The Role of Financial Ratios in Signalling Financial Distress - Ev

Southern Cross University
ePublications@SCU
Theses
2013
The role of financial ratios in signalling financial

distress: evidence from Thai listed companies
Sumeth Tuvadaratragool
Southern Cross University
Publication details
Tuvadaratragool, S 2013, 'The role of financial ratios in signalling financial distress: evidence from Thai listed companies', DBA thesis,
Southern Cross University, Lismore, NSW.
Copyright S Tuvadaratragool 2013
ePublications@SCU is an electronic repository administered by Southern Cross University Library. Its goal is to capture and preserve the intellectual
output of Southern Cross University authors and researchers, and to increase visibility and impact through open access to researchers around the
world. For further information please contact epubs@scu.edu.au.
The Role of Financial Ratios in Signalling Financial
Distress: Evidence from Thai Listed Companies.
Bachelor of Business Administration, Ramkhamhaeng University, Thailand
Master of Business Administration, Mount Saint Marys University, USA
A thesis submitted to the Graduate College of Management, Southern Cross University,

Australia, in partial fulfilment of the requirements for the degree of Doctor of Business
Administration
February 2013
Statement of original authorship
I certify that the substance of this thesis has not been submitted for any degree and is
not currently being submitting for any other degree.
I also certify that to the best of my knowledge any help received in preparing this
thesis, and all sources used have been acknowledged in this thesis.
Signed: .
Date: ..
i
Acknowledgements
During my candidature for this degree, I have had the support, assistance, and encouragement
of a number of people. I would like to express my gratitude to everyone for their kind
assistance.
First of all, I owe my principal supervisor, Associate Professor Keith Sloan, and my co-
supervisor, Dr Albert Wijeweera, a great debt of gratitude. They kindly provided suggestions,
guidance, encouragement and support while I wrote this thesis. Without them, this thesis
could not have been completed in the way that it has.
Secondly, a debt of gratitude is also owed to Dr Don McMurray and Dr Margo Poole for their
support with statistical techniques and suggestions.
Thirdly, my sincere appreciation is addressed to staff of the Graduate College of Management

and the International Office, including Sue White, Susan Riordan, Ashley Cooper, Rosemary
Graham and Sarah Rogers.
Last but not least, I am extremely grateful to my family for believing in me and for their
encouragement while I have been a candidate for this degree. Without moral support from
them, undertaking the DBA at Southern Cross University (Australia) would not have been
possible, and I could not have completed it within the target period.
ii
Abstract
This thesis investigates the relationship between company characteristics and business failure
among publicly listed companies on the Stock Exchange of Thailand during 20032008. The
characteristics used are financial statement information/ratios. In other words, this thesis aims
to examine whether financial statement ratios can be adequately used to signal business
failure in the Thai context in normal economic circumstances.
The motivation behind this thesis derives from the literature, particularly as it applies to
Thailand. This thesis used quarterly financial statement data so as to take into account
seasonal/cyclical changes, which previous studies have ignored. The underlying reason for
applying quarterly financial statement data is that it is believed that business failure is a
dynamic process (Hossari 2006; and Shumway 2001). The methods used in this thesis are the
Integrated Multi-Measure (IMM) approach (which comprises the Emerging Market Score
model, comparative ratio analysis, and ratio trend analysis) and the logit model as a
benchmarking measure. The successful classification rates of each individual measure are
similar (approximately 57 per cent). The classification accuracy rate for the IMM approach is
64 per cent for the financially distressed firms while the logit model provides the
classification accuracy rate of up to 86 per cent for the financially distressed firms.
The empirical results show that financial statement information can be used to adequately
signal business failure in the Thai context in normal economic circumstances. This is
consistent with the recent study by Beaver, McNichols and Rhie (2005) conducted in a
western setting. Given this, the empirical results of this study could be used as a stepping
stone for future researchers who are interested in finding the best predictors for failure by
developing a ratiobased prediction model and assessing its classification accuracy.
Key words: financial statement information/ratios, business failure/financial distress, Thai

listed companies, normal economic circumstances, ratiobased prediction model.
iii
Table of Contents
Statement of original authorship ......................................................................................... i

Acknowledgements .............................................................................................................. ii
Abstract .............................................................................................................................. iii
Table of Contents ............................................................................................................... iv
List of Tables .................................................................................................................... viii
List of Figures...................................................................................................................... x
Glossary of Abbreviations.................................................................................................. xi
Chapter 1: Introduction ...................................................................................................... 1

1.1 Introduction ..................................................................................................................... 1
1.2 Background to the research.............................................................................................. 3
1.3 Research issues ............................................................................................................... 4
1.3.1 Research problem ..................................................................................................... 5
1.3.2 Research question ..................................................................................................... 5
1.3.3 Research objective .................................................................................................... 6
1.4 Justification for the research ............................................................................................ 7
1.5 Methodology ................................................................................................................... 8
1.6 Principal definitions used .............................................................................................. 11
1.7 Outline of this study ...................................................................................................... 12
Chapter 2: The Thai Environment ................................................................................... 15

2.1 Introduction ................................................................................................................... 15
2.2 Thailand in general ........................................................................................................ 16
2.2.1 Geography .............................................................................................................. 17
2.2.2 Population............................................................................................................... 18
2.2.3 Government/politics ................................................................................................ 18
2.2.4 Economy................................................................................................................. 18
2.3 Thai financial markets and key market players .............................................................. 19
2.4 Thai business failure ...................................................................................................... 24
iv
2.5 Conclusion .................................................................................................................... 26
Chapter 3: Literature Review ........................................................................................... 28

3.1 Introduction ................................................................................................................... 28
3.2 The important aspects of business failure prediction ...................................................... 30
3.2.1 The definition of generic terms used in business failure prediction .......................... 30
3.2.2 The costs of business failure ................................................................................... 32
3.2.3 The usefulness of business failure prediction ........................................................... 34
3.3 The evolution of business failure prediction................................................................... 35
3.3.1 Theoretical framework overview ............................................................................. 35
3.3.2 Previous research overview ..................................................................................... 37
3.3.2.1 International studies.......................................................................................... 39
3.3.2.2 Thai studies ...................................................................................................... 50
3.4 Limitations of the previous research .............................................................................. 53
3.5 Development of the analytical approach ........................................................................ 55
3.6 Conclusion .................................................................................................................... 56
Chapter 4: Research Methodology ................................................................................... 58

4.1 Introduction ................................................................................................................... 58
4.2 Research problem identification and formulation ........................................................... 60
4.3 Research design............................................................................................................. 61
4.3.1 Dimensions of research ........................................................................................... 61
4.3.2 Typology of research .............................................................................................. 62
4.3.3 Research methods ................................................................................................... 63
4.4 Sampling design ............................................................................................................ 65
4.4.1 Specification of population ..................................................................................... 65
4.4.2 Sample selection ..................................................................................................... 69
4.5 Data collection .............................................................................................................. 78
4.6 Analytical approach....................................................................................................... 81
4.6.1 Emerging Market Score (EMS) Model .................................................................... 88
4.6.2 Comparative ratio analysis ...................................................................................... 90
4.6.3 Ratio trend analysis ................................................................................................. 92
v
4.7 Development of the Integrated MultiMeasure (IMM) approach ................................... 94
4.8 Application of logit-based models ................................................................................. 96
4.9 Conclusion .................................................................................................................. 103
Chapter 5: Data Analysis ................................................................................................ 104

5.1 Introduction ................................................................................................................. 104
5.2 Description data of sample used .................................................................................. 106
5.2.1 Descriptive statistics definition ............................................................................. 106
5.2.2 Profile of companies used ..................................................................................... 107
5.2.3 Profile of financial statements/ratios used ............................................................. 109
5.3 Applying the EMS model and its results ...................................................................... 115
5.4 Applying comparative ratio analysis and its results ...................................................... 120
5.5 Applying ratio trend analysis and its results ................................................................. 125
5.6 Applying the Integrated MultiMeasure approach and its results ................................. 130
5.7 Applying the logit-based model and its results ............................................................. 136
5.7.1 Testing for correlation among the 16 financial ratios ............................................. 137
5.7.2 Testing for normality of the 16 financial ratios ...................................................... 139
5.7.3 Discussion of logit model results based on the entire 20 quarters sample ............... 140
5.7.4 Discussion of logit model results based on the split sample ................................... 166
5.8 Conclusion .................................................................................................................. 176
Chapter 6: Conclusions and Implications ...................................................................... 179

6.1 Introduction ................................................................................................................. 179
6.2 Summary ..................................................................................................................... 181
6.2.1 Summary about the research question ................................................................... 183
6.2.2 Summary about the research problem.................................................................... 187
6.3 Implications ................................................................................................................ 188
6.3.1 Implications for theory .......................................................................................... 188
6.3.2 Implications for policy and practice ...................................................................... 190
6.4 Contributions of the study ........................................................................................... 192
6.5 Limitations of this study .............................................................................................. 195
6.6 Implications for further research .................................................................................. 196
vi
6.7 Conclusion .................................................................................................................. 199
List of References ............................................................................................................ 200
Appendices ...................................................................................................................... 210

Appendix A: Multivariate outliers test ............................................................................... 210
Appendix B: Statistical result for between group ttest ...................................................... 211
vii
List of Tables
Table 2.1 SET statistical highlights from 2006 to 2009 ........................................................ 22
Table 2.2 Numbers of dissolutions for juristic persons between 2000 and 2009 ................... 25
Table 2.3 Number of dissolutions classified by business category (December 2009) ........... 26
Table 3.1 Comparison of key features of the international studies........................................ 49
Table 3.2 Comparison of key features of the Thai studies .................................................... 52
Table 4.1 Descriptive statistics of the event of failure in Thailand from 1975 to 2008.......... 68
Table 4.2 Descriptive statistics of initially failed sample group during the period of 2003
2008 .................................................................................................................................... 70
Table 4.2a Number of failed firms and their counterparts in the sample ............................... 73
Table 4.3 List of failed firms based on the matching criteria (20032008) ........................... 74
Table 4.4 List of failed firms and their nonfailed mates (20032008) ................................ 75
Table 4.5 Breakdown of companies in the sample by 20quarter investigation period ......... 77
Table 4.6 Original list of the explanatory variables .............................................................. 79
Table 4.7 Summary of the 16 financial ratios ...................................................................... 87
Table 5.1 Descriptive profile of companies used ............................................................... 109
Table 5.2 Descriptive profile of the availability of financial statements ............................. 110
Table 5.3 Descriptive profile between both groups for the EMS model ............................. 111
Table 5.4 Descriptive profile between both groups for the 15 ratios ................................... 113
Table 5.5 Results of the EMS model application for the failed group................................. 117
Table 5.6 Results of the EMS model application for the nonfailed group ......................... 119
Table 5.7 Results of the comparative ratio analysis application for the failed group ........... 122
Table 5.8 Results of the comparative ratio analysis application for the nonfailed group ... 124
Table 5.9 Results of the ratio trend analysis application for the failed group ...................... 127
Table 5.10 Results of the ratio trend analysis application for the nonfailed group ............ 128
Table 5.11 Successful classification rate for each measure ................................................ 130
Table 5.12 Results of the IMM approach application for the failed group .......................... 131
Table 5.13 Results of the IMM approach application for the nonfailed group .................. 134
Table 5.14 Successful classification rate for the IMM approach ........................................ 136
Table 5.15 Testing for correlation among the 16 financial ratios ........................................ 138
Table 5.16 Testing for normality of the 16 financial ratios ................................................. 139
Table 5.16a The structure of the classification table .......................................................... 143
Table 5.17 The classification table of Model I ................................................................... 145
Table 5.18 The Overall classification accuracy and the type I error of Model I .................. 148
Table 5.19 The classification table of Model II .................................................................. 150
Table 5.20 The Overall classification accuracy and the type I error of Model II ................. 151
Table 5.21 The classification table of Model III ................................................................ 154
Table 5.22 The Overall classification accuracy and the type I error of Model III ............... 155
Table 5.23 The classification table of Model IV ................................................................ 158
Table 5.24 The Overall classification accuracy and the type I error of Model IV ............... 159
Table 5.25 The classification table of Model V ................................................................. 162
Table 5.26 The Overall classification accuracy and the type I error of Model V ................ 163
viii
Table 5.27 Comparison of overall classification accuracy and type I error (Model I to Model
V) ..................................................................................................................................... 164
Table 5.28 The classification table of Model VI ................................................................ 168
Table 5.29 The Overall classification accuracy and the type I error of Model VI ............... 170
Table 5.30 The classification table of Model VII ............................................................... 172
Table 5.31 The Overall classification accuracy and the type I error of Model VII .............. 174
Table 5.32 Comparison of overall classification accuracy and type I error (Model VI and
Model VII) ........................................................................................................................ 175
ix
List of Figures
Figure 1.1 The structure of Chapter 1 .................................................................................... 3
Figure 1.2 Outline of this study ........................................................................................... 12
Figure 2.1 The structure of Chapter 2 .................................................................................. 16
Figure 2.2 Map of Thailand ................................................................................................. 17
Figure 3.2 Analytical approach of this study........................................................................ 56
Figure 4.2 Correlation between research issues and research methodology .......................... 60
Figure 4.3 Comparison of the major components of research design and the focus of this
study ................................................................................................................................... 64
Figure 4.4 Relationship between a binomial dependent variable and an explanatory variable
in the logit model .............................................................................................................. 100
Figure 5.1 The structure of Chapter 5 ................................................................................ 105
Figure 5.2 Logit results for Model I ................................................................................... 141
Figure 5.3 Logit results for Model II ................................................................................. 149
Figure 5.4 Logit results for Model III ................................................................................ 153
Figure 5.5 Logit results for Model IV ................................................................................ 157
Figure 5.6 Logit results for Model V ................................................................................. 160
Figure 5.7 Logit results for Model VI ................................................................................ 167
Figure 5.8 Logit results for Model VII ............................................................................... 171
Figure 6.1 The structure of Chapter 6 ................................................................................ 180
x
Glossary of Abbreviations
ACMF ASEAN Capital Markets Forum
AFET Agricultural Futures Exchange of Thailand
ASEAN Association of Southeast Asian Nations
BEX Bond Electronic Exchange
BOT Bank of Thailand
CACL Current Assets/Current Liabilities
CAS Current Assets/Sales
CATA Current Assets/Total Assets
CCL Cash/Current Liabilities
CFCL Cash Flow/Current Liabilities
CFTA Cash Flow/Total Assets
CFS Cash Flow/Sales
CFTL Cash Flow/Total Liabilities
CLTA Current Liabilities/Total Assets
CLTE Current Liabilities/Total Equity
CS Cash/Sales
CTA Cash/Total Assets
DE Debt/Equity
EBITI Earnings before Interest and Taxes/Interest
EBITS Earnings before Interest and Taxes/Sales
EBITTA Earnings before Interest and Taxes/Total Assets
EBITTE Earnings before Interest and Taxes/Total Equity
EBTTA Earnings before Taxes/Total Assets
EMS Emerging Market Score
FX Foreign Exchange
GDP Gross Domestic Products
IMF The International Monetary Fund
IMM The Integrated Multi-Measure
InvS Inventory/Sales
InvWC Inventory/Working Capital
IOSCO International Organization of Securities Commissions
LTLTA LongTerm Liabilities/Total Assets
LTLTE LongTerm Liabilities/Total Equity
M Mean value
MAI Market for Alternative Investment
MDA Multiple Discriminant Analysis
NIS Net Income/Sales
NITA Net Income/Total Assets
NITE Net Income/Total Equity
NNA Neural Network Analysis
OTC OverTheCounter
QR Quick Assets/Current Liabilities
QAS Quick Assets/Sales
QATA Quick Assets/Total Assets
RETA Retained Earnings/Total Assets
ROA Net Income/Total Assets
ROE Net Income/Shareholders Equity
xi
SD Standard Deviation
SEC The Securities and Exchange Commission
SET Stock Exchange of Thailand
SFA Sales/Fixed Assets
SINV Sales/Inventory
SME Small and Medium Enterprise
STA Sales/Total Assets
STE Sales/Total Equity
TELTL Total Equity/LongTerm Liabilities
TES Expenses/Sales
TETA Total Equity/Total Assets
TETL Total Equity/Total Liabilities
TFEX Thailand Futures Exchange
TLTA Total Liabilities/Total Assets
TLTE Total Liabilities/Total Equity
WCS Working Capital/Sales
WCTA Working Capital/Total Assets
xii
Chapter 1: Introduction
1.1 Introduction
There are a number of reasons why business entities disappear from the marketplace. They
may be financially distressed, or liquidated, or they may be acquired by another company.
Stakeholders, for instance employees, bank creditors, stockholders, community, and
government (Fitzpatrick 1931) are likely to suffer from their investment no matter what the
event that triggered the entitys disappearance. Given that business failure can cause
significant trauma (i.e. high costs and heavy losses) to these stakeholders, its prediction is
highly beneficial. This motivates researchers to find a tool to detect unfavourable symptoms
before the entity disappears. Fitzpatrick (1931) and Merwin (1942) were the first researchers
who attempted to identify the potential of financial ratios as indicators of financial distress.
Subsequently, Altman (1968) introduced the more complex and sophisticated approach of
multivariate analysis using financial ratios as a tool to signal financial distress.
The purpose of this study is to examine the ability of financial ratios to signal financial
distress utilising techniques which are well developed and have met with wide acceptance in
the discipline. To this end, the investigation employs comparative ratio analysis, ratio trend
analysis and the Altmans Emerging Market Score (EMS). In addition, the logit-based model
is employed as a benchmarking measure. It should be noted that this study does not aim to
find the best indicators for Thai financial distress even though a number of specifications of
the logit model are estimated to find the best explanatory power in predicting the probability
of financial distress.
A number of theoretical concepts relevant to financial distress inform the analysis. The study
applies descriptive and inferential statistics and develops the Integrated MultiMeasure (IMM)
approach together with the logit model to evaluate the ability of financial ratios to signal
financial distress. Data for Thai listed firms covering the period 1998 - 2007 formed the basis
of the study. This chapter presents an introduction and lays the foundation for the subsequent
chapters.
1
The chapter is organised into nine sections as depicted in Figure 1.1. Section 1.1 outlines the
studys objective together with the structure of the chapter. Section 1.2 discusses the
background to the research. Research issues are described in Section 1.3, followed by a
justification for the research in Section 1.4. An overview of the methodology is presented in
Section 1.5. After principal definitions used in this study are presented in Section 1.6, the
outline of this study is discussed in Section 1.7.
2
Figure 1.1 The structure of Chapter 1
1.1 Introduction
(Objectives and structure of the chapter)
1.2 Background to the research
1.3 Research issues
1.3.1 Research problem 1.3.3 Research objective
1.3.2 Research question
1.4 Justification for the research
1.5 Methodology
1.6 Principal definitions used
1.7 Outline of this study
Source: Developed for this thesis
1.2 Background to the research

It is believed that formalisation of attempts to detect financial symptoms of unsuccessful
businesses began in the 1930s (for instance, in the studies of Fitzpatrick 1931; and Merwin
1942). However, prediction of corporate distress events in the US gathered momentum from
1970 onwards (Altman & Hotchkiss 2006). The frequently quoted studies in this field are
Beavers (1966) study and Altmans (1968) study. These two financial ratiobased studies
aim to introduce and/or develop an appropriate instrument for signalling business failure
before the unfavourable event happens.
3
In Thailand, the number of unsuccessful business entities can be identified by the number of
company dissolutions. According to the Ministry of Commerces business registration
statistics, the number of dissolutions for Thai companies lay between 20,000 and 30,000 in
each year during the period 2000 to 2008. With reference to these figures, it can be deduced
that the problem is persistent and critical in Thailand. The dissolution types include:
dissolved, defunct, bankrupt, and others (Ministry of Commerce 2009). More discussion on
this issue is provided in Chapter 2.
In the context of studying Thai business failure, most extant studies focus on the period of the
1997 financial crisis (for instance, the studies of Persons 1999; Reynolds et al. 2002; and
Tirapat & Nittayagasetwat 1999). Like western research (for instance, the studies of Altman
1968; Beaver, McNichols & Rhie 2005; Hossari 2006; and Shumway 2001), Thai studies use
both financial statement ratios and a combination of financial statement ratios and non
financial factors to search for the best indicators for business failure. This type of research is
dominant in Thailand. In other words, there is little empirical research focussing on
examining the ability of financial statements to signal business failure, particularly in the
Thai context. In addition, to date only annual financial statements have been used in the field.
Given this, this study is designed to investigate the informative ability of financial statements
to signal early warnings of financial distress for Thai business entities during periods of
normal economic circumstances. Furthermore, this study applies quarterly financial statement
data, instead of annual financial statements, so as to take into the account seasonal and/or
cyclical changes, and to access a more uptodate data set and potentially provide earlier
indications of company financial distress.
1.3 Research issues

This section identifies the research issues of this study. The issues help outline the research
boundary and provide an appropriate guideline for the development of analytical approach of
this research. In this study, research issues constitute the research problem, research question,
and research objective.
4
1.3.1 Research problem
The problem addressed in this research is:
Whether financial statement information can be adequately used to predict

financial distress for Thai firms in normal economic circumstances
Within the context of Thailand, a combination of financial statement ratios and non
financial factors has been used in order to find the best predictor for business failure.
Also, previous Thai studies have focused on periods of financial crisis. Given this,
this study is designed to examine whether financial statement information can be
adequately used to predict financial distress for Thai firms in normal economic
circumstances. The approaches utilised were comparative ratio analysis, ratio trend
analysis, the EMS, and estimated logit models. Adopting the conceptual framework
for a multimeasure design, this study introduces the IMM approach as an analytical
tool to test the power of financial statement ratios in signalling financial distress. The
IMM approach consists of integrating the results of a sophisticated prediction model
(the EMS) and two traditional analytical measures (comparative ratio analysis and
ratio trend analysis). In other words, the IMM approach is a means of obtaining
triangulation. Employing the triangulation approach is likely to provide a broader
perspective for the phenomena being studied. The development of the IMM approach
will be discussed later in the thesis.
1.3.2 Research question

The question addressed in this research is:
Can financial statement information be used to discriminate between potentially

failing and nonfailing firms in the context of Thailand in normal economic
circumstances?
The research question is developed in response to the research problem. In addition,

the question helps determine more clearly the research boundary.
5
1.3.3 Research objective
With reference to the research problem and question above, the objective of this study
is:
To investigate whether financial statement information can adequately classify

potentially failing and nonfailing firms in the context of Thailand in normal
economic circumstances
Addressing the issues in this study provides contributions that will be presented in
Section 6.4 of Chapter 6. In summary, this research made a number of contributions:
The study expands the body of knowledge about financial statementsbased

modelling of business failure in Thailand, one of a number of emerging
market countries. Little empirical research has been conducted on this problem
in emerging market countries.
This study has refined a methodology which uses ratiobased signalling for
predicting failure by adopting quarterly financial statement data rather than
annual financial data.
Another methodological refinement is that this study developed the Integrated
MultiMeasure (IMM) approach for data analysis. The IMM approach is a
combination of a sophisticated ratiobased model and traditional analytical
tools for financial statement analysis. This approach to combining the
indicators is not reported in previous studies reviewed in the literature. The
underlying reason for developing the IMM approach is that no one measure is
significantly more accurate than the others. Therefore, the IMM approach is
useful as a means of triangulation and helps confirm that the findings were
thoroughly examined.
As far as the author has been able to establish, this study is the first attempt to
examine the predictive ability of financial statements in normal economic
circumstances in the Thai context.
The descriptive statistical findings of this study provide a broader perspective
on the incidence of failure.
6
Stakeholders such as regulatory bodies, managers, and analysts could use the
studys findings as a foundation to further develop efficacious prediction
models for financial distress.
1.4 Justification for the research

According to statistics provided by the Ministry of Commerce (2009), the number of
companies which disappeared from the registry ranged between 20,000 and 30,000 in most
years of the recent decade. In consequence, company dissolution is a serious problem in
Thailand and adversely affects the economy and stakeholders of the failed firms. Since
business failure is a persistent problem and a critical issue in Thailand, this study investigates
a small but crucial sub-set of those firms, that is, firms listed on the Stock Exchange of
Thailand. The rationale for excluding non-listed companies is because of data accessibility
constraints.
Even though research on signalling symptoms of financial distress can be traced back to the
1930s, most studies have been conducted in western countries. Little research on financial
distress has been conducted in emerging markets like Thailand. To date, there has been a
relative neglect of empirical research in investigating the ability of financial statement
information to signal financial distress in Thailand during normal economic circumstances.
Most Thai studies, in the past, focused on finding the best predictors for failure by using both
financial statement ratios and nonfinancial factors. Also the studies concentrated on periods
of financial crisis. To take into the account seasonal and/or cyclical changes, quarterly
financial statements were employed in this study.
The results of this study are likely to be useful to stakeholders such as regulatory bodies,
managers, and analysts. These stakeholders could use the studys findings to further develop
efficacious prediction models for failure. For example, Thai regulatory bodies are obliged to
prevent, rather than protect, the potentially adverse incidence not only in the period of
financial crisis. This studys findings could equip Thai authorities with alternative methods to
detect potential corporate failure of regulated business entities before the event happens.
7
Another example is that the findings of this study could not only help Thai private sector
managers improve internal management but could also help them keep a track record of a
companys performance. In addition, this studys findings could provide financial analysts
with an alternative analytical tool for deciding whether to invest in companies.
1.5 Methodology
This section is designed to provide an introductory overview of the methodology employed in
this study. The indepth details of the methodology are described in Chapter 4. This section
also introduces the data collection and data analysis methods. Finally, the findings from the
data analysis are briefly discussed.
Data collection method

To satisfy the purpose of the study, historical financial statement information of Thai
firms was collected. Out of the 44 financial ratios identified as potential predictors for
failure by previous studies, 16 ratios were used in this study. The 16 ratios are divided
into two groups; that is, the original four ratios for the Emerging Market Score model,
and 15 selected ratios for the two traditional analytical measures. The selection of the
15 ratios is based on their frequency of use in previous studies. The ratios are
classified into four meaningful areas of financial performance. They include liquidity,
turnover/performance, leverage/solvency, and profitability (Blum 1974; Brigham &
Ehrhardt 2008; Persons 1999; and Ross et al. 2007). More details of the 16 ratios are
discussed in Chapter 4.
Like most extant studies, the target samples in this study are publicly listed companies.
The rationale for choosing listed companies is data accessibility. The required data are
derived from the Stock Exchange of Thailands database. The cutoff period between
2003 and 2008 was chosen because it warrants investigating most recent business
failure of Thai companies in a normal economic period. As mentioned earlier, the
samples consist of a failed group and a nonfailed group. The paired sample
technique comparing a failed firm to a nonfailed firm (1:1 ratio) is adopted. The
8
justification for choosing the pair sample technique is to ensure that the sample of
firms is as similar as possible in all aspects except for their financial ratios (Norton &
Smith 1979, p. 73). In addition, the technique is frequently used by previous
researchers in estimating prediction models.
However, unlike previous studies, this study uses quarterly financial statement data in
place of annual financial statement data to better observe shifts in the values of ratios
over the investigation period. According to the SET database, the quarterly financial
statement data used in this study are audited. It is apparent that quarterly financial
statement data have been relatively neglected by previous research. The underlying
reason for adopting quarterly financial statement data is to detect any unfavourable
shifts of financial characteristics as soon as they happen. Finally, 14 failed firms and
14 nonfailed firms were included in the final sample for analysis. The data set
employed the entire population of listed firms which met the analytical criteria for the
period 2003 to 2008 and a look-back period of five years financial statement data to
1998 was incorporated. More discussion on the data collection method is presented in
Chapter 4.
Data analysis method

This study is designed to explore the ability of financial statement information to
predict business failure in the Thai context during normal economic circumstances.
The analytical techniques used were comparative ratio analysis, ratio trend analysis
and the EMS. Moreover, logit models were developed for comparative purposes. Like
previous studies, descriptive and inferential statistics are used to describe financial
characteristics of the sample set. The sample is divided into two groups, that is, failed
firms and nonfailed firms. Descriptive statistics used in this study include the mean
as the measure of central tendency and the standard deviation as the measure of
variation while inferential statistics include the ttest for differences between groups.
The underlying reasons for choosing these measures are that extreme values and data
dispersion are taken into consideration, and that these are the most frequently used
measures in the literature. More discussion of the measures is provided in Chapter 5.
9
This study developed an analytical tool to examine the ability of financial ratios in
signalling financial distress. The analytical tool in this study called the Integrated
Multi-Measure technique and this technique combined established approaches used in
the previous studies. They are comparative ratio analysis, ratio trend analysis and the
EMS. The first two represent traditional analytical measures while the last one
represents a sophisticated emerging market prediction model. To date, different ratio
based measures have never been used in combination to examine the predictive ability
of financial statements in Thailand. The integrated use of these measures provides
data triangulation aiming to observe broader dimensions of the phenomena being
studied (Easterby-Smith, Thorpe & Lowe 1991). To validate the results of the IMM
approach, logit models were also estimated in this study. Indepth discussion (i.e. the
justification for selecting these measures, and the criteria used in each measure) is
provided in Chapter 4.
Brief summary of the findings

When each of the analytical approaches was applied individually, the rates of
successful classification were similararound 57 per cent. This finding is consistent
with the literatures finding that no specific technique is dominant over its
counterparts in the context of signalling business failure
The IMM approach using ratiobased measures was adopted in this study so as to
triangulate and confirm the findings. The results provided by the IMM approach are
promising. The successful classification rate increased from 50 per cent (the lowest
rate based on the individual measure approach) to 64 per cent for the financially
distressed firms. Moreover, the results derived from the estimated logit models
revealed correct classification of the financially distressed firms of up to 86 per cent.
More discussion of the findings is provided in Chapter 5.
10
1.6 Principal definitions used
In the context of signalling business failure, definitions adopted by researchers are not
uniform and depend on the specific interest or condition of the firms being studied
(Dimitras, Zanakis & Zopounidis 1996, p. 487). Therefore, it is imperative to elaborate key
terms used so as to establish the positions taken in this study. The key terms used in this
study are as follows:
Business failure
A number of failure definitions are used in the literature. For instance, Altman and
Hotchkiss (2006, p. 4) define failure as existing when the realized rate of return on
invested capital is significantly and continually lower than prevailing rates on similar
investments. Beaver (1966, p. 71) defines failure as the inability of a firm to pay
its financial obligations as they mature.
The determination of business failure in this study is more subtle than in most
previous studies because most previous studies use definitions which are too narrow
and exclusive. In this study, the definition of failure by Altman and Hotchkiss (2006)
is favoured and adopted.
Financial distress
Like failure, a multitude of definitions of financial distress have emerged in the
literature. Researchers identify distressed companies based on several financial
dimensions. Financial distress conditions are represented by business restructuring
or reorganisation (Routledge & Gadenne 2000), failure to pay annual listing fees
(Jones & Hensher 2004), debt default criteria (Kahya & Theodossiou 1999), going
private for a publicly listed company (Queen & Roll 1987), liquidation and
acquisition (Coats & Fant 1993) and the like.
As can be seen, there are a number of conditions representing financial distress in

the literature. This is consistent with Coats and Fants (1993, p. 147) notion that
11
bankruptcy is one outcome of financial distress. In other words, distressed
companies are financially weak but do not always become bankrupt (Gilbert, Menon
& Schwartz 1990; and Hill, Perry & Andes 1996). Thus, the broader definitions of
financial distress used by the aforementioned researchers are employed in this study.
The adoption of this broader concept should be more meaningful to decisionmakers
in Thailand. In this study, both business failure and financial distress are used
interchangeably.
1.7 Outline of this study

This study is presented in six chapters as depicted in Figure 1.2. Each chapter is briefly
discussed below.
Figure 1.2 Outline of this study
Chapter 1: Introduction
Chapter 2: The Thai Environment
Chapter 3: Literature Review
Chapter 4: Research Methodology
Chapter 5: Data Analysis
Chapter 6: Conclusions and Implications
Chapter 1 provides an introductory overview of the study. Background to the research is

firstly discussed, followed by research issues constituting the research problem, research
question, and research objective. After that, the justification for the research and
12
methodology are described. The studys outline is provided. In the last section, key
definitions used and limitations of the study are provided.
Chapter 2 is designed to shed the light on Thailands environment to set the context of the
research. The chapter starts by presenting general information on Thailand including
geography, population, government/politics, and economy. Then, the discussion turns to Thai
financial markets and key market players, for instance, the Stock Exchange of Thailand, the
Bank of Thailand, and the Securities and Exchange Commissions. In the end, the chapter
sheds the light on recent Thai business failures.
Chapter 3 reviews the literature in the field of business failure. The chapter commences with
the key aspects of business failure prediction (e.g. generic terms used in previous studies, the
cost of business failure, and the usefulness of business failure models). To find gaps in
business failure studies in Thailand, the evolution of business failure prediction models both
in Thailand and elsewhere, together with an overview of the theoretical frameworks, is
provided. After the gaps are identified, the chapter addresses the research problem, research
question, and research objective. Finally, the discussion turns to the development of the
analytical approach used in this study.
Chapter 4 discusses the methodology used in this study. The justification for choosing the
research design (for example, exploratory research, and secondary data technique) based
upon the purpose of the study is provided. The sampling design is described before a
discussion of data collection. Next, the chosen three ratiobased measures (the Emerging
Market Score model, comparative ratio analysis, and ratio trend analysis), together with the
criteria used in each measure, are presented. After that, a discussion of the development of
the IMM approach along with the criteria used is presented. Finally, a discussion of utilising
the logit model and its justification concludes the chapter.
13
Chapter 5 is designed to present the findings of this study. The chapter commences with
descriptive data about the sample used. Both descriptive profiles of companies and financial
ratios used are discussed. Descriptive statistics measures (for instance, means, and standard
deviation) and inferential statistics (e.g. ttest) are adopted. The data are analysed in
accordance with the methodology described in Chapter 4. In this respect, individual
applications and their results are presented first, followed by the IMM approach along with
its results. After that, the tests for correlation and normality of the 16 ratios are conducted
before estimating a number of specifications of the logit model. A discussion of logit model
results concludes the chapter.
To finish, Chapter 6 discusses the conclusions and implications of this study. The chapter
begins with conclusions concerning the research question and problem based on the findings
reported in Chapter 5. Implications for theory and for policy and practice are provided prior
to a discussion of the limitations of this study. Finally, the chapter concludes with a
discussion of implications for further research.
14
Chapter 2: The Thai Environment
2.1 Introduction
An introductory overview of this study was presented in Chapter 1. Since this study focuses
on business failure in Thailand, it is imperative to provide an overview of Thailand for people
who are not familiar with that country. As a result, this chapter sheds light on the Thai
environment. It includes (1) basic information about Thailand such as geography, population,
government/politics, and economy, (2) Thai financial markets (i.e. money markets and capital
markets), and (3) specific aspects of business failure in Thailand.
The chapter is organised into five sections as depicted in Figure 2.1. Section 2.1 outlines an
objective and the structure of the chapter. Section 2.2 describes Thailand in general,
particularly its geography, population, government/politics, and economy. After that, Thai
financial markets and key market players are described in Section 2.3. Section 2.4 presents
aspects of business failure in Thailand and a conclusion is drawn in Section 2.5.
15
2.1 Introduction
2.2 Thailand in general
2.2.1 Geography 2.2.3 Government/politics
2.2.2 Population 2.2.4 Economy
2.3 Thai financial markets and key market players
2.4 Thai business failure
2.5 Conclusion
2.2 Thailand in general

This section provides general information about Thailands geography (Subsection 2.2.1),
population (Subsection 2.2.2), government/politics (Subsection 2.2.3), and economy (Sub
section 2.2.4). The presentation of these aspects is intended to provide some insights into
characteristics of Thailand.
16
2.2.1 Geography
Thailand, or officially known as The Kingdom of Thailand, is situated in the heart of
Southeast Asia. Thailand is bordered by Myanmar to the west and northwest, the
Lao Peoples Democratic Republic to the northeast and east, Cambodia and the Gulf
of Thailand to the southeast, Malaysia to the south, and the Andaman Sea to the
southwest. Figure 2.2 is a map of Thailand.
Figure 2.2 Map of Thailand
Source: World Bank (n.d.)
Bangkok, known to Thais as Krung Thep, is the capital city of Thailand and is
located in the central region. Thailands climate is tropical, with high temperatures
and humidity, and is dominated by monsoons (United Nations Thailand n.d.).
17
2.2.2 Population
The population of Thailand is estimated to be approximately 64 million. Around 15
per cent (or approximately 9.3 million people) reside in Bangkok and its surrounds
(United Nations Thailand n.d.). Thai people are famous for their courtesy and
hospitality. Thais are predominantly Buddhist and Buddhism is the national religion.
The Thai language is the official language of the nation.
2.2.3 Government/politics
In an early period of its history, Thailand adopted a paternalistic form of government
(Government and Politics in Thailand n.d., p. 1). The major shift in the Thai political
system was made by King Chulalongkorn (18681910). Nowadays, Thailand is a
constitutional democratic monarchy. The Prime Minister is the head of government
while the monarch is the head of state and the head of the armed forces.
2.2.4 Economy
In its early periods, Thailand was an agricultural economy and at that time its major
economic products were rice, cassava, rubber, and sugarcane. The boom years for
Thailands economy were from the 1980s to the mid1990s, resulting from a shift to
industrialisation, government stability, and exportoriented policies (Encyclopedia of
the Nations n.d.). Likewise, the World Bank (n.d., p. 1) states that In the decade that
ended in 1995, the Thai economy was one of the worlds fastestgrowing at an
average rate of 89 per cent a year.
The boom period ended in 1997 when the 1997 Asian financial crisis emerged.
Thailand was one of the hardest hit Asian countries. According to Sookying (n.d., p.
175), the 1997 financial crisis stemmed from the rapid expansion in real estate,
construction and financial sectors that generated overinvestment and a high level of
external shortterm loans. Thailand had to ask for assistance from the International
18
Monetary Fund (IMF) to mitigate the impact of the crisis. In 2003, a final repayment
of the IMF debt was made by the Thai government (AccessMyLibrary n.d.).
From 2002 to 2004, Thailand was one of the best performers among East Asian
countries with an average of more than six per cent annual real GDP growth (Central
Intelligence Agency n.d.). Since the Thai economy is exportdependent, exports of
goods and services played a major role in helping Thailand recover from the 1997
crisis and in increasing the GDP growth rate (Tourism Authority of Thailand n.d.).
The main export markets of Thailand include the United States, Japan, and Europe.
Besides the traditional export markets, Thailand is attempting to penetrate new export
markets such as India, China, and the Middle East (Travel Document Systems n.d.).
Nevertheless, a declining GDP growth rate of 2.6 per cent in 2008 was reported,
resulting from domestic political chaos and the global financial crisis (Central
Intelligence Agency n.d.).
Having provided an overview of Thailand, the next task is to discuss Thai financial
markets and key market players.
2.3 Thai financial markets and key market players

This section is designed to shed the light on the overall characteristics of Thai financial
markets and major market players. Like financial markets in other countries, Thai financial
markets are regarded as a critical mechanism that drives the Thai economy. In Thailand,
financial markets can be categorised into five major markets as follows:
The Foreign Exchange Market

The four basic instruments of this market include spot, forward, Foreign Exchange
(FX) swap, and FX derivatives (BOT n.d.). The key market players are authorised
banks which have been awarded FX licences from the Bank of Thailand (BOT n.d.).
19
The Money Market
Unsecured interbank borrowing, trading for shortterm financial instruments (i.e.
treasury bills, promissory notes and bills of exchange), and repurchase agreements are
the most popular types of transactions in this market (BOT n.d.). Major market
players are banks, financial institutions, large corporations, and state enterprises (BOT
n.d.).
The Derivatives Market

The Thailand Futures Exchange (TFEX), a subsidiary of the Stock Exchange of
Thailand, was founded in 2004 aiming to be a world class exchange for the trading
and hedging of Thai Derivatives products (TFEX n.d., p. 1). SET50 Index Futures
was launched in 2006 and SET50 Index Options was unveiled in 2007. TFEX
launched Stock Futures and Gold Futures in 2008 and 2009 respectively (TFEX n.d.).
However, before the TFEX was established, the Agricultural Futures Exchange of
Thailand (AFET) was set up in 1999 (AFET n.d.). AFET is supervised by the
Agricultural Futures Trading Commission, and initially launched its products in 2004,
that is, Natural Rubber Ribbed Smoked Sheet No.3 and White Rice 5% Broken. In
2005, tapioca starch premium grade was launched into the market (AFET n.d.).
The Debt Market

This market facilitates debt instruments trading for business entities. After the 1997
Asian financial crisis, bonds gained popularity in the market (BEX n.d.). Major
players in this market are rather heterogeneous. They include financial institutions,
large business entities, and retail investors (BOT n.d.). At present, the official
secondary debt market of Thailand is the Bond Electronic Exchange (BEX). BEX was
established in 2003 by the Stock Exchange of Thailand to develop the Thai bond
market (BEX n.d.).
20
The Equity Market, or better known as the stock market, (hereafter referred to as
stock market)
In the Thai stock market, equity products can be divided into six categories. They
include ordinary shares, preferred shares, warrants, unit trusts, nonvoting depository
receipts, and depository receipts (SET n.d.). Equity holders typically have the right to
vote and participate in major business decisions of a company that issues the security
(SET n.d., p. 1). In Thailand, there are two stock marketplaces. One is the Stock
Exchange of Thailand and the other is the Market for Alternative Investment. The
Market for Alternative Investment is presented first, followed by the Stock Exchange
of Thailand.
o The Market for Alternative Investment (MAI)

The MAI was officially set up under the Securities Exchange of Thailand Act and
the presence of the MAI enables Small and Medium Enterprise (SME) operators
to access alternative longterm funding sources (SET n.d.).
SME operators who are eligible to register in the MAI must have paidup capital
of not less than 20 million baht but less than 300 million baht (SET 2003).
Applicants who are eligible to register in the Stock Exchange of Thailand must
have paidup capital of at least 300 million baht (SET 2001). MAIs procedures
regarding security trading, settlement and trading hours are similar to those of the
Stock Exchange of Thailand (Vichyanond 2002).
o The Stock Exchange of Thailand (SET)

In 1975, the Securities Exchange of Thailand was established under the Securities
Exchange of Thailand Act., B.E. 2517 (1974) and commenced operation. From
1991 onwards, its name was changed to the Stock Exchange of Thailand (SET
n.d.). According to SET (n.d., p. 2), the three principal roles of SET are:
21
1) To serve as a centre for the trading of listed securities, and to provide the
essential systems needed to facilitate securities trading;
2) To undertake any business relating to the Securities Exchange, such as a
clearing house, securities depository centre, securities registrar, or similar
activities;
3) To undertake any other business approved by the Securities and Exchange
Commission (SEC).
The Board of Governors of the SET has 11 members. The major responsibilities
of the Board are to formulate the SET policies with the approval of the SEC and
to supervise the Exchanges operations (SET n.d.). The SET represents Thailand
in three international organisations. They include the Asian and Oceanian Stock
Exchanges Federation (joined in 1982), the World Federation of Exchanges
(joined in 1990), and the International Organization of Securities Commissions
(IOSCO) (joined in 1990).
Finally, the most recent reported years of the SET statistical highlights are
illustrated in Table 2.1 below.
Table 2.1 SET statistical highlights from 2006 to 2009
2006 2007 2008 2009

SET index 680 858 450 735
Market capitalization (billion 5,079 6,636 3,568 5,873
baht)
Daily average turnover 16,281 17,097 15,870 17,854
(million baht)
Market dividend yield (%) 4.24 3.31 6.57 3.65
Market P/E ratio 9.44 17.03 7.01 25.56
Number of listed companies 476 475 476 475
Number of delisted companies 4 8 9 8
Number of listed securities 584 581 580 588
Note: On 23/07/09, THB 28.05 was equal to AU$ 1, according to the Bank of
Thailand.
Source: Market Statistics (SET n.d.)
22
Having presented the Thai financial markets along with key market players, the
discussion now turns to the two key regulatory bodies of Thai financial markets.
Like other countries, to ensure wellfunctioning financial markets in Thailand,

the Bank of Thailand (BOT) and the Securities and Exchange Commission (SEC)
were set up by the government. The major aims of the two regulatory bodies are
to supervise, monitor, and develop the financial markets. BOT is responsible for
the Thai monetary market while the SEC regulates the countrys capital market.
o The Bank of Thailand (BOT)

BOT was first established as the Thai National Banking Bureau and commenced
operations in 1942 (BOT n.d., p. 1). The BOT board has 13 members (BOT n.d.).
The presence of BOT in the international organisations not only helps promote
the stability and soundness of the global financial and economic system but also
enhances financial cooperation and macroeconomic surveillance amongst
member countries (BOT n.d., p. 1).
o The Securities and Exchange Commission (SEC)

The SEC was set up in 1992 under the Securities and Exchange Act B.E. 2535
(1992). The SEC is responsible for supervising and developing both primary and
secondary capital markets. The key roles of SEC are to (1) maintain orderly
market, (2) ensure investor protection, (3) foster business innovation, and (4)
promote competition (SEC 2009b, p. 1).
SEC is made up of 10 members (SEC 2009c). SEC has joined the ASEAN Capital
Markets Forum (ACMF), which was founded in 2004 as a cooperative venture by
the ASEAN Finance Ministers (SEC 2009a). Besides ACMF, SEC is a member of
the International Organisation of Securities Commissions, a leading international
forum for securities agents (SEC 2009a).
23
This concludes the description of the Thai financial markets and key market players. As
stated earlier, this chapter would be remiss if it did not mention Thai business failure. This
issue is discussed in the next section.
2.4 Thai business failure

This section examines aspects of business failure in Thailand. There are a number of reasons
why companies disappear from the public marketplace. They include going bankrupt, being
liquidated, and even going private (Queen & Roll 1987). During the 1997 financial crisis in
the Southeast Asia, Thailand was one of the hardesthit countries. At that time, banks and
financial institutions were not willing to grant any loans to customers, mainly due to concerns
about nonperforming loans. This action of the banks and financial institutions exacerbated
the economic conditions of financially troubled companies.
In the context of studying business failure, a number of indicators (e.g. company bankruptcy,
company liquidation, and even publicly listed company delisting) were used in previous
studies to indicate the event (see review in Chapter 3). This study adopts company dissolution
as an indicator of business failure. Figures provided by the Ministry of Commerce show the
numbers of dissolutions for juristic persons during the postcrisis period (20002009) and
they are depicted in Table 2.2. A juristic person is defined as a limited company, limited
partnership, ordinary partnership, or public limited company (Ministry of Commerce 2009).
24
Table 2.2 Numbers of dissolutions for juristic persons between 2000 and 2009
Dissolution
Year Dissolved Defunct Bankrupt Others Total
2000 8,327 20,872 249 72 29,520
2001 10,944 9,369 259 78 20,650
2002 16,106 9,940 365 147 26,558
2003 18,690 7,851 536 216 27,293
2004 17,913 6,721 581 291 25,506
2005 18,421 7,868 493 272 27,054
2006 17,466 2,625 555 358 21,004
2007 16,050 8,635 525 465 25,675
2008 16,580 13,803 433 594 31,410
2009 17,073 45,509 429 1,367 64,378
Total 157,570 133,193 4,425 3,860
Source: Ministry of Commerce (2009)
According to Table 2.2, the number of companies which disappeared from the registry ranged
between 20,000 and 30,000 in most years of the recent decade. The chosen study period is
between 2003 and 2008 (excluding the 2009 period) and a range of 10 years financial
statements (1998-2007) was examined. The reason for including the 2009 figures in this table
is to show the most recent number of companies which disappeared from the registry. It can
be inferred that the problem is persistent and critical in Thailand. The dissolutions are
classified into four common groups: dissolved groups, defunct groups, bankrupt groups, and
an other group (those that do not lie in any of the previous groups). Apparently, the most
common form of the dissolution is the dissolved group while the least form is other.
Defunct comes second and bankrupt comes third.
Besides the four common forms, dissolutions can be classified according to business category.
This provides information relevant to business failure. The recent data on dissolution based
on business category is for December 2009, and is depicted in Table 2.3.
25
Table 2.3 Number of dissolutions classified by business category (December 2009)
Business category Number of company dissolution
Wholesale and retail trade; repair of motor vehicles,
motorcycles and personal and household goods 2,664
Real estate, renting and business activities 1,174
Manufacturing 719
Construction 708
Transport, storage and communications 393
Source: Ministry of Commerce (2009)
Table 2.3 shows the top five categories for Thai company dissolutions in December 2009. As
can be seen, the most common category is wholesale and retail trade; repair of motor vehicles,
motorcycles and personal and household goods with 2,664 dissolutions. The smallest
category in the top five is transport, storage and communications with 393 dissolutions.
Having presented the aspects of business failure in Thailand, the discussion now turns to the
conclusion of this chapter.
2.5 Conclusion
This chapter provides an overview of Thailand for those who are not familiar with the
country. Basic information about Thailands geography, population, government/politics, and
economy is provided in the beginning section. After describing this basic information, the
discussion turned to Thai financial markets and key market players. The principal five
markets were discussed. They include the foreign exchange market, the money market, the
derivatives market, the debt market, and the equity market.
Then, the two main stock marketplaces were presented. They are the Market for Alternative
Investment (MAI) and the Stock Exchange of Thailand (SET). The MAI was established to
enable SME operators to access longterm funding sources while the SET was established to
assist corporations to access funding sources. Also, the recent SET key statistical highlights
were shown. Besides the two bourses, the two key regulatory bodies for Thai financial
markets were discussed. These are the Bank of Thailand (BOT) and the Securities and
26
Exchange Commission (SEC). The BOT is responsible for Thai financial markets while the
SEC is in charge of Thai capital markets.
Finally, the aspects of business failure in Thailand were discussed. Dissolution was used as
the indicator of business failure. The data for dissolutions are derived from the Ministry of
Commerce. The number of dissolutions during 20002009 was discussed, as was the number
of the dissolutions in five business categories. The next chapter will present a review of the
literature on business failure.
27
Chapter 3: Literature Review
3.1 Introduction
In financial research, the analysis of historical financial statement data to determine
distinguishing characteristics of firms which disappeared from the market is a well-
established practice. The use of past financial statement data for discriminating between
failed and nonfailed firms started in the 1930s (for example, in the studies of Fitzpatrick
1931; and Merwin 1942). In the 1960s, however, the seminal studies in the field were
conducted by Beaver (1966) and Altman (1968). Their findings reveal that financial
statement data are useful in predicting business failure. Subsequently, these works were
extended by studies using either the same or different statistical techniques to search for
efficacious predictors of business failure (Sung, Chang & Lee 1999).
This chapter reviews the literature on utilising either financial statement data or a
combination of financial statement data and nonfinancial factors (e.g. interest rates, and
inflation rates) as predictors for business failure. This review covers traditional studies of
business failure prediction, recent business failure prediction studies, international and Thai
business failure prediction studies. The main objectives of this chapter are to review extant
studies that use financial statement data and/or nonfinancial factors in developing failure
prediction models, to review international studies and Thai studies in business failure
prediction, and, finally, to identify research gaps. Many of the studies reviewed below focus
on development of models to predict business financial distress. These studies are reviewed
to inform our choice of the best methods to use for our analysis of whether financial
statement information remains useful for signalling the potential for financial distress in Thai
firms in normal economic periods.
The chapter is made up of seven sections as shown in Figure 3.1. Section 3.1 presents an
introduction to the literature review together with the objectives and structure of the chapter.
Section 3.2 discusses the important aspects of business failure prediction. Section 3.3
presents the evolution of business failure prediction. Section 3.4 identifies limitations of the
28
previous research. Section 3.5 develops the analytical approach of this study and Section 3.6
draws a conclusion to the chapter.
3.1 Introduction
3.2 The important aspects of business failure prediction
3.2.1 The definition of generic terms 3.2.3 The usefulness of business failure
used in business failure prediction prediction
3.2.2 The costs of business failure
3.3 The evolution of business failure prediction
3.3.1 Theoretical framework overview 3.3.2 Previous research overview
3.3.2.1 International studies
3.3.2.2 Thai studies
3.4 Limitations of previous research
3.5 Development of the analytical approach
3.6 Conclusion
29
3.2 The important aspects of business failure prediction
This section discusses the definition of generic terms used in business failure prediction
(Subsection 3.2.1), the costs of business failure (Subsection 3.2.2), and the usefulness of
business failure prediction models (Subsection 3.2.3). This section aims to depict the
important elements of business failure prediction.
3.2.1 The definition of generic terms used in business failure prediction

Since the 1960s, both bankruptcy and failure have been commonly used in the
study of corporate failure prediction. Even though these terms are often used
interchangeably, failure together with financial distress is preferred in this study
because the event of business failure and/or financial distress implies that firms are
financially weak but do not always become legally bankrupt (Gilbert, Menon &
Schwartz 1990; Hamer 1983; Hill, Perry & Andes 1996; and Scott 1981). In this
respect, failure and/or financial distress provide a broader dimension of the
phenomenon being studied and, importantly, these terms suit the intent of this study,
which is to focus on periods of normal economic circumstances. The following
discussion reviews evidence that bankruptcy provides a restrictive perspective.
Balcaen and Ooghe (2006) found that most studies used legal interpretations of the
relevant terms to discriminate failed and nonfailed firms. This is consistent with the
finding of Charitou, Neophytou and Charalambous (2004, p. 473) that a legal
definition of failure is widely applied because it provides an objective criterion that
allows researchers to easily classify the population of firms being examined. In
addition, Karels and Prakash (1987, p. 575) state that bankruptcy is a process which
begins financially and is consummated legally. Even though this notion might be
misleading because not all financially distressed/troubled firms go bankrupt,
bankruptcy is the most popular term chosen in failure definition (Balcaen & Ooghe
2006). Some studies, nevertheless, employed nonlegal financial dimensions such as
cumulative negative earnings, reductions in dividends, violations of debt covenants,
and the like to distinguish failed and nonfailed firms.
30
Researchers who use legal definitions of failure in their study include Altman (1968);
Beaver, McNichols and Rhie (2005); Charitou, Neophytou and Charalambous (2004);
Chi and Tang (2006); Gentry, Newbold and Whitford (1985); Hamer (1983); He and
Kamath (2006); McKee (2000); Norton and Smith (1979); Platt and Platt (1990;
1991); Pongsatat, Ramage and Lawrence (2004); Shumway (2001); Sung, Chang and
Lee (1999); and Wilcox (1973).
On the other hand, researchers who explicitly define business collapse by employing
economic, or nonlegal, criteria in their studies include Beaver (1966, 1968a, 1968b);
Fitzpatrick (1931); Gilbert, Menon and Schwartz (1990); Hill, Perry and Andes
(1996); Kahya and Theodossiou (1999); and Merwin (1942). Finally, researchers who
employ both legal definitions and economic criteria in defining corporate collapse
include Blum (1974); Hing and Lau (1987); Jones and Hensher (2004); Leksrisakul
( 2004); Reisz and Perlich (2007); and Ugurlu and Aksoy (2006).
It is worth noting that the four common terms widely and consistently used in the
extant studies are failure, insolvency, default and bankruptcy (Altman & Hotchkiss
2006). Despite being used interchangeably, these terms, according to Altman and
Hotchkiss (2006), differ in their formal usage.
Failure, in terms of economic criteria, is defined as:
The realized rate of return on invested capital is significantly and continually

lower than prevailing rates on similar investments. It includes insufficient
revenues to cover costs and where the average return on investment is below
the firms cost of capital (Altman & Hotchkiss 2006, p. 4).
Insolvency is another economic criterion. Generally an insolvent firm is defined as

one that is not able to service its current debts due to a lack of liquidity and often
culminates in a declaration of bankruptcy. Insolvency, from a bankruptcy perspective,
31
nonetheless, denotes an excess of a firms total liabilities over total assets (Altman &
Hotchkiss 2006). The aforementioned definitions are confirmed by Charitou,
Neophytou and Charalambous (2004).
Default is another term that is frequently used in business failure prediction studies.
The relationship between debtor and creditor is used to determine whether or not
default exists. For instance, a debtor, who cannot maintain mutually agreed financial
ratios with creditors during an agreement period, is considered to be in technical
default (Altman & Hotchkiss 2006). Another example is that a debtor who cannot
service either an interest payment or the principal repayment of a scheduled loan/bond,
which is considered to be in legal default (Altman & Hotchkiss 2006).
Bankruptcy is defined both legally and economically by Altman and Hotchkiss (2006).
Technical bankruptcy, by economic criteria, is defined as when a firm experiences a
negative net worth position. Legal bankruptcy is indicated by a firms declaration of
bankruptcy in a court. In the US legal context, an ailing firm might file either for
liquidating its assets (Chapter 7) or for business reorganisation (Chapter 11) (Altman
& Hotchkiss 2006).
The costs of business failure are one of the important issues discussed in the study of
business failure prediction. Altman and Hotchkiss (2006, p. 93) state that the costs of
financial distress and legal proceedings under the Bankruptcy Code are important for
obvious practical reasons. Business failure costs, therefore, are briefly discussed
below.
3.2.2 The costs of business failure

Parties involved with a financially distressed company on the verge of business failure
are inevitably affected. Workers/managers, for example, suffer from low morale as
they need secure employment and a financially distressed company might lose good
32
staff and be unable to retain sufficient workers. Creditors suffer from potential
financial loss while the company is likely to be forced to repay its loans before
maturity; suppliers might reduce the terms of trade credit or even stop supplying
goods and services while the company may bear high costs for goods sold or cease
production; customers tend to shift their purchases to more stable firms to ensure
prompt delivery of goods and services while the company is likely to encounter low
volumes of sales; shareholders and investors may suffer from potential financial or
investment loss while the relationship between shareholders/investors and the
company is likely to be damaged (Chen & Merville 1999; Fitzpatrick 1931; and
Hertzel et al. 2006). The loss of relationships with key workers/managers, main
suppliers and big customers, and the loss of confidence of creditors and shareholders
are further examples of the financial and/or economic costs of business failure. To
provide empirical evidence on how to assess the costs of business failure, the study by
Chen and Merville (1999) is discussed next.
Chen and Merville (1999) examine the indirect costs of financial distress for 1,041
firms during 19821992. The indirect costs of financial distress are defined as the
opportunity costs suffered by the firm as the degree of solvency declines (Chen &
Merville 1999, p. 277). Examples include sales or profit loss resulting from loss of
customer confidence, investment opportunity loss and the loss of important supplier
relationships and key managers. Chen and Merville (1999) found that the average
profits loss was eight per cent of firm market value and it can be as high as 80 per
cent of market value.
The literature search conducted for this study found no evidence regarding costs of
business failure in Thailand. The area has been researched in the West so costs of
business failure in western countries were discussed above. Since business failure
prediction can be actively used for recovery by a firms management (Altman 1993),
these financial distress costs are preventable. It is obvious that business failure
prediction is advantageous to numerous parties. The following section, therefore,
discusses who benefits from business failure prediction.
33
3.2.3 The usefulness of business failure prediction
It is apparent that business failure can cause significant trauma to stakeholders.
Fitzpatrick (1931) states that seven groups of stakeholders affected by business failure
are workers, community, bondholders, merchandise and bank creditors, stockholders,
boards of directors, and government. Notwithstanding this, the term stakeholder is
defined extensively in subsequent studies but still includes the aforementioned seven
groups. Effective business failure prediction can help these stakeholders to avoid
and/or at least to reduce the adverse impact of business failure.
It is generally accepted that the business failure prediction is widely used for different
purposes by stakeholders (Charitou, Neophytou & Charalambous 2004; Dimitras,
Zanakis & Zopounidis 1996; Hamer 1983; Jones & Hensher 2004; Scott 1981; Tamari
1966; and Zavgren 1983). Managers/owners, for instance, use business failure
prediction as an early warning tool and can take appropriate action to avoid business
failure. Investors/shareholders use financial distress prediction to prevent the
considerable financial losses they incur when firms become insolvent. Creditors (i.e.
bank or financial institutions) use failure prediction techniques to examine the degree
of their portfolio risk or to assess the probability of failure for new borrowers.
Auditors use failure prediction approaches to evaluate whether or not a firm is on the
verge of failure. Finally, regulators use various techniques to monitor the financial
position of regulated firms.
These are the important aspects of business failure prediction. The next section
discusses the evolution of business failure prediction. The emphasis is on two key
issues: the theoretical framework (Subsection 3.3.1) and methodologies used in
previous research (Subsection 3.3.2).
34
3.3 The evolution of business failure prediction
3.3.1 Theoretical framework overview

Neuman (2006, p. 74) defines a theoretical framework as being:
at the widest range and the opposite extreme from empirical generalizations. It
is more than a formal or substantive theory, and includes many specific formal
and substantive theories that may share basic assumptions and general
concepts in common.
Financial ratio theory in the general sense is discussed first, followed by the
theoretical framework of financial ratios relevant to business failure prediction.
Financial ratios are computed from a pair of items in financial statements. Brigham
and Ehrhardt (2008) state that the four basic financial statements are the balance sheet,
the income statement, the statement of retained earnings and the statement of cash
flow. Financial statements provide essential information about the operating
performance and/or financial position of a company (Altman 1982; Brigham &
Ehrhardt 2008; Fitzpatrick 1931; Gibson & Frishkoff 1986; and Stickney, Brown &
Wahlen 2007). Since there are a large number of items in financial statements, it is
not easy to analyse how a company is performing by looking at these items. In this
respect, ratios are extracted from financial statements to enable analysts to assess the
strengths and weaknesses of a company (Palat 1989).
Theoretically and practically, a great number of financial ratios could be generated

from financial statements. To make interpretation of financial ratios easier and more
precise, it is necessary to group ratios based on the financial attributes being
represented. For example, if an analyst would like to assess whether a company has
enough cash to pay its current obligations, he can quickly look at liquidity ratios.
Another example is that if analysts want to examine whether a company makes profits,
they can look at profitability ratios. Nevertheless, the ratios themselves tell analysts
nothing unless they are benchmarked against something else (Gibson & Frishkoff
35
1986). In this study, the ratios used are selected based on their usefulness in previous
studies and grouped into four classes. They include liquidity ratios,
turnover/performance ratios, leverage/solvency ratios, and profitability ratios. The
details of the ratios, selection procedures, and benchmarking are discussed in the next
chapter. Having already presented an overview of financial ratios theory, the
discussion turns to the theoretical framework of financial ratios in signalling business
failure.
In the context of predicting business failure, the introduction of financial ratios for
detecting symptoms of failed firms can be traced back to the 1930s (for example, the
studies of Fitzpatrick 1931; and Merwin 1942). According to these studies, financial
ratios proved useful in predicting business failure. As a result, subsequent studies
were conducted using financial ratios as an additional tool for signalling business
failure. Beaver (1966, p. 80) explains the theory of financial ratio analysis by
comparing the relationship between cash flow concepts and the ratios as follows:
1) The larger the reservoir (of liquid assets), the smaller the probability of
failure.
2) The larger the net liquidasset flow from operations, the smaller the
probability of failure.
3) The larger the amount of debt held, the greater the probability of failure.
4) The larger the fund expenditures for operations, the greater the probability
of failure.
The propositions are used to investigate the ability of financial ratios to predict failure
in his study. Beaver (1966, p. 102) concludes that ratio analysis can be useful in the
prediction of failure for at least five years before failure.
36
Early studies such as Altman (1968), Beaver (1966) and Tamari (1966) use only
financial ratios to make predictions of business failure while a number of recent
studies rely on a combination set of financial ratios and nonfinancial factors in their
predictions. This indicates a conceptual shift in modelling business failure.
Notwithstanding this, it is evident that financial ratios are predominantly used in the
recent prediction models. It is therefore reasonable to hypothesise that financial ratios
still provide useful information in signalling business failure in recent years.
To substantiate the aforementioned theoretical frameworks related to business failure

prediction, previous studies are reviewed here on the basis of methodology used and
are dealt with in chronological order.
3.3.2 Previous research overview

Early attempts to use financial ratios in signalling corporate failure began in the 1930s
(e.g. the studies of Fitzpatrick 1931; and Merwin 1942). Due mainly to an absence of
sophisticated statistical tools, techniques used in the earlier studies are not
complicated. To demonstrate how earlier researchers conducted their studies
regarding corporate failure, the study of Fitzpatrick (1931) is briefly discussed.
Fitzpatrick (1931) attempted to explore symptoms of corporate failure by

investigating the financial statements of 20 failed firms engaged in manufacturing and
trading industries during the years 19201929 in the United States. The technique
used in this study is a comparison of ratios and trend percentages of the failed firms.
Note that the trend percentage method used in this study focuses on some particular
items in financial statements such as receivables, inventories, total debt, net worth,
and the like. A total of 13 ratios were selected based upon frequent usage by the
leading analysts at that time. The findings suggest that the best ratios for predicting
failure from the last annual statement prior to business failure are net profit to net
worth, net worth to fixed assets, net worth to debt, and the quick ratio. Fitzpatrick
(1931) acknowledges that ratios and trend percentages are not the only techniques
37
which can be used for predicting corporate failure but he suggests that they would be
useful tools to assist stakeholders to identify symptoms of corporate failure.
In the 1960s, the seminal studies promoting predictive techniques and/or models for
business failure based on financial ratios were conducted by Beaver (1966) and
Altman (1968) (see Bei and Liu (2005); Chi and Tang (2006); and Pongsatat, Ramage
and Lawrence (2004)). Furthermore, the 1960s is a critical period in which there was
a methodological change from univariate methods (considering only one financial
ratio at a time when investigating an incidence of business failure) to multivariate
methods (considering a group of financial ratios when investigating an incidence of
business failure) (Dambolena & Khoury 1980; He & Kamath 2006; Pongsatat,
Ramage & Lawrence 2004; Tirapat & Nittayagasetwat 1999; Ugurlu & Aksoy 2006;
and Zavgren 1983).
After its emergence, the multivariate approach became widely used by numerous
researchers who attempted to develop an efficient business failure prediction model
(for instance, the studies of Altman 1968; Bei & Liu 2005; Blum 1974; Chi & Tang
2006; Coats & Fant 1993; Dambolena & Khoury 1980; Deakin 1972; Dimitras et al.
1999; Pongsatat, Ramage & Lawrence 2004; and Tirapat & Nittayagasetwat 1999).
Thanks to the innovation of sophisticated statistical tools, there are numerous studies
which have used the multivariate approach for developing business failure prediction
models. The prominent analytical tools of this approach are Multiple Discriminant
Analysis (for example, earlier significant works of Altman 1968; Blum 1974; Deakin
1972; and Edmister 1972), Logit Analysis (Gentry, Newbold & Whitford 1985; and
Ohlson 1980), and Neural Network Analysis (Coats & Fant 1993).
This section briefly reviews the studies employing either different approaches
(univariate or multivariate approach) or different statistical techniques (e.g. Multiple
Discriminant Analysis, Logit, and the like). International studies are reviewed in
Section 3.3.2.1 whilst Thai studies are presented in Section 3.3.2.2.
38
3.3.2.1 International studies
In an early period, Beaver (1966) applied univariate analysis in his study to
investigate the predictive ability of financial ratios rather than to find the best
predictor of failure (Beaver 1966, p. 100). The statistical tools used in his
study include profile analysis, dichotomous classification tests, and likelihood
ratios analysis. Applying pairedsample design criteria (for firms in the same
industry and of similar asset size), a sample of 79 failed firms and 79 non
failed firms was derived from Moodys Industrial Manual for the period from
1954 to 1964. Noncorporate firms, privately held firms, and nonindustrial
firms (e.g. public utilities, transportation companies, and financial
institutions) were not included in the sample (Beaver 1966, p. 72). A total of
30 ratios were selected based on their use in previous studies.
The findings reveal that there were significant differences in the ratios
between failed and nonfailed firms over the observation period. For instance,
failed firms had lower cash flow to total debt ratios and net income to total
assets ratios than nonfailed firms. Finally, the study suggests that ratio
analysis for at least five years before failure takes place is useful for signalling
business failure.
Two years later, Beaver (1968a) conducted a study evaluating alternative

accounting measures as predictors of failure. A replication on a population of
79 failed and 79 nonfailed firms was employed and a total of 14 ratios were
used. The findings indicate that the nonliquid assets variables are superior to
liquid assets variables for predicting corporate failure.
Notwithstanding this, it should be noted that ratio analysis using univariate

approaches might be misleadingly interpreted and has potentially dubious
classification accuracy. The difficulty in interpretation occurs when financial
39
ratios provide conflicting predictions for a firm (Dimitras, Zanakis &
Zopounidis 1996). For example:
a firm with a poor profitability and/or solvency record may be regarded

as a potential bankrupt. However, because of its above average
liquidity, the situation may not be considered serious (Altman 1968, p.
591).
To address these shortcomings, multivariate approaches were developed and

introduced into the field. Multiple Discriminant Analysis (MDA) is a
statistical technique used to classify an observation into one of several a priori
groupings dependent upon the observations individual characteristics
(Altman 1968, p. 591). It enables researchers to find the optimal set of
financial ratios that best distinguishes the observation.
The MDA technique is selected as the most appropriate one for business
failure study by both academics and financial practitioners (Frydman, Altman
& Kao 1985). Moreover, it is frequently used as a benchmark for comparison
(for example, in the studies of Coats & Fant 1993; Frydman, Altman & Kao
1985; Gentry, Newbold & Whitford 1985; Hing & Lau 1987; Pongsatat,
Ramage & Lawrence 2004; Sung, Chang & Lee 1999; and Ugurlu & Aksoy
2006). Therefore, the studies utilising the MDA technique are discussed below.
Altman (1968) is the pioneer who introduced multivariate approaches into the
field, resulting in a methodological change in ratiobased modelling of
business failure (Balcaen & Ooghe 2006; and Hossari 2005). In 1968, Altman
developed the Z score model (a five variables model). Samples of 33 bankrupt
and 33 nonbankrupt firms in the US during the period 19461965 were
compared. The firms were in the same industry (manufacturing) and were of
similar assets size. A total of 22 variables (financial ratios) were selected on
40
the basis of frequent use in previous studies as the best indicators of predicting
corporate failure. The Z score model (a five variables model) is as follows:
Z 1.2 X 1 1.4 X 2 3.3 X 3 0.6 X 4 1.0 X 5 ......(3.1)
Where:
X 1 =Working Capital/Total Assets
X 2 =Retained Earnings/Total Assets
X 3 =Earnings before Interest and Taxes/Total Assets
X 4 =Market Value of Equity/Book Value of Total Liabilities
X 5 =Sales/Total Assets
Z =Overall score
Zones of discrimination:
Z > 2.99: Safe zone
1.81 < Z <2.99: Zone of ignorance or Grey zone
Z <1.81: Distress zone
Source: Altman (1968, p. 594 and 606)
Altmans (1968) findings are very encouraging with 95 per cent predictive
accuracy of the total sample for one year prior to bankruptcy. Later, the model
was modified to use for private firms and for nonmanufacturing firms. To
apply the model to nonlisted firms, Altman (1993) revised a numerator of the
fourth variable ( X 4 ) to book value of equity from market value of equity. This
results in a slight change to all of the weighting factors and zones of
discrimination as follows:
41
Z ' 0.717 X 1 0.847 X 2 3.107 X 3 0.420 X 4 0.998 X 5 ...... (3.2)
Where:
X 4 =Book Value of Equity/Total Equity
X 5 =Sales/Total Assets
Z ' =Overall score
Z ' > 2.90: Safe zone
1.23 < Z ' <2.90: Grey zone
Z ' <1.23: Distress zone
Source: Altman (1993, pp. 203-4)
However, the latest modified version is called the Emerging Market Score
(EMS) model (Altman & Hotchkiss 2006), developed for firms operating in
developing countries. The EMS model was developed in the mid1990s to
achieve improved accuracy when predicting the probability of default of non
manufacturing firms and corporations in emerging markets. The EMS model
includes only four variables along with a constant value as follows:
EMS 3.25 6.56 X 1 3.26 X 2 6.72 X 3 1.05 X 4 ......(3.3)
42
Where:
X 4 =Book Value of Equity/Total Liabilities
EMS =Overall score
EMS > 5.85 : Safe zone
4.15 < EMS < 5.85 : Grey zone
EMS <4.15 : Distress zone
Source: Altman & Hotchkiss (2006, pp. 267-8)
The EMS model was applied to test both the Enron and WorldCom cases and
the scores showed there were warning signs before the bad news was exposed.
Enrons EMS score was below the safe zone in June 2001 before filing for
bankruptcy under Chapter 11 of the United States Bankruptcy Code on
December 2001 whilst WorldComs EMS score was below the safe zone at the
end of the first quarter of 2002 before filing for bankruptcy under Chapter 11
of the United States Bankruptcy Code in midJuly 2002 (Altman & Hotchkiss
2006). This demonstrates an impressive result for the EMS model when
applied to nonmanufacturing firms. In addition, it is evident that the models
application in emerging markets e.g. Mexico, Brazil and Argentina yields
successful results in predicting corporate collapse (Altman & Hotchkiss 2006).
43
Deakin (1972) employs the same 14 ratios that Beaver (1968a) used in his
study. A sample of 32 failed US firms matched with 32 nonfailed US firms
between 1962 and 1966 were selected in this study. The matching is based on
three criteria: industry classification, asset size, and year of financial data
available. The 14 ratios were selected to represent a set of predictors in this
study. The results suggest that the overall predictive accuracy of the model is
97 per cent when using the last financial statement prior to failure.
Edmister (1972) conducted his study focusing on financial ratio modelling for
small business firms failure. The sample was derived from the data of the
Small Business Administration and Robert Morris Associates between 1954
and 1969. A sample of 42 US firms were selected and a total of seven
financial ratios were chosen. Edmisters (1972) findings indicate that the
sevenratio discriminant function correctly classifies 39 out of 42 firms. That
is, the overall predictive power of the model is 93 per cent.
In Blums (1974) study, a matched sample of 115 US failed and 115 US non
failed firms during the years 19541968 is used in his study. The matching
criteria are industry, sales, number of employees, and fiscal year. Blums
(1974) discriminant function constitutes 12 variables. The results suggest that
the overall predictive accuracy of the model stands at approximately 94 per
cent using oneyear statements prior to failure.
Bei and Liu (2005) used paired sample design with 62 Chinese listed firms
consisting of 31 failed and 31 nonfailed firms in the development of a
discriminant function. A total of 30 financial ratios were selected based on
their use in previous studies. To search for an optimal set of financial ratios, a
forward stepwise approach was used in this study and the final model
contained 3 ratios. The overall classification accuracy of the model is up to 79
per cent.
44
Hossari (2007) conducted his study using a multiperiod (or dynamic)
business failure model instead of a oneperiod (or static) model. The
comparative accuracy of multiperiod model and oneperiod model is
examined in this study. Applying paired sample design, dynamic ratiobased
models were developed based on a total sample of 494 US publicly listed
firms (247 failed and 247 nonfailed firms) during the period 1986 to 2004. In
this study, 9 ratios were retained in the final models.
The findings confirm that the heuristic model generates higher predictive
ability than the static model during the last two years prior to failure. For
example, the overall classification accuracy of the dynamic model one year
before failure was 71.3 per cent. In comparison, the overall classification
accuracy of the static model one year before failure was 63.4 per cent.
Although MDA was used in the majority of the earlier studies (for instance, in
the studies of Altman 1968; Blum 1974; Deakin 1972; Edmister 1972; and
Norton & Smith 1979), the validity of its results was criticised because of its
statistical assumptions such as a normal distribution of ratios and equality of
variancecovariance matrices of ratios for both groups (Balcaen & Ooghe
2006; Coats & Fant 1993; Dimitras et al. 1999; Frank, Massy & Morrison
1965; Goudie 1987; Hamer 1983; Ohlson 1980; and Taffler 1982). The
restrictive assumptions on the application of discriminant analysis led
researchers to develop other multivariate statistical tools to overcome these
limitations. After the 1980s, MDA began to lose its popularity with
researchers and was replaced by less demanding statistical techniques. One of
the statistical techniques that requires fewer restrictive assumptions is logit
analysis (Balcaen & Ooghe 2006).
Ohlson (1980) applied a probabilistic estimate of failure to his study and

developed the socalled logit model from 105 US bankrupt firms and 2,058
45
US nonbankrupt firms during the period 19701976. Ohlsons (1980) model
was comprised of nine explanatory variables. The findings suggest that the
logitbased model provided classification accuracy of 96 per cent one year
prior to failure.
Gentry, Newbold and Whitford (1985) employed cashbased funds flow ratios
as indicators for discriminating between failed and nonfailed firms as an
alternative to financial ratios. Their final logitbased model contains eight net
funds flow variables. A total of 33 failed US firms were matched with 33 US
nonfailed firms based on asset size and sales during the period 19701981.
The overall predictive power of the logitbased model using cashbased funds
flow ratios was 83 per cent one year before failure.
Ugurlu and Aksoy (2006) used both a logit model and an MDA model to
investigate corporate failure and to demonstrate the comparative
discrimination ability of the models in an emerging market, Turkey. A sample
of 27 failed and 27 nonfailed firms listed on the Istanbul Stock Exchange
during 1996 to 2003 was selected. Financial data for 1996 to 2003 were
collected to develop explanatory variables. 11 variables were included in the
final logitbased model and 10 variables were included in the final MDA
based model. The findings suggest that the logitbased model outperformed
the MDAbased model. For example, the overall predictive power of the
former was 95.6 per cent in contrast to 85.9 per cent of the latter. The
difference of approximately 10 per cent was considered statistically
significant.
Chi and Tang (2006) examined publicly listed firms in export industries in
seven AsiaPacific nations; Hong Kong, Japan, Korea, Malaysia, Singapore,
Thailand, and the Philippines. During the years 20012003, 60 bankrupt
exportoriented firms were selected and matched with 180 nonbankrupt
46
exportoriented firms based on the same asset size and industry. The final
logitbased models were comprised of 10 variables. The findings suggest that
the overall classification accuracy of the models ranges between 79 per cent
and 86 per cent.
He and Kamaths (2006) study examines the efficacy of Ohlsons (1980)

model and Shumways (2001) model and investigates the comparative
classification accuracy of the two models when applied in the US retail
industry. In the period 19901999, 354 OverTheCounter (OTC) mixed
industry US firms (177 failed and 177 nonfailed) were selected as an overall
sample group. Of the 354 OTC mixed industry firms, only 40 firms were
identified as firms in the retail industry.
The OTC market is defined as:
A securities market that is conducted by dealers throughout the country

through negotiation of price over the phones rather than through the
use of an auction system as represented by a stock exchange (SET n.d.,
p. 8).
The findings indicate that Shumways (2001) model outperformed Ohlsons

(1980) model in year one before failure. The overall classification accuracy of
Shumways (2001) model for retail firms in the first year before failure was 88
per cent. In comparison, the overall classification accuracy of Ohlsons (1980)
model for retail firms in the first year before failure was 80 per cent.
Like Logit Analysis, Neural Network Analysis (NNA) has been developed as
an alternative tool to overcome MDAs restrictive requirements.
47
Coats and Fant (1993) developed NNAbased models and matched 94 failed
firms with 188 nonfailed firms covering the period 19701989 from the
Standard & Poors COMPUSTAT database. The models consisted of a set of
Altmans (1968) five ratios. Furthermore, MDAbased models were built as
the benchmark to compare classification accuracy between NNAbased
models and MDAbased models. The findings suggest that the NNAbased
models generate a higher overall prediction accuracy than MDAbased
models. For instance, the overall classification accuracy of the NNAbased
model one year before failure is 95 per cent compared to 88 per cent for the
MDAbased model.
To compare and contrast how the international studies were conducted, their
major features are illustrated in Table 3.1.
48
Table 3.1 Comparison of key features of the international studies
Researcher(s) Features
(year of publication) Method Number Study Matching criteria Overall
used of firms periods classification
(F/NF) accuracy
Beaver (1966) Univariate 79/79 19541964 Industry and assets n.a.

size
Beaver (1968a) Univariate 79/79 19541964 Industry and asset size n.a.
Altman (1968) MDA 33/33 19461968 Industry and asset size 95%
Deakin (1972) MDA 32/32 19621966 Industry, assets size, 97%
and year of financial
available
Edmister (1972) MDA 42 19541969 Small businesses 93%
Blum (1974) MDA 115/115 19541968 Industry, sales, 94%
number of employees,
and fiscal year
Bei and Liu (2005) MDA 31/31 20022003 79%
Hossari (2007) MDA 247/247 19862004 Industry and asset size 71%
Ohlson (1980) Logit 105/2,058 19701976 no 96%
Gentry, Newbold and Logit 33/33 19701981 Asset size and sales 83%
Whitford (1985)
Ugurlu and Aksoy Logit and 27/27 19962003 Industry and asset size 96%;86%
(2006) MDA
Chi and Tang (2006) Logit 60/180 20012003 Industry and asset size 86%
He and Kamath Logit 20/20 19901999 Industry, asset size, 88%;80%
(2006) and fiscal year
Coats and Fant (1993) NNA 94/188 19701989 No 95%
Note:MDA=Multiple Discriminant Analysis; NNA= Neural Network Analysis; F/NF=
Failed/NonFailed.
There have been numerous studies using financial ratios to predict corporate
failure in many countries. Researchers in each country, particularly in
emerging countries like Thailand, have attempted to establish an efficient
prediction model suitable for their own country instead of applying the US
based model. On the evidence of these studies, multivariate analysis is widely
and consistently employed to develop models to distinguish between
financially unhealthy firms and financially healthy firms. Like the
international studies, the Thai studies consider the same range of variables (i.e.
liquidity, turnover, leverage, profitability and cash flow) to construct the
discriminant function.
49
3.3.2.2 Thai studies
Tirapat and Nittayagasetwats study (1999)
In Tirapat and Nittayagasetwats (1999) study, macroeconomic factors and

financial ratios are included in the prediction model. The study was conducted
in Thailand following the emergence of Thailands 1997 financial crisis. The
logitbased model is based on a sample of 55 failed and 341 nonfailed firms
listed on the Stock Exchange of Thailand. The findings suggest that the overall
classification accuracy is between 64 per cent and 76 per cent at the optimal
cutoff point of 0.1 generating less Type I errors than Type II errors.
Persons study (1999)
Persons (1999) study focuses on both financial statement and nonfinancial

statement modelling for Thai finance firms failure. The sample was made up
of 41 finance companies (26 defunct and 15 surviving) listed on the Stock
Exchange of Thailand. Their financial statements between 1993 and 1996
were used to construct the logitbased model and only four financial ratios
were retained in the final model. The findings indicate that the degree of
classification accuracy of the model is encouraging. For instance, 96 per cent
of defunct finance firms and 87 per cent of surviving finance firms were
correctly classified in the first year preceding failure.
Reynolds et al.s study (2002)
Like Persons (1999) study, this study examines the probability of business
failure of Thai financial firms. A total sample of 91 financial firms consisting
of 56 failed and 35 nonfailed companies was selected. Financial and non
financial data for 1993 to 1996 obtained from the Bank of Thailand were
collected to develop probabilistic models. They included a probit model
(Model 1), a logit model (Model 2) and a cumulative logit model (Model 3).
The models were made up of six explanatory variables. The findings suggest
50
that the overall prediction ability of the three models is somewhat similar. In
other words, no model is significantly more accurate than its counterparts. The
average overall prediction ability of the models was 68 per cent.
Pongsatat, Ramage and Lawrences study (2004)
To investigate whether Ohlsons (1980) logitbased model and Altmans

(1968) MDAbased model can be applied to Thai firms, Pongsatat, Ramage
and Lawrence (2004) conducted their study based upon 60 failed and 60 non
failed firms listed on the Stock Exchange of Thailand. Like other studies,
paired sample design was employed in this study using firms in the same
industry that were of similar asset size. Data collection (financial statements
for 1998 to 2003) was derived from the elibrary of the Stock Exchange of
Thailand. The findings reveal that there is no significant difference in
predictive power between Ohlsons (1980) model and Altmans (1968) model.
The overall classification accuracy of the two models was between 59 per cent
and 75 per cent.
Puagwatana and Gunawardanas study (2005)
Puagwatana and Gunawardanas (2005) study investigates the probability of

business failure of limited companies in the technology industry in Thailand.
A total sample of 24 nonlisted technology firms consisting of 12 failed and
12 nonfailed companies was selected. Financial data from the Department of
Business Development of the Ministry of Commerce were collected to
develop a binomial logit model. The final model was developed based on
Altmans (1968) model and one new ratio (net income (loss) to amount of
shares) was added. The findings indicate that the overall discrimination
accuracy of the model was 77.8 per cent.
51
Sookhanaphibarn et al.s study (2007)
This study focuses on financial firms listed on the Stock Exchange of Thailand
and their financial statements between 1993 and 1996. A total of 41 financial
firms listed on the Stock Exchange of Thailand during 1993 to 2003 were
selected. This study selected 30 financial variables and seven ownership
variables (for example, family and control rights, and the like) to develop the
NNAbased model. Three NNA approaches (i.e. Learning Vector
Quantisation, Probabilistic Neural Network, and Feedforward Network with
Backpropagation) were used to examine the classification ability of the
models. The findings indicate that, in all three approaches, the optimal set of
inputs is between five and nine variables. The overall classification accuracy
of the variables set ranges between 100 per cent and 89 per cent.
Again, to compare and contrast how the Thai studies were conducted, their
major features are illustrated in Table 3.2.
Table 3.2 Comparison of key features of the Thai studies

Researcher(s) Features
(year of publication) Method Number of Study Matching Overall
used firms periods criteria classification
(F/NF) accuracy
Tirapat and Logit 55/341 No 64%76%
Nittayagasetwat (1999)
Persons (1999) Logit 26/15 19931996 Industry 96%
Reynolds et al. (2002) Logit 56/35 19931996 Industry 68%
Pongsatat, Ramage and Logit and 60/60 19982003 Asset size 59%75%
Lawrence (2004) MDA
Puagwatana and Logit 12/12 Industry, 78%
Gunawardana (2005) location, group
size, and
financial year.
Sookhanaphibarn et al. NNA 26/15 19931996 Industry 89%100%
(2007)
Note: MDA=Multiple Discriminant Analysis; NNA= Neural Network Analysis; F/NF=
Failed/NonFailed.
52
3.4 Limitations of the previous research
It is evident over several decades that most researchers in this field have attempted to
construct an efficient predictive model for business failure. Both new explanatory variables
and new statistical techniques have been used to search for the best predictors for failure.
Only the studies of Beaver (1966), Beaver , McNichols and Rhie (2005) and Fitzpatrick
(1931) have addressed the ongoing ability of financial statement data to predict business
failure as the objective of the research. Notwithstanding this, these studies were conducted in
the western environmental setting. In this regard, there are a number of explicit limitations in
the previous researchwhich have to be addressed:
1) Most wellknown studies in this field are conducted in western countries like the US
and, notwithstanding the studies reviewed above in this thesis, little empirical
research has been conducted in emerging markets like Thailand.
2) What is scarce in the literature on the topic is the application of quarterly financial
statement information. Numerous international and Thai studies have used annual
financial statement data to investigate the occurrence of business failure. Coats and
Fant (1993) suggest that using quarterly financial statement data together with several
time periods to observe the characteristics of a firm is likely to offer potential for
accurate failure prediction. This is consistent with the notion of Dimitras, Zanakis and
Zopounidis (1996) that investigation of the ratios value should be made over time to
monitor the progress of a firm since the event of failure is dynamic.
3) To date, most empirical work in Thailand has been focused on studying business
failure in periods of economic crisis. Relatively little research has been conducted
under normal economic circumstances. According to the recent investigations of Thai
business failure reviewed in Chapter 2, the number of Thai company dissolutions
remained problematic over the period 2000 to 2009. This confirms that business
failure occurs at any time, not only in periods of economic crisis.
4) The major stream of research investigating business failures in Thailand has aimed to
find the best predictors (a combination of financial ratio and nonfinancial factors) of
failure by estimating a discriminant model and comparing the predictive accuracy of
the model. There has not been evidence in the Thai literature that financial statement
data are adequately informative to predict Thai business failure after the post 1997
financial crisis. Recently, Beaver, McNichols and Rhie (2005) conducted their study
53
to examine whether financial statements have become less informative for failure
prediction but this study was conducted in a western setting. In this regard, this study
attempts to explore whether financial statement data can be used to adequately signal
business failure in Thailand.
5) To date, relatively little empirical research has been focused on investigating the
ability of financial statement information to identify failed and nonfailed companies
in Thailand. The results of the present study should have more relevance to Thai
decisionmakers and could be used as a foundation for future research into failure
prediction in the Thai environment.
To overcome the aforementioned limitations, this study employs quarterly financial statement
data to investigate the ability of financial statement information to predict business failure in
Thailand under normal economic circumstances. Simply stated, this study makes a
preliminary attempt at empirically investigating the predictive ability of financial ratios under
normal economic circumstances in Thailand. Furthermore, the three analytical tools (the
Emerging Market Score model, comparative ratio analysis, and ratio trend analysis) used in
this study have not previously been collaboratively applied in the Thai environment.
Consequently, the findings of this research could be used to help fill the knowledge gap and
broaden the body of knowledge in this area.
Even though there is no unique set of financial ratios used in business failure prediction, the
financial ratios selected by numerous researchers have a lot in common. In the literature, the
most frequently used criteria for selecting financial ratios are the usage of the ratios (in both
practice and prior research) and their predictive ability. The same criteria are applied in this
study. The selected financial ratios are classified into four meaningful groups: liquidity,
turnover/performance, leverage/solvency, and profitability (Blum 1974; Brigham & Ehrhardt
2008; Ohlson 1980; and Tirapat & Nittayagasetwat 1999). Finally, concerning sample
selection, the Stock Exchange of Thailands financial database will be utilised as the primary
data source of this study. More discussion concerning these issues will be presented in
Chapter 4.
54
3.5 Development of the analytical approach
Based on the literature review, the development of the research problem, research question,
and research objective were presented in the previous section. This section discusses how the
analytical approach of this study was developed so as to search for solutions to the research
issues.
In this research, the sample is divided into two groups (failed and nonfailed firms). A failed
firm is paired with its nonfailed mate based on being in a similar industry, being closest in
terms of asset size, and having similar financial statement date. The firms characteristics, or
financial ratios, will be identified based on four financial performance areas. They include
liquidity, turnover/performance, leverage/solvency, and profitability. The ratios used as
proxies for the four financial areas are selected on the basis of their usage and their predictive
ability in the previous studies. The predictive ability of financial statement information will
be examined by using three ratiobased measures. They are the Emerging Market Score
model, comparative ratio analysis, and ratio trend analysis. The first measure is a
sophisticated ratiobased model while the last two measures are traditional financial
statement analytical tools. In addition, the logit-based model is employed as a benchmarking
measure. The predictive results will be compared and contrasted in order to answer the
research question and problem. The analytical approach applied in this study is depicted in
Figure 3.2.
55
Figure 3.2 Analytical approach of this study
Financial ratios Firm status

Liquidity Failed
Turnover Non-failed
Leverage/solvency
Profitability
Examine the predictive ability of financial ratios of the three
ratio-based measures
Develop an Integrated Multi-Measure approach and examine its

result
Apply logit-based model as a benchmarking measure and

examine its result
Conclusion
As assessment of the predictive ability of financial statement
information in the context of Thailand setting
Source: Developed for this thesis.
3.6 Conclusion
As mentioned in the beginning of the chapter, its main objectives were to review extant
studies that use financial statement data and/or nonfinancial factors as analytical tools for
signalling business failure, to describe the important aspects of failure prediction, and to
present the evolution of predictive models based on the methods and concepts used. In
addition, the chapter sought to review studies of business failure prediction, and identify
research gaps.
This chapter began by discussing the key terms widely used in business failure prediction,
which are failure, insolvency, default, and bankruptcy. Then the costs of business failure
56
were presented. This led to a discussion of the importance of business failure study. This
section reviewed both international and Thai studies of business failure prediction, classifying
them according to the methodology employed. According to the findings of these studies,
there is reason to believe that financial ratios can be effective for signalling the likelihood of
corporate failure. It is apparent that different statistical techniques such as MDA, logit
analysis and NNA provide similar levels of classificatory accuracy. From the previous
literature, it can be concluded that there is no consensus among researchers that any specific
technique is dominant over its counterparts.
Examination of the literature has provided a basis for examination of the research objective of
this thesis. Both comparative ratios analysis and trend analysis (i.e. trend percentage method
and trend breaks method) have been utilised as analytical tools to signal financial distress in
previous studies (for example. in the studies of Beaver 1966; Blum 1974; and Fitzpatrick
1931). Altman (1968)s Z Score model and its further developments including the EMS
model have formed the predominant basis of more sophisticated approaches to developing
measures for signalling financial distress and have been the preferred benchmark for
researchers attempting to develop improved prediction approaches. The resiliency of these
three approaches over time and their widespread acceptance lead to the selection of the three
approaches in this study.
Having already presented the research gaps and the research issues, the discussion now turns
to methodology used in this study. The rationales and details of the methodology used will be
provided in Chapter 4.
57
Chapter 4: Research Methodology
4.1 Introduction
Chapter 3 reviewed literature that investigates the use of either financial statement data or a
combination of financial statement data and nonfinancial factors as potential predictors for
business failure. As can be seen, irrespective of the statistical methodology used, studies both
in western countries and in emerging markets indicate that financial ratios can be used to
signal business failure. At the end of Chapter 3, after the research problem and other research
issues were identified, an analytical approach for an assessment of the ability of financial
statement data to signal business failure was developed.
This chapter describes the research methodology and design used for collecting and analysing
data in satisfying the research objective. The primary objectives of this chapter are to
describe and justify the research methodology and the research design used in this study, and
to illustrate how data were collected and analysed.
The chapter comprises eight sections as shown in Figure 4.1. Section 4.1 provides an
introduction to the research methodology along with the objectives and structure of the
chapter. After a discussion on research problem identification and formulation in Section 4.2,
the research design of this study is elaborated in Section 4.3. Then a detailed description of
sampling design is provided in Section 4.4. Section 4.5 discusses data collection. The
analytical approach is described in Section 4.6 and the development of the Integrated Multi
Measure approach is described in Section 4.7. Section 4.8 discusses the logit-based model as
a benchmarking measure. Finally, the conclusion to this chapter is drawn in Section 4.9.
58
4.1 Introduction
4.2 Research problem identification and formulation
4.3 Research Design
4.3.1 Dimensions of research 4.3.3 Research methods
4.3.2 Typology of research
4.4 Sampling design
4.4.1 Specification of population 4.4.2 Sample selection
4.5 Data collection
4.6 Analytical approach
4.6.1 Emerging Market Score 4.6.3 Ratio trend analysis

(EMS) model
4.6.2 Comparative ratio analysis
4.7 Development of the Integrated Multi-Measure (IMM)

approach
4.8 Application of logit-based models
4.9 Conclusion
59
4.2 Research problem identification and formulation
Researchers in two schools of thought, social and business research, share a common view
that identifying the research problem plays a crucial role before the research investigation is
carried out (Davis 2005; and Zikmund 2003). Davis (2005, p. 100 and 4) notes that problem
identification is the start of the research process whilst problem formulation sets the direction
and scope of the research endeavours that follow. Likewise, Zikmund (2003, p. 8) states that
the first and foremost stage in business research is identifying problems or opportunities.
It is worth noting that a research problem, as it is termed, does not always mean something
has gone wrong: A research problem could simply indicate an interest in an issue and a sense
that finding the right answers might help to improve an existing situation (Cavana, Delahaye
& Sekaran 2001, p. 62). Thus, a restatement of the research issues, as discussed in Section 3.5,
is made. Figure 4.2 depicts the correlation between the research issues and the research
methodology of this study.
Figure 4.2 Correlation between research issues and research methodology
Research problem Whether financial statement information can be

adequately used to predict financial distress for Thai
firms in normal economic circumstances
Research question Can financial statement information be used to

discriminate between potentially failing and non-failing
firms in the context of Thailand in normal economic
circumstances?
Research Research Data Conclusion

methodology design and collection and reports
sampling and analysis
design techniques
60
4.3 Research design
After the research problem is thoroughly identified and formulated, the research design needs
to be developed. Research design is a master plan specifying the methods and procedures for
collecting and analysing the needed information (Zikmund 2003, p. 65). This is consistent
with Davis (2005, p. 134) notion that research design is considered as the road map for
researchers. Simply stated, research design equips a researcher with appropriate means and
methods for solving the research problem (Davis 2005, p. 135). Likewise, Punch (2006, p.
47) states that research design helps researchers connect the research questions to data. This
view is shared by Denzin and Lincoln (2008). This section, therefore, elaborates on the
dimensions of the research (Subsection 4.3.1), the types of research (Subsection 4.3.2), and
research methods (Subsection 4.3.3). Finally, the focus of this study is presented.
4.3.1 Dimensions of research

In the history of western philosophy, there are extensive schools of thought (for
instance, positivism, antipositivism, rationalism) that try to describe how people
come to know what they know (McMurray 2007, p. 11). These notions have been
extensively adopted by researchers. Nevertheless, Neuman (2006, p. 24) classifies
researchers into two groups: some who use research to advance general knowledge
are engaged in basic research, whereas others who use research to solve specific
problems are engaged in applied research. This view is shared by Cavana, Delahaye
and Sekaran (2001), Davis (2005), and Zikmund (2003).
Applied research aims to answer questions about specific problems or to make

decisions about a particular course of action or policy decision (Zikmund 2003, p. 7).
The view is shared by Cavana, Delahaye and Sekaran (2001, p. 12) who argue that
research done with the intention of applying the results to solving specific problems
currently being experienced in the business is called applied research. As indicated
by the research problem above, this study is designed to address the problem of
business failure prediction. Stated clearly, this study aims to examine whether
financial statement data can be adequately used to predict business failure for Thai
61
companies in recent years. Therefore, applied research is more suitable for this study.
To provide a more complete picture of the choice of research method, types of
research are further discussed.
4.3.2 Typology of research

It is generally accepted by researchers that there are three types of research based on
its purpose or goal (Davis 2005; Neuman 2006; and Zikmund 2003). It should be
noted that one research type can have different names. For example, Neuman (2006, p.
35) and McMurray (2007, p. 21) use the term explanatory research whilst Davis
(2005, p. 145) and Zikmund (2003, p. 56) refer to the same research type as causal
research. Cavana, Delahaye and Sekaran (2001, p. 111) call this type of research
hypothesis testing. According to Neuman (2006, pp. 33-5), the three types of
research are exploratory research, descriptive research and explanatory research.
Exploratory research is conducted when little is known in the specific area of interest
(McMurray 2007). This is consistent with the notion of Cavana, Delahaye and
Sekaran (2001, p. 108) that exploratory studies are undertaken to better comprehend
the nature of the problem that has been the subject of very few studies. Rudimentary
knowledge and understanding about a phenomenon is offered and might be expanded
by subsequent researchers. In other words, exploratory research may be the first stage
in a sequence of studies (Neuman 2006, p. 33).
Descriptive research, as the term implies, aims to provide descriptive information (for
instance, age, gender, social status and the like) of a problem situation by supplying
numerical data (Cavana, Delahaye & Sekaran 2001; Punch 2006; and Zikmund 2003)
while explanatory research aims to identify the causes of the phenomenon being
studied (Punch 2006; and Zikmund 2003).
62
To date, no study in Thailand has been conducted to investigate the financial distress
predication ability of financial statements using quarterly financial data. Furthermore,
the three ratiobased measures have never been used together in the context of recent
company financial distress in Thailand. Finally, this study employed the logit-based
model as a benchmarking measure. The results provided by this study could be used
as a foundation for future research. Therefore, exploratory research is the most
appropriate descriptor for this study.
As previously noted, the identification and formulation of the research problem have
an influence on research design and another component of research design is how data
are collected. Subsequently, it is necessary to discuss selecting the pertinent research
methods.
4.3.3 Research methods

According to Zikmund (2003, p. 65), research methods can be divided into four types:
surveys, experiments, secondary data studies, and observation. Each method has its
own strengths and weaknesses (Neuman 2006). A brief description of the methods
needs to be presented to help justify the selection of the appropriate research methods
for this study.
Surveys are a commonly and widely used technique for gathering information from
primary data sources by means of questionnaires and/or interviews (Davis 2005;
Neuman 2006; and Zikmund 2003). Experiments are another systematic method of
gathering data. They can be laboratory experiments or field experiments (Cavana,
Delahaye & Sekaran 2001; Davis 2005; Neuman 2006; and Zikmund 2003).
Secondary or historical data are data originally gathered by previous studies for some
purpose whilst primary data are data gathered by current studies for a specific purpose
(Cavana, Delahaye & Sekaran 2001; Davis 2005; Neuman 2006; and Zikmund 2003).
Finally, observation is a systematic attempt to collect and assemble data. Conducting
63
an observation study, a researcher has to decide to be either a participantobserver or
a nonparticipantobserver (Cavana, Delahaye & Sekaran 2001).
As can be seen in Chapter 3, financial ratios are one of the critical components of the
study in this field. Financial ratios are defined as a quotient of two numbers, where
both numbers consist of financial statement items (Beaver 1966, pp. 71-2). A large
number of financial ratios are derived from financial statement data in the balance
sheet, and the income statement. Like previous studies, this research uses historical
financial statement information and examines its ability to predict failure in the Thai
environment. Therefore, secondary data study is the most applicable type of data for
this study.
Since the target population is public companies listed on the Stock Exchange of
Thailand, the data required are readily accessible. Figure 4.3 provides a comparison of
the major components of research design and the focus of this study.
Figure 4.3 Comparison of the major components of research design and the focus of this
study
The focus of this study
Dimensions of Basic research Applied research

research
Typology of Exploratory Descriptive Explanatory

research research research research
Surveys Secondary
Research methods Experiments
data Observation
Studies
64
4.4 Sampling design
All of the powerful research techniques discussed above might be useless in producing the
anticipated results if a researcher does not specify an appropriate population (Black 1999; and
Cavana, Delahaye & Sekaran 2001). All research inevitably employs sampling as researchers
cannot study everyone everywhere doing everything (Miles & Huberman 1994, p. 27). The
salient objective of sampling, according to Zikmund (2003, p. 369), is to make conclusions
about the whole population by using a small number of items or parts of a larger population.
This view is echoed by Davis (2005) and Neuman (2006). It is, therefore, imperative to
discuss the specification of the population (Subsection 4.4.1) and sample selection (Sub
section 4.4.2) of this study.
4.4.1 Specification of population

When the term population is used in the context of sampling, it refers to any group of
units of interest that have something in common and are examined by researchers
(Black 1999; Cavana, Delahaye & Sekaran 2001; Davis 2005; and Zikmund 2003).
All business failure prediction research requires a definition of failure, specification of

the population (e.g. industrial or nonindustrial companies), sample selection and data
collection (e.g. the studies of Altman 1968; Beaver 1966; Beaver, McNichols & Rhie
2005; Bei & Liu 2005; Blum 1974; Chi & Tang 2006; Jones & Hensher 2004;
Sookhanaphibarn et al. 2007; Tirapat & Nittayagasetwat 1999; and Ugurlu & Aksoy
2006). This sequence is considered the traditional way to structure empirical studies
in this field.
As mentioned in Subsection 3.2.1, a legal definition of business failure is used in

most business failure prediction studies as it helps researchers clearly differentiate the
population being investigated (Balcaen & Ooghe 2006; and Charitou, Neophytou &
Charalambous 2004). Notwithstanding this, researchers define failure differently
based on their own rationale and/or specific interest because there is no consensus on
what constitutes failure (Ohlson 1980, p. 111). This notion is supported by Dimitras,
Zanakis, and Zopounidis (1996, p. 487) who argue that failure can be defined in
65
many ways, depending on the specific interest or condition of the firms under
examination. This explains why there is a multitude of definitions of business failure
in the literature.
However, most extant studies narrowly define the term failure, resulting in bankrupt
firms dominating in their sample. For example, Altmans (1968) study focuses on
failed firms that filed a bankruptcy petition under the national bankruptcy act. This
narrower definition leads to a limitation of sample size in their studies (Leksrisakul
2004).
Focusing on the period of normal economic circumstances, this study investigates

firms in financial distress and/or fragility in the context of Thailand in recent years.
Given this, the determination of failure criteria in this study is more subtle than in
previous ones. Adopting a narrow legal definition such as the filing of a bankruptcy
petition would result in fewer firms in the failed group.
According to Queen and Roll (1987, p. 9):
for a publicly listed company, bankruptcy is not the only route to mortality. A
firm may go private, be acquired and the like. Firm mortality could be an
occasion for mourning on the part of other interested groups such as
employees, managers, clients, suppliers.
Likewise, Coats and Fant (1993, p. 147) state that bankruptcy is only one outcome of
financial distress. Furthermore, Routledge and Gadenne (2000, p. 233) note that
many companies choose reorganisation to deal with financial distress. Thus, it can
be inferred that distressed firms are financially weak but do not always become
bankrupt.
66
Therefore, a potentially failed firm is defined as a firm that has been delisted from the
Stock Exchange of Thailand (SET). This broader definition of failure is preferable
because it satisfies the purpose of the study and leads to the selection of dependent
variables. The SET outlines the conditions for delisting as follows:
1) When shareholder equity in a listed company is less than zero.

2) When shareholder equity in a listed company is greater than zero, but the
auditors report includes a qualified opinion, a disclaimer of opinion, or an
adverse opinion. If so, the companys financial statements must be adjusted in
accordance with the auditors opinion. If the resulting adjustment causes
shareholder equity to decrease to less than zero, the firm meets this delisting
criterion.
3) In considering shareholder equity, only losses arising from foreign debt that
existed before any change to the managed float system was announced will be
deemed by the SET as unrealized foreign exchange losses. If a listed company
would like this condition waived, it must provide details of any losses
mentioned in the management report, which is reviewed by an independent
auditor and is submitted to the SET.
4) The SET will not announce that a company needs to prepare a rehabilitation
plan if the company has been able to eliminate the problem by increasing
shareholder equity to more than zero. The company must submit financial
statements as of the date the problem is resolved, reviewed by auditors, or a
report showing that the causes that required the rehabilitation plan have been
eliminated, together with its audited financial statements, which are SET
requirement.
Source: SET (2006, pp. 1-3)
In this study, a firm is identified as failed if it is delisted from the SET. Otherwise, it
is identified as nonfailed. Like previous studies (e.g. Flagg, Giroux & Wiggins Jr.
1991; and Gentry, Newbold & Whitford 1985), any failed firm which is delisted for
merger is eliminated from this study.
67
Consistent with previous studies, the population of firms in this study consists of all
firms listed on the SET and classified as industrial and/or nonindustrial (excluding
banks and other financial institutions). The reason why publicly listed firms are
selected is that there is difficulty in accessing to financial statement data of Thai
private firms whose stocks are not publicly traded. Banks and other financial
institutions are excluded from the study because their financial statements are
prepared on a different basis from those of industrial firms (Bryant 1997; Charitou,
Neophytou & Charalambous 2004; Flagg, Giroux & Wiggins Jr. 1991; He & Kamath
2006; and Ohlson 1980).
Last but not least, the frequency of firms delisting in Thailand is examined to provide
descriptive statistics of the event of failure. The data used to determine the percentage
of all failed firms in Thailand is derived from the delisting securities file as of 20 June
2008, which is published by the SET. Details of the delisting securities have been
provided since 1975. Table 4.1 shows the total number of failed firms in each year
from 1975 to 2008 and the percentage of delisted firms that delisted each year.
Table 4.1 Descriptive statistics of the event of failure in Thailand from 1975 to 2008
Year Number of delisting firms (a) Percentage of delisting firms
19751993 (b) 20 12.42%
1994 1 0.62%
1995 1 0.62%
1996 2 1.24%
1997 28 17.40%
1998 14 8.70%
1999 26 16.15%
2000 13 8.07%
2001 6 3.73%
2002 11 6.83%
2003 4 2.48%
2004 7 4.35%
2005 10 6.21%
2006 4 2.48%
2007 8 4.97%
2008 (c) 6 3.73%
Total 161 100.00%
Note: a) Includes banks and other financial institutions
b) Financial statement data have been available since 1994
c) As of 20 June 2008
Source: SET (2008)
68
The lowest number of firms that delisted in one year between 1975 and 1993 is zero
and the highest number is five. It is obvious that 1997 was a critical year for the Thai
capital market. The number of delisting firms in Thailand significantly surged from
two firms in 1996 to 28 firms in 1997. Of the 28 delisting firms in 1997, 25 were
financial institutions. That Thailand was hit hard by the 1997 financial crisis is self
evident.
4.4.2 Sample selection

The term of sample refers to a smaller set of cases a researcher selects from a larger
pool and generalizes to the population (Neuman 2006, p. 219). Stated simply by
Zikmund (2003, p. 369), a sample is a subset, or some part, of a larger population.
This view is shared by Cavana, Delahaye and Sekaran (2001) and Davis (2005).
It is apparent that a sample in business failure prediction studies is composed of both

failed and nonfailed firms. The dichotomous classification method is generally
accepted by researchers for failure prediction (Dimitras, Zanakis & Zopounidis 1996,
p. 490). Firms incorporated in the failed sample group for this study had to satisfy the
following criteria:
a) Firms had to be delisted from the SET. Notwithstanding this, firms which
were delisted for merger were eliminated from the failed sample group.
b) Firms had to be classified as industrial and/or nonindustrial. Banks and other
financial institutions were excluded from the dataset because of differences in
financial statement reporting.
Firms had to have failed between 2003 and 2008. This period was chosen because this
study aims to investigate recent corporate failure of Thai firms in the postfinancial
crisis period. The aforementioned period represents the period in which firms failed
but financial statement data of these firms up to five years before failure are collected.
Therefore, a range of 10 years financial statement data (19982007) was used in this
study.
69
An initial list of the failed firms was identified by reviewing the SET delisting
securities files as of 20 June 2008. As mentioned earlier, banks and other financial
institutions were eliminated from the sample data due mainly to differences in
financial structure. Failed firms which were delisted for merger were excluded. In
addition, firms had to have failed during the period of 20032008. As a result, 20
firms were identified from among the total of 39 firms that failed in the specified
period. Table 4.2 depicts the number of failed sample firms in a particular year and
the percentage of the total sample that failed in that year.
Table 4.2 Descriptive statistics of initially failed sample group during the period of 2003
2008
Year Number of failed firms Percentage of failed firms
2003 4 20.00%
2004 2 10.00%
2005 6 30.00%
2006 1 5.00%
2007 3 15.00%
2008 4 20.00%
Total 20 100.00%
Source: SET (2008)
To investigate the predictive ability of financial statement data for business failure,
both failed and nonfailed firms are required. A sample that contained only failed
firms would be inadequate for this study because the traditional ratiobased measures
used, i.e. comparative ratio analysis and ratio trend analysis, need to compare and
contrast the financial ratios of both groups. The principal reason is that the symptoms
of failed firms can be more effectively assessed by comparing and contrasting their
financial ratios to those of nonfailed firms. This approach is supported by Palats
(1989, p. 3) notion that:
it is important to remember that for comparative analysis, a peer group

(companies similar in business and size) in the same industry should be used,
otherwise the conclusions drawn will be incorrect.
70
Selection of nonfailed firms in this study was based on a paired sample technique.
The paired sample technique, as the term implies, compares a failed firm to a non
failed firm in the sample. The underlying reason for employing this technique is to
ensure that the sample of firms is as similar as possible in all aspects except for their
financial ratios (Norton & Smith 1979, p. 73). The paired sample technique has been
extensively used by a number of studies (for example, Altman 1968; Back et al. 1996;
Beaver 1966; Bei & Liu 2005; Blum 1974; Charitou, Neophytou & Charalambous
2004; Dambolena & Khoury 1980; Darayseh, Waples & Tsoukalas 2003; Deakin
1972; Dimitras et al. 1999; Gentry, Newbold & Whitford 1985; Ginoglou, Agorastos
& Hatzigagios 2002; He & Kamath 2006; McKee 2000; and Ugurlu & Aksoy 2006).
This observation is confirmed by Gilbert, Menon and Schwartz (1990). Most business
failure studies employ an equal proportion of bankruptcy and nonbankruptcy firms
to develop their discriminating model (Gilbert, Menon & Schwartz 1990, p. 161).
Apart from the paired sample technique, the criteria for selecting the firms in the
matching samples need to be discussed. In this study, each failed firm was matched to
a nonfailed mate which was in a similar industry, which was the closest in asset size,
and which had the same statement date. The underlying reason for applying these
criteria is to attempt to control different factors when examining the ability of
financial statement data to predict failure. Beaver (1966, p. 74) states that differences
exist among industries that prevent the direct comparison of firms from different
industries. For example, a current ratio of firm X in the consumer products industry
might noticeably differ from that of firm Y in the property and construction industry.
The difference is because the two firms belong to two different industries. As a result,
comparing them would lead to a wrong interpretation. It is widely accepted by
researchers that financial ratios vary across industry sectors. Therefore, it is
imperative to match companies in similar industries so as to make the paired
companies more comparable in terms of their business activities.
Besides the industry in which a firm operates, a number of studies confirm that firm
size influences its financial ratios (for instance, the study of Osteryoung, Constand &
71
Nast 1992). Like previous studies, this study employs asset size to represent firm size.
The justification of using asset size is that asset size, rather than net sales, tends to be
more stable across time (Norton & Smith 1979, p. 73). Having the same statement
date was used as a matching criterion because this practice helps ensure that the
comparison was of companies in the same period and therefore of companies
operating under the same business conditions. This was necessary because business
conditions (e.g. depression, recovery, and prosperity) change over time (Fitzpatrick
1931). In this respect, being in a similar industry, being the closest firm in asset size,
and having the same statement date provided measurable compatibility among the
companies studied. Therefore, these criteria were used in this study. Having described
the matching criteria used in this study, the discussion now turns to the procedure for
selecting nonfailed firms.
The procedure for selecting of nonfailed firms in this study was as follows:
1) Match each failed firm to a nonfailed firm using the industry code
determined by the SET.
2) Calculate the average asset size for each failed firm based on the data
availability over the investigation period.
3) Calculate the average asset size for each nonfailed firm in the same industry
and similar investigation period.
4) Select the nonfailed firm that has the closest asset size to that of a failed firm
as a paired counterpart.
Source: Adapted from Beaver (1966) and Hossari (2007)
After the procedure for selecting nonfailed firms was set, a failed firm was paired to
a nonfailed mate based on the matching criteria. Of the 20 failed firms in Table 4.2,
only 14 were included in the final sample. There were two main reasons why six
failed firms were dropped. First, three failed firms had no paired match on the basis of
similar industry, closest asset size, and the same statement date. Second, for the other
72
three failed firms, at least three quarters of financial statement data were not available
from the SET database.
Table 4.2a Number of failed firms and their counterparts in the sample
Year Number of failed firms Number of non-failed firms
2003 1 1
2004 2 2
2005 5 5
2006 1 1
2007 3 3
2008 2 2
Total 14 14
Source: SET (2008)
Besides the qualitative criteria used to identify the failed firms, a quantitative criterion
was applied to ensure that all selected failed firms belonged to the same population.
The quantitative criterion provided data triangulation. To do so, this study adopted the
theoretical concept of the outliers test to examine whether observations would
statistically be likely to have been sampled from another population (Manning &
Munro 2007, p. 49). In accounting research, multivariate outliers should be identified
to enhance statistical conclusion validity and to improve the effectiveness of decision
models (Watson 1990, p. 682).
Multivariate outliers can be detected by computing the Mahalanobis distance for each
observation. If the value of the Mahalanobis distance is greater than a critical value
(recommended p<.001), they are deemed to be multivariate outliers (Manning &
Munro 2007, p. 55). According to a test for multivariate outliers (shown in Appendix
A), Mahalanobis distance scores for the selected failed firms are not greater than the
critical value (p<.001). This means no multivariate outliers were identified. Therefore,
it can be concluded that the failed firms all belong to the same population. The 14
failed firms are listed in Table 4.3 with their industry groups.
73
Table 4.3 List of failed firms based on the matching criteria (20032008)
No Security Company Name Date of Date of de Type of Industry
listed listed delisting groups (N=6)
1 PPPC PHOENIX PULP & PAPER PCL 14/12/1989 07/01/2005 V Industrials
2 TLI THAI LIFT INDUSTRIES PCL 31/08/1994 23/09/2005 V Industrials
3 UBC UNITED BROADCASTING 27/02/1992 11/04/2006 V Services
CORPORATION PCL
4 ASTL AMERICAN STANDARD 22/01/1988 01/02/2005 V Property and
SANITARYWARE (THAILAND) Construction
PCL
5 UFC UNIVERSAL FOOD PCL 16/12/1988 28/10/2004 V Agro and Food
6 UCOM UNITED COMMUNICATION 01/02/1994 14/09/2007 V Technology
INDUSTRY PCL
7 HT HUA THAI MANUFACTURING 29/07/1991 15/01/2008 V Consumer
PCL Products
8 APSP ALCAN PACKAGING 16/05/1978 29/11/2005 V Industrials
STRONGPACK PCL
9 UTL UNITHAI LINE PCL 23/06/1994 13/12/2003 V Services
10 RHC RAJADAMRI HOTEL PCL 09/04/1990 12/07/2007 V Services
11 MFG THE MINOR FOOD GROUP PCL 18/12/1991 05/01/2005 V Agro and Food
12 MPT MAGNECOMP PRECISION 08/06/1995 20/05/2008 V Technology
TECHNOLOGY PCL
13 TGP THAI GYPSUM PRODUCTS PCL 17/11/1980 05/10/2004 V Property and
Construction
14 AF AAPICO FORGING PCL 26/01/2005 21/06/2007 V Industrials
Note: Type of delisting: M=Mandatory, V=Voluntary
Source: SET (2008)
As mentioned earlier, the sample of nonfailed companies is necessary for

investigating the ability of financial statements to be used to signal business failure.
The screening process of nonfailed mate selection was carefully chosen based on the
approach established in previous studies to ensure that the paired companies were
comparable to the failed companies. The rationale for the matching procedure has
already been discussed above.
The data set contains all of the firms that met the criteria for analysis established in
this study for the period 2003 to 2008. The sample of failed firms and their paired
nonfailed mates is listed in Table 4.4 together with their business sectors.
74
Table 4.4 List of failed firms and their nonfailed mates (20032008)
Failed firms Nonfailed firms Business sectors
Security Company Name Security Company Name
PPPC PHOENIX PULP & TCP THAI CANE PAPER PCL Paper and Printing
PAPER PCL Materials
TLI THAI LIFT TCJ T.C.J. ASIA PCL Machinery and

INDUSTRIES PCL Equipment
UBC UNITED BEC BEC WORLD PCL Media and Publishing
BROADCASTING
CORPORATION PCL
ASTL AMERICAN RCI THE ROYAL CERAMIC Construction
STANDARD INDUSTRY PCL Materials
SANITARYWARE
(THAILAND) PCL
UFC UNIVERSAL FOOD TWFP THAI WAH FOOD Food and Beverage
PCL PRODUCTS PCL
UCOM UNITED JAS JASMINE Information and
COMMUNICATION INTERNATIONAL PCL Communication
INDUSTRY PCL Technology
HT HUA THAI TPCORP TEXTILE PRESTIGE Fashion
MANUFACTURING PCL
PCL
APSP ALCAN PACKAGING NEP NEP REALTY AND Packaging
STRONGPACK PCL INDUSTRY PCL
UTL UNITHAI LINE PCL TTA THORESEN THAI Transportation and
AGENCIES PCL Logistics
RHC RAJADAMRI HOTEL ROH ROYAL ORCHID Tourism and Leisure
PCL HOTEL (THAILAND)
PCL
MFG THE MINOR FOOD UFM UNITED FLOUR MILL Food and Beverage
GROUP PCL PCL
MPT MAGNECOMP KCE KCE ELECTRONICS Electronic
PRECISION PCL Components
TECHNOLOGY PCL
TGP THAI GYPSUM UMI THE UNION MOSAIC Construction
PRODUCTS PCL INDUSTRY PCL Materials
AF AAPICO FORGING PCL SMC MOTORS PCL Automotive
A further data sampling issue, that of the investigation period for the companies
studied, needs to be discussed. There is not unanimous agreement among researchers
as to the optimal investigation period which should be used in the study of business
failure prediction. Nevertheless, it is apparent that a large number of studies,
including the seminal works of Beaver (1966) and Altman (1968), use an
investigation period of five years prior to business failure. The findings suggest that
the predictive ability of any model decreases as the lead time increases. Hamer (1983,
p. 295) notes that financial ratios are useful in predicting failure as much as five
years before the event. This notion is shared by Beaver (1966). Moreover, Hossari
(2006, p. 222) states that:
75
The objective of any corporate collapse prediction model is to signal collapse
before it happens. If the reporting period were too short, it would be too late to
take corrective action and try to turn the company around. Likewise, if the
reporting period were too long, then the prediction model might not detect any
signs of impending collapse.
According to Hossaris (2006) study, 22 out of the 60 studies during 19682004 were
found to employ the investigation period of five years. Given this, it is reasonable to
infer that an investigation period of more than or less than five years might be
ineffective in signalling business failure. In other words, the optimal investigation
period for business failure is five years. Therefore, a five-year investigation period
was adopted for this study.
Coats and Fant (1993, p. 152) suggest using quarterly data in place of annual data
and using several time periods in conjunction to look for time series patterns in the
magnitude of changes in ratios. Since the event of failure is dynamic, the value of
ratios should be inspected over time to provide full information about the progress of
a firm (Dimitras, Zanakis & Zopounidis 1996, p. 488). Given this, quarterly financial
statement data are likely to provide rich detail about shifts in financial characteristics.
For example, using quarterly data enables researchers to detect the shift as soon as it
exists, before the end of the accounting period.
In the context of prediction model development, quarterly financial data enables

researchers to use more recent financial data than does annual financial data. For
example, the average lead time of financial statements prior to failure in Altmans
(1968) study and Ohlsons (1980) study is approximately seven and onehalf months,
and thirteen months, respectively. Consequently, this study refines the methodology
by examining quarterly financial data rather than annual financial data.
76
Since this study examines quarterly data rather than annual data, the five years
financial statements are broken down to 20 quarters. The first quarter before failure is
defined as the most recent quarter prior to the quarter in which the firm failed. For
example, if one firm failed in quarter 1 of 2003 and another failed in quarter 4 of
2007, the first quarter before failure is quarter 4 2002 for the first firm and quarter 3
2007 for the second firm. The second, third, fourth, , and twentieth quarters are
similarly defined. The breakdown of data availability for the investigation period up
to 20 quarters prior to failure is shown in Table 4.5.
Table 4.5 Breakdown of companies in the sample by 20quarter investigation period

Investigation period (prior to failure)
Failed Nonfailed Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
firms firms 1 2 3 4 5 6 7 8 9 1 1 1 1 1 1 1 1 1 1 2
0 1 2 3 4 5 6 7 8 9 0
PPPC TCP
TLI TCJ
UBC BEC
ASTL RCI
UFC TWFP
UCOM JAS
HT TPCORP
APSP NEP
UTL TTA
RHC ROH
MFG UFM
MPT KCE
TGP UMI
AF SMC
In the Table 4.5, the shaded areas indicate that quarterly financial statements of each
pair in the sample are available for the specific investigation period. In the SET
database, only 17 quarters of financial statement data are available for TGP and only
10 quarters financial statement data are available for AF.
77
This concludes the discussion of the sampling design. After the sample group was
identified, the next task was to specify the explanatory variables that were used by the
three ratiobased analytical tools and the logit-based model in this study.
4.5 Data collection

After the population was specified and firms were selected, financial statement data were
collected (Altman 1968, 1993; and Altman & Hotchkiss 2006). A fundamental purpose of
financial statements is to provide not only information concerning the results of an entitys
operations, but also information useful for making economic decisions (Altman 1982, pp. 5-
6). A large number of financial ratios are derived from financial statement data in the balance
sheet and the income statement. Financial ratios are defined as a quotient of two numbers,
where both numbers consist of financial statement items (Beaver 1966, pp. 71-2). Ratio
analysis enables researchers to compare a companys financial performance with that of other
companies (Beal, Goyen & Shamsuddin 2008; Brigham & Ehrhardt 2008; Gibson &
Frishkoff 1986; and Palat 1989). Brigham and Ehrhardt (2008, p. 141) define the comparative
technique as benchmarking and the firms used for the comparison as benchmarking
companies.
Financial ratios play a crucial role in signalling business failure despite the fact that
alternative discriminating variables (i.e. GDP growth, interest rates, inflation, stock returns,
and the like) are also influential. Recently, Beaver, McNichols and Rhie (2005) studied the
ability of financial ratios to predict business failure during the period 19622002. Their
findings confirm that the ability of financial ratios to predict business failure is robust over
time. The study was conducted in the western environmental setting. 1 However, there has not
been research conducted in the Thai environment on the ability of financial ratio analysis to
signal financial distress.
1
It is noted that the sample period and analytical technique used in Beaver, McNichols and Rhies (2005) study
and this study are different.
78
It is apparent that there are a lot of financial ratios which may be derived from financial
statements. The question raised is which ratios should be used in studying business failure.
When it comes to selecting financial ratios, the usefulness of ratios should be taken into
consideration (Altman 1968; Beaver 1966; Beaver, McNichols & Rhie 2005; Dambolena &
Khoury 1980; Edmister 1972; Frydman, Altman & Kao 1985; Gentry, Newbold & Whitford
1985; Jones & Hensher 2004; and McKee 2000). The usefulness of financial ratios is gauged
by examining how often particular ratios are employed in classifying corporate failure in
previous studies.
Hossaris (2006) survey of 84 studies which use financial ratios to predict corporate failure
for the period 19682004 identifies 44 financial ratios as potentially useful indicators of
corporate failure. These studies include Altman (1968, 1973); Charitou, Neophytou and
Charalambous (2004); Coats and Fant (1993); Dambolena and Khoury (1980); Deakin
(1972); Dimitras et al. (1999); Edmister (1972); Fydman, Altman and Kao (1985); Hamer
(1983); Jones and Hensher (2004); Norton and Smith (1979); Ohlson (1980); Taffler (1982);
Wilson and Sharda (1994); Zapranis and Ginoglou (2000); and Zmijewski (1984). According
to Dimitras, Zanakis and Zopounidis (1996, p. 492), the ratios found useful in earlier studies
were the first under consideration by many researchers and subsequently used in later
studies. Each ratio is grouped based on certain category measures, which are possession,
ownership, liquidity, performance, and peripheral activities. The 44 financial ratios are
summarised in Table 4.6.
Table 4.6 Original list of the explanatory variables

No Category Acronym Full description
1 Possession (a) C/TA Cash/Total Assets
2 Possession CA/TA Current Assets/Total Assets
3 Possession CF/TA Cash Flow/Total Assets
4 Possession CL/TA Current Liabilities/Total Assets
5 Possession EBT/TA Earnings Before Taxes/Total Assets
6 Possession EBIT/TA Earnings Before Interest and Taxes/Total Assets
7 Possession FA/TA Fixed Assets/Total Assets
8 Possession LTL/TA LongTerm Liabilities/Total Assets
9 Possession NI/TA Net Income/Total Assets
10 Possession QA/TA Quick Assets/Total Assets
11 Possession RE/TA Retained Earnings/Total Assets
12 Possession S/FA Sales/Fixed Assets
79
13 Possession S/TA Sales/Total Assets
14 Possession TE/TA Total Equity/Total Assets
15 Possession TL/TA Total Liabilities/Total Assets
16 Possession WC/TA Working Capital/Total Assets
17 Ownership (b) CL/TE Current Liabilities/Total Equity
18 Ownership EBIT/TE Earnings Before Interest and Taxes/Total Equity
19 Ownership FA/TE Fixed Assets/Total Equity
20 Ownership LTL/TE LongTerm Liabilities/Total Equity
21 Ownership NI/TE Net Income/Total Equity
22 Ownership S/TE Sales/Total Equity
23 Ownership TL/TE Total Liabilities/Total Equity
24 Liquidity (c) C/CL Cash/Current Liabilities
25 Liquidity CA/CL Current Assets/Current Liabilities
26 Liquidity CF/CL Cash Flow/Current Liabilities
27 Liquidity CF/TL Cash Flow/Total Liabilities
28 Liquidity MVE/TL Market Value of Equity/Total Liabilities
29 Liquidity QR Quick Assets/Current Liabilities
30 Liquidity TE/LTL Total Equity/LongTerm Liabilities
31 Liquidity TE/TL Total Equity/Total Liabilities
32 Performance (d) C/S Cash/Sales
33 Performance CA/S Current Assets/Sales
34 Performance CF/S Cash Flow/Sales
35 Performance EBIT/S Earnings Before Interest and Taxes/Sales
36 Performance Exp/S Expenses/Sales
37 Performance Inv/S Inventory/Sales
38 Performance NI/S Net Income/Sales
39 Performance QA/S Quick Assets/Sales
40 Performance WC/S Working Capital/Sales
41 Peripheral activity (e) AR/Inv Accounts Receivable/Inventory
42 Peripheral activity EBIT/I Earnings Before Interest and Taxes/Interest
43 Peripheral activity Inv/WC Inventory/Working Capital
44 Peripheral activity S/Inv Sales/Inventory
Note: a) All ratios in this category have assets as denominators.
b) All ratios in this category have equity as denominators.
c) All ratios in this category have liabilities as denominators.
d) All ratios in this category have sales as denominators.
e) All ratios in this category are those that do not lie in any of the previous categories.
Source: Adopted from Hossari (2006)
The original list of the 44 financial ratios was used as a starting point for selecting financial
ratios in this study. Unlike Hossaris (2006) study, this study classifies financial ratios based
on four key meaningful areas of financial performance. They include liquidity,
turnover/performance, leverage/solvency, and profitability (Blum 1974; Brigham & Ehrhardt
2008; Ohlson 1980; Persons 1999; Ross et al. 2007; and Tirapat & Nittayagasetwat 1999).
The financial ratios chosen for the three ratiobased measures in this study are discussed in
the next section.
80
Another key issue about financial ratio is the availability of data. During the 20 quarters
investigation period, not all selected financial ratios of each company in the data set could be
computed due to missing financial items. In such cases, the problematic ratios were ignored
for that particular company. The underlying reason for casting out the problematic ratios in
this study is that the situation and symptoms of failed firms can be assessed by comparing
and contrasting their financial ratios to those of their nonfailed counterparts. Another reason
is that the total number of 28 firms (14 failed and 14 nonfailed firms) in this study is
considered quite a small number compared to that of previous studies. A decision to reduce
the number of sample firms, instead of ignoring the problematic ratios, might have called into
question the representativeness of the data set.
The financial ratios of the failed firms were calculated for up to 20 quarters depending on
data availability. The results are displayed so that the first quarter is the one immediately
prior to the one in which the firm failed. Likewise, the financial ratios of each nonfailed firm
are computed and displayed with the first quarter being the one immediately before the
failure of the firm with which is linked. Having described the principal source of the data set
that is used, the next step is to discuss the three ratiobased measures applied in this study.
4.6 Analytical approach

After describing the data set of companies and the original list of financial ratios, the
discussion now turns to the analytical approach of this study. As can be seen in previous
studies, the selection of the appropriate analytical approach depends mainly on the research
purpose, the researchers rationale, and the influence of data constraints.
As stated earlier, this study aims to explore the predictive ability of financial ratios under
normal economic circumstances in Thailand (rather than to find the best predictors of failure).
The analytical approach used in this study applies from the earlier work of Fitzpatrick (1931),
Beaver (1966) and Blum (1974) in utilising comparative ratio analysis and ratio trend
analysis; augmented by the use of the Emerging Market Score model to classify financially
distressed and stable firms. The Integrated Multi-Measure approach was developed for this
81
study. The approach combined the Emerging Market Score model, comparative ratio analysis
and ratio trend analysis. In addition, the logit-based model was employed to verify the results
of the IMM approach. Details of the criteria used in the analytical approach are described
below.
Before presenting the Integrated MultiMeasure approach used in this study, the derivation of
the financial ratios used in the three measures needs to be discussed.
In the Emerging Market Score model, the original four ratios of the model are used in this
study. They are: working capital to total assets, retained earnings to total assets, earnings
before interest and tax to total assets, and total equity to total liabilities. Except for the ratio
of total equity to total liabilities, the rest are in the most frequently used group of the 44
financial ratios. The definition of these ratios will be discussed further.
Concerning comparative ratio analysis and ratio trend analysis, the financial ratios used in the
two measures are classified into four meaningful areas of financial performance, which have
already been discussed. Of the 44 ratios, 15 (including the ratios in the Emerging Market
Score model, except the total equity to total liabilities ratio) were selected based on their use
in previous studies. The underlying reasons for eliminating the total equity to total liability
ratio are (1) that the ratio is the inverse form of the total liabilities to total equity ratio, and (2)
that the ratio of total liabilities to total equity has a higher overall usage rate than that of the
total equity to total liabilities. Like the ratios in the Emerging Market Score model, most of
the selected 15 ratios are those that were most frequently used in the 84 studies during 1968
2004. Furthermore, most of the selected ratios are frequently referred to as common ratios in
the four financial groups by academics (e.g. Brigham & Ehrhardt 2008; Gibson & Frishkoff
1986; and Palat 1989). The four financial groups are defined as follows:
Liquidity: This measures a companys ability to meet its shortterm obligations. Simply
stated, it indicates how well a firm can pay off its currently maturing liabilities. If a firm
82
could not pay off its current obligations, it might culminate in business discontinuity
(Gibson & Frishkoff 1986). Consequently, firms with relatively higher liquidity ratios are
less likely to suffer financial distress. The three commonly used measures in this category
are described as follows:
Current ratio: This ratio is calculated by dividing current assets by current

liabilities. Current assets consist of cash, marketable securities, receivables, and
inventories. Current liabilities are made up of payables, current maturities of
longterm debts, and other liabilities payable within one year (Brigham &
Ehrhardt 2008; and Palat 1989).
Quick ratio: This ratio is computed by subtracting inventories from current assets
and then dividing by current liabilities. The idea behind this ratio is that
inventories have less liquid condition to be converted to cash and it is sometimes
called the acid test ratio (Brigham & Ehrhardt 2008; and Gibson & Frishkoff
1986).
Working Capital/Total Assets: Being one of the four ratios in the Emerging
Market Score model, this ratio compares the net liquid assets to the total assets. In
other words, it measures liquidity relative to the size of a company. Working
capital is calculated as current assets minus current liabilities (Beal, Goyen &
Shamsuddin 2008; Brigham & Ehrhardt 2008; and Frino, Amelia & Chen 2009).
Turnover/performance: This measures the capacity of a company to manage its assets.

Ratios in this group deal with the question: How effectively is a company managing its
assets? (Brigham & Ehrhardt 2008; and Palat 1989). The earningsgenerating capacity of
assets can be described by dividing sales by the value of assets. Higher values of total
assets and inventory turnover ratios and lower values of operating ratios are likely to be
associated with a lower probability of financial distress. The selected ratios in this group
include:
83
Total assets turnover ratio: This ratio is derived by dividing sales by total assets.
The relationship between a companys size and the volume of sales is revealed in
this ratio (Palat 1989). Since this ratio indicates the productivity of total assets, an
increase of the ratio over time is regarded as a satisfactory result and a decrease
over time is regarded as unsatisfactory.
Operating ratio: This ratio is calculated by dividing total operating expenses by

sales. It measures how efficiently a company manages funds used as operating
expenses in managing its operation. This ratio focuses on the costs of doing
business compared to sales. A decrease of the ratio indicates that a company has
efficiently managed the costs of operating its business.
Inventory turnover ratio: This ratio is defined as sales divided by inventories. It

measures how quickly a companys inventory is sold out, or turned over (as it is
termed). The ratio is important because the faster the inventories turn, the greater
the profit (Brigham & Ehrhardt 2008).
Leverage/solvency: This group of ratios measures to what extent a company operates on

its own funds and/or depends on funds from outside sources. In other words, most ratios
in this group aim to examine a companys financial structure. It provides a realistic
picture of the level to which a companys assets are financed by funds borrowed from
creditors and funds invested from owners. The ratios in this group are:
Debt ratio: This ratio is calculated by dividing total liabilities by total assets. It
measures the proportion of total debts a company has relative to total assets.
Industry factors affect the level of debt used in a company. For example, a firm in
a capitalintensive industry tends to use greater debt financing. This problem can
be compensated for by comparing a companys debt ratio with that of its peers in
the same industry. Gibson and Frishkoff (1986, p. 238) state that from a long
84
term debtpaying ability perspective, the lower this ratio, the better the companys
position.
Debt to Equity (D/E) ratio: This ratio is computed by dividing total liabilities by
total equity. The D/E ratio gauges to what extent a company is operating on funds
invested by shareholders and to what extent it is dependent upon funds borrowed
from creditors. The debt ratio and the D/E ratio both examine longterm debt
paying ability (Brigham & Ehrhardt 2008). The underlying rationale of selecting
these two ratios is that both ratios are frequently used in the literature. Like the
debt ratio, the higher the D/E ratio, the greater the companys risk.
Interest coverage ratio: This ratio is determined by dividing earnings before

interest and tax by interest expenses. The ratio, sometimes called timesinterest
earned ratio, measures a companys ability to pay interest. A low or decreasing
ratio is a sign of weakness and potential financial distress.
Profitability: This group of ratios measures a companys ability to generate profit from
its operations. These ratios draw stakeholders attention because money earned from
profit can be reinvested in a company for the coming period, paid out as dividends to
shareholders, and used to pay off funds loaned by creditors (Gibson & Frishkoff 1986).
Consequently, higher values of profitability ratios are likely to be associated with lower
probability of financial distress. These ratios include:
Operating profit margin: This ratio is determined by dividing earnings before

interest and tax by sales. It measures to what extent a company is able to make a
profit from its operations. The idea of this ratio is that nonoperating expenses
should not be considered in the return on sales.
85
Net profit margin: This ratio is calculated by dividing net income by sales. The
ratio measures the profitability of a company after the costs of financing and tax
are deducted. In other words, this is the true profit for the shareholders of a
company.
Operating return on assets: Being one of the four ratios in the Emerging Market
Score model, this ratio is computed by dividing earnings before interest and tax by
total assets. It measures the earnings power of assets from an operational
perspective. The costs of financing and paying tax are excluded in order to assess
the true earningsgenerating capacity of assets from operations (Brigham &
Ehrhardt 2008; and Palat 1989). The true earnings capacity of total assets is the
key factor for a companys survival (Altman 1968).
Net return on assets (ROA): This ratio is determined by dividing net income by
total assets. It assesses how efficiently a company manages its assets to generate
earnings. Unlike the ratio of operating return on assets, ROA indicates the
earningsgenerating capacity of assets after the costs of financing and tax are
deducted.
Return on equity (ROE): This ratio is calculated by dividing net income by total
equity. ROE measures how efficiently a company uses owners equity to generate
profit. In addition, ROE enables shareholders to determine whether the return
made by a company from invested funds is acceptable (Palat 1989).
Retained Earnings/Total Assets: Being one of the four ratios in the Emerging
Market Score model, this ratio measures to what extent a company uses internally
accumulated funds from its operations (lowrisk capital). Retained earnings
represent the cumulative amount of earnings of a company (Altman 1968; and
Gibson & Frishkoff 1986). A high, or increasing, ratio is a positive sign, showing
a company can retain more earnings.
86
These are definitions of the 15 selected ratios (including the ratios in the Emerging Market
Score model, except the total equity to total liabilities ratio) used in comparative ratio
analysis and ratio trend analysis. Due to the fact that the ratio of total equity to total liabilities
is excluded in the two measures, the definition of this ratio is separately provided as follows:
Total Equity to Total Liabilities ratio: Being one of the four ratios in the
Emerging Market Score model, this ratio measures to what extent a companys
operations rely on funds invested by owners and borrowed from creditors. As
mentioned earlier, this ratio is the inverse form of the D/E ratio so it represents the
level of financial risk borne by a company.
The aforementioned ratios are used for data analysis in this study. A total of 16 ratios are
summarised in Table 4.7 below.
Table 4.7 Summary of the 16 financial ratios

Category Ratio (N=16) Ratio description
(N=4)
Liquidity Current ratio Current Assets/Current Liabilities
Quick ratio Current AssetsInventories/Current
Liabilities
Working Capital/Total Assets* Working Capital/Total Assets
Turnover/ Total Assets turnover Sales/Total Assets
Performance Operating ratio Total Expenses/Sales
Inventory turnover Sales/Inventories
Leverage/ Debt ratio Total Liabilities/Total Assets
Solvency D/E ratio Total Liabilities/Total Equity
Interest coverage Earnings before Interest and Tax/Interest
Total Equity/Total Liabilities* Total Equity/Total Liabilities
Profitablility Operating profit margin Earnings before Interest and Tax/Sales
Net profit margin Net Income/Sales
Operating return on assets* Earnings before Interest and Tax/Total
Assets
Net return on assets (ROA) Net Income/Total Assets
Return on Equity (ROE) Net Income/Total Equity
Retained Earnings/Total Retained Earnings/Total Assets

Assets*
Note: * ratios used in the EMS model
87
Having described the financial ratios used in this study, the discussion now turns to the three
analytical tools adopted. After careful consideration of the purpose of the study, three
analytical tools, which are the key components for testing the predictive ability of financial
statement information, were selected based on their ability to distinguish between financially
distressed firms and nonfinancially distressed firms. The three measures included the
Emerging Market Score model, comparative ratio analysis, and ratio trend analysis. Each of
the three measures used relies on financial ratios, which satisfies the purpose of this study. In
other words, they are ratiobased measures. The Emerging Market Score model is discussed
first, followed by comparative ratio analysis and ratio trend analysis.
4.6.1 Emerging Market Score (EMS) Model

The EMS model was chosen as the first analytical tool because it explicitly aims to
assess the financial health of companies in emerging markets. The model also
represents a sophisticated ratiobased measure used in this study. The EMS model is
applicable for manufacturing and nonmanufacturing companies in emerging markets
(Altman 2005). The model takes account of critical factors of emerging markets i.e.
currency vulnerability, industry risk, and competitive position (Altman & Hotchkiss
2006). The EMS model uses four variables along with a constant value as follows:
EMS 3.25 6.56 X 1 3.26 X 2 6.72 X 3 1.05 X 4 ......(4.1)
Where:
X1 =Working Capital/Total Assets
X2 =Retained Earnings/Total Assets
X3 =Earnings before Interest and Taxes/Total Assets
X4 =Book Value of Equity/Total Liabilities
EMS =Overall score
88
EMS> 5.85 : Safe zone
4.15 <EMS< 5.85 : Grey zone
EMS<4.15 : Distress zone
Since its advent, the EMS model has been widely applied to companies in emerging
markets like Brazil, Argentina, and a number of Southeast Asian countries. The
findings are impressive (Altman & Hotchkiss 2006). The key reasons for including
the model in this study are (1) that the model has a proven track record in predicting
business failure in emerging markets, (2) that this model is a ratiobased function,
which satisfies the studys objective, and (3) that the model is regarded as a
sophisticated failure prediction instrument.
As can be seen above, the original model has three zones of discrimination. They are:
a safe zone (score>5.85), a grey zone (5.85>score>4.15), and a distress zone
(score<4.15). Including a grey zone in this study could have resulted in faulty
interpretations. False interpretations could occur because, if the score of a company
falls into the range of the grey zone, the company could be viewed as being either
failed or nonfailed. If this were the case, the ability of financial statement data to
classify failed and nonfailed companies would be questionable. Another factor is
that, in practice, the cost of incorrectly classifying a failed firm as a nonfailed firm
(Type I error) is considered higher than the cost of misclassification a nonfailed firm
as failed firm (Type II error) (Balcaen & Ooghe 2004). Also, a dichotomous measure
is required for comparative purpose in this study. Given this, the cutoff score of 5.85
is selected in this study, which is the cutoff for the safe zone in the original
classification. This means that a company with a score below 5.85 is identified as
failed while a company which has score of 5.85 or above is identified as a surviving
firm.
89
For the purposes of this study, correct classifications of firms as failing prior to actual
failure for the failed group is observed and presented over the investigation period.
Thus, analytical evidence of the ability of methods utilising financial ratios for each
quarter in the five years prior to failure is presented to correctly predict firm failure. A
recent correct classification is identified when financial ratios are able to correctly
classify the firm in the quarter immediately before actual failure. A recent correct
classification demonstrates the ability of financial statement information to predict
business failure of the data set in this study. This is similar to the approach used in
previous studies.
4.6.2 Comparative ratio analysis

This technique is one of the most commonly used techniques for analysing financial
statements (Brigham & Ehrhardt 2008; Gibson & Frishkoff 1986; and Temte 2004).
This analysis can be used within a company or between companies. From an intra
company perspective, it enables researchers to examine the economic condition of a
business entity. For example, it indicates whether a company has enough cash to pay
off current obligations and/or long term bills; whether a company efficiently manages
its own resources; whether a company is a profit maker or loss sufferer; and whether
shareholders gain or lose from their investment. On the other hand, from the inter
company view, the analysis points out whether a company performs better or worse
than its competitors in the same industry. Stated simply, the intercompany analysis
provides information regarding the relative competitive position of two or more
companies. When doing so, the strengths and weaknesses of a companys financial
position are determined by comparing the companys financial ratios to those of its
peers.
Comparing a companys financial ratios to industry averages and/or to those of its

peers is called benchmarking (Brigham & Ehrhardt 2008, p. 141). Unlike the
situation in western countries, sources that provide information on industry
comparisons in Thailand are scarce. Therefore, comparing the ratios of a firm with
90
those of a benchmarking firm in the same industry is preferred. In addition, this
approach is predominantly applied in predicting business failure.
Some explicit potential shortcomings of applying comparative ratio analysis are the
effects of different industries/sectors, time (for example, the investigation period), and
company size factors (Brigham & Ehrhardt 2008; Gibson & Frishkoff 1986; and Palat
1989). Beaver (1966) states that such problems can be solved by using the paired
sample technique, which is adopted in this study. Finally, there are a number of
considerations that account for the use of comparative ratio analysis in this study.
Firstly, this technique was used in early research in this field by Fitzpatrick (1931)
and Merwin (1942). According to their findings, the technique is a useful tool for
identifying the deficiencies of failed firms. Secondly, the technique is currently
accepted in both academic and professional realms as a useful analytical tool for
financial statement analysis (Brigham & Ehrhardt 2008; and Temte 2004). Finally,
comparative ratio analysis is more meaningful when the effects of type of industry,
company size, and period of time are eliminated (Brigham & Ehrhardt 2008; and
Temte 2004).
The criteria for the selection of the 15 ratios used in this measure have already been
discussed. The ratios of a failed company are compared and contrasted to those of its
nonfailed mate over the investigation period. Also, descriptive statistics (e.g. the
mean, and standard deviation) and inferential statistics (e.g. ttest) were employed to
help describe financial characteristics of both groups.
For some ratios used in comparative ratio analysis (e.g. current ratio, return on assets,
return on equity and the like), a higher score was taken as a sign of strength while for
other ratios (e.g. debt to equity, total debt to total assets, and operating ratio) a higher
score was taken as a sign of weakness. For example, if a failed firm in a pair had a
higher current ratio than that of its nonfailed mate, this was taken as a sign of
91
strength for the failed firm and as a sign of weakness for its nonfailed mate (because
the nonfailed firm had lower current ratio).
When more than half of the 15 ratios for each pair indicated signs of relative
weakness for the failed firm in that pair compared to its nonfailed counterpart,
correct classification was assigned to the ability of financial ratios in signalling
financial distress.
As is the case for the EMS model, recent correct classifications using this measure
which indicate firm failure are observed and presented.
4.6.3 Ratio trend analysis

Like the comparative ratio analysis, this technique is a widely used technique for
financial statement analysis (Brigham & Ehrhardt 2008; Gibson & Frishkoff 1986;
and Temte 2004). Being regarded as a traditional financial analytical instrument, it
provides important clues for detecting financial symptoms and observing
managements performance. Brigham and Ehrhardt (2008) ascertain that this
technique is as important as the ratio analysis technique in the context of financial
statement analysis. The technique is more powerful when applied together with other
methods (Firminger 2003).
Though not as popular as ratio analysis for studies about predicting business failure,
this analysis is widely used in different contexts e.g. consumer behaviour, business
planning, education, health, and politics. Firminger (2003, p. 2) states that there are a
number of different types of trend analysis i.e. historical trend analysis, content
analysis, and cyclical pattern analysis. In the context of financial statement analysis,
both academics and professionals agree that one can gauge the financial status of a
company by using historical ratios and looking at trends of the ratios. To do so, as
Brigham and Ehrhardt (2008) explain, one simply plots the values of the ratios over
the investigation period.
92
The technique was used in Blums (1974) study in a western context. However, there
is no record of this technique being used for predicting business failure in Thailand.
Therefore, this study takes the opportunity to explore ratio trend analysis as an
analytical tool for signalling failure. Like comparative ratio analysis, trends of ratios
of a company are compared to those of a benchmark firm in the same industry.
The factors that justify the use of ratio trend technique in this study are as follows.
Firstly, it is widely accepted by researchers in this field that the failure process is
dynamic, not static (Altman 1970; Beaver, McNichols & Rhie 2005; Hossari 2007;
Shumway 2001; and Theodossiou 1993). Therefore, a dynamic analytical tool may be
more appropriate to detect warning signs of failure. By its nature, this technique
employs dynamic or historical time series data to assess the financial health of
companies. Secondly, there has been a recent shift in the methodology used for
modelling business failure prediction. Recent work such as Hossaris (2007) study
and Shumways (2001) study show that dynamic predictive models outperform
static/traditional models. Static/traditional models are models that use snapshot and/or
singleperiod data for estimation, while dynamic models are derived from multiple
period data. This technique satisfies these recent changes in approach. Thirdly,
dynamic models gather momentum each day. So applying this technique is likely to
add a broader perspective to signalling business failure in the Thai environment.
In our application of ratio trend analysis, the same 15 ratios were used as for
comparative ratio analysis. Also, the base quarter is employed as an index to assess
the financial status of a company by comparing values of ratios in each quarter with
the base quarter over the investigation period. To do so, the trend in the ratios of firms
in both groups can be determined in order to assess whether its financial situation was
improving or deteriorating. For example, if the net income to sales ratio of a failed
firm is declining, this is a sign of deteriorating financial situation. Where the trend of
more than half of the ratios for the failed firm were less favourable than those of the
benchmark partner firm, correct classification was assigned to the ability of trend in
financial ratios to signal financial distress. A less favourable trend would occur where
93
the financial ratios (say for the failed firm) were deteriorating and those of the
benchmark firm were improving, where the rate of deterioration for the failed firm
was greater than that of the benchmark firm, and where the rate of improvement was
less than that of the benchmark firm.
The selected ratios and the criteria used for the three measures have now been
presented. The next step is to discuss the development of the Integrated Multi
Measure approach.
4.7 Development of the Integrated MultiMeasure (IMM) approach

This section discusses the development of the IMM approach. The motivation for the
development of the IMM approach comes from the conceptual framework for multimethod
design in the literature. In social research, a number of researchers have adopted multi
method designs to provide more comprehensive understanding of social phenomena (Jacobs
2005). Studies employing the multimethod design combine different methods and/or
different approaches to data analysis in the same study. The combination can be a mix of
quantitative and qualitative methods, a mix of different quantitative methods, or a mix of
different qualitative methods (Brannen 2005; and Jacobs 2005). Since there are many ways to
combine different research methods, Spratt, Walker and Robinson (2004) classify studies
using multimethod design into two groups: multimethod study and mixed method study.
Spratt, Walker and Robinson (2004, pp. 6-7) state that multimethod studies use different
methods of data collection and analysis within a single research paradigm while mixed
method studies attempt to bring together methods from different paradigms.
However, the two terms (multimethod and mixed method) are used interchangeably by
researchers (for example, the studies of Birnberg, Shields & Young 1990; Gable 1994; and
Pager & Quillian 2005). These studies advocate that using a multimethod approach not only
helps compensate for the weaknesses of each method but also provides a more
comprehensive understanding of the subject being studied. In other words, a multimethod
design enables a researcher to crosscheck the findings by examining the subject from
94
different perspectives. This leads to the use of the term triangulation. Gable (1994, p. 123)
states that multimethod design is useful as a source of rich detail and a further means of
triangulation. In other words, broader and deeper dimensions of the findings can be expected
(Easterby-Smith, Thorpe & Lowe 1991). Likewise, Spratt, Walker and Robinson (2004, p. 6)
note that using multiple approaches provides more comprehensive answers to research
questions, going beyond the limitations of a single approach.
With reference to the conceptual framework for multimethod design, different approaches to
data analysis are used to thoroughly examine the power of financial statements in predicting
financial distress. Simply stated, investigation from different perspectives is conducted to
provide a more comprehensive understanding of the subject being studied. This leads to the
development of the IMM approach in this study. The IMM approach includes the EMS model,
comparative ratio analysis, and ratio trend analysis. The approach represents a combination of
a sophisticated ratiobased prediction model for business failure and two traditional
analytical tools for financial statement analysis. Integrating the sophisticated ratiobased
model with the traditional analytical tools is regarded as a means of obtaining triangulation.
The criteria used in each of the measures in the IMM approach are similar to those used in the
individual measures, which are as follows:
For the EMS model, a company that has a score below 5.85 is identified as a
potentially financially distressed firm while a company which has a score of 5.85
or above is identified as a financially healthy firm. Correct classification for a
failed firm is assigned when a firms computed score is less than the critical score
(5.85). Similarly, correct classification as a nonfailed firm is assigned when a
firms computed score is equal to or above the critical score.
For comparative ratio analysis, correct classification is assigned when more than
half of the 15 ratios for each pair indicated that the failed firm in a pair was
showing signs of relative weakness compared to its nonfailed counterpart.
For ratio trend analysis, classification of one firm as failed and the other as non
failed is assigned when the trends in more than half of the 15 ratios for each pair
95
point to this conclusion (ie. When the trends in the majority of financial ratios are
less favourable for the failed than the surviving firm).
The aforementioned criteria are integrated as the criteria for classification of target firms and
benchmarking firms in the IMM approach. The IMM approach combines the results of the
three measures to test the power of financial statements in signalling failure. Correct
classification in the IMM approach is assigned when any TWO of the measures correctly
classify potentially failed and nonfailed firms. To answer the research question of this study,
correct classification of the failed firm group by the IMM approach prior to actual failure is
observed and presented over the investigation period. If a firm is correctly identified as failed
prior to actual failure, this is evidence of the predictive ability of financial statement
information. This is similar to the reasoning used in previous studies.
According to the review of the literature dealing with Thailand, the three measures have not
previously been used in combination for signalling business failure. This study takes this
opportunity to explore the ability of financial statement information to predict business
failure by using the IMM approach.
Having already described the IMM approach, a discussion now turns to an additional
financial modelling technique employed in this study for testing the ability on financial ratios
in signalling financial distress.
4.8 Application of logit-based models

Logit models are estimated in this study to further evaluate and benchmark the results of
foregoing approaches on the adequacy of financial ratios in signalling financial distress. The
rationale for selecting the logit-based model, as a benchmarking measure, rather than as an
alternative model, is presented below. As indicated in the literature review, logit-based
approaches have previously been used to develop financial distress prediction models. In this
study, the models are developed as a further means of evaluating the worth of the underlying
financial data as represented in the model variable in the prediction of financial distress.
96
In financial distress prediction studies, there are extensive range of methodologies employed
such as MDA, logit analysis, and the like. The underlying rationale for developing corporate
failure prediction models is to search for the best efficaciously predictive model (Balcaen &
Ooghe 2006). In other words, most studies have aimed to find the best predictors for failure
by searching for an optimal set of explanatory variables and model specifications. As was
discussed in section 3.3, each technique possesses its own assumptions and makes significant
contribution to the field of forecasting corporate failure. From the review of Chapter 3, it is
possible to conclude that the selection of analytical technique depends mainly on the research
purpose, the researchers rationale and the influence of data constraints.
With reference to sub-section 3.3.2, the two prominent analytical techniques used in
modelling business failure prediction are MDA and logit analysis. This is consistent with
Dimitras, Zanakis and Zopounidiss (1996) observation that a large number of previous
studies employed either MDA or logit analysis as an analytical technique for signalling
business failure. To select an appropriate analytical technique for testing the contribution of
financial ratios to the signalling of financial distress, MDA and logit analysis are discussed
below.
In the context of business failure, MDA enables researchers to classify a firm into one of the
two groups, which are the failed group and the non-failed group; based on an optimal set of
financial ratios (Altman 1968). A large number of financial ratios are selected but only
financial ratios that are statistically significant are included in the final model. According to
studies using MDA technique, it is obvious that the two crucial steps are (1) testing the
significance of an optimal set of financial ratios used in the MDA function and (2)
classification (see the studies of Altman 1968; Bei & Liu 2005; and Hossari 2007). In the
field of business failure prediction, MDA was firstly applied by Altman (1968), who
developed the well-known Z score model. Also, Altmans (1968) study promotes MDA as
the most popular predictive technique for business failure between the 1960s and the 1980s
(Balcaen & Ooghe 2006).
Although MDA was used in the majority of the earlier studies (for instance, in the studies of
Altman 1968; Blum 1974; Deakin 1972; Edmister 1972; and Norton & Smith 1979), the use
of MDA was criticised because of its statistical assumptions such as a normal distribution of
97
ratios and equality of variancecovariance matrices of ratios for both groups (Balcaen &
Ooghe 2006; Coats & Fant 1993; Dimitras et al. 1999; Frank, Massy & Morrison 1965;
Goudie 1987; Hamer 1983; Ohlson 1980; and Taffler 1982). It is evident that financial ratios
are likely to be not normally distributed (Brooks 2008; Eisenbeis 1977; Sori et al. 2006; and
Watson 1990). MDA, therefore, is not an appropriate analytical technique in this study.
To overcome the MDA limitations, a less demanding statistical technique was developed to
improve prediction results (Balcaen & Ooghe 2006; and Dimitras, Zanakis & Zopounidis
1996). Thanks to sophisticated mathematical and statistical techniques, logistic regression
analysis (or logit model, which is hereafter referred to as logit model) was developed and
introduced into the field.
The logit model is derived from the cumulative logistic probability function. The logit model
overcomes the restriction of normality. The equation of logit model can be written as follows:
e zi
Pi ........(4.2)
1 e zi
e zi
For the estimation purpose, the above formulation is multiplied by zi . Therefore, the
e
estimate of the probability that Yi 1 is as follows:
1
Pi .........(4.3)
1 e zi
Where
Pi = probability of an occurrence of an event of interest
= ( i X i i ), and i are unknown parameters, X i is a set of independent variables,
and i is the disturbance term.
Source: Brooks (2008, p. 514)
The interpretation of i (usually called the regression coefficient) in the logit model is
different from that in the linear probability model because of the non-linear logistic function.
98
According to the equation above, the relationship between the set of independent variables
and the probability can be interpreted as a one unit change in X i will lead to a change in the
probability of the outcome by i F ( X i ) (Brooks 2008). The term i F ( X i ) is known as
Marginal effect (Brooks 2008; and Cameron & Trivedi 2005). For example, suppose the
following logit model is estimated from two explanatory variables.
1
Pi ( 0.5 0.2 X i 1 0.7 X i 2 )
.........( 4.4)
1 e
As can be seen, the unknown parameters include = 0.5, 1 = 0.2, and 2 = 0.7. To estimate
F ( X i ) , Brooks (2008) stated that mean values of explanatory variables are required. Again,
suppose that means of X 1 and X 2 are 1.5 and 2, respectively. The estimate of F ( X i ) will be
as follows:
1 1
Pi ( 0.5 0.2*1.5 0.7*2 )
0.90 .........(4.5)
1 e 1 e 2.2
With reference to the above assumption, it can be stated that one unit increase in X 1 will lead
to an increase in the probability of the outcome of Yi 1 by Likewise, an
increase in the probability of the outcome of Yi 1 for variable X 2 is A
positive regression coefficient ( ) tells that the explanatory variable increases the probability
of the outcome while a negative regression coefficient ( ) indicates that the explanatory
variable decreases the probability of that outcome (Suriya 2011).
Cameron and Trivedi (2005) stated that the regression coefficient ( ) can be interpreted in
terms of the odds ratio or relative risk. Odds represent the relative chance or likelihood that
different events occur (Dayton 1992). In the binary outcome dependent, there are only two
probability outcomes, that is, the event of interest is likely to happen or not. The odds of the
event of interest happening is the probability that the event will occur ( Pi ) divided by the
99
Pi
probability that it will not occur (1 Pi ) . Simply stated, the odds are the
(1 Pi )
proportion of the probability of two different events. The odds ratio is used to compare the
odds for two groups, lets say, the financially distressed group and the non-financially
distressed group in the context of financial distress studies. It is possible to infer that (1) the
odds ratio of 1 tells that both events are equally likely to happens, (2) the odds ratio of greater
than 1 indicates that the first event is more likely to happen, and (3) the odds ratio of less than
1 expresses that the happening of the event of interest is less likely.
In the logit model, the property of the outcome is mutually exclusive and exhaustive. In
practice, the event of interest is denoted as 1 while the event of the other alternative is
denoted as 0. For instance, a researcher would like to study whether companies are likely to
be financially distressed or not. In this case, the probability outcome of a financially
distressed firm is denoted as 1 ( Pr(Y 1) ) and 0 otherwise (Pr(Y 0)) . Consequently, the
logit model helps researchers predict the probability of an occurrence of an event of interest,
rather than predict the expected value of dependent variables, as is the case for MDA.
Figure 4.4 Relationship between a binomial dependent variable and an explanatory variable
in the logit model
Probability of an occurrence of an event of interest
Pr(Y 1)
Pr(Y 0) X
Note:
Pr(Y 1) = Probability of an occurrence of an event of interest
Pr(Y 0) = Probability of a non-occurrence of an event of interest
X = an explanatory variable
100
Figure 4.4 shows that the probability values of Y , or the probability outcomes, relative to an
observation on X lie within the (0,1) space. As can be seen, the best fitted regression line is
an S shape, rather than a linear line. In addition, the estimated probability values of Y lie
within the (0,1) interval due to the use of the cumulative probability function (Brooks 2008;
and Pindyck & Rubinfeld 1998). According to Figure 4.4, it can be inferred that the logit
model is an analytical technique of non-linear regression probability.
In the logit model, the relationship between binary outcomes and a set of explanatory
variables is estimated by a maximum likelihood approach (Cameron & Trivedi 2005; and
Pindyck & Rubinfeld 1998). Myung (2003) suggested that maximum likehood estimation is
an appropriate tool to estimate parameters for a conditional probability model. Likewise,
Flagg, Giroux and Wiggins Jr. (1991), Ginoglou, Agorastos and Hatzigagios (2002), and
Zavgren (1983) stated the non-linear maximum likelihood estimation is used in logit analysis
to find the regression coefficient of explanatory variables.
In maximising the likelihood equation, logarithmic transformation (or data transformation) is

required (Cameron & Trivedi 2005; Greene 2003; Manning & Munro 2007; Pindyck &
Rubinfeld 1998). The logarithmic transformation method is applied to transform the
explanatory variables and re-estimate the best fitted regression coefficients of the logit model.
Recently, the study of Sori et al. (2006) confirmed that the natural logarithmic transformation
is a more appropriate method to deal with distributional properties of financial ratios.
According to previous financial distress modelling, the two traditional principal method
designs (cross-sectional method and time-series method) are applied (Cybinski 2001). The
cross-sectional method design is used to compare characteristics of financially distressed
firms and non-financially distressed firms while the time-series method design is used to
study a sequence of data points. As a result, this study will employ the two method designs in
developing the logit models. By doing so, the financial ratios of each financially distressed
firm and non-financially distressed firm over the observation period are used in estimating
the logit models. After that, the estimated coefficients are used to calculate the probability of
financial distress of each firm in the sample. This test is called the estimation sample test,
which provides the explanatory/ex-post results. Moreover, the logit models in this study are
estimated by employing data from one sample period and the estimated coefficients are used
101
to examine the probability of financial distress in another sample period. This test is called
the out-of-sample test, which provides the predictive/ex-ante results. This notion is derived
from the study of Beaver, McNichols and Rhie (2005). The logit models in this study are
derived from the EVIEWS program.
When it comes to the issue of goodness to fit for the logit model, the EVIEWS program
provides a number of measures to examine the goodness to fit of the model such as
McFadden R-Squared, Log likelihood and the like. Pindyck and Rubinfeld (1998) suggested
that the likelihood ratio index is a preferable inferential statistical test to the R 2 measure.
Likewise, Brooks (2008) stated that the commonly reported measures of goodness to fit for
the logit model are the percentage of values correctly predicted and McFadden R-Squared. In
financial distress modelling, the widely used measure of goodness to fit is the percentage of
values correctly predicted, which is presented in terms of a classification table. Therefore,
this study will employ a classification table, as a conventional measure, to examine the
goodness to fit of the estimated logit models.
In previous financial distress studies, the logit model has been the more often used method
among other probability models (such as the linear probability model and the probit model)
since the 1980s (Balcaen & Ooghe 2006). One of the forefront users of logit analysis in the
this field is Ohlson (1980) (please refer to sub-section 3.3.2.1). The more appealing property
of the logit model is likely to be that the logit analysis has less rigorous constraints on the
estimated variables (Flagg, Giroux & Wiggins Jr. 1991, p. 72). This assumption enables
researchers in this field to overcome the normality limitation of financial ratios as potential
predictors for financial distress. According to Sub-section 3.3.2.2, the logit model has been
predominantly employed in Thai studies in relation to financial distress prediction.
Since the logit function is developed on the basis of cumulative logistic probability
distribution (Brooks 2008; and Pindyck & Rubinfeld 1998), the logit model is a preferable
method to use as a benchmarking measure in examining the ability of financial ratios in
signalling financial distress in this study.
Having discussed the benchmarking measure used in this study, the next section brings this
chapter to a conclusion.
102
4.9 Conclusion
As stated in the introduction to this chapter, the principal aims of the chapter were to describe
and justify the research methodology together with the research design, and to explain how
data was collected and analysed.
The chapter began by discussing the problem identification and formulation of research. Then
key components of research design such as the dimension of the research, the typology of the
research and research method were elaborated. This led to a discussion of sampling design.
The identification of firms as being potentially failed or nonfailed relies on a number of
criteria stated in the sample selection section. As a result, a total number of 28 firms,
consisting of 14 failed firms and 14 nonfailed firms, were used in this study. The rationale
behind their selection was discussed. Also, the underlying reasons for adopting five years as
an optimal study period were provided. Unlike previous studies, this study used quarterly
financial statements to investigate the predictive ability of financial ratios. Therefore, a 20
quarters sample period was used in this study.
Following the selection of failed and nonfailed firms, the collection of financial statement
data was required. Financial ratios are a critical part in modelling corporate failure as they are
the foundation instrument and predominant predictors for signalling business failure. The
financial ratios were generally selected based on the predictive ability they demonstrated in
previous studies. As a consequence, a total number of 44 ratios were identified as potential
predictors, from which a total of 16 ratios were selected.
The discussion then turned to how the methodology for this study was developed. The
rationale for selecting the three ratiobased analytical tools was elaborated. The three
measures included the EMS model, comparative ratio analysis, and ratio trend analysis. After
that, the discussion turned to why the IMM approach was developed. To verify the results
provided by the IMM approach, logit-based models were employed as a benchmarking
measure. The properties of logit-based models were presented. The rationale for selecting the
logit models has been described. The task of data analysis is discussed in Chapter 5.
103
Chapter 5: Data Analysis
5.1 Introduction
Since the early period of studying the prediction of business failure, financial ratios have
played a major role as a useful tool to indicate early signs of failure. The two explicit
approaches for studying failure are to detect the symptoms of failure before the failure occurs
(e.g. the studies of Fitzpatrick 1931; and Merwin 1942) and to develop a prediction model for
failure (for instance, the studies of Altman 1968; Hossari 2007; Ohlson 1980; Persons 1999;
and Shumway 2001).
To date, in the Thai literature, there has not been evidence of studies investigating the ability
of financial statements to adequately predict failure in the Thai environment in the post 1997
financial crisis period. This leads to an exploration of the failure prediction ability of
financial statements in the Thai environment. The question raised is whether or not financial
statement data can be adequately used to predict potentially failed and nonfailed firms in the
Thai context in recent years.
This chapter aims to apply the methodology outlined in the previous chapter together with the
data collected to answer the above question. The results of the descriptive statistics, the
individual application of each measure, the application of the IMM approach together with
the logit models are discussed and presented in this chapter. Finally, the results provide the
necessary empirical evidence. The results can be used as a foundation for future research
focusing on estimation of ratiobased prediction models.
This chapter is presented in seven sections as depicted in Figure 5.1. Section 5.1 outlines the
objective and the structure of the chapter. Section 5.2 provides descriptive statistics
definitions and summarises descriptive data of the sample used in this study. Section 5.3
presents the results of the EMS model. Section 5.4 presents the results of comparative ratio
analysis and then the results of ratio trend analysis are presented in Section 5.5. The results of
104
the IMM approach are presented in Section 5.6. Section 5.7 presents the results of logit-based
models. Finally, Section 5.8 brings the chapter to a conclusion.
5.1 Introduction
5.2 Description of sample used
5.2.1 Descriptive statistics

definition 5.2.3 Profile of financial
statements/ratios used
5.2.2 Profile of companies
used
5.3 Applying the EMS model and its results
5.4 Applying ratio comparative analysis and its results
5.5 Applying ratio trend analysis and its results
5.6 Applying the Integrated Multi-Measure approach and its results
5.7 Applying the logit-based model and its results
5.7.3 Discussion of logit model

5.7.1 Testing for correlation results based on the entire 20
among the 16 financial ratios quarters sample
5.7.2 Testing for normality 5.7.4 Discussion of logit model

of the 16 financial ratios results based on split sample
5.8 Conclusion
105
5.2 Description data of sample used
This section provides the definitions of descriptive statistics and summarises the
characteristics of the data sample used in this study. The definition of descriptive statistics is
discussed first (Subsection 5.2.1), followed by a profile of the companies used (Subsection
5.2.2). Finally, a profile of the financial statements/ratios used (Subsection 5.2.3) is
discussed.
5.2.1 Descriptive statistics definition

In the general sense, statistics is a set of tools and techniques that is used for
describing, organizing and interpreting information or data (Salkind 2008, p. 7).
Likewise, Neuman (2006, p. 346) states that statistics can mean a set of collected
numbers as well as a branch of applied mathematics used to manipulate and
summarize the features of numbers. In the research realm, a statistic is used as the
summary descriptor of a given variable in the sample (Davis 2005, p. 230). This
notion is consistent with Neumans (2006) perspective.
There are two common types of statistics, that is, descriptive statistics and inferential
statistics (Berenson et al. 2007; Levine, Krehbiel & Berenson 2003; Levine et al. 2008;
and Salkind 2008). Descriptive statistics are used to collect, summarise, and present a
set of data, while inferential statistics are used to draw conclusions about a
population (Berenson et al. 2007, p. 6). Neuman (2006, p. 347) notes that researchers
use descriptive statistics to describe basic patterns in the data. In other words,
descriptive statistics enable researchers to gain insight into the characteristics of the
data being studied. Descriptive statistics can be presented in either numerical form (i.e.
a raw data or a percentage frequency distribution) or graphic form (for example, using
histograms, and bar charts) (Davis 2005; Levine, Krehbiel & Berenson 2003; Neuman
2006; and Salkind 2008).
106
Levine, Krehbiel and Berenson (2003, p. 101) state that there are three major
properties that describe a set of numerical data: central tendency, variation, and shape.
The three measures of central tendency include the arithmetic mean, the median, and
the mode(Levine, Krehbiel & Berenson 2003, p. 101), while the five measures of
variation are the range, the interquartile range, the variance, the standard deviation,
and the coefficient of variation (Levine, Krehbiel & Berenson 2003, p. 109).
In this study, the mean is used as the measure of central tendency and the standard
deviation (SD) is employed as the measure of variation. The underlying reasons for
applying the mean are that (1) it takes into consideration the extreme values in a data
set, and (2) it is the most common measure used in previous studies. The mean
represents a balance point in a data set with extreme values (Berenson, Levine &
Krehbiel 2009, p. 56). Extreme values or anomalies of the financial statement data are
regarded as critical contributors to business failure. Likewise, Hossari (2006, p. 217)
states that companies that are in financial trouble are expected to have ratios that are
outofline from those of healthy firms. The reasons for adopting the SD is that the
SD, as the most practical and most commonly used measure of variation, takes data
dispersion into consideration (Levine, Krehbiel & Berenson 2003, p. 110).
Having presented the definition of descriptive statistics together with the justification
for using it in this study, the discussion now turns to exploring numerical data and the
properties of the sample set.
5.2.2 Profile of companies used

The final sample constitutes 28 companies with 14 firms in each of the two groups.
The failed group was selected based on the criteria discussed in Chapter 4. During the
study period (from 2003 to 2008), the year of most failures is 2005, in which five
companies failed from a total of 14 failed firms. There was only one occurrence of
failure in each of 2003 and 2006. Based on the most recent quarterly balance sheet
prior to the date of failure, the average asset size of failed firms was THB 3.73 billion
107
(on 23/07/09, THB 28.05 was equal to AU$ 1, according to the Bank of Thailand).
However there were significant differences in size, with the groups range of assets
varying between THB 728.30 million and THB 15.75 billion. The 14 failed firms
operated in 12 different business sectors. Of the 12 business sectors, 10 had only one
failed firm. The most frequently represented sectors were construction materials, and
food and beverages, which included two firms each. The average age of firms prior to
failure was 28 years, with a range of between 16 years and 47 years. In summary, the
group is heterogeneous in terms of year of failure, asset size, business sectors, and
firms age.
The nonfailed group was chosen based on the criteria discussed in Chapter 4, which
are:
1) Match each failed firm to a nonfailed firm using the industry code
determined by the SET.
2) Calculate the average asset size for each failed firm based on the data
availability over the investigation period.
3) Calculate the average asset size for each nonfailed firm in the same industry
and investigation period.
4) Select the nonfailed firm that has the closest asset size to that of a failed firm
as a paired counterpart.
The data for each nonfailed firm was collected from the same time period that was
used for its failed counterpart. The average asset size of the nonfailed firms was
THB 3.24 billion, with a range of between THB 717.20 million and THB 10.85
billion. The average age of firms in this group was 31 years, and ranged between 16
years and 52 years. It should be noted that the age of each nonfailed firm was
derived by subtracting the delisting date of its failed counterpart from the
establishment date of the nonfailed firm. Like the failed group, this group is not
completely homogeneous. According to the descriptive data, the average size of the
nonfailed group was smaller than that of the failed group but the nonfailed groups
age was longer than that of the failed group, as shown in Table 5.1.
108
Table 5.1 Descriptive profile of companies used
Failed group Nonfailed group
Number of firm 14 14
Average asset size* THB 3.73 billion THB 3.24 billion
Average age (years) 28 31
Number of delisted firm(s) and their counterpart(s) per year
2003 1 1
2004 2 2
2005 5 5
2006 1 1
2007 3 3
2008 2 2
Number of failed and nonfailed firm(s) based on business
sectors
1) Paper and Printing Materials 1 1
2) Machinery and Equipment 1 1
3) Packaging 1 1
4) Automotive 1 1
5) Transportation and Logistics 1 1
6) Media and Publishing 1 1
7) Tourism and Leisure 1 1
8) Construction Materials 2 2
9) Food and Beverages 2 2
10) Information and Communication Technology 1 1
11) Electronic Components 1 1
12) Fashion 1 1
Note: *1) Data based on 20quarter financial statements prior to failure and its availability,
2) On 23/07/09, THB 28.05 was equal to AU$ 1, according to the Bank of Thailand.
Source: Data analysis for this thesis, and SET (2008)
This concludes the discussion of the profile of companies used in this study. The next
task is to present a profile of the financial statements/ratios used in the three analytical
tools.
5.2.3 Profile of financial statements/ratios used

In the context of signalling business failure, financial ratios play a critical role. They
enable researchers to find symptoms and/or early warning signs of corporate failure.
In practice, if symptoms of financial distress/fragility can be detected early, corrective
measures could be implemented by management to prevent failure. Financial ratios
are derived from the financial statements of companies, and this explains why the
109
selection of companies is done first. This section discusses a profile of the financial
statements/ratios used along with the statistics obtained for them in this study.
Because the ratios are calculated from financial statements, the availability of
financial statements directly affects the derivation of the ratios. In the SET database,
12 out of the 14 pairs had financial statements available for the specified investigation
periods, which were 20 quarters prior to failure. One pair had 17 quarterly financial
statements available and another had 10 quarterly financial statements available. This
uncontrollable phenomenon is consistent with previous studies. The availability of
financial statements is shown in Table 5.2.
Table 5.2 Descriptive profile of the availability of financial statements

The availability of financial statements Number of pairs
10 quarters 1
17 quarters 1
20 quarters 12
Total 14
Source: Data analysis for this thesis
After the financial statements of the sample set were obtained, the computation of all
financial ratios used for the three analytical techniques were started. The ratios used in
the EMS model are now discussed. Following this, the selected ratios used in
comparative ratio analysis and ratio trend analysis are described.
Ratios used in the EMS model
The ratios used in the EMS model in this study are based on Altmans original four
ratio EMS model. They include working capital to total assets (WCTA), retained
earnings to total assets (RETA), earnings before interest and tax to total assets
(EBITTA), and book value of equity to total liabilities (TETL). All four ratios were
110
computed for the 14 pairs of the sample set. The descriptive statistics of these ratios
for both groups is illustrated in Table 5.3 (please refer to Appendix B for details).
Table 5.3 Descriptive profile between both groups for the EMS model
Ratios
Groups WCTA RETA EBITTA TETL
Failed Mean=0.0871 ** Mean=-0.1003 Mean=0.0153 Mean=2.8145
SD=0.1568 SD=0.6172 SD=0.0441 SD=3.0383
Nonfailed Mean=0.0037 ** Mean=-0.1050 Mean=0.0185 Mean=2.8196
SD=0.2655 SD=0.5336 SD=0.0406 SD=4.2891
Note: **Significant at p<0.05

*Significant at p<0.10
SD=Standard Deviation
Table 5.3 shows the differences between the means of the ratios for the failed and
nonfailed groups for the entire investigation period with the statistical significance at
5 per cent and 10 per cent. Mean values of the ratios refer to the arithmetic averages
of the values for the entire investigation period. To compute the mean values, all
values in a set of data are added up and then divided by the number of values in the
dataset. Of the four ratios, only the means of WCTA for failed and nonfailed firms
were significantly different at a 5 per cent significance level. This result is also true
when investigating at a 10 per cent significance level2.
Having presented the descriptive profile of the EMS models ratios, the discussion
now turns to the descriptive profile of the selected ratios used in comparative ratio
analysis and ratio trend analysis.
2
The fact that the means were not statistically different on average for the whole group over the entire
observation period should not be taken to mean that the ratios would not differ in critical periods prior to
failure.
111
Ratios used in comparative ratio analysis and ratio trend analysis
A total of 15 ratios were used in the comparative ratio analysis and ratio trend
analysis techniques. They were selected on the basis of their popularity in previous
studies. The selected ratios are classified into four groups: liquidity,
turnover/performance, leverage/solvency, and profitability. They include current
assets to current liabilities (CACL), quick assets to current liabilities (QR), sales to
total assets (STA), expenses to sales (TES), sales to inventory (SINV), total liabilities
to total assets (TLTA), total liabilities to total equity (TLTE), earnings before interest
and tax to interest (EBITI), earnings before interest and tax to sales (EBITS), net
income to sales (NIS), earnings before interest and tax to total assets (EBITTA), net
income to total assets (NITA), net income to total equity (NITE), working capital to
total assets (WCTA), and retained earnings to total assets (RETA).
Not all the selected ratios could be calculated due to missing financial items from the
financial statements of some companies during the observation period. As a result, the
problematic ratios were not used in the comparative ratio analysis and ratio trend
analysis. The rationale for ignoring the problematic ratios has already been discussed
in Chapter 4. The descriptive statistics of these ratios for both groups is illustrated in
Table 5.4 (please refer to Appendix B for details).
112
Table 5.4 Descriptive profile between both groups for the 15 ratios
Groups
Ratios Failed Nonfailed
CACL M=1.6185 ** M=2.1543 **
SD=1.1768 SD=1.9644
QR M=1.2507 ** M=1.5151 **
SD=1.0795 SD=1.6514
WCTA M=0.0871 ** M=0.0037 **
SD=0.1568 SD=0.2655
STA M=0.1859 ** M=0.1531 **
SD=0.0738 SD=0.0923
TES M=1.0133 M=0.9864
SD=0.3639 SD=0.2281
SINV M=4.9429 M=6.3383
SD=6.2897 SD=15.6251
TLTA M=0.3780 ** M=0.5221 **
SD=0.1932 SD=0.3306
TLTE M=1.0528 * M=1.3922*
SD=0.9394 SD=2.2052
EBITI M=28.8920 ** M=0.6579 **
SD=94.5910 SD=7.0329
EBITS M=0.0670 ** M=0.3784 **
SD=0.3648 SD=1.4797
NIS M=0.0383 ** M=0.3547 **
SD=0.3636 SD=1.5592
EBITTA M=0.0153 M=0.0185
SD=0.0441 SD=0.0406
NITA M=0.0102* M=0.0228*
SD=0.0432 SD=0.0970
NITE M=0.0112 ** M=0.0305 **
SD=0.0805 SD=0.0696
RETA M=-0.1003 M=-0.1050
SD=0.6172 SD=0.5336
Note: **Significant at p<0.05
*Significant at p<0.10
M=Mean
SD=Standard Deviation
Table 5.4 shows the difference between the mean values for the failed and nonfailed
groups for the entire investigation period with the statistical significance at 5 per cent
and 10 per cent. The calculation of the mean values has already been described. At a 5
per cent significance level, nine ratios displayed a statistically significant difference
between the mean values for the failed and nonfailed groups. In addition, out of the
15 ratios, 11 ratios demonstrated a significant difference between the mean values for
both groups at a 10 per cent significance level. This illustrates a priori indication of
113
the ability of the selected ratios to identify financially distressed firms. For the
remaining ratios where the difference is not significant, the expected relationship
between the ratios was found to display a similar pattern. However, both groups may
have been affected by the small scale of the data set, the impact of industry
differences, and the averaging of values in the data set for the entire investigation
period.
In terms of liquidity, two mean ratios (CACL and QR) of the failed group are less
than those of the nonfailed group. According to Gibson and Frishkoff (1986), CACL
is more informative than WCTA when measuring the ability of a company to meet its
current obligations. Means of CACL, QR, and WCTA are significantly different.
Concerning turnover/performance, the failed group tends to enjoy higher earnings
generating power of total assets. Only the mean of STA is significantly different.
When it comes to leverage/solvency, the failed group was less leveraged than the
nonfailed group. As can be seen, the means of TLTA and TLTE for the failed group
were lower than those of the nonfailed group.
Last but not least, four mean ratios (EBITS, NIS, NITA, and NITE) of profitability for
the failed group are less than those of the nonfailed group. Given that, the failed
group tends to experience a lower ability to generate profit from its operations than
the nonfailed group. Means of EBITS, NIS, NITA and NITE are significantly
different.
In summary, according to the descriptive profile of the 15 ratios, it can be inferred

that the failed group during the specified period was likely to have less liquidity, as
shown by lower CACL and QR, and lower profitability, as illustrated by lower EBITS,
NIS, NITA, and NITE. For the remaining ratios where the difference was not
significant, the expected relationship between the ratios was similar. It appears that
the failed group was less financially leveraged, as demonstrated by lower TLTA,
TLTE, and EBITI. In this regard, the average overall economic condition of the failed
114
group deteriorated over the specified period. The financial characteristics of the failed
group would be expected to deteriorate before the actual failure happened.
Having presented a description of the sample used in this study, the discussion now
turns to the application of each individual ratiobased measure. The application of the
EMS model and the results obtained are presented first, followed by the application of
the two traditional measures and their results.
5.3 Applying the EMS model and its results

This section demonstrates how the EMS model can be used to examine the predictive power
of financial statement information. The EMS model is selected as a proxy of sophisticated
ratiobased models. The underlying reason for choosing this model was described earlier.
The model was developed and modified by Altman in the mid1990s (Altman & Hotchkiss
2006). In this study, the critical score was determined. The rationale for using the critical
score has already been discussed in Sub-section 4.6.1. The following is the formula for the
EMS function together with the critical score applied for this study.
EMS 3.25 6.56 X 1 3.26 X 2 6.72 X 3 1.05 X 4 ....... (5.1)
Where:
X1 =Working Capital/Total Assets
X2 =Retained Earnings/Total Assets
X3 =Earnings before Interest and Taxes/Total Assets
X4 =Book Value of Equity/Total Liabilities
EMS=Overall score
115
Safe zone=5.85 or more
Distressed zone<5.85
Source: Developed for this study
If the computed score of a company is equal to or more than the critical score (5.85), the
company is identified as nonfailed. On the other hand, a company is identified as failed
when their computed score is less than the critical score. In this regard, correct classification
of the model can be explicitly determined by comparing the actual status with the predicted
status of each company in both groups. If the model has predictive power, then it should be
able to correctly classify the firm in the quarter immediately before actual failure. Also, it
would be expected that indications of likely failure would occur in earlier quarters. This is
similar to the findings in the literature that the predictive accuracy of the model increases
when the period of failure approaches.
In the context of predicting financial distress, both Type I and Type II error are frequently
referred to. Employing the Type I and Type II errors is another way of measuring the
predictive ability of financial distress prediction models. A Type I error is a misclassification
of a failed firm as a non-failed firm (as represented in red colour in Table 5.5) while a Type II
error is a misclassification of a non-failed firm as a failed firm (as presented in red colour in
Table 5.6). After the criteria of the model were set and the required ratios were collected, the
calculation of EMS scores for both the failed group and the nonfailed group was conducted.
The results of the EMS model application for the failed group are shown in Table 5.5.
116
Table 5.5 Results of the EMS model application for the failed group
Prior to failure
Failed Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
firm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
28.78 24.78 12.49 20.13 21.32 26.19 13.44 14.46 10.83 9.24 10.21 8.74 8.63 7.59 6.94 6.63 6.11 5.43 4.24 3.45
PPPC
4.08 3.93 4.04 5.13 5.93 6.53 6.88 6.89 7.13 7.36 7.89 8.61 9.79 9.29 10.35 10.14 10.91 10.08 11.53 10.82
TLI
4.98 0.75 0.28 -0.48 -0.96 -0.89 -1.06 -2.32 -3.12 -2.70 -3.05 -3.30 -0.99 -0.93 -1.22 -1.40 -1.76 -1.20 -1.05 -0.85
UBC
12.63 12.80 12.83 12.86 11.91 11.84 12.33 12.15 10.85 10.37 10.20 10.17 10.15 9.95 9.66 9.41 9.04 8.06 7.47 7.39
ASTL
3.36 3.34 3.79 3.44 3.63 3.68 3.87 3.56 4.42 4.34 4.39 4.60 4.44 4.30 3.82 4.57 4.69 5.39 4.27 4.94
UFC
6.03 -3.45 5.49 5.06 6.71 6.52 5.56 4.91 5.05 4.80 4.62 4.51 4.57 4.42 4.14 3.86 3.87 4.08 3.57 3.24
UCOM
4.63 5.10 5.54 8.83 8.28 7.89 9.68 8.49 7.04 6.20 6.38 6.48 5.01 4.54 4.31 3.80 4.77 4.40 4.69 5.04
HT
9.71 8.89 8.50 8.52 10.04 10.29 11.35 11.15 9.37 8.84 9.08 9.71 8.87 8.14 8.06 7.37 6.91 6.36 7.36 8.54
APSP
11.66 10.51 7.57 8.62 7.38 7.11 7.07 6.65 4.74 3.81 3.63 3.67 0.09 1.23 1.27 1.95 1.25 1.74 1.88 2.24
UTL
13.51 12.99 11.97 13.98 9.89 9.59 8.38 8.06 9.74 9.46 8.37 7.78 8.36 7.87 7.45 7.23 8.16 7.84 6.17 6.11
RHC
5.03 4.81 4.69 4.53 3.92 3.54 3.94 4.81 4.87 4.90 4.69 4.36 4.17 2.50 2.98 3.76 4.15 4.67 5.53 5.66
MFG
1.74 3.17 0.94 2.89 3.28 4.55 3.11 5.41 6.02 4.53 4.63 6.46 -3.04 -0.36 2.68 3.77 4.04 3.83 2.82 4.20
MPT
17.01 17.22 15.64 15.47 14.46 14.10 14.01 12.22 12.33 12.22 11.32 11.13 10.10 9.15 8.08 6.79 6.01
TGP
4.78 3.17 4.24 4.35 4.49 4.54 4.78 5.45 5.81 4.10
AF
Note: Q=Quarter
Green colour=Correct classification
Red colour=Misclassification
Table 5.5 illustrates the EMS results for the 14 failed firms. A green square represents a
correct classification (where the EMS score computed for the company leads to its
classification as failed) whilst a red square represents a misclassification (where the EMS
score computed for the company does not classify it as failed). Of the 14 failed firms, 4
firms (UBC, UFC, MFG, and AF) were correctly classified over the whole study period and
four firms (ASTL, APSP, RHC, and TGP) were erroneously classified over the specified
period. In case of UCOM and MPT, only three quarters and two quarters, respectively, were
misclassified. In addition, the model could correctly classify 11 quarters of HT and 12
quarters for UTL. Also, three quarters for PPPC and four quarters for TLI were correctly
classified.
117
As stated earlier in the context of the predictive power of financial statement data, recent
correct classification prior to failure is expected. This notion is similar to the notion in the
previous study that predictive accuracy of the model increases when the period of failure
approaches. According to Table 5.5, there were seven firms for which there was a recent
correct classification prior to failure. The model correctly identified these four failed firms
over the entire period before failure for UBC, UFC, MFG, and AF; for the eight quarters just
before failure for MPT; for the four quarters just before failure for TLI; and for the three
quarters just before failure for HT. Therefore, the successful prediction rate of the EMS
model using financial statement data was 50 per cent (7 out of the 14 failed firms).
If a cutoff score of 4.15 (the cutoff point for the original distress zone) had been used, 4
out of 14 failed firms (ASTL, APSP, RHC, and TGP) would have been incorrectly classified
over the whole study period. However, the following correct classifications would have been
made: 19 quarters for UBC, 13 quarters for MPT, and 11 quarters for UTL. If it had used a
cutoff score of 4.15, the model would have correctly classified nine quarters for UFC and
seven quarters for UCOM and MFG. Also, three quarters for TLI, two quarters for AF, and
one quarter for PPPC and HT were correctly classified. Finally, there were only three failed
firms (TLI, UFC, and MPT) for which a recent correct classification before the failure could
be detected. This may result in a misleading estimate of the model accuracy because failed
firms could have a score of more than 4.15 and less than 5.85 (grey zone). Another principal
rationale for not adopting 4.15 as a cutoff score is that the cost of incorrectly classifying of
failed firms as nonfailed firms is higher than the cost of incorrectly classifying nonfailed
firms as failed firms (Balcaen & Ooghe 2004).
Having presented the successful prediction rate for failed firms, the discussion now turns to
the successful rate for nonfailed firms. The results of the EMS model application for the
nonfailed group are depicted in Table 5.6.
118
Table 5.6 Results of the EMS model application for the nonfailed group
Prior to failure
Non- Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
Failed
firm 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
5.93 5.61 5.80 5.72 5.83 5.71 4.77 1.05 0.39 0.30 -0.12 -0.20 -0.06 -0.08 -0.04 -0.20 -0.17 0.04 2.44 2.63
TCP
5.67 5.61 4.84 2.90 2.59 4.89 4.74 -6.71 -6.30 -5.63 -6.83 1.97 2.30 3.22 3.23 3.39 3.62 4.05 4.21 4.66
TCJ
21.32 21.88 22.13 7.94 21.64 18.88 27.96 10.91 13.85 29.46 24.40 7.57 19.64 20.85 19.82 7.63 24.37 23.81 20.95 8.05
BEC
7.58 7.30 7.02 6.77 6.55 6.30 6.24 4.67 3.12 2.98 3.20 3.29 3.73 3.75 3.65 3.28 -2.73 -2.23 -1.95 -1.46
RCI
9.47 16.13 13.85 12.82 13.10 20.95 19.75 13.80 25.11 23.65 27.65 15.20 14.18 11.56 11.80 11.54 10.72 10.13 10.12 6.94
TWFP
8.01 9.50 7.34 7.35 7.30 7.46 7.13 7.00 6.59 6.85 6.86 6.28 6.07 6.22 5.22 4.38 0.38 0.48 0.16 -0.10
JAS
13.18 12.63 11.69 13.47 12.48 12.08 10.28 10.41 9.24 9.71 9.27 9.59 9.20 8.35 7.95 8.21 8.35 8.25 7.82 7.79
TPCORP
4.81 4.82 4.65 4.50 2.15 2.23 1.74 2.71 1.70 1.43 1.63 1.43 2.28 2.28 2.40 1.48 -0.17 -2.20 -2.24 -2.06
NEP
4.46 3.82 3.47 3.45 3.04 3.90 3.53 3.30 2.01 2.12 1.55 1.29 0.71 1.20 1.76 2.28 0.11 2.71 2.03 2.28
TTA
13.36 13.14 14.34 11.73 13.38 12.19 13.44 11.97 12.82 11.89 13.18 11.92 13.56 12.44 11.52 13.44 10.27 9.95 10.00 9.86
ROH
3.82 3.77 3.54 3.62 2.80 2.66 2.45 2.38 1.81 1.68 1.55 1.56 1.39 1.05 0.92 0.64 0.47 1.35 0.87 0.60
UFM
3.28 3.62 3.61 3.27 4.58 4.44 4.73 5.32 5.90 5.62 5.01 5.04 5.77 5.49 5.35 5.64 4.80 4.64 4.45 4.72
KCE
6.46 7.03 7.75 7.51 4.31 -1.77 -1.90 -2.04 -1.75 -2.10 -3.34 -1.47 -1.63 -1.86 -1.15 -0.56 -0.22
UMI
-3.03 -2.37 -1.56 -0.96 -0.75 0.14 0.62 0.84 1.51 1.04
SMC
Note: Q=Quarter
A green square represents a correct classification (where the EMS score computed for the
company leads to its classification as non-failed) whilst a red square represents a
misclassification (where the EMS score computed for the company does not classify it as
non-failed). Table 5.6 shows the EMS results for the 14 nonfailed firms. Of the 14 non
failed firms, 4 firms (BEC, TWFP, TPCORP, and ROH) were correctly classified over the
whole study period. Five firms (TCJ, NEP, TTA, UFM, and SMC) were incorrectly classified
over the specified period. For seven quarters, RCI was correctly classified and for 14 quarters
JAS was correctly classified. The model also correctly classified one quarter for TCP and
KCE and four quarters for UMI.
119
As was the case for the observation of the failed group, recent correct classifications of the
nonfailed group is observed. As can be seen, the model correctly classified: BEC, TWFP,
TPCORP, and ROH in 20 quarters; JAS in the most recent 14 quarters; RCI in the most
recent seven quarters; UMI in the most recent four quarters; and TCP in the most recent
quarter. Out of the 12 business sectors, 10 sectors were found to have correct classifications
in common. As a result, the successful rate for prediction of nonfailed firms was 57 per cent
(8 out of the 14 nonfailed firms).
If 4.15 had been used as the cutoff score to this group, four nonfailed firms (BEC, TWFP,
TPCORP, and ROH) would have been correctly classified over the whole study period. The
model would have correctly classified 16 quarters for JAS and KCE, eight quarters for RCI,
and seven quarters for TCP and TCJ. Also, five quarters for UMT, four quarters for NEP and
one quarter for TTA would have been correctly classified. Nonetheless, the model would
have incorrectly classified UFM and SMC if the 4.15 cutoff had been used. Finally, 11 out
of the 14 nonfailed firms would have showed recent correct classification (successful rate of
79 per cent). The reason for not employing the 4.15 cutoff has already been discussed.
To ensure that the investigation of predictive ability of financial statements is performed

rigorously, triangulation was adopted in this study. This led to employing the traditionally
frequentlyused analytical measures for financial statements analysis. They are comparative
ratio analysis and ratio trend analysis. The next task was to apply ratio comparative measure
and observe the results.
5.4 Applying comparative ratio analysis and its results

The analysis enables researchers to examine the strengths and weaknesses of a companys
financial performance by comparing a companys financial ratios to those of its peers.
Generally, the potential shortcomings of using this technique are the effects of industry/sector,
time (for example, the investigation period), and company size. The analysis is more
meaningful if the effect of these factors are eliminated (Brigham & Ehrhardt 2008; Gibson &
Frishkoff 1986; and Palat 1989). In the context of signalling business failure, Beaver (1966)
120
states that such effects can be removed by utilising the pair sample technique, which is
adopted in this study. The technique has been found to be a useful tool in identifying
deficiencies of the failing firms (see for example, the studies of Fitzpatrick 1931; and Merwin
1942).
This section is designed to show how comparative ratio analysis can be used to examine the
predictive power of financial statements. The ratios for each failed firms were compared and
contrasted to those of its nonfailed counterpart over the investigation period. The ratios were
compared to detect signs of relative strength or weakness. Like the analysis performed for the
EMS model, the incidence of correct classification during the investigation period was
observed. In this technique, a correct classification was assigned when more than half of the
ratios in each pair correctly indicated a failed or nonfailed firm. After that, the incidence of
recent correct classifications prior to failure, which represents the predictive ability of
financial statement data, was observed. The results of applying comparative ratio analysis to
the failed firms are illustrated in Table 5.7.
121
Table 5.7 Results of the comparative ratio analysis application for the failed group
Prior to failure
firm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
PPPC
TLI
UBC
ASTL
UFC
UCOM
HT
APSP
UTL
RHC
MFG
MPT
TGP
AF
Note: Q=Quarter
A correct classification of a failing firm is assigned to quarters in which more than half of its
financial ratios indicate relative weakness in comparison to its paired firm, and is represented
by the colour green. Table 5.7 shows the results of comparative ratio analysis of the 14 failed
firms. Of the 14 failed firms, PPPC and TGP were erroneously classified over the whole
study period. HT was correctly classified over the specified period. UBC and UFC were
correctly classified over 19 quarters while UCOM and RHC were correctly classified over 15
quarters. MPT was correctly classified over 14 quarters and UTL was correctly classified
over 11 quarters. Also, MFG was correctly classified in eight quarters and TLI was correctly
classified for seven quarters. Finally, both ASTL and APSP were correctly classified for three
quarters and AF was correctly classified for two quarters.
122
To gain a further indication of the predictive power of financial statement data, the recent
classification of the failed firms prior to failure was observed. There were eight firms for
which correct recent classification before failure was detected. The measure correctly
classified HT for 20 quarters before failure; UFC for the 19 most recent quarters; RHC for the
14 most recent quarters before failure; TLI for the seven most recent quarters before failure;
UCOM for the four most recent quarters before failure; UBC and ASTL for the three most
recent quarters before failure; and MPT for the most recent quarter before failure (and in a
number of earlier quarters). Hence, the success rate of predictive ability of comparing
financial ratios between the failed firms and the benchmarking firms was 57 per cent (8 out of
the 14 failed firms).
Next, the results of the application of the comparative ratio analysis for the nonfailed group
are presented. The results of the comparative ratio analysis application for the nonfailed
group are shown in Table 5.8.
123
Table 5.8 Results of the comparative ratio analysis application for the nonfailed group
Prior to failure
Non Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
Failed
firm 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
TCP
TCJ
BEC
RCI
TWFP
JAS
TPCORP
NEP
TTA
ROH
UFM
KCE
UMI
SMC
Note: Q=Quarter
A correct classification of a non-failed firm is assigned to quarters in which more than half of
its financial ratios indicate relative strength in comparison to its paired firm, and is
represented by the colour green. The identical pattern of classification for both groups as
shown in Table 5.7 and Table 5.8 is expected. This is because the same ratio of each pair (a
failed firm and a non-failed firm) are compared and contrasted. If a ratio correctly indicates
the sign of weakness/deterioration for a failed firm in a pair, it means the ratio can detect a
strength/improvement sign of a non-failed mate. Table 5.8 illustrates the results of
comparative ratio analysis for the 14 nonfailed firms. Of the 14 nonfailed firms, TCP and
UMI were erroneously classified over the whole study period. TPCORP was correctly
classified over the whole period. BEC and TWFP were correctly classified over 19 quarters
124
while JAS and ROH were correctly classified over 15 quarters. KCE was correctly classified
for 14 quarters; TTA for 11 quarters; UFM for eight quarters; and TCJ for seven quarters.
Lastly, RIC and NEP were both correctly classified for three quarters and SMC was correctly
classified for two quarters.
There were eight surviving firms for which the recent correct classification was detected. The
measure correctly classified TPCORP for 20 quarters; TWFP for the 19 most recent quarters;
ROH for the 14 most recent quarters; TCJ for the most recent seven quarters; JAS for the
most recent four quarters; BEC and RIC for the most recent three quarters; and KCE for the
most recent quarter (and in a number of earlier quarters). Out of the 12 business sectors, 8
sectors were found to have correct classification in common. There were two sectors
(packaging, and transportation and logistics) for which misclassification occurred when using
the EMS score model and when using the comparative ratio analysis. This characteristic will
be further investigated when discussing the next analytical technique. Thus, the successful
classification rate for the nonfailed firms was 57 per cent (8 out of the 14 nonfailed firms).
Having presented the results of the two measures used in this study, the discussion now turns
to ratio trend analysis application and results.
5.5 Applying ratio trend analysis and its results

Ratios trend analysis is another widely used tool for financial statement analysis (Brigham &
Ehrhardt 2008; Gibson & Frishkoff 1986; and Temte 2004). Brigham and Ehrhardt (2008)
claim that this technique is as important as ratio analysis in the context of financial statement
analysis. Also, the technique is more powerful when applied with other methods (Firminger
2003). Both academics and professionals agree that one can observe the financial position of
a company by using historical ratios and looking at trends of the ratios. To do so, as
Brighham and Ehrhardt (2008) point out, the values of ratios are simply plotted over the
study period.
125
Like comparative ratio analysis, trends in the ratios of a failed firm are compared and
contrasted to those of a benchmarking firm. In this regard, the value of each ratio in the
earliest quarter prior to failure is used as a base figure. The values of the ratios in the
subsequent quarters are compared to the base figure in order to examine the trend over the
investigation period. If the same trend is shared by both firms, the rate of increase and/or
decrease is applied to differentiate between the target firms and the benchmarking firms. The
15 ratios used in this measure are the same as those used in the comparative ratio analysis
measure.
Like the two preceding sections, this section illustrates how ratio trend analysis can be used
to examine the predictive power of financial statements. The results of applying ratio trend
analysis to the failed group are shown in Table 5.9.
126
Table 5.9 Results of the ratio trend analysis application for the failed group
Prior to failure
firm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
PPPC
TLI
UBC
ASTL
UFC
UCOM
HT
APSP
UTL
RHC
MFG
MPT
TGP
AF
Note: Q=Quarter
Yellow colour=base quarter
A correct classification is assigned for a failed firm in quarters for which the majority of its
financial exhibited trends that were less favourable than that of its paired firm, and is
indicated by a green coloured square. Table 5.9 reveals the results of ratio trend analysis for
the 14 failed firms. Of the 14 failed firms, three firms (UFC, APSP, and MFG) were correctly
classified over the whole study period while only TGP was erroneously classified over the
specified period. Over 17 quarters for RHC and 16 quarters for TLI were correctly classified.
The measure also correctly classified 15 quarters for UCOM and 12 quarters for ASTL. Eight
quarters for MPT and seven quarters for UTL were correctly classified. Finally, the measure
correctly classified five quarters for PPPC; four quarters for UBC; and three quarters for HT
and AF.
127
There were eight firms that revealed recent correct classification before failure. The measure
correctly classified for all 20 quarters for UFC, APSP, and MFG; for the most recent 14
quarters before failure for RHC; the most recent 11 quarters before failure for ASTL; the
most recent 10 quarters before failure for TLI; the most recent four quarters before failure for
UCOM; and one quarter before failure for PPPC. Consequently, the successful rate of
predictive ability was 57 per cent (8 out of the 14 failed firms).
Next, the application of ratio trend analysis for the nonfailed group is presented. The results
of the ratio trend analysis application for the nonfailed group are shown in Table 5.10.
Table 5.10 Results of the ratio trend analysis application for the nonfailed group
Prior to failure
Non Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
Failed
firm 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
TCP
TCJ
BEC
RCI
TWFP
JAS
TPCORP
NEP
TTA
ROH
UFM
KCE
UMI
SMC
Note: Q=Quarter
128
A correct classification is assigned for a non-failed firm in quarters for which the majority of
its financial ratios exhibited trends that were more favourable than that of its paired firm, and
is indicated by a green coloured square. The identical pattern of classification for both groups
as shown in Table 5.9 and Table 5.10 is expected. This is because the same ratio of each pair
(a failed firm and a non-failed firm) are compared and contrasted. If a ratio correctly indicates
the sign of weakness/deterioration for a failed firm in a pair, it means the ratio can detect a
strength/improvement sign of a non-failed mate. Table 5.10 shows the results of ratio trend
analysis for the 14 nonfailed firms. Of the 14 nonfailed firms, three firms (TWFP, NEP,
and UFM) were correctly classified over the whole study period. Only UMI was incorrectly
classified over the specified period. Over 17 quarters for ROH and 16 quarters for TCJ were
correctly classified by this measure. In addition, the measure correctly classified 15 quarters
for JAS and 12 quarters for RCI. Also, eight quarters for KCE and seven quarters for TTA
were correctly classified. Five quarters for TCP, four quarters for BEC, and three quarters for
TPCORP and SMC were correctly classified. The pattern of the misclassification of the non
failed firms is similar to that of the failed firms.
Concerning recent correct classification before failure, the measure can discriminate eight
surviving firms from their failed counterparts. The measure correctly classified over the
whole study period for TWFP, NEP, and UFM; the most recent 14 quarters for ROH; the
most recent 11 quarters for RCI; the most recent 10 quarters for TCJ; the most recent four
quarters for JAS; and one quarter for TCP. Out of the 12 business sectors, 7 sectors were
found to have correct classifications in common. There is only one sector (transportation and
logistics) for which misclassification was common across all three measures. Since a similar
pattern of correct classification between the two groups were observed, the successful
classification rate for the nonfailed group accounts for 57 per cent (8 out of the 14 non
failed firms).
The successful classification rate for each measure, which demonstrates the predictive ability
of financial statements, is almost identical. Table 5.11 shows the successful classification rate
for each measure.
129
Table 5.11 Successful classification rate for each measure
EMS model 50% 57%
Comparative ratio analysis 57% 57%
Ratio trend analysis 57% 57%
Table 5.11 concludes the application together with the result for each measure used in this
study. Now the discussion turns to the Integrated MultiMeasure approach application and its
results.
5.6 Applying the Integrated MultiMeasure approach and its results

As mentioned earlier, the adoption of the Integrated MultiMeasure (IMM) approach aims to
thoroughly examine the power of financial statements in predicting financial distress. By
adopting the IMM approach as well as employing quarterly financial statement data, this
study attempts to provide initial evidence on the ability of financial statements to signal
failure in the Thailand context in recent years.
The introduction of the IMM approach in this study is motivated by the notion of multi
method design. Using a multimethod design enables researchers to gain more indepth
detail and to triangulate the findings (Easterby-Smith, Thorpe & Lowe 1991; and Gable
1994). The IMM approach uses the EMS model, comparative ratio analysis, and ratio trend
analysis. It represents a combination of a sophisticated ratiobased prediction model and two
traditional analytical techniques for financial statement analysis. The distinguishing
characteristics of the three measures have been provided above. The criteria used in each of
the measures in the IMM approach are similar to those used in the individual measures,
which have been discussed in Chapter 4. The IMM approach combines the results of the three
measures used to test the power of financial statements in signalling failure. In other words, if
any one of the three measures correctly classifies the failed and nonfailed firms, this is
adopted as the classification of the IMM approach. Table 5.12 depicts the results of the IMM
approach application for the failed group.
130
Table 5.12 Results of the IMM approach application for the failed group
Prior to failure
Failed measure Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
firm 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
EMS
PPPC CRA
RTA
EMS
TLI CRA
RTA
EMS
UBC CRA
RTA
EMS
ASTL CRA
RTA
EMS
UFC CRA
RTA
EMS
UCOM CRA
RTA
EMS
HT CRA
RTA
EMS
APSP CRA
RTA
EMS
UTL CRA
RTA
EMS
RHC CRA
RTA
EMS
MFG CRA
RTA
EMS
MPT CRA
RTA
EMS
TGP CRA
RTA
EMS
AF CRA
RTA
Note: Q=Quarter
EMS=EMS model
CRA=Comparative Ratio analysis
RTA=Ratio Trend analysis
There are a number of characteristics that can be identified from Table 5.12. With the
exception of TGP, early warning signs of financial distress could be detected for the failed
131
group. The early warning sign for each individual measure is different. For instance, the EMS
model, which is a sophisticated ratiobased measure, could not detect symptoms of financial
distress over the study period for APSP but ratio trend analysis, which is a traditional ratio
based measure, could indicate such symptoms over the corresponding period. In this case, if a
practitioner depended exclusively on the EMS model, they might be misled and form the
opinion that the overall economic condition of APSP was good. This implies that
practitioners should not rely solely on one measure to detect early warning signs for financial
distress. In other words, the IMM approach enables the practitioners to see a boarder
dimension of potential financial distress.
Another characteristic is the ability of the selected measures in the IMM approach to
distinguish the failed firms from the nonfailed firms. There were nine failed firms for which
all three measures in the IMM approach provided correct classification. They included TLI,
UBC, UFC, UCOM, HT, UTL, MFG, MPT, and AF. In addition, there were four failed firms
for which at least two of the three measures gave correct classification. They included PPPC,
ASTL, APSP, and RHC. It is reasonable to conclude that the IMM approach is a useful tool
for triangulation when studying financial distress.
Another explicit characteristic from Table 5.12 is that quarterly financial statement data can
provide indications of unfavourable financial characteristics as soon as they begin. The shifts
are considered one of the early warning indications of problems in a firms financial position.
To elaborate this notion, ASTL is a good sample. The unfavourable shifts of financial
characteristics of ASTL, represented by the green colour, appear in several periods during the
14 quarters before failure. However, the shifts occur more frequently when the period of
failure approaches. There were 11 unfavourable quarters illustrating that financial conditions
deteriorated. The early detection of the shifts could be useful to management to promptly take
corrective measures.
Another characteristic emerging from Table 5.12 is the recent correct classification before
failure by the IMM approach which demonstrates the power of the financial statements in
132
predicting financial distress. As mentioned earlier, that any one of the measures is capable of
correctly identifying the failed firms and nonfailed firms is used as the basis of the IMM
approach. There were 12 failed firms for which the IMM approach provided the recent
correct classification before failure. They included PPPC, TLI, UBC, ASTL, UFC, UCOM,
HT, APSP, RHC, MFG, MPT, and AF. Out of the 12 failed firms, there were nine failed
firms for which at least two out of the three measures in the IMM approach provided a recent
correct classification. The success rate for the IMM approach in identifying the failed group
was 64 per cent (9 out of the 14 failed firms). For the target group, the IMM approach helped
reveal a broader dimension of the ability of financial statements in signalling failure. As can
be seen, the successful classification rate increased significantly to 64 per cent (or nearly 2/3)
from 50 per cent (the lowest rate based on the individual measure approach).
Having presented the application and results of the IMM approach for the failed group, the
discussion now turns to the application together with its results of the corresponding
approach for the nonfailed group. Table 5.13 shows the results of the IMM approach
application for the nonfailed group.
133
Table 5.13 Results of the IMM approach application for the nonfailed group
Prior to failure
Non measure Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
failed 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
firm
EMS
TCP CRA
RTA
EMS
TCJ CRA
RTA
EMS
BEC CRA
RTA
EMS
RCI CRA
RTA
EMS
TWFP CRA
RTA
EMS
JAS CRA
RTA
EMS
TPCORP CRA
RTA
EMS
NEP CRA
RTA
EMS
TTA CRA
RTA
EMS
ROH CRA
RTA
EMS
UFM CRA
RTA
EMS
KCE CRA
RTA
EMS
UMI CRA
RTA
EMS
SMC CRA
RTA
Note: Q=Quarter
EMS=EMS model
CRA=Comparative Ratio analysis
RTA=Ratio Trend analysis
According to the criteria for financial distress classification (please refer to Section 4.7, page
96), a financial distress classification is adopted for any firm when any TWO of the three
134
measures criteria are satisfied (for both failed firms and non-failed firms). From Table 5.13,
there were 12 nonfailed firms that demonstrated a recent correct classification. They
included TCP, TCJ, BEC, RCI, TWFP, JAS, TPCORP, NEP, ROH, UFM, KCE, and UMI.
Out of the 12 nonfailed firms, there were 8 nonfailed firms for which two measures in the
IMM approach gave a recent correct classification, while all three measures gave a recent
correct classification for the remaining 4 nonfailed firms. The success rate for the IMM
approach in identifying the nonfailed group was 57 per cent (8 out of the 14 failed firms).
According to Table 5.12 and Table 5.13, there are a number of characteristics that can be
identified. The EMS measure shows non-identical successful classification pattern for both
groups. It is due to the fact that a company is identified as failed when their computed score
is less than the critical score (5.85), otherwise it is identified as non-failed. In this regard,
correct classification of the model can be explicitly determined by comparing the actual
status with the predicted status. Consequently, an identical successful classification pattern
for both failed and non-failed firms is not expected.
Unlike the successful classification pattern of the EMS measure, the pattern of comparative
ratio analysis and ratio trend analysis is identical for both groups. It is because (1), for
comparative ratio analysis, a correct classification of a failing firm (or a non-failed firm) is
assigned to quarters in which more than half of its financial ratios indicate relative weakness
(or relative strength) in comparison to its paired firm, and is represented by the colour green,
and (2), for ratio trend analysis, a correct classification is assigned for a failed firm (or a non-
failed firm) in quarters for which the majority of its financial ratios exhibits trends that were
less favourable (or more favourable) than that of its paired firm and is indicated by a green
coloured square. As a result, an identical successful classification pattern is expected for the
two measures.
Another characteristic is that the IMM approach provided a better successful classification
rate for the failed group than those provided by each individual measure. Table 5.14 shows
the successful classification rate for the IMM approach.
135
Table 5.14 Successful classification rate for the IMM approach
The IMM approach 64% 57%
As one can see, using an individual measure to investigate the ability of financial statements
to predict failure might be unreliable and the result is likely to mislead. This is consistent
with the literature which suggests that no specific technique is dominant over the others. On
the other hand, applying the IMM approach enables researchers to thoroughly examine the
predictive ability of financial statement information.
Again, the focal point of adopting the IMM approach is to explore whether financial
statements could adequately signal an occurrence of failure during the specified period in
Thailand. Consequently, the result is rather promising. Note that the ratios used in the two
traditional measures were selected as representatives of potential indicators for failure rather
than of the optimal set of the best predictors. The selection is based on their popularity in
previous studies. Given this, the finding could be used as a foundation for future research to
find the best predictors of failure by developing a discriminant model and comparing the
predictive accuracy of the model.
5.7 Applying the logit-based model and its results

As mentioned earlier, this study attempts to examine evidence of the ability of financial
statements to signal failure in the Thailand context in recent years. Besides the
aforementioned approach, utilising the logit model together with quarterly financial statement
data will help thoroughly examine the power of financial statements in signalling financial
distress. Simply stated, this section provides a triangulation perspective on the ability of
financial ratios in signalling financial distress. Details of the logit model were discussed in
Chapter 4.
136
With reference to the sample set, there are 28 firms comprising 14 financially distressed firms
and 14 non-financially distressed firms. Also the observation period is 20 quarters but
financial ratios of some financially distressed firms are not available for this long. This
explains why the sample size gradually decreases to 13 financially distressed firms in quarter
11 from 14 financially distressed firms in quarter 10 and 12 financially distressed firms in
quarter 18 from 13 financially distressed firms in quarter17. The selected 16 financial ratios,
including Altmans Z score 4 ratios (please refer to Table 4.7), are used as a starting point.
The selection criteria for these ratios have been already discussed in Section 4.6.
The logit model in this study is estimated by using the EVIEWS program. Not only does this
study estimate the logit model based on the total 16 financial ratios, but also a number of the
logit models based on selected financial ratios out of the 16 ratios are estimated. In the logit
model context, the dependent variable is either one if a firm is financially distressed or
zero if it is not. A pooled data set of all 28 firms financial ratios during the 20 quarters
observation period is used in an estimation of the logit model. As a result, the total number of
observations is 558 (please refer to Figure 5.2 below). Descriptive statistics between
financially distressed group and non-financially distressed group have been previously
presented in Section 5.2. Before estimating the logit model, both correlation tests and
normality tests among the 16 ratios are performed.
5.7.1 Testing for correlation among the 16 financial ratios

As mentioned earlier, the 16 financial ratios used can be classified into four
categories, that is, liquidity group, turnover group, leverage group, and profitability
group. Since the results of correlation testing either by the specified groups or by the
entire 16 ratios are quite identical, the correlation testing by the whole 16 ratios are
presented in order to display both intra-group correlation and inter-group correlation.
Table 5.15 shows the correlation testing among the 16 financial ratios.
137
Table 5.15 Testing for correlation among the 16 financial ratios
CACL QR WCTA STA TES SINV TLTA TLTE EBITI TETL EBITS NIS EBITTA NITA NITE RETA
CACL 1 0.866029 0.624787 -0.06238 -0.03874 0.069614 -0.22326 -0.10692 0.17468 0.585299 0.051268 0.069105 0.106417 0.086974 0.025804 0.120925
QR 0.866029 1 0.530165 -0.14263 -0.03023 0.179251 -0.35071 -0.06327 0.232009 0.662561 0.080217 0.098658 0.217263 0.138577 0.036682 0.154913
WCTA 0.624787 0.530165 1 0.130422 -0.0918 0.13434 -0.47422 -0.10918 0.125538 0.364002 0.083031 0.094721 0.124079 0.138672 0.089849 0.176695
STA -0.06238 -0.14263 0.130422 1 -0.09342 0.085576 0.058841 -0.02638 0.009962 -0.02906 -0.03648 -0.0344 0.043524 0.01346 -0.02602 0.05074
TES -0.03874 -0.03023 -0.0918 -0.09342 1 -0.0216 0.109666 -0.02586 -0.01585 -0.04703 -0.77583 -0.73286 -0.30174 -0.14705 0.006288 -0.06847
SINV 0.069614 0.179251 0.13434 0.085576 -0.0216 1 -0.23362 -0.00786 0.005763 0.176581 0.02991 0.017755 0.190301 0.065057 0.005557 0.177878
TLTA -0.22326 -0.35071 -0.47422 0.058841 0.109666 -0.23362 1 0.005529 -0.18081 -0.59183 -0.0576 -0.07093 -0.2817 -0.18992 -0.01886 -0.5318
TLTE -0.10692 -0.06327 -0.10918 -0.02638 -0.02586 -0.00786 0.005529 1 -0.00498 -0.01446 0.010915 0.013254 -0.00734 -0.00491 -0.49174 -0.01953
EBITI 0.17468 0.232009 0.125538 0.009962 -0.01585 0.005763 -0.18081 -0.00498 1 0.296074 0.024193 0.020146 0.150415 0.07453 0.008852 0.18073
TETL 0.585299 0.662561 0.364002 -0.02906 -0.04703 0.176581 -0.59183 -0.01446 0.296074 1 0.049573 0.043616 0.302447 0.151044 0.015678 0.335847
EBITS 0.051268 0.080217 0.083031 -0.03648 -0.77583 0.02991 -0.0576 0.010915 0.024193 0.049573 1 0.889698 0.486164 0.249104 0.067613 0.061731
NIS 0.069105 0.098658 0.094721 -0.0344 -0.73286 0.017755 -0.07093 0.013254 0.020146 0.043616 0.889698 1 0.4008 0.405876 0.089637 0.067759
EBITTA 0.106417 0.217263 0.124079 0.043524 -0.30174 0.190301 -0.2817 -0.00734 0.150415 0.302447 0.486164 0.4008 1 0.483848 0.159705 0.359426
NITA 0.086974 0.138577 0.138672 0.01346 -0.14705 0.065057 -0.18992 -0.00491 0.07453 0.151044 0.249104 0.405876 0.483848 1 0.228442 0.222795
NITE 0.025804 0.036682 0.089849 -0.02602 0.006288 0.005557 -0.01886 -0.49174 0.008852 0.015678 0.067613 0.089637 0.159705 0.228442 1 0.023839
RETA 0.120925 0.154913 0.176695 0.05074 -0.06847 0.177878 -0.5318 -0.01953 0.18073 0.335847 0.061731 0.067759 0.359426 0.222795 0.023839 1
As can be seen, the correlation level among liquidity ratios (CACL, QR, and WCTA)
is relatively high compared to that of ratios in other groups. The level of correlation
ranges between 0.53 and 0.87, with only positive correlation observed. On the other
hand, turnover ratios (STA, TES, SINV) are low correlated as the correlation levels
are close to zero. Both positive and negative correlations are observed in this group,
as is the case for the leverage ratios group. The overall level of correlation among
leverage ratios (TLTA, TLTE, EBITI, and TETL) is relatively low. However, one pair
(TLTA and TETL) shows a high level of negative correlation (-0.59). Finally, for the
profitability ratios (EBITS, NIS, EBITTA, NITA, NITE, and RETA), a low level of
correlation is observed, except the pair of EBITS and NIS (0.89). Only positive
correlation is observed in this group, as is the case for the liquidity ratios.
Traditionally, when financial ratios are found highly correlated, only one is kept for
model estimation. This practice has been observed in previous studies which aim to
find the best set of predictors for financial distress. Since this study aims to examine
the ability of financial ratios in signalling financial distress, a comprehensive list of
the 16 ratios are initially retained for model derivation to examine their results. After
138
that, a number of logit models are developed based on the results found in the initial
stage model, which will be explained in more detail below.
5.7.2 Testing for normality of the 16 financial ratios

This sub-section is to test the distributional property of financial ratios. The test for
normality is carried out on all 16 financial ratios of 28 firms during the observation
period. The results from the EVIEWS program are presented in Table 5.16.
Table 5.16 Testing for normality of the 16 financial ratios

Ratios Jarque-Bera Probability Observations
CACL 1390.313 <0.05 558
QR 3167.076 <0.05 558
WCTA 132.9537 <0.05 558
STA 21.7197 <0.05 558
TES 1958023 <0.05 558
SINV 13202.91 <0.05 558
TLTA 47.19569 <0.05 558
TLTE 406017.9 <0.05 558
EBITI 702518.8 <0.05 558
TETL 2019.769 <0.05 558
EBITS 1188365 <0.05 558
NIS 834264.3 <0.05 558
EBITTA 9010.901 <0.05 558
NITA 378973.5 <0.05 558
NITE 347024.9 <0.05 558
RETA 231.5793 <0.05 558
In Table 5.16, the first column (Ratios) displays the financial ratios that are tested for
normality. The second column (Jarque-Bera) specifies a measure of the normality
distribution of explanatory variables in the EVIEWS program. The third column
(Probability) provides P value for the null hypothesis that the series is normally
distributed. A low significant value (p<0.05) suggests that the financial ratio is not
normally distributed. The fourth column (Observation) specifies the number of
observations. The results confirm that financial ratios are not normally distributed.
This is consistent with previous reports that financial ratios are likely to be not
139
normally distributed (Brooks 2008; Eisenbeis 1977; Sori et al. 2006; and Watson
1990).
As a result, it is imperative to employ an estimation approach that assumes non-

normality. Having discussed the testing for both correlation and normality of the 16
financial ratios, the next step is to discuss the logit model results.
5.7.3 Discussion of logit model results based on the entire 20 quarters

sample
Primarily, pool data of the 28 firms 16 financial ratios for the entire 20 quarters
period are used to estimate the logit model (hereafter called Model I). Consequently,
Model I is constructed based on the 558 observations, which is equally divided into
279 observations in the financially distressed group and 279 observations in the non-
financially distressed group. The dependent variable is either one if a firm is
financially distressed or zero if it is not. Figure 5.2 presents the logit results for
Model I3.
3
It should be noted that a 15-ratio logit model (with TETL removal) has been estimated and its results do not
significantly differ from that of the reported estimation of Model I, except that the TLTA ratio variable has a
significant coefficient while the TLTE ratio variable coefficient remained not significant in this model variant.
140
Figure 5.2 Logit results for Model I
Variable Std. Error z-Statistic Prob.
C 1.258466 0.338501 3.717760 0.0002

CACL -2.491273 0.382557 -6.512162 0.0000
QR 1.596787 0.336334 4.747620 0.0000
WCTA 7.236997 1.079198 6.705905 0.0000
STA 5.200526 1.326888 3.919341 0.0001
TES 0.012408 0.024353 0.509503 0.6104
SINV -0.042390 0.010442 -4.059671 0.0000
TLTA -1.056868 0.701647 -1.506268 0.1320
TLTE -0.002800 0.009209 -0.304046 0.7611
EBITI -0.000546 0.000233 -2.338215 0.0194
TETL 0.172823 0.059056 2.926423 0.0034
EBITS -0.157892 0.122093 -1.293205 0.1959
NIS 0.156793 0.129736 1.208550 0.2268
EBITTA 14.46323 10.13570 1.426958 0.1536
NITA -20.65791 10.17428 -2.030405 0.0423
NITE -0.038468 0.212547 -0.180984 0.8564
RETA -0.280761 0.240752 -1.166183 0.2435
McFadden R-squared 0.234582 Mean dependent var 0.500000

S.D. dependent var 0.500449 S.E. of regression 0.422506
Akaike info criterion 1.122026 Sum squared resid 96.57448
Schwarz criterion 1.253772 Log likelihood -296.0452
Hannan-Quinn criter. 1.173477 Restr. log likelihood -386.7761
LR statistic 181.4618 Avg. log likelihood -0.530547
Prob(LR statistic) 0.000000
Obs with Dep=0 279 Total obs 558

Obs with Dep=1 279
According to Figure 5.2, the equation of Model I can be written as follow:
Pi F ( X i )
X i 1.258466 2.491273CACL 1.596787QR 7.236997WCTA 5.200526 STA

0.012408TES 0.042390 SINV 1.056868TLTA 0.0028TLTE 50.000546 EBITI
0.172823TETL 0.157892 EBITS 0.156793NIS 14.46323EBITTA
20.65791NITA 0.038468 NITE 0.280761RETA ....... Model I (5.2)
Where
Pi = Conditional probability of financial distress
F ( X i ) = Cumulative logistic probability function
141
According to the descriptive statistics of Model I, 8 ratios out of the 16 ratios are
statistically significant (p<0.5). Among the statistically significant 8 ratios, there are
three ratios (CACL, QR, and WCTA) in the liquidity group; two ratios (STA and
SINV) in the turnover group; two ratios (EBITI and TETL) in the leverage group; and
one ratio (NITA) in the profitability group. When it comes to Z statistic value, the
ratios that provide the highest absolute Z statistic value in each group are WCTA in
the liquidity group, SINV in the turnover group; TETL in the leverage group; and
NITA in the profitability group.
Generally, the direction of relationship between the change in explanatory variables

and the change in the probability of financial distress is understood by the sign of
coefficients. For instance, the negative coefficient of CACL indicates that an increase
in CACL would decrease the probability of financial distress. However, another
explicit characteristic from Figure 5.2 is that the sign of coefficient estimates for some
ratios can cause misleading interpretations in terms of marginal effect. For example, a
one unit increase in QR and WCTA (liquidity ratios) will cause an increase in the
probability of financial distress, which is against the theory of financial ratios
analysis. Theoretically, an increase in liquidity ratios will cause a decrease in the
probability of financial distress. However, this characteristic is quite often the case in
logit-based studies, for example, the studies of Chi and Tang (2006), Darayseh,
Waples, and Tsoukalas (2003), Reynolds et al. (2002), and Ugurlu and Aksoy (2006).
With reference to the goodness to fit measures, the EVIEWS program provides a
number of measures to examine the goodness of fit of the model such as McFadden
R-Squared, Log likelihood and the like. Brooks (2008) stated that the commonly
reported measures of goodness of fit are the percentage of values correctly predicted
and McFadden R-Squared. In financial distress modelling, the widely used measure of
goodness of fit is the percentage of values correctly predicted, which is presented in
terms of a classification table. Therefore, this study will employ a classification table,
as a conventional measure, to examine the goodness of fit of the estimated logit
model.
142
The coefficients are the odds ratio but not the probability. The probability of financial
distress has been calculated by the following formula:
Where
= probability of an occurrence of financial distress
= unknown parameters
= a set of independent variables
Since the estimated probability of the logit model will lie between 0 and 1, whether a firm is
classified as financially distressed (or not) depends on comparing the estimated probability to
a critical value (the cut-off score). Since the probability is between 0 and 1, the cut-off score
is determined as half of the entire value of the probability. Again, in this study, 1 represents
financially distressed firms and 0 represents non-financially distressed firms. Consequently,
firms with estimated probability of 0.5 or above are be classified as financially distressed
whilst firms with estimated probability of below 0.5 are classified as non-financially
distressed.
Before presenting the classification table, the structure of the table needs to be discussed. The
table shows period, actual status, predicted status, and total number of firms of both groups as
follows:
Table 5.16a The structure of the classification table

Actual Predicted status Total
Period status 0 % 1 % firms
1 0 Hit Miss
1 Miss Hit
Source: Adopted from Altman (1968)
Column 1 represents the period of investigation.

143
Column 2 represents actual status of firms to be investigated, which can be divided into two
groups, that is, non-financially distressed group (0) in row 1 and financially distressed group
(1) in row 2.
Column 3 represents predicted status of firms, which can be divided into four columns as
follows:
Column 3.1:
Row 1 represents the number of correct classification for non-financially distressed

group.
Row 2 represents the number of incorrect classification for financially distressed

group, or type I error.
Column 3.2:
Row 1 represents the percentage of correct classification for non-financially distressed

group.
Row 2 represents the percentage of incorrect classification for financially distressed

group.
Column 3.3:
Row 1 represents the number of incorrect classification for non-financially distressed

group, or type II error.
Row 2 represents the number of correct classification for financially distressed group.
Column 3.4:
Row 1 represents the percentage of incorrect classification for non-financially

distressed group.
Row 2 represents the percentage of correct classification for financially distressed

group.
Column 4 represents the total number of firms in each group.
144
Having discussed the criteria used in the estimated logit model and presented the structure of
the classification table, the next step is to present the classification power of Model I. Table
5.17 exhibits the classification accuracy of Model I in signalling financial distress.
Table 5.17 The classification table of Model I

Predicted status
Period Actual status 0 % 1 % Total firms
Q_1 0 6 43% 8 57% 14
1 8 57% 6 43% 14
Q_2 0 3 21% 11 79% 14
1 8 57% 6 43% 14
Q_3 0 4 29% 10 71% 14
1 11 79% 3 21% 14
Q_4 0 4 29% 10 71% 14
1 9 64% 5 36% 14
Q_5 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_6 0 3 21% 11 79% 14
1 9 64% 5 36% 14
Q_7 0 3 21% 11 79% 14
1 10 71% 4 29% 14
Q_8 0 6 43% 8 57% 14
1 12 86% 2 14% 14
Q_9 0 4 29% 10 71% 14
1 11 79% 3 21% 14
Q_10 0 2 14% 12 86% 14
1 13 93% 1 7% 14
Q_11 0 2 15% 11 85% 13
1 13 100% 0 0% 13
Q_12 0 3 23% 10 77% 13
1 12 92% 1 8% 13
Q_13 0 3 23% 10 77% 13
1 11 85% 2 15% 13
Q_14 0 3 23% 10 77% 13
1 10 77% 3 23% 13
Q_15 0 4 31% 9 69% 13
1 9 69% 4 31% 13
Q_16 0 4 31% 9 69% 13
1 11 85% 2 15% 13
Q_17 0 2 15% 11 85% 13
1 11 85% 2 15% 13
Q_18 0 3 25% 9 75% 12
1 11 92% 1 8% 12
Q_19 0 3 25% 9 75% 12
1 11 92% 1 8% 12
Q_20 0 4 33% 8 67% 12
1 9 75% 3 25% 12
Note: 0=Non-financially distressed firms
1=Financially distressed firms

145
According to equation from Figure 5.2 (Model I) on pages 141, values of ratios of
firms for each period are multiplied with their coefficients. Then, a calculation is
made to find the estimated probability of the firms. After that, the estimated
probability is compared with the cut-off score (0.5). Firms with estimated probability
of 0.5 or above are classified as financially distressed whilst firms with estimated
probability of below 0.5 are classified as non-financially distressed. This
interpretation applies in the following classification tables. From Table 5.17, the
results are from quarter 1 (the most recent quarter prior to financial distress) to quarter
20. A firm status of 0 denotes financial non-distressed firms while 1 represents
financially distressed firms. Due to data availability, financial ratios of 14 financially
distressed firms and 14 non-financially distressed firms are calculated between quarter
1 and quarter 10. During quarter 11-17, financial ratios of 13 financially distressed
firms and 13 non-financially distressed firms are calculated. Finally, financial ratios
of 12 financially distressed firms and 12 non-financially distressed firms are
calculated between quarter 18 and quarter 20.
How to read the Table 5.17 is as follows (please refer to quarter 1):
1) Out of the 14 non-financially distressed firms, the model correctly classified 6,

which accounts for 43 per cent accuracy rate.
2) Out of the 14 non-financially distressed firms, the model erroneously classified 8.
Simply stated, the model classified 8 non-financially distressed firms as
financially distressed firms. This misclassification is called a type II error (a
misclassification of non-financially distressed firms as financially distressed
firms). Thus, the type II error is 57 per cent.
3) Out of the 14 financially distressed firms, the model correctly classified 6,
accounting for 43 per cent accuracy rate.
4) Out of the 14 financially distressed firms, the model incorrectly classified 8 as
non-financially distressed firms. This misclassification is referred to as a type I
error (a misclassification of financially distressed firms as non-financially
distressed firms) and, in this case, the type I error is 57 per cent.
146
The descriptive statistics in Table 5.17 reveal the classification accuracy together with
type I error and type II error between both groups. The highest accuracy rate for the
non-financially distressed group is 43 per cent (quarter 1 and quarter 8) whilst the
lowest accuracy rate for this group is 14 per cent (quarter 10). Thus, the highest and
the lowest type II errors are 86 per cent (quarter 10) and 57 per cent (quarter 1 and
quarter 8), respectively.
The highest and the lowest of accuracy rates for the financially distressed group is 43
per cent (quarter 1 and quarter 2) and 0 (zero) per cent (quarter 11), respectively.
Consequently, the highest and the lowest type I errors are 100 per cent (quarter 11)
and 57 per cent (quarter 1 and quarter 2), respectively. One characteristic emerging in
Table 5.17 is that classification accuracy rates for both groups indicate a steady
decline when the time away from the financial distress increases.
In order to assess the overall classification accuracy of Model I, the average of the
probabilities of correctly classified firms in both groups is calculated. Also, the type I
error (a misclassification of financially distressed firms as non-financially distressed
firms) is focused. The main reason that the type I error is the area focused on is that
this study aims to examine the ability of financial ratios in signalling financial
distress, rather than to find to the optimal set of predictors for financial distress.
Therefore, both overall classification accuracy and type I error of Model I are
observed, which is demonstrated in Table 5.18.
147
Table 5.18 The Overall classification accuracy and the type I error of Model I
Overall classification
Period accuracy* Type I error**
Q_1 43% 57%
Q_2 32% 57%
Q_3 25% 79%
Q_4 32% 64%
Q_5 29% 71%
Q_6 29% 64%
Q_7 25% 71%
Q_8 19% 86%
Q_9 25% 79%
Q_10 11% 93%
Q_11 8% 100%
Q_12 15% 92%
Q_13 19% 85%
Q_14 23% 77%
Q_15 31% 69%
Q_16 23% 85%
Q_17 15% 85%
Q_18 17% 92%
Q_19 17% 92%
Q_20 29% 75%
Note:*Overall classification accuracy is the summation of correctly classified percentage
between both groups divided by 2.
**Type I error is a misclassification of financially distressed firms (1) as non-
financially distressed firms (0).
In Table 5.18, the overall classification accuracy is between 43 per cent (the highest
level in quarter 1) and 8 per cent (the lowest level in quarter 11). The type I error
varies between 100 per cent (the highest level in quarter 11) and 57 per cent (the
lowest level in quarters 1 and 2). Like the pattern emerging in Table 5.17 above, a
steady decline in the overall classification accuracy is observed. In summary, the
Model Is overall classification accuracy rate is less than 45 per cent with the highest
type I error of 100 per cent.
As stated earlier, Model I estimation is used at the initial stage. This study expands
descriptive statistics found in Model I by estimating another two logit models. They
are Model II (comprising the statistically significant 8 ratios (out of 16 ratios)) and
Model III (comprising one ratio in each financial category (e.g. liquidity group,
148
turnover group, leverage group, and profitability group) that provides the highest
absolute Z statistic value).
Model II is developed based on statistically significant 8 ratios out of the 16 ratios.

Model II is estimated by 558 observations, which is equally divided into 279
observations in the financially distressed group and 279 observations in the non-
financially distressed group. Figure 5.3 displays the logit results for Model II.
Figure 5.3 Logit results for Model II
Variable Coefficient Std. Error z-Statistic Prob.
C 0.964019 0.248562 3.878384 0.0001

CACL -2.700131 0.358810 -7.525241 0.0000
QR 1.722071 0.327334 5.260902 0.0000
WCTA 7.850285 0.911308 8.614303 0.0000
STA 5.090225 1.245981 4.085315 0.0000
SINV -0.040482 0.010138 -3.993080 0.0001
EBITI -0.000591 0.000232 -2.546382 0.0109
TETL 0.226993 0.046705 4.860146 0.0000
NITA -8.396136 2.493495 -3.367216 0.0008


Obs with Dep=1 279
According to Figure 5.3, the Model IIs equation can be presented as follows:
Pi F ( X i )
X i 0.964019 2.700131CACL 1.722071QR 7.850285WCTA 5.090225STA

0.040482SINV 0.000591EBITI 0.226993TETL 8.396136 NITA ....... Model II (5.3)
Where

149
After deriving Model II, the probability of the sample is calculated in the same
procedures as those described in Model I. Also the cut-off score and the number of
total firms should be considered in the same manner as those in Model I. The results
of classification accuracy between both groups for Model II are exhibited in Table
5.19.
Table 5.19 The classification table of Model II

Predicted status
Q_1 0 6 43% 8 57% 14
1 9 64% 5 36% 14
Q_2 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_3 0 5 36% 9 64% 14
1 11 79% 3 21% 14
Q_4 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_5 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_6 0 4 29% 10 71% 14
1 9 64% 5 36% 14
Q_7 0 3 21% 11 79% 14
1 11 79% 3 21% 14
Q_8 0 4 29% 10 71% 14
1 11 79% 3 21% 14
Q_9 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_10 0 3 21% 11 79% 14
1 13 93% 1 7% 14
Q_11 0 2 15% 11 85% 13
1 13 100% 0 0% 13
Q_12 0 3 23% 10 77% 13
1 11 85% 2 15% 13
Q_13 0 3 23% 10 77% 13
1 11 85% 2 15% 13
Q_14 0 3 23% 10 77% 13
1 10 77% 3 23% 13
Q_15 0 4 31% 9 69% 13
1 9 69% 4 31% 13
Q_16 0 4 31% 9 69% 13
1 10 77% 3 23% 13
Q_17 0 2 15% 11 85% 13
1 10 77% 3 23% 13
Q_18 0 3 25% 9 75% 12
1 10 83% 2 17% 12
Q_19 0 3 25% 9 75% 12
1 11 92% 1 8% 12
Q_20 0 4 33% 8 67% 12
1 9 75% 3 25% 12
150
From Table 5.19, the highest accuracy rate for the non-financially distressed group is
43 per cent (quarter 1) whilst the lowest accuracy rate for this group is 15 per cent
(quarter 11 and quarter 17). Thus, the highest and the lowest type II errors are 85 per
cent (quarter 11 and quarter 17) and 57 per cent (quarter 1), respectively.
The highest and the lowest of accuracy rates for the financially distressed group are
36 per cent (quarter 1 and quarter 6) and 0 (zero) per cent (quarter 11), respectively.
Consequently, the highest and the lowest type I errors are 100 per cent (quarter 11)
and 64 per cent (quarter 1 and quarter 6), respectively. The between groups results
generated by Model II is quite similar to those of Model I. Next, overall classification
accuracy and type I error are observed.
Table 5.20 The Overall classification accuracy and the type I error of Model II
Period Overall classification accuracy* Type I error**

Q_1 39% 64%
Q_2 29% 71%
Q_3 29% 79%
Q_4 29% 71%
Q_5 29% 71%
Q_6 32% 64%
Q_7 21% 79%
Q_8 17% 79%
Q_9 29% 71%
Q_10 14% 93%
Q_11 8% 100%
Q_12 19% 85%
Q_13 19% 85%
Q_14 23% 77%
Q_15 31% 69%
Q_16 27% 77%
Q_17 19% 77%
Q_18 21% 83%
Q_19 17% 92%
Q_20 29% 75%
151
In Table 5.20, the overall classification accuracy ranges between 39 per cent (the
highest level in quarter 1) and 8 per cent (the lowest level in quarter 11). The type I
error stays between 100 per cent (the highest level in quarter 11) and 64 per cent (the
lowest level in quarters 1 and 6). Like the pattern emerging in Model I, a steady
decline in the overall classification accuracy is observed with increasing time to
financial distress. In summary, the overall classification accuracy rate of Model II is
less than 40 per cent with the highest type I error of 100 per cent.
As can be seen, the overall classification accuracy rate of Model II is lower than that
of Model I but both models generate the same highest type I error. As a result, it can
be inferred that removing non-statistically significant ratios makes no significant
difference in the classification accuracy in this sample set.
Model III is developed based on one ratio in each financial category that provides the
highest absolute Z statistic value in Model I. Since there are four financial groups
described above, Model III is made up of four ratios. Model III is estimated by 560
observations, which is equally divided into 280 observations in the financially
distressed group and 280 observations in the non-financially distressed group. Figure
5.4 presents the logit results for Model III.
152
Figure 5.4 Logit results for Model III
C 0.057158 0.111603 0.512153 0.6085

WCTA 2.228925 0.460974 4.835254 0.0000
SINV -0.001268 0.007716 -0.164372 0.8694
TETL -0.033317 0.025785 -1.292100 0.1963
NITA -4.285623 1.805144 -2.374117 0.0176


Obs with Dep=1 280
According to Figure 5.4, the equation of Model III can be written as follows:
Pi F ( X i )
X i 0.057158 2.228925WCTA 0.001268SINV 0.033317TETL

4.285623NITA ....... Model III (5.4)
Where
After deriving Model III, the probability of the sample is calculated in the same
total firms should be considered in the same manner as those in Model I. The
following is the results of classification accuracy between both groups generated by
Model III.
153
Table 5.21 The classification table of Model III
Predicted status
Q_1 0 8 57% 6 43% 14
1 6 43% 8 57% 14
Q_2 0 7 50% 7 50% 14
1 8 57% 6 43% 14
Q_3 0 9 64% 5 36% 14
1 7 50% 7 50% 14
Q_4 0 8 57% 6 43% 14
1 6 43% 8 57% 14
Q_5 0 8 57% 6 43% 14
1 6 43% 8 57% 14
Q_6 0 5 36% 9 64% 14
1 7 50% 7 50% 14
Q_7 0 6 43% 8 57% 14
1 7 50% 7 50% 14
Q_8 0 6 43% 8 57% 14
1 8 57% 6 43% 14
Q_9 0 5 36% 9 64% 14
1 9 64% 5 36% 14
Q_10 0 3 21% 11 79% 14
1 7 50% 7 50% 14
Q_11 0 4 31% 9 69% 13
1 7 54% 6 46% 13
Q_12 0 6 46% 7 54% 13
1 7 54% 6 46% 13
Q_13 0 7 54% 6 46% 13
1 9 69% 4 31% 13
Q_14 0 7 54% 6 46% 13
1 6 46% 7 54% 13
Q_15 0 6 46% 7 54% 13
1 6 46% 7 54% 13
Q_16 0 5 38% 8 62% 13
1 7 54% 6 46% 13
Q_17 0 5 38% 8 62% 13
1 9 69% 4 31% 13
Q_18 0 5 42% 7 58% 12
1 7 58% 5 42% 12
Q_19 0 5 42% 7 58% 12
1 7 58% 5 42% 12
Q_20 0 6 50% 6 50% 12
1 5 42% 7 58% 12
Note:0=Non-financially distressed firms
Referring to Table 5.21, the highest accuracy rate for the non-financially distressed
group is 64 per cent (quarter 3) whilst the lowest accuracy rate for this group is 21 per
154
cent (quarter 10). When it comes to type II error for this group, the highest level is 79
per cent (quarter 10) and the lowest level is 36 per cent (quarter 3).
58 per cent (quarter 20) and 31 per cent (quarter 13 and quarter 17), respectively. The
highest type I error is 69 per cent (quarter 13 and quarter 17) while the lowest type I
error is 42 per cent (quarter 20). Among the first three models, Model III yields the
best results in terms of the classification accuracy rate and the type I error between
both groups.
Table 5.22 The Overall classification accuracy and the type I error of Model III
Q_1 57% 43%
Q_2 46% 57%
Q_3 57% 50%
Q_4 57% 43%
Q_5 57% 43%
Q_6 43% 50%
Q_7 46% 50%
Q_8 29% 57%
Q_9 36% 64%
Q_10 36% 50%
Q_11 38% 54%
Q_12 46% 54%
Q_13 42% 69%
Q_14 54% 46%
Q_15 50% 46%
Q_16 42% 54%
Q_17 35% 69%
Q_18 42% 58%
Q_19 42% 58%
Q_20 54% 42%
155
Pertaining to Table 5.22, the overall classification accuracy ranges between 57 per
cent (the highest level in quarters 1, 3, 4 and 5) and 29 per cent (the lowest level in
quarter 8). The type I error is between 69 per cent (the highest level in quarters 13 and
17) and 42 per cent (the lowest level in quarter 20). Like the pattern emerging in the
first two models, a steady decline in the overall classification accuracy is observed. In
summary, overall classification accuracy rate of Model III is 57 per cent with the
highest type I error of 69 per cent.
Comparing the three models (Model I, II and III), Model III yields superior results of
overall classification accuracy and type I error. Even though the overall classification
accuracy rate increases to nearly 60 per cent (Model III) from 39 per cent (Model II),
the type I errors provided by the three models are considered high. This leads to the
development of another two models, that is, Model IV (comprising nine ratios that
provide the statistically significant difference of mean value (p<0.05) between the two
groups (please refer to Table 5.3 and 5.4) and Model V (comprising Altmans Z score
four ratios).
With reference to Table 5.3 and 5.4, the between group mean values of nine ratios are
statistically significant different (p<0.05). As result, Model IV is made up of nine
ratios. Model IV is estimated by 558 observations, which is equally divided into 279
observations in the financially distressed group and 279 observations in the non-
financially distressed group. Figure 5.5 reveals the logit results for Model IV.
156
Figure 5.5 Logit results for Model IV
C 1.223180 0.299801 4.079968 0.0000

CACL -1.605370 0.268015 -5.989847 0.0000
QR 0.986804 0.267116 3.694296 0.0002
WCTA 5.398969 0.849202 6.357697 0.0000
STA 3.724077 1.148318 3.243073 0.0012
TLTA -1.299479 0.440881 -2.947458 0.0032
EBITI -0.000349 0.000199 -1.750236 0.0801
EBITS 0.035247 0.110184 0.319893 0.7490
NIS -0.068208 0.106130 -0.642683 0.5204
NITE -0.125601 0.149489 -0.840206 0.4008


Obs with Dep=1 279
According to Figure 5.5, the Model IVs equation can be written as follows:
Pi F ( X i )
X 1.223180 1.605370CACL 0.986804QR 5.398969WCTA 3.724077 STA

1.299479TLTA 0.000349 EBITI 0.035247 EBITS 0.068208 NIS
0.125601NITE ....... Model IV (5.5)
Where
After deriving Model IV, the probability of the sample is calculated in the same
total firms should be considered in the same manner as those in Model I. Table 5.23
illustrates the results of classification accuracy between the financially distressed
group and the non-financially distressed group.
157
Table 5.23 The classification table of Model IV
Predicted status
Q_1 0 6 43% 8 57% 14
1 9 64% 5 36% 14
Q_2 0 4 29% 10 71% 14
1 10 71% 4 29% 14
Q_3 0 4 29% 10 71% 14
1 11 79% 3 21% 14
Q_4 0 6 43% 8 57% 14
1 9 64% 5 36% 14
Q_5 0 5 36% 9 64% 14
1 9 64% 5 36% 14
Q_6 0 3 21% 11 79% 14
1 8 57% 6 43% 14
Q_7 0 4 29% 10 71% 14
1 11 79% 3 21% 14
Q_8 0 5 36% 9 64% 14
1 13 93% 1 7% 14
Q_9 0 5 36% 9 64% 14
1 13 93% 1 7% 14
Q_10 0 4 29% 10 71% 14
1 13 93% 1 7% 14
Q_11 0 3 23% 10 77% 13
1 12 92% 1 8% 13
Q_12 0 3 23% 10 77% 13
1 12 92% 1 8% 13
Q_13 0 3 23% 10 77% 13
1 12 92% 1 8% 13
Q_14 0 3 23% 10 77% 13
1 11 85% 2 15% 13
Q_15 0 3 23% 10 77% 13
1 11 85% 2 15% 13
Q_16 0 4 31% 9 69% 13
1 12 92% 1 8% 13
Q_17 0 4 31% 9 69% 13
1 12 92% 1 8% 13
Q_18 0 3 25% 9 75% 12
1 10 83% 2 17% 12
Q_19 0 3 25% 9 75% 12
1 11 92% 1 8% 12
Q_20 0 6 50% 6 50% 12
1 10 83% 2 17% 12
50 per cent (quarter 20) whilst the lowest accuracy rate for this group is 21 per cent
158
(quarter 6). When it comes to type II error for this group, the highest level is 79 per
cent (quarter 6) and the lowest level is 50 per cent (quarter 20).
The highest and the lowest accuracy rates for the financially distressed group are 43
per cent (quarter 6) and 7 per cent (quarters 8, 9 and 10), respectively. The highest
type I error is 93 per cent (quarters 8, 9 and 10) while the lowest type I error is 57 per
cent (quarter 6).
Table 5.24 The Overall classification accuracy and the type I error of Model IV
Q_1 39% 64%
Q_2 29% 71%
Q_3 25% 79%
Q_4 39% 64%
Q_5 36% 64%
Q_6 32% 57%
Q_7 25% 79%
Q_8 14% 93%
Q_9 21% 93%
Q_10 18% 93%
Q_11 15% 92%
Q_12 15% 92%
Q_13 15% 92%
Q_14 19% 85%
Q_15 19% 85%
Q_16 19% 92%
Q_17 19% 92%
Q_18 21% 83%
Q_19 17% 92%
Q_20 33% 83%
159
From Table 5.24, the overall classification accuracy ranges between 39 per cent (the
highest level in quarters 1 and 4) and 14 per cent (the lowest level in quarter 8). The
type I error is between 93 per cent (the highest level in quarters 8, 9 and 10) and 57
per cent (the lowest level in quarter 6). Like the pattern emerging in previous logit
models estimated in this study, a steady decline in the overall classification accuracy
is observed. In summary, overall classification accuracy rate of Model IV is 39 per
cent with the highest type I error of 93 per cent.
Since this study previously employs the EMS model as an analytical tool in
examining the ability of financial ratios in signalling financial distress, four financial
ratios of the EMS model are used to estimate Model V (a logit-based model).
Therefore, Model V is made up of four ratios. Model V is estimated by 560
observations, which is equally divided into 280 observations in the financially
distressed group and 280 observations in the non-financially distressed group. Figure
5.6 illustrates the logit results for Model V.
Figure 5.6 Logit results for Model V
C 0.042217 0.117205 0.360198 0.7187

WCTA 2.070704 0.453799 4.563041 0.0000
TETL -0.037192 0.026821 -1.386708 0.1655
EBITTA -2.161907 2.283391 -0.946797 0.3437
RETA 0.008784 0.169899 0.051701 0.9588


Obs with Dep=1 280
160
According to Figure 5.6, the Model Vs equation can be presented as follows:
Pi F ( X i )
X i 0.042217 2.070704WCTA 0.037192TETL 2.161907 EBITTA

0.008784 RETA ......ModelV (5.6)
Where
After deriving Model V, the probability of the sample is calculated in the same
shows the results of classification accuracy between the financially distressed group
and the non-financially distressed group.
161
Table 5.25 The classification table of Model V
Predicted status
Q_1 0 7 50% 7 50% 14
1 7 50% 7 50% 14
Q_2 0 8 57% 6 43% 14
1 6 43% 8 57% 14
Q_3 0 9 64% 5 36% 14
1 6 43% 8 57% 14
Q_4 0 9 64% 5 36% 14
1 6 43% 8 57% 14
Q_5 0 8 57% 6 43% 14
1 6 43% 8 57% 14
Q_6 0 6 43% 8 57% 14
1 7 50% 7 50% 14
Q_7 0 6 43% 8 57% 14
1 8 57% 6 43% 14
Q_8 0 7 50% 7 50% 14
1 8 57% 6 43% 14
Q_9 0 5 36% 9 64% 14
1 9 64% 5 36% 14
Q_10 0 5 36% 9 64% 14
1 7 50% 7 50% 14
Q_11 0 4 31% 9 69% 13
1 7 54% 6 46% 13
Q_12 0 7 54% 6 46% 13
1 7 54% 6 46% 13
Q_13 0 8 62% 5 38% 13
1 8 62% 5 38% 13
Q_14 0 7 54% 6 46% 13
1 6 46% 7 54% 13
Q_15 0 8 62% 5 38% 13
1 6 46% 7 54% 13
Q_16 0 7 54% 6 46% 13
1 5 38% 8 62% 13
Q_17 0 5 38% 8 62% 13
1 9 69% 4 31% 13
Q_18 0 5 42% 7 58% 12
1 6 50% 6 50% 12
Q_19 0 5 42% 7 58% 12
1 7 58% 5 42% 12
Q_20 0 6 50% 6 50% 12
1 5 42% 7 58% 12
162
64 per cent (quarters 3 and 4) whilst the lowest accuracy rate for this group is 31 per
cent (quarter 11). The highest type II error is 69 per cent (quarter 11) and the lowest
type II error is 36 per cent (quarter 3 and 4).
62 per cent (quarter 16) and 31 per cent (quarter 17), respectively. For type I error in
this group, the highest level is 69 per cent (quarter 17) while the lowest level is 38 per
cent (quarter 16).
Table 5.26 The Overall classification accuracy and the type I error of Model V
Q_1 50% 50%
Q_2 57% 43%
Q_3 61% 43%
Q_4 61% 43%
Q_5 57% 43%
Q_6 46% 50%
Q_7 43% 57%
Q_8 31% 57%
Q_9 36% 64%
Q_10 43% 50%
Q_11 38% 54%
Q_12 50% 54%
Q_13 50% 62%
Q_14 54% 46%
Q_15 58% 46%
Q_16 58% 38%
Q_17 35% 69%
Q_18 46% 50%
Q_19 42% 58%
Q_20 54% 42%
163
From Table 5.26, the overall classification accuracy ranges between 61 per cent (the
highest level in quarters 3 and 4) and 31 per cent (the lowest level in quarter 8). The
type I error is between 69 per cent (the highest level in quarter 17) and 38 per cent
(the lowest level in quarter 16). Over the entire observation period, the overall
classification accuracy rate remains in the interval of 40 per cent and 60 per cent. In
summary, overall classification accuracy rate of Model V is just above 60 per cent
with the highest type I error of 69 per cent.
To compare and contrast the ability of financial ratios in classifying financially

distressed firms and non-financially distressed firms of Model I to Model V, the
overall classification accuracy and type I error are presented in Table 5.27.
Table 5.27 Comparison of overall classification accuracy and type I error (Model I to Model
V)
Logit-based model Overall classification accuracy Type I error

High Low High Low
Model I (16 ratios) 43% 8% 100% 57%
Model II (8 ratios) 39% 8% 100% 64%
Model III (4 ratios) 57% 29% 69% 42%
Model IV (9 ratios) 39% 14% 93% 57%
Model V (Altmans EMS 4 ratios) 61% 31% 69% 38%
In terms of the accepted level prediction for the estimated models in this study, the
minimum level of accuracy is at 50 per cent. In other words, the estimated models are
considered useful when the models correctly classify more than half of the firms in
both groups (to ensure that they provide a better than random classification (50/50) of
the firms). From Table 5.27, the overall classification accuracy of the estimated
models is relatively poor. For example, Models I, II, and IV correctly classified less
than half of firms in both groups (43 per cent and 39 per cent, respectively).
Nevertheless, Models III and V correctly classified just above half of the firms in both
groups (57 per cent and 61 per cent, respectively). Consequently, Models III and V
provide a better than minimum criteria of acceptability and are considered relatively
useful.
164
Concerning type I error, Model V provides the lowest type I error of 38 per cent. This
means Model V correctly predicts up to 62 per cent of failed firms. Model III provides
the minimum level of type I error of 42 per cent but the rest provide an above 50 per
cent minimum level of type I error.
As mentioned above in Sub-section 3.2.2 (pages 32-33), there are financial and/or
economic costs of financial distress, which may be the consequential cost of not
correctly identifying a potential financial distress of firms. The costs of not
identifying non-financial distress - or classifying financially robust firms as being
likely to suffer financial distress (or type II error) may be associated with investors
(or investment agencies) deciding not to invest (or recommending investors not to
invest) in a company. However, in the context of studying financial distress, the
minimum level of type I error is widely referred to as an appropriate criterion, besides
the overall classification accuracy. It is because of the argument that costs of not
identifying financial distress leads to more significant trauma than that of not
identifying non-financial distress. Since this study aims to examine the ability of
financial statement data in signalling financial distress, type I error (not identifying
financial distress) is mainly focused on in Table 5.27. This explains why type II error
(not identifying non-financial distress) is not included and discussed in Table 5.27.
That more than half of financially distressed firms are correctly classified is employed
in order to judge whether the models are useful or not in terms of financial distress
prediction.
According to the criteria aforementioned, both Model III and Model V yield the better
overall classification accuracy rate (the highest accuracy rate) of above 55 per cent
while the rest provide overall classification accuracy rates below 45 per cent. When it
comes to type I error (misclassification of financially distressed firms as non-
financially distressed firms), Model III and Model V provide a lower type I error of
42 per cent and 38 per cent, respectively. Out of the five models, Model V provides a
better result in terms of both overall classification accuracy (61 per cent) and type I
error (38 per cent).
165
It is noticeable that the results in Table 5.27 are derived from the estimation sample
test. It is because each model is developed based on the entire 20 quarters sample
period and the estimated coefficients are used to classify financial distress over the 20
quarters sample period. Beaver, McNichols and Rhie (2005) state that a more
challenging test of classification ability of financial distress models is an out-of-
sample test. To perform the out-of-sample test, coefficients are estimated from one
period and the estimated coefficients are used to signal financial distress in another
period (Beaver, McNichols & Rhie 2005). Thus, this study will investigate and
present the classification results of out-of-sample test.
5.7.4 Discussion of logit model results based on the split sample

This sub-section is designed to examine the classification results of out-of-sample test.
In other words, the predictive ability of financial ratios for financial distress is
observed. By doing so, coefficients estimated from one sample period are used to
differentiate financially distressed firms and non-financially distressed firms in
another sample period. Since Model V (Altmans EMS four ratios) generates a
superior result, the four ratios are used to estimate coefficients in Model VI and
Model VII below.
Model VI is developed by employing only quarter 1 sample period of the 28 firms

financial data. This explains why Model VI is estimated by 28 observations, which is
equally divided into 14 observations in the financially distressed group and 14
observations in the non-financially distressed group. Figure 5.7 presents the logit
results for Model VI.
166
Figure 5.7 Logit results for Model VI
C -0.273525 0.575335 -0.475419 0.6345

WCTA -0.701573 2.287420 -0.306709 0.7591
EBITTA 5.097333 12.11501 0.420745 0.6739
RETA 0.359796 0.863821 0.416517 0.6770
TETL 0.044618 0.090900 0.490849 0.6235


Obs with Dep=1 14
According to Figure 5.7, the equation of Model VI can be presented as follows:
Pi F ( X i )
X i 0.273525 0.701573WCTA 5.097333EBITTA 0.359796 RETA

0.044618TETL ......Mode VI (5.7)
Where
After a derivation of Model VI, the probability of the sample is calculated in the same
reveals the results of classification accuracy between the financially distressed group
and the non-financially distressed group.
167
Table 5.28 The classification table of Model VI
Predicted status
Q_1 0 6 43% 8 57% 14
1 8 57% 6 43% 14
Q_2 0 8 57% 6 43% 14
1 7 50% 7 50% 14
Q_3 0 7 50% 7 50% 14
1 6 43% 8 57% 14
Q_4 0 7 50% 7 50% 14
1 7 50% 7 50% 14
Q_5 0 7 50% 7 50% 14
1 5 36% 9 64% 14
Q_6 0 8 57% 6 43% 14
1 5 36% 9 64% 14
Q_7 0 7 50% 7 50% 14
1 5 36% 9 64% 14
Q_8 0 7 50% 7 50% 14
1 4 29% 10 71% 14
Q_9 0 8 57% 6 43% 14
1 5 36% 9 64% 14
Q_10 0 8 57% 6 43% 14
1 4 29% 10 71% 14
Q_11 0 7 54% 6 46% 13
1 4 31% 9 69% 13
Q_12 0 6 46% 7 54% 13
1 5 38% 8 62% 13
Q_13 0 7 54% 6 46% 13
1 4 31% 9 69% 13
Q_14 0 7 54% 6 46% 13
1 4 31% 9 69% 13
Q_15 0 5 38% 8 62% 13
1 3 23% 10 77% 13
Q_16 0 5 38% 8 62% 13
1 3 23% 10 77% 13
Q_17 0 6 46% 7 54% 13
1 4 31% 9 69% 13
Q_18 0 8 67% 4 33% 12
1 4 33% 8 67% 12
Q_19 0 4 33% 8 67% 12
1 4 33% 8 67% 12
Q_20 0 5 42% 7 58% 12
1 4 33% 8 67% 12

168
According to Table 5.28, quarter 1 provides the estimation sample test while the rest
(quarter 2 to quarter 20) provides the out-of-sample test. For the estimation sample
test, the classification accuracy rate for both groups is similar (43 per cent in quarter
1) with the same rate of type I and type II errors (57 per cent).
For the out-of-sample test, the highest classification accuracy rate for the non-
financially distressed firms is 67 per cent (quarter 18) whilst the lowest accuracy rate
for this group is 33 per cent (quarter 19). The highest type II error is 67 per cent
(quarter 19) and the lowest type II error is 33 per cent (quarter 18).
The highest and the lowest accuracy rates for the financially distressed group are 77
per cent (quarters 15 and 16) and 50 per cent (quarters 2 and 4), respectively.
Concerning the type I error in this group, the highest level is 50 per cent (quarters 2
and 4) while the lowest level is 23 per cent (quarters 15 and 16).
As can be seen, the out-of-sample results of Model VI are quite promising compared
to those of Model I to Model V. The between group classification accuracy rate
reaches 77 per cent for financially distressed firms and 67 per cent for non-financially
distressed firms. Table 5.29 exhibits the overall classification accuracy results and
type I error of Model VI.
169
Table 5.29 The Overall classification accuracy and the type I error of Model VI
Q_1 43% 57%
Q_2 54% 50%
Q_3 54% 43%
Q_4 50% 50%
Q_5 57% 36%
Q_6 61% 36%
Q_7 57% 36%
Q_8 40% 29%
Q_9 61% 36%
Q_10 64% 29%
Q_11 62% 31%
Q_12 54% 38%
Q_13 62% 31%
Q_14 62% 31%
Q_15 58% 23%
Q_16 58% 23%
Q_17 58% 31%
Q_18 67% 33%
Q_19 50% 33%
Q_20 54% 33%
The estimation sample test and out-of-sample test should be read in the same manner
as those in Table 5.28 above. Concerning the estimation sample test, the overall
classification accuracy is 43 per cent while type I error stands at 57 per cent. For the
out-of-sample test, the overall classification accuracy ranges between 67 per cent (the
highest level in quarter 18) and 40 per cent (the lowest level in quarter 8). The type I
error is between 50 per cent (the highest level in quarters 2 and 4) and 23 per cent (the
lowest level in quarters 15 and 16). It is noticeable that Model VIs out-of-sample
results (both overall classification accuracy rate and type I error) are considered better
than those of Model I to Model V.
170
Model VII is developed by employing quarter 1 to quarter 4 sample periods of the 28
firms financial data. This explains why Model VII is estimated by 112 observations,
which is equally divided into 56 observations in the financially distressed group and
56 observations in the non-financially distressed group. Figure 5.8 presents the logit
results for Model VII.
Figure 5.8 Logit results for Model VII
C 0.032501 0.284555 0.114218 0.9091

WCTA -0.755071 1.185753 -0.636786 0.5243
EBITTA -10.15856 6.291717 -1.614593 0.1064
RETA 0.147316 0.376413 0.391369 0.6955
TETL 0.072563 0.054529 1.330704 0.1833


Obs with Dep=1 56
According to Figure 5.8, the equation of Model VII can be written as follows:
Pi F ( X i )
X i 0.032501 0.755071WCTA 10.15856 EBITTA 0.147316 RETA

0.072563TETL ......Mode VII (5.8)
Where
After a derivation of Model VII, the probability of the sample is calculated in the
same procedures as those described in Model I. Also the cut-off score and the number
171
of total firms should be considered in the same manner as those in Model I. Table
5.30 reveals the results of classification accuracy between the financially distressed
group and the non-financially distressed group.
Table 5.30 The classification table of Model VII

Predicted status
Q_1 0 6 43% 8 57% 14
1 6 43% 8 57% 14
Q_2 0 7 50% 7 50% 14
1 9 64% 5 36% 14
Q_3 0 5 36% 9 64% 14
1 7 50% 7 50% 14
Q_4 0 5 36% 9 64% 14
1 7 50% 7 50% 14
Q_5 0 9 64% 5 36% 14
1 9 64% 5 36% 14
Q_6 0 5 36% 9 64% 14
1 6 43% 8 57% 14
Q_7 0 7 50% 7 50% 14
1 8 57% 6 43% 14
Q_8 0 7 50% 7 50% 14
1 5 36% 9 64% 14
Q_9 0 5 36% 9 64% 14
1 2 14% 12 86% 14
Q_10 0 6 43% 8 57% 14
1 4 29% 10 71% 14
Q_11 0 8 62% 5 38% 13
1 4 31% 9 69% 13
Q_12 0 7 54% 6 46% 13
1 5 38% 8 62% 13
Q_13 0 8 62% 5 38% 13
1 5 38% 8 62% 13
Q_14 0 5 38% 8 62% 13
1 7 54% 6 46% 13
Q_15 0 9 69% 4 31% 13
1 7 54% 6 46% 13
Q_16 0 8 62% 5 38% 13
1 7 54% 6 46% 13
Q_17 0 9 69% 4 31% 13
1 8 62% 5 38% 13
Q_18 0 6 50% 6 50% 12
1 4 33% 8 67% 12
Q_19 0 9 75% 3 25% 12
1 7 58% 5 42% 12
Q_20 0 7 58% 5 42% 12
1 5 42% 7 58% 12
172
According to Table 5.30, quarter 1 to quarter 4 provide the estimation sample test
while the rest (quarter 5 to quarter 20) provides the out-of-sample test. For the
estimation sample test, the highest classification accuracy for the non-financially
distressed firms is 50 per cent (quarter 2) and the lowest accuracy rate for this group is
36 per cent (quarters 3 and 4). The type II error for non-financially distressed firms
ranges between 64 per cent (the highest level in quarters 3 and 4) and 50 per cent (the
lowest level in quarter 2). The classification accuracy for financially distressed firms
is between 57 per cent (the highest level in quarter 1) and 36 per cent (the lowest level
in quarter 2). The type I error is between 64 per cent (the highest level in quarter 2)
and 43 per cent (the lowest level in quarter 1).
For the out-of-sample test, the highest classification accuracy rate for the non-
financially distressed firms is 75 per cent (quarter 19) whilst the lowest accuracy rate
for this group is 36 per cent (quarters 6 and 9). The highest type II error is 64 per cent
(quarters 6 and 9) while the lowest type II error is 25 per cent (quarter 19). The
highest and the lowest accuracy rates for the financially distressed group are 86 per
cent (quarter 9) and 36 per cent (quarter 5), respectively. The type I error in this group
stays between 64 per cent (the highest level in quarter 5) and 14 per cent (the lowest
level in quarter 9).
Between Model VI and Model VII, Model VII provides better out-of-sample results
with the highest classification accuracy of 75 per cent for the non-financially
distressed firms and 86 per cent for the financially distressed firms. Moreover, the
between group out-of-sample results of Model VII are superior to those of Model I to
Model VI. Nevertheless, the overall classification accuracy of Model VII needs to be
examined. Table 5.31 exhibits the overall classification accuracy results and type I
error of Model VII.
173
Table 5.31 The Overall classification accuracy and the type I error of Model VII
Q_1 50% 43%
Q_2 43% 64%
Q_3 43% 50%
Q_4 43% 50%
Q_5 50% 64%
Q_6 46% 43%
Q_7 46% 57%
Q_8 38% 36%
Q_9 61% 14%
Q_10 57% 29%
Q_11 65% 31%
Q_12 58% 38%
Q_13 62% 38%
Q_14 42% 54%
Q_15 58% 54%
Q_16 54% 54%
Q_17 54% 62%
Q_18 58% 33%
Q_19 58% 58%
Q_20 58% 42%
The estimation sample test and out-of-sample test should be read in the same manner
as those in Table 5.30 above. For the estimation sample test, the overall classification
accuracy is between 50 per cent (the highest level in quarter 1) and 43 per cent (the
lowest level in quarters 2, 3 and 4) while type I error ranges between 64 per cent (the
highest level in quarter 2) and 43 per cent (the lowest level in quarter 1).
For the out-of-sample test, the overall classification accuracy ranges between 65 per
cent (the highest level in quarter 11) and 38 per cent (the lowest level in quarter 8).
The type I error is between 64 per cent (the highest level in quarter 5) and 14 per cent
(the lowest level in quarter 9). To compare and contrast the results of Model VI and
Model VII, Table 5.32 is presented below.
174
Table 5.32 Comparison of overall classification accuracy and type I error (Model VI and
Model VII)
Type of model and sample test accuracy Type I error
High Low High Low
Model Estimation sample test 43% na 57% na
VI Out-of-sample test 67% 40% 50% 23%
Model Estimation sample test 50% 43% 64% 43%
VII Out-of-sample test 65% 38% 64% 14%
From Table 5.32, the overall classification accuracy of the estimated models in the
estimation sample test is relatively poor. For example, Model VI correctly classified
43 per cent firms in both groups while Model VII correctly classified 50 per cent
firms in both groups. However, the overall classification accuracy of the models in the
out-of-sample test significantly improves. For example, Model VI correctly classified
firms in both groups at 67 per cent while Model VII correctly classified firms in both
groups at 65 per cent.
Concerning type I error, Model VI provides a minimum level of type I error of 23 per
cent while Model VII provides minimum level of type I error as low as 14 per cent.
This means Model VI correctly predicts up to 77 per cent of failed firms while Model
VII correctly predicts up to 86 per cent. To determine whether the models are useful
or not, the criteria stated earlier (page 164) is applied. As a result, the predictive
ability in terms of identifying financial distress is relatively impressive, that is, up to
86 per cent for Model VII and up to 77 per cent for Model VI. It should be noted that
a number of characteristics emerging in Model VI and Model VII are that both models
employ the same four financial ratios of Altmans EMS model and that the estimated
coefficients of each model are different because of different sample periods used in
estimating the logit model.
175
According to the results in Table 5.27 and Table 5.32, it can be inferred that the
model estimated by the most recent data sample period provides superior results to
models developed by a pool data sample set. Again, to support the research question
of this study, the correct classification of the financially distressed firms or the type I
error plays a crucial role. In Model VI, the correct classification of the financially
distressed firms is up to 77 per cent (the lowest type I error of 23 per cent) while, in
Model VII, the correct classification of the financially distressed firms is up to 86 per
cent (the lowest type I error of 14 per cent).
It should be noted that, besides the most recent 1 quarter model (Model VI), a number
of logit models (for instance, the recent 2 quarter model, and the recent 3 quarter
model) have been estimated but their results are not as strong as that of Model VI.
Likewise, the recent 8 quarter model, the recent 12 quarter model, and the recent 16
quarter model have been estimated but their results are not as strong as that of the
recent 4 quarter model (Model VII). Therefore, Model VI and Model VII have been
discussed in this study.
With reference to the results presented above, it can be inferred that financial ratios
could provide useful information for signalling an occurrence of financial distress
during the specified period in Thailand.
5.8 Conclusion
As mentioned at the beginning of this chapter, this chapter aimed to apply the methodology
together with the data collected to answer the research question.
The chapter began with the descriptive profile of companies, financial statements, and ratios
used for the analysis. In particular, significant differences in financial ratios were identified to
illustrate the characteristics of the target companies and the benchmark companies over the
investigation period. Both descriptive and inferential statistics were employed to support the
176
identification of these groups. After presenting the key attributes/variables of both groups, the
discussion turned to the individual application of each predictive measure and its results. The
performance of each measure for the target group and the benchmark group was discussed.
The measures adopted in this study can be divided into two groups. They are (1) the EMS
model representing sophisticated prediction measures, and (2) comparative ratio analysis and
ratio trend analysis representing traditional measures. The underlying reasons for employing
the three measures were provided. The IMM approach and its results were discussed and
presented. Likewise, the rationale for adopting the IMM approach has been discussed in
Section 5.6.
To provide a triangulation perspective, a set of logit models were estimated based on pool
data of 28 firms 16 financial ratios with the entire 20 quarters sample period. Prior to a
derivation of logit-based models, the 16 financial ratios were tested for both correlation and
normality. The results for correlation and normality tests were presented and discussed. Then,
seven logit models were estimated with different criteria. They include Model I (comprising
the 16 financial ratios), Model II (comprising the statistically significant 8 ratios (out of 16
ratios)), Model III (comprising one ratio in each financial category (e.g. liquidity group,
turnover group, leverage group, and profitability group) that provides the highest absolute Z
statistic value), Model IV (comprising nine ratios that provide the statistically significant
difference of mean value (p<0.05) between the two groups), and Model V to Model VII
(comprising Altmans Z score four ratios).
Model I to Model V were estimated based on the entire data sample set, which provides the
results of the estimation sample test. Alternatively, Model VI and Model VII were estimated
on the basis of a spilt data sample, which provides the results of the out-of-sample test. To
perform the out-of-sample test in this study, coefficients estimated from one sample period
were used to differentiate financially distressed firms and non-financially distressed firms in
another sample period.
Finally, results of each model were presented and discussed. They include the between group
classification accuracy rate, the rate of type I error and type II error, and the overall
classification accuracy rate. Moreover, a comparison of overall classification accuracy
177
together with type I error among Model I to Model V was presented and discussed, as was the
case of Model VI to Model VII.
Having conducted the data analysis, the discussion now turns to conclusions and implications
of the findings. The details of these issues will be discussed in Chapter 6.
178
Chapter 6: Conclusions and Implications
6.1 Introduction
Data analysis and results were shown in Chapter 5. This chapter discusses the findings
derived from the data analysis. Also a discussion of the implications and limitations of this
study is presented. Before discussing the aforementioned aspects, it is imperative to restate
the purpose of this study. This study investigates the ability of financial statement
information to signal business failure in Thailand in recent years. Specifically, this study
examines whether financial statements can adequately be used to discriminate between
potentially failed and nonfailed Thai firms in normal economic circumstances. As
mentioned earlier, this research did not focus on finding the best predictors for failure. The
underlying reason behind this study is that there has not been any empirical study
investigating the ability of financial statements to signal business failure in Thailand,
particularly in the period following the 1997 financial crisis. Also the Integrated Multi
Measure (IMM) approach and logit modelling used together with quarterly financial
statements had not been previously considered in the analysis of financial distress in the
Thailand environment.
The chapter comprises seven sections as depicted in Figure 6.1. Section 6.1 outlines the
objective and the structure of the chapter. Section 6.2 provides summary about research
issues, which were developed in Chapter 1. Implications of this research are presented in
Section 6.3. Section 6.4 outlines contributions of the study, followed by limitations of this
study in Section 6.5. Section 6.6 suggests implications for further research. Finally, the
conclusions of this chapter are drawn in Section 6.7.
179
6.1 Introduction
6.2 Summary
6.2.1 Summary about the 6.2.2 Summary about the

research question research problem
6.3 Implications
6.3.1 Implications for 6.3.2 Implications for policy

theory and practice
6.4 Contributions of the study
6.5 Limitations of this study
6.6 Implications for further research
6.7 Conclusion
180
6.2 Summary
In investigating the ability of financial statement information to signal business failure, this
study initially attempted to resolve a number of debated issues in the context of ratiobased
modelling for business failure. This section provides an overall conclusion aiming to link the
focal sections of this study.
First of all, it is imperative to understand generic terms used in failure definition and these
terms were presented in Chapter 3. There are a multitude of definitions of business failure in
the literature. This is because there is no agreement among researchers on this issue (Ohlson
1980). According to previous studies, it is reasonable to conclude that the two main
approaches to defining failure reflect the specific interests of the researchers and the
conditions of the companies being examined (Dimitras, Zanakis & Zopounidis 1996). To
satisfy the purpose of this study, consideration of the term failure, together with financial
distress was deemed preferable because business failure and/or financial distress reflects the
fact that firms which are financially weak do not always become legally bankrupt (Gilbert,
Menon & Schwartz 1990; Hamer 1983; Hill, Perry & Andes 1996; and Scott 1981). The
discussion of adopting a definition of failure in this study was made in Chapter 4. Without a
formal definition of business failure, the integrity of the sample becomes questionable
(Hossari 2006, p. 298). Business failure in this study is defined as including firms that are
financially weak (but not always become bankrupt) and which have been delisted. This is
consistent with previous studies. Once the generic terms were presented, the theoretical
framework was discussed, followed by a review of the previous research.
In the literature review, the methodological developments were explicitly reviewed. One
example is that, before the advent of the multivariate approach, the univariate approach was
used. Another example is that some methodologies incorporate nonfinancial statement
information (i.e. inflation, interest rates and the like) in their modelling of business failure.
The findings of the literature review helped identify the research gap and led to the
development of the research problem and question of this study. Again, the research problem
is whether financial statement information can be adequately used to predict financial distress
for Thai firms in normal economic circumstances. The research question is can financial
181
statement information be used to discriminate between potentially failing and nonfailing
firms in the context of Thailand in normal economic circumstances?
Another core issue in studying business failure is the selection of the sample group.
Following many previous studies, this study uses both failed and nonfailed firms. A sample
that contains only failed firms is inadequate for this study because the measures (for instance,
comparative ratio analysis, ratios trend analysis, and the IMM approach) used need to
compare and contrast financial ratios of both groups. Furthermore, the pair sample technique
and the matching criteria (i.e. similar industry, asset size, and statement date) were applied.
The rationale for adopting this technique and the criteria were provided in Chapter 4.
The investigation period is one of the controversial issues in this field because there is no
unanimous view among researchers concerning the optimal investigation period in the study
of business failure prediction. Notwithstanding this, a large number of studies investigate the
fiveyear period prior to failure. Given this, it can be inferred that a study period of more
than or less than five years might be ineffective. As a result, this study adopted five years as
an investigation period.
Another important issue is the selection of financial ratios. In earlier studies, 44 potential
ratios were identified as potentially useful indicators for business failure so they were
considered in this study as a starting point. Of these, a total of 16 ratios were selected in this
study. The 16 ratios can be divided into two groups: one is the four ratios required for the
EMS model and the other is the 15 ratios selected for the two traditional measures (including
some of the EMS models ratios). The 15 ratios used in the two measures were selected based
on their use and technical merit in previous studies. In addition, the 15 ratios were divided
into four meaningful areas of financial performance, that is, liquidity, turnover/performance,
leverage/solvency, and profitability. The definition of each selected ratio was provided in
Chapter 4. After that, the discussion turned to the three measures used for failure prediction
and the rationale for selecting the measures. The three measures include the EMS model,
comparative ratio analysis, and ratio trend analysis. The first measure is a representative of
182
sophisticated failure prediction ratiobased models while the last two measures represent
traditional ratiobased measures. The discussion together with justification for the
development of the Integrated MultiMeasure approach (IMM) is presented. Moreover, the
discussion along with justification for utilising logit-based models for benchmarking is also
provided.
Chapter 5 reported the data analysis. The descriptive profiles of all data used were provided
prior to the results from applying individual measures and then the IMM approach. They
included the profile of companies, the profile of financial statements, and the profile of ratios.
Furthermore, both descriptive statistic (i.e. mean and standard deviation) and inferential
statistic (i.e. ttest) were used to help describe the financial characteristics of both groups.
Subsequently, the results of the analysis were presented based on the criteria employed in
each measure. Both early and recent correct classifications before failure illustrated warning
signs. Early correct classification before failure would be helpful for management to take a
corrective action to prevent an occurrence of failure. Importantly, recent correct classification
before failure illustrates the predictive ability of financial statements in this study, which is
similar to the idea used in previous studies. Concerning the logit-based model, test results for
correlation and for normality of the 16 financial ratios were presented followed by estimation
results for a number of the logit models. Finally, discussion of the logit model results was
presented.
Having reviewed the overall conclusions, the discussion now turns to the conclusions about
the research question (Subsection 6.2.1), followed by the conclusions about research
problem (Subsection 6.2.2).
6.2.1 Summary about the research question

In this sub-section, the findings from Chapter 5 are discussed in the context of the
research question developed in Chapter 3. There was one research question developed
in response to the research problem and the research question is restated as follows:
183
Can financial statement information be used to discriminate between potentially
failing and nonfailing firms in the context of Thailand in normal economic
circumstances?
The research question examines whether financial statement data can be used to
classify potentially failed and nonfailed firms in Thailand in recent years. The
development of the question was motivated by the fact that a combination of financial
ratios and nonfinancial ratios has attracted attention in recent studies in modelling of
business failure. Also, most empirical work in Thailand has been focused on periods
of economic crisis, rather than normal economic circumstances. Using the figures
announced by the Ministry of Commerce, it can be inferred that financial distress is
persistent and critical in Thailand.
As presented in Chapter 4, the failed firms and their benchmark firms were paired by
similar industry, asset size, and statement date. The 16 ratios used can be divided into
two groups: one is the four EMS models required ratios and another is the 15 ratios
selected for the two traditional measures. The 15 ratios used in the two measures were
selected based on their usefulness in previous studies. In addition, the 16 ratios were
used in estimation of the unrestricted logit model and variously in estimation of a
number of restricted logit models. The ratios represent four meaningful financial
characteristics of business entities. They include liquidity, turnover/performance,
leverage/solvency, and profitability. Descriptive and inferential statistics, which were
discussed in Chapter 5, were employed to gain insight into the preliminary
characteristics of the phenomena being studied.
Before addressing the research question above, it is imperative to discuss the

approach used in this study. This study adopted the IMM approach to investigate the
ability of financial statements to signal business failure and employed the logit-based
model to validate the results of the IMM approach. The IMM approach is represented
by one sophisticated ratiobased model and two traditional ratiobased measures. The
184
underlying rationale for adopting the IMM approach is to comprehensively broaden
the perspective about the phenomena being studied (Easterby-Smith, Thorpe & Lowe
1991). To answer the above research question, the graphical table of the IMM result
for the failed group needs to be referred to.
According to the Table 5.12 (on page 139), the findings reveal that, out of the 14
failed firms, there are 13 firms for which financial statements could reveal symptoms
of financial distress/fragility, which is represented by green colour. The financially
unhealthy symptoms for targeted firms manifested themselves either for every quarter
or for some quarters during the investigation period. Another finding which emerged
is that different measures provide different indications. For instance, in case of the
firm called ASTL, there are 12 quarters that are correctly determined by ratio trend
analysis measure; three quarters by comparative ratio analysis measure; and none by
EMS model. This finding is consistent with the results of previous studies which
found that no special measure is dominant over its counterparts in terms of
classificatory accuracy. With reference to the IMM criteria (page 99), there are 9 out
of the 14 financially distressed firms were correctly classified. Based on the results, it
is reasonable to conclude that financial statements can be used to identify financially
unhealthy firms in the Thai context in the specified period.
To examine whether financial statements can be used to identify financially healthy

firms in the Thai context in the specified period, the IMM approach together with its
result for the nonfailed group is presented. Consequently, the graphical table of the
IMM approach application result for the nonfailed group is referred to.
To interpret the results of non-failed firms, it should be noted that a financial distress
classification is adopted when any TWO of the three measures criteria are satisfied.
The criteria for financial distress classification have been mentioned in Section 4.7
(page 96). Table 5.13 (on page 134) reveals that financial statements could correctly
indicate overall financial health for all 14 benchmarking firms. This is represented by
185
green colour. The correct indication pattern of an individual measure was different
during the investigation period. For example, in the case of RCI, there are 12 quarters
that are correctly indicated by ratio trend analysis; seven quarters by EMS model; and
three quarters by comparative ratio analysis. This is similar to the results found in
previous research that different measures provide different indications. With reference
to the IMM criteria, there are 8 out of the 14 non-financially distressed firms were
correctly classified. Finally, the findings from the nonfailed group confirm that
financial statements can be used to indicate financially healthy firms in the Thai
context in the specified period.
To provide a triangulation perspective in examining the ability of financial ratios in

signalling financial distress, this study employed the logit-based model as a
benchmarking measure. A number of specifications of the logit model were estimated
based on several assumptions. In terms of the accepted level of the estimated models
in this study, the minimum level of accuracy is at 50 per cent. In other words, the
estimated models are considered useful when the models correctly classify more than
half of the firms in both groups. With reference to Table 5.30 (on page 172), the
correct classification of the financially distressed firms is up to 86 per cent (please
refer to quarter 9).
According to the results aforementioned, it can be inferred that financial statement

data can be used to classify potentially failed and nonfailed firms in the context of
Thailand in recent years. This helps explain why financial ratios have been
predominantly used in modelling of business failure in previous studies.
Having answered the research question, the discussion now turns to the conclusions
about research problem.
186
6.2.2 Summary about the research problem
Like the previous subsection, this subsection discussed the findings from Chapter 5
relevant to the research problem developed in Chapter 3. The research problem is
restated as follows:
Whether financial statement information can be adequately used to predict

financial distress for Thai firms in normal economic circumstances
The study of business failure has attracted a great deal of attention from researchers
and analysts for many decades (Altman 1968; Blum 1974; Coats & Fant 1993; Deakin
1972; Edmister 1972; Ohlson 1980; Pongsatat, Ramage & Lawrence 2004;
Puagwatana & Gunawardana 2005; and Ugurlu & Aksoy 2006). In the early period of
this research, researchers such as Altman (1968) and Beaver (1966) used financial
statement data as their principal tool to predict business failure. However, a number of
recent researchers, as presented in Chapter 3, have used both financial statement data
and nonfinancial factors as predictors for failure. With reference to Table 3.2 in
Chapter 3, the overall classification accuracy for the recent models indicates that there
is no significant difference in the classification accuracy of ratiobased models and
nonratiobased models. Stated simply, employing financial statement information is
likely to be adequate for predicting failure. This led to the development of the
research question above.
In the literature, the concept of recent correct classification before failure is used to
test the power of prediction models. Like previous studies, the recent correct
classification before failure is adopted in this study to examine the predictive ability
of financial statement information for financial distress. The findings reveal that for 9
out of the 14 failed firms, recent correct classification emerged. This accounts for 64
per cent of failed firms (9 out of the 14). Likewise, the results from the nonfailed
group show recent correct classification for 8 out of the 14 surviving firms. The
successful rate for the IMM approach in classifying the nonfailed group is 57 per
cent (8 out of the 14 nonfailed firms). Nearly 66 per cent (or 2/3) successful
187
classification rate for the failed group is promising. Furthermore, the results from the
logit model reveal that the correct classification of the financially distressed firms is
up to 86 per cent resulting in the lowest type I error of 14 per cent. This evidence
helps answer the research problem above. It is reasonable to conclude that financial
statement information can be adequately used to predict financial distress for Thai
firms in the specified period.
Having presented the conclusions about the research question and research problem,
the next task focuses on a number of the implications of this study.
6.3 Implications
The distinguishing characteristics of this study lie in five areas. They include (1) that the
study focuses on financial statement information as the potential predictor for business failure,
(2) that quarterly financial statement data, rather than annual financial statement data, were
used, (3) that triangulation of the three methods and the logit-based model was adopted to
examine the predictive ability of financial statement data, (4) that the Thai setting, an
emerging market environment instead of developed market environment, was the area studied,
and (5) that normal economic circumstances, instead of a period of economic crisis, were
studied. The implications from the findings in Chapter 5 are discussed in terms of theory, and
policy and practice. The implications for theory are discussed first, followed by the
implications for policy and practice.
6.3.1 Implications for theory

Generally, in the study of business failure, traditional theory indicates that most failed
firms in previous studies were forced into a declaration of bankruptcy in a court.
However, a number of previous studies (e.g. Gilbert, Menon & Schwartz 1990;
Hamer 1983; Hill, Perry & Andes 1996; and Scott 1981) argue that the event of
business failure and/or financial distress implies that firms are financially troubled but
do not always become bankrupt. Bankruptcy is only one possible outcome of business
188
failure and others include business restructuring, liquidation, and even that publicly
listed firms go private (Coats & Fant 1993; and Queen & Roll 1987). These
phenomena indicate the deteriorated financial conditions of companies.
As reported in Chapter 5, most failed firms financial conditions deteriorated before

failure. Particularly, net profit margin (net income/sales), return on assets (net
income/total assets), and return on equity (net income/total equity) for the failed firms
were significantly different from those of the nonfailed firms. This is likely to cause
shareholders of the failed firms to exercise their right to protect their interest in the
firm before the financial conditions get worse. The findings were that the failed firms
had deteriorated financial condition and decided to delist from the SET. This confirms
that the firms are financially weak but do not become bankrupt. This is consistent
with the evidence in previous studies.
In the literature, a number of researchers such as Theodossiou (1993) and Shumway

(2001) have attempted to develop time series models. The models principally aim to
detect changes to the financial characteristics of firms. In the context of business
failure, the time series patterns of a firms ratios should be observed to assess the
progress of the firm (Coats & Fant 1993). However, previous researchers used annual
financial statement data to estimate their time series models. If the objective of the
time series models is to identify unfavourable changes to the financial characteristics
of a firm as soon as they take place, the use of quarterly financial statement data
(three months period) should prove more appropriate than annual financial statement
data (12 months period).
Since the study aims to investigate the ability of financial statements to signal
business failure, the findings could be used as additional evidence to support the
recent findings of relevant ratiobased studies such as the study of Beaver, McNichols
and Rhie (2005). Also, the findings indicate a direct theory implication that financial
statement information can be assessed in terms of predictive ability for failure during
189
normal economic circumstances, not only for periods of economic crisis. Moreover,
as the study was focused on Thailand, researchers in emerging market countries
whose capital market conditions are similar to Thailands can use the findings to
compare and contrast the patterns of predictive ability of financial ratios. Above all,
the findings can be used as a stepping stone to further develop predictive models for
signalling business failure.
In addition to implications for theory, the findings have implications for policy and
practice. These are discussed next.
6.3.2 Implications for policy and practice

Due to the fact that this study was focused on Thailand, the results could have
implications for Thai regulatory bodies, Thai private sector management and analysts.
The implications for Thai authorities are discussed first, followed by the implications
for Thai private sector management and analysts.
When business failure happens before being found by Thai regulatory bodies, one of
questions is the appropriateness of the predictive measures used by the authorities.
Studies like this one would provide helpful instruments to Thai regulatory bodies to
assist in the detection of impending corporate failures. Since the services provided by
some industries involve either the majority of people, or the macroeconomic
environment of the country, Thai regulatory bodies are obliged to prevent potentially
adverse incidents, rather than protect stakeholders from the consequences once failure
has occurred. For instance, the Bank of Thailand has an interest in the solvency
situation of commercial banks in Thailand. Not only is this due to regulation of
monopoly, and fair trade practice, but also the need for the public service provided.
This study, like previous ones, could equip Thai authorities with alternative means to
detect impending corporate failure.
190
In the context of signalling business failure, the findings of this study could be used as
preliminary evidence on the usefulness of financial statements in signalling business
failure. The results could be expanded by developing an optimal set of financial ratios
that is considered the best predictor for failure. It is likely that the failure of Thai
listed companies could tarnish Thailands economic image to both domestic and
overseas investors. An early warning system like a ratiobased prediction model is
likely to be one of the alternative instruments available to Thai regulatory bodies to
mitigate the problem.
Besides the aforementioned implications, the findings also have implications for Thai
private sector management and analysts. As noted in Chapter 3, business failure has
adverse impacts on stakeholders. They include shareholders/investors, creditors,
suppliers, managers, workers, and customers (Chen & Merville 1999; Fitzpatrick
1931; and Hertzel et al. 2006). The loss of key workers/managers, main suppliers, big
customers, and the loss of confidence of creditors and shareholders/investors could be
prevented if the management have an appropriate early warning system for signalling
impending failure.
Studies like this one could help boost internal management monitoring systems and
provide a basis for management to keep a track record of a companys performance.
For the IMM approach, management who are not comfortable with the use of
sophisticated statistics can replicate the methodology used in this study to create a
suitable tool for monitoring both company performance and the event of impending
business failure. For the logit model, management who are familiar with statistical
package programs like EVIEWS or SPSS can also replicate the methodology
employed in this study to develop a financial model for their own purpose.
The Integrated MultiMeasure technique and the logit model are likely to be
analytical instruments used in the field of credit analysis, financial analysis, and
investment analysis. In reality, both individuals and institutions encounter investing
191
and lending decisions that have to assess the riskiness of their future returns. It is
accepted that failure inevitably incurs both direct costs (i.e. money out of pocket) and
indirect costs (i.e. opportunity costs). To prevent and/or alleviate the costs, analysts
should have an appropriate tool for detecting impending failure. Examining financial
statements together with other qualitative data helps analysts to identify the strengths
and weaknesses of a company. In this regard, the methodology of this study enables
analysts to monitor early warning signs of failure. If financial fragility is detected,
remedial actions could be taken before failure occurs.
However, it is not suggested that managers and/or analysts exclusively take financial
statements into account when studying business failure. Other nonfinancial statement
data and qualitative data should be included. According to the results of this study,
financial statement information could be used as an additional tool for signalling
business failure. Furthermore, recent research conducted by Beaver, McNichols and
Rhie (2005) finds that financial statement data have not become less informative over
time in the context of signalling failure. In other words, the predictive ability of
financial ratios is robust over time. As can be seen, the findings of this study are
consistent with the findings of Beaver, McNichols and Rhies (2005) study even
though the studies were done in different environments, using different analytical
techniques and in different periods.
Having presented the implications for theory and for policy and practice, the
discussion now turns to contributions of this study.
6.4 Contributions of the study

There are a number of contributions that this study has made to the body of knowledge in the
context of signalling business failure. The contributions of this study are discussed as follows:
192
The study expands the body of knowledge about financial statementsbased
modelling of failure in Thailand, one of the emerging market countries. As noted
in Chapter 3, most well-known studies in this field are conducted in western
countries like the US and little empirical research has been conducted in emerging
market countries. In particular, as far as the author has been able to establish, an
attempt to investigate the predictive ability of financial statement data for business
failure in Thailand has not been previously undertaken. Most extant studies in
Thailand concentrate on finding the best predictors of failure.
This study has refined the methodology used in ratiobased signalling for failure
by adopting quarterly financial statement data rather than annual financial data.
Previous international and Thai studies used annual financial statement data in the
modelling of failure. Adopting quarterly financial statements enables researchers
to take into account cyclical and seasonal changes. This methodological
refinement is consistent with the recent shift from static to dynamic modelling of
failure. Static/traditional models represent the models that use snapshot or
singleperiod data to estimate such models, while dynamic models are derived
from multipleperiod data or time series data. The recent shift has resulted from
the notion that the failure process is dynamic, not static (Altman 1970; Hossari
2007; Shumway 2001; and Theodossiou 1993). Given this, the use of quarterly
financial statement data (threemonth periods) rather than annual financial
statement data (12month periods) should be more appropriate for detecting
unfavourable changes of financial characteristics of a firm as soon as they begin.
To date, most work in Thailand has been focused on studying business failure in
periods of economic crisis. This study attempts to examine the predictive ability
of financial statements in normal economic circumstances in Thailand. This might
explain why some findings of this study are consistent with the literature, but
some are not. In Chapter 2, the discussion of recent Thai business failure reflected
that business failure occurs any time, not only in periods of economic crisis. This
led to choosing normal economic circumstances as the sampling period in this
study.
193
The descriptive statistical findings of this study provide a broader perspective on
the event of failure during the specified period. Referring the four areas of
financial performance, the failed group in previous studies tends to have lower
profitability, lower liquidity, lower asset quality, and higher leverage level. This
studys findings illustrated some difference, that is, the failed group is likely to
have lower profitability, lower liquidity, lower asset quality, and lower financial
leverage. That financially distressed firms do not always become bankrupt is clear.
The reasons for this inconsistency might be that (1) previous studies narrowly
define failure resulting in legally bankrupt firms dominating in their samples, and
(2) most Thai literature focuses on periods of economic crisis but this study
concentrates on normal economic circumstances.
Applying the Integrated MultiMeasure technique and the logit model for
analysing the data confirmed that the findings were thoroughly examined. The
results of the study indicate that financial statements are still informative for Thai
companies in the context of signalling failure. Stated clearly, using financial
statement data is adequate for signalling Thai business failure. The view that
financial statements have become less informative was not supported by the
results of this study. Given this, concerned parties should pay more attention to
financial statement data than they have in the past. Furthermore, concerned parties
such as regulatory bodies, managers, and analysts could use the studys findings
as evidence to further develop efficacious prediction models for failure.
Since the findings illustrate that Thai business failure relates to a firms financial
characteristics, the study helps fill the gap between conceptual development and
real world practice. The procedure (i.e. sample selection, variable selection, the
techniques used, and the assessment of the result) can be replicated by interest
groups involved in decisionmaking affected by a firms financial characteristics,
particularly those in emerging market countries.
The next task is to discuss the limitations of this study.
194
6.5 Limitations of this study
Like all extant research, this study is not without limitations. There are a number of
limitations that emerged during the researchs progression. These are outlined below.
As can be seen, there are various methodological paradigms in modelling of

business failure utilising a statistical approach that could be observed in Chapter
3. One of these paradigms uses financial statement data for detecting symptoms of
financial fragility and/or early warning signs of failure. Another adopts non
financial factors or a combination of both financial statements and nonfinancial
factors for signalling business failure. According to Beaver, McNichols and
Rhies study (2005), the predictive ability of approaches incorporating non-
financial factors was not significantly more accurate than those reliant on
financial statement data. A limitation of this study is that it focused on information
from financial statements in the prediction of business failure and excluded
consideration of nonfinancial factors.
This study employed three established ratiobased measures along with the logit-
based model as analytical techniques and 16 financial ratios as proxies for the four
areas of financial performance. The three measures are represented by one
sophisticated ratiobased model and two traditional ratiobased measures. The
rationale for adopting the three measures was provided in Chapter 4. Nevertheless,
there are a number of other techniques and potential financial ratios that could be
drawn from the literature for examining the predictive ability of financial
statements. A limitation is that the results of this study were derived from
adopting the three measures, the logit model, and the selected financial ratios.
Due to the limited number of firms in Thailand delisting during the available
sample timeframe, the sample size is relatively small compared to previous studies.
However, this study applied an appropriate matching procedure in order to make
the results more reliable. They included matching firms from similar industries,
195
matching firms closest in asset size, and matching firms which had similar
statement dates.
Due to data availability constraints, the sample of failed and nonfailed companies
was public companies. The financial statement data of the sample were derived
from the SET database. The data set came from the entire population of listed
firms which met the analytical criteria during the investigation period. A
limitation of this study is that it did not examine the predictive ability of financial
statements for private companies.
The reasons for delisting provided to the SET are similar and follow a somewhat
standard wording pattern. The probability is that some failed firms might request
their delisting for strategic reasons. Inclusion of these firms may have distorted
the results. In principle, this matter should be checked, but this is beyond the
scope of this study. Nonetheless, in this study, both qualitative and quantitative
criteria were applied to ensure that all selected failed firms belonged to the same
population.
The Altman EMS model was developed using a different definition of failure to
that employed in the current research.
Notwithstanding this, the limitations are acknowledged. It is asserted that the limitations do
not unduly detract from the significance of this research. This study employed both a well-
tested sophisticated ratiobased model and traditional ratiobased measures and the selection
of criteria for financial ratios used in this study was based on their usefulness in the literature.
This concludes the limitations section. The next section discusses the implications for further
research.
6.6 Implications for further research

This section aims to assist future researchers to design and/or select topics and make
suggestions about the methodology used for further study. Simply stated, there are a number
196
of opportunities that emerge in this study for followup research. The implications for further
research are discussed as follows:
This study focuses mainly on financial statements as potential predictors for

business failure in Thailand. As noted earlier, there are various methodological
paradigms concerning the explanatory variables such as employing nonfinancial
factors and/or a combination of financial statements and nonfinancial factors. As
a matter of fact, financial statement information is one of an extensive group of
potential indicators for business failure. Therefore, a recommendation for future
research is to investigate the ability of nonfinancial factors and/or a combination
of financial statements and nonfinancial factors for signalling business failure.
In Chapter 3, there are a number of methodological designs for modelling of

business failure in terms of both the statistical tools and the approaches used.
Concerning the statistical technique, this study used both descriptive and
inferential statistics to describe the characteristics of the target group and of the
benchmarking group. Similarly, the Integrated MultiMeasure approach and the
logit model were adopted to test the predictive ability of financial statements. The
underlying reasons for employing the two approaches have already been discussed.
Given that, a recommendation for future research would be to apply different
methodological designs such as MDA, Probit analysis and the like to evaluate the
results and/or to improve the validity of the results.
This study focuses on Thailand, one of the emerging market countries. The
characteristics of all companies in the sample, that is, financial statement data, are
derived from the SET database. Any research conducted in similar stock market
environments could help validate and/or refine the results of this study. As such, a
recommendation is that followup research could be replicated in other emerging
market countries whose capital market conditions are similar to Thailand.
197
Like most extant studies in Thailand, this study investigates large and/or public
companies instead of small/medium and/or private companies. The underlying
reason for not focusing on the small/medium and/or private companies is the
inaccessibility of data. If a followup researcher could manage to access the
financial statement data of these companies, it would help expand the body of
knowledge about signalling business failure in the Thailand context. In this
respect, a recommendation for future research would be to replicate this study by
employing financial statement data of small/medium and/or private companies in
Thailand.
This study was limited in terms of sample size due mainly to the availability of
data and a specified sampling period. Expanding the size of the sample would in
principle enhance the validity of results in investigating the predictive ability of
financial statements. Unlike the established capital markets in western countries,
Thailands capital market is at an early development stage, and so not many
companies are listed in the market. Moreover, the percentage of delisting firms
during the normal economic circumstances was noticeably different from that in
the 1997 economic financial crisis. With the passing of a period of time there may
be the opportunity to access a larger sample size of Thai firms.
Finally, since the findings suggest that financial statements can be adequately used
to predict business failure for Thai firms in normal economic circumstances, they
could be used as a stepping stone to make progress in the context of signalling
business failure. A followup researcher who is interested in finding the best
predictors for failure could develop predictive models and compare the prediction
accuracy of the models. One potential area for further research flowing form this
thesis is that development of ratio based models that triangulate that results of
comparative static ratios, trend in ratios and indicators like the EMS model
derived from logit analysis. Development of these models may be further
facilitated as a more comprehensive firm database utilising quarterly data becomes
available with the passage of time.
198
Having discussed the implications for further research, the conclusion of this chapter is drawn
next.
6.7 Conclusion
This study examines the research problem: Whether financial statement information can be
adequately used to predict financial distress for Thai firms in normal economic
circumstances. An exploratory research design was adopted because (1) the predictive
ability of financial statement information for Thai business failure during normal economic
circumstances has not been previously appraised, (2) quarterly financial statement data have
not been used in the Thai literature on predicting business failure, and (3) the Integrated
MultiMeasure technique for data analysis had not been previously employed.
The methodology used in this study is that of secondary data study. The source of the data
is the SET database. According to the literature, a number of recent Thai studies use a
combination of financial statement data and nonfinancial factors as predictors for business
failure. In addition, recent Thai business failure indicates that business failure occurs any
time, not only in periods of economic crisis. These considerations led to the development of
the research problem. Finally, with the Integrated MultiMeasure technique and the logit
model, the findings indicate that financial statement information can be adequately used to
predict financial distress for Thai firms in normal economic circumstances.
199
List of References
AccessMyLibrary n.d., Thailand pays off IMF debt [Online], Available:
http://www.accessmylibrary.com/article-1G1-106543052/thailand-pays-off-imf.html,
[Accessed 1 October 2009].
AFET n.d., Introduction-The Agricultural Futures Exchange of Thailand [Online], Available:

http://www.afet.or.th/v081/english/corporate/introduction.php, [Accessed 6 October
2009].
, Vision and Mission [Online], Available:

http://www.afet.or.th/v081/english/corporate/vision_and-mission.php, [Accessed 6
October 2009].
Altman, EI 1968, 'Financial Ratios, Discriminant Analysis and the Prediction of Corporate
Bankruptcy', Journal of Finance, vol. 23, no. 4, pp. 589-610.
1970, 'Ratio Analysis and the Prediction of Firm Failure: A Reply', The Journal of
Finance, vol. 25, no. 5, pp. 1169-72.
1973, 'Predicting Railroad Bankruptcies in America', Bell Journal of Economics and

Management Science, vol. 4, no. 1, pp. 184-212.
1982, 'Accounting Implications of Failure Prediction Models', Journal of Accounting

Auditing and Finance, vol. 4, no. 19, pp. 4-19.
1993, Corporate Financial Distress and Bankruptcy: A Complete Guide to Predicting

and Avoiding Distress and Profiting from Bankruptcy, 2nd edn, John Wiley & Sons,
Inc., USA.
2005, 'An Emerging Market Credit Scoring System for Corporate Bonds', Emerging
Markets Review, vol. 6, no. 4, pp. 311-23.
Altman, EI & Hotchkiss, E 2006, Corporate Financial Distress and Bankruptcy: Predict and
Avoid Bankruptcy, Analyze and Invest in Distress Debt, 3rd edn, John Wiley & Sons,
Inc., USA.
Back, B, Laitinen, T, Sere, K & Wezel, MV 1996, 'Choosing Bankruptcy Predictors Using
Discriminant Analysis, Logit Analysis, and Genetic Algorithms', Technical Report,
vol. 40, pp. 1-18.
Balcaen, S & Ooghe, H 2004, '35 Years of Studies on Business Failure: An Overview of the
Classic Statistical Methodologies and Their Related Problems', EconPapers.
2006, '35 Years of Studies on Business Failure: An Overview of the Classic Statistical
Methodologies and Their Related Problems', The British Accounting Review, vol. 38,
no. 1, pp. 63-93.
200
Beal, D, Goyen, M & Shamsuddin, A 2008, Introducing Corporate Finance, 2 edn, John
Wiley & Sons Australia, Ltd., Australia.
Beaver, WH 1966, 'Financial Ratios as Predictors of Failure', Empirical Research in

Accounting: Selected Studies, Supplement to Journal of Accounting Research, vol. 4,
pp. 71-111.
1968a, 'Alternative Accounting Measures As Predictors of Failure', The Accounting

Review, January, pp. 113-22.
1968b, 'Market Prices, Financial Ratios, and the Prediction of Failure', Journal of
Accounting Research, Autumn, pp. 179-92.
Beaver, WH, McNichols, MF & Rhie, JW 2005, 'Have Financial Statements Become Less
Informative? Evidence from the Ability of Financial Ratios to Predict Bankruptcy',
Review of Accounting Studies, vol. 10, pp. 93-122.
Bei, Z & Liu, L 2005, 'Empirical Study on Financial Failure Prediction of Listed Companies',
Chinese Business Review, vol. 4, no. 8, pp. 73-8.
Berenson, ML, Levine, DM & Krehbiel, TC 2009, Basic Business Statistics: Concepts and
Applications, Australasian and Pacific edn, Pearson Education Australia.
Berenson, ML, Levine, DM, Krehbiel, TC, Watson, J & Turner 2007, Basic Business
Statistics: Concepts and Applications, Australasian and Pacific edn, Pearson
Education Australia.
BEX n.d., Background of Thai Bond Market [Online], Available: http://www.set.or.th,

Birnberg, JG, Shields, MD & Young, SM 1990, 'The Case of Multiple Methods in Empirical
Management Accounting Research (With an Illustration from Budget Setting)',
Journal of Management Accounting Research (Fall), pp. 33-66.
Black, TR 1999, Doing Quantitative Research in the Social Sciences, SAGE Publications Ltd,
London.
Blum, M 1974, 'Failing Company Discriminant Analysis', Journal of Accounting Research,

vol. 12, no. 1, pp. 1-25.
BOT n.d., About BOT [Online], Available:

http://www.bot.or.th/ENGLISH/ABOUT/Pages/index.aspx?, [Accessed 5 October
2009].
, The Bank of Thailand Board [Online], Available:

http://www.bot.or.th/English/AboutBOT/Committees/Pages/CourtOfDirectors.aspx,
201
, Financial Market Development [Online], Available:
http://www.bot.or.th/English/FinancialMarkets/Fin_Mkt_Development/Pages/Financi
al..., [Accessed 1 October 2009].
, Representation in International Fora [Online], Avaliable:

http://www.bot.or.th/English/AboutBOT/index/Pages/representative.aspx, [Accessed
12 October 2009].
Brannen, J 2005, Mixed Methods Research: A Discussion Paper [Online], Available:

http://www.ncrm.ac.uk/research/outputs/publications/documents/MethodsReviewPape
rNCRM-005.pdf, [Accessed 4 May 2010].
Brigham, EF & Ehrhardt, MC 2008, Financial Management: Theory and Practice, 12 edn,
South-Western, USA.
Bryant, MS 1997, 'A Case-Based Reasoning Approach to Bankruptcy Prediction Modeling',

Journal of Intelligent Systems in Accounting, Finance and Management, vol. 6, pp.
195-214.
Bulow, JI & Shoven, JB 1978, 'The Bankruptcy Decision', The Bell Journal of Economics,
vol. Autumn, pp. 437-56.
Cavana, RY, Delahaye, BL & Sekaran, U 2001, Applied Business Research: Qualitative and
Quantitative Methods, John Wiley and Sons Australia Ltd, Australia.
Central Intelligence Agency n.d., Economy:Thailand [Online],

Available:http://www.cia.gov/library/publications/the-world-factbook/geos/th.html,
[Accessed 30 September 2009].
Charitou, A, Neophytou, E & Charalambous, C 2004, 'Predicting Corporate Failure:

Empirical Evidence for the UK', European Accounting Review, vol. 13, no. 3, pp.
465-97.
Chen, GM & Merville, LJ 1999, 'An Analysis of the Underreported Magnitude of the Total
Indirect Costs of Financial Distress', Review of Quantitative Finance and Accounting,
vol. 13, pp. 277-93.
Chi, LC & Tang, TC 2006, 'Bankruptcy Prediction: Application of Logit Analysis in Export
Credit Risks', Australian Journal of Management, vol. 31, no. 1, pp. 17-28.
Coats, PK & Fant, LF 1993, 'Recognizing Financial Distress Patterns Using a Neural
Network Tool', Financial Management, vol. 22, no. 3, pp. 142-56.
Dambolena, IG & Khoury, SJ 1980, 'Ratio Stability and Corporate Failure', Journal of
Finance, vol. 35, no. 4, pp. 1017-26.
Darayseh, M, Waples, E & Tsoukalas, D 2003, 'Corporate Failure for Manufacturing

Industries Using Firms Specifics and Economic Environment with Logit Analysis',
Managerial Finance, vol. 29, no. 8, pp. 23-36.
202
Davis, D 2005, Business Research for Decision Making, 6th edn, Thomson South-Western,
USA.
Deakin, EB 1972, 'A Discriminant Analysis of Predictors of Business Failure', Journal of

Accounting Research, vol. 10, no. 1, pp. 167-80.
Denzin, NK & Lincoln, YS 2008, Strategies of Qualitative Inquiry, 3rd edn, Sage
Publications, Inc., USA.
Dimitras, AI, Slowinski, R, Susmaga, R & Zopounidis, C 1999, 'Business Failure Prediction
Using Rough Sets', European Journal of Operational Research, vol. 114, no. 2, pp.
263-80.
Dimitras, AI, Zanakis, SH & Zopounidis, C 1996, 'A Survey of Business Failures with an
Emphasis on Prediction Methods and Industrial Applications', European Journal of
Operational Research, vol. 90, no. 3, pp. 487-513.
Easterby-Smith, M, Thorpe, R & Lowe, A 1991, Management Research: An Introduction,

Sage, London.
Edmister, RO 1972, 'An Empirical Test of Financial Ratio Analysis for Small Business
Failure Prediction', Journal of Financial and Quantitative Analysis, vol. 7, no. 2, pp.
1477-94.
Encyclopedia of the Nations n.d., Thailand: Overview of Economy [Online], Available:

http://www.nationsencyclopedia.com/economies/Asia-and-the-Pacific/Thailand-OVE,
[Accessed 30 September 2009].
Firminger, L 2003, Trend Analysis: Methods and Problems, Strategic Planning Services,
Swinburne University of Technology TAFE Divison.
Fitzpatrick, P 1931, Symptoms of Industrial Failure, Catholic University of America Press.,

USA.
Flagg, JC, Giroux, GA & Wiggins Jr., CE 1991, 'Predicting Corporate Bankruptcy Using
Failing Firms', Review of Financial Economics, vol. 1, no. 1, pp. 67-78.
Frank, RE, Massy, WF & Morrison, DG 1965, 'Bias in Multiple Discriminant Analysis',
Journal of Marketing Research, vol. 11, pp. 250-8.
Frino, A, Amelia, H & Chen, Z 2009, Introduction to Corporate Finance, 4 edn, Pearson,
Australia.
Frydman, H, Altman, EI & Kao, D 1985, 'Introducing Recursive Partitioning for Financial
Classification: The Case of Financial Distress', The Journal of Finance, vol. 40, no. 1,
pp. 269-92.
203
Gable, GG 1994, 'Intergrating Case Study and Survey Research Methods: An Example in
Information Systems', European Journal of Information Systems, vol. 3, no. 2, pp.
112-26.
Gentry, JA, Newbold, P & Whitford, DT 1985, 'Classifying Bankrupt Firms with Funds Flow
Components', Journal of Accounting Research, vol. 23, no. 1, pp. 146-61.
Gibson, CH & Frishkoff, PA 1986, Financial Statement Analysis:Using Financial

Accounting Information, 3 edn, Kent Publishing Company, USA.
Gilbert, LR, Menon, K & Schwartz, KB 1990, 'Predicting Bankruptcy for Firms in Financial
Distress', Journal of Business Finance and Accounting, vol. 17, no. 1, pp. 161-71.
Ginoglou, D, Agorastos, K & Hatzigagios, T 2002, 'Predicting Corporate Failure of

Problemmatic Firms in Greece with LPM Logit Probit and Discriminant Analysis
Models', Journal of Financial Management and Analysis, vol. 15, no. 1, pp. 1-15.
Goudie, AW 1987, 'Forecasting Corporate Failure: The Use of Discriminant Analysis within
a Disaggregated Model of the Corporate Sector', Journal of the Royal Statistical
Society, vol. 150, no. 1, pp. 69-81.
Government and Politics in Thailand n.d. , An Overview of Government and Politics in

Thailand [Online], Available: http://thaiembdc.org/politics/govpol.htm, [Accessed 29
September 2009].
Hamer, MM 1983, 'Failure Prediction: Sensitivity of Classification Accuracy of Alternative

Statistical Methods and Variable Sets', Journal of Accounting and Public Policy, vol.
2, no. 4, pp. 289-307.
He, Y & Kamath, R 2006, 'Business Failure Prediction in Retail Industry: An Empirical
Evaluation of Generic Bankruptcy Prediction Models', Academy of Accounting and
Financial Studies Journal, vol. 10, no. 2, pp. 97-110.
Hertzel, MG, Li, Z, Officer, MS & Rodgers, KJ 2006, Inter-firm Linkages and the Wealth
Effects of Financial Distress along the Supply Chain, [Online], Available:
http:/lcb.uoregon.edu/departments/fin/HLOR_7_28, [Accessed 21/07/09].
Hill, NT, Perry, SE & Andes, S 1996, 'Evaluating Firms in Financial Distress: An Event
History Analysis', Journal of Applied Business Research, vol. 12, no. 3, pp. 60-71.
Hing, A & Lau, L 1987, 'A Five-State Financial Distress Prediction Model', Journal of
Hossari, G 2006, 'A Ratio-Based Multi-Level Modelling Approach for Signalling Corporate
Collapse: A Study of Australian Corporations', PhD Thesis, Swinburne University of
Technology, Australia.
2007, A Heuristic Framework for Ratio-Based Modelling of Corporate Collapse

[Online], Available:
204
http://www.melbournecentre.com.au/Finsia_MCFS/2007/Ghassan_Hossari.pdf,
2005, 'Methodological Shifts in Ratio-Based Modelling of Corporate Collapse',

Australasian Journal of Business and Behavioural Sciences, vol. 1, no. 2, pp. 101-15.
Jacobs, JA 2005, Multiple Methods in American Sociological Review [Online], Available:

http://www2.asanet.org/footnotes/dec05/indextwo.html, [Accessed 3 May 2010].
Jones, S & Hensher, DA 2004, 'Predicting Firm Financial Distress: A Mixed Logit Model',
Accounting Review, vol. 79, no. 4, pp. 1011-38.
Kahya, E & Theodossiou, P 1999, 'Predicting Corporate Financial Distress: A Time-Series

CUSUM Methodology', Review of Quantitative Finance and Accounting, vol. 13, pp.
323-45.
Karels, GV & Prakash, AJ 1987, 'Multivariate Normality and Forecasting of Business

Bankruptcy', Journal of Business Finance and Accounting, vol. 14, no. 4.
Leksrisakul, P 2004, 'A Model of Corporate Bankruptcy in Thailand: MDA and ANN
Systems', DBA Thesis, Southern Cross University, Australia.
Levine, DM, Krehbiel, TC & Berenson, ML 2003, Business Statistics: A First Course, 3rd
edn, Pearson Education, Inc.
Levine, DM, Stephan, DF, Krehbiel, TC & Berenson, ML 2008, Statistics for
Managers:Using Microsoft Excel, 5th edn, Pearson Education International.
Manning, M & Munro, D 2007, The Survey Researcher's SPSS Cookbook, 2nd edn, Pearson
Education Australia.
McKee, TE 2000, 'Developing a Bankruptcy Prediction Model via Rough Sets Theory',
International Journal of Intelligent Systems in Accounting, Finance and Management,
vol. 9, pp. 159-73.
McMurray, D 2007, Qualitative Research Methods, 3rd edn, Southern Cross University,
Australia.
Merwin, CL 1942, Financing Small Corporations in Five Manufacturing Industries, 1926-36,

National Bureau of Economic Research, New York.
Miles, MB & Huberman, AM 1994, Qualitative Data Analysis, 2nd edn, Thousand Oaks, CA.
Ministry of Commerce 2009, Business Registrations Statistic [Online], Available:

http://www.dbd.go.th, [Accessed 30 November 2009].
Neuman, WL 2006, Social Research Methods: Qualitative and Quantitative Approaches, 6th
edn, Pearson Education, Inc., USA.
205
Norton, CL & Smith, RE 1979, 'A Comparison of General Price Level and Historical Cost
Financial Statements in the Prediction of Bankruptcy', Accounting Review, vol. 54, no.
1, pp. 72-87.
Ohlson, JA 1980, 'Financial Ratios and the Probabilistic Prediction of Bankruptcy', Journal of
Osteryoung, J, Constand, RL & Nast, D 1992, 'Financial Ratios in Large Public and Small
Private Firms', Journal of Small Business Management (Journal Article Excerpt), vol.
30.
Pager, D & Quillian, L 2005, 'Walking the Talk? What Employers Say Versus What They
Do', American Sociological Review, vol. 70, no. 3, pp. 355-80.
Palat, R 1989, Understanding Ratios: A Practical Guide for Business, Finance and Banking,
Kogan Page, UK.
Persons, OS 1999, 'Using Financial Information to Differentiate Failed vs. Surviving Finance
Companies in Thailand: An Implication for Emerging Economies', Multinational
Finance Journal, vol. 3, no. 2, pp. 127-45.
Platt, HD & Platt, MB 1990, 'Development of a Class of Stable Predictive Variables: The
Case of Bankruptcy Prediction', Journal of Business Finance and Accounting, vol. 17,
no. 1, pp. 31-51.
1991, 'A Note on the Use of Industry-Relative Ratios in Bankruptcy Prediction', Journal
of Banking and Finance, vol. 15, pp. 1183-94.
Pongsatat, S, Ramage, J & Lawrence, H 2004, 'Bankruptcy Prediction for Large and Small
Firms in Asia: A Comparison of Ohlson and Altman', Journal of Accounting and
Corporate Governance, vol. 1, no. 2, pp. 1-13.
Puagwatana, S & Gunawardana, KD 2005, 'Business Failure Prediction Model: A Case Study
of Technology Industry in Thailand', IEEE International Engineering/Management
Conference, vol. 1, pp. 246-9.
Punch, KF 2006, Developing Effective Research Proposals, 2nd edn, SAGE Publications Ltd,
London.
Queen, M & Roll, R 1987, 'Firm Mortality: Using Market Indicators to Predict Survival',
Financial Analysts Journal, vol. 43, no. 3, pp. 9-26.
Reisz, AS & Perlich, C 2007, 'A Market-based Framework for Bankruptcy Prediction',
Journal of Financial Stability, vol. 3 pp. 85-131.
Reynolds, S, Fowles, R, Gander, J, Kunaporntham, W & Ratanakomut, S 2002, 'Forecasting

the Probability of Failure of Thailand's Financial Companies in the Asian Financial
Crisis', Economic Development and Cultural Change, vol. 51, no. 1, pp. 237-46.
206
Ross, SA, Thompson, S, Chistensen, M, Westerfield, RW & Jordan, B 2007, Fundamentals
of Corporate Finance, 4th edn, McGraw-Hill Irwin.
Routledge, J & Gadenne, D 2000, 'Financial Distress, Reorganization and Corporate

Performance', Accounting and Finance, vol. 40, pp. 233-60.
Salkind, NJ 2008, Statistics for People Who (Think They) Hate Statistics, 3rd edn, SAGE
Publications.
Scott, J 1981, 'The Probability of Bankruptcy: A Comparison of Empirical Predictions and

Theoretical Models', Journal of Banking and Finance, vol. 5, no. 3, pp. 317-44.
SEC 2009a, International Relations [Online], Available:

http://www.sec.or.th/sec/Content_0000000018.jsp?categoryID=CAT0000431&lang=
en, [Accessed 14 October 2009].
2009b, Introduction to the SEC [Online], Available:

2009c, SEC Board [Online], Available:

SET 2001, Listing of Ordinary Shares or Preferred Shares as Listed Securities, The Stock
Exchange of Thailand.
2003, Listing of Securities on the Market for Alternative Investment, The Stock
Exchange of Thailand.
2006, Procedures and guidelines for listed companies if their operations or financial
conditions call for possible delisting, Listing and Disclosure Department; Listed
Companies Development Department, The Stock Exchange of Thailand.
2008, Delisting Securities files, The Stock Exchange of Thailand.
SET n.d., About MAI [Online], Available:

http://www.set.or.th/en/about/overview/about_mai_vision.html, [Accessed 8 October
2009].
, Board of Governors [Online], Avaliable:

http://www.set.or.th/en/about/overview/board_p1.html, [Accessed 25 September
2009].
, Equities [Online], Available:

http://www.set.or.th/en/products/equities/equities_p1.html, [Accessed 8 October
2009].
207
, Glossary [Online], Available:
http://www.set.or.th/en/products/bonds/files/Glossary_eng.pdf, [Accessed 10
September 2010].
, History and Roles [Online], Available:

http://www.set.or.th/en/about/overview/history_p1.html, [Accessed 25 September
2009].
, Market Statistic [Online], Available:

http://www.set.or.th/en/market/market_statistics.html, [Accessed 13 October 2009].
Shumway, T 2001, 'Forecasting Bankruptcy More Accurately: A Simple Hazard Model',

Journal of Business, vol. 74, no. 1, pp. 101-24.
Sookhanaphibarn, K, Polsiri, P, Choensawat, W & Lin, FC 2007, 'Application of Neural

Networks to Business Bankruptcy Analysis in Thailand', Internation Journal of
Computational Intelligence Research, vol. 3, no. 1, pp. 91-6.
Sookying, S, n.d., The Department of Special Investigation (DSI): Countermeasures in

regard to the Investigation of Economic Crimes and Special Crimes in Thailand
[Online], Available: http://www.unafei.or.jp/english/pdf/PDF_rms/no66/L_p171-
p178.pdf, [Accessed 1 October 2009].
Spratt, C, Walker, R & Robinson, B 2004, Module A5: Mixed Research Method [Online],
Available: http://www.col.org/SiteCollectionDocuments/A5.pdf, [Accessed 4 May
2010].
Stickney, CP, Brown, PR & Wahlen, JM 2007, Financial Reporting, Financial Statement
Analysis, and Valuation: A Strategic Perspective, 6 edn, Thomson South-Western,
USA.
Sung, TK, Chang, N & Lee, G 1999, 'Dynamics of Modeling in Data Mining: Interpretive
Approach to Bankruptcy Prediction', Journal of Management Information Systems,
vol. 16, no. 1, pp. 63-86.
Taffler, RJ 1982, 'Forecasting Company Failure in the UK Using Discriminant Analysis and
Financial Ratio Data', Journal of the Royal Statistical Society, vol. 145, no. 3, pp.
342-58.
Tamari, M 1966, 'Financial Ratios as a Means of Forecasting Bankruptcy', Management

International Review, 4, pp. 15-21.
Temte, A 2004, Financial Statement Analysis, Dearborn Trade, A Kaplan Professional

Company.
TFEX n.d., About Thailand Futures Exchange [Online], Available:

http://www.tfex.co.th/en/about/about.html, [Accessed 6 October 2009].
208
Theodossiou, PT 1993, 'Predicting Shifts in the Mean of a Multivariate Time Series Process:
An Application in Predicting Business Failures', Journal of the American Statistical
Association, vol. 88, no. 422, pp. 441-9.
Tirapat, S & Nittayagasetwat, A 1999, 'An Investigation of Thai Listed Firms' Financial
Distress Using Macro and Micro Variables', Multinational Finance Journal, vol. 3, no.
2, pp. 103-25.
Tourism Authority of Thailand n.d., About Thailand [Online], Available:

http://www.tourismthailand.org/about-thailand/about-thailand-58-1.html, [Accessed
29 September 2009].
Travel Document Systems n.d., Thailand Economy [Online], Available:

http://www.traveldocs.com/th/economy.htm, [Accessed 30 September 2009].
Ugurlu, M & Aksoy, H 2006, 'Prediction of Corporate Financial Distress in an Emerging

Market: the Case of Turkey', An International Journal, vol. 13, no. 4, pp. 277-95.
United Nations Thailand n.d., Thailand Info [Online], Available:

http://www.un.or.th/thailand/geography.html, [Accessed 29 September 2009].
Vichyanond, P 2002, 'Capital Market Development in Thailand', Thailand Development

Research Institute.
Watson, CJ 1990, 'Multivariate Distributional Properties, Outliers, and Transformation of

Financial Ratios', The Accounting Review, July, vol. 65, no. 3, pp. 682-95.
Wilcox, JW 1973, 'A Prediction of Business Failure Using Accounting Data', Empirical
Research in Accounting: Selected Studies, Supplement to Journal of Accounting
Research, vol. 11, pp. 163-79.
Wilson, RL & Sharda, R 1994, 'Bankruptcy Prediction Using Neural Networks', Decision
Support Systems, vol. 11, pp. 545-57.
World Bank n.d., Map of Thailand [Online], Available:

http://siteresources.worldbank.org/INTTHAILAND/Images/333199-
1192650956297/THA32651.jpg, [Accessed 29 September 2009]. .
Zapranis, A & Ginoglou, D 2000, 'Forecasting Corporate Failure with Neural Network
Approach: The Greek Case', Journal of Financial Management and Analysis, vol. 13,
no. 2, pp. 11-20.
Zavgren, C 1983, 'The Prediction of Corporate Failure: the State of the Art', Journal of
Accounting Literature, vol. 2, no. 1, pp. 1-37.
Zikmund, WG 2003, Business Research Methods, 7th edn, Thomson South-Western, USA.
Zmijewski, M 1984, 'Methodological Issues Related to the Estimation of Financial Distress

Prediction Models', Journal of Accounting Research, vol. 22, pp. 59-82.
209
Appendices
Appendix A: Multivariate outliers test
ID MAH_1 MAH_2 MAH_3

1 12.07143 11.07692 10.08333
2 12.07143 11.07692 10.08333
3 12.07143 11.07692 10.08333
4 12.07143 11.07692 10.08333
5 12.07143 11.07692 10.08333
6 12.07143 11.07692 10.08333
7 12.07143 11.07692 10.08333
8 12.07143 11.07692 10.08333
9 12.07143 11.07692 10.08333
10 12.07143 11.07692 10.08333
11 12.07143 11.07692 10.08333
12 12.07143 11.07692 10.08333
13 12.07143 11.07692 Na
14 12.07143 na na
Note:
MAH_1 (14 firms, 10 quarters, p<0.001)
210
Appendix B: Statistical result for between group ttest
BETWEEN GROUP T TEST (Significance at 5 per cent)
CACL QR
Variable 1 Variable 2 Variable 1 Variable 2
Mean 1.618481948 2.15428382 Mean 1.250746375 1.515123734

Variance 1.384808224 3.858812703 Variance 1.165340898 2.727113699
Observations 267 267 Observations 267 267

Hypothesized Mean Difference 0 Hypothesized Mean Difference 0
df 435 df 458
t Stat -3.823351386 t Stat -2.189616786

P(T<=t) one-tail 7.54359E-05 P(T<=t) one-tail 0.014527922
t Critical one-tail 1.648364062 t Critical one-tail 1.648187415
P(T<=t) two-tail 0.000150872 P(T<=t) two-tail 0.029055844
t Critical two-tail 1.96543235 t Critical two-tail 1.965157018
STA ExpS
Mean 0.18592002 0.153055518 Mean 1.013280603 0.986445789


df 469 df 414
t Stat 4.371449984 t Stat 0.981999794


P(T<=t) two-tail 1.52132E-05 P(T<=t) two-tail 0.326673465
SINV TLTA
Mean 4.942886855 6.338269063 Mean 0.378000131 0.522111775

df 271 df 429
t Stat -1.191916382 t Stat -6.149457206
P(T<=t) one-tail 0.117168375 P(T<=t) one-tail 8.90767E-10

P(T<=t) two-tail 0.234336751 P(T<=t) two-tail 1.78153E-09
211
TLTE EBITI
Mean 1.052807235 1.392163118 Mean 28.89201228 0.657916267

df 228 df 90
t Stat -1.845940903 t Stat 2.82389457

P(T<=t) one-tail 0.033098577 P(T<=t) one-tail 0.002920614
EBITS NIS
Mean 0.066958012 0.378391324 Mean 0.038346016 0.354687883

df 276 df 273
t Stat -3.211723928 t Stat -3.105381986

EBITTA NITA
Mean 0.015295846 0.018452064 Mean 0.010202483 0.022769449

Pooled Variance 0.001794663 Hypothesized Mean Difference 0
Hypothesized Mean Difference 0 df 367

df 532 t Stat -1.933348386
t Stat -0.86082734 P(T<=t) one-tail 0.026981056
P(T<=t) one-tail 0.194860467 t Critical one-tail 1.649016151

t Critical one-tail 1.64772288 P(T<=t) two-tail 0.053962112
P(T<=t) two-tail 0.389720934 t Critical two-tail 1.966448874
t Critical two-tail 1.964433045
212
NITE WCTA
Mean 0.011211206 0.0304674 Mean 0.08710939 0.003654652

df 331 df 431
t Stat -2.358954428 t Stat 4.421956074

RETA TETL
Mean -0.100282169 -0.104997301 Mean 2.81447625 2.819641084

df 519 df 479
t Stat 0.094256889 t Stat -0.016056103

BETWEEN GROUP T TEST (Significance at 10 per cent)

CACL QR

Mean 1.618481948 2.15428382 Mean 1.250746375 1.515123734
df 435 df 458
t Stat -3.823351386 t Stat -2.189616786
213
STA ExpS

Mean 0.18592002 0.153055518 Mean 1.013280603 0.986445789
df 469 df 414
t Stat 4.371449984 t Stat 0.981999794
P(T<=t) two-tail 1.52132E-05 P(T<=t) two-tail 0.326673465
SINV TLTA

Mean 4.942886855 6.338269063 Mean 0.378000131 0.522111775
df 271 df 429
t Stat -1.191916382 t Stat -6.149457206
TLTE EBITI
Mean 1.052807235 1.392163118 Mean 28.89201228 0.657916267
df 228 df 90
t Stat -1.845940903 t Stat 2.82389457
214
EBITS NIS

Mean 0.066958012 0.378391324 Mean 0.038346016 0.354687883
df 276 df 273
t Stat -3.211723928 t Stat -3.105381986
EBITTA NITA

Mean 0.015295846 0.018452064 Mean 0.010202483 0.022769449
Pooled Variance 0.001794663 Hypothesized Mean Difference 0
Hypothesized Mean Difference 0 df 367
df 532 t Stat -1.933348386
t Stat -0.86082734 P(T<=t) one-tail 0.026981056
P(T<=t) one-tail 0.194860467 t Critical one-tail 1.28386258
t Critical one-tail 1.283144909 P(T<=t) two-tail 0.053962112
P(T<=t) two-tail 0.389720934 t Critical two-tail 1.649016151
t Critical two-tail 1.64772288
NITE WCTA

Mean 0.011211206 0.0304674 Mean 0.08710939 0.003654652
df 331 df 431
t Stat -2.358954428 t Stat 4.421956074
215
RETA TETL

Mean -0.100282169 -0.104997301 Mean 2.81447625 2.819641084
df 519 df 479
t Stat 0.094256889 t Stat -0.016056103
216

The Role of Financial Ratios in Signalling Financial Distress - Ev

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Role of Financial Ratios in Signalling Financial Distress - Ev

Uploaded by

Copyright:

Available Formats

Southern Cross University

The role of financial ratios in signalling financial

Bachelor of Business Administration, Ramkhamhaeng University, Thailand

Master of Business Administration, Mount Saint Marys University, USA

A thesis submitted to the Graduate College of Management, Southern Cross University,

Thirdly, my sincere appreciation is addressed to staff of the Graduate College of Management

Key words: financial statement information/ratios, business failure/financial distress, Thai

Statement of original authorship ......................................................................................... i

Chapter 1: Introduction ...................................................................................................... 1

Chapter 2: The Thai Environment ................................................................................... 15

Chapter 3: Literature Review ........................................................................................... 28

Chapter 4: Research Methodology ................................................................................... 58

Chapter 5: Data Analysis ................................................................................................ 104

Chapter 6: Conclusions and Implications ...................................................................... 179

List of References ............................................................................................................ 200

Appendices ...................................................................................................................... 210

1.2 Background to the research

1.3 Research issues

1.3.1 Research problem 1.3.3 Research objective

1.3.2 Research question

1.4 Justification for the research

1.6 Principal definitions used

1.7 Outline of this study

Source: Developed for this thesis

1.2 Background to the research

1.3 Research issues

Whether financial statement information can be adequately used to predict

1.3.2 Research question

Can financial statement information be used to discriminate between potentially

The research question is developed in response to the research problem. In addition,

To investigate whether financial statement information can adequately classify

The study expands the body of knowledge about financial statementsbased

1.4 Justification for the research

Data collection method

Data analysis method

Brief summary of the findings

As can be seen, there are a number of conditions representing financial distress in

1.7 Outline of this study

Figure 1.2 Outline of this study

Chapter 2: The Thai Environment

Chapter 3: Literature Review

Chapter 4: Research Methodology

Chapter 5: Data Analysis

Chapter 6: Conclusions and Implications

Source: Developed for this thesis

Chapter 1 provides an introductory overview of the study. Background to the research is

2.2 Thailand in general

2.2.1 Geography 2.2.3 Government/politics

2.2.2 Population 2.2.4 Economy

2.3 Thai financial markets and key market players

2.4 Thai business failure

Source: Developed for this thesis

2.2 Thailand in general

Figure 2.2 Map of Thailand

Source: World Bank (n.d.)

2.3 Thai financial markets and key market players

The Foreign Exchange Market

The Derivatives Market

The Debt Market

o The Market for Alternative Investment (MAI)

o The Stock Exchange of Thailand (SET)

Table 2.1 SET statistical highlights from 2006 to 2009

2006 2007 2008 2009