Market risk versus credit risk of the selected countries in the trans pacific partnership agreement

This study is conducted to measure and rank the market risk level of 10 industries/sectors for selected courtiers in the Asia Pacific region: Vietnam, Malaysia, Australia and New Zealand

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

MARKET RISK VERSUS CREDIT RISK OF SELECTED COUNTRIES IN THE TRANS-PACIFIC PARTNERSHIP

AGREEMENT

BY

QUANG VAN TUAN

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY December 2017

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

MARKET RISK VERSUS CREDIT RISK OF SELECTED COUNTRIES IN THE TRANS-PACIFIC PARTNERSHIP

AGREEMENT

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

ABSTRACT

At the time this study is finalized, the future of the so-called Trans-Pacific Partnership Agreement (TPP) is still uncertain after the US Present Donald Trump walked away from his predecessor Barack Obama’s commitment A different version of TPP, or

to be called the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP), may be formed without the US presence Among these member countries, Vietnam and Malaysia (in the ASEAN), together with Australia and New Zealand, in the Pacific Ocean, are generally considered closely competitive nations for various industries,

in particular for Agriculture; Food and Beverage and Tourism

This study is conducted to measure and rank the market risk level of 10 industries/sectors for selected courtiers in the Asia Pacific region: Vietnam, Malaysia, Australia and New Zealand Two periods are considered in market risk, including: (i) the GFC period (2007-2009); and (ii) the post-GFC period (2010-2016) The market risk level

is measured using the parametric approach and the historical approach for both Value at Risk (VaR), the potential losses in the future over the given time period (day or month) at

a given confidential level, and Conditional Value at Risk (CVaR), which is designed to

estimate the risk of extreme loss

Findings from this study confirm that Vietnamese sectors are relatively riskier than their counterparts in Malaysia, Australia and New Zealand In addition, market risk level across sectors in all countries has substantially reduced in the post-GFC period Financials including Banks, Diversified Financials, and Insurance have been largely ignored from the Vietnamese Government’s focus Interestingly, IT industry is considered very low risk in Vietnam whereas this sector belongs to a group of high market risk in Malaysia, Australia, and New Zealand

This study is then extended to measure and rank the credit risk level for all industries for Vietnam as the case study Credit risk is generally defined as the risk that is determined

on a credit requirement from the default Findings from this empirical study indicate that Industrials, Energy and Consumer Discretionary sectors have had the worst ranking performance in relation to their credit risk Utilities, Financials and IT have achieved a substantial improvement in the post-post GFC periods In addition, this study also

demonstrated an important link between market risk and credit risk, which can provide an important insight to develop for further issues integrating these aspects.

With the ambition to be a financial hub in the Asia Pacific region in the regional integration and a modern industrial economy, a shift of the attention to this particular and important sector in Vietnam is the near future is strongly recommended

Key words: Market risk; Credit risk; Sectors; VaR; CVaR; DD; Vietnam; Malaysia,

Australia, New Zealand

DECLARATION

I hereby declare that the thesis entitled “Market risk versus credit risk of selected countries in the Trans-Pacific Partnership Agreement” written and submitted by me in fulfillment of the requirements for the degree of Master of Art in Development Economics

to the Vietnam – Netherlands Programme This is also my original work and conclusions drawn are based on the material collected by me

I further declare that this work has not been submitted to any other university for the award of any other degree, diploma or equivalent course

HCMC, December 2017

Quang Van Tuan

ACKNOWLEDGEMENTS

First of all, I would like to express my gratitude to my supervisor Dr Vo Hong Duc,

for his knowledge, motivation, support and for providing me enormous, valuable

opportunities His guidance helped me at all the time of research and writing of this thesis,

without him, this thesis would have never been completed

In addition, I would like to thank Prof Nguyen Trong Hoai, Dr Pham Khanh Nam,

Dr Truong Dang Thuy who have provided me the valuable knowledge in the first step of

research

Furthermore, I would also like to thank all lecturers, staff and Mr Pham Ngoc Thach

at the Vietnam Netherlands Programme

Finally, I wish to express my greatest gratitude to my parents, my aunt and my

younger sister for their unconditional encouragement, support and love on the way I have

chosen

Quang Van Tuan

Ho Chi Minh City, Vietnam

CONTENTS

ACKNOWLEDGEMENTS iii

CONTENTS iv

LIST OF TABLES vi

LIST OF FIGURES vi

ABBREVIATIONS vii

CHAPTER 1 8

INTRODUCTION 8

1.1 Problem statement 8

1.2 The research objectives 11

1.3 Research questions 11

1.4 A choice of the countries in the Asia Pacific Region in this study 12

CHAPTER 2 13

LITERATURE REVIEW 13

2.1 Theoretical review 13

2.1.1 Basel II 13

2.1.1.1 Categories of risk 15

2.1.2 Value at Risk 16

2.1.2.1 Introduction 16

2.1.2.2 The Historical method 16

2.1.2.3 The Monte Carlo simulation 17

2.1.2.4 The Variance-Covariance method 18

2.1.2.5 Comparison of VaR Methodologies 20

2.1.2.6 Limitations of VaR 21

2.1.3 Conditional Value at Risk 22

2.1.4 Correlation 23

2.1.5 Distance to Default 25

2.1.5.1 KMV-Morton Model 25

2.1.5.2 Steps in the KMV-Merton model 27

2.2 Empirical literature 28

2.2.1 Empirical evidences on the market risk 28

2.2.2 Empirical evidences on credit risk 29

CHAPTER 3 31

METHEDOLOGY AND DATA 31

3.1 Methodology 31

Value at Risk 31

Conditional Value at Risk 31

Equity model 32

Distance to Default 33

Hypothesis Testing 34

Test selection 34

Spearman Rank Correlation Test 34

3.2 Data 36

CHAPTER 4 38

EMPIRICAL RESULTS 38

4.1 Data descriptions 38

4.2 Market Risk by VaR and CVaR Results 41

4.2.1 In the GFC period (2007 - 2009) 41

4.2.2 In the post-GFC (2010 – 2016) 44

4.2.3 Ranking Shifts in Vietnam 40

4.3 Credit Risk by Distance to Default Results for Vietnam 44

4.4 Market risk versus Credit risk outcomes 45

CHAPTER 5 48

CONCLUDING REMARKS AND POLICY IMPLICATIONS 48

5.1 Concluding remarks 48

5.2 Policy implications 49

5.2.1 The implications for practitioners and investors 50

5.2.2 The implications for Vietnamese government 50

5.3 The limitations and further research 51

Reference 52

LIST OF TABLES

Table 1 Comparison of VaR methods 20

Table 2 Matrix Variance-Covariance Calculation for a Two-Asset Portfoli 33

Table 3 Spearman Rank Correlation Test 35

Table 4 Sector Breakdown 37

Table 5 Daily commodity market price movements in Vietnam and Malaysia (2007– 2016) 39

Table 6 Daily commodity market price movements in Australia and New Zealand (2007–2016) 40

Table 7 The level of market risk proxied by VaR using Parametric and Historical approaches for Vietnam, Malaysia, Australia and New Zealand in the GFC period (2007-2009) 42

Table 8 The level of market risk proxied by CVaR using Parametric and Historical approaches for Vietnam, Malaysia, Australia and New Zealand in the GFC period (2007-2009) 43

Table 9 The level of market risk proxied by VaR using Parametric and Historical approaches for Vietnam, Malaysia, Australia and New Zealand in the GFC period (2010-2016) 38

Table 10 The level of market risk proxied by CVaR using Parametric and Historical approaches for Vietnam, Malaysia, Australia and New Zealand in the GFC period (2010-20016) 39

Table 11 VaR Ranking Shifts in Vietnam 41

Table 12 CVaR Ranking Shifts in Vietnam 43

Table 13 DD Ranking Shifts in Vietnam 44

Table 14 Market Risk proxied by Parametric and Credit Risk proxied by DD Comparison in post-GFC (2010 – 2016) 46

Table 15 Market Risk proxied by Historical and Credit Risk proxied by DD Comparison in post-GFC (2010 – 2016) 47

LIST OF FIGURES

Figure 1 Distribution of daily returns of NASDAQ 100 – Ticker: QQQ 17

Figure 2 Monte Carlo simulation 100 random trials 18

Figure 3 Distribution of daily returns of NASDAQ 100 – Ticker: QQQ 19

Figure 4 VaR, CVaR, Deviations 22

Figure 5 VaR Values Changes in Vietnam 41

Figure 6 CVaR Values Changes in Vietnam 43

ABBREVIATIONS

Basel II Basel II Accord 2004: BIS revised capital adequacy framework

CVaR Conditional Value at Risk: Value at risk on condition

DD Distance to Default: approach developed by KMV – Merton

GICS Global Industry Classification Standard

VaR Value at Risk

CHAPTER 1 INTRODUCTION

1.1 Problem statement

In recent years, the global financial market has undergone a huge change At the early stage of the 2000s, the European Union indicated an important inclination of the European financial markets Moreover, the international crisis of 2007 - 2008, which was originated from the US, caused the negative effect on the global system In addition, Duffie & Pan (1997) presented the idea to minimalize the loss by recognizing and calculating the risk, and certify the financial market and economy system securely

Vietnam has emerged as a new economic engine for the Southeast Asian region with many important industries The three pillars contributing the most value to the Vietnamese economy over the last decade or so are agriculture, manufacturing, and food & beverage Among these key pillars, for example, agriculture is a key industry, which has consistently contributed 20 percent to the national GDP in 2015.1 In addition, Vietnam becomes an official member of The Trans-Pacific Partnership (TPP) with 11 other countries including Australia, Brunei, Canada, Chile, Japan, Malaysia, Mexico, Peru, New Zealand, Singapore, and the United States According to statistics,2 total gross domestic product (GDP) of the current TPP parties is approximately $28 trillion, comprises 40 percent of global GDP and one third of world trade TPP members account for 11.3 percent of world population and 25.9 per cent world trade

Even though the full TPP Agreement is dead under the water after the US Present Donald Trump walked away from his predecessor Barack Obama’s commitment A different version of TPP, or to be called the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP), may be formed without the US presence This new CPTPP agreement will bring both opportunities and challenges for Vietnam in the future

In order to maximize the potential benefits from this new partnership, it is time to recognize the important role of sectorial risk, in particular, for key sectors (industries) relatively to

1 https://www.gso.gov.vn/default.aspx?tabid=621&ItemID=15507

2 http://www.nytimes.com/interactive/2016/business/tpp-explained-what-is-trans-pacific-partnership.html

similar sectors from other members

In addition, the so-called Asia-Pacific region has emerged as the economic and political powerhouse for the last decade or so It is time for Vietnam to go beyond the borders of the ASEAN The common market has been getting larger and larger and competition has been becoming more challenging Malaysia and Thailand are no longer the only competitors for certain sectors in Vietnam It is time for Vietnam, as well as other identical ASEAN nations including Thailand and Malaysia, to recognize and prepare a response for competition from other countries, who are not the members of the ASEAN, from the Asia-Pacific region, in particular for Australia and New Zealand

Risks may subsist in every movement of life and risk estimation is the essential activity Moreover, this movement may support the lenders forecasting and avoiding bad debts In business activities, risk may arise from various sources: the volatility of the business environment; business cycle, changes in government policies and especially in the financial markets In the event, a number of companies passively accept the risk However, others attempt to manage risk by utilizing the proactive methods Either of circumstances, risk should be carefully monitored because of its potentially harmful effects Jorion (2007)

According to Basel II Accord (Bank for International Settlements, 2004), risk may

be split into three groups including credit risk, operational risk, and market risk First,

credit risk is generally defined as risk of loss because of the payment default of borrower

Second, operational risk presents the risk due to internal process failures, systems and people Third, market risk is considered as a change in the price of financial assets due to

the changes in interest rates, stock prices, exchange rates and commodity risk

To be more precise, one of the most effective techniques to estimate the market risk

is Value at Risk (VaR) In 1995, The Basel Accord recommended banks to calculate the capital requirements for market risk by employing certain parameters with value at risk model VaR laid the groundwork for resolving a numerous aspect of financial risk Jorion (1996) and Pritsker (1997) indicated that VaR, as a risk management method, can estimate the maximum expected loss that may occur over a given period, at a given confidence level

In addition, VaR is an uncomplicated risk management and 𝛼-percentile of a distribution

is easily figured out The tempting simplicity of the VaR concept can conduct its implementation as an ordinary risk measure for financial actualities entailed in not only operations, banks, insurances, but also an institutional investment, and nonfinancial enterprises As such, VaR tends to become the prevailing method for measuring risk in majority of the industries and countries

Despite its popularity, VaR remains unsatisfactory mathematical properties Artzner

et al (1999) analyzed risk measures and concluded that an arrangement of axioms that should be suitable for any risk measures, which persuades these axioms, is considered to

be “coherent” The four axioms are Monotonicity, Translation Equivariance, Subadditivity, and Positive Homogeneity However, there is no precise quantity of VaR and each measure

stretch with its limitations and VaR is not able to catch the subadditivity axiom Therefore,

VaR is not recognized for a coherent risk measure This makes VaR optimization a challenging computational problem

Acerbi and Tasche (2002) demonstrated that Condition Value at Risk (CVaR) could

convince all the above axioms and especially subadditivity axiom As such, CVaR is a

coherent risk measure CVaR is considered as a good measurement of the extreme losses

in the tail of the distribution as it is conditioned on the returns exceeding VaR First, CVaR

is usually measured as a percentage If VaR is measured at a 95% confidence level, CVaR

will be the average of the worst 5% of observations Second, CVaR has been related to

sector risk and economic periods to measure modifications in risk for extensive product categories by Powell, Vo, and Pham (2016c); Pflug (2000) presented that CVaR is a rational measure, not containing the disadvantageous properties of VaR such as

subadditivity Third, CVaR does measure tail risk as it appraises those returns beyond VaR

This research aims to emphasize on market risk and credit risk, which have attracted great attention from academia, investment bankers, and policymakers Although numerous studies such as Allen at el (2014); Powell, Vo, and Pham (2016a, 2016b, 2016c) investigated the market and credit risk for various countries over the world, there has been

no study concerning, comparing Value at Risk (VaR) and Conditional Value at Risk (CVaR) rankings for various sectors for Vietnam and other members from the Asia-Pacific region As a result, this study is conducted to fill this gap In addition, this study will

demonstrate the link, if any, between the market risk estimates when the VaR/CVaR and the Distance to Default techniques are adopted in the context of selected countries in the Asia-Pacific region including Vietnam

A careful examination indicates that Vietnam and Malaysia (the ASEAN members) and Australia and New Zealand (the non-ASEAN members) are relatively identical in relation to the key industries that contribute substantially to the national economies According to Bloomberg where all the data required for this study are extracted, economic activities in any economy can be allocated into 11 different sectors These sectors include

Energy; Materials; Industrials; Consumer Discretionary; Consumer Staples; Healthcare; Financials; Information Technology; Telecommunication Services; Utilities; and Real estate The choice of these four selected countries is arbitrary albeit interesting While

Vietnam and Malaysia are developing countries, Australia and New Zealand are developed economies All these selected countries are members of the Asia-Pacific Economic Cooperation

1.2 The research objectives

This study is conducted to achieve the following research objectives:

 First, estimating the market risk for all sectors for Vietnam and selected other

countries in the Asia-Pacific region where data is available Furthermore, this study will closely focus on the differences of the estimates between the crisis and non-crisis periods

 Second, estimating the credit risk using the Distance to Default (DD) structural

approach and providing the link, if any, between the VaR and the DD model

1.3 Research questions

The following research questions have been raised in this study:

 What is the currently prevailing market risk level of various sectors from Vietnam and selected other countries in the Asia-Pacific region using VaR and CVaR?

 Is there any link between the estimates of the market risk using VaR techniques

and the credit risk using the DD model?

1.4 A choice of the countries in the Asia Pacific Region in this study

Due to time constraint, conducting research for all members of the Asia Pacific Region may become excessive As such, it is the intention of this study is to focus on the members in the Australasian region In this region, members include Malaysia and Viet Nam (ASEAN) and Australia and New Zealand A preliminary search provides evidence

to confirm that Brunei may not be on the final list because of its financial market size and its level of economic development It is worth noting that, for example, agriculture is one

of the top three industries for Vietnam, Malaysia, Australia and New Zealand A scrutiny will be conducted

CHAPTER 2 LITERATURE REVIEW

This chapter is to review the theoretical and empirical literature on both market risk and credit risk There are two main parts in this chapter:

i The basis theories in market risk and credit risk

ii The empirical evidences on market risk and credit risk

2.1 Theoretical review

The literature review re-examines the Basel Accords II, Value at Risk methodologies, Conditional Value at Risk methodology and correlation techniques

 The Basel II framework establishes the minimum standards for management

on a sensitive risk: Market Risk, Operational Risk and Credit Risk

 Value at Risk is one of the most effective techniques to estimate maximum expected loss on the market risk In addition, there are three main methods to calculate VaR First, the historical method is relied on the actual historical data Second, the Monte Carlo method simulation that randomly creates multiple scenarios Third, the Variance-Covariance (parametric) method estimates VaR

on the assumption of a normal distribution

 Conditional value at Risk is a coherent risk approach and satisfies the desirable characteristics that are the shortcomings of Value at Risk

 Correlation is an estimate the level of one variable’s value is related to the value

of another It is particularly important for practitioners interested in minimize risk to measure the correlation between variables

2.1.1 Basel II

Basel Capital Accord (Basel I) initially indicated by the Group of Ten (G10) countries in 1988, to retain the adequate capital that provide a support against unexpected losses Therefore, Value at Risk (VaR) was selected as a method proposed to predict the maximum expected loss over a given period time and confidence level Moreover, Basel

Accord stipulated a standardized approach, which is utilized to calculate the capital for credit risk and market risk, for all institutions However, this method may contain several adequacies, the most important of which were its traditionalism and its failure to remunerate organizations with higher risk administration

The Basel Accord was enhanced in April 1995 Basel II permitted associations to utilize internal models to specify their VaR and the required capital expenses Nevertheless, organizations desire to use their internal models, which were originated from regulator, to employ back-testing method Following the Australian Prudential Regulatory Authority (APRA), Basel Accord was chosen as the national mechanism of financial markets The normalized procedure is built on ratings from qualified external rating operation

The Basel II structure entails of three Pillars:

 Pillar 1: Capital Requirements This Pillar recommends minimum capital

obligations for market risk, operation risk and credit risk The value of total capital ratio to risk weighted is determined to 8% In addition, this ratio has the same result with the Basel Accord I, but the only difference is the modified alternatives for measuring risk-weighted assets

 Pillar 2: Supervisory Review Managers are involved to certify that all the

capital essential requirements, standard procedures and schemes are similarly structured in banks to determine the capital requirements Moreover, the supervisor in Australian Prudential Regulatory Authority (APRA) initiates general discussion with bank manager, and confirms that systems are utilized

in practice

 Pillar 3: Market Discipline Those achieves the possibility to highlight capital

regulation and alternative endeavors, delivers relative safety to financial systems and banks In addition, banks are required to expose their risks and systems specifically

The Basel II gratefully acknowledges the outstanding position of Value at Risk as a standard risk measurement and capital estimation In particular, with the appearance of Basel II, VaR approach can clearly distinguish the different of the market risk from the

credit and other risks Moreover, market VaR methodology has become very meaningful

to banks to estimate the daily basis with the Basel requirement This considerable benefit banks can clarify a reduced capital requirement

2.1.1.1 Categories of risk

Total capital requirements contain the sum of the capital that is required for the following risks:

Market Risk, also known of as “systematic risk” that is originated from elements

influencing a large indefinite number of assets in the entire market Besides, the risk of loss initiated from the opposite trend in market factors there are four different kinds of market risk: foreign exchange risk, interest rate risk, equity price risk and commodity price risk The value of market risk for capital charge can be measured by utilizing the Regulator’s standard method or employing the internal model method Hogan et al (2004) demonstrated the great advantage of utilizing internal models as the standard method to produce extra outstanding qualities for the capital charge Moreover, market risk is also measured by using the Value at Risk (VaR) method

Operational Risk is not the formal section of Basel I This encloses processing risks,

procedural risks and transactional risks In addition, Banks may deal with these future scenarios by utilizing either standardized method or Internal Ratings Based (IRB) method Moreover, the standardized method includes segmentation of the common enterprise along typical lines of business after that determining beta for each line, regulator measures the beta and that beta function is represented for whole industry By contrast, with IRB method that bank with highly complex and advanced systems may be obtainable In addition, Operational risk can be estimated based on past plans and experience by bank

Credit Risk refers to the risk that is determined on a credit requirement from the

default It naturally arises due to the expectation of the borrowers, who desire to pay current debt by utilizing future cash flow It is completely impossible to certify that borrowers will reimburse their debts A reward of investor, which is originated from the interest payments for debt obligation of borrower, is assuming credit risk In addition, this research aims to emphasize on credit risk, which has attracted great attention from academia, investment bankers, and policymakers

2.1.2 Value at Risk

2.1.2.1 Introduction

Following the introduction of RiskMetrics Technical paper by J.P Morgan in 1994 and updated information by J.P Morgan & Reuters (1996) reveal that Value at Risk (VaR) approach is a well-known and widely employed metric for estimating the risk in recent years VaR is significantly different to other approaches Harper (2004) demonstrated that VaR based on the historical information to calculate the potential losses

in the future over the given time period (day or month) at a given confidential level (typically 95% or 99%)

For instance, the expectation of a portfolio that may lose no more than $1 million, with 95% confidence level of the time (30 days) has a Value at Risk of $1 million The negative aspect is that 5% of the period time or 1 day out of 30, the portfolio could lose at least $1 million VaR is also utilized for estimating the governing capital investment Moreover, one of the most great advantages of VaR is it may compile several both related risks and unrelated risks into general method that is expressed in currency terms of an enterprise or portfolio

In particular, VaR may be estimated by three methods: The Historical method, the Monte Carlo simulation and the Variance-Covariance method (or correlation or parametric method)

2.1.2.2 The Historical method

The historical approach that collects actual historical losses in portfolio from top (best) to bottom (worse) after that estimating the VaR value based on the assumption of history information tends to be occurred repeatedly Harper (2004) denotes the QQQ began

to trade in Mar 1990, the big data from QQQ, which is collected and calculated by daily, will show in bar chart on Figure 1 to compare and analysis

Figure 1 Distribution of daily returns of NASDAQ 100 – Ticker: QQQ

Source: Investopia (2016)

The red bars spreading out, from -4% to -8% It describes the highest amount of daily loss will not over 4% with 95% level of confidence Moreover, if the investors uses $1000

to invest, they expect that their will loss not higher than $40

VaR may not calculate the absolute value, instead of estimating the probabilistic value for the portfolio In addition, when raising the level of confidence to 99%, the interval from -7% to -8% will be represented for the daily returns at 1% If the investor uses $1000

to invest, they expect that their will loss not higher than $70 with 99% level of confidence

2.1.2.3 The Monte Carlo simulation

The Monte Carlo Simulation refers to approach that randomly creates experiments

In addition, a random number generates the changes of portfolio value to particular simulation conducting the Monte Carlo simulation However, the results for particular conduct will provide the dissimilar value despise of the small differences among values The result conducted 100 trials of monthly returns for QQQ that will show into a bar chart

Figure 2 Monte Carlo simulation 100 random trials

Source: Investopia (2016)

According to the statistics on Figure 2, there are two results in the range from -15%

to -20% and three results in the range from -20% to -25% This outcome reveals the worst five results (that complies the worst 5%) may be less than -15% Therefore, the Monte Carlo tends to provide the VaR-type conclusion: the loss expectation during given month for the QQQ not excess than 15% with 95% level of confidence

2.1.2.4 The Variance-Covariance method

The returns, which are normally distributed; the correlations between risk factors are the essential requirements for assumption of this approach Specifically, the Variance-Covariance method requires two important factors to calculate the VaR for an asset: the mean and a standard deviation that allow sketching the distribution curve

Figure 3 Distribution of daily returns of NASDAQ 100 – Ticker: QQQ

Source: Investopia (2016)

The actual daily standard deviation for QQQ is 2.64% based on the blue curve on Figure 3 After that, using standard distribution tables by Statsoft Inc (2003) to obtain the value of the “worst” 5% and 1% Obviously, the results can be achieved by plugging the standard deviation into the formulas below:

On the low side, zero tends to bound in the lognormal distribution Choudhry (2004) reveals the appropriateness of lognormal distribution for estimating financial time series observations including extreme negative values (values that cannot be practically observed

with share prices)

2.1.2.5 Comparison of VaR Methodologies

Among these methodologies, this is remained the question for VaR methodologies:

“which method is best?” As Linsmeier and Pearson (2000) indicated that it is not easy to find out a satisfied answer for that complicated question Each method differs in ability to measure and capture the risk of different scenarios The best choice may be determined by risk managers for their organization and evaluation Table 1 will discuss how the three methods differ depending on precise dimensions

Table 1 Comparison of VaR methods

Attribute Historical method The

Variance-Covariance method

The Monte Carlo simulation

Is it able to measure the

risk including option in

the portfolio?

holding period time with limited option in the portfolio

Yes

Is it easy to apply? Yes if the past data are

readily available

Depending on the complexity of the instruments and data

Extremely difficult to implement

Are the computations

3 http://financetrain.com/calculating-var-using-monte-carlo-simulation/

Shahabuddin (2000) argued that the estimation of VaR for Big Data (huge portfolios) faces

a tradeoff between speed and accuracy Hence, Monte Carlo simulation spends much time

to operate and too slow to be practical

2.1.2.6 Limitations of VaR

Although VaR has a number of advantages, this approach remains the majority of argument and is accused of providing the considerable loss in both the pre- and post-financial crisis In addition, Artzner et al (1999) demonstrated that VaR is not an appropriate “coherent” approach of risk because it do not satisfy the following axioms:

In specific, the loss is not only represented by random variable X and Y, but also

𝑐 𝜖 ℝ that is a scalar In addition, 𝜌 is a risk function, i.e it demonstrates the random variable X (or Y) to ℝ Moreover, the risk is associated with X (or Y)

 Monotonicity reveals that a lower loss asset will produce a lower risk measure

A risk measure 𝜌 is monotone:

𝑋 ≤ 𝑌 ⇒ 𝜌(𝑋) ≤ 𝜌(𝑌)

 Translation Equivariance demonstrates that if adding more one additional risk,

it may generate more risk and more constant to random variable in order to stability in its variability A risk is measured by 𝜌:

combining two positions is not excess than sum of them In addition, principle of diversification is a fundamental of subadditivity Albanese (1997) demonstrate that when employing the VaR to manage the credit portfolio risk, leading to a higher value of credit risks due to the deficiency of subadditivity for VaR Therefore, return of portfolios that are associated with default risk may be asymmetrical by using credit instruments, and fat tails may be appeared in these return distributions

Another problem is considered by McKay and Keefer (1996); Mauser and Rosen (1999) It is complicated to make VaR optimize when estimate for these scenario It cannot simply estimate the function of portfolio In addition, it can display multiple local extrema that is not certain to be successful to specify the VaR of a particular mix Generally, CVaR can enhance all the above disadvantages of VaR presented in this part

2.1.3 Conditional Value at Risk

Conditional Value at Risk (CVaR) is highly relevant to VaR The value estimation

of CVaR is equal or higher than VaR’s value Furthermore, CVaR was designed to estimate the risk of extreme loss; CVaR is an upgraded approach of VaR that can provide the total amount of expected loss Obviously, the user of VaR would ask this question with VaR:

“How often may the portfolio lose at least $1 million?” with CVaR the users could have a question “When the portfolio loses higher than $1 million, and how much it would lose?”

In addition, the relationship between VaR and CVaR is explained in the Figure 3 below:

Figure 4 VaR, CVaR, Deviations

The mean excess loss or tail of VaR is the specified name of CVaR α-VaR is a probability value of the loss and (1-α)*100 is the mean value of the worst losses for CVaR Uryasev (1999) reveals that when estimating for VaR at 95% level of confidence (α = 0.95), the average of the 5% of worst loss is the value of CVaR

As the discussion the part limitations of VaR, it did not satisfy the requirements of

“coherent” risk measures and it was not simple to optimize VaR By contrast, CVaR measures the losses that are occurrence at the end of the tail distribution Pflug (2000) demonstrate that CVaR is a coherent risk approach and satisfies the desirable characteristics that VaR cannot Moreover, VaR cannot estimate on the excess or extreme losses value, but CVaR can deal with this estimation Rockafellar and Uryasev (2000) indicate that CVaR will be easier than CaR by utilizing the nonparametric method to estimate the portfolio losses

VaR does not provide any information on the excess/extreme losses, but this is calculated by CVaR Since CVaR value is greater than or equal to VaR, portfolios with a high VaR also have a high CVaR Therefore, this is the reason why CVaR has the equal or higher values than VaR, and portfolios have a high VaR value estimation will have a high CVaR value

However, CVaR also has its disadvantage Yamai and Yoshiba (2002) demonstrate that CVaR requires a huge number of observations to produce a trustworthy estimation value, and VaR tends to provide the stable results than CVaR Moreover, the consistency

of CVaR mainly bases on the precision of the tail model; hence, CVaR users should fully comprehend that

2.1.4 Correlation

The discussions on this section have an intimate connection to calculation of VaR for individual asset Each of three models that have been chosen for further examine also combine a portfolio approach and assets relating to each other; for instance, the correlation between the assets

It is particularly important for practitioners interested in minimize risk to measure the correlation between variables Correlation is an estimate the level of one variable’s

value is related to the value of another The correlation coefficient is a single number with the value determines the strength and the sign reveals the relative directions for comparing

in two values of instruments In addition, the coefficient’s value ranging from -1 to +1, depend on the relationship of two instrument’s values For instance, the correlation result

is 0.9; it means the one asset has the same direction by 90% with another Furthermore, the instrument is uncorrelated when correlation value is equal; hence, their direction is independent of each other

Single asset VaR has been examined in Section 2.1.3 Choudhry (2004); Powell (2007) considered the approach can deal with 2 assets

 Step 1: Obtaining relative weightings

 Step 2: Obtaining Standard Deviation for each asset (𝜎) by multiplying the standard deviation of daily price relatives by the square root of the number of trading days per annum (usually 250 – the number of days data required by Basel to measure market risk)

 Step 3: Obtaining variance for each asset (𝜎2)

 Step 4: Obtaining correlation coefficient between the two assets by function

 Step 7: Portfolio Standard Deviation by square root Portfolio Variance

This study will present consideration for a different view to other previous studies

In particular, McAleer and da Veiga (2004) considered for forecasting accuracy that determined by backtesting three multivariate GARCH models via four international portfolios By contrast, this study will deal with whether the inclusion of correlations has

an effect upon relative industry VaR To be more precise, this thesis will re-examine Equity

model that follows the variance-covariance parametric approach to calculate VaR by using undiversified approach (equally weighted) and diversified approach (market value weighted for each company basing on market capitalization) The method for the Equity model is discussed in Section 3.1.3

2.1.5 Distance to Default

Black and Scholes (1973) demonstrated the pricing of options theory Merton (1974) extended the accomplishment of Back and Scholes in the default prediction of the firm under the strong assumption and Merton’s model widely employed to forecast the fault of the firms In 1980s, KMV Corporation developed Merton’s model, it is called KMV-Merton model relying on the firm’s equity implied as an option on the underlying value of the firm’s asset in a given point in time

2.1.5.1 KMV-Morton Model

The KMV-Merton model is likely to produce a probability of default for each firm

in a certain time horizon To measure the probability of default, KMV-Merton model obtains the difference between the face value of the firm’s debt and the market value of the firm, after that divides this difference by the estimation of the volatility of the firm In that, the market value of the firm is calculated by sum of the value of firm’s equity and the market value of firm’s debt The estimation default probabilities would look effortless if both these quantities were readily available While, equity values are readily observable, the market value of firm debt is not easily obtainable

In addition, KMV – Merton model applies the Merton bond pricing model to estimate the market value of debt There are two important assumptions in the Merton model The first assumption is the total value of a firm following geometric Brownian motion:

𝑑𝑉 = 𝜇𝑉𝑑𝑡 + 𝜎𝑉𝑉𝑑𝑊 Where:

 V: The market value of firm’s assets

 𝜇 : The expected continuously compounded return on V

 𝜎𝑉: The volatility of firm value

 dW: A standard Weiner process

The second key assumption of the Merton model is that discount bond, which the firm has issued just only one in T periods According to these assumptions, the equity value

of the firm (E) can be inferred as a call option with risk free rate (r) and strike price on V Furthermore, on the one hand, the face value of firm debt is lower than the value of firm’s assets, the option will be exercised and the equity holders will payoff the bond holders On the other hand, the face value of firm debt is greater than the value of its assets, the option will be expired and the equity holders will transfer ownership of the firm to debt holders The firm’s equity under these conditions is produced by The Black-Scholes equation:

Where:

 E: The market value of firm’s equity

 F: The face value of the firm’s debt

 r: Instantaneous risk-free rate

 𝒩(∙) The cumulative standard normal distribution function

 𝑑1 𝑎𝑛𝑑 𝑑2 are given by:

𝑑1 = ln (

𝑉𝐹) + (𝑟 + 0.5𝜎𝑣2)𝑇

𝜎𝑣√𝑇

𝑑2 = 𝑑1− 𝜎𝑣√𝑇 The KMV – Merton model employs two important equations including the Black-Scholes equation and volatility of the firm’s equity In according Mertons’s assumption value of equity is a function following directly from Ito’s lemma:

and its equity can be show

𝜎𝐸 = (𝑉

In sum, The KMV – Merton model fundamentally employ two nonlinear equation (1) and (2) to produce the probability of default by translating the value and volatility of a firm’s equity In most application, the Black-Scholes-Merton model can illustrate the unobserved value of four variables (underlying asset price, strike price, risk-free rate and time-to-maturity) as a function and estimate one variable (volatility) Instead of this point, the value of the firm’s equity in KMV – Merton model is implied as the value of the option, while this model is not able to obtain directly the value of underlying asset Therefore, while E is easily obtainable by multiplying the firm’s shares outstanding with its current stock price, V must be inferred; and volatility of underlying firm to do likewise

2.1.5.2 Steps in the KMV-Merton model

 The first step is to collect the market equity of the firm and the risk-free rate

 The second step in the KMV-Merton model is to calculate 𝜎𝐸 which can obtain from historical stock returns data

 The third step is to select a time horizon and estimate the face value of firm’s debt For example, if data of historical returns to estimate 𝜎𝐸 assuming a time horizon of one year, this model will create T=1 and obtain the book value of the firm’s total liabilities by the face value of its debt

 When we have sufficient value for each the variables in equation (1) and (2) except V (the total value of the firm) and the 𝜎𝑉 (the volatility of firm value) The fourth step and this is the most important step in applying this model This part will solve simultaneously equation (1) and (2) numerically for values for value V and 𝜎𝑉 Once this solution is acquired, the distance to default is illustrated as

𝐷𝐷 =ln (

𝑉𝐹) + (𝜇 − 0.5𝜎𝑣2)𝑇

𝜎𝑣√𝑇

However, it is not simple to solve these equations (1) and (2) to obtain the value V and 𝜎𝑉 and then DD values Crosbie and Bohn (2003) revealed that it spent a lot of time to obtain the reasonable results provided by equation (2) from the market leverage in practice

To deal with this issue, the author will apply a complicated interactive procedure by KMV

To be precise, author makes a proposal for an initial value of 𝜎𝑣 = 𝜎𝐸(𝐸+𝐹𝐸 ) and employs the value of 𝜎𝑣 to calculate the market value of each firm’s assets every day in equation (1) After that, estimating the log return on assets for every day and utilizing that series returns to produce new 𝜎𝑣 𝑎𝑛𝑑 𝜇 This calculation for 𝜎𝑣 will be iterated in this manner until it converges (this means the absolute difference among in contiguous 𝜎𝑣𝑠 is less than

10−3)

2.2 Empirical literature

2.2.1 Empirical evidences on the market risk

A few studies have concentrated on examining VaR or CVaR approach For example, Mausser and Rosen (1999) demonstrated instruments containing the computation of VaR contribution, marginal VaR and risks for portfolio of European options Authors examine tools by utilizing parametric, delta-normal versions and upgrade these tools to the non-parametric and simulation approach Gourieroux et al (2000) examined the ability of VaR

by allocating two companies listed on the Paris Bourse (546 observations) The results provided the explanation to employ statistical inference and perform a particular analysis

of VaR and the expected loss revealing the loss is greater than a given loss quantile Powell, Vo, and Pham (2016a) focused on the particular agricultural product that has illustrated consistent supremacy mastery over others This study generates the competition

of various economic circumstances over a twelve years period (divide into four stages: crisis, crisis, post-crisis and post-post crisis) to find out a winner, which contains three elements as returns, resilience (ability to overcome the risk), and teamwork (significant contribution to portfolio optimization) To achieve this objective, the authors utilized the CVaR to estimate the extreme risk to determine the overall winner However, in this study, there was no consistency in return rankings from one period to another, with individual commodities shifting from having among the best returns in one period to having among

pre-the worst in opre-thers There was found to be a much higher level of consistency in relative risk, with commodities displaying similar risk rankings from one period to another period

As several studies have been investigated the association between VaR and CVaR For instance, Powell (2007) inspected VaR and CVaR by utilizing a group of Australian industries In addition, authors compare the VaR and CVaR values among these industries over time Moreover, diversified and undiversified VaR also employed as well as parametric and nonparametric CVaR approaches by the author The study has also established an important link between credit and market risk, which can provide a springboard for the development of further models integrating these aspects

Gaivoronski and Pflug (2000) also computed the optimization for the portfolio by utilizing VaR approach and compared these results with CVaR approach Moreover, the authors indicated that it is more feasible to calculate for portfolio based on historical data than CVaR or variance optimization Furthermore, Allen at el (2012) employed CVaR to calculate, compare the extreme risk value in mining share portfolios among seven leading mining areas and to demonstrate CVaR ability optimizing the portfolios and minimize the extreme risk Authors also argued considerable differences between countries utilizing VaR and CVaR via comparing risk rankings Furthermore, the result indicates that investors using traditional VaR will not minimize the risk of portfolios

Rockafellar and Uryasev (2000) developed the model based on CVaR to calculate the credit risk optimization for portfolio of bonds The findings presented that CVaR results are quite similar to the results acquired with VaR, the expected loss and especially with the minimum expected regret approach In addition, this model was adopted in the emerging market bonds

The review has shown that there are very few VaR (and even fewer CVaR) studies

in Vietnam context, particularly as regards the topic of industry risk The few notable studies that have been undertaken in Vietnam have either been on international portfolios,

or focusing on different aspects to this study

2.2.2 Empirical evidences on credit risk

Majority of researchers concentrate on demonstrating the important factor of default