to estimate the relationship between capital structure and firm’s performance.

The model used return on assets (ROA) and market-to-book ratio (MBR) to measure

performance. In which, the independent variable used to represent capital structure

included total debt ratio (TDR), long-term debt ratio (LDR) and short-term debt ratio

(SDR), and the control variable included firm size (SIZE), the ratio of tangible assets

(ATNG) and growth opportunities (GROW).

(2.6) TobQit = 𝛼i + 𝛽1.TD/TAit + 𝛽2.LD/TAit + 𝛽3.SD/TAit + 𝛽4.Sit + 𝜇it

(2.7) ROAit = 𝛽0 + 𝛽1.CAPit + 𝛽2.SIZEit + 𝛽3.ATNGit + 𝛽4.GROWit + 𝜀it

Empirical results showed that all capital structure measures (total debt ratio, long-term

debt ratio and short-term debt ratio) were negatively correlated with return on assets in

all regressions. In addition, the total debt ratio and the long-term debt ratio both had a

negative relationship with the market value to book ratio according to the OLS model,

while these measures positively affect the market-to-book ratio under the fixed-effects

model. Short-term debt ratio was positively related to market-to-book ratio in all

(2.8) MBRit = 𝛽0 + 𝛽1.TDRit + 𝛽2.SIZEit + 𝛽3.ATNGit + 𝛽4.GROWit + 𝜀it

19

regression methods, however, the relationship was not statistically significant. As far as

the control variables were used, the ratio of tangible assets had a negative effect, while

firm size and growth rate were positively related to firm’s performance. The study

concluded that the relationship between capital structure and firm’s performance was

inverse. Furthermore, this finding was inconsistent with the proposition made by

Modigliani and Miller (1958). Furthermore, the negative correlation between capital

structure and performance may lead companies to adopt more than appropriate

proportions in their capital structure. This excessive disparity could increase the

negative impact on performance, thereby limit the ability to manage operations

effectively.

The study by Pouraghajan et al. (2012) examined the impact of capital structure on the

performance of companies listed on the Tehran Stock Exchange. For this purpose, the

authors researched and experimented with 80 companies listed on the Tehran Stock

Exchange and selected among 12 industry groups for 5 years from 2006 to 2010. In the

study, return on assets (ROA) and return on equity (ROE) were used to measure the

performance of companies. For capital structure, the author used the debt ratio (DR) to

represent and the control variables were used in the article included total assets turnover

ratio (TURN), firm size (SIZE), age (AGE), tangible asset ratio (TANG), growth

opportunity (GROW). In addition, the study also used a dummy variable of business

industry (IND) to explain the difference between enterprises.

ROAi,t = a0 + b1.DRi,t + b2.TURNi,t + b3.SIZEi,t + b4.AGEi,t + b5.TANGi,t (2.9) + b6.GROWi,t + μe

ROEi,t = a0 + b1.DRi,t + b2.TURNii,t + b3.SIZEi,t + b4.AGEi,t + b5.TANGi,t (2.10) + b6.GROWi,t + μe

ROAi,t = a0 + b1.DRi,t + b2.TURNi,t + b3.SIZEi,t + b4.AGEi,t + b5.TANGi,t (2.11) + b6.GROWi,t b7.IND + μe

20

The results indicated that there was a significant and negative relationship between the

debt ratio and the performance measure of Iranian firms (ROA and ROE). The debt ratio

would help determine the financial health because it represented the risk ratio for the

company. Research indicated that a company with a high debt ratio would have a

negative impact on operating performance and corporate value. With the average debt

ratio higher than 65%, the company’s performance would be affected. In particular,

when Iranian companies reduced their debt ratio, it could increase profits and improve

ROA and ROE. In addition, the results showed that there was a positive and statistically

significant relationship between the total assets turnover ratio, firm’s size, tangible

assets ratio and growth opportunities with other performance measures (ROA and ROE).

In which, the effect of company’s age (operating history) on performance was not

statistically significant. As a result, the long operating history of Iranian companies did

not affect their performance. Meanwhile, the results of the model showed that firms in

the other non-metallic mining; food and beverage; base metals; auto parts and

manufacturing sectors had a negative influence, and companies in the chemical products

and materials industry had a negative impact on performance.

Research by Gill et al. (2011) showed the relationship between capital structure and

profitability as two important factors because improving profitability is necessary for

long-term viability of the enterprise. This paper examined the effect of capital structure

on the profitability of manufacturing and service firms in the United States. The author

conducted a sample set of 272 companies listed on the New York Stock Exchange

during the 3-year period from 2005 to 2007. Correlation and regression analysis were

used to estimate the relationship between profitability (as measured by return on equity)

with a measure of capital structure. The study uses book value to measure variables,

with return on equity (ROE) as a measure of profitability, in which the independent

variable represents capital structure including the ratio of short-term debt to total assets

(SDA), the ratio of long-term debt to total assets (LDA) and the ratio of total debt to

total assets (DA). In addition, the study further used the firm’s size (SIZE) measured by

ROEi,t = a0 + b1.DRi,t + b2.TURNii,t + b3.SIZEi,t + b4.AGEi,t + b5.TANGi,t (2.12) + b6.GROWi,t + b7.IND + μe

21

taking the company’s revenue logarithm, the revenue growth (SG) measured by

calculating the revenue growth rate and the industry variable with the convention that 1

is the manufacturing industry and 0 is the remaining industries.

The findings of this paper showed a positive correlation between firstly short-term debt

to total assets and profitability; secondly long-term debt to total assets; and thirdly total

debt to total assets and profitability in the manufacturing industry. This article provides

useful insights for owners, operators and lenders based on empirical evidence.

Ramachandran et al. (2011) analyzed how capital structure affects the profitability of IT

companies in India. The study established a hypothetical relationship on the degree of

influence of capital structure on firm’s revenue and tested the relationship between

capital structure and profitability. The author selected the sample based on two criteria

were revenue and size. Small, medium and large companies by revenue and small,

medium and large companies by size. The study used Multi - stage sampling technique

to select a sample of 102 companies. Data for an 8-year period from 1999–2000 to 2006–

2007 were collected and reviewed for analysis. Regression analysis method used to

analyze the impact of capital structure on profitability is OLS regression, Pearson

correlation coefficient is used to analyze the relationship between capital structure and

profitability, multicollinearity test among independent variables, in addition, descriptive

statistics such as mean, standard deviation and proportion were used. To measure

profitability, the author uses ROA and ROCE criteria. Capital structure was measured

through independent variables including debt to total assets ratio (TD_TA), debt to

equity ratio (DER), expense to income ratio (EXP_INC), current asset ratio (CA). In

which, the author used the expense to income ratio (EXP_INC) as a control variable.

(2.13) ROEi,t = b0 + b1.CAP + b3.SIZE + b3.SG + 𝜇i,t

ROAe = a + b1.EXP_INC + b2.TD_TA + b3.CR + b4.DER + e (2.14)

Regarding the research results, based on business revenue, (i) low-income companies

with low costs would have high profits but profits are independent to debt ratio in capital

ROCEe = a + b1.EXP_INC + b2.TD_TA + b3.CR + b4.DER + e (2.15)

22

structure. Therefore, profitability represented by ROCE had a negative relationship with

costs but was independent to the capital structure of firms with low turnover. (ii) A mid-

income company that generates substantial revenue with low debt. The profitability of

this company was significantly affected by its capital structure. The higher the debt ratio

in the capital structure, the lower the profitability for the mid-income company. (iii) The

higher use of debt in the capital structure would negatively impact on profitability

through the use of assets in this group. An increase in the debt to assets ratio would

reduce the profitability represented by ROCE while increasing costs and the use of

current assets.

For the size factor, (i) the profitability of a small-size company would be negatively

affected by the increase in costs and in total debt to total assets. Capital structure had

the only effect on profitability when there was a significant negative effect of total costs

on profitability. From the regression results, profitability (measured by ROCE) of this

group of companies was negatively affected by the use of debt in capital structure. (ii)

A medium-size company with a net profit of 10% of total assets and used capital, had a

smaller proportion of debt. Therefore, increasing the use of debt would reduce

profitability. (iii) Large-size companies do not depend much on the use of debt in their

capital structure. These companies had a higher yield than possible profits without debt.

The research results show that the more debt was used, the lower the net profit measured

by the total assets of large-size companies in India.

For the revenue factor, research showed that an increase in total debt relative to total

assets would tend to decrease net income relative to capital employed, when there is an

increase in total costs and an in the use of short-term assets in the high revenue group.

As for the size factor, the study showed that small-size companies were not efficient in

generating revenue. Profitability is negatively related by an increase in total costs and

in the ratio of total debt to total assets respectively. Capital structure has a unique and

significant effect on profitability when there is a negative effect of total cost on

profitability for small-size firms. The study demonstrated that there was a strong 1-1

relationship between capital structure and profitability, return on assets (ROA) and

return on capital employed (ROCE). Capital structure had a significant effect on

23

profitability and increasing use of debt in capital structure would tend to reduce the net

profit of companies listed on Bombay Stock Exchange in India.

Margaritis & Psillaki (2010) studied the relationship between capital structure,

ownership structure and performance using a sample of manufacturing firms in France.

The authors use the data envelopment analysis (DEA) method to measure a company’s

performance. The study examines whether increasing the debt ratio in the capital

structure helps to increase the efficiency of enterprises. The study tests the negative

relationship between performance and capital structure based on two hypotheses:

efficiency-risk hypothesis and franchise-value hypothesis. The results show a positive

impact of financial leverage on the performance of the business. However, the

dependent variable in the model does not use the profitability of the business, but rather

the performance as measured by the DEA method. In addition, the independent variable

representing capital structure in the model is only total debt to total assets, does not

mention the short-term debt ratio and long-term debt ratio.

2

+ a3Zi,t-1 + ui,t

(2.16)

EFFi,t = a0 + a1LEVi,t-1 + a2LEVi,t−1

Ebaid et al. (2009) studied the impact of capital structure choices on the performance of

listed companies in Egypt, one of the emerging or transitioning economies. Three

reasons that the author did the research in Egypt, firstly, Egypt had moved to capitalism

and open markets, decision making was still limited by the old management of

government support about the financial leverage of companies, especially public

companies before 1990, which were partially or fully private. Secondly, the capital

market in Egypt was not yet complete, so the level of asymmetric information would be

more than in developed countries. And thirdly, the capital market in Egypt was still the

equity market, the debt market structure was still young, leading to inefficient and

inaccurate financial decisions. The dataset included 64 non-financial enterprises

selected from 10 industries. These enterprises were listed on the Egyptian stock

exchange in the period 1997-2005. The study was based on sample data of listed

companies in Egypt and used three of the performance measures return on assets (ROA),

return on equity (ROE) and gross profit margin (GM). In which, the independent

24

variables to measure capital structure included the ratio of short-term debt to total assets,

the ratio of long-term debt to total assets, the ratio of total debt to total assets together

with the control variable was the firm’s size used by to build the research model.

(2.17)

ROAi,t = 𝛽0 + 𝛽1.STDi,t + 𝛽2.logSi,t + eii,t

(2.18)

ROEi,t = 𝛽0 + 𝛽1.LTDi,t + 𝛽2.logSi,t + eii,t

(2.19)

GMi,t = 𝛽0 + 𝛽1.TTDi,t + 𝛽2.logSi,t + eii,t

The empirical test results showed that capital structure (especially short-term debt and

total debt) had a negative impact on the performance as measured by ROA. On the other

hand, capital structure (short-term debt, long-term debt and total debt) had no impact on

performance as measured by ROE or measured by GM. Therefore, the choice of capital

structure, in general, had a weak to zero effect on the performance of listed companies

in Egypt. The direction for further research is to take into account the determinants of

capital structure of Egyptian enterprises such as size, growth rate, business risk and

compared with the results achieved in developed markets. The relationship between

financial leverage and the value of Egyptian firms also needed empirical testing.

Subsequent studies may examine the relationship between the debt maturity structure

and its decisions and performance. Finally, future researches may consider deeper firstly

the impact of capital structure and ownership structure on firm’s performance because

a large number of Egyptian companies were family companies. Secondly is to include

the variables of business size, development opportunities, business risks, etc. in the

capital structure components to compare with developed countries, and thirdly, to test

the relationship between debt maturity structure and performance.

2.2.2. Domestic researches

Tran Thi Kim Oanh and Hoang Thi Phuong Anh (2017) studied the impact of capital

structure on performance by analyzing data of 60 processing industry enterprises listed

on the Stock Exchange Vietnam in the period 2009-2015. With array data processing

and percentile regression techniques, the research showed that performance was affected

by capital structure, corporate income tax, business risk and solvency with different

25

levels on different percentiles. At the higher percentile, with all other factors constant,

capital structure had a strong impact on performance. Conversely, the lower the

percentile, the greater the decline in business activity, and the higher the solvency,

corporate income tax, and business risk, the lower the enterprise’s performance.

Doan Vinh Thang (2016) examined the effect of capital structure on profitability of

2,888 state-owned joint stock companies in Vietnam. Estimation results by OLS

regression model show that capital structure and profitability, represented by ROA and

ROE ratios, have an inverted U-shaped relationship. In addition, this study also shows

the interactive influence of State ownership on the relationship between capital structure

and performance, in which, the effect of debt to total assets ratio (capital structure) on

profitability is even weaker in the case of state-dominated enterprises. However, this

study has not taken into account the influence of short-term debt and long-term debt on

the profitability, therefore, the trade-off between these two items has not been explained.

Le Thi Phuong Vy (2015) used unbalanced panel data from all listed non-financial

institutions in Vietnam for the period 2007–2012 and applied the pooled ordinary least

squares method (OLS), random effects estimation (REM), fixed effects estimation

(FEM) and a dynamic panel generalized method of moments (GMM) for data analysis.

The author had taken many different approaches and all gave consistent results.

Specifically, the study showed that while the foreign ownership variable had a negative

effect on financial leverage, the state ownership variable had a positive effect. Although

there was a positive relationship between manager ownership, the effect of owning a

high debt ratio was not conclusive. In addition, the results showed that foreign

ownership affected ownership inside the company, which would have an impact on

financial decisions. In particular, the foreign ownership ratio would reduce the positive

effect of the ownership’s level on the debt ratio of managers.

PERit = 𝛽0 + 𝛽1.TLEVit + 𝛽2.MTLEVit + 𝛽3.GROit + 𝛽4.INVit + 𝛽5.LIQit (2.20) + 𝛽6.RISKit + 𝛽7.DIVit + 𝛽8.CFit + 𝜀it

26

2 +

PERit = 𝛽0 + 𝛽1.LLEVit + 𝛽2.MLLEVit + 𝛽3.GROit + 𝛽4.INVit + 𝛽5.LIQit (2.21) + 𝛽6.RISKit + 𝛽7.DIVit + 𝛽8.CFit + 𝜀it

2 + 𝛽3.SLEVit + 𝛽4.SLEVit

PERit = 𝛽0 + 𝛽1.TLEVit + 𝛽2.TLEVit

2 + 𝛽7.GROit + 𝛽8.INVit + 𝛽9.LIQit +

(2.22) 𝛽5.LLEVit + 𝛽6.LLEVit

The research results also showed that the ratios of long-term debt, short-term debt and

total debt at book value and market value were all statistically significant and had a

negative relationship with the return on assets, return on equity and Tobin's Q. The non-

linear relationship between financial leverage and financial performance of an enterprise

was evident only when performance was measured by return on equity and capital

structure was measured by total debt ratio and short-term debt ratio. The study also

pointed to some limitations for future studies. First, the duration of the study sample was

relatively short. The observation period lasted only 6 years from 2007 to 2012, which

may have influenced the significance of the results. Second, the author only tested

experimentally in Vietnam, still not guaranteeing the convincingness of the research

results. Finally, although a number of different methods, including OLS, REM, FEM

and GMM had been applied in the study, there were some defects of the model such as

heteroscedasticity, potential problems with Endogeneity had not been completely

controlled.

Quang et al. (2014) used a set of 134 samples of non-financial companies listed on the

Ho Chi Minh City Stock Exchange in the period 2009–2012 to study and analyze the

impact of ownership structure, capital structure on the performance in the context of a

transition economy. The authors performed regression methods for all 3 research models

measuring financial leverage using the ratio of total debt to total assets, the ratio of long-

term debt to total assets and the ratio of short-term debt to total assets. To measure

financial performance, the study used return on assets (ROA) and return on equity

(ROE). Meanwhile, the part of the ownership structure was represented by the

ownership of the members of the Board of Directors and the State ownership is shown

𝛽10.RISKit + 𝛽11.DIVit + 𝛽12.CFit + 𝜀it

27

in the model. In addition, the study also used a number of control variables such as firm

size (FSIZE), growth opportunity (GROW) and the ratio of tangible assets (TANG).

ROAit = 𝛽0 + 𝛽1.TDAit + 𝛽2.MaOWit + 𝛽3.STATEit + 𝛽4.GROWit

(2.23)

+ 𝛽5.TANGit + 𝛽6.FSIZEit + uit

ROEit = 𝛽0 + 𝛽1.TDAit + 𝛽2.MaOWit + 𝛽3.STATEit + 𝛽4.GROWit

(2.24)

+ 𝛽5.TANGit + 𝛽6.FSIZEit + uit

The results of the study showed that capital structure was negatively related to the

performance (using ROA and ROE measures). However, the variable, which represents

manager’s ownership, had an ambiguous and non-statistically significant impact on

performance for all companies in general, while for State ownership’s firms, there was

a negative relationship between manager ownership and performance as measured by

ROE. This showed that the level of authorization of managers in SOEs was higher than

other types. Especially, in the context of Vietnam, State ownership had a positive impact

on the performance based on ROA and ROE measures. Therefore, enterprises with a

high percentage of State ownership in the ownership structure would have a higher level

of performance. Regarding developed and emerging economies, like many other

previous foreign reseaches, this study also showed evidence of the influence of capital

structure factors on performance such as the ratio of tangible assets, growth

opportunities and firm size. Evidence of a negative relationship between capital

structure and performance would further explain the pecking order theory, showing that

Vietnamese firms would prioritize using internal sources, then debt and finally stocks if

necessary.

Do Van Thang and Trinh Quang Thieu (2010) analyzed the empirical relationship

between firm value (measured by Tobin's Q index) and capital structure of companies

listed on The Ho Chi Minh City Stock Exchange. The study was conducted on the

unbalanced array data of 159 non-banking firms, with 407 observations from 2006

to 2009. The general regression results show that there is a close relationship between

firm value and financial structure: (1) firm value has a ternary relationship with

debt/equity ratio; (2) when the debt ratio increases and is less than 105%, the firm

28

value increases with the same direction of capital structure, but when the debt ratio

is greater than 105%, the opposite result will be obtained, (3) the capital structure

will be optimal at the debt ratio of 105%. This result is a strong support for the trade-

off theory. However, this study only takes the total debt to total assets ratio as the

independent variable representing capital structure. Moreover, with the data of 407

observations from 2006 to 2009, the results show that the regression coefficient of

the total debt to total assets ratio (independent variable) order 1 in the quadratic

regression model is not significant statistic. In addition, some studies also show that

the impact of capital structure on profitability metrics such as return on assets (ROA)

and return on equity (ROE) can be different from the effect on Tobin’s Q.

2.2.3. Researches’ gaps

The inefficient use of capital structure, especially debt capital, will cause the opposite

effect of financial leverage. The impact of capital structure on performance has been

verified by many previous studies. Studies have also produced mixed results

including that capital structure has a positive, negative, or no statistically significant

effect on performance. According to the trade-off theory, an increase in debt will

generate a tax shield benefit, but at the same time, the cost of distress will also

increase (Myers, 1977). That means to a certain extent, the increase in the debt ratio

will lead to the disadvantage for the company. If the regression coefficients of the

squared variables are not statistically significant, the relationship will return to linear

form. In order to consider the non-linear relationship with U-shaped or inverted U-

shape, previous studies have included the debt squared ratio variable in the regression

model. According to previous researches, some authors consider the non-linear

impact of capital structure on performance to determine the optimal point in capital

structure (Margaritisa and Psillaki, 2010; Do Van Thang and Trinh Quang Thieu,

2010; Doan Vinh Thang, 2016). Most of studies only squared the proportion of total

debt representing capital structure without considering the maturity of the debt.

Therefore, the thesis will build a model with the square of the long-term debt ratio

and the short-term ratio as independent variables in the model.

29

Chapter 2’s summary

In this chapter, the author has introduced a number of concepts related to capital

structure including the traditional view, M&M theory, trade-off theory and pecking

order theory. Meanwhile, the thesis briefly reviewed previous studies including

foreign researches and in Vietnam about the impact of capital structure on the

performance of enterprises. Each study will have different results due to differences

in economic conditions and characteristics of each country. Based on a review of

domestic and international studies, the authors use the variable total debt ratio to

represent capital structure. Most of the previous studies only squared the proportion

of total debt representing capital structure without considering the maturity of the

debt. In addition, most of the studies only consider the linear relationship between

capital structure and performance, but have not exploited much to the nonlinear

relationship. Therefore, the study will build a model with the square of the long-term

debt ratio and the short-term ratio as independent variables in the model. At the end

of chapter 2, the thesis introduces the factors in performance that capital structure

affects based on a dataset of 47 listed material manufacturing enterprises in Vietnam.

30

CHAPTER 3: RESEARCH METHOD

3.1. Research hypothesis

According to author Berzkalne (2015), there is a non-linear relationship between

capital structure and profitability. Based on the research gap in chapter 2, the study

hypothesizes the impact of capital structure on performance.

ROA is calculated as profit after tax on average total assets, the use of debt will

reduce profit after tax due to interest expense but does not affect the denominator.

Most firms that use debt will benefit from a tax shield so debt is expected to have a

positive effect to a certain extent on ROA, but to have a negative effect beyond that

level. The hypothesis is put forward as follows:

H1: Short-term debt ratio (STD) impacts ROA in an inverted-U shape.

ROA is calculated as profit after tax on average total assets, the use of long-term debt

will reduce profit after tax due to long-term interest expense but does not affect the

denominator. So, ROA is expected to decrease with debt ratio or in other words, the

relationship between debt ratio and ROA is linear. The hypothesis put forward by the

author is as follows:

H2: Long-term debt ratio (LTD) has a negative effect on ROA.

For ROE, the use of short-term debt or long-term debt will reduce profit after tax in

the numerator due to an increase in interest expense but decrease equity in the

denominator due to the use of debt instead of equity. Most firms that use debt will

benefit from a tax shield so debt is expected to have a positive effect to a certain

extent on ROE but to have a negative effect beyond that level. Therefore, the study

hypothesized the following:

H3: Short-term debt ratio (STD) impacts ROE in an inverted-U shape.

ROE is calculated by profit after tax on average total assets, the use of long-term debt

will reduce profit after tax due to long-term interest expense but does not affect the

31

denominator. So, ROE is expected to decrease with debt ratio or in other words, the

relationship between debt ratio and ROE is linear. The hypothesis put forward by the

author is as follows:

H2: Long-term debt ratio (LTD) has a negative effect on ROE.

3.2. Description of variables in the research model

3.2.1. Dependent variable

The notion of performance is a controversial concern due to its multiple meanings. The

study of the company’s performance is rooted in organization theory and strategic

management (Murphy et al., 1996). Measures of performance are related to both

financial statement metrics and market value. Performance

is explained by

Chakravarthy (1986) as the core of profit maximization, on assets and shareholder

benefits. Performance is measured via sales growth and market share growth and sales

per employee, a broader definition than only focusing on metrics of financial statements

(Hoffer and Sandberg, 1987). The usefulness of performance measures can be affected

by the company’s target strategy, which leads to the choice of measures as well as the

development of the stock and capital markets are different. For example, if research in

the stock market does not develop, then measures of the performance of the market will

not indicate positive results. Kimberly et al. (2010) argue that performance measures

include financial performance and operational efficiency. Profit maximization,

shareholder wealth maximization and return on assets maximization are good examples

of financial performance.

Return on assets (ROA) and return on equity (ROE)

The common metrics in measuring performance are return on assets (ROA) and return

on equity (ROE). These accounting measures, taken from balance sheetsand business

reports, are implemented by many researchers (Demsetz and Lehn, 1985; Gorton and

Rosen, 1995; Ang, Cole and Line, 2000). According to Jonathan and Peter (2014)

Analysts typically evaluate a company’s return on investment by comparing the earnings

using ratio such as return on equity (ROE). ROE provides a measure of the return that a

company has earned on its previous investments. A high ROE may indicate that the

32

company is likely to find very profitable investment opportunities. Another popular

metric is return on assets (ROA). As a performance measure, ROA is more beneficial

because it is less sensitive to leverage than ROE. However, ROA is sensitive to working

capital, for example, an equal increase in a company’s accounts receivable and payable

will increase total assets and therefore lower ROA. Based on the literature and empirical

evidence, it is possible to explain the effect of debt on a firm’s performance using ROA

and ROE. The ROA measurement is based on researchers Mathur et. al (2001) and Abor

(2007), where ROA is calculated by dividing earning after tax by total assets and ROE

is calculated by dividing earning after tax by equity according to Abor (2005).

3.2.2. Independent variables

Capital structure is the combination of debt and equity that a company uses to finance

its business (Gill et al., 2009; Azhagaiah and Gavoury, 2011; Shubita and Alsawalhah,

2012). The debt ratio is one of the measures used to express capital structure, which

calculated by dividing total debt to total assets (Abor, 2005). Several studies measure

the debt ratio using an alternative formula, such as total liabilities to total assets (Gill

and Mathur, 2011) and long-term debt to total assets (Anderson and Reeb, 2004).

However, it is not convincing enough to measure capital structure only by one measure

or one attribute. This can lead to false conclusions about the capital structure of the firm

(Shubita and Alsawalhah, 2012). Therefore, many previous researchers selected more

than one measure as a proxy for capital structure, such as the combination of total debt

to total assets, short-term debt to total assets, and long-term debt to total assets (Ahmad

and Abdul Rahim, 2013). Firms tend to use short-term debt more than long-term debt.

While raising short-term debt may be easy, it may affect company when profits drop

suddenly. When there is a problem of sudden decrease in profit during the year, firms

difficultly settle short-term debt which leads to profitability tend to decline (Abor, 2005;

Margaritis and Psillaki, 2010; Salim and Yadav, 2012).

According to asymmetric information theory, business managers have more information

about their operations, future cash flows, profitability opportunities, etc. about the

business they are managing than outside investors or debtors. Therefore, outside

investors or creditors see the increase or decrease in debt (capital structure adjustment)

33

as a signal of information being held by managers. Pettit & Singer (1985) discussed the

problem of asymmetric information and agency costs that affect the cost and availability

of credit to small businesses. They argue that smaller firms will have higher asymmetric

information because the quality of their financial statements is not as high as that of

large firms. On the other hand, the cost of raising capital (especially issuing shares) of

small businesses is often higher than the cost of raising capital of larger enterprises. In

addition, the issuance of new shares to raise capital will dilute the ownership rights of

existing shareholders, which is especially serious for SMEs as shareholders currently

face loss of control or even annexation or merger. The long-term debt-to-total assets

ratio is a measurement of the percentage of a company's assets that are financed with

long-term debt, including loans or other debt obligations that last more than one year.

The short-term debt ratio will be measured by dividing the sum of all loans or other

liabilities lasting less than one year by the total assets.

Long-term debt

Long-term debt is a composite indicator that reflects the total value of long-term debts

of an enterprise, including debts with the remaining payment term of twelve months or

more or over a normal business cycle usually at the time of reporting. Long-term

liabilities are liabilities that are due after one year or during the normal operation of the

company. The normal operating period is the amount of time it takes for a company to

turn inventory into cash. On the classified balance sheet, liabilities are segregated

between short-term and long-term to help users assess the financial position of the

company in the short-term and long-term. Long-term debt meets long-term investment

needs such as: building factories, production lines, building infrastructure (roads,

seaports, airports...), improving and expanding products large scale export. Because the

investment term is often long, long-term credit often applies the form of disbursement

many times according to the project progress. In general, long-term credit carries a great

deal of risk, because the longer the term, the greater the potential for unforeseen

fluctuations, such as the company falling into a long-term loss of money leading to loss

of ability to repay the loan. Long-term debt provides viewers with financial statements

and balance sheets more information about a company’s long-term development, while

34

short-term debt informs debt users that the company is owed during present. On the

balance sheet, accounts are listed in order of liquidity, so long-term debt comes after

short-term debt.

Short-term debt

Current liabilities are short-term financial obligations of a company that must be paid

within one year or within a normal operating cycle. The operating cycle, also known as

the cash conversion cycle, is the time it takes for a company to purchase inventory and

turn it into cash from a sale. An example of a current liability is a balance owed to a

supplier in the form of accounts payable. Current liabilities are short-term financial

obligations of a company that must be paid within one year or within a normal operating

cycle. Current liabilities are usually paid off with current assets, which are used up

within a year. Examples of current liabilities include accounts payable, current

liabilities, dividends, and accounts payable as well as income tax payable. Current

liabilities are usually paid off with current assets, which are used up within a year.

Current assets include cash or accounts receivable, which is money owed by customers

to make sales. The ratio of current assets to current liabilities is an important factor in

determining a company's ability to pay its ongoing liabilities as they come due. Accounts

payable is typically one of the largest short-term liabilities on a company’s financial

statements, and it represents a supplier's outstanding invoices. Companies try to match

payment dates so that their receivables are collected before they are payable to suppliers.

For example, a company may have a 60-day period on money owed to their supplier,

which results in asking their customers to pay within the 30-day period. Short-term debt

can also be paid off by creating a new current liability, such as a new short-term debt

obligation.

3.2.3. Control variables

Firm size (SIZE)

As the business grows, management becomes more diversified and the likelihood of

bankruptcy decreases. In addition, capital market approaches become easier (Gruber and

Warner, 2012). Therefore, large enterprises will increase the use of debt at a relatively

35

low cost. In the case of a larger size, it helps to boost the performance of the company.

As a result, the company’s performance is enhanced by its larger scale. However, from

a negative point of view, the larger the firm size, the more difficult it can be managed

and thus leading to a decrease in performance (Amran and Che-Ahmad, 2011). The size

of a company is measured by the logarithm of total assets. Firm size has been indicated

by previous authors to have a positive effect on performance, as bankruptcy cost

decreases with size. Therefore, firm’s size is expected to have a positive effect on

performance. Gleason et al. (2001) have also shown that firm size has a positive and

statistically significant effect on ROA. In contrast, many other researchers such as

Mudambi and Nicosia (1998); Lauterbach and Vaninsky (1999); Durand and Coeuderoy

(2001); Tzelepis and Skuras (2004) made the comment about the insignificant influence

of size on performance. Therefore, the study carries out hypothesis as following:

H5: Firm size (SIZE) has a positive effect on ROA and ROE.

Revenue growth (GROWTH)

Sales growth is one of the determinants of capital structure. This was concluded by

Nadaraja et al. (2011), in which sales growth is one of the company-specific factors has

important significance and is consistent with capital structure theory. Businesses with

high growth rates can be seen as having more diversified investment opportunities than

others. According to Jensen and Meckling (1976), firms with many growth opportunities

tend to maintain a low debt level to avoid conflicts of interest with creditors because

they have a wide choice of financing sources for new investment opportunities (Jensen

and Meckling, 1976; Kim and Sorensen, 1986). Since the increasing cost of debt due to

growth is seen as an agency cost that will reduce the business’s value, managers should

prefer equity over loans to avoid incurring this additional costs. Several studies have

used changes in total assets to measure sales growth (Gill et al., 2011). However, this

study adopts the study of Shubita and Alsawalhah (2012), where the sales growth is

measured by the current year’s sales minus the previous year’s sales divided by the

current year’s sales and also found that sales growth and performance has a positive

relationship. With the expectation that companies with high revenue growth will be

36

more efficient, as companies with growth will generate returns on investment.

Therefore, revenue growth is expected to positively affect the company’s performance.

H6: Revenue growth (GROWTH) has a positive effect on ROA and ROE.

Table 3.1: Description of variables in the research model

Source: Author’s summary

3.3. Theoretical research model

The topic inherits the research model of authors Abor Joshua (2005), author Gill

Amarjit et al (2011), author Ebaid Ibrahim El-Sayed (2013) and author Nguyen Thi

Thanh Vinh (2021). The model will include two independent variables, namely

short-term debt ratio and long-term debt ratio, two control variables are firm size and

revenue growth. Based on the traditional view, also known as the optimal capital

structure theory, there is an optimal capital structure. According to mathematical

theory, to determine the extreme point, the thesis will square the independent variable

to have a quadratic function y=ax + bx + c. Therefore, the thesis will build a research

model with squared capital structure variable to determine the optimal capital

structure for short-term debt ratio and long-term debt ratio. The theoretical research

model will have the following form:

37

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

(3.1)

ROAi,t = β0 + β1*STDi.t + β2*STDi.t

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

(3.2)

ROEi,t = β0 + β1*STDi.t + β2*STDi.t

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

(3.3)

ROAi,t = β0 + β1*LTDi.t + β2*LTDi.t

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

(3.4)

ROEi,t = β0 + β1*STDi.t + β2*STDi.t

In which,

ROA: Return on assets

ROE: Return on equity

STD: Short-term debt ratio

LTD: Long-term debt ratio

SIZE: Firm size

GROWTH: Growth opportunity

β0: Intercept

β0,… β6: Coefficients

i, t: Corresponding to each business and each year

ε: Error

The four equations above describe the impact of capital structure on performance

including the impact of control variables. In the above model, we see the presence of

squared variable of capital structure. The presence of both squared and first-order

variables in capital structure in the same model will cause autocorrelation and

multicollinearity. Therefore, we replace them with averaged variables (NSTD: averaged

short-term debt ratio, NLTD: averaged long-term debt ratio) to minimize autocorrelation

and multicollinearity. These variables are calculated by subtracting their mean from

their respective debt ratios. The explanatory meaning is still preserved. This part will be

presented in more detail in the regression section of the thesis.

Thus, the model after adding the variable of averaged short-term debt ratio and averaged

long-term debt ratio will have the following form.

38

(3.5)

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

ROAi,t = β0 + β1*NSTDi.t + β2*NSTDi.t

(3.6)

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

ROEi,t = β0 + β1*NSTDi.t + β2*NSTDi.t

(3.7)

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

ROAi,t = β0 + β1*NLTDi.t + β2*NLTDi.t

(3.8)

2 + β3*SIZEi.t + β4*GROWTHi,t + εi.t

ROEi,t = β0 + β1*NSTDi.t + β2*NSTDi.t

In which,

ROA: Return on assets

ROE: Return on equity

NSTD: STD minus the average of STD

NSTD2: Square of NSTD

NLTD: LTD minus the average of LTD

NLTD2: Square of NLTD

SIZE: Firm size

GROWTH: Growth opportunity

β0: Intercept

β0,… β6: Coefficients

i, t: Corresponding to each business and each year

ε: Error

3.4. Research data

The study will carry out sample, based on secondary data of listed material

manufacturing enterprises on the Ho Chi Minh Stock Exchange (HoSE). Listed firms

have a fairly significant effect on economies in term of scale or business activities (Choi

et al., 2014). Using data from audited financial statements of listed companies will

ensure the reliability and transparency of the information disclosed by the company.

Based on data extracted from HoSE website, the total number of enterprises in the

material manufacturing industry is 67 enterprises, ranking second among the overall of

406 listed ones, so the research sample will be guaranteed. In which, the research period

39

is 10 years from 2012 to 2021 that ensures the representativeness of the selected sample

to explain the research results (Chadha and Sharma, 2015; Choi et al., 2014).

According to Tabachnick and Fidell (2007), for statistical significance results in

regression analysis, the number of observations is estimated with the formula:

n >= 50 + 8p

(3.9)

Where,

n: The number of observations

p: The number of variables

In the model (3.2) there are six independent variables used, applied to the formula

will have 50+8*6= 98 observations. Therefore, to get the most suitable regression

model, there must be at least 98 observations or more. The number of samples is

based on the list of 67 listed companies in the material manufacturing industry from

HoSE’s official website updated to August 31, 2022. The data is extracted from

annual reports and financial statements from 67 of these enterprises. However, in the

period from 2012 to 2021, the data set includes only 47 enterprises because 20

enterprises do not have complete data for the above period. These are Binhthuan

Agriculture Services Joint Stock Company, An Phat Holdings Joint Stock Company,

Create Capital Viet Nam Joint Stock Company, South Basic Chemicals Joint Stock

Company, Petro Viet Nam Ca Mau Fertilizer Joint Stock Company, FLC Mining

Investment and Asset Management Joint Stock Company, Viet Nam Rubber Group

- Joint Stock Company, HAI Agrochem Joint Stock Company, Hai Phong Hoang Ha

Paper Joint Stock Company, Binh Dien Fertilizer Joint Stock Company, Hoang Minh

Finance Investment Joint Stock Company, An Tien Industries Joint Stock Compnay,

Lao Cai Mineral Exploitation and Processing Joint Stock Company, Hanoi Plastics

Joint Stock Company, Pha Le Plastics Manufacturing and Technology Joint Stock

Company, Thuan Duc Joint Stock Company, Thang Long Urban Development and

40

Construction Investment Joint Stock Company, Thanh Nam Group Joint Stock

Company, Tien Bo Group Joint Stock Company and Yen Bai Industry Mineral Joint

Stock Company. The number of observations obtained from these 47 enterprises

accounts for most of the total assets of 67 raw material manufacturing enterprises in

Vietnam, helping to ensure the representativeness of the thesis. In addition,

secondary data is also obtained from the publications on websites vietstock.vn,

Cafef.vn. Through the collection of related parameters and the use of calculation

formulas, the study will describe the independent variables STD, LTD, STD2, LTD2,

SIZE and GROWTH.

3.5. Research method

The thesis uses a combination of qualitative and quantitative research methods to

solve the initially set objectives.

Qualitative research method used in the thesis includes approaching and analyzing

the synthetic theory, combined with references and discussions of previous related

empirical studies both in domestic and foreign countries. Then, the thesis designs the

research model and interprets variables, giving hypothesis for each independent

variable. Finally, the study discusses the research results, gives relevant suggestions

and recommendations to the subjects.

Quantitative research method is to examine the extent of impact of capital structure’s

factors on performance of enterprises. It includes specific technical methods such as

descriptive statistics, correlation analysis and panel data regression. Firstly,

descriptive statistics are used to provide general information about the variables in

the research model, the descriptive statistics include mean value, minimum value,

maximum value, standard deviation and number of observations. Secondly,

correlation analysis is used to determine the degree of strong, weak, positive or

negative correlation between the variables in the research model. In addition,

correlation analysis also suggests whether multicollinearity occurs, according to

Gujarati et al. (2011), if the correlation coefficient of any pair of independent

41

variables has the absolute value is higher than 0.8, then the model may experience

serious multicollinearity error. According to Gujarati (2011), there are three ways

that can deal with multicollinearity: (i) remove variables with high correlation with

other variables, (ii) use principal components analysis, and (iii) remains the same, in

which, the second method is especially effective when dealing with many

independent variables models. Finally, regression analysis of the balance data to test

the trend and extent of capital structure’s factors affecting the performance of listed

companies, using Pooled OLS regression, fixed-effects model (FEM) and random-

effects model (REM).

In order to choose the model with the highest explanatory level and the most suitable

for the research data, the thesis uses Hausman test to choose between FEM and REM

with the hypothesis H0 is accepting REM, and H1 is accepting FEM; use the

Redundant Fixed Effects test to choose between FEM and Pooled OLS with the

hypothesis that H0 is accepting Pooled OLS, and H1 is accepting FEM, also using

Lagrange Multiplier (LM) test to choose between REM and Pooled OLS with the

hypothesis H0 is to accept Pooled OLS, and H1 is to accept REM. To test the research

hypotheses according to the proposed model, the topic uses the t-test or the F-test

with the significant level of 1%, 5% and 10% to determine the reliability level of the

influence of the independent variable, and based on the regression coefficient to

explain the trend and the degree of influence of the independent variable on the

dependent variable. In addition, the study tests the defects of the regression model,

including severe multicollinearity, heteroskedasticity and autocorrelation, whereby

(i) severe multicollinearity will be tested and concluded through the Variance-

Inflating Factor (VIF), if VIF is greater than 10, the model has serious

multicollinearity and vice versa. (ii) The phenomenon of heteroskedasticity will be

tested and concluded by White's test with the following hypothesis: H0 is that there

is no heteroskedasticity, and H1 is that there is heteroskedasticity. (iii) The

phenomenon of autocorrelation will not be tested if the regression results are selected

FEM because FEM is only interested in individual differences contributing to the

42

model, so there is no autocorrelation. In contrast, the autocorrelation phenomenon

will be tested and concluded by the Wooldridge test with the following hypothesis:

H0 is the absence of autocorrelation, and H1 is the autocorrelation phenomenon. If

there are still defects, the final regression results will be determined by the general

least squares method (GLS); otherwise, the final regression result will be determined

according to Pooled OLS, or FEM, or REM depending on the results of the

aforementioned selection test.

3.6. Data analysis sequence

After collecting secondary data from reports published on enterprises’ websites, the

thesis analyzes the obtained data and proceeds to build a model using Stata 16.0 software

in the following sequence descriptive statistical system analysis, correlation analysis

between variables, testing multicollinearity phenomenon, regression according to

estimated models. After that, the discussion conducted the selection test to choose the

estimated model, check the model’s defects and if the model violates those tests, the

thesis will deal with the defect remedies.

3.6.1. Descriptive statistics

Descriptive statistics are coefficients that describe briefly or summarize a given data set,

which represents the whole or a sample of a population. Descriptive statistics review

essential demography for the research to reflect generally the object, describe the factors

to explain hypothesis. This analysis expresses the data’s characteristics via the average

values, examines the fluctuation of those variables through the minimum and maximum

figures also the median and standard deviation values. From these statistical criteria, we

will have some initial judgments about the research data sequence.

3.6.2. Correlation analysis

The correlation coefficient is a coefficient that measures the extent of correlation

between two variables with values range from -1 to 1. The value of -1 indicates a

perfectly negative correlation and vice versa the coefficient of 1 explains a completely

positive correlation. For example, for two variables x and y, when the correlation

43

coefficient is greater than zero, meaning x increases when y increases or x decreases

when y decreases. Otherwise, when the correlation is less than zero, x increases if y

decreases or x decreases if y increases. The greater the coefficient is, the stronger the

variation of x on y and vice versa. If the correlation coefficient is 0, there is no linear

relationship between the two variables. In many coefficient types, Pearson r is most

commonly used. In correlation analysis, Pearson is used to analyze linear relationship

between variables without distinguishing independent or dependent ones. If the

independent variables are strongly correlated, the multicollinearity problem must be

considered when implementing regression analysis.

According to Hoang Trong and Chu Nguyen Mong Ngoc (2008), the variables

correlation in the model is expressed by the Pearson coefficient (r) as follows.

Coefficient r > 0.8 is a very strong linear correlation; 0.6 < r < 0.8 is a strong linear

correlation; 0.4 < r < 0.6 is a linear correlation; 0.2 < r < 0.4 is weak linear correlation

and r < 0.2 is very weak linear correlation or there is no linear correlation.

3.6.3. Regression Model Selection

Regression analysis is to test the influence of independent variables on the dependent

one, which explains the direction and magnitude of the impact. The thesis conducts

regression analysis to examine the dependence of CAR on SIZE, ROE, DEP, NPL, NIM,

LDR and CIR. The regression results are empirical evidence to consider the impact. The

thesis applies the multivariate regression model with panel data method in combination

with: Pooled OLS, Fixed effect, Random effect. Firstly, we consider the advantages and

disadvantages of each model to elect the most appropriate estimation model.

Pooled OLS model

The OLS model is the first model which is used to perform regression. OLS is an

estimation over a dataset over a period of time, so all coefficients are constant across

various units and the same over time (Gujarati, 2011). Therefore, this model ignores the

differences between each unit. This weakness may lead to correlation between

independent variables in a model with many explanatory ones. Consequently, for a

44

higher efficient estimation, the thesis will continue to regress with fixed effect model

(FEM) and random effect model (REM).

Fixed effect model (FEM)

The fixed effect model (FEM), assuming each bank has its own characteristics that can

affect the explanatory variables, FEM analyzes this correlation between the residual of

each unit and the explanatory variables. Therefore, the model controls and distinguishes

the influence of individual characteristics (constant over the time) from the explanatory

variables so that we can estimate the real effects of these variables on the dependent

variable. To select a more meaningful model between Pooled OLS and FEM, the thesis

conducts the F-test, to test whether the intercept of the regression function in each bank

is different.

Random effect model (REM)

The random effect model (REM) is different from the FEM in the variation between

units. If the fluctuation between units is correlated with the independent variables - the

explanatory variables in the fixed effect model then in the random effect model, the

variation between banks is assumed to be random and uncorrelated with explanatory

variables. Therefore, if these differences affect the dependent variable then REM is more

preferable than FEM. Particularly, the residual of each entity (uncorrelated with the

explanatory variable) is considered a new explanatory variable. The appropriateness of

REM was verified by Hausman test when compared with FEM and LM test (Breusch-

Pagan Lagrange multiplier test) when compared with Pooled OLS model.

3.6.4. Model’s deficiencies test

The nature of Pooled OLS, FEM, REM is the average parameter estimation of the

regression function based on the least squares method. In order for these estimations do

not

to change (reliable estimation results),

the regression does not have

heteroskedasticity, auto-correlation and multi-collinearity. Therefore, the thesis detects

the model’s deficiencies. Firstly, the thesis performs multi-collinearity test through the

VIF (Variance Inflation Factors). If the VIF coefficients of each factor in the model are

less than 10, it proves that the research model does not have the multi-collinearity

45

phenomenon (Hoang Trong and Chu Nguyen Mong Ngoc, 2008). Then, the auto-

correlation phenomenon is determined by the Wooldridge test with the null hypothesis

H0 that there is no correlation between the observations’ components of the

chronologically ordered series. Thirdly, when studying the classical linear regression

model, the hypothesis is that the variance of each ui under the given value of the

explanatory variable Xi is constant. With respect to the homoscedasticity, the thesis

performs the Modified Wald test and finally the endogeneity phenomena is examined

through the Wu-Hausman test with the hypothesis H0 that there is no correlation between

the independent variables and the residual.

3.6.5. Model’s deficiencies remedy

In case the model has some deficiencies, the thesis will remedy by using the Feasible

Generalized Least Squares (FGLS) method (Gujarati, 2011). This is a way to remedy

deficiencies for models with large samples (from 30 samples or more) and without

lagged variables. The feasible generalized least squares method (FGLS) is actually the

ordinary least square method (OLS) applied to variables that have been transformed

from a model that does not meet the classical assumptions to the new one with no

deficiencies. Therefore, the estimation parameters in the new model will be more

reliable to perform.

46

Chapter 3’s summary

In chapter 3, the author made hypotheses related to the impact of capital structure on

performance. Meanwwhile, the author describes the variables used in the model and

from that, the author sets out a theoretical research model consisting of four quadratic

equations of the short-term debt ratio and the long-term debt ratio. The author’s

dataset used in the study is 47 listed material manufacturing enterprises in the 10-

year period from 2012 to 2021. Based on the research model and data, the author

describes the research method and detailed sequence for data analysis

47

CHAPTER 4: RESEARCH RESULTS AND DISCUSSIONS

Based on the research sequence presented in chapter 3, the thesis conducts

descriptive statistics and analyzes the correlation between variables to get general

comments about the research data set. Next, the thesis analyzes regression,

implements tests to select the most suitable model, and while conducts tests to check

the model’s defects, remedy them to lead to the research results.

4.1. Descriptive statistics

Through a sample of data collected from 47 listed companies in the material

manufacturing industry in Vietnam in the period from 2012 to 2021, Table 4.1

presents descriptive statistical results of the variables studied in the article.

Descriptive statistics results of the dependent variable and six independent variables

including: mean, maximum, minimum, and standard deviation.

Table 4.1: Descriptive statistics of variables

Variables Observations Mean

Standard deviation

Minium

Maximum

ROA

470

0.0683746

0.0741837

-0.1459692

0.4134419

ROE

470

0.1273453

0.1362202

-0.4087784

0.6452121

STD

470

0.3799659

0.1859466

0.0416457

0.8661429

LTD

470

0.0832982

0.1181578

0

0.6182065

SIZE

470

12.07926

0.5746812

11.10985

14.251

GROWTH

470

0.200572

2.299531

-0.8767009

48.81374

Source: Analysis results from Stata 16.0

According to Table 4.1, descriptive statistics show that most enterprises use short-

term debt more than long-term one. On average, short-term debt accounts for 38%

of total capital and long-term debt accounts for only 8%, less than 10%. The current

SIZE variable is taken according to the common logarithm of the corresponding

value in units of hundreds of billions of Dong. The size of the enterprise has an

average value of about one trillion Dong, with total assets ranging from ten billion

to about one hundred trillion. In terms of revenue, enterprises in the material

manufacturing industry have an average rate about 20%/year, of which there are

48

enterprises achieve a relatively high level of about 49%/year. Performance is

measured via return on average assets (ROA) and return on average equity (ROE).

In which, some enterprises have high average ROA ratio such as Nui Nho Stone Joint

Stock Company, Lamdong Minerals and Building Materials Join Stock Company,

enterprises with high average ROE ratio such as Nui Nho Stone Joint Stock

Company, Phu Tai Joint Stock Company and Hoa Phat Group Joint Stock Company.

According to Table 4.1 statistics show that the standard deviation of ROA is about

7% and ROE is about 13%.

Average rate of return for the period 2012 - 2021

20%

15%

10%

5%

0%

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

ROA ROE

Chart 4.1: Average rate of return over 10 years

Source: Author's summary

Chart 4.1 shows the performance through the average rate of return over 10 years of

47 enterprises in the period from 2012 to 2021. The data in Chart 4.1 shows the

profitability ratio in the period in 2016 and 2017 reached the highest value. In 2016,

ROE reached 18% and ROA reached 9%, in 2017, ROE reached 16% and ROA

reached 8%.

49

Average debt ratio for the period 2012 - 2021

50%

40%

30%

20%

10%

0%

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

STD

LTD

Chart 4.2: Average debt ratio over 10 years

Source: Author's summary

Chart 4.2 shows the proportion of short-term debt and long-term debt of 47 raw

material manufacturing in the period from 2012 to 2021. The overall figure shows

that most enterprises use more short-term capital, in which short-term debt accounts

for about 40% of total capital, the rest of long-term debt accounts for less than 10%.

For the size variable (SIZE), the average value of 47 firms is about one trillion Dong.

In which, Hoa Phat Group Joint Stock Company has total assets of about 62 trillion

Dong and Hoa Sen Group Joint Stock Company has total assets of about 14 trillion

Dong. These are the leading enterprises in the market as well as diversifying their

business in many fields, so they have quite large scale. The smallest enterprise is Do

Thanh Technology Corporation with an average size of about 152 billion Dong.

For the revenue growth variable (GROWTH), the average growth rate of 47

enterprises is about 20% per year. In which, TNT Group Joint Stock Company

achieved the highest revenue growth rate with an average rate of about 35% per year.

50

4.2. Correlation analysis

Correlation analysis is used to quantify the relationship between variables in the

research model. Statistical results describing the matrix of correlation coefficients

between pairs of independent variables are shown in the table below.

Table 4.2: Correlation between variables

ROA

ROE

STD

STD2

LTD

LTD2

SIZE

GROWTH

1.0000

ROA

0.8663

1.0000

ROE

-0.2830

-0.0370 1.0000

STD

-0.2870

-0.0628 0.9670

1.0000

STD2

-0.1359

-0.0266

-0.2038

-0.2078 1.0000

LTD

-0.1295

-0.0561

-0.2084

-0.2061 0.9146

1.0000

LTD2

0.0135

0.1461

0.2213

0.2313

0.3829

0.2564

1.0000

SIZE

0.0277

-0.0700

-0.0470

-0.0272

-0.0153

-0.0141 1.0000

GROWTH 0.0381

Source: Analysis results from Stata 16.0

The pairs of independent variables showing short-term debt ratio and short-term debt

ratio squared, dlong-term debt ratio and long-term debt ratio squared are all

correlated above 0.9. Thus, if these two pairs of variables are put in the same model,

there will be multicollinearity and autocorrelation. To overcome the above situation,

we calculate the variables NSTD equals STD minus the mean of STD, NLTD equals

LTD minus the mean of LTD. In the table, the other pairs of variables all have

correlation coefficients less than 0.8, so it can be seen that there is no

multicollinearity when the above variables are included in the model. After

calculating the above values, we consider the correlation between the new pairs of

independent variables as follows.

Based on the results of the correlation test for the dependent variable ROA and ROE,

there are all four independent variables NSTD, NSTD2, NLTD, NLTD2 and two

control variables are SIZE and GROWTH. In which, all four independent variables

have a negative correlation, while two control variables have a positive correlation

51

with ROA and ROE. Therefore, these variables will be included in the model to

explain the dependent variable ROA and ROE, in which, there are four variables that

have opposite effects, namely NSTD, NSTD2, NLTD, NLTD2, and two variables

that have the same direction are SIZE and GROWTH.

Table 4.3: Correlation between variables (after adjustment)

ROA

ROE

NSTD NSTD2 NLTD NLTD2 SIZE

GROWTH

1.0000

ROA

0.8663

1.0000

ROE

-0.2830

-0.0370 1.0000

NSTD

-0.0869

-0.1096 0.1235

1.0000

NSTD2

-0.1359

-0.0266

-0.2038

-0.0668

1.0000

NLTD

-0.1168

-0.0682

-0.1960

-0.0277

0.8026

1.0000

NLTD2

0.0135

0.1461

0.2213

0.0947

0.3829

01690

1.0000

SIZE

0.0277

-0.0700 0.0718

-0.0272

-0.0077

-0.0141 1.0000

GROWTH 0.0381

Source: Analysis results from Stata 16.0

Based on the results of the correlation test for the dependent variable ROA and ROE,

there are all 6 independent variables NSTD, NSTD2, NLTD, NLTD2, SIZE and

GROWTH that are correlated. Therefore, these variables will be included in the

model to account for the dependent variable ROA and ROE.

4.3. Regression analysis

Based on the results of the correlation test above, the thesis conducts regression for

ROA and ROE based on four independent variables NSTD, NSTD2, NLTD, NLTD2

and two control variables SIZE and GROWTH. According to the research sequence

presented in Chapter 3, the three models that the thesis will perform regression are

Pooled OLS, fixed effect model FEM and random effect model REM.

52

4.3.1. The impact of short-term debt on ROA

Table 4.4: Model regression results of short-term debt for the ROA

Coefficient Variable Pooled OLS FEM REM

-0.1167341*** -0.1170367*** -0.1167341*** NSTD

-0.1153818 -0.1200088 -0.1153818 NSTD2

0.0108871* 0.0123145** 0.0108871* SIZE

0.0007464 0.0005559 0.0007464 GROWTH

-0.0593028 -0.0763465 -0.0593028 Constant

R2 = 9.01% R2 = 9.19% R2 = 9.17%

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

For the results of Pooled OLS regression, two variables NSTD and NSTD2 have a

negative effect on ROA with 99% confidence, while the variable NSTD2 has no

statistical significance. Both SIZE and GROWTH variables have the same effect on

ROA but GROWTH is not statistically significant, SIZE variable has the same effect

with 90% confidence. Regarding the level of influence, NSTD is the variable that

has the strongest impact on the dependent variable with a β coefficient of -0.1167341.

With R2 equal to 9.01%, this means that 9.01% of the variation in the dependent

variable is explained by the independent variable. In general, the explanatory level

of the OLS model is low.

For the FEM fixed-effects model, two variables NSTD and NSTD2 have a negative

effect on ROA with 99% confidence, while the variable NSTD2 has no statistical

significance. Both SIZE and GROWTH variables have the same effect on ROA, but

GROWTH is not statistically significant, SIZE variable has the same effect with 95%

confidence. Regarding the level of influence, NSTD is the variable that has the

strongest impact on the dependent variable with a β coefficient of -0.1170367. The

FEM model has the same level of explanation as the Pooled OLS model with R2 of

53

9.19%, showing that 9.19% of the variation of the dependent variable is explained

by the NSTD and SIZE variables.

For the REM random effects model, two variables NSTD and NSTD2 have a

negative effect on ROA with 99% confidence, while the variable NSTD2 has no

statistical significance. Both SIZE and GROWTH variables have the same effect on

ROA but GROWTH is not statistically significant, SIZE variable has the same effect

with 90% confidence. Regarding the level of influence, NSTD is the variable that

has the strongest impact on the dependent variable with a β coefficient of -0.1167341.

The R2 results show that 9.17% of the variation of the dependent variable is explained

by the variables NSTD and SIZE.

4.3.1.1. Model selection test

As described in Chapter 3, the thesis will perform tests to choose between Pooled

OLS, FEM and REM models. However, the estimation of the research model

according to the Pooled OLS method does not accurately reflect the separate and

specific impacts of each different enterprise. Therefore, the author will perform

Hausman test to choose the FEM or REM model that will be more suitable for the

research data. Table 4.5 shows the results of the selection test between the two

models.

Table 4.5: Model’s selection for short-term debt on the ROA

Dependent variable

Test

Statistical value

P-value

Choose between

Prob>chi2 =

FEM and REM

Chi2(4) = 4.78

ROA

Hausman

0.3112*

 Choose REM

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

54

 Hausman test to choose between FEM and REM

Hausman test hypotheses:

H0: There is no correlation between variables and residual so REM is more

appropriate.

H1: At least one variable is correlated with the residual so FEM is more relevant.

For the impact of short-term debt on ROA, the test results show that p-value = 0.3112

is greater than the 10% significance level, so the hypothesis H0 is not rejected. So the

REM model will be a more suitable model than the FEM (Hausman, 1978).

Thus, through the Hausman test, it can be concluded that, for the impact of short-

term debt on ROA, the REM regression model is the most suitable model for the

research data. On the other hand, it is necessary to continue to consider the defect

tests of each model before a final conclusion can be drawn about the accuracy of this

model.

4.3.1.2. Model deficiencies test

 Multicollinearity test

Table 4.6: VIF test result for short-term debt on the ROA

ROA

Variables Variance Inflation Factors (VIF)

1.07 NSTD

1.06 SIZE

1.03 NSTD2

1.01 GROWTH

Source: Analysis result from Stata 16.0

In the assumptions of the linear regression model, the independent variables have no

linear relationship. If this assumption is violate, there will be multi-collinearity in the

model, which means the independent variables are interdependent and expressed as

55

a function. Therefore, before regression variables, the thesis conducts the multi-

collinearity test.

Compared to the correlation matrix analysis method, Variance Inflation Factors

(VIF) is more commonly used in the multi-collinearity test. The result in Table 4.6

shows the VIF coefficient of each factor is less than 10, indicating that the research

model does not violate the multi-collinearity phenomenon (Hoang Trong and Chu

Nguyen Mong Ngoc, 2008). Therefore, the variables chosen for the regression model

are appropriate.

 Heteroskedasticity test

Table 4.7: Heteroskedasticity test result for short-term debt on the ROA

Modified Wald test Statistic value P-value

Chi2(01) = 0.00 Prob>chibar2 = 1.0000* ROA

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

The Modified Wald is performed to test the heteroskedasticity in the FEM model

(Christopher, 2001). LM test – Breusch and Pagan Lagrangian Multiplier will be

used to test the heteroskedasticity in the REM regression model (Baltagi, 2008). This

is a phenomenon that occurs when the residual variance is not a constant but

increases or decreases as the independent variable increases. When conducting the

linear regression model, the hypothesis is given that the variance of each random

variable ui under the given value of the explanatory variable Xi is constant. The

appearance of the heteroskedasticity indicates that the regression model is not

accurate or because the model has omitted some independent variables which is

important to the dependent variable (Nguyen Van Ngoc, 2006).

The hypotheses of the Modified Wald test include:

H0: The model does not have the heteroskedasticity phenomenon.

H1: The model has the heteroskedasticity phenomenon.

56

For ROA, the test results show that p-value = 1.0000 is greater than the 10%

significance level, so the hypothesis H0 is not rejected. Thus, the REM model does

 Auto-correlation test

not have the heteroskedasticity phenomenon.

Table 4.8: Auto-correlation test result for short-term debt on the ROA

Wooldridge test Statistic value P-value

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

F(1,9) = 17.438 Prob>F = 0.0024*** ROA

Source: Analysis result from Stata 16.0

The auto-correlation phenomenon is recognized by Wooldridge test. The term auto-

correlation can be understood as the correlation between the components of a series

observations arranged in chronological order. This phenomenon makes R-squared

results, variances and standard errors of prediction become ineffective. The linear

regression model assumes that the auto-correlation does not exist in ui residual.

The hypotheses of the Wooldridge test are set out as follows:

H0: The model does not have the auto-correlation.

H1: The model has the auto-correlation.

For ROA, the test results show that p-value = 0.0024 is less than 1% significance

level, so hypothesis H0 is rejected. Thus, for the impact of short-term debt on ROA,

the REM model has auto-correlation.

4.3.1.3. Remedy model’s deficiencies

From the above tests, the model has one deficiency is the auto-correlation

phenomenon. According to Gujarati (2011), how to overcome the above defect for a

model with a large sample (from 30 samples or more) is to select the regression

model according to the feasible generalized least squares method (FGLS). The FGLS

method is actually the ordinary least squares method (OLS) applied to variables that

57

have been transformed from a model that violates the classical assumptions. Thus,

the estimated parameters from the new model will be more reliable.

Table 4.9 shows that three out of four independent variables are included in the

regression model and are statistically significant, namely NSTD, NSTD2 and SIZE.

The three variables NSTD, NSTD2 and SIZE all have p-values less than 10%

significance level with values of 0%, 5.7% and 8.1%, respectively, so they will have

statistical significance for the model.

Table 4.9: FGLS regression model for short-term debt on the ROA

Variables

Constant

NSTD

NSTD2

SIZE

GROWTH

-0.048

-0.108***

-0.156*

0.01*

0.001

Coefficient

Number of observations: 470

Prob > chi2 = 0.0000

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

About the influence of the variables show that only the SIZE variable has the same

effect on the ROA ratio, while the other two variables NSTD and NSTD2 have the

opposite effect on ROA. Specifically, ranked in order of the influence of the

independent variables on the ROA variable, the first is the NLTD2 variable with

15.6%, the second is the NSTD2 variable with 10.8%, and finally the least impact is

the SIZE with 1%. For the GROWTH variable, there is an impact on ROA but it is

not statistically significant, so it cannot be explained for the model.

Therefore, the regression model of the thesis is presented as follows:

ROA = − 0.0479383 − 0.108076*NSTD − 0.1560113*NSTD2 + 4.1 0.0099409*SIZE + ε

4.3.1.4. Discussion about research results

Based on the research results, there are all four independent variables including

NSTD, NSTD2, NLTD and SIZE that affect the performance (ROA) of the listed

material manufacturing enterprises in Vietnam. The degree of impact of each

58

independent variable on the dependent variable is different, presented in detail in

order from most to least.

Short-term debt ratio (STD)

Based on the regression results, the original of this squared adjusted short-term debt

ratio variable is the short-term debt ratio, so the author will discuss the short-term

debt ratio variable. The effect of short-term debt on ROA is essentially non-linear.

Although the debt ratio of the squared short-term debt ratio variable is statistically

significant, the effect of the short-term debt ratio on ROA will follow an inverted U-

shaped parabol. The regression coefficient corresponding to the variable short-term

debt ratio is statistically significant at 1% and has a value of -0.108 while the

regression coefficient corresponding to the squared short-term debt ratio is

statistically significant at 10% and is equal to -0.156. We have a graph of the

quadratic function y = ax2 + bx + c with a<0 and b<0 (a is the regression coefficient

of the squared short-term debt ratio -0.203 and b is the regression coefficient of the

b

short-term debt ratio -0.145). Regression results show that both of these regression

2a

(−0.108)

, which coefficients are negative leading to the extreme point for NSTD equal to −

2.(−0.156)

means the value is equal to − = − 0.346. However, since the NSTD variable

has been adjusted, the value of the extreme point for the STD variable will be

recalculated by adding the mean of the short-term debt variable to 0.3799659 in

Table 4.1. Thus, the extreme point value will be -0.346 + 0.380 = 0.0336. Thus, an

increase in the short-term debt ratio will increase the ROA to the extreme point,

which will decrease. In fact, in the short term, an increase in short-term debt will

increase profits for businesses, but to a certain extent, in the long term, increasing

debt will increase the burden on businesses and reduce enterprise’s profits.

Firm size (SIZE)

The variable of firm size is the one that has the least impact on the dependent variable

ROA, but this is the variable most authors include in the test and most are statistically

significant. The coefficient of the variable SIZE greater than 0 shows the positive

59

impact of the firm size factor on performance. According to the regression results,

SIZE has the same impact with the author’s initial expectation of having the positive

effect and the coefficient is 0.0099409. The results obtained are similar to those of

Mudambi and Nicosia (1998), Lauterbach and Vaninsky (1999), Durand and

Coeuderoy (2001) and Tzelepis and Skuras (2004). This positive effect is explained

that the larger the size of the business, the greater the total assets that the enterprise

owns, the more effective the performance will be. Large-scale enterprises will have

resources to optimize capital to create better performance than small-sized ones.

The revenue growth variable has p-values outside the 0%, 5% and 10% significance

levels, so they will not be included in the explanatory model. However, the thesis

will still provide discussions on revenue growth factors to provide policy suggestions

for managers.

Revenue growth (GROWTH)

Revenue growth variable has a positive impact with the dependent variable of

performance. In terms of impact, revenue growth is the weakest independent variable

among the independent ones included in the model. With the impact dimension

greater than 0, the revenue growth variable has the right sign of the author’s initial

expectation. Regression results show similar to the predictions of Jensen and

Meckling (1976), Nadaraja et al. (2011), Gill et al (2011). According to the results

of the model, the higher the revenue growth rate, the better the performance of the

business. Positive revenue growth is a good sign that the business can take advantage

of capital to increase profits.

60

4.3.2. The impact of long-term debt on ROA

Table 4.10: Regression results for long-term debt on the ROA

Coefficient Variable Pooled OLS FEM REM

-0.1073244** -0.1130533** -0.1096662** NLTD

0.0154275 0.0180987 0.0164316 NLTD2

0.0100969 0.0117065* 0.0107602 SIZE

0.0011158 0.0009151 0.0010305 GROWTH

-0.0540276 -0.0734673 -0.0620365 Constant

R2 = 2.47% R2 = 2.69% R2 = 2.69%

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

For Pooled OLS regression results, two variables NLTD and NLTD2 have a negative

impact on ROA with 95% confidence, while NLTD2 variable has no statistical

significance. Both SIZE and GROWTH variables have the same effect on ROA but

are not statistically significant. Regarding the degree of influence, NLTD is the

variable with the strongest impact on the dependent variable with the coefficient β of

-0.1073244. With R2 equal to 2.47%, this means that 2.47% of the variation in the

dependent variable is explained by the independent variable. In general, the

explanatory level of the OLS model is low.

For the FEM fixed-effects model, two variables NLTD and NLTD2 have a negative

impact on ROA with 95% confidence, while the variable NLTD2 has no statistical

significance. Both SIZE and GROWTH variables have the same effect on ROA, but

GROWTH is not statistically significant, SIZE variable has a positive effect with

10% significance level. Regarding the level of influence, NLTD is the variable that

has the strongest impact on the dependent variable with the coefficient β of -

0.1130533. The FEM model has the same level of explanation as the Pooled OLS

61

model with R2 of 2.69%, showing that 2.69% of the variation of the dependent

variable is explained by the NLTD and SIZE variables.

For the REM random effect model, two variables NLTD and NLTD2 have a negative

effect on ROA with 99% confidence, while the variable NLTD2 has no statistical

significance. Both SIZE and GROWTH variables have the same effect on ROA but

are not statistically significant. Regarding the degree of influence, NLTD is the

variable with the strongest impact on the dependent variable with the coefficient β of

-0.1096662. The R2 results show that 2.69% of the variation of the dependent

variable is explained by the NSTD variable.

4.3.2.1. Model selection test

As described in Chapter 3, the thesis will perform tests to choose between Pooled

OLS, FEM and REM models. To choose between Pooled OLS and FEM models, the

author performs F-test. And the fit of REM model is verified through Hausman test

when compared with FEM model and LM test (Breusch-Pagan) Lagrange multiplier

test) when compared with the OLS model. Table 4.12 shows the results of the

selection test between the three models.

Table 4.11: Model’s selection for long-term debt on the ROA

Dependent variable

Test

Statistical value

P-value

Choose between

Prob>chi2 =

FEM and REM

Chi2(4) = 2.09

ROA

Hausman

0.7187*

 Choose REM

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

 Hausman test to choose between FEM and REM

Hausman test hypotheses:

H0: There is no correlation between variables and residual so REM is more

appropriate.

H1: At least one variable is correlated with the residual so FEM is more relevant.

62

For ROA, the test results show that p-value = 0.7187 is greater than the 10%

significance level, so hypothesis H0 is not rejected. So the REM model will be a more

suitable model than the FEM model (Hausman, 1978).

Thus, through the Hausman test, it can be concluded that, for the impact of long-term

debt on ROA, the REM regression model is the most suitable model for our research

data. On the other hand, it is necessary to continue to consider the defect tests of each

model before a final conclusion can be drawn about the accuracy of this model.

4.3.2.2. Model deficiencies test

 Multicollinearity test

Table 4.12: VIF test result for long-term debt on the ROA

ROA

Variables Variance Inflation Factors (VIF)

3.42 NLTD

3.00 NLTD2

1.25 SIZE

1.00 GROWTH

Source: Analysis result from Stata 16.0

In the assumptions of the linear regression model, the independent variables have no

linear relationship. If this assumption is violate, there will be multi-collinearity in the

model, which means the independent variables are interdependent and expressed as

a function. Therefore, before regression variables, the thesis conducts the multi-

collinearity test.

Compared to the correlation matrix analysis method, Variance Inflation Factors

(VIF) is more commonly used in the multi-collinearity test. The result in Table 4.12

shows the VIF coefficient of each factor is less than 10, indicating that the research

model does not violate the multi-collinearity phenomenon (Hoang Trong and Chu

63

Nguyen Mong Ngoc, 2008). Therefore, the variables chosen for the regression model

are appropriate.

 Heteroskedasticity test

Table 4.13: Heteroskedasticity test result for long-term debt on the ROA

LM – Breusch and Pagan

Statistic value P-value

Lagrangian Multiplier test

Chi2(01) = 0.64 Prob>chibar2 = 0.2119* ROA

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

Modified Wald test was performed to test the heteroskedasticity in the FEM

regression model (Christopher, 2001). LM – Breusch and Pagan Lagrangian

Multiplier test will be used to test the heteroskedasticity in the REM regression

model (Baltagi, 2008). This is a phenomenon that occurs when the variance of the

error is not constant but increases or decreases as the independent variable increases.

When studying the classical linear regression model, the hypothesis is made that the

variance of each random variable ui under the condition that the given value of the

regressor Xi is constant. The appearance of the heteroskedasticity shows that the

regression model is not accurate or because the model has omitted some important

independent variables for the dependent variable (Nguyen Van Ngoc, 2006).

The hypotheses of the LM – Breusch and Pagan Lagrangian Multiplier test:

H0: The REM model does not have the heteroskedasticity phenomenon.

H1: The REM model has the heteroskedasticity phenomenon.

For ROA, the test results show that p-value = 0.2119 is greater than the 10%

significance level, so hypothesis H0 is not rejected. Thus, the REM model does not

have the heteroskedasticity phenomenon.

64

 Auto-correlation test

Table 4.14: Auto-correlation test result for long-term debt on the ROA

Wooldridge test Statistic value P-value

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

F(1,9) = 13.446 Prob>F = 0.0052*** ROA

Source: Analysis result from Stata 16.0

The auto-correlation phenomenon is recognized by Wooldridge test. The term auto-

correlation can be understood as the correlation between the components of a series

observations arranged in chronological order. This phenomenon makes R-squared

results, variances and standard errors of prediction become ineffective. The linear

regression model assumes that the auto-correlation does not exist in ui residual.

The hypotheses of the Wooldridge test are set out as follows:

H0: The model does not have the auto-correlation.

H1: The model has the auto-correlation.

For ROA, the test results show that p-value = 0.0052 is less than 1% significance

level, so hypothesis H0 is rejected. Thus, for ROA, the REM model has auto-

correlation

4.3.2.3. Remedy model’s deficiencies

From the above tests, the model has one deficiency is the auto-correlation

phenomenon. According to Gujarati (2011), how to overcome the above defect for a

model with a large sample (from 30 samples or more) is to select the regression

model according to the feasible generalized least squares method (FGLS). The FGLS

method is actually the ordinary least squares method (OLS) applied to variables that

have been transformed from a model that violates the classical assumptions. Thus,

the estimated parameters from the new model will be more reliable.

65

Table 4.15 shows that two out of four independent variables are included in the

regression model and are statistically significant, namely NSTD and SIZE. The two

variables NSTD and SIZE both have p-values less than 10% significance level with

values of 0% and 7.5%, respectively, so they will have statistical significance for the

model.

Table 4.15: FGLS regression model for long-term debt on the ROA

Variables

Constant

NLTD

NLTD2

SIZE

GROWTH

-0.068

-0.128***

0.033

0.011*

0.000

Coefficient

Number of observations: 470

Prob > chi2 = 0.0000

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

About the influence of the variables show that only the SIZE variable has the same

direction that affects the ROA ratio, while the other variable NLTD has the opposite

effect on ROA. Specifically, in order of the influence of the independent variables

on the ROA variable, the first is the NLTD variable with 12.8% and the second is the

SIZE variable with 1.1%. For variables NLTD2 and GROWTH both have an impact

on ROA but are not statistically significant, so it cannot be explained for the model.

Therefore, the regression model of the thesis is presented as follows:

ROA = − 0.0683894 − 0.1285862*NLTD + 0.0111925*SIZE + ε 4.2

4.3.2.4. Discussion about research results

Based on the research results, there are all two independent variables including

NLTD and SIZE that affect the performance (ROA) of listed material manufacturing

enterprises in Vietnam. The degree of impact of each independent variable on the

dependent variable is different, presented in detail in order from most to least.

66

Long-term debt ratio (LTD)

Based on the results of the research model, the squared long-term debt variable is not

statistically significant, so the impact of the long-term debt ratio on ROA is linear.

Based on the regression results, only the squared adjusted long-term debt ratio is not

statistically significant, but the source of this independent variable is the long-term

debt ratio, so the author will conduct a discussion long-term debt ratio. For long-term

debt, most companies producing raw materials will often depend on short-term debt,

because the business cycle is short, working capital needs are more than long-term

capital, so long-term debt does not affect the rate of return. However, according to

the regression model, the impact of long-term debt is in the same direction as ROA,

showing that the use of long-term debt will have a positive effect on ROA when long-

term debt is at a certain rate. If a business uses debt beyond a certain level, the more

long-term debt is used, the more it will reduce the financial performance of the

business.

Firm size (SIZE)

The variable of firm size is the variable that has the least impact on the dependent

variable ROE, but this is the variable most authors include in the research and most

of them have statistical significance. The coefficient of the variable SIZE greater than

0 shows the positive impact of the firm size factor on performance. According to the

regression results, SIZE has the same impact with the author’s initial expectation of

having the positive effect and the coefficient of the effect is 0.0111925. The results

obtained are similar to the study of Mudambi and Nicosia (1998), Lauterbach and

Vaninsky (1999), Durand and Coeuderoy (2001) and Tzelepis and Skuras (2004).

This positive effect is explained that the larger the size of the business, the greater

the total assets that the enterprise owns, the more effective the performance will be.

Large-scale enterprises will have resources to optimize capital to create better

performance than small-sized enterprises.

The two variables revenue growth have p-values outside the significance levels of

0%, 5% and 10%, so they will not be included in the explanatory model. However,

67

the thesis will still provide discussions on revenue growth factors ratio to provide

policy suggestions for business managers.

Revenue growth (GROWTH)

Revenue growth variable has a positive impact with the dependent variable of

performance. In terms of impact, revenue growth is the weakest independent variable

among the independent variables included in the model. With the impact greater than

0, the revenue growth variable has the same sign of the author’s initial expectation.

Regression results show similar to the predictions of Jensen and Meckling (1976),

Nadaraja et al. (2011), Gill et al (2011). According to the results of the model, the

higher the revenue growth rate, the better the performance of the business. Positive

revenue growth is a good sign that the business can take advantage of capital to

increase profits.

4.3.3. The impact of short-term debt on ROE

Table 4.16: Model regression results of short-term debt for the ROE

Coefficient Variable Pooled OLS FEM REM

-0.0420573 -0.0404933 -0.0415076 NSTD

-0.4253643*** -0.4385748*** -0.4297584*** NSTD2

0.0404848*** 0.0428972*** 0.0412875*** SIZE

0.0020619 0.0017465 0.0019548 GROWTH

-0.3474189*** -0.3760393*** -0.3569409*** Constant

R2 = 4.13% R2 = 4.47% R2 = 4.47%

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

For the results of Pooled OLS regression, two variables NSTD and NSTD2 have a

negative impact on ROE with 99% confidence, while the NSTD variable has no

statistical significance. Both SIZE and GROWTH variables have the same effect on

ROA but GROWTH is not statistically significant, SIZE variable has the same effect

68

with 99% confidence. Regarding the level of influence, NSTD2 is the variable with

the strongest impact on the dependent variable with a β coefficient of -0.4253643.

With an R2 of 4.13%, this means that 4.13% of the variation in the dependent

variable is explained by the independent variable. In general, the explanatory level

of the OLS model is at a relative level.

For the FEM fixed-effects model, two variables NSTD and NSTD2 have a negative

impact on ROE with 99% confidence, while the NSTD variable has no statistical

significance. Both SIZE and GROWTH variables have the same effect on ROE but

GROWTH is not statistically significant, SIZE variable has the same effect with 99%

confidence. Regarding the level of influence, NSTD2 is the variable that has the

strongest impact on the dependent variable with a β coefficient of -0.4385748. The

FEM model has the same level of explanation as the Pooled OLS model with R2 of

4.47%, showing that 4.47% of the change in the dependent variable is explained by

the variables NSTD2 and SIZE.

For the REM random effects model, two variables NSTD and NSTD2 have a

negative impact on ROE with 99% confidence, while the NSTD variable has no

statistical significance. Both SIZE and GROWTH variables have the same effect on

ROE but GROWTH is not statistically significant, SIZE variable has the same effect

with 99% confidence. Regarding the level of influence, NSTD2 is the variable that

has the strongest impact on the dependent variable with a β coefficient of -0.4297584.

The R2 results show that 4.47% of the variation of the dependent variable is explained

by the variables NSTD2 and SIZE.

4.3.3.1. Model selection test

As described in Chapter 3, the thesis will perform tests to choose between Pooled

OLS, FEM and REM models. However, the estimation of the research model

according to the Pooled OLS method does not accurately reflect the separate and

specific impacts of each different enterprise. Therefore, the author will perform

Hausman test to choose the FEM or REM model that will be more suitable for the

69

research data. Table 4.5 shows the results of the selection test between the two

models.

Table 4.17: Model’s selection for short-term debt on the ROE

Dependent variable

Test

Statistical value

P-value

Choose between

Prob>chi2 =

FEM and REM

Chi2(4) = 4.63

ROE

Hausman

0.3276*

 Choose REM

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

 Hausman test to choose between FEM and REM

Hausman test hypotheses:

H0: There is no correlation between variables and residual so REM is more

appropriate.

H1: At least one variable is correlated with the residual so FEM is more relevant.

For the impact of short-term debt on ROE, the test results show that p-value = 0.3276

is greater than the 10% significance level, so the hypothesis H0 is not rejected. So the

REM model will be a more suitable model than the FEM. (Hausman, 1978).

Thus, through the Hausman test, it can be concluded that, for the impact of short-

term debt on ROE, the REM regression model is the most suitable model for the

research data. On the other hand, it is necessary to continue to consider the defect

tests of each model before a final conclusion can be drawn about the accuracy of this

model.

70

4.3.3.2. Model deficiencies test

 Multicollinearity test

Table 4.18: VIF test result for short-term debt on the ROE

ROE

Variables Variance Inflation Factors (VIF)

1.07 NSTD

1.06 SIZE

1.03 NSTD2

1.01 GROWTH

Source: Analysis result from Stata 16.0

In the assumptions of the linear regression model, the independent variables have no

linear relationship. If this assumption is violate, there will be multi-collinearity in the

model, which means the independent variables are interdependent and expressed as

a function. Therefore, before regression variables, the thesis conducts the multi-

collinearity test.

Compared to the correlation matrix analysis method, Variance Inflation Factors

(VIF) is more commonly used in the multi-collinearity test. The result in Table 4.6

shows the VIF coefficient of each factor is less than 10, indicating that the research

model does not violate the multi-collinearity phenomenon (Hoang Trong and Chu

Nguyen Mong Ngoc, 2008). Therefore, the variables chosen for the regression model

are appropriate.

 Heteroskedasticity test

Table 4.19: Heteroskedasticity test result for short-term debt on the ROE

Modified Wald test Statistic value P-value

Chi2(01) = 4.15 Prob>chibar2 = 0.0209** ROE

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

71

The Modified Wald is performed to test the heteroskedasticity in the FEM model

(Christopher, 2001). LM test – Breusch and Pagan Lagrangian Multiplier will be

used to test the heteroskedasticity in the REM regression model (Baltagi, 2008). This

is a phenomenon that occurs when the residual variance is not a constant but

increases or decreases as the independent variable increases. When conducting the

linear regression model, the hypothesis is given that the variance of each random

variable ui under the given value of the explanatory variable Xi is constant. The

appearance of the heteroskedasticity indicates that the regression model is not

accurate or because the model has omitted some independent variables which is

important to the dependent variable (Nguyen Van Ngoc, 2006).

The hypotheses of the Modified Wald test include:

H0: The model does not have the heteroskedasticity phenomenon.

H1: The model has the heteroskedasticity phenomenon.

For ROE, the test results show that p-value = 0.0209 is less than the 5% significance

level, so the hypothesis H0 is rejected. Thus, the REM model has the

 Auto-correlation test

heteroskedasticity phenomenon.

Table 4.20: Auto-correlation test result for short-term debt on the ROE

Wooldridge test Statistic value P-value

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

F(1,9) = 15.368 Prob>F = 0.0035*** ROE

Source: Analysis result from Stata 16.0

The auto-correlation phenomenon is recognized by Wooldridge test. The term auto-

correlation can be understood as the correlation between the components of a series

observations arranged in chronological order. This phenomenon makes R-squared

results, variances and standard errors of prediction become ineffective. The linear

regression model assumes that the auto-correlation does not exist in ui residual.

72

The hypotheses of the Wooldridge test are set out as follows:

H0: The model does not have the auto-correlation.

H1: The model has the auto-correlation.

For ROE, the test results show that p-value = 0.0035 is less than 1% significance

level, so hypothesis H0 is rejected. Thus, for the impact of short-term debt on ROE,

the REM model has auto-correlation.

4.3.3.3. Remedy model’s deficiencies

From the above tests, the model has one deficiency is the auto-correlation

phenomenon. According to Gujarati (2011), how to overcome the above defect for a

model with a large sample (from 30 samples or more) is to select the regression

model according to the feasible generalized least squares method (FGLS). The FGLS

method is actually the ordinary least squares method (OLS) applied to variables that

have been transformed from a model that violates the classical assumptions. Thus,

the estimated parameters from the new model will be more reliable.

Table 4.21 shows that two out of four independent variables are included in the

regression model and are statistically significant, namely NSTD2 and SIZE. Two

variables NSTD2 and SIZE all have p-values less than 1% significance level with

values of 0.2% and 0%, respectively, so they will have statistical significance for the

model.

Table 4.21: FGLS regression model for short-term debt on the ROE

Variables

Constant

NSTD

NSTD2

SIZE

GROWTH

-0.312**

-0.037

-0.485***

0.037***

0.002

Coefficient

Number of observations: 470

Prob > chi2 = 0.0002

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

73

About the influence of the variables show that only the SIZE variable has the same

effect on the ROE ratio, while the other variable NSTD2 has the opposite effect on

ROE. Specifically, ranked in order of the influence of the independent variables on

the ROE variable, the first is the NLTD2 variable with 48.5% and the second is the

SIZE variable with 3.7%. For variables GROWTH and NLTD both have an impact

on ROE but are not statistically significant, so it cannot be explained for the model.

Therefore, the regression model of the thesis is presented as follows:

ROE = − 0.3116454 − 0.4847542*NSTD2 + 0.0021431*SIZE + ε 4.3

4.3.3.4. Discussion about research results

Based on the research results, there are all two independent variables including

NSTD2 and SIZE that affect the performance (ROE) of listed material manufacturers

in Vietnam. The degree of impact of each independent variable on the dependent

variable is different, presented in detail in order from most to least.

Short-term debt ratio (STD)

Based on the regression results, the original of this squared adjusted short-term debt

ratio variable is the short-term debt ratio, so the author will discuss the short-term

debt ratio variable. The effect of short-term debt on ROE is essentially non-linear.

Although the debt ratio of the squared short-term debt ratio variable is statistically

significant, the effect of the short-term debt ratio on ROE will follow an inverted U-

shaped parabol. Since the regression coefficient for the variable short-term debt ratio

is not statistically significant, we have a quadratic function graph y = ax2 + bx + c

with a<0 and b=0 (a is the regression coefficient of the squared short-term debt ratio

-0.485 and b is the regression coefficient of the short-term debt ratio 0). Thus, the

extreme point of the parabolic graph is at the point O, that is, the maximum point of

the above model is 0. However, since the NSTD variable has been adjusted, the value

of the extreme point for the STD variable will be recalculated by adding the mean of

the short-term debt variable to 0.3799659 in Table 4.1. Thus, the extreme point value

will be 0 + 0.380 = 0.380. Thus, an increase in the short-term debt ratio will increase

74

the ROE to the extreme point of 0.380 and will decrease. In fact, in the short term,

an increase in short-term debt will increase profits for businesses, but to a certain

extent, in the long term, increasing debt will increase the burden on businesses and

reduce profits. In contrast to ROA, the use of debt will negatively affect the ROE

variable because the firm substituting debt for equity will further reduce the

denominator of ROE. Therefore, increasing the long-term debt ratio will reduce the

financial efficiency of enterprises.

Firm size (SIZE)

The variable of firm size is the one that has the least impact on the dependent variable

ROA, but this is the variable most authors include in the test and most are statistically

significant. The coefficient of the variable SIZE greater than 0 shows the positive

impact of the firm size factor on performance. According to the regression results,

SIZE has the same impact with the author’s initial expectation of having the positive

effect and the coefficient is 0.0021431. The results obtained are similar to those of

Mudambi and Nicosia (1998), Lauterbach and Vaninsky (1999), Durand and

Coeuderoy (2001) and Tzelepis and Skuras (2004). This positive effect is explained

that the larger the size of the business, the greater the total assets that the enterprise

owns, the more effective the performance will be. Large-scale enterprises will have

resources to optimize capital to create better performance than small-sized ones.

The revenue growth variable has p-values outside the 0%, 5% and 10% significance

levels, so they will not be included in the explanatory model. However, the thesis

will still provide discussions on revenue growth factors to provide policy suggestions

for managers.

Revenue growth (GROWTH)

Revenue growth variable has a positive impact with the dependent variable of

performance. In terms of impact, revenue growth is the weakest independent variable

among the independent ones included in the model. With the impact dimension

greater than 0, the revenue growth variable has the right sign of the author’s initial

75

expectation. Regression results show similar to the predictions of Jensen and

Meckling (1976), Nadaraja et al. (2011), Gill et al (2011). According to the results

of the model, the higher the revenue growth rate, the better the performance of the

business. Positive revenue growth is a good sign that the business can take advantage

of capital to increase profits.

4.3.4. The impact of long-term debt on ROE

Table 4.22: Regression results for long-term debt on the ROE

Coefficient Variable Pooled OLS FEM REM

-0.0452866 -0.0521368 -0.0496091 NLTD

-0.2478944 -0.2512625 -0.250277 NLTD2

0.0409222*** 0.0437869*** 0.0427551*** SIZE

0.0016923 0.0013488 0.0014685 GROWTH

-0.3638505 ** -0.3983384*** -0.3859127*** Constant

R2 = 3.15% R2 = 3.52% R2 = 3.52%

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

For the Pooled OLS regression results, two variables NLTD and NLTD2 have a

negative impact on ROE, but both NLTD and NSTD2 variables are not statistically

significant. Both SIZE and GROWTH variables have the same effect on ROE but

GROWTH is not statistically significant, SIZE variable has the same effect with 99%

confidence. Regarding the level of influence, SIZE affects the dependent variable

with a β coefficient of 0.0409222. With an R2 of 3.15%, this means that 3.15% of the

variation in the dependent variable is explained by the independent variable. In

general, the explanatory level of the OLS model is low.

For the FEM fixed-effects model, two variables NLTD and NLTD2 have a negative

impact on ROE, however, both NLTD and NLTD2 variables are not statistically

significant. Both SIZE and GROWTH variables have the same effect on ROE but

76

GROWTH is not statistically significant, SIZE variable has the same effect with 99%

confidence. Regarding the level of influence, SIZE is the variable that has an impact

on the dependent variable with a β coefficient of 0.0437869. The FEM model has the

same level of explanation as the Pooled OLS model with an R2 of 3.52%, showing

that 3.52% of the variation of the dependent variable is explained by the SIZE

variable.

For the REM random effects model, two variables NLTD and NLTD2 have a

negative effect on ROE, but both NLTD and NLTD2 variables are not statistically

significant. Both SIZE and GROWTH variables have the same effect on ROE but

GROWTH is not statistically significant, SIZE variable has the same effect with 99%

confidence. Regarding the level of influence, SIZE is the variable that has an impact

on the dependent variable with a β coefficient of 0.0427551. The R2 results show that

3.52% of the variation of the dependent variable is explained by the SIZE variable.

4.3.4.1. Model selection test

As described in Chapter 3, the thesis will perform tests to choose between Pooled

OLS, FEM and REM models. To choose between Pooled OLS and FEM models, the

author performs F-test. And the fit of REM model is verified through Hausman test

when compared with FEM model and LM test (Breusch-Pagan) Lagrange multiplier

test) when compared with the OLS model. Table 4.12 shows the results of the

selection test between the three models.

Table 4.23: Model’s selection for long-term debt on the ROE

Dependent variable

Test

Statistical value

P-value

Choose between

Prob>chi2 =

FEM and REM

Chi2(4) = 0.86

ROE

Hausman

0.9296*

 Choose REM

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

77

 Hausman test to choose between FEM and REM

Hausman test hypotheses:

H0: There is no correlation between variables and residual so REM is more

appropriate.

H1: At least one variable is correlated with the residual so FEM is more relevant.

For ROE, the test results show that p-value = 0.9296 is greater than the 10%

significance level, so hypothesis H0 is not rejected. So the REM model will be a more

suitable model than the FEM model (Hausman, 1978).

Thus, through the Hausman test, it can be concluded that, for the impact of long-term

debt on ROE, the REM regression model is the most suitable model for our research

data. On the other hand, it is necessary to continue to consider the defect tests of each

model before a final conclusion can be drawn about the accuracy of this model.

4.3.4.2. Model deficiencies test

 Multicollinearity test

Table 4.24: VIF test result for long-term debt on the ROE

ROE

Variables Variance Inflation Factors (VIF)

3.42 NLTD

3.00 NLTD2

1.25 SIZE

1.00 GROWTH

Source: Analysis result from Stata 16.0

In the assumptions of the linear regression model, the independent variables have no

linear relationship. If this assumption is violate, there will be multi-collinearity in the

model, which means the independent variables are interdependent and expressed as

78

a function. Therefore, before regression variables, the thesis conducts the multi-

collinearity test.

Compared to the correlation matrix analysis method, Variance Inflation Factors

(VIF) is more commonly used in the multi-collinearity test. The result in Table 4.12

shows the VIF coefficient of each factor is less than 10, indicating that the research

model does not violate the multi-collinearity phenomenon (Hoang Trong and Chu

Nguyen Mong Ngoc, 2008). Therefore, the variables chosen for the regression model

are appropriate.

 Heteroskedasticity test

Table 4.25: Heteroskedasticity test result for long-term debt on the ROE

LM – Breusch and Pagan

Statistic value P-value

Lagrangian Multiplier test

Chi2(01) = 4.26 Prob>chibar2 = 0.0196** ROE

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

Modified Wald test was performed to test the heteroskedasticity in the FEM

regression model (Christopher, 2001). LM – Breusch and Pagan Lagrangian

Multiplier test will be used to test the heteroskedasticity in the REM regression

model (Baltagi, 2008). This is a phenomenon that occurs when the variance of the

error is not constant but increases or decreases as the independent variable increases.

When studying the classical linear regression model, the hypothesis is made that the

variance of each random variable ui under the condition that the given value of the

regressor Xi is constant. The appearance of the heteroskedasticity shows that the

regression model is not accurate or because the model has omitted some important

independent variables for the dependent variable (Nguyen Van Ngoc, 2006).

79

The hypotheses of the LM – Breusch and Pagan Lagrangian Multiplier test:

H0: The REM model does not have the heteroskedasticity phenomenon.

H1: The REM model has the heteroskedasticity phenomenon.

For ROE, the test results show that p-value = 0.0196 is less than the 5% significance

level, so hypothesis H0 is rejected. Thus, the REM model has the heteroskedasticity

phenomenon.

 Auto-correlation test

Table 4.26: Auto-correlation test result for long-term debt on the ROE

Wooldridge test Statistic value P-value

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

F(1,9) = 3.347 Prob>F = 0.1006* ROE

Source: Analysis result from Stata 16.0

The auto-correlation phenomenon is recognized by Wooldridge test. The term auto-

correlation can be understood as the correlation between the components of a series

observations arranged in chronological order. This phenomenon makes R-squared

results, variances and standard errors of prediction become ineffective. The linear

regression model assumes that the auto-correlation does not exist in ui residual.

The hypotheses of the Wooldridge test are set out as follows:

H0: The model does not have the auto-correlation.

H1: The model has the auto-correlation.

For ROE, the test results show that p-value = 0.1006 is greater than 10% significance

level, so hypothesis H0 is not rejected. Thus, for the impact of long-term debt on

ROE, the REM model has auto-correlation

80

4.3.4.3. Remedy model’s deficiencies

From the above tests, the model has one deficiency is the auto-correlation

phenomenon. According to Gujarati (2011), how to overcome the above defect for a

model with a large sample (from 30 samples or more) is to select the regression

model according to the feasible generalized least squares method (FGLS). The FGLS

method is actually the ordinary least squares method (OLS) applied to variables that

have been transformed from a model that violates the classical assumptions. Thus,

the estimated parameters from the new model will be more reliable.

Table 4.27 shows that the only independent variable included in the regression model

and statistically significant is SIZE. SIZE variables all have p-values less than 1%

significance level with a value of 0.1%, so it will be statistically significant for the

model.

Table 4.27: FGLS regression model for long-term debt on the ROE

Variables

Constant

NLTD

NLTD2

SIZE

GROWTH

-0.346**

-0.047

-0.276

0.039***

0.002

Coefficient

Number of observations: 470

Prob > chi2 = 0.0030

Note: ***, **, * denote significant level at 1%, 5%, 10% respectively

Source: Analysis result from Stata 16.0

Comments on the direction of the influence of the variables show that, only the SIZE

variable has the same effect on the ROE ratio, while the other two variables NLTD

and NLTD2 have the opposite effect on ROE but not statistically significant for the

model. Regarding the level of impact, the SIZE factor with 1% has a coefficient of

0.039. For the GROWTH variable, there is an impact on ROE but it is not statistically

significant, so it cannot be explained for the model.

Therefore, the regression model of the thesis is presented as follows:

ROE = − 0.3465297 + 0.0394443*SIZE + ε 4.4

81

4.3.4.4. Discussion about research results

Based on the research results, there is only one variable SIZE that affects the financial

performance (ROE) of the listed material manufacturers in Vietnam.

Firm size (SIZE)

The variable of firm size is the variable that has the least impact on the dependent

variable ROE, but this is the variable most authors include in the research and most

of them have statistical significance. The coefficient of the variable SIZE greater than

0 shows the positive impact of the firm size factor on performance. According to the

regression results, SIZE has the same impact with the author’s initial expectation of

having the positive effect and the coefficient of the effect is 0.0394443. The results

obtained are similar to the study of Mudambi and Nicosia (1998), Lauterbach and

Vaninsky (1999), Durand and Coeuderoy (2001) and Tzelepis and Skuras (2004).

This positive effect is explained that the larger the size of the business, the greater

the total assets that the enterprise owns, the more effective the performance will be.

Large-scale enterprises will have resources to optimize capital to create better

performance than small-sized enterprises.

Long-term debt ratio (LTD)

Based on the regression results, the long-term debt variable has a negative impact on

ROE but is not statistically significant. For long-term debt, most companies

producing raw materials will often depend on short-term debt, because the business

cycle is short, working capital needs are more than long-term capital, so long-term

debt does not affect the rate of return. However, according to the regression model,

the impact direction of long-term debt is in the same direction as ROE, showing that

the use of long-term debt will have a positive effect on ROE when long-term debt is

at a certain rate. If a business uses debt beyond a certain level, the more long-term

debt is used, the more it will reduce the financial performance of the business.

The two variables revenue growth have p-values outside the significance levels of

0%, 5% and 10%, so they will not be included in the explanatory model. However,

82

the thesis will still provide discussions on revenue growth factors ratio to provide

policy suggestions for business managers.

Revenue growth (GROWTH)

Revenue growth variable has a positive impact with the dependent variable of

performance. In terms of impact, revenue growth is the weakest independent variable

among the independent variables included in the model. With the impact greater than

0, the revenue growth variable has the same sign of the author’s initial expectation.

Regression results show similar to the predictions of Jensen and Meckling (1976),

Nadaraja et al. (2011), Gill et al (2011). According to the results of the model, the

higher the revenue growth rate, the better the performance of the business. Positive

revenue growth is a good sign that the business can take advantage of capital to

increase profits.

Thus, based on the regression results, the impact of short-term debt ratio on financial

performance is a non-linear relationship. This result is similar to the original

hypothesis of the author. This shows that the increase in short-term debt will help

increase the financial efficiency of the business in terms of profitability. However,

the increase will reach a certain level which is the optimal capital structure, where,

as the business increases the short-term debt, the profitability will decrease. For long-

term debt, because most enterprises use a very small ratio of long-term debt, only

about 10%, the increase in long-term debt also has an impact on financial

performance but follows a linear relationship. . The more long-term debt a business

has, the lower its profitability. Long-term debt is used at a higher cost than short-

term debt, thereby increasing costs and reducing the profitability of the business. For

two control variables are firm size and revenue growth. Both of these variables show

a positive impact on financial performance. This can be explained when the business

increases in size or has good revenue growth, which will affect its reputation and

competitiveness in the market. Therefore, creating better profitability than other

businesses.

83

Table 4.28: Overall comparison results

Significance

Dependent variable

Independent variables

Research hypothesis

Research results

STD

Statistical significance

ROA

LTD SIZE GROWTH

Statistical significance Statistical significance No statistical significance

STD

No statistical significance

ROE

LTD SIZE GROWTH

U-shaped reverse Linear Positive Positive U-shaped reverse Linear Positive Positive

U-shaped reverse Linear Positive Positive U-shaped reverse Linear Positive Positive

No statistical significance Statistical significance No statistical significance

Source: Author’s summary

84

Chapter 4’s summary

In chapter 4, the thesis presents and analyzes research results, including descriptive

statistics, correlation matrix analysis between variables in the model and regression

by Pooled OLS, FEM, REM methods. After choosing a model for this research, the

thesis tests some defects of the model and remedies them by using the FGLS model.

Thus, based on the regression results, the impact of short-term debt ratio on financial

performance is a non-linear relationship. For long-term debt, because most

enterprises use a very small ratio of long-term debt, only about 10%, the increase in

long-term debt also has an impact on financial performance but follows a linear

relationship. The more long-term debt a business has, the lower its profitability. For

two control variables are firm size and revenue growth. Both of these variables show

a positive impact on financial performance. In addition, the author also determines

the extreme point of the model to indicate the optimal capital level of the enterprise

when using debt in the capital structure.

85

CHAPTER 5: CONCLUSION AND POLICY

RECOMMENDATION

In this chapter, the thesis will summarize the research results in chapter 4, and based

on those conclusions to suggest reasonable policies for businesses. The last part of

the chapter will be the limitations of the thesis when implementing this topic.

5.1. Conclusion

With 470 observations from 47 material manufacturing enterprises listed on HoSE

in the period 2012 to 2021, the regression model results are as follows. The impact

of short-term debt ratio on financial performance is the non-linear relationship. This

result is similar to the original hypothesis of the author. This shows that the increase

in short-term debt will help increase the financial efficiency of the business in terms

of profitability. However, the increase will reach a certain level which is the optimal

capital structure, where, as the business increases the short-term debt, the

profitability will decrease. This result is similar to the research results of Margaritis

and Psillaki (2010), Do Van Thang and Trinh Quang Thieu (2010); Doan Vinh

Thang, 2016) and Berzkalne (2015). For long-term debt, because most enterprises

use a very small ratio of long-term debt, only about 10%, the increase in long-term

debt also has an impact on financial performance but follows a linear relationship.

The more long-term debt a business has, the lower its profitability. This result is

similar to the research results of Sedeaq Nassar (2021), Mireku et al. (2014) and

Sheikh et al. (2013). Long-term debt is used at a higher cost than short-term debt,

thereby increasing costs and reducing the profitability of the business. For two

control variables are firm size and revenue growth. Both of these variables show a

positive impact on financial performance. This can be explained when the business

increases in size or has good revenue growth, which will affect its reputation and

competitiveness in the market. Therefore, creating better profitability than other

businesses.

86

For short-term debt, according to the traditional point of view, the optimal capital

structure has been proven and shown through the research results of the thesis. In the

short term, the increase in short-term debt will increase the profit for the business,

but to a certain extent, in the long term, the increase in debt will increase the burden

on the business and reduce the profit of the business.

For short-term debt, the thesis has demonstrated and clarified the trade-off theory.

This theory indicates that firms must weigh the benefits from the tax shield and the

costs of financial distress to choose the one that maximizes the firm’s value.

However, this theory also has limitations that (1) it is difficult to accurately determine

the costs of financial distress, especially indirect costs such as loss of customers, loss

of suppliers, loss of reputation and reputation; (2) in practice many large and

successful firms still use debt much lower than the optimal debt ratio determined by

theory. The increased use of debt will generate income from tax shield benefits,

however, up to a certain percentage, the increase in debt will become a burden for

the business in the long run.

Based on the regression results, the impact of short-term debt ratio on financial

performance is a non-linear relationship. This result is similar to the original

hypothesis of the author. This shows that the increase in short-term debt will help

increase the financial efficiency of the business in terms of profitability. However,

the increase will reach a certain level which is the optimal capital structure, where,

as the business increases the short-term debt, the profitability will decrease. For long-

term debt, because most enterprises use a very small ratio of long-term debt, only

about 10%, the increase in long-term debt also has an impact on financial

performance but follows a linear relationship. The more long-term debt a business

has, the lower its profitability. Long-term debt is used at a higher cost than short-

term debt, thereby increasing costs and reducing the profitability of the business. For

long-term debt, most companies producing raw materials will often depend on short-

term debt, because the business cycle is short, working capital needs are more than

long-term capital, so long-term debt does not affect the rate of return.

87

5.2. Recommendations on capital structure for enterprises

The author will make some recommendations based on factors affecting capital

structure including the short-term debt ratio, the long-term debt ratio, the firm size

and the revenue growth.

 Short-term debt

As the statistics description in chapter 3, short-term debt accounts for the majority of

the capital structure of each Vietnamese enterprise. Specifically, short-term debt

accounted for 40%, while long-term debt only accounted for about 10%. Although

the use of short-term debt brings benefits in terms of capital mobilization at low cost,

enterprises must ensure business continuity. Working capital rotation ensures

efficiency and accuracy to help businesses not face liquidity risks. According to the

regression results, short-term debt has a non-linear effect on both the dependent

variables ROE and ROA. The use of debt will help the business create a tax shield

but if used at an optimal rate. Therefore, businesses need to consider a reasonable

debt ratio to optimize performance.

 Long-term debt

Since most of the raw material manufacturing firms use more short-term debt, most

of the long-term debt in the study does not show a clear impact on both ROA and

ROE dependent variables. When there is a need to use long-term debt, businesses

need to meet specific conditions in terms of collateral, which are also difficulties for

newly established companies. In addition, long-term debt has a high interest rate and

a long payback period, so there will be certain risks. Therefore, the use of long-term

debt can negatively affect the profitability of the business if the business does not

manage cash flow well. Thus, businesses need to consider and have a specific plan

when using long-term debt to avoid incurring liquidity risks.

 Firm size

According to the research results, firm size has a positive impact on performance, as

measured by profitability ratio. The larger the size of the enterprise, the higher the

88

degree of risk diversification, thereby reducing the risk for the business. Enterprises

need to invest in a portfolio of high-return assets with an acceptable level of risk, in

accordance with the risk appetite of the organization. The increase in the assets of

the enterprise also needs to be ensured in accordance with the existing capital sources

and in accordance with the development goals of each period.

 Revenue growth

According to the research results, the growth rate of revenue is not statistically

significant for the profit margin. Therefore, the author does not have enough basis to

conclude on the relationship between revenue growth variable and performance for

raw material manufacturing enterprises. However, in fact, increasing revenue at a

stable rate over the years will ensure the reputation of the business in the market. The

increase in prestige will make it easier for businesses to raise capital, thereby

strengthening their competitive advantage in the market. This benefit will increase

the profit for the business.

5.3. Recommendations for relevant organizations

Banks are one of the organizations that have a close relationship with almost any

business. Businesses need to be facilitated to access the bank’s capital to the fullest

extent. Therefore, banks need to develop the most diversified lending products to

create conditions for enterprises to have many financing options for business

activities from time to time. When businesses receive capital support from banks at

a reasonable cost, they will also stabilize their long-term business activities and

increase financial efficiency. In addition to providing a variety of loan packages,

banks also need to have a comprehensive financial management suitable for the

enterprises condition to help them expand their scale and ensure long-term and

sustainable operations. For each business, a good arrangement of funding sources

will ensure the efficiency because good liquidity and appropriate costs will be

decisive factors for the existence and success of the business. In order to have the

best coordination, the State also needs to issue preferential policies for new, small-

89

scale enterprises or those with technological development orientation. The support

from the State is equally important for every enterprise in Vietnam.

5.4. Limitations of the topic

In the process of implementation, the thesis has some limitations and these

limitations will open new directions for further researches. Firstly, the scope of the

study is limited. In this topic, the thesis only conducts research for one type of

business industry, which is the field of raw material production, has not been tested

for other industries. Secondly, is the limitation on the tested factors. The thesis only

tests on factors directly related to capital structure, which is the debt ratio variable,

for performance, the state ownership ratio also has an impact on performance.

Thirdly, performance in this thesis only mentions one factor, profitability, because

this is a fairly broad definition, so future studies can add value efficiency of

enterprises on the market.

90

Chapter 5’s summary

Based on the research results of chapter 4, a number of recommendations have been

made for businesses and regulators, in order to better control the performance of

business operations, namely ROA and ROE based on four influencing factors are

short-term debt ratio, long-term debt ratio, business size and revenue growth. In

addition, the thesis has given some suggestions to help improve profits for businesses

and help businesses get a reasonable capital structure by using an appropriate ratio

of short-term debt and long-term debt and through the influencing factors author

propose policies to improve the performance of Vietnamese enterprises. And the last

part of the thesis is the limitations of the topic to open up new directions for further

researches.

i

REFERENCES

English references

1. Abor Joshua (2005). The effect of capital structure on profitability: an empirical

analysis of listed firms in Ghana. The Journal of Risk Finance, ISSN: 1526-5943,

Vol. 6 (5), pp. 438 – 445.

2. Abor Joshua (2007). Debt policy and performance of SMEs: Evidence from

Ghanaian and South African firms. Journal of Risk Finance, ISSN: 1526-5943, Vol.

8 (4), pp. 364 – 379.

3. Ahmad Noryati and Rahim Fahmi Abdul (2013). Theoretical Investigation on

Determinants of Government-Linked Companies Capital Structure. Journal of

Accounting, Finance and Economics, Vol. 3 (2), pp. 72 – 85.

4. Ahmad Zuraidah, Abdullah Norhasniza Mohd Hasan and Roslan Shashazrina

(2012). Capital Structure Effect on Firms Performance: Focusing on Consumers and

Industrials Sectors on Malaysian Firms. International Review of Business Research

Papers, Vol. 8 (5), pp. 137 – 155.

5. Amran Noor Afza and Ahmad Che Ahmad (2011). Board Mechanisms and

Malaysian Family Companies’ Performance. Asian Journal of Accounting and

Governance 2, ISSN 2180-3838, pp. 15 – 26.

6. Anderson Ronald C. and Reeb David M. (2004). Board Composition: Balancing

Family Influence in S&P 500 Firms. Administrative Science Quarterly, Vol. 49 (2),

pp. 209 – 237.

7. Ang James S., Rebel Allen Cole, and James Wuh Lin (2000). Agency Costs and

Ownership Structure. The Journal of Finance, Vol. 55 (1), pp 81 – 106.

ii

8. Azhagaiah Ramachandran and Gavoury Candasamy (2011). The Impact of Capital

Structure on Profitability with Special Reference to IT Industry in India vs. Domestic

Products. Managing Global Transitions, Vol. 9 (4), pp. 371 – 392.

9. Baltagi Badi H. (2008). Econometric Analysis of Panel Data. John Wiley & Sons

Ltd., Chichester.

10. Berk Jonathan Berk and DeMarzo Peter (2014). Corporate finance, Pearson,

Boston.

11. Berzkalne Irina (2015). Company’s Capital Structure and Value: a Panel

Threshold Regression Analysis. Applied Economics: Systematic Research, Vol. 9

(1), pp 77 - 94.

12. Chadha Saurabh and Sharma Anil K. (2015). Capital Structure and Firm

Performance: Empirical Evidence from India. Vision, Vol. 19 (4), pp. 295 – 302.

13. Chakravarthy Balaji S. (1986). Measuring Strategic Performance. Strategic

Management Journal, Vol. 7 (5), pp. 437 – 458.

14. Cheng Yu-Shu, Liu Yi-Pei, and Chien Chu-Yang. (2010). Capital structure

and firm value in China: A panel threshold regression analysis. African Journal of

Business Management, ISSN 1993-8233, Vol. 4 (12), pp. 2500 – 2507.

15. Choi Jae-Kyu, Yoo Seung-Kyu, Kim Ju-Hyung and Kim Jae-Jun (2014).

Capital Structure Determinants among Construction Companies in South Korea: A

Quantile Regression Approach. Journal of Asian Architecture and Building

Engineering, Vol. 13 (1), pp. 93 – 100.

16. Christopher F. Baum (2001). Residual diagnostics for cross-section time

series regression models. Stata Journal, Vol. 1 (1), pp. 101 - 104.

iii

17. Coeurderoy Regis and Durand Rodolphe (2001). Leveraging the advantage of

early entry: Proprietary technologies versus cost leadership. Journal of Business

Research, Vol. 57 (6), pp. 583 – 590.

18. Damodaran Aswath (2011). Applied Corporate Finance : A User’s Manual

Ed. 3, John Wiley & Sons.

19. Demsetz Harold and Lehn Kenneth (1985). The Structure of Corporate

Ownership: Causes and Consequences. Journal of Political Economy, Vol. 93 (6),

pp. 1155 – 1177.

20. Ebaid Ibrahim El-Sayed (2009). The Impact of Capital Structure Choice on

Firm Performance: Empirical Evidence from Egypt. Journal of Risk Finance,

ISSN: 1526-5943, Vol. 10 (5), pp. 477 – 487.

21. Gill Amarjit, Biger Nahum and Mathur Neil (2011). The effects of capital

structure on profitability: Evidence from United States. International Journal of

Management, Vol. 28 (4), pp. 3 – 15.

22. Gill Amarjit and Mathur Neil (2011). Factors that Influence Financial

Leverage of Canadian Firms. Journal of Applied Finance & Banking, ISSN: 1792-

6580, Vol. 1 (2), pp. 19 - 37.

23. Gill Amarjit, Biger Nahum, Pai Chenping and Bhutani Smita (2009). The

Determinants of Capital Structure in the Service Industry: Evidence from United

States. The Open Business Journal, Vol. 2 (1), pp 48 – 53.

24. Gleason Kimberly C., Knowles Lynette Mathur and Mathur Ike (2000). The

Interrelationship Between Culture, Capital Structure, and Performance: Evidence

from European Retailers. Journal of Business Research, Vol. 50 (2), pp. 185 – 191.

25. Gorton Gary and Rosen Richard (1995). Corporate Control, Portfolio Choice,

and the Decline of Banking. The Journal of Finance, Vol. 50 (5), pp. 1377 – 1420.

iv

26. Gruber Martin J. and Warner Jerold B. (1977). Bankruptcy Costs: Some

Evidence. The Journal of Finance, Vol. 32 (2), pp. 337 – 347.

27. Gujarati (2011). The Fundamentals of Corporate Finance. Paperback.

28. Hausman J. A. (1978). Specification tests in Econometrics. Econometrica,

Vol. 46 (6), pp. 1251 - 1271.

29. Hofer Charles W. and Sandberg William R. (1987). Improving New Venture

Performance: Some Guidelines for Success. American Journal of Small Business,

Vol. 12 (1), pp. 11 – 26.

30. Jensen Michael C. and Meckling William H. (1976). Theory of the firm:

Managerial behavior, agency costs and ownership structure. Journal of Financial

Economics, Vol. 3 (4), pp. 305 – 360.

31. Kim Wi Saeng and Sorensen Eric H. (1986). Evidence on the Impact of the

Agency Costs of Debt on Corporate Debt Policy. The Journal of Financial and

Quantitative Analysis, Vol. 21 (2), pp. 131 – 144.

32. Lauterbach Beni and Vaninsky Alexander (1999). Ownership Structure and

Firm Performance: Evidence from Israel. Journal of Management and Governance,

Vol. 3, pp. 189 – 201.

33. Margaritis Dimitris and Psillaki Maria (2010). Capital structure, equity

ownership and firm performance. Journal of Banking & Finance, Vol. 34 (3), pp.

621 – 632.

34. Mathur Ike, Singh Manohar, and Gleason Kimberly C. (2001). The evidence

from Canadian firms on multinational diversification and performance. The

Quarterly Review of Economics and Finance, Vol. 41 (4), pp. 561 – 578.

35. Mehran Hamid (1995). Executive compensation structure, ownership, and

firm performance. Journal of Financial Economics, Vol. 38 (2), pp. 163 – 184.

v

36. Mireku Kwame, Samuel Mensah and Ogoe Emmanuel (2014). The

Relationship between Capital Structure Measures and Financial Performance:

Evidence from Ghana. International Journal of Business and Management, ISSN

1833-3850, Vol. 9 (6), pp. 151.

37. Modigliani Franco and Miller Merton H. (1958). The Cost of Capital,

Corporation Finance and the Theory of Investment. The American Economic Review,

Vol. 48 (3), pp. 261 – 297.

38. Mudambi Ram and Nicosia Carmela (1998). Ownership structure and firm

performance: evidence from the UK financial services industry. Applied Financial

Economics, Vol. 8 (2), pp. 175 – 180.

39. Murphy Gregory B., Trailer Jeff W. and Hill Robert C. (1996). Measuring

performance in entrepreneurship research. Journal of Business Research, Vol. 36 (1),

pp. 15 – 23.

40. Nadaraja Punitharaja, Zulkafli Abdul Hadi and Masron Tajul Ariffin (2011).

Family Ownership, Firm’s Financial Characteristics and Capital Structure: Evidence

from Public Listed Companies in Malaysia. Economia Seria Management, Vol. 14

(1), pp. 141 – 156.

41. Nassar Sedeaq (2021). The Impact of Capital Structure on Financial Firm

Performance of Palestinian Listed Companies. International Journal of Academic

Research in Accounting Finance and Management Sciences, Vol. 5 (8), pp. 56 – 61.

42. Olokoyo Felicia Omowunmi (2013). Capital Structure and Corporate

Performance of Nigerian Quoted Firms: A Panel Data Approach. African

Development Review, Vol. 25 (3), pp. 358 - 369.

43. Pettit R. Richardson and Singer Ronald F. (1985). Small Business Finance: A

Research Agenda. Financial Management, Autumn, pp. 47 – 61.

vi

44. Pouraghajan Abbasali and Malekian Esfandiar (2012). The Relationship

between Capital Structure and Firm Performance Evaluation Measures: Evidence

from the Tehran Stock Exchange. International Journal of Business and Commerce,

ISSN: 2225-2436, Vol. 1 (9), pp. 166 – 181.

45. Quang Do and Xin Wu (2014). The Impact of Ownership Structure and

Capital Structure on Financial Performance of Vietnamese Firms. International

Business Research, ISSN: 1913-9004, Vol. 7 (2), pp. 64.

46. Salim Mahfuzah and Yadav Raj Dr. (2012). Capital Structure and Firm

Performance: Evidence from Malaysian Listed Companies. Procedia - Social and

Behavioral Sciences, Vol. 65 (3), pp. 156 – 166.

47. Samuel Antwi, Atta Mills E. and Zhao Xicang. (2012). Capital structure and

firm value: Empirical evidence from Ghana. International Journal of Humanities

and Social Science, Vol. 3 (22), pp. 103 – 111.

48. Sheikh Nadeem Ahmed and Wang Zongjun (2013). The impact of capital

structure on performance: An empirical study of non-financial listed firms in

Pakistan. International Journal of Commerce and Management, ISSN: 1056-9219,

Vol. 23 (4), pp. 354 - 368.

49. Shubita Mohammad Fawzi and Alsawalhah Jaafer Maroof (2012). The

Relationship between Capital Structure and Profitability. International Journal of

Business and Social Science, Vol. 3 (16).

50. Tzelepis Dimitris and Skuras Dimitris (2004). The effects of regional capital

subsidies on firm performance: An empirical study. Journal of Small Business and

Enterprise Development, ISSN: 1462-6004, Vol. 11 (1), pp. 121 – 129.

Vietnamese references

51. Hoang Trong và Chu Nguyen Mong Ngoc (2008). Analyze research data with

SPSS. Hong Duc Publisher.

vii

52. Le Thi Phuong Vy (2015). Ownership structure, capital structure and firm

performance : a study of Vietnamese listed firms, Australian Economic Papers, Vol.

55 (4), pp. 319 – 344.

53. Nguyen Van Ngoc (2006). Macroeconomic lecture. Publisher of National

Economics University.

54. Oanh Tran Thi Kim and Anh Hoang Thi Phuong (2017). Capital structure and

business performance of processing industry in Vietnam. Asian Journal of

Economics and Banking, Vol. 132, pp. 44 – 44.

55. Thang Doan Vinh (2016). Moi quan he giua cau truc von va hieu qua kinh

doanh: Vai tro tuong tac cua so huu nha nuoc. AGU International Journal of

Sciences, ISSN: 0866-8086, Vol. 10 (2), pp. 20 - 31.

56. Thang Do Van and Thieu Trinh Quang (2010). Anh huong cua cau truc von

len gia tri doanh nghiep cua cac cong ty niem yet tren san chung khoan Thanh pho

Ho Chi Minh. Journal of Asian Business and Economic Studies, Vol. 238, pp. 36 –

41.

57. Annual reports, financial statements of enterprises in the research sample from

2012 to 2021 according to Appendix 01

viii

APPENDIX 1

LIST OF LISTED MATERIAL MANUFACTURING ENTERPRISES

ON HO CHI MINH CITY STOCK EXCHANGE

Name

No. Code 1 2 3 4 5 6 7 8

9

CTI

AAA AN PHAT BIOPLASTICS JSC ABS BINHTHUAN AGRICULTURE SERVICES JOINT STOCK COMPANY ACC ACC BINH DUONG INVESTMENT AND CONSTRUCTION JSC APH AN PHAT HOLDINGS JOINT STOCK COMPANY BFC BINH DIEN FERTILIZER JOINT STOCK COMPANY BMC BINH DINH MINERALS JOINT STOCK COMPANY CRC CREATE CAPITAL VIET NAM JOINT STOCK COMPANY CSV SOUTH BASIC CHEMICALS JOINT STOCK COMPANY CUONG THUAN IDICO DEVELOPMENT INVESTMENT COPORATION

10 CVT CMC JOINT STOCK COMPANY 11 DCM PETRO VIET NAM CA MAU FERTILIZER JOINT STOCK COMPANY 12 DHA HOA AN JOINT STOCK COMPANY 13 DHC DONG HAI JOINT STOCK COMPANY OF BENTRE 14 DHM DUONG HIEU TRADING AND MINING JOINT STOCK COMPANY 15 DPM PETROVIETNAM FERTILIZER AND CHEMICALS CORPORATION 16 DPR DONG PHU RUBBER JOINT STOCK COMPANY 17 DTL DAI THIEN LOC CORPORATION 18 DTT DO THANH TECHNOLOGY CORPORATION 19 DXV DA NANG BUILDING MATERIAL VICEM JOINT STOCK COMPANY 20

FCM FECON MINING JOINT STOCK COMPANY

21 GAB

FLC MINING INVESTMENT & ASSET MANAGEMENT JOINT STOCK COMPANY

22 GTA THUAN AN WOOD PROCESSING JOINT STOCK COMPANY 23 GVR VIET NAM RUBBER GROUP - JOINT STOCK COMPANY 24 HAI HAI AGROCHEM JOINT STOCK COMPANY 25 HAP HAPACO CORPORATION 26 HHP HAI PHONG HOANG HA PAPER JOINT STOCK COMPANY HII AN TIEN INDUSTRIES JOINT STOCK COMPNAY 27 28 HMC VNSTEEL – HO CHI MINH CITY METAL CORPORATION 29 HPG HOA PHAT GROUP JOINT STOCK COMPANY 30 HRC HOA BINH RUBBER JOINT STOCK COMPANY 31 HSG HOA SEN GROUP 32 HT1 HATIEN CEMENT JOINT-STOCK COMPANY 33 HVX VICEM HAI VAN CEMENT JOINT STOCK COMPANY 34 KPF HOANG MINH FINANCE INVESTMENT JOINT STOCK COMPANY

35 KSB

36 LBM

37 LCM

BINH DUONG MINERAL AND CONSTRUCTION JOINT STOCK COMPANY LAMDONG MINERALS AND BUILDING MATERIALS JOIN STOCK COMPANY LAO CAI MINERAL EXPLOITATION & PROCESSING JOINT STOCK COMPANY

ix

38 MCP MYCHAU PRINTING AND PACKAGING CORPORATION 39 NAV NAMVIET JOINT STOCK COMPANY 40 NHH HANOI PLASTICS JOINT STOCK COMPANY 41 NKG NAM KIM STEEL JOINT STOCK COMPANY 42 NNC NUI NHO STONE JOINT STOCK COMPANY 43

PHR PHUOC HOA RUBBER JOINT STOCK COMPANY

44

PLP

PHA LE PLASTICS MANUFACTURING AND TECHNOLOGY JOINT STOCK COMPANY

POM POMINA STEEL CORPORATION PTB PHU TAI JOINT STOCK COMPANY

45 46 47 QBS QUANG BINH IMPORT EXPORT JOINT STOCK COMPANY 48 RDP RANG DONG HOLDING PLASTIC JOINT STOCK COMPANY SFG THE SOUTHERN FERTILIZER JOINT STOCK COMPANY 49 SMC SMC TRADING INVESTMENT JOINT STOCK COMPANY 50 SVI BIEN HOA PACKAGING JOINT STOCK COMPANY 51 TDP THUAN DUC JOINT STOCK COMPANY 52

53

THG

TLD

54

TIEN GIANG INVESTMENT AND CONSTRUCTION JOINT STOCK COMPANY THANG LONG URBAN DEVELOPMENT AND CONSTRUCTION INVESTMENT JOINT STOCK COMPANY

55 56 57 58 59 60

TLH TIEN LEN STEEL GROUP JOINT STOCK COMPANY TNI THANH NAM GROUP JOINT STOCK COMPANY TNT TNT GROUP JOINT STOCK COMPANY TPC TAN DAI HUNG PLASTIC JOINT STOCK COMPANY TRC TAY NINH RUBBER JOINT STOCK COMPANY TTB TIEN BO GROUP JOINT STOCK COMPANY

61 VAF

VAN DIEN FUSED MAGNESIUM PHOSPHATE FERTILIZER JOINT STOCK COMPANY

62 VCA VNSTEEL - VICASA JOINT STOCK COMPANY 63 VFG VIET NAM FUMIGATION JOINT STOCK COMPANY

64 VID

VIEN DONG INVESTMENT DEVELOPMENT TRADING CORPORATION

VIS VIET NAM - ITALY STEEL JOINT STOCK COMPANY

65 66 VPS VIETNAM PESTICIDE JOINT STOCK COMPANY 67 YBM YEN BAI INDUSTRY MINERAL JOINT STOCK COMPANY

x

APPENDIX 2

DATA PROCESSING RESULTS BY STATA SOFTWARE 16.0

1. Descriptive statistics

2. Correlation analysis

3. Auto-correlation test (adjusted)

xi

4. OLS regression for short-term debt on the ROA

5. FEM model for short-term debt on the ROA

xii

6. REM model for short-term debt on the ROA

7. Choosing between FEM and REM model for short-term debt on the ROA

xiii

8. Multicollinearity test for short-term debt on the ROA

9. Heteroskedasticity test for short-term debt on the ROA

10. Auto-correlation test for short-term debt on the ROA

xiv

11. FGLS model for short-term debt on the ROA

xv

12. OLS regression for long-term debt on the ROA

13. FEM model for long-term debt on the ROA

xvi

14. REM model for long-term debt on the ROA

15. Choosing between FEM and REM model for long-term debt on the ROA

xvii

16. Multicollinearity test for long-term debt on the ROA

17. Heteroskedasticity test for long-term debt on the ROA

18. Auto-correlation test for long-term debt on the ROA

xviii

19. FGLS model for long-term debt on the ROA

xix

20. OLS regression for short-term debt on the ROE

21. FEM model for short-term debt on the ROE

xx

22. REM model for short-term debt on the ROE

23. Choosing between FEM and REM model for short-term debt on the ROE

xxi

24. Multicollinearity test for short-term debt on the ROE

25. Heteroskedasticity test for short-term debt on the ROE

26. Auto-correlation test for short-term debt on the ROE

xxii

27. FGLS model for short-term debt on the ROE

xxiii

28. OLS regression for long-term debt on the ROE

29. FEM model for long-term debt on the ROE

xxiv

30. REM model for long-term debt on the ROE

31. Choosing between FEM and REM model for long-term debt on the ROE

xxv

32. Multicollinearity test for long-term debt on the ROE

33. Heteroskedasticity test for long-term debt on the ROE

34. Auto-correlation test for long-term debt on the ROE

xxvi

35. FGLS model for long-term debt on the ROE