Southern Cross University

ePublications@SCU
Theses

2005

Cash flows and accrual accounting in predicting future cash flows of Thai listed companies
Porntip Chotkunakitti
Southern Cross University, [email protected]

Suggested Citation
Chotkunakitti, P 2005, 'Cash flows and accrual accounting in predicting future cash flows of Thai listed companies', DBA thesis, Southern Cross University, Lismore, NSW. Copyright P Chotkunakitti 2005

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 [email protected].

Cash Flows and Accrual Accounting in Predicting Future Cash Flows of Thai Listed Companies

Porntip Chotkunakitti BBA (Accounting), MBA (Accounting)

A thesis submitted to the Graduate College of Management of Southern Cross University, Australia, in partial fulfilment of the degree of Doctor of Business Administration. 2005

Statement of Original Authorship

The work contained in this dissertation has not been previously submitted for a degree or diploma at any other higher education institution.

I also certify that to the best of my knowledge, the dissertation contains no material previously published or written by another person except where references have been given, and any help received has been acknowledged.

Signed: Porntip Chotkunakitti

Date:..

ii

Acknowledgments
I wish to express my profound gratitude to all who have supported the completion of this thesis. First, I would like to especially thank my supervisor, Associate Professor Michael D. Evans for his valuable advice, full support, patience and encouragement. Without him, this thesis would not have been completed. Additionally, I indebted to Dr Mark Manning for his immense help in statistical techniques and data analysis.

My special thanks also go to the staff of the SCU Graduate College of Management at Tweed campus, Sue White, Susanna, Di Clarke for their administrative help.

My sincere gratitude goes to Rosemary Graham for her editing and recommendation on English in earlier draft of my thesis. Additionally I appreciate Gita Sankaran for kindly editing and providing suggestions on English.

I acknowledge Dr Pranee Leksrisakul for providing secondary data involving the Stock Exchange of Thailand. I also acknowledge my friends in the DBA program for the good friendship and help during my study in Australia.

I wish to thank Associate Prof. Dr Aekkachai Nittayagasetwat, Assistant Prof. Dr Pradit Withisupakorn, Assistant Prof. Dr Sasivimon Sricharanjit and Dr Rojanasak Chomvilailuk for their helpful encouragement and suggestions. I also wish to thank my College for help and support while I was writing my thesis in Thailand. Also the partial grant from Rangsit University is acknowledged with thanks.

My deepest gratitude goes to my parents for their caring and unbounded support. Without their support, I would have had no chance to study this program.

iii

Abstract
Cash flow prediction is involved in a number of economic decisions, particularly in investment. Previous research conducted in the United States has provided inconsistency in the results of investigating accounting data, cash flow and accrual accounting data in predicting future cash flows. No published research has studied cash flow prediction in Thailand. The current study investigates the ability of accrual and cash flows accounting data to predict future cash flows of Thai listed companies.

Three regression models are constructed namely earnings, cash flows, accrual components and cash flows models. In addition, cash flow ratios are investigated to predict future cash flows by using a stepwise regression. Data used in this study is collected from the financial statements of non-financial companies listed on the Stock Exchange of Thailand from 1994 to 2002. Cash flow data are selected directly from the cash flow statements.

The empirical results show that past earnings, cash flows, cash flow and accrual component of earnings can be used to predict future cash flows of Thai listed companies and cash flows have better predictive power than past earnings. Additionally, the cash flow model and the cash flow and accrual components of earnings model have better predictive power than the earnings model. The findings of testing the models in an out-of-sample period suggest that the cash flow model is a better predictor of future cash flows than the other models. Furthermore, additional year lags of accounting data can improve the predictive power of the model. However, the results indicate that cash flow ratios are not a good predictor of future cash flows. In addition, this study finds that the Asian economic crisis had an impact on the predictive power of accounting data.

Key Words: Cash flow prediction, Cash flow, Accrual, Earnings

iv

Abbreviations

BOT CFO EIU FASB IAS IASC ICAAT NI NIDPR SEC SET TAS TBDC UK US $ USA WCFO

The Bank of Thailand Cash Flow from Operations The Economist Intelligence Unit Limited The Financial Accounting Standards Board The International Accounting Standard The International Accounting Standard Committee The Institute of Certified Accountants and Auditors of Thailand Net Income Net Income plus Depreciation and Amortisation The Securities and Exchange Commission The Stock Exchange of Thailand Thai Accounting Standard The Thai Bond Dealing Center United Kingdom US dollar (The United States of America currency) United States of America Working Capital from Operations

Table of Contents
Statement of Original Authorship ...........................................................ii Acknowledgments ....................................................................................iii Abstract ....................................................................................................iv Abbreviations ............................................................................................v Table of Contents .....................................................................................vi List of Figures ...........................................................................................x List of Tables............................................................................................xi

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

1.1 Introduction.................................................................................................1 1.2 Background to the Research.......................................................................1 1.3 Research Issues............................................................................................4 1.4 Justification of the Research.......................................................................5 1.5 Methodology................................................................................................6 1.5.1 Data collection ...................................................................................7 1.5.2 Sample selection ................................................................................7 1.5.3 Variables and measurement................................................................7 1.5.4 Prediction model ................................................................................8 1.6 Keywords and Abbreviations .....................................................................9 1.7 Contribution of the Research....................................................................10 1.8 Limitations of the Proposed Research......................................................11 1.9 Structure of the Thesis ..............................................................................11

Chapter 2: Thai Literature ................................................................... 14

2.1 Introduction...............................................................................................14 2.2 General Background to Thailand.............................................................16 2.2.1 Country profile.................................................................................16 2.2.2 Thailands economy .........................................................................17 2.3 Thai Accounting Standards ......................................................................21

vi

2.3.1 The Institute of Certified Accountants and Auditors of Thailand (ICAAT)..........................................................................................21 2.3.2 Development of Accounting Standards in Thailand ..........................21 2.4 The Stock Exchange of Thailand (SET) ...................................................25 2.4.1 Roles of the Stock Exchange of Thailand .........................................25 2.4.2 Information Disclosure in the SET ...................................................28 2.5 Conclusion .................................................................................................30

Chapter 3: Literature Review ............................................................... 32

3.1 Introduction...............................................................................................32 3.2 Cash Flow Prediction ................................................................................34 3.3 Accrual Accounting Data..........................................................................38 3.3.1 Accrual accounting basis..................................................................38 3.3.2 Usefulness of accrual accounting data ..............................................41 3.4 Cash Flow Accounting Data .....................................................................44 3.4.1 Cash flow accounting data................................................................44 3.4.2 Cash flow statement .........................................................................45 3.4.3 Usefulness of cash flow data ............................................................48 3.4.4 Cash flow ratios ...............................................................................51 3.5 Application of Cash Flows and Accrual Accounting Data in Predicting Future Cash Flow......................................................................................56 3.5.1 Accrual based and cash flow based data in predicting future cash flow .................................................................................................56 3.5.2 Combined approach: cash flow and accrual-based approach.............64 3.5.3 Other studies ....................................................................................74 3.6 Research Gaps, Research Issues and Model of Conceptual Framework 78 3.6.1 Research gaps ..................................................................................78 3.6.2 Research problem.............................................................................79 3.6.3 Research questions...........................................................................79 3.6.4 Research hypothesis .........................................................................80 3.6.5 Conceptual framework .....................................................................81 3.7 Conclusion .................................................................................................83

vii

Chapter 4: Methodology........................................................................ 84
4.1 Introduction...............................................................................................84 4.2 Justification of the Research Paradigm....................................................86 4.3 Research Design ........................................................................................89 4.3.1 The purpose of the study ..................................................................89 4.3.2 The basic research method ...............................................................90 4.4 Variables and Measurement .....................................................................90 4.4.1 Dependent variable: future cash flows ..............................................91 4.4.2 Independent variables.......................................................................92 4.5 Model Building..........................................................................................95 4.5.1 Earnings model ................................................................................97 4.5.2 Cash flows model.............................................................................98 4.5.3 Cash flows and accrual components of earning model......................99 4.5.4 Cash flow ratios model................................................................... 101 4.6 The Explanatory Ability of Models ........................................................ 102 4.7 Evaluation of the Predictive Ability of the Models in Out-of-sample ... 102 4.8 Hypothesis Testing .................................................................................. 103 4.9 Data Specification ................................................................................... 105 4.9.1 Testing periods............................................................................... 105 4.9.2 Sampling selection ......................................................................... 106 4.9.3 Sources of data............................................................................... 107 4.10 The Validity of the Research................................................................. 107 4.11 Conclusion ............................................................................................. 108

Chapter 5: Data Analysis .....................................................................109

5.1 Introduction............................................................................................. 109 5.2 Regression Equations and Models.......................................................... 111 5.3 Data Preparation..................................................................................... 112 5.3.1 Data arrangement ........................................................................... 112 5.3.2 Missing data................................................................................... 113 5.3.3 Outliers .......................................................................................... 115 5.3.4 Multicolinearity ............................................................................. 115 5.4 Preliminary Analysis............................................................................... 116

viii

5.4.1 Descriptive statistics ...................................................................... 116 5.4.2 Correlation analysis........................................................................ 119 5.5 Regression Results................................................................................... 125 5.5.1 Earnings model .............................................................................. 125 5.5.2 Cash flows model........................................................................... 131 5.5.3 Cash flows and accrual components of the earnings model............. 137 5.5.4 Cash flow ratios model................................................................... 145 5.6 Comparison of Explanatory Ability among Models .............................. 148 5.7 Evaluation of the Predictive Ability of the Models with Different Approaches .............................................................................................. 150 5.8 Test of Hypotheses .................................................................................. 155 5.9 Additional Analysis ................................................................................. 158 5.9.1 Test for the predictive ability of disaggregated accrual components of earnings..................................................................................... 158 5.9.2 Pooled-year analysis excluding prediction year 1998 (economic crisis)............................................................................................. 163 5.10 Conclusion ............................................................................................. 167

Chapter 6: Conclusion..........................................................................169
6.1 Introduction............................................................................................. 169 6.2 Conclusion of the Research Findings ..................................................... 171 6.3 Implications for Theory .......................................................................... 181 6.4 Implications for Policy and Practice....................................................... 182 6.5 Limitations of the Research Study ......................................................... 183 6.6 Suggestions for Future Research ............................................................ 184 6.7 Conclusion ............................................................................................... 185

References

ix

List of Figures
Figure 1.1 Outline of the Thesis ..................................................................................13 Figure 2.1 Structure of Chapter 2: Thai Literature.......................................................15 Figure 3.1 Structure of Chapter 3: Literature Review...................................................33 Figure 3.2 Model of Predicting Future Cash Flows using Earnings..............................81 Figure 3.3 Model of Predicting Future Cash Flows using Cash Flows .........................82 Figure 3.4 Model of Predicting Future Cash Flows using Cash Flows and Accrual Components of Earnings..............................................................................82 Figure 3.5 Model of Predicting Future Cash Flows using Cash Flow Ratios................83 Figure 4.1 Structure of Chapter 4: Methodology .........................................................85 Figure 5.1 Structure of Chapter 5: Data Analysis ...................................................... 110 Figure 6.1 Structure of Chapter 6: Conclusions and Implications .............................. 170

List of Tables
Table 2.1 Thai Economic Indicators 1994-2003 ..........................................................18 Table 2.2 Summary comparing between Thai Accounting Standards and International Accounting Standard, as of 30 January 2002 (www.icaat.or.th) ...................23 Table 2.3 Outstanding Trading Statistics of SET from 1994 to 2002 ...........................27 Table 3.1 Summary of Studies Investigating Earnings and Cash Flow Data as a Predictor of Future Cash Flows ...................................................................63 Table 3.2 Summary of Studies using Cash flow and Accrual Based Approach, to Predict Future Cash Flow ............................................................................72 Table 4.1 Summary of Four Research Paradigms ........................................................89 Table 4.2 Cash Flow Ratios, Symbols and Calculation................................................95 Table 4.3 Annual Data-matching in Different Periods between the Predictors and Future Cash Flows..................................................................................... 106 Table 4.4 Number of Companies in Each Year.......................................................... 107 Table 5.1 Summary of Regression Equations for All Models in Data Analysis.......... 111 Table 5.2 Matching Years in the Analysis ................................................................. 113 Table 5.3 Number of Companies in the Sample for Each Year .................................. 114 Table 5.4 Number of Cases in Data Analysis for Each Prediction Year ..................... 114 Table 5.5 Descriptive Statistics of Cash Flow from Operations, Earnings and Accrual Component of Earnings. ............................................................................ 118 Table 5.6 Pearson Correlation Coefficients of Cash Flow from Operations for Each Year During 1994-2002............................................................................. 120 Table 5.7 Pearson Correlation Coefficients Among Earnings for Each Year During 1994- 2001 ................................................................................................ 121 Table 5.8 Pearson Correlation Coefficients Among Accrual Component of Earnings for Each Year During 1994- 2001.............................................................. 121 Table 5.9 Pearson Correlation Coefficients Between Variables for Each Year During 1994-2002 ................................................................................................. 122 Table 5.10 Pearson Correlation Coefficients of a Pair of Variables for Pooled Year During 1994- 2002 .................................................................................... 124 Table 5.11 Summary Statistics from Regressions of Future Cash Flow on Earnings During 1994 to 2002.................................................................................. 129

xi

Table 5.12 Summary Statistics from Regressions of Future Cash Flows on Cash Flows During 1994 to 2002 ....................................................................... 136 Table 5.13 Summary Statistics from Regressions of Future Cash Flows on Cash Flows and Accrual Components of Earnings.............................................. 142 Table 5.14 Summary Statistics from Regressions of All Models (Pooled-year Data) . 144 Table 5.15 Summary Statistics from Stepwise Regressions of Future Cash Flows on Cash Flow Ratios ...................................................................................... 147 Table 5.16 Summary of Coefficient of Determination (Adjusted R2) for Each Model 150 Table 5.17 r2 (r) and MAPE for Each Model.............................................................. 151 Table 5.18 Summary of the Values of r2 (r) and MAPE for Each Model (pooled year)154 Table 5.19 Summary of Statistical Results of Regression Analysis from the Cash Flows and Disaggregated Accrual Components of Earnings Model for Each Prediction Year from 1995 to 2002............................................................ 161 Table 5.20 Summary of the Statistical Results from Pooled Year Analysis of Cash Flows and Disaggregated Accrual Components of Earnings Model ........... 162 Table 5.21 Comparison of Adjusted R2 between Cash Flows and Aggregated Accrual Components of Earnings Model and Cash flows and Disaggregated Accrual Components of Earnings Model................................................................. 162 Table 5.22 Summary Statistics from Regressions of Future Cash Flows on Predictors (Pooled-year Data Excluding Prediction Year 1998).................................. 166 Table 5.23 Comparison of Adjusted R2 between Analysis of All Prediction Years and Analysis Excluding Prediction Year 1998.................................................. 167

xii

Chapter 1
Introduction to the Study
1.1 Introduction This chapter provides an overview and describes the structure of this thesis. Nine sections are contained in this chapter. Section 1.1 introduces the chapter. Section 1.2 provides background to the research and the context and the research objective which establishes the purpose of the study. Section 1.3 presents research issues including the research problem, questions and hypotheses. Section 1.4 justifies the research study. Section 1.5 outlines the research methodology. Section 1.6 defines keywords and abbreviations used specifically for this study. Section 1.7 outlines the contribution of the research findings while section 1.8 considers potential limitations. Section 1.9 delineates the structure of the thesis.

1.2 Background to the Research

Cash flow prediction is an important task since it is involved in various economic decisions. Investors, for example, need information about future cash flows, because the value of their investment is the present value of the future cash flows to them, through investing in a company. In the same way, the ability of a company to generate cash flows is reflected in the value of its shares. Research done in Thailand suggested that cash flows have value relevance to stock prices in the Thai stock market (Narktabtee 2000). Thus predicting future cash flows allows investors to predict stock prices. In addition, the Financial Accounting Standard Board (FASB) suggested that financial reporting can help users assess future cash flows (FASB 1978). Cash flow plays a pivotal role in all of these issues. Most researchers have attempted to investigate the predictive ability of earnings under an accrual accounting basis and cash flows, in predicting future cash flows (Neill et al. 1991). FASB (1978) asserted that earnings are a better predictor of future cash flows than cash flows themselves. However, previous research findings have shown inconclusive results. Some research has concluded that the predictive ability of earnings outperforms that of

cash flows in forecasting future cash flows, for example, the studies of Greenberg, Johnson and Ramesh (1986), and Dechow, Kothari and Watts (1998). In contrast, some findings showed conflicting results in which cash flows are the better predictor of future cash flows, such as the studies of Finger (1994), Bowen, Burgstahler and Daley (1986), Percy and Stokes (1992). However the study by McBeth (1993) rejected both conclusions and claimed that neither cash flows nor earnings are a good predictor of future cash flows. In addition to single variable testing, some researchers have focused on multiple variables, such as the components of earnings including cash flow and accrual accounting data (Barth, Cram & Nelson 2001; Stammerjohan & Nassiripour 2000/2001). Barth, Cram and Nelson (2001) used a simple time series model to test the relationship between accrual components of earnings and future cash flows. They concluded that each accrual component reflected different information relating to future cash flows. Stammerjohan & Nassiripour (2000/2001) replicated Barth, Cram and Nelsons study and both these studies provided evidence that models based on cash flows and total accruals obtained a superior predictor of future cash flows over models based solely on earnings. However Stammerjohan and Nassiripours study provided weak evidence that prediction models based on cash flow and total accruals outperform models based only on cash flows.

Moreover, most research has focused narrowly on operating cash flow, earnings and accrual components of earnings. Those previous studies have ignored the potential of other cash flow variables, particularly cash flow ratios. Cash flow ratios are calculated by using data from both the cash flow statement prepared on a cash basis and the income statement and balance sheet based on the accrual basis. A cash flow ratio is a tool for analysing a firms performance (Giacomino & Mielke 1988; Mills & Yamamura 1998; Plewa & Friedlob 1995; Reilly 1994), particularly cash flow return ratios that indicate the ability of a firm to generate its cash flows (Plewa & Friedlob 1995). In addition, cash flow ratios can be used in predicting bankruptcy, determining bond ratings and estimating stock return (Donleavy 1994). No research has investigated the ability of cash flow ratios to predict future cash flows.

In respect to the cash flow data investigated in previous research, some researchers have estimated cash flows by adjusting income from income statements, whereas little research has used actual cash flow data directly derived from statements of cash flow (McBeth 1993; Stammerjohan & Nassiripour 2000/2001). To investigate the usefulness of the cash flow statement, the cash flow data employed in research analysis should be directly derived from the cash flow statements instead of proxy cash flow measures calculated by using data from accrual based statements. Cash flow prediction studies have been mainly based on data from the United States. There is little Thai research providing evidence on the usefulness of cash flow statements particularly in regard to the cash flow prediction issue. United States companies have been mandated to report the statement of cash flow since 1988 under the Financial Accounting Standard No. 95 (FAS 95). In the case of Thailand, Thai listed companies have disclosed cash flow statements since 1994 based on Thai accounting standard No. 25 (TAS 25). In practice, there are many differences in reporting cash flow in each country (Donleavy 1994). In the case of Thailand, Thai accounting standards are based on International Accounting Standards. Although international standards are likely to be similar to the United States standards, in practice, some of the international accounting standards differ to the US standards, such as the presentation of cash flows from operating activities and components of cash and cash equivalents. Consequently, the available cash flow data are likely to differ in each country and the findings of Thai research may not correspond directly with that of United States-based research.

This research aims to fill the gaps of prior research as mentioned above. The study reexamines the predictive ability of cash flows and accrual accounting data and extends the analysis by investigating the ability of cash flow ratios to predict future cash flows. The subject of this research is companies listed on the Stock Exchange of Thailand. Cash flow data used in the analysis is derived from cash flow statements.

1.3 Research Issues Research problem

This research intends to replicate previous studies by focusing on cash flow ratios and using cash flow information directly from cash flow statements. It aims to solve the following research problem:

How can cash flow and accrual accounting data be used to predict future cash flow of Thai listed companies?

Research Questions Several research questions established from the research problem are defined below. Research Question 1. Are past earnings significant predictors of future cash flows of Thai listed companies?

Research Question 2. Are past cash flows significant predictors of future cash flows of Thai listed companies?

Research Question 3. Are past cash flows and accrual components of earnings significant predictors of future cash flows of Thai listed companies?

Research Question 4. Are there different predictive powers between three prediction models, earnings, cash flows and cash flows and accrual components of earnings models?

Research Question 5. Are past cash flow ratios significant predictors of future cash flows of Thai listed companies?

Research hypotheses The following research hypotheses have been formulated in relation to the research questions.

Research Hypothesis 1: Past earnings have significant predictive power in predicting future cash flows of Thai listed companies. Research Hypothesis 2: Past cash flows have significant predictive power in predicting future cash flows of Thai listed companies. Research Hypothesis 3: Past cash flows and accrual components of earnings have significant predictive power in predicting future cash flows of Thai listed companies.

Research Hypothesis 4: There are different predictive powers between three prediction models: earnings, cash flows and cash flows and accrual components of earnings models.

Research Hypothesis 5: Past cash flow ratios are significant predictors of future cash flows of Thai listed companies.

1.4 Justification of the Research

Cash flow prediction is an important purpose for reporting financial statements, as suggested by the Financial Accounting Standard Board (FASB 1978). Moreover, cash flows are involved in various economic decision contexts such as liquidity and solvency evaluation,

performance evaluation, monitoring evaluation and in a prediction function (Jones & Widjaja 1998). Research conducted in Thailand suggested that cash flows have value relevance to stock price in the Thai Stock Market (Narktabtee 2000). Therefore if investors can predict future cash flows of a company, they can predict share prices of that company.

Many studies have investigated the ability of cash flow and accrual accounting data to predict future cash flows, however, their findings are inconsistent. In addition, most research has been undertaken in the United States. There has been no published research undertaken in Thailand.

In the case of Thailand, since 1994 Thai listed companies have been mandated by the Stock Exchange of Thailand to release a statement of cash flow based on Thai Accounting Standard No. 25. Companies outside the stock market are not required to report this statement because of doubt regarding the usefulness of cash flow statements and the cost of preparation, as there is little evidence on the usefulness of the cash flow statement, in the case of Thailand. This makes users of financial statements unwilling to use cash flow information in their decisions. Cash flow prediction research will provide an example of the usefulness of cash flow data, in the case of Thailand. Furthermore, standard-setters may be able to use the results of this research to develop or design their standards. Also, the findings of this research can help the Thai government regulate companies and consider whether unlisted companies should report statements of cash flow.

These uses underscore the need to precisely define the ability of cash flow and accrual accounting data to predict future cash flows, in the case of Thai listed companies.

1.5 Methodology This research aims to test the stated hypotheses and employs a quantitative method. Because the research aims to investigate the ability of accounting data to predict future cash flows, secondary data analysis will be used as the basis of the study to develop the prediction models. Secondary data have been collected for other purposes and can be used for the present

research (Ticehurst & Veal 2000). Usually, secondary data are used in research that aims to test theory, particularly in accounting and finance areas.

1.5.1 Data collection

Data is collected from financial statements of companies listed on the Stock Exchange of Thailand (SET) obtained on the SETs Listed Company Info CD-ROM for annual reports 1994 to 2002. To focus on the statements of cash flows, cash flows from operating activities are selected directly from the cash flow statements. Earnings are derived from income statements. Total assets, sales and other variables are selected from balance sheets and income statements.

1.5.2 Sample selection The sample used for this research comprises non-financial companies listed on the stock exchange of Thailand, except for companies under a rehabilitation plan, from 1994 to 2002. The samples are selected according to the following criteria: The companies must have completed data for all variables such as cash flow statements, income statements and balance sheets. The companies must have operated during the fiscal year ended December 31.

1.5.3 Variables and measurement Based on the research hypotheses, this study focuses on five major variables: future cash flows, earnings, cash flows, accrual accounting data and cash flow ratios. In the prediction model, future cash flows are investigated as a dependent variable caused by independent variables. Since this research intends to investigate the usefulness of cash flow statements based on Thai Accounting Standard No. 25 (TAS 25), in the analysis, future cash flows are defined as net cash flows from operation (CFO) reported on cash flow statements for the current year. Independent variables are earnings, cash flows, accrual components of earnings and cash flow ratios. Net income before extraordinary items and discontinued operations derived from income statements for the previous years is used as the measure of past earnings 7

(Cheng, Liu & Schaefer 1997; Greenberg, Johnson & Ramesh 1986). The definition of past cash flow variable is net cash flow reported on statements of cash flows for the previous years. Past accrual components of earnings (ACR) are defined in two measurements, aggregated accrual components and disaggregated accrual components similar to Barth, Cram and Nelson (2001) and Stammerjohan and Nassiripour (2000/2001). The accrual components of earnings are obtained from cash flow statements for the previous years. This research aims to investigate the abilities of cash flow ratios to predict future cash flow. Even though there is no empirical evidence supporting these ratios, this study initiated these ratios as a sign of future cash flows. Ten cash flow ratios are tested: cash flow adequacy ratio; debt coverage ratio; repayment of borrowings ratio; dividend payment ratio; reinvestment ratio; cash flow on revenues; cash flow to net income ratio; cash flow return on assets; cash flow return on stockholders equity ratio and cash flow per share ratio (see Chapters 3 and 4). These cash flow ratios are calculated by using data from income statements, balance sheets and cash flow statements.

1.5.4 Prediction model The relationship between the dependent and independent variables is analysed by using regression models. Regression models have been used by a number of researchers such as Barth, Cram and Nelson (2001), Finger (1994), Greenberg, Johnson and Ramesh (1986), McBeth (1993), Murdoch and Krause (1989) and Stammerjohan and Nassiripour (2000/2001). This research builds four models based on the literature review. The first model investigates the relationship between earnings and future cash flows. The second model examines whether cash flows from operations are significant in predicting future cash flows. The third model tests whether cash flow and accrual components of earnings have a positive, significant capacity for predicting future cash flows. This model consists of two dependent variables: cash flows from operations and accrual components of earnings. The final model consists of ten cash flow ratios. This procedure aims to investigate whether cash flow ratios provide a good predictor of future cash flows. The technique of stepwise regression is employed to select and place cash flow ratios into the model.

After analysing the relationship between predictor variables and future cash flows, the comparison between the predictive abilities of the three models (Models 1, 2 and 3) is conducted. Correlation between actual and predicted cash flows and mean absolute percentage errors, calculated by using data from out-of-samples, are used to assess predictive ability. The model producing the highest correlation and smallest mean absolute percentage error is interpreted as being a better predictor.

1.5.5 Analytical techniques

This research utilises quantitative methods in which the data is analysed based on statistical techniques, which include descriptive statistics, Pearsons correlation and regression analysis. These descriptive statistics provide an initial summary data of the essential features of the sample. The correlation analysis is used to fundamentally examine the relationship between dependent and independent variables. Regression analysis, both simple linear and multiple regression, is applied to test the prediction models depending upon the ability of predictor variables to explain future cash flows. All analytical techniques use the computer software package Statistical Processing for Social Scientists (SPSS) Version 11.5.

1.6 Keywords and Abbreviations Key definitions relevant to this research study are briefly identified and listed in alphabetical order as follows:

Accrual basis: a principal accounting assumption dealing with the accounting process of recognising the effects of financial transactions in the period in which events occur rather than when the cash is received or paid.

Accrual component of earnings (ACR): earnings can be disaggregated into cash flow and the component of accruals. Accrual component include: change in accounts receivable;

change in accounts payable; change in inventory; change in other short-term assets and liabilities and depreciation and amortisation.

Cash flow (CFO): net cash received and cash paid from operations.

Cash basis: a principal accounting assumption dealing with the accounting process of recognising the effects of financial transactions when cash is received or paid.

Cash flow prediction: forecasting future cash flow from operations by using past accounting information.

Cash flow ratios (CFRs): financial ratios based on cash flow data from cash flow statements.

Cash flow statement: a report of the amounts of cash received or paid during a fiscal period.

Earnings (EARN): net income before extraordinary items and discontinued operations are presented on income statements.

Future cash flow: future net cash inflow and cash outflow from operations of a company.

Listed company: a company of which ordinary shares are listed on the Stock Exchange of Thailand and which has its ordinary shares listed on the Stock Exchange and is required by the Stock Exchange to have its securities traded on the market for alternative investment. 1.7 Contribution of the Research The benefits stemming from this study are as follows: 1. The results of this research will assist users of financial information to make decisions. Investors, creditors and auditors can use cash flow prediction models to assess future cash flows of companies in evaluating their future liquidity, solvency and other performance. 2. The results of this research will provide evidence of the usefulness of the cash flow statements of Thai listed companies.

10

3. Regulatory agencies, such as the Securities and Exchange Commission (SEC) can use the results of the study to formulate policy and decide what information should be provided to the general public. 4. This research will provide a theoretical and practical guide for future research.

1.8 Limitations of the Proposed Research There are three possible limitations of this research, which call for clarification in the following specific areas. These limitations may affect the generalisability and validity of this research.

Firstly, due to its focus on actual cash flow data from the statement of cash flows, this research studies only Thai companies listed on the Stock Exchange of Thailand. These companies have reported statements of cash flow since 1994. The data employed in this research is only available for the years ended 31 December 1994 to 2002. Therefore, this research may experience problems due to the inadequacy of the data.

Secondly, this research focuses solely on companies listed on the Stock Exchange of Thailand, which are the companies required by the Stock Exchange of Thailand (SET) to publish cash flow statements. Thai companies which are not listed on the stock exchange have not reported statements of cash flow. Therefore, the results of this research may not be generalised to unlisted Thai companies. Finally, regression analysis is used to construct a prediction model because it has been used in much of the prior research. However, time series analysis (Box & Jenkins 1976) is another approach that can be used in prediction research. This proposed research will not provide evidence on time series properties analysed by Box & Jenkins methodology. 1.9 Structure of the Thesis This research study contains six chapters. The first chapter provides an introduction to the study. The theoretical framework upon which the study is based is presented in the next two

11

chapters. Thai literature on the general background of Thailand, the Thai economy, Thai accounting standards and the Stock Exchange of Thailand is reviewed in Chapter 2, which provides the specific details of Thai listed companies. Thai accounting standards involving cash flow statements and the Thai accounting framework are also discussed in Chapter 2. Chapter 3 reviews the concepts relevant to the cash flows and accrual accounting data, as well as prior empirical research in predicting future cash flows. In these chapters, existing theories and concepts enable the researcher to identify research gaps and lead to the development of the research issues.

The methodology chosen to test the model is described in Chapter 4 in which the data correction and research design are explained. In addition, the analytical techniques used to gain the results are also provided. The actual results of the data analysis and hypothesis testing are presented in Chapter 5. These results contribute to the findings of research which are discussed in Chapter 6. The conclusion of the research findings and implications for the research are also discussed in Chapter 6, including areas for possible further research. Finally, Chapter 6 contains the conclusion to the study. The organisation of this thesis is depicted in Figure 1.1.

The next chapter provides a general background to Thailand including its economy, accounting standards and Stock Exchange.

12

Figure 1.1 Outline of the Thesis

Chapter 1 Introduction

Chapter 2 Thai Literature

Chapter 3 Literature Review

Chapter 4 Methodology

Chapter 5 Data Analysis

Chapter 6 Conclusion

13

Chapter 2
Thai Literature
2.1 Introduction The aim of this research is to determine the way in which cash flow and accrual accounting data can be used to predict future cash flow in the case of Thai listed companies. The purpose of this chapter is to provide information about Thailand and Thai companies. As, this study deals with accounting data prepared under Thai accounting standards, the background of Thai accounting standards are also introduced in this chapter. Also, as this research focuses on using information drawn from financial statements particularly cash flow statements, some concepts and standards, including the accounting framework and cash flow statement, are reviewed as well. The chapter has five sections, as summarised in Figure 2.1. The first part introduces the structure of the chapter. Next, Section 2.2 briefly discusses Thailands geography and population as an introduction to the general background of Thailand, followed by a discussion of the Thai economy. Section 2.3 deals with Thai accounting standards. In addition, Section 2.3.1 provides the background to the Institute of Certified Accountants and Auditors of Thailand which issues Thai accounting standards. The standard dealing with the cash flow statement is reviewed and compared with the US standard in Section 2.3.2. The sample of this study is Thai listed companies, therefore Section 2.4 discusses the regulatory environment of Thai listed companies including the Stock Exchange of Thailand (SET) and information disclosure on SET. Finally, Section 2.5 provides the conclusion of the chapter.

14

Figure 2.1 Structure of Chapter 2: Thai Literature

2.1 Introduction

2.3 Thai Accounting Standards

15

2.2 General Background to Thailand 2.2.1 Country profile A unified Thai kingdom was established in the mid-14th century. It was known as Siam until 1939. The capital and largest city is Bangkok, with about 7,200,000 people. The population of Thailand is approximately 64.27 million (in 2003), with an annual growth rate of about 1.0% of the population. Most people are Buddhist (EIU 2004). Other details about geography, government and the economy are discussed below.

Thailands physical geography Thailand is located in Southeast Asia, bordering the Andaman Sea and the Gulf of Thailand. It borders Myanmar on the north and west, Laos on the north and northeast, Cambodia on the east and Malaysia on the south. Its area is 514,000 square kilometres. It can be split up into four distinct geographical regions, the Northern, the North-eastern, the Central and the Southern regions. The Northern region contains mountains, forests and fertile valleys. The North-eastern region is high, semi-arid, raised ground used mainly for cattle and growing crops such as rice and maize. The Central region is mainly flat and fruitful with the large river, the Chao Phraya River passing through the area. Bangkok, the capital city, is situated in this area. The Southern region is hilly and mountainous, covered mainly in rainforest. This area also receives the most annual rainfall and has a hot climate (Tourism Authority of Thailand 2003). Thai government system Thailand has been governed under a constitutional monarchy since 1932, with a parliamentary form of government. The King is the head of state. The head of government is the Prime Minister, elected from among the members of the House of Representatives. The Parliament comprises 200 members of the Senate and 500 elected members of the House of Representatives (EIU 2004).

16

Thailand is divided into 76 provinces, each administered by an appointed governor. The provinces are further divided into districts, sub-districts, tambons or groups of villages and villages. In the governmental system, all important decisions originate from the traditionally powerful bureaucratic elite in Bangkok. Composed of senior members of the civil and military wings of the bureaucracy, this elite dominated the governmental process from the national level down to the district level. In this process, the Ministry of Interior continues to play a key role as the administrative framework of the state, resisting reforms and changes (EIU 2004). 2.2.2 Thailands economy

The character and structure of the economy

The Thai economy is essentially a free enterprise system. The government owns and operates particular services such as power generation, transportation and communications. Agricultural products are produced in such quantities that in many commodities the country ranks as the worlds number one supplier. In addition to tapioca and rice, it is a leader in the production of frozen shrimp, canned pineapple, natural rubber and sugar. Thailands industrial sector produces a wide range of goods ranging from textiles, including the famous Thai silk, and handy garments to integrated circuits, plastics, jewellery, footwear, knocked-down furniture and fibreglass yachts. In recent years, in fact, manufacturing has exceeded agricultural products in Thailands GNP, while tourism has replaced agricultural products as Thailands largest source of foreign exchange (Mahidol University 1997).

Economic performance during the study period (1994-2002)

The period of this study was 1994 to 2002. In order to differentiate the effect of the Asian economic crisis during 1997 and 1998, the economic performance is reviewed in three parts, before, during, and after the Asian economic crisis.

17

Before the Asian economic crisis Thailands economy has had continuous and rapid economic growth since 1985. During the prosperous period years economic growth averaged more than 7% annually and remarkably approximately 9% in 1987-1995. Inflation as measured by the consumer price index (CPI), averaged 5% per year between 1992 and 1996. The total of goods and service exports grew in volume terms by an average of 15.2% per year between 1989 and 1995. The manufacturing sectors share of output had grown by an annual average of 13.5% per year in the period 19871995 (EIU 2004).

Table 2.1 Thai Economic Indicators 1994-2003 Economic Growth (%) 1994 9.0 1995 9.2 1996 5.9 1997* -1.4 1998* -10.5 1999 4.4 2000 4.6 2001 1.8 2002 4.8 2003** 4 137 106.0 1.8 2.0 71.2 16.1 67.1 15.2 6.75 2.00 42.69

GDP at market price(US$ Billion) 144.3 168.0 181.9 150.7 111.5 124.3 124.5 115 128 Inflation CPI 78.2 82.8 87.6 92.5 100.0 100.3 101.9 103.5 104.2 Inflation rate (%) 5.0 5.8 5.9 5.6 8.1 0.3 1.5 1.6 0.7 International trade Trade balance -8.7 -14.7 -16.1 -4.6 12.3 9.3 5.5 2.5 3.5 (US $Billion) Export 44.7 55.7 54.7 56.7 52.9 56.8 67.9 63.2 66.8 (US $Billion) % change 22.1 24.8 -1.9 3.8 -6.8 7.4 19.5 -6.9 5.7 Import 53.4 70.4 70.8 61.3 40.7 47.5 62.4 60.7 63.4 (US $Billion) % change 31.9 0.6 -13.4 -33.8 16.9 31.3 -2.8 4.4 18.4 Interest rate Prime rate (%) 11.75 13.75 13.25 15.25 12.00 8.50 8.25 7.50 7.00 Fixed deposits 10.25 11.00 9.25 13.00 6.00 4.25 3.50 3.00 2.00 Average currency exchange rate 25.15 24.97 25.39 31.48 41.59 37.96 40.27 44.48 43.00 (Baht: US$) Source: National Economic and Social Department Board, Bank of Thailand, Ministry of Finance. * Period during economic crisis ** Estimation

During the economic crisis The economy was destroyed by a severe recession that started in late 1997. A slowdown began in 1996 with economic growth of 5.5%. In 1997, negative growth by 1.4% and by 10.5% in 1998. The crisis of 1997 and 1998 began in the countrys financial sector and spread

18

throughout the economy. Because of the crisis, the Bank of Thailand (BOT) decided to float the baht freely on international currency markets on 2 July 1997. Prior to that decision, the bahts value was pegged to a basket of currencies heavily weighted to the United States (US) dollar, with the bahts value stable at approximately 25:1 compared to the US dollar. The baht reached its lowest point of 56 to the US dollar in January 1998 (Bank of Thailand 2003d).

During 1997, businesses started failing. A number of Thai companies were affected by the economic crisis. In 1997 and 1998, the number of failed firms was 9,925 and 12,409 respectively. Unstable financial institutions were closed. The root of the collapse occurred prior to 1997, when many companies were mismanaged and borrowed money from financial institutions operating with unhedged borrowing in foreign currencies. This led to non performing loan problems. Moreover the manufacturing sector suffered from overcapacity, low technology and dependence on imported components (Glassman 2001). In that time, government policy cooperated closely with the International Monetary Fund (IMF) in an attempt to recover the economy. As a result, new bankruptcy and foreclosure laws were passed. A lot of firms laid off their labour force to stay viable during the crisis (The Economist Intelligence Unit Limited 2004). This led to an increasing unemployment rate with an annual rate of 4.4% in 1998. Although in 1998 the devaluation of the baht led to import price inflation, domestic demand contained the increase in the CPI to an annual rate of 8.1% (Bank of Thailand 2003d).

After the economic crisis Thailand entered a recovery stage in 1999. GDP growth rates for 1999 were strong at 4.4%. Beginning in 1999 the baht stabilised and inflation and interest rates began to come down. The average inflation rate in 1999 decreased to 0.3%, owing to a more stable baht and declining world commodity prices. The unemployment rate declined to 4.2% (Bank of Thailand 2003c).

In 2000, the Thai economy remained stable, with growth at 4.5%. However, exports expanded to 19.6%. In 2001, headline inflation was reported at 1.6%, only slightly higher than it was in 2000 (Arnold 2002). Increases in the inflation rate were caused by a short period of baht

19

weakness and volatile international oil prices. The baht depreciated by more than 10% in 2000 (Bank of Thailand 2003b). In 2001 the labour force expanded by 1.8% and the unemployment rate dropped slightly to 3.3%, compared with 3.6% in 2000. In the export sector, Thailands exports experienced a sharp decline of 6.9% in US dollar terms, due mainly to a slowdown in the world economy. However, the trade balance in 2001 remained in surplus with US$2.5 billion. The current account recorded a surplus of US$6.1 billion, declining from a surplus of US$9.3 billion in 2000. The unemployment rate declined gradually to 3.6% and 3.36% in 2000 and 2001 respectively (Bank of Thailand 2002). The Thai economy grew faster than expected in the first part of the year 2002 with the rate at approximately 4.8%. Domestic demand increased steadily and exports rose sharply in the second half of the year. Economic stability was satisfactory with a low inflation rate, moderate trade and current account surpluses and a quite stable baht. The baht strengthened to Baht 43 per dollar from Baht 44.48 in 2001, and exports in US dollar terms increased (Bank of Thailand 2003a).

The increased growth in domestic demand was primarily caused by the governments economic stimuli comprising chiefly a sizable budgetary deficit, village fund and real estate business revival measures. Private sector consumption and investment made significant progress, influenced by the low level of interest rates, while public sector expenditure, both in investment and in purchasing goods and services had decreased. Additionally, the growth of Thai exports was partly supported by an improvement in market demand for global electrical appliances and economic improvement in the countrys main trading partners such as the United States and Japan (Bank of Thailand 2003a).

The key Thai economic indicators during 1994-2003 are summarised in Table 2.1. As shown in the table, Thai economic growth had increased rapidly before the economic crisis and severely decreased during the crisis. The growth rate however, has gone up again since 1999 to approximately 4% on average. In line with the growth rate, GDP rose dramatically before the crisis, declined in 1998, and then increased slightly after the crisis. In international trade,

20

Thailand started to experience a trade balance surplus after the crisis. balance surplus has decreased considerably since 1999.

However the trade

In addition, it can be seen from the table that the inflation rate had increased greatly during the crisis to about 8% but began to decrease in 1999. The crisis affected the Thai baht value which fell roughly by half, compared with the US dollar, from 1995 to 2003. Overall, as revealed by key Thai economic indicators, the economic crisis during 1997 and 1998 seriously affected the Thai economy. 2.3 Thai Accounting Standards 2.3.1 The Institute of Certified Accountants and Auditors of Thailand (ICAAT)

The Institute of Certified Accountants and Auditors of Thailand (ICAAT) is an independent organization, established on 13 October 1948. Its main objective is to enhance and develop Thai professional accounting, particularly Thai accounting standards. Members of the Institute Board are accounting professionals and have a variety of functional backgrounds such academics, auditors, accountants and users of financial statements. The Institute deals with accounting including developing accounting standards, internal auditing and publishing knowledge in the accounting area (ICAAT 1999a).

2.3.2 Development of Accounting Standards in Thailand Thai accounting standards have been developed continuously since the foundation of ICAAT. In 1997, ICAAT initiated a large project in the development of Thai accounting standards, due to the emergence of new financial businesses in Thailand. Meanwhile, the economic crisis caused many firms to fail, particularly financial institutions and real estate firms. Financial institutions faced non-performing loans (NPL) and investors lost a lot of value from investment in those firms. There was an argument that accounting information did not provide enough useful information for decision making. As a result, ICAAT reviewed Thai accounting standards to improve the reliability, credibility and usefulness of accounting information. Also, at that time, it was necessary to have new accounting standards suitable for the

21

economic situation. Some standards were revised or superseded by others and new standards were issued immediately (ICAAT 1999b).

ICAAT has set (as ICAATs pronouncement no. 010/1997-1999) Thai accounting standards by referring to the International Accounting Standard (IAS), or the US Statements of Financial Accounting Standards (FASB) if no IAS is comparable. Before the foundation of IAS, Thai accounting standards were set based only on the FASB of the United States. Recently, most Thai Accounting Standards (TAS) were developed based on IAS (see Table 2.2). By 2002, the number of legally effective TAS was 29, excluding the accounting framework. All Thai firms are legally required to prepare and report their financial information under these standards. However, unlisted companies are not ruled by some standards such as TAS Nos. 24, 25, 36, 44, 45, 47 and 48 (see Table 2.2).

Accounting framework

The accounting framework involves the concepts that emphasise the preparation and presentation of financial statements for external users. The framework outlines the objectives of financial statements, the qualitative characteristics that determine the usefulness of information in financial statements and the definition, recognition and measurement of the elements of the financial statements and is adopted from the IASs framework. The framework replaced TAS Number 1, involving accounting assumptions. This framework changed the concept of the purpose of preparation and reporting of accounting information to be not only for taxation but also to help external users in their decision making. In addition, some concepts in the measurement of the elements of financial statements and of capital and capital maintenance differ from those on TAS Number 1. Consequently, the information reported on financial statements may provide content different to that provided prior to the development of the accounting framework (IASC 2000).

According to the framework, providing financial information about financial position, performance and change in financial position is the main purpose of financial statements. The required set of financial statements includes a balance sheet, an income statement and a

22

Table 2.2 Summary Comparing between Thai Accounting Standards and International Accounting Standard, as of 30 January 2002 (www.icaat.or.th) TAS No. 7 8 11 14 21 24* 25* 26 27 29 Title /Content Accounting Framework Leases Construction Contracts Bad Debt Research and Development Costs Events after the Balance Sheet Date Segment Reporting Cash Flow Statement Revenue for Real Estate Disclosures in the Financial Statements of Banks and Similar Financial Institutions Accounting for leasing Effective Date 25 Feb 1999 28 Feb 1987 1 Jan 1988 1 July 1989 1 Jan 1990 31 Dec 1991 1 Jan 1994 1 Jan 1994 1 April 1994 1 Jan 2000 1 Jan 1996 (for listed companies) 1 Jan 1999 (for others) 1 Jan 1996 1 Jan 1997 1 Jan 1999 1 Jan 1999 30 Sep 1998 1 Jan 2002 1 Jan 1999 1 Jan 1999 1 Jan 1999 1 Jan 1999 1 Jan 1999 1 Jan 1999 1 Jan 1999 1 Jan 2000 1 Jan 2000 1 Jan 2000 1 Jan 2000 1 Jan 2000 1 Jan 2000 1 Jan 2004 1 Jan 2004 1 Jan 2006 1 Jan 2004 Based on IAS / SFAS IAS Framework IAS 17 IAS 11 SFAS IAS 9 IAS 10 IAS 14 IAS 7 IAS 30 IAS 17

30 31 32 33 34 35 36* 37 38 39

The Effective of Changes in Foreign Exchange Inventories Property, Plant and Equipment Borrowing Costs

Accounting for Troubled Debt Restructurings Presentation of Financial Statement Impairment of Assets Revenue Earnings per Share Net Profit or Loss for the Period, Fundamental Errors and Changes in Accounting Policies 40 Accounting for Certain Investments in Debt and Equity Securities 41 Interim Financial Reporting 42 Accounting for Investment Companies 43 Business Combinations 44* Consolidated Financial Statements and Accounting for Investments in Subsidiaries 45* Accounting for investment in Associates 46 Financial Reporting of Interest in Joint Ventures 47* Related Party Disclosures 48* Financial Instruments: Recognition and Measurement 51** Intangible Assets 53** Provisions, Contingent Liabilities and Contingent Assets 54** Discontinuing Operations 55** Accounting for government Grants and Disclosure of Government Assistance * except unlisted companies ** not effective

IAS 21 IAS 2 IAS 16 (1998) IAS 23 (1993) SFAS 15,114 IAS 1 IAS 36 IAS 18 IAS 33 IAS 8 IAS 25 IAS 34 AICPA IAS 22 IAS 27 IAS 28 IAS 31 IAS 24 IAS 32, 39 IAS 38 IAS 37 IAS 35 IAS 20

23

statement of cash flows. Financial statements are prepared on an accrual basis and under the going concern assumption. That is, under the accrual basis, the effects of transactions and other events are recorded when they occur. Under the going concern assumption, an enterprise is assumed to continue in operation for the predictable future. Moreover, the framework guides the qualitative characteristics of financial statements. The principal qualitative characteristics that make information useful to users are understandability, relevance, reliability and comparability (IASC 2000).

TAS 25: Cash flow statement

Thai Accounting Standard No. 25 dealing with preparation and presentation of the cash flow statement was issued in 1994. The standard regulates that Thai listed companies will release a statement of cash flow as part of their financial statements. The cash flow statement is the accounting report that provides information about cash receipts, cash payments and net change in cash balances during a period. Cash flow must be identified with three main activities of enterprises, operating, investing and financing activities. These activities are outlined below.

Operating activities are the main activities involving the revenue-producing activities of the company and other activities that are not investing and financing activities. Investing activities involve the acquisition and disposal of long-term assets and other investment except shortterm investments. Financing activities are activities that result in changes in the size and composition of the equity capital and borrowings of the enterprise.

The requirements of TAS 25, adapted from IAS 7, are generally similar to that of FAS 95. Nevertheless, TAS 25 provides some different concepts, as discussed below.

1) Meaning of cash. Under FAS 95, companies can choose to include cash equivalent as well as cash that corresponds with cash shown on their balance sheet. In contrast, TAS has defined cash flow to mean change in both cash and cash equivalents.

24

2) Classifications in the statement of cash flows. Both TAS 25 and FAS 95 require the classification of cash flow into three categories: operating, investing and financing activities. However, under TAS some items are categorized more flexibly. For example, based on FAS 95, interest paid must be reported in operating activities. However, TAS 25 does not mandate any particular classification. Accordingly, this item can be included in either operating or financing activities. 3) Format of reporting cash flow from operations. Cash flow from operations can be presented in two different ways, directly or indirectly. The direct method shows receipts from customers and payments to suppliers, employees, government and other creditors, while the indirect method starts with net profit or loss based on the accrual basis and adjusts for the effects of non-cash transactions.

FAS and TAS agree that the direct method should be used to report cash flow from operating activities. The direct method is not mandated. Most United States and Thai companies prefer to use the indirect method, because it is easier to prepare (Dennis 1994).

2.4 The Stock Exchange of Thailand (SET) 2.4.1 Roles of the Stock Exchange of Thailand The Stock Exchange of Thailand (SET) is a non-profit organisation under the supervision of the Securities and Exchange Commission (SEC). It was established under the Securities Exchange of Thailand Act (SEA) in 1974 and began trading on 30 April 1975 (SET 2000c).

The SET has the primary function of a centre for trading listed securities and providing the essential systems necessary to facilitate securities trading. In addition, the SET has a role in any business relating to the Securities Exchange, eg, as a clearing house, securities depository centre, securities registrar, or similar activities and any other business approved by the SEC (SET 2000d).

In the SET, only member companies of the exchange are allowed to buy or sell securities. Securities firms which have obtained a securities licence from the Ministry of Finance

25

(following recommendation from the SEC) to engage in the securities business as stock brokers may apply for membership of the SET. Membership status is obtained once approval is granted by the SETs Board of Governors. As of 31 December 2002, the SET had 28 member firms (SET 2000d).

Thai capital market

Capital markets are the markets for trading long-term debt and corporate stocks (Reilly 1994). A main role of the stock market is to distribute capital among public companies and between those companies and other assets (Ball et al. 1989). The SEC is responsible for the development of the Thai capital market including primary and secondary markets. The primary market, regulated and managed by the SEC, is the capital market for issuing new securities to the public. Companies wishing to issue new securities, perform an initial public offering (IPO) or wishing to offer additional securities to the public, must first apply for SEC approval and comply with its filing requirements (SET 2000b).

Secondary markets are the markets dealing with outstanding securities (Reilly 1994). In Thailand, the secondary market is comprised of the Stock Exchange of Thailand (SET) and the Thai Bond Dealing Center (TBDC). The SET operates the secondary market for trading capital securities, whereas TBDC deals with debt securities. Securities of companies that have applied for and been approved by the SET can be traded in the secondary market (SET 2000b).

Secondary markets provide significant benefits to investors. Investors who acquired securities in the primary market can receive liquidity by selling those securities in the secondary market. In addition, secondary markets are important to issuers because trading securities on the primary market depends on the current market price and yield of their securities in the secondary market (Reilly 1994).

Securities traded on the SET include common shares, preferred shares, unit trusts, warrants, debentures and convertible debentures. Common shares are the majority of shares traded on

26

the SET. At the end of December 2002, there were 289 common share issues, 11 preferred stocks issues, 11 unit trusts issues, 59 warrants issues and 1 derivative warrant on the SET (Source: Publication Department, The Stock Exchange of Thailand 2003).

Stock market performance Before the Asian economic crisis, trading in the Stock Exchange of Thailand was at a peak. The SET index was very high at about 1,360 and 1,280 in 1994 and 1995 respectively. During the Asian economic crisis, the SET faced a lack of quality listings of new companies and liquidity shortages, causing the number of listed stocks to drop by 30% in 1998-2000.

Nevertheless, in 2001, the Stock Exchange of Thailand revived significantly with improved price stability and active trading activities. The SET index recorded 303.85 points at the end of the year, compared with 269.19 points in 2000. Total turnover value in the SET increased substantially to 1,577.75 billion baht, up by 70.81% from the previous year, reflecting the fact that investor confidence had returned and the Thai economy was on the path to recovery.

The government has focused on the development of the Thai capital market in recent years, aiming to reinforce the strength of the Thai capital market as well as increasing its role as an alternative source of funds for the private sector. By early in 2002, a master plan had been formulated to cover all factors of the development of the Thai capital market. At the end of 2002 the SET index increased steadily to 356 points with total turnover value of 2,047 billion baht. The trading statistics of the SET are summarised in Table 2.3.
Table 2.3 Outstanding Trading Statistics of SET from 1994 to 2002 Item SET Index 1994 1360 1995 1280 1996 832 1997 373 1998 356 1999 482 1,610 6,571 2,193 392 450 26 2000 269 924 3,740 1,279 381 438 13 2001 304 1,578 6,440 1,607 382 449 6 2002 356 2,047 8,357 1,986 389 471 11

Total Trading T urnover 2,113 1,535 1,303 930 855 (Billion Baht) Average Daily Turnover 8,628 6,239 5,341 3,764 3,505 (Million Baht) Total market Capitalization 3,300 3,564 2,559 1,133 1,268 Number of Listed 389 416 454 431 418 Companies Number of Listed Securities 494 538 579 529 494 Number of Delisted 1 1 1 28 14 companies Source: Publication Department, the Stock Exchange of Thailand 2003

27

2.4.2 Information Disclosure in the SET The information disclosure regulation for both the primary and secondary markets is important for investors, as investors need information to make decisions and some information affects market price and yield of securities. In particular, an efficient capital market is one in which security prices adjust rapidly to new information, that is, investors immediately adjust their decisions according to available information (Reilly 1994).

The SET regulation requires companies to release information according to two categories, financial and non-financial, as discussed below (SET 2000f). Financial information or financial statements show the financial position and performance of companies including balance sheets, income statements and statements of cash flow. The financial statements must be prepared under Thai accounting standards in order to ensure the information is comparable among companies. Quarterly financial statements must be reviewed by an auditor. Financial statements of each accounting period must be audited and accompanied by the auditors report. Financial information is an important source of data for investment decisions. This is suggested by a study by Narktabtee (2000) on the application of accounting information to the Thai stock market. That research found that investors value the quality of accounting information as a whole. In addition, it has been found that cash flow statements are useful for the investor. Non-financial information must also be provided by the companies, including financial analyses explaining the financial position and performance of the company, risk factors, the internal control system of the company and the structure of business groups.

Listed companies are required to prepare and distribute financial statements to investors not only during their listing periods, but are also required to submit their financial statements to the SET when applying to be listed.

Since 2001, reports and submissions by the listed companies have been made through the Electronic Company Information Disclosure (ELCID) system. In addition, listed companies

28

have to simultaneously report and submit information to the SET both in Thai and English (SET 2000a). Disclosure procedure

The SET has implemented a full disclosure policy ensuring investors receive accurate, adequate and timely information. Disclosure of significant information is concurrently broadcasted by facsimile and on-line through the SET information system.

Listed companies must make any disclosure of material information at least one hour prior to the commencement of each trading session or after the close of the days trading. Trading sessions are from 10.00 am to 12.30 pm and from 2.30 pm to 5.00 pm. (SET 2000e).

Information dissemination

Information on companies and trading in their securities is distributed through various channels to meet different user requirements. These channels include the SET Information Service (SETINFO), internet, mass media and publications (SET 2000a).

SETINFO (SET Information Services) has been developed by SET in order to support efficient trading. By means of the Data Dissemination (DD) system, the SET directly receives news and information from listed companies and trading information from the exchanges own computerised trading system. It then updates information in the SET Information Management Systems (SIMS). The information alternatives provided are outlined below (SET 2000a).

SET Information Management Systems (SIMS) provides extensive background information on listed companies, relevant news updates, financial statements and historical trading data.

Price Reporting System (PRS) provides real time information on securities trading and current trading or market news.

29

Through information download, users can download information dealing with daily quotations at the end of a trading day plus key ratios such as the price-earnings ratio and dividend yield. The financial statements of listed companies are also available to download.

CD-ROMs provide information involving listed companies and securities trading data.

2.5 Conclusion

In this chapter, the background of Thailand including geography, government and economy was briefly reviewed. Furthermore, some details of Thai accounting standards were described, as an introduction to Thai accounting standards and to highlight the differences to US accounting standards. This chapter also provides some information about the Thai capital market and information disclosure requirements.

In summary, since the year 2000, the Thai economy has recovered from the economic crisis and economic growth has increased. The Thai government has initiated policies to advance the economic growth of Thailand and has aimed to improve the strength of the economy by developing the Thai stock market using the SET to enhance stock market performance. One objective of the SET plan is to enhance the expansion of the investor base. Information disclosure is a crucial factor in ensuring that investors will receive information useful for their investment decisions. In terms of accounting, information which is one type of financial information required by SET, ICAAT supports SET policies by improving and developing accounting standards in reporting financial statements. Thai accounting standards have been reviewed and issued in order to meet the requirements of users with respect to the development of the Thai stock market. The objective of reporting financial statements is to provide financial information to help users in their decision making including investors who are a major group of financial information users. In the meantime, investors or users of financial information do not understand how to use accounting information and the usefulness of accounting information is still doubtful. In the case of Thailand, research is required on ways to ensure the relevance and reliability of accounting information.

30

Cash flow statements have been required by SET since 1994 in addition to balance sheets and income statements. Under TAS No 25 a main objective of these statements is to provide cash flow information to help users in their decision making, particularly investors. Investors expect to receive high returns from their investments, and information which helps them to access information on future cash flows of companies in which they invest, is necessary. Research into the usefulness of accounting information for predicting future cash flows would be of benefit to them, by providing access to future cash flows of companies which issue shares trading in the Thai stock market. Creditors or analysts also require access to cash flows of a company for their particular purposes.

The next chapter will provide a review of the literature regarding two parent disciplines, cash flows and accrual accounting bases. Also, previous studies undertaken in the United States and other countries, relating to cash flow predictions have been reviewed, in order to isolate research gaps and construct research issues, in the case of Thailand.

31

Chapter 3
Literature Review
3.1 Introduction

Chapter two provided background information on Thailand. Some essential features of the economy of the country were discussed and some details of Thai accounting standards were introduced. Furthermore, the regulatory environment of Thai listed companies was discussed.

The main purpose of this chapter is to provide the literature review by considering the key theoretical issues related to the research proposal of using accrual accounting and cash flow data in predicting future cash flows and to present models of cash flow prediction.

There are seven sections in this chapter as shown in figure 3.1. Section 3.1 introduces the objective and structure of this chapter. The concept of cash flow prediction is discussed in Section 3.2 to assist in defining the boundaries of this research. The basic themes of the two parent disciplines, accrual accounting and cash flow data are presented in Section 3.3 and 3.4 respectively. The literature related to these parent disciplines is reviewed in this section as well. These parent disciplines providing the background for the immediate discipline of applying cash flow and accrual accounting data to extend the body of knowledge on predicting cash flow are in Section 3.5. Building from the previous literature cash flow and accrual accounting data models for predicting cash flow are presented in Section 3.6. Finally, the conclusion of the chapter is provided in Section 3.7.

32

Figure 3.1 Structure of Chapter 3: Literature Review

3.1 Introduction

3.2 Cash Flow Prediction

3.3 Accrual Accounting Data

3.4 Cash Flow Accounting Data

3.5 Application of Cash flows and Accrual Accounting Data in Predicting Future Cash Flows

3.6 Research Gaps, Research Issues and Model of Conceptual Framework

3.7 Conclusion

33

3.2 Cash Flow Prediction

Prediction is an important part of the decision-making process, because decision making reflects what will happen in the future. In economic decision making, financial prediction is a prime activity (Foster 1986). Every economic decision entails a choice among alternative means of achieving a given goal. Each alternative involves an expectation of receiving greater benefits in the future. The decision makers need to predict the consequences of the alternatives and choose the alternative which is expected to provide the greater benefits.

Cash flow prediction is a task required in various economic decisions, because cash flows play a vital role in almost all the decision making of many parties including security analysts, creditors and managers. Additionally, the decision makers are interested in a firms cash flows because they expect that current cash flows may affect their future cash flows (Neill et al. 1991) and they will have an interest in assessing the firms future cash flows to the extent that these provide a clearer indication of the firms cash flow in the future. In other words, a general objective of fundamental analysis is a forecast of the enterprises future cash flows (Staubus 2004), because the cash flows are the basis for dividends, interest payments and repayment of debt.

Particularly in security investment decisions, investors or security analysts need to estimate the cash return from their investment in the capital markets. The cash return includes cash from either share dividends or capital gains when shares are sold. This decision deals with which shares to buy, retain, or sell and the appropriate time for purchase or sale of those shares. The ability of a company to pay dividends is reflected by the ability of the company to generate its future cash flows. Therefore, in making investment decisions, predicting the cash flows of a company issuing shares is a primary task in indicating the companys ability to pay dividends for the future period (Frigo & Graziano 2003; Neill et al. 1991).

In creditors lending decisions, predicting bankruptcy problems of a client or customer can help creditors prevent losses due to bad debts. There are a number of early warning signs indicating that a company is experiencing financial distress. Cash flow is an important

34

financial indicator of a financial problem (Zwaig & Pickett 2001). A decline in cash flow can provide an early warning signal of a bankruptcy to creditors and other interested parties (Epstein & Pava 1992; Zwaig & Pickett 2001).

With respect to business management, cash flow is viewed as the lifeblood of a business (Schaeffer 2002) as cash must be available when it is needed. Therefore a companys ability to manage cash is vital to survival and wealth (Sharma, R & Jones 2000). Predicting cash flows of future periods can help a manager identify future financial problems (Kelly & O'Connor 1997). Cash flow prediction allows the company to know its cash position and to make the necessary expenditures for such items as debt repayment, acquisitions and payment of expenses (Plewa & Friedlob 1995). In addition, the difference between forecast and actual cash flows needs to be analysed to understand and measure a firms performance.

In internal capital investment, capital budget analysis also involves cash flow prediction (Foster 1986). The capital investment deals with investment projects such as new products, replacement of existing assets, or expansion of product lines (Bierman 1988). The projects can be evaluated by various methods including net present values (NPV) and internal rate of return (IRR) (Brigham & Gapenski 1999). In assessing projects, managers or analysts need

to estimate the investment outlays and annual cash flows from operations after projects have been operating over the forecast period (Weaver & Michelson 2003). Then, a manager has to make a decision whether the project should be accepted or rejected under an assumption of each method. In terms of NPV, for instance, the project is accepted if the present value of the cash flows for the operations exceeds the initial investment cost or the project provides a positive net present value of cash flows (Brigham & Gapenski 1999). However, the cash flow prediction is made very difficult because the future cash flows generated by the investment cannot be estimated with perfect certainty at the time of initial outlays (Giaccotto 1990). Although the annual cash flows from operations can be predicted by using historical data, in the literature on capital budgeting the modeling and forecasting of future cash flows have not been identified (Giaccotto 1990). Studies in cash flow prediction may be applicable for the purpose of capital budgeting.

35

In terms of the importance of cash flow data, cash flows as a measure of a firms performance is less subject to distortion than is the net profit figure (Bernstein 1993), because the calculation of cash flows from operations removes the causes of the distortion such as depreciation methods, deferred taxes, and goodwill amortization (Giacomino & Mielke 1987, 1988), whereas the determination of net profit under accrual accounting requires approximations, deferrals, allocations and valuations. These procedures allow managers to manipulate their companys profits. That is, they can choose an accounting method, from among various methods of calculating depreciation and valuing inventories, to produce high or low profit as they want. As a result, analysts usually consider cash flow to evaluate firm performance in addition to the profit (Boyd & Cortese-Danile 2000/2001; Kremer & Rizzuto 2000).

In addition, cash flows are expected to be a natural alternative performance indicator to evaluate a firms performance because earnings are not informative when they are transitory and extreme and cash flows are already available to be used in cash flow statements (Cheng, Yang & Clubb 2003). The investors also consider cash flow a sustainable performance

measure for a firms valuation.

Moreover, the difference between earnings and cash flows from operations can be used as an important signal of potentially fraudulent financial reporting that auditors and other analysts should consider, in addition to other factors such as leverage, retained earnings and market value (Lee, Ingram & Howard 1999). The excess of earnings over cash flows indicates the fraud risk in the coming years. This is because the fraudulent firms often have poor financial performance but they conceal their performance by overstating earnings.

Cash flows from operations is used to calculate free cash flow. Free cash flow is money earned from operations after provision for capital expenditures at the end of an accounting period. It is basically defined as net cash flow from operating activities less capital expenditures and dividends on preferred stock (Lees & Leibman 2000; Williamson 2003). It shows the ability of the company to generate cash from its operations after spending money on the capital expenditures (Chang 2002). Without free cash flow, it is difficult for a business to

36

pursue new opportunities, acquire other businesses or pay dividends. Free cash flow analysis helps managers identify the capital available for reinvestment in enhancing the companys growth. In turn, analysing free cash flow can separate the firms with a high ability to internally grow from firms with a low ability to grow. In addition to reinvestment, the company can distribute free cash flow to pay dividends to shareholders (Hackel, Livnat & Rai 1994). As a result, the free cash flow may be considered to assess the ability of companies to pay dividends on common stock.

With respect to the importance of the cash flow prediction, some academics have advocated revealing cash flow forecasts in order to assist investors and analysts predict future dividend streams (Climo 1976; Lawson 1992; Lee 1993). They suggested that cash flow forecasts should be compared with actual cash flow in order to provide more useful information for investment decisions. That is, the valuation of investment is assumed to be determined by discounting the future cash flow expected by the investors at a rate of interest acceptable to them. So the investors should be provided with the information relating to future cash flow from investment (Climo 1976).

The importance of cash flow prediction is supported by statements of accounting standards. Both the Financial Accounting Standard Board (FASB) and the International Accounting Standard Committee (IASC) provided a fundamental guideline for preparing and presenting financial statements, that the objective of reporting financial statements is to provide financial information for users to predict the amount, timing and uncertainty of the future cash flow of a company.

The primary objective of accounting data is to provide information to help present and potential investors, creditors and others assess the amounts, timing and uncertainty of prospective net cash inflows to the related enterprise (FASB 1978, paragraph 37).

Information about economic resources controlled by the enterprise and its capacity in the past to modify these resources is useful in predicting the ability of the enterprise to generate cash and cash equivalents in the future (IASC 1995, p. 44-45).

37

These statements suggest that accounting information from financial statements is useful in predicting future cash flows of a company. Consequently, the usefulness of accounting information has been investigated by a number of researchers in terms of their ability to predict future cash flows (Ashton 1974; Neill et al. 1991).

In general, accounting information is seen as input data for financial prediction models (Godfrey, Hodgson & Holmes 2003; Riahi-Belkaoui & Jones 2002). Accounting information is reported under the accrual and cash accounting bases. Income statements and balance sheets report information on an accrual basis and cash flow statements are on a cash basis.

3.3 Accrual Accounting Data

3.3.1 Accrual accounting basis The accrual accounting basis is a basic accounting assumption dealing with the accounting process of recognising the effects of financial transactions in the period in which events occur, rather than focusing only on cash receipts or payments (IASC 2000). The transactions are recorded and reported in financial statements of the period they occur whether or not cash has been received or paid (Riahi-Belkaoui & Jones 2002). As a result, accounting information reported in financial statements consists of both the effect of credit and cash transactions.

A main reason for the development of accrual accounting systems is the mitigation of timing and matching problems inherent in cash flows, in order to measure a firms performance (Cheng, Liu & Schaefer 1997b; Dechow 1994). A demand for information used for evaluating a firms performance arises from measuring and rewarding management. Additionally, the

need for accrual accounting systems is derived from the industrial revolution which precipitated business growth causing the complexity of firms transactions (Sharma, DS 2001). Although the success of a firm can be measured by its ability to generate cash, reporting of cash receipts and payments has timing and matching problems that cause cash flows to be a noisy measure of firm performance. Since under the continuing entity assumption, companies are assumed to operate without discontinuing, in order to measure

38

their performance the companies need to slice time into small segments and report their progress for each specific period of recording. Companies which record only cash transactions would have problems when the transaction involves more than one period of recording. This is because companies usually deal with credit transactions. For example, if a company purchases assets on credit, the payment will be made in the next period which does not match the period of the purchase. If the company records only cash transactions, the report will not show the cost at the time the business purchased the assets. The incomplete information would give an inaccurate measurement of the firms performance.

In the accrual accounting process, the timing of cash flow recognition is ignored in recognising earnings (Dechow 1994). That is, accrual accounting recognises events in which related cash flows occur in previous or subsequent accounting periods. Earnings is measured by an excess of revenue over expenses in each period of recording. Accrual accounting records the purchase of assets or resources used to operate a business and the provision of goods and services made by a company during a period which does not match the cash receipts and payments, by recognising revenues and related increases in assets and expenses and related increases in liabilities, for amounts expected to be received or paid, usually in cash in the future.

The accrual concept recognises assets, liabilities, revenues and expenses to record the trade transactions, including cash and credit transactions. An asset refers to a resource that belongs to the company as a result of past financial transactions (IASC 2000). Assets represent future benefits including cash that is expected in the future. They can be divided into two categories according to their longevity, current and non-current assets. In addition to cash, current assets include accounts receivable, inventories, prepayments and other assets that will be converted into cash within twelve months of the reporting date. In contrast, non-current assets refer to assets that will not be converted into cash with in next twelve months after the end of the financial year, such as land and buildings, plant and equipment and intangible assets, including goodwill (Henderson & Peirson 2000).

39

A liability refers to a present obligation of a company as a result of past financial transactions and which the company is expected to pay for in cash, or other economic benefits in the future (IASC 2000). Liabilities can also be categorised into current and non-current (Henderson & Peirson 2000). Current liabilities include accounts payable, short-term debt and other liabilities that will be paid within one year, while non-current liabilities include long-term debts that will not be paid within the next twelve months (Kimmel, Weygandt & Kieso 2000).

Accrual accounting involves concepts of historical cost and matching revenue to expenses. Assets or liabilities are recorded under the basis of historical cost. That is, the value of the assets of a company is measured by using their original acquisition cost, from which is deducted a proportion of this cost in the form of amortisation or depreciation. Additionally, recognition of revenue and expenses is based on a matching concept (Godfrey, Hodgson & Holmes 2003). Under the revenue recognition principle, firms recognise revenue when they have performed all, or a substantial portion of, services that have to be rendered, and cash receipts from the transaction are reasonably certain, and under the matching principle, firms recognise all expenses associated with revenues in the same period in which the revenues have been recognised.

In addition, accrual accounting merges the concepts of allocation, amortisation, and realisation (Trotman & Gibbins 1998). Allocation is the accounting process of assigning or distributing an amount based on a plan or formula. Some assets and liabilities will be allocated to expenses and revenues based on the length of time of use of the assets or to match revenue and expenses. Amortisation and depreciation mean the systematic reduction of an accounting amount related to the utilisation of long-lived assets or non-current assets, in order to allocate the costs of these assets to the time periods in which the asset is utilised (Sorter, Ingberman & Maximon 1990). For example, the cost of property, plant and equipment will be written down to expenses called depreciation based on the period of useful life of them and the methods used. Realisation is the process of changing non-cash resources and rights into money. This concept involves sales of assets for cash or claims of cash. For example, gains from sales are identified as revenues and losses are identified as expenses (Trotman & Gibbins 1998).

40

3.3.2 Usefulness of accrual accounting data

In practice, accounting financial information on an accrual basis is supported for many reasons. First, it is considered relevant in measuring a firms performance (Godfrey, Hodgson & Holmes 2003). For example, a manager needs data on past transactions in order to evaluate past operating performance. Accounting information from past transactions can be used as a measure of the past performance. Secondly, the cost of assets recorded based on historical cost is derived from actual transactions, not estimated (Godfrey, Hodgson & Holmes 2003). Thirdly, the concept of matching expenses and revenue reflects the uses of assets in generating revenue. This can measure the efficiency of utilising the assets of the company. Fourthly, it reports assets or future benefits and liabilities or obligations of a company, allowing the company to estimate future cash receipts and payments. In addition, reporting financial statements on an accrual basis meets taxation requirements (Lawson 1992).

In addition, with respect to earnings or profit as one of the key measures of a companys performance (Calabrese & Rafferty 2003; Dechow 1994; Henderson & Peirson 2000), an advantage of measuring with earnings is that performance cannot be distorted by uncertain variations in cash flows (Kremer & Rizzuto 2000) because financial transactions are recorded by historical cost. Furthermore, allocation, amortisation, depreciation and other accrual

processes allow a company to spread the cost of assets over the assets useful lives or time of their benefits. Otherwise the income statement will show a low profit or loss for the first period of acquiring the assets which may not be reasonable as those assets are used to generate income during more than one period.

Since under the revenue recognition and matching principles, accrual accounting recognises events in which related cash flows occur in previous or subsequent accounting periods in which cash flows are transformed into accounting earnings, earnings are seen as an indicator of firms valuations (Cheng, Liu & Schaefer 1997b; Dechow 1994). In addition, the accrual earnings and values of net assets are expected to equal net cash flows over the life of a business (Sharma, DS 2001). Consequently, studies in the usefulness of earnings in the

capital market are based on the hypothesis that earnings are a good surrogate for a firms

41

future cash flows (Board & Day 1989; Watts & Zimmerman 1986). A number of researchers have studied the relationship between accounting earnings and stock price and suggested that earnings have an implication for future cash flows of companies (Ball & Brown 1968; Bamber & Cheon 1995; Lipe 1990)
.

Furthermore, accrual earnings are indicators of future cash flow for many reasons. Firstly, because historical cost accounting earnings is the standard reported earnings measure and is the most common variable to be analysed in the press and accounting literature, it is used to indicate future cash flows (Board & Day 1989). Secondly, earnings are supported by the assumption that earnings provide information about the future dividend-paying ability of firms (Bierman 1992; Lipe 1990). Furthermore, accrual earnings are seen as a more relevant basis for assessing cash flow return than cash flow, because dividend payouts are based on accrual earnings (Board & Day 1989).

The IASC (1995) suggested that under the accrual basis, accounting information including financial position and financial performance of a company is useful for users in their decisionmaking. In particular, in regard to prediction, the IASC claims that this information can be useful in predicting the ability of the enterprise to generate or pay cash and other benefits in the future. Furthermore, the FASB suggested that earnings under an accrual basis have a better predictive ability than cash flow information. The FASB statement is below.

Information about enterprise earnings based on accrual accounting generally provides a better indication of an enterprises present and continuing ability to generate favourable cash flows than information limited to the financial aspects of cash receipts and payments (FASB 1978, paragraph ix).

In the same way, IASC (1995) claims that information about financial performance, position and change in financial position reported under an accrual basis aids users in assessing ability to generate cash flows:

42

Users are better able to evaluate this ability to generate cash and cash equivalents if they are
provided with information that focuses on the financial position, performance and change in financial position of an enterprise (IASC 1995, p.44).

However, there has been an argument on the quality of accrual-based data, particularly, profit as a measure of company performance (Dechow 1994; Hodgson & Stevenson-Clarke 2000; Sharma, DS 2001). It was argued that earnings suffer from flexible accounting techniques, subjective judgement and manipulative practices (Bernard & Stober 1989; Lee 1992). Therefore, the financial information from the accrual accounting process may be misleading, making earnings a less reliable measure of a firms performance (Dechow 1994).

Creative accounting is an issue as users of financial statements claim that it allows managers or preparers to manipulate financial reports (Charitou 2000; Henderson & Peirson 2000; Hill, Benston & Hartgraves 2003). Creative accounting involves making a choice of accounting methods to ensure that the financial statements reveal the information that impresses the preparers or managers (Blake et al. 2000). There are four creative techniques in the accounting process which can achieve the desired financial report (Henderson & Peirson 2000). The first method is to choose an accounting policy, or change an accounting policy, between permitted alternative accounting policies (Blake et al. 2000). For example, preparers may choose first-in, first-out (FIFO) methods to assign costs to inventory if they want to defer expenses, instead of using the last-in, first-out (LIFO) method of valuing inventory. The second method is to estimate or predict future events in a way biased towards a desired result. For instance, accountants may estimate asset life for calculating depreciation in order to manipulate depreciation expenses. Consequently, the firm may overstate the values of its assets or the amount of its income (Lundholm 1999). The third is a use of time transactions to take advantage of the assumption in matching revenues and expenses. Finally, through disclosure of transactions or events preparers may misrepresent financial statements by concealing some items that would show unsatisfactory features of an entitys operations. Additionally they may present the format of financial statements in a way that may distract attention from some items and attract attention to others.

43

Another question regarding the measurement of earnings arises from the historical cost principle under accrual accounting (Brigham & Gapenski 1999). For some conditions, notably inflation, use of a traditional historical cost system tends to overstate profits and understate asset values (Board & Day 1989; Lawson 1992; Lee 1993). That is, increasing rates of inflation will lead to increases in interest rates; assets are still reported in historical cost in which, in the real world, their values are high; and expenses are recorded at a low cost, as historical cost.

Moreover, earnings under accrual accounting could themselves suffer from timing and matching problems that may contribute to errors in assessment of a firms value (Dechow 1994). In some cases, accruals may include poorly-estimated receivable collections, depreciation, equity-method income that does not approximate market changes, and current recognition of previous periods increases in market value.

As a result, under these circumstances accrual earnings may be less effective in their ability to predict future cash flows, and, users of accounting information turn to consider cash flow instead (Bierman 1988; Sharma, DS 2001; Wang & Eichenseher 1998).

3.4 Cash Flow Accounting Data

3.4.1 Cash flow accounting data

Cash flow accounting (CFA) is defined by Lee (1981) as below.

CFA is the term used to denote a system of financial reporting which describes performance of an entity in cash terms. It is based on a matching of periodic cash inflows and cash outflows, free of credit transactions and arbitrary accounting allocations. Inflows include cash from trading operations and providers of long-term finance; and outflows include payments for replacement and growth investment, taxation, interest, and distributions (Lee 1981, p. 63).

Historically, the CFA system has been operating since the beginning of the 18th century. However, users of financial information did not support this system until the early 1960s (Lee 44

1993). At that time, financial information users were concerned about the use of cash flow data for decision-making purposes as a substitute for income. Cash flow was represented by profit plus depreciation (Paton 1963 and Drebin 1964 cited in Lee 1981, p. 64). Lawson (1992) and Lee (1985) suggested that cash flow accounting may be helpful to investor decision- making. The cash flow accounting concept advocated by academics and researchers in the UK and USA has appeared to influence financial reporting practice since the 1980s (Lee 1992; Staubus 1989). In the USA, FASB required a statement of cash flow to be included in companies financial reporting instead of fund flow statements.

Cash flow accounting can avoid uncertain accounting allocations present in the accrual system, produce more objective financial information and provide users with fundamental and critical financial data (Lee 1993), because cash flow accounting does not involve allocation and matching problems. Payments and receipts are recorded when the transaction of receipts or payments are made. As a result, it is expected that cash flow is less vulnerable to manipulation than accrual information (Ali 1994; Sharma, DS 2001). For similar reasons, cash flow is seen as the superior instrument for predictive purposes, particularly for predicting future cash flows (Charitou & Ketz 1991; Lee 1993).

3.4.2 Cash flow statement

The cash flow statement is the accounting report that provides information about cash receipts, cash payments and net change in cash balances during a period. In the United States, companies are required to disclose a statement of cash flow, as a result of Financial Accounting Standard No. 95 (FAS No.95) as regulated by the Financial Accounting Standard Board (FASB) in November 1987, effective for fiscal years ending after 15 July 1988. In Thailand, the Institute of Certified Accountants and Auditors of Thailand (ICAAT) issued Thai Accounting Standard No.25 (TAS 25), which is comparable to the International Accounting Standard No. 7 (IAS 7). As a result, a statement of cash flow is reported as part of a companys financial statements for all business enterprises in the USA and many countries whose accounting standards are based on these standards (Brown 1994). In Thailand, only Thai listed companies need to release cash flow statements.

45

Previously, companies were required to present a fund flow statement that reported sources and uses of funds. Funds can be defined in three ways, including cash, working capital and total resources (Henderson & Peirson 1994). Funds presented on fund statements are interpreted as working capital which is measured as current assets less current liabilities, whereas funds reported on cash flow statements refer to cash. The meanings of cash and cash equivalents presented on statements of cash flow are defined in IAS7 as below:

Cash comprises cash on hand and demand deposits. Cash equivalents are short-term, highly liquid investments that are readily convertible to known amounts of cash and which are subject to an insignificant risk of changes in value (IASC 2000, p. 111).

Cash flow statements replaced the fund flow statement for two main reasons. First, it resolved disputes over the definition of funds, and the purposes and presentation of the fund flow statement. The fund flow statements were not considered to provide sufficient information for investors and other financial statement users due to form and the definition of funds used (Zega 1988). Secondly, they improved the reliability and usefulness of reported financial information (Pizzey 1991). Moreover, they omit the effects of some transactions that may be very important. One example is intra-working capital transactions that affect components of working capital, but leaves total working capital unchanged. Another type is the transactions that do not affect the components of working capital (Henderson & Peirson 1994). In addition, the accounting standard for fund flow statements had not been revised for many years.

Cash flows on the cash flow statement must be identified with three main activities of enterprises as required by FAS No. 95 (and IAS 7). These are, cash flow from operating, investing and financing activities. The basis for the classification is derived from finance theory (Nurnberg 1993), that is, enterprises derive the cash used for investing activities and settlement of outstanding financial obligations in an accounting period from internal and external sources. Internal cash resources originate from the net cash generated from current operations. External cash resources come from financing activities such as borrowing and

46

receiving cash from the sale of equity shares (Wallace, Choudhury & Pendlebury 1997). Operating activities are the main activities involved in the revenue producing activities of the company and other activities that are not investing and financing activities. Investing activities involve the acquisition and disposal of long-term assets and other investment except shortterm investments. Financing activities are activities that result in changes in the size and composition of the equity capital and borrowings of the enterprise (IASC 2000).

Cash flows from operating activities can be reported by two methods, the direct or indirect method (IASC 2000). The direct method shows cash receipts from customers and cash payments to suppliers, employees, government and other creditors. The indirect method starts with net profit or loss based on the accrual basis and adjusts for the effects of non-cash transactions such as depreciation and amortization expenses, and changes in current assets and liabilities. The indirect method is preferred over the direct method (Sondhi, Sorter & White 1988; Wallace, Choudhury & Pendlebury 1997).

The indirect method reflects conversion from accrual-basis profit to cash-basis profit (Moorehead 2001). In other words, it shows the association between the cash flow statement and two financial statements based on accrual basis. That is, cash flows from operations relate to revenues and expenses on income statements, and current assets and liabilities on balance sheets (Bernstein 1993). Therefore, a reason for supporting the indirect method is that it is more informative than the direct method because it emphasises the difference between net income and operating cash flow, which can reduce the ability of management to manipulate the income statement numbers (Wallace, Choudhury & Pendlebury 1997).

Cash flow from operations is often seen as the most important category among the three categories because it results from the main income-producing activity. Cash generated from the operating activity provides an indication of the capacity of the company to produce cash from its main activity. The company must generate sufficient cash from its operating activities to finance its daily activities (Boyd & Cortese-Danile 2000/2001). Moreover cash flows from operations primarily support capital expenditures and dividends (Grossman & Pearl 1988). If the company cannot generate any cash to repay loans, pay dividend or make new investment,

47

the company would lend cash from external sources, causing future cash outflows. Cash available for investments and external financing also shows the firms ability to make new investments and external financing shows the firms ability to make new investments. It also indicates to investors the dividend-paying ability of the firm (Charitou & Ketz 1991).

In addition, the cash flows from operations can be used to evaluate the quality of profits on income statements. The difference between net cash flow from operations and net profit is helpful in interpreting the quality of earnings (Gallinger 2000). A large difference between net profits and cash flow from operations reflects a low quality of profits- perhaps net income has increased without an increase in cash flows from operations. This may result from increases in sales on credit, causing increases in accounts receivable, indicating that the company may have a cash collection problem in the future.

3.4.3 Usefulness of cash flow data

Nowadays, the cash flow statement is accepted as a necessary component of complete financial reporting by national and international accounting standard setters, because financial statement users note that the balance sheet, income statement and retained earnings statement do not always show the whole financial condition of a company (Kimmel, Weygandt & Kieso 2000). The balance sheets show the variety of assets owned by a company and the manner in which they were financed at the end of period but the source of activity related to those items during the period are not provided. Also profit in the income statement does not reflect an increase in cash. Moreover, the profitability and financing issues are reported separately on income statements and balance sheets respectively. This causes misleading and confusing results to users (Gombola et al. 1987).

The requirements of cash flow statements are based on the assumption that past cash flows are useful for assessing future cash flows and the cash flow statement supplements and presents information differently to the information provided in the other financial statements (Charitou & Ketz 1991). Accounting standard setters claimed that the cash flow statement used in

48

conjunction with other financial statements, the balance sheet and income statement, provides the following perceived benefits:

It presents an insight into the changes in net assets of a company, financial structure (including its liquidity and solvency). It shows the ability of a company to generate cash and cash equivalents. It can be used in developing models to assess and compare the present value of the future cash flows of different companies. It also enhances the comparability of the reporting of operating performance by different enterprises because it eliminates the effects of using different accounting policies in accrual accounting for the same transaction and events. It is usually used as a sign of the amount, timing and certainty of future cash flow. It is also useful in checking the accuracy of past assessments of future cash flows and in examining the relationship between profitability and net cash flow and the impact of changing prices (IASC 1995, p. 115).

Cash flows are advocated as very useful information for investors and creditors (Hodgson & Stevenson-Clarke 2000). Many advocates argued that the main benefit of cash flows to users is that the information overcomes many limitations associated with accrual accounting measurement procedures manifested in traditional financial statements (Lee 1993). In addition, because cash flows represent the most objective measure of capacity to consume and command resources and cash portrays the best measure of liquidity, they are not contaminated by measurement problems, and facilitate the prediction of future dividends and credit and loan payments (Hodgson & Stevenson-Clarke 2000; Lancaster, Stevens & Jennings 1998; Wertheim & Robinson 1993a, 1993b).

The issuance by the accounting standard setters on the statement of cash flow has made cash flows a major issue to users of financial statements (Dambolena & Shulman 1988; Jones 1998). The number of capital market-based research studies has increased, to investigate the usefulness of cash flow data in decision making (Lee 1993; O'Bryan, Quirin & Berry 1999). In addition, with respect to the suggestion of FASB (1978) that accrual accounting data provides a better indication for decision making than cash flow data and arguments for quality

49

of earnings from accrual basis, the usefulness of cash flow data is compared with that of accrual accounting data across many issues (Bierman 1994; Jones & Widjaja 1998). The relative usefulness of cash flow versus accrual data in providing information to decision makers has been examined in many contexts, such as relevance to stock prices, insolvency and bankruptcy (Mossman et al. 1998; Sharma, DS & Iselin 2003; Sloan 1996; Wertheim & Robinson 1993b). Mossman et al. (1998) suggested that cash flows can be used as an early warning of potential financial distress. Some researchers indicated that cash flow and income together are useful for evaluation of growth and growth prospects with implications for the value and the survival of business (Ingram & Lee 1997).

Moreover, the predictive ability of cash flow versus accrual accounting data in predicting economic events has appeared as a topic of interest. The predictive ability of accounting data in predicting future cash flow is an important issue because of the suggestion of the accounting standard setters and the relevancy of cash flows to a wide range of decision making (Neill et al. 1991).

There is a need for research into the usefulness of cash flows versus accrual accounting. Although there have been a number of studies in the context of the usefulness of cash flows and accrual accounting data, the results of those studies were inconsistent and inconclusive, and the contradiction between cash and accrual based data still exists (Penman 2001).

Furthermore, studies from an international perspective found evidence that the relative and information content of earnings and cash flows are different across countries. Bartov,

Goldberg and Kim (2001) examined the relative and information content of earnings and cash flow for equity valuation in the cases of the United States, the United Kingdom, Canada, Germany and Japan. Their study indicated that earnings are more important than cash flows in the United States, the United Kingdom and Canada, but not in Germany and Japan. They concluded that the national reporting regime is a factor which has an impact on the information content of earnings and cash flows. Therefore, studies using data in different countries may provide different results.

50

3.4.4 Cash flow ratios

Financial ratio analysis is a tool used in financial statement analysis (Giacomino & Mielke 1993). Financial ratios can be used to predict financial variables and to evaluate relative performance such as predicting bankruptcy, stock prices and the probability of loan defaults (Sylvestre & Urbancic 1994). Ratios are developed to help users of financial statements compare performances of companies on a year-to-year basis and across companies.

Cash flow statements provide new measures to evaluate firm performance (Figlewicz & Zeller 1991). The concept of cash-based performance ratios had been used in financial analyses before the regulation of reporting cash flow statements. In that time, surrogates of cash flows were used, such as net income plus depreciation, resulting in a lack of uniformity and misdirected analysis. Currently, statements of cash flow have the ready availability of cash flow data with consistent performance measures of cash flow from operations (Figlewicz & Zeller 1991).

Cash flow ratios are based on the cash flow from the operations (CFO) of the company. Also, ratios can contain accrual-based accounting data. The cash flow ratios provide a clearer picture of a companys performance, highlighting an organisations cash flow strengths and weaknesses (Carslaw & Mills 1991). Cash flow ratios could be a better measure of firm performance than financial ratios from income statements and balance sheets, because cash flows from operations as a main component of the ratios, exclude the effect of non-cash flow items such as depreciation expenses and gain or loss on the sale of operating assets (Kelly & O'Connor 1997; Plewa & Friedlob 2002). Profit reported on income statements may be a subjective distortion because it includes these items (Giacomino & Mielke 1993; Hoggett, Edwards & Medlin 2003). It has been argued that traditional ratios from income statements and balance sheets such as the liquidity ratio and quick ratio may not provide a comprehensive measure of a companys ability to retire its debts because current assets, including accounts receivable and inventory, may not be converted into cash.

51

There has been evidence that cash flow ratios contain additional information which is not contained in the accrual based figures and ratios (Gombola & Ketz 1983a; Salmi, Virtanen & Yli-Olli 1990; Yli-Olli & Virtanen 1989). Gombola and Ketz (1983a) found that cash flow ratios fill a separate and distinct factor not captured by any other ratio group from accrual based financial statements such as profitability ratios. The cash flow ratios are advocated because they can give users better insight into the financial performance of a company (Dennis 1994).

Cash flow ratios may be categorised into two groups; cash flow sufficiency and cash flow return ratios as described below (Figlewicz & Zeller 1991; Gombola & Ketz 1983b; Plewa & Friedlob 1995).

1) Cash flow sufficiency ratios

Cash flow sufficiency ratios are aimed at assessing a companys relative ability to generate sufficient cash to meet its cash flow needs. All ratios indicate whether a companys cash flows are sufficient for the payment of debt, acquisitions of assets and payment of dividends. These ratios are cash flow adequacy, debt coverage, repayment of borrowing and dividend payment ratios. Cash flow adequacy ratio The cash flow adequacy ratio is an attempt to assess the entitys ability to produce sufficient operating cash flows to cover its main cash requirement, specifically, the payment of debt, the acquisition of assets, and the payment of dividends.

Cash flow adequacy =

Cash flow from operations Repayment of borrowings + Assets acquired + Dividends paid

52

 Debt coverage ratio The debt coverage ratio shows the ability of a company to generate cash flow from operating activities to pay its long-term debt commitment.

Debt coverage ratio

Total debt Cash flow from operations

Repayment of borrowings ratio This ratio indicates the ability of a firm to generate cash from operating activities for the purpose of covering long-term debt commitments in the current year.

Repayment of borrowings ratio

Repayment of borrowings Cash flow from operations

Dividend payment ratio The dividend payment ratio presents the ability of a company to generate cash from operating activities for the purpose of covering dividend commitments to both ordinary and preference shareholders. If the ratio is greater, it means that the company paid a smaller portion of its cash from operating activities in dividend payments.

Dividend payment ratio

Dividends paid Cash flow from operations

Reinvestment ratio The reinvestment ratio presents the ability of a company to generate cash from operating activities for the purpose of covering asset acquisition payments.

53

Reinvestment ratio

Payment for property, plant and equipment Cash flow from operations

2) Cash flow return ratios

This group is sometimes called efficiency ratios. It shows the ability of a company to generate operating cash flows. Cash flow efficiency ratios are used to assess the relationship between items in the income statement and balance sheet with cash flow from operations as disclosed in the cash flow statement. These ratios are as follows. Cash flow on revenues ratio This ratio is aimed at showing the ability of the company to turn revenue into cash. The higher the ratio, the better the ability. This ratio employs information provided by the statement of cash flow and the income statement. It is computed by dividing cash from operating activities by revenues.

Cash flow to revenues =

Cash flow from operations Revenues

Cash flow to net income ratio This ratio is sometimes called the operating index. It compares the companys profit with cash flow from operations and attempts to provide an index of the cash-generating productivity of operations. It is calculated as cash flows from operations divided by profit after income tax.

Operations index

Cash flow from operations Profit

54

 Cash flow return on assets ratio This ratio attempts to measure the companys return on assets in term of the cash flow generated from operations.

Cash flow return on assets = Cash flow from operations + Income tax paid + Interest paid Average total assets Cash flow return on stockholders equity ratio This ratio shows the ability of the company to generate a sufficient cash return for stockholders.

Cash flow return on stockholders equity ratio =

Cash flow from operations Average Stockholders Equity

Cash flow per share ratio This ratio indicates the operating cash flow attributable to each common share. It is defined as cash available to common stockholders divided by the weighted average number of common shares outstanding.

Cash flow per share ratio = Cash flow from operations Preferred Dividends Average number of shares of Common Stock outstanding

These ratios are advocated as being able to help users of financial statements in making decisions. For example, cash flow on revenue ratio assists credit managers in analysing the credit risk of the firm. Generally, the higher the ratio, the better the credit risk (Dennis 1994). In addition, it is suggested that the cash flow on revenue ratio and the debt coverage ratio are considered in analytical procedures to detect financial statement fraud (Fleming 2004).

55

3.5 Application of Cash Flows and Accrual Accounting Data in Predicting Future Cash Flow

Published studies relating to cash flow prediction are reviewed in this section. The first part of this section discusses research focusing on earnings and cash flow data as a single predictor of future cash flows. The second part reviews studies highlighting accrual accounting and cash flow data. The last part provides a review of other research on cash flow prediction.

3.5.1 Accrual based and cash flow based data in predicting future cash flow

Following the assertion of FASB (1978) that earnings provide a better indicator of future cash flows than cash flows, a number of researchers have studied the ability of earnings and cash flows to predict future cash flows.

Greenberg, Johnson & Ramesh (1986) provided evidence supporting FASBs statement regarding the importance of earnings. They sought to test empirically whether current earnings or current cash flows are the better predictor of future cash flow. A sample of 157 industrial companies selected from the Compustat database for the period 1963-1982 was employed in their study. The operating cash flow variable used in this study was approximated by indirectly adjusting earnings for non-cash items and changes in current assets and current liabilities (excluding the current portion of long-term debt). The average linear relationship between each companys current cash flow and its previous cash flow and the average linear relationship between each companys current cash flow and its previous earnings were estimated by using ordinary least-squares regression. Not only models of one years data but also multi-year data, including two and three years, were examined. They reported that current earnings have the ability to predict future cash flows better than current cash flows for each lag period of one to five years and for each multi-lagged period of two or three years.

Bowen, Burstahler and Daley (1986) examined the relationship between earnings and various CF measures. Additionally, they compared the predictive ability between cash flow variables

56

and earnings to forecast future cash flow. They focused on the differences among various definitions of CF as follows.

1. The traditional CF measure (NIDPR) calculated by adding back depreciation and amortization (DPR) to net income before extraordinary items and discontinued operations (NIBEI), that is: NIDPR = NIBEI + DPR 2. Working capital from operations (WCFO), the second traditional CF measure, calculated by adjusting earnings to remove the effects of gains and losses on asset sales, gains and losses on investments accounted for by the equity method, amortization of bond premiums or discounts, and deferred taxes. WCFO = NIDPR + adjustments for other element of NIBEI not affecting working capital 3. Cash flow from operations (CFO) calculated by adjusting WCFO by changes in noncash current assets and current liabilities (excluding change in cash, notes payable and the current portion of long term debt) 4. Cash flow after investment, before financing (CFAI) which equalled CFO plus proceeds from the sale of property, plant and equipment and investment, minus amount of capital expenditures during the period, and new investment. 5. Change in cash (CC) during the period.

They obtained data from financial statements of 324 firms on Compustat for the period 19711981. The results showed that accrual earnings have a high correlation with traditional measures (NIDPR and WCFO) of CF, but low correlation with alternative measures (CFO, CFAI and CC). Simple random walk prediction models were considered in examining the relative performance of each CF variable to FCF. Findings provided preliminary evidence that net income plus depreciation and amortization, and working capital from operations, provide the best predictive power of cash flow from operating activities. Finally, the findings do not clearly confirm that earnings are a better signal for predicting future cash flows than CF variables.

57

Murdoch and Krause (1990) addressed three questions:

1. Are current years earnings or cash flows from operations a better predictor of future cash
flows from operations? 2. Are the current or non-current components of earnings more important in predicting future cash flows from operations? 3. Does using earnings or cash flow data over a long period provide a more accurate forecast than those over a short period for cash flow prediction? (Murdoch & Krause 1990, p. 3)

Their study emphasised the percentage changes in annual cash flow return. Net income, working capital and cash flow from operations were the main variables. Cash flows from operating and working capital were measured by adjusting net income. Data on the Compustat tapes for the years 1966-1985 (20 years) were employed to compute the variables. Companies were selected with respect to size, industry categorisation, fiscal year and other factors. In order to control the difference in sizes and changes in purchasing power of the dollar over time, every variable was deflated by the firms common equity. Then every independent variable (cash flow return, working capital return and return on equity) was analysed in the form of percentage changes to forecast percentage changes of cash flow return.

In terms of their question 1, the Murdoch and Krause (1990) conclusion supported the assertion of FASB that earnings are a better predictor than cash flow from operations. In answer to question two, they found that the current component of earnings included in the measurement of working capital was a better predictor than the non-current component included in measuring earnings. Finally, they concluded that the accuracy prediction of the model can be improved by utilising a long period of data.

Percy and Stokes (1992) replicated the test of Bowen, Burstahler and Daley (1986) studying two traditional cash flow measures (net income plus depreciation and amortization, and working capital from operations) and a more refined measure (working capital from operations plus additional adjustments for changes in non-cash current assets and current liabilities) and extended their study by analysing the relationship between cash flows and earnings across industries using Australian data. They employed data from the Australian 58

Graduate School of Management (AGSM) Annual Report File for the year 1974-1985 for 107 companies, a period comparable to that used by Bowen, Burstahler and Daley (1986). For the industry analysis, companies were classified in industry categories using the Australian Stock Exchange Industry grouping. Companies that changed industry grouping were excluded. Ninety-nine companies from 23 industries were examined in the industry analysis. The results of their study confirmed the evidence from Bowen, Burstahler and Daley (1986) in that the traditional cash flow measures showed more correlation with accrual income than the more refined cash flow measure. In addition, the correlation between the traditional cash flow measures and the more refined cash flow measure was low. These results were not different across firms. The results of the test of predicting future cash flow corresponded with Bowen, Burstahler and Daley (1986) in that the traditional cash flow measure provided more accurate predictions than did either earnings or the more refined cash flow measure for either forecasting one or two years ahead. However, the result indicated that the relative predictive abilities differed across industry. These results may have been caused by the limitations on the sample sizes used in the analysis.

McBeth (1993) examined the ability of cash flows and earnings to predict future cash flow by using cash flows from operations directly from the statement of cash flow and net income from the income statement. A potential sample of 4,415 companies on COMPACT DISCLOSURE was selected by limiting the sample of the reported cash flow statement in each of the years 1988, 1989, 1990 and those that employed a December 31 year end. There were only three years of data because companies had only been required to disclose a cash flow statement since 1988.

In simple regression analysis, current cash flows from operations were the dependent variable, whereas net income and /or cash flows from operations for last one or two years were the independent variables. In conclusion, McBeth (1993) suggested that neither past net income nor past cash flows from operations provide a better predictor of future cash flows.

Finger (1994) used a time series model to test firm-specific predictive ability for future cash flow over the entire time period. Annual data for the 50 sample firms spanning the years 1935-

59

1987 was obtained from the Compustat Annual Industrial File from 1968 to 1987 and supplemented with hand-gathered Annual Report information from 1935 to 1967. Those firms were members of the 1988 Fortune 500. Cash flows from operations were approximated by adjusting income before extraordinary items for depreciation, deferred taxes, changes in noncash current assets, and changes in current liabilities excluding current maturities of longterm debt. Earnings were represented by net income before extraordinary items.

Finger indicated that earnings used either alone or together with cash flow, were an important predictor of future cash flows. However, the results revealed that current cash flows were a superior predictor of future cash flows compared with current earnings for short-term prediction.

Quirin et al. (1999) re-examined the relative ability of earnings-based and cash flow-based measures to predict one-year ahead operating cash flows using actual cash flow data from the cash flow statement for an eight-year period. They collected sample observations from the 1997 version of Compustat PC Plus. A firm was selected if it had complete accounting information for all years of the sample period (1988-1996). After the screening procedure 1,442 firms per year over the 8-year prediction period of 1989-1996 were included in the study. They studied four different predictors of actual cash flows from operations: net income before extraordinary items (NIBEI), net income plus depreciation (NIDPR), working capital from operations (WCFO) and actual cash flow from operations (ACFO). Simple ordinary least squares regression procedures were used in the analysis.

The results of four models for each year from 1989 to 1996 were inconsistent. Actual cash flows from operations were the best predictor of future cash flow in five of the eight years. WCFO was a better predictor than others in 1989 and 1990, whereas NIDPR had the highest predictive power in 1996. NIBEI was not a superior predictor in all eight years. In addition, the result for the pooled sample supported that ACFO was the best predictor for the 19891996 time period followed by WCFO, NIDPR and NIBEI respectively. This study concluded that accrual based earnings provided a lower predictive ability than cash flow based predictors.

60

Quirin, OBryan and Berry (2000) replicated Bowen, Burgstahler and Daley (1986). They used the same four distinct predictors of actual cash flow including NIBEI, NIDPR, WCFO and actual cash flow from operations but focused only on oil and gas companies. Data were collected from the 1998 version of Compustat PC Plus for period of 1988-1997. Simple ordinary least squares regression was employed to predict cash flows with the one-year lagged predictors.

The result of the research provided evidence that NIDPR and WCFO were the best predictors, followed by ACFO and NIBEI for the pooled data. Meanwhile, for the annual results, WCFO was the best predictor, followed by ACFO, NIDPR and NIBEI respectively.

Jordan and Waldron (2001) evaluated the ability of accrual based measures versus cash based measures in predicting future cash flows. Five regression models were investigated in their study. Each predictor in each model is shown below.

1. Income before extraordinary items and discontinued operations (NI); 2. NI plus depreciation and amortization expenses on noncurrent operating assets (NIDPR); 3. NIDPR plus or minus adjustments to earnings to remove the effects of gains or losses on transactions impacting earnings but not effecting cash flows (WCFO); 4. WCFO plus or minus changes in the current operating assets and liabilities (CFO), this meanings is cash flows from operations presented on the cash flow statements under SFAS No. 95; 5. Net change in cash and cash equivalents during the year.

Each model was used to predict one year ahead cash flows with 10 years of quarterly data pooled for 30 companies in the petroleum industry derived from Compustat files. The results of regression analysis indicated that each model provided significant predictive power for predicting future cash flows. In addition, both coefficients of determination (r2) and

61

percentage prediction errors suggested that NIDPR model produced more predictive power than other models.

In summary, previous research has shown that cash flows and earnings have a role in predicting future cash flows, however the results are still unclear and it is not possible to conclude whether cash flows or earnings provide a better predictor. Some researchers indicated that cash flows are a better predictor (Finger 1994) whereas some reported contrasting results, such as Murdoch and Krause (1990).

Researchers have measured the cash flow variable in many ways including cash flows from operations (such as Finger 1994, Greenberg, Johnson & Ramesh 1986, e.g.), cash flows calculated by adding back depreciation and amortization to net income before extraordinary items and discontinued operations (NIDPR), working capital from operations (WCFO), cash flows after investment and change in cash (Bowen, Burstahler & Daley 1986). These studies found that those different measures provided different predictive ability.

Most studies focused on cash flows from operations which were approximated by adjusting income on income statements (Bowen, Burstahler & Daley 1986, Finger 1994, Greenberg, Johnson & Ramesh 1986). After FASB required firms to present statements of cash flows in 1988, some researchers attempted to test the predictive ability of cash flow data reported on the cash flow statements instead of cash flow estimated by adjusting income (McBeth 1993; Quirin et al. 1999; Quirin, OBryan & Berry 2000). However, their results were inconsistent. Quirin et al. (1999) reported actual cash flows from cash flow statements were a better predictor than earnings. On the other hand, McBeths (1993) findings revealed that neither cash flows nor earnings provided a good predictor. The summary of those studies and their findings are shown in Table 3.1.

62

Table 3.1 Summary of Studies Investigating Earnings and Cash Flow Data as a Predictor of Future Cash Flows Study Test Data Dependent Variables Greenberg, Johnson and Ramesh (1986) 157 industrial companies on Compustat data for 1963-1982. Bowen, Burstahler and Daley (1986) 324 firms on Compustat for 1971-1981. Earnings, NIDPR, WCFO, CFO, CFAI CC Murdoch and Krause (1990) Compustat data for 20 Years (19661985). Data deflated by common equity. Percentage change of Net Income, WCFO, &CFO divided by common equity. Earnings were a better predictor than CFO. WCFO were a better predictor than earnings. The number of year of data improved the accuracy of prediction. Percy and Stokes (1992) 107 firms on the AGSM annual report file for 19741985. McBeth (1993) 4,415 firms on COMPACT DISCLOSURE for 1988-1990. Finger (1994) 50 firms of the 1988 Fortune 500. Annual data from Compustat for 1968-1987 and hand-gathered annual report for 1935-1967. Earnings CFO Either earnings used alone or with cash flow were a significant predictor. CF was a better shortterm predictor. NI, Actual CFO from cash flow statements Neither NI nor CFO provided a better predictor. NIBEI (Earnings) NIDPR, WCFO, CFO NIDPR and WCFO provided a better predictive power than Earnings and CFO. Overall earnings did not provide a better predictor than CF. NIDPR were the best predictor of CF. Earnings, CFO calculated by adjusting earnings. Earnings were a better predictor than CF. Results

63

Table 3.1 (Continued) Quirin et al. (1999) 1,442 firms on Compustat for 1988-1996. Actual CFO from cash flow statement, NIBEI, NIDPR, WCFO Actual CFO was the best predictor in five of eight years for each year and the pooled sample. NIBEI (Earnings) provided a lower predictive power. Quirin, OBryan and Berry (2000) Firms on Compustat for 1988-1997. Actual CFO from cash flow statement, NIBEI, NIDPR, WCFO NIDPR and WCFO provided a good predictor, for the pooled data. For the annual results, WCFO was the best predictor.

Source: Developed for this study

3.5.2 Combined approach: cash flow and accrual-based approach

In addition to the prediction models based on only earnings or only cash flows, some researchers have incrementally studied accrual components of earnings by combining them with cash flow data in predicting future cash flows (Murdoch & Krause 1989; Barth, Cram & Nelson 2001; Dechow 1994; Dechow Kothari & Watts 1998; Stammerjohan & Nassiripour 2000/2001).

Murdoch and Krause (1989) investigated the ability of accrual accounting and cash flow measures to predict future cash flows. Furthermore, in addition to investigating only one variable, accrual accounting or cash flow, they examined a combination of accrual and cash flow data in predicting future cash flows. They developed hypotheses by considering accrual revenues and expense.

Murdoch and Krause (1989) employed 20 years of data from the Compustat tape to construct variables including income before extraordinary items (IBEI), working capital from operations (WCO), net sales (NS), and cash flows from operations (CFO). Each variable was deflated by

64

the average market value of common equity (MVE) to control for serial correlation that often occurs in relationships between monetary amounts measured over different time periods. Bivariate and multivariate regressions were employed to test seven hypotheses using the following regression models:
= + = + = +

CFRit CFRit CFRit CFRit CFRit CFRit CFRit

1ROE1,t-1 + it

(1) (2) (3) (4) (5) (6) (7)

1WCR1,t-1 + it 1ROS1,t-1 + it

= + 1CFR1,t-1 + it

= + 1CFR1,t-1 + 2 ROE1 , t-1 + it = + 1CFR1,t-1 + 2 ROS1 , t-1 + it CFRit ROE1,t-1 WCR1,t-1 ROS1,t-1 1

= + 1CFR1,t-1 + 2 WCR1 , t-1 + it

where

= Cash flow return for firm i period t = Return on equity for firm i period t-1 = Working capital return for firm i period t-1 = Return on sales for firm i period t-1 = intercept; = coefficient of first (or only) independent variable in regression;

= coefficient of second independent variable in regression; = error term

Murdoch and Krause (1989) found that accrual earnings are a better predictor of operating cash flows than operating cash flows themselves. Also, sales and working capital are better predictors of cash flows than earnings. In summary, the empirical results of their research showed that operating cash flows are not important variables in predicting cash flow.

Dechow, Kothari and Watts (1998) developed a model of earnings, cash flows and accruals. They started the model by assuming that sales generate the accounting cycle related to

65

accounts receivable, accounts payable, and inventory. The Dechow, Kothari and Watts (DKW) model explained why current earnings are a better predictor of future operating cash flows than current operating cash flows. For their data analysis, they used annual financial data obtained from the Compustat Annual Industrial and Annual Research tapes from 1963 to 1992. The final sample consisted of 667 surviving firms. Earnings were represented by earnings before extraordinary items and discontinued operations. The cash flow variable was calculated as operating income before depreciation minus interest minus taxes minus change in non-cash working capital. Operating accrual was earnings minus cash flow. Simple and multiple regression analysis were used to establish the cash flow prediction models.

The findings showed that current earnings were a better forecaster of future cash flow than current cash flow. In addition, earnings provided incremental power in predicting future cash flow. The difference in the ability of current earnings and current cash flows to predict future cash flows was a positive function of the firms expected operating cash cycle.

Barth, Cram and Nelson (2001; 2002) extended the DKW model to investigate the role of the components of accruals change in accounts receivable (AR), change in accounts payable (AP), change in inventory (INV), depreciation, amortization and other accruals. They tested the predictive ability of earnings, cash flow and the six-accrual components using regression analysis as in the following equations: CFi,t+1 = + t- EARNi, t- + ui, t
k =0

accruals for predicting future cash flows by disaggregating earnings into cash flow (CF) and

CFi,t+1

= + CFCFi,t + ARARi,t+ IINVi,t + APAPi,t + DDEPRi,t + AMAMORTi,t + OOTHERi,t + ui, t

Where CF AR AP EARN = cash flow from operations = earnings = change in accounts receivable = change in accounts payable 66

INV DEPR AMORT OTHER i t- , k

= change in inventory = depreciation = amortization = the aggregate of other accruals = firm = year and k ranges from 0 to 6 = coefficient of independent variables in regression

Data used in the analysis was from the Compustat annual industrial and research file between 1987 and 1996. EARN was represented by income before extraordinary items and discontinued operations. CF was net cash flow from operations adjusted for the accrual portion of extraordinary items and discontinued operations. The accrual components were from the cash flow statements and computed from balance sheet data. Financial service firms were excluded from the analysis.

The results of their research indicated that each accrual component reflects different information relating to future cash flow. They concluded that long-term accruals have a role in predicting future cash flows. Additionally, several past aggregate earnings have incremental descriptive power to aggregate current earnings for predicting future cash flows, whereas disaggregated current earnings have considerably more predictive capacity than several lags of aggregate earnings. Furthermore, they confirmed DKWs analysis that the superior predictive ability of aggregate earnings relative to cash flows varies with firms operating cash cycles. In comparing each model, they indicated that the cash flow and components of accruals specification has the most predictive ability for cash flows up to four years in the future, followed by cash flow and aggregate accruals, cash flow only and multiple lags of aggregate earnings. Also the findings implied that current earnings only, and current year accrual components, have predictive ability less than cash flows only.

Stammerjohan and Nassiripour (2000/2001) studied the evidence resulting from the research of Barth, Cram and Nelson (2001). The difference between their studies is that Stammerjohan and Nassiripour (2000/2001) presented results from out of sample period versions of the

67

model in the within-sample-period metrics and added two more accrual components of earnings, changes in taxes payable (TXP), and deferred tax expense (TXDI). The increment content of deferred tax expense in predicting future cash flows was proven by Cheung, Krishnan and Min (1997).

The study employed data of firms listed on the June 1999 Compustat during 1988 and 1997. Firms were included in the sample if they met the requirements. In the pool sample, they needed to have a complete time series of information during 1988 and 1997 and also meet other criteria including having total assets more than $1 million, no involvement in merger and acquisition activities, reporting income before extraordinary items or discontinued operations and reporting a cash flow statement. The final pooled sample had 944 firms. For the cross-sectional model, the number of firms in three-year samples and five-year samples were between 1,887 and 3,603 firms.

Stammerjohan and Nassiripours research results were consistent with Barth, Cram and Nelson (2001). That is, prior cash flows are a better predictor of future cash flow than prior earnings. The components-of-earnings model, including both cash flows and total accruals, provide better prediction of future cash flows than does the model based only on prior earnings. However, their study does not precisely prove that the two components of the model have better predictive power than models based only on prior cash flows as found by Barth, Cram, and Nelson (2001).

Cheung, Krishnan, and Min (1997) investigated the incremental information of deferred income tax. First, they tested the ability of deferred tax to predict future tax payment, and predicted future operating cash flows. They examined the predictive ability by adding the deferred tax variable into the Lorek and Willinger (1996) model. In their study, operating cash flows are an independent variable, and were calculated by operating income before depreciation, minus interest expense, current portion of income tax expense and increase in net working capital other than cash and securities, net of short-term debt. This test showed that the addition of the deferred tax variable improves the prediction of future cash flow. In particular, it is more useful if companies have large amounts of deferred tax.

68

Krishnan and Largay (2000) examined the predictive ability of the direct method cash flow format in predicting future cash flows, comparing cash flow information from the indirect method format in various ways, focusing on firms which reported operating cash flows by the direct method. In the analysis, they used cash flow data from the NAARS database during 1988 to 1993. The sample included all firms which reported operating cash flows by the direct method, without limitations on year-end or the age of the firms.

Firstly, they tested whether the cash flow information of the direct method provides more accurate predictions than information from the indirect method. They compared two prediction models, one forecasted by the direct method information predictor and one by the indirect method cash flow information, using pooled time-series and cross-section based models. The findings suggested that the direct method cash flow format provides more accurate predictors than the indirect method data.

Secondly, they investigated whether including direct method information, particularly cash collected from customers and cash paid to suppliers and employees, enhances the predictive power of the model. They employed the Lorek and Willinger (1996) multivariate, time-series cash flow model as a benchmark model. The research results provided evidence that gross amounts of cash receipts and cash payments are more relevant than information about their net amounts.

Thirdly, they tested whether direct method cash flow information can be determined indirectly without causing excessively onerous costs. For this purpose, they estimated the amount of cash collected from customers and cash paid to suppliers and employees using income statement and balance sheet data derived from Compustat. They concluded that direct method cash flow should be directly observed from the cash flow statement. Krishnan and Largay (2000) also examined prediction errors to measure the predictive ability of accrual and cash flow data. For the final observation, they concluded that past direct method cash flows provide a better predictor of future cash flows than earnings and accrual information alone.

69

In the case of Asian countries, Supriyadi (1998) studied the ability of accounting information to predict future cash flows of Indonesian firms. Regression analysis was used to construct prediction models. Earnings, cash flows and accounting information were derived from financial statements. The analysis was performed on both specific firm and pooled cross sectional year data. Five prediction models were developed as below: = + 1CFOt-1 + 2CFOt-2 + 2 D + t = + 1EAt-1 + 2EAt-2 + 2 D + t = + k=1 to 9 (kWik) + t = CFOt-1 = + 1CFOt-1 + 2CFOt-2 + 1EAt-1 + 2EAt-2 + 2 D + t

1. CFOt 2. CFOt 3. CFOt 4. CFOt 5. CFOt where, CFOt CFOt-i EAt-i t D Wik

= cash flows from operations, = the lagged values of cash flows from operations, = the lagged values of earnings, = time variable measured semi-annually, = dummy variable, equal to 1 for 1995-1997 and 0, otherwise, = a vector of 8 independent variables of lagged values (t-1 and t-2) for; earnings, revenues, current accruals, cash flows from operations and a dummy variable.

The first model was comprised of two year lags of cash flows, the second model was comprised of two year lags of earnings, the third model was comprised of two year lags of both earnings and cash flows, the fourth model was comprised of two year lags of earnings, revenues, current accruals and cash flows from operations and the last model was a random walk model containing one year lag of cash flows from operations. In the analysis, due to the study period involving two different accounting standards, a dummy variable was utilised to indicate the effect of different accounting standards on the accounting variables.

The study contained two sets of analysis. First, the 1990-96 period data set consisting of semiannual data from the first semi-annual reporting period of 1990 to the second semi-annual

70

reporting period of 1996 was employed to build the five prediction models, and the regression equation of each prediction model from the regression analysis were then used to predict cash flows from operations for the first semi-annual reporting period of 1997. Second, the 1990-97 period data set from the first semi-annual reporting period of 1990 to the first semi-annual report period of 1997 was used to build the five models and the regression equations generated from the analysis were employed to predict cash flows from operations for the second semiannual report period of 1997. The predictive ability of each model was measured by mean absolute percentage errors when each model was applied to prediction of cash flows of the first and second semi-annual reporting period of 1997.

The study found that for both firm-specific and pooled cross-sectional levels, the cash flow model (Model 1) outperformed the earnings model (Model 2) and the model containing both earnings and cash flows (Model 3) provided more significant predictive power than the model based only on either cash flows (Model 1), earnings (Model 2), or various accounting information (Model 4). However the predictive ability among the models was not significantly different and the predictive errors increased for the prediction of cash flows of the second semi-annual report period of 1997. The researcher explained that this may result from the impact of the Asian economic crisis occurring in the middle of year 1997 on accounting information of Indonesian firms. The results of model 4 showed that adding current accruals and revenues into the prediction model did not significantly provide more predictive power than only cash flows. This conclusion did not support the Accounting Standards Committee of the Indonesian Accountants Association (KPSAK) assertion that a set of accounting information provides more information to assess future cash flows than earnings or cash flows alone. Moreover this study indicated that regression models outperformed random walk models for predicting future cash flows. In conclusion, focusing on cash flows and the accrual components of earnings as combined predictors of future cash flow, previous research findings revealed that the components of earnings model including both cash flows and total accruals provide better prediction of future cash flows than does the model based only on earnings (Barth, Cram & Nelson 2001; Stammerjohan & Nassiripour 2000/2001). However they could not precisely confirm that the

71

Table 3.2 Summary of Studies using Cash Flow and Accrual Based Approach Study Test Data Dependent Variables Murdoch and Krause (1989) 20 years of data from Compustat. IBEI (Earnings), WCO, NS, CFO, Each variable was deflated by MVE. Dechow, Kothari and Watts (1998) Barth Cram and Nelson (2001) 667 surviving firms from Compustat for 1963-1992. from Compustat for 1987-1996. Earnings, ActualCFO, Components of INV, AP, DEPR, AMORT, OTHER. accruals: AR, Earnings, CFO. Earnings are a better predictor than CF. Earnings were incrementally useful in predicting future cash flow. Actual CF and components of accruals had the most predictive ability for FCF up to four years. Current earnings-only and current year accrual components have predictive ability less than CF only. Long- term accruals have a role in predicting FCF. Stammerjohan and Nassiripour (2000/2001) Data from Compustat for 1988-1997. 944 firms in the pooled sample and between 1,887 and 3,603 firms in the cross-sectional samples. Cheung, Krishnan and Min (1997) Data from Compustat for 1975-1994. CFO were calculated by operating income before depreciation, minus interest expense, current portion of income tax expense and increase in net working capital. Earnings, Actual cash flows, Components of earnings: AR, INV, AP, DEPR, TXDI. ACF is a better predictor of future cash flow than earnings. The components of earnings model including both CF and total accruals provide better prediction of FCF only earnings. The study did not precisely prove that the two components of the model have better predictive power than models based only on CF as found by Barth Cram and Nelson (2001) The deferred tax variable improved the prediction of future cash flows. Earnings are a better predictor than CFO. NS and WCO are better predictors than Earnings. CFO is not important in predicting cash flows. Results

AMORT, TXP and

72

Table 3.2 (continued) Krishnan and Largay (2000) Data from the NAARS database during 1988 to 1993. Actual CF, Variables in calculating CFO based on direct method and indirect method. Past direct method cash flows provide a better predictor of FCF than earnings and accrual data alone. The direct method cash flow format provides more accurate predictors than the indirect method data.

Supriyadi (1998)

Semi-annual data of Indonesian firms during 1990-1997

Earnings, Cash flows from operations, Revenues, and current accruals.

Model containing both cash flows and earnings provided the best predictor and cash flow model provided a better predictor than earnings model. Adding current accrual and revenues into model did not improve the predictive of the model.

Source: Developed for this study

combined predictors provided more power than a single predictor in predicting future cash flows such as in the Stammerjohan and Nassiripour study. Further they could not provide strong evidence that the model comprised of cash flows and accrual component of earnings provided a better predictor than the only cash flow model.

In comparing the predictive ability between cash flows and earnings, the results of Stammerjohan and Nassiripour (2000/2001) and Barth, Cram and Nelson (2001) rejected the statement of FASB (1978) that earnings provided a better predictor than cash flows. Their result differed from the results of Murdoch and Krause (1989) and others previous research discussed in Section 3.4.1 (such as Greenberg, Johnson & Ramesh (1986).

73

The Krishnan and Largay study (2000) highlighted the usefulness of information on the cash flow statement, by comparing the direct and indirect methods. The findings supported the cash flow statement reported by the direct method. The summary of studies in this section is shown in Table 3.2.

3.5.3 Other studies

In addition to cash flow and accrual components of earnings, other accounting data have been investigated to construct predictors for cash flow. Espahaodi (1988) aimed to identify the important predictors of cash flow by considering 29 accounting measures, comprised of 23 absolute accounting balances, four financial ratios and two dummy variables. A regression model was used to establish the relationship between cash flows and change in the explanatory variables. All independent variables were lagged by one period. A sample of 114 companies in four industries for five years data (1973-1978) was employed in the analysis. Accounting data were obtained from the annual Industrial Compustat data file. It was found that 21 variables were important in the cash flow prediction, however none of the signs of the coefficients for any variable were statistically significant.

Some research has focused on comparing the predictive ability of statistical techniques instead of comparing the ability of the predictors. Those studies are reviewed below.

Khumawala, Polhemus & Liao (1981) developed prediction models of future cash flow by utilising the Box-Jenkins methodology and compared the model with other five models. The competitive models are as below:

Nave models:

t (1) = Zt + (Zt-3 - Zt-4) t (1) = Zt + (Zt-3 /Zt-4)

t (1) = Zt + [(Zt-3 - Zt-4) + (Zt-7 - Zt-8)]/ 2 t (1) = Zt + [(Zt-3 /Zt-4) +(Zt-7 - Zt-8)]/ 2

(1) (2) (3) (4)

74

Financial analysts model:

t (n) = 1/105[28Zt + 25Zt-1 +22Zt-2+ 19Zt-3 +16Zt-4 +13Zt-5 + 10Zt-6 + 7Zt-7 +4Zt-8 +Zt-92Zt-10 - 5Zt-11 - 8Zt-12 - 11Zt-13 -14Zt-14] Where, t (n) Zt Zt-1, Zt-2, Zt-3, Zt-14 (5)

= predicted cash flow for time period t+n = the value of cash flows at quarter t = the value of cash flows at first, second, third,, and fourteenth quarters.

Their study focused on the airline industry, using quarterly data for the period 1965 to 1976 obtained from the Air Carrier Financial Statistics. The sample contained twenty-nine firms.

The prediction models were built for individual firms and each category of operation. The operation categories were trunk, local service, helicopter, intra-Alaska, intra-Hawaii, and all cargo carriers. The results indicated that the four nave models performed poorly and that they were not useful for predicting future cash flows. The results of individual firms showed that the predictive power of the Box-Jenkins model was equal to the financial analysts model, while the result of the prediction models developed for each operating category showed that the Box-Jenkins model was better than other models in every category. Moreover, they found that the model built for each category produced a better predictor than the model for individual companies. Their study provided evidence that past cash flows can be used to predict future cash flows.

Wertheim (1989) investigated the predictive ability among various cash flow prediction models. He also examined the difference in predictive power of the model among industries. A sample of 1,185 firms was selected from the annual Compustat data file for the years 1973 through 1987. Firms were classified into industry groups based on SIC codes. In examining cash flow prediction among industry groups, the industry groups that had less than 25 firms were eliminated. Cash flows from operations used in the analysis were calculated by

adjusting working capital from operations, for changes in working capital accounts.

75

Six models were investigated by using annual information including random walk, three year moving average, mean reverting, single exponential smoothing, double exponential smoothing and Holts two parameter smoothing model. It was found that the random walk model provided the largest error. Holts two parameter smoothing model had the smallest error. In addition, the results indicated that the accuracy of prediction among industries was significantly different.

Lorek, Schaefer and Willinger (1993) compared the ability of statistical patterns of cash flows and working capital series. They aimed to compare the accuracy of future fund flow predictions generated by univariate time-series models versus those obtained from the multivariate cross-sectional models. Models for the cash flow from operations (CF) series comprised a seasonal autoregressive model (SAR), a seasonally differenced, seasonal autoregressive model (SMA) and a seasonal random walk model (SRW). Working capital flow from operations (WCF) series models included an autoregressive and seasonal autoregressive (ASA) ARIMA model and the Griffin-Watts (GW) characterization.

They used data from the annual and quarterly Compustat databases from 1976 to 1986 to calculate two fund flow measures, working capital flow from operations (WCF) and cash flow from operations (CF). Quarterly WCF and CF time-series were constructed beginning in the first quarter of 1976 and ending in the fourth quarter of 1985 for 109 sample firms.

Lorek, Schaefer and Willinger concluded that univariate time-series models of cash flow and WCF provide more accurate predictors than the multivariate cross-sectional regression models. In addition, the SAR autoregressive-integrated-moving-average (ARIMA) model was suitable as a prediction model for the CF series and both the ASA and GW ARIMA models perform well on the WCF series.

Lorek and Willinger (1996) were interested in comparing the predictive ability of future cash flows between the multivariate cross-sectional models and univariate autoregressiveintegrated-moving-average models in addition to examining the time-series properties of quarterly cash flows.

76

They used data from both the annual and quarterly Compustat databases between 1979 and 1991. Cash flows from operations were calculated by using information from income statements and balance sheets. Three forms of cash flows including undeflated cash flows, cash flows per share and cash flows deflated by total assets were investigated. Only large and successful firms were selected for the sample. As a result, the findings of this research may not be generalised to newly-formed firms and failed firms. Predictive ability was considered for five cash flow prediction models, including three univariate time series models: the seasonal autoregressive (SAR), seasonal moving average (SMA), and firm-specific ARIMA model, Wilsons multivariate cross-sectional model and a multivariate time-series model. The independent variables in Wilsons model were comprised of current and lagged values of sales revenues, net earnings, cash flows from operations, current and non-current accruals and the most recent annual capital expenditures, whereas the independent variables in the multivariate time-series regression model consisted of the cash flows from operations, operating income before depreciation, accounts receivable, inventory and accounts payable variables. All models were estimated using data beginning with the second quarter of 1979 and ending with the fourth quarter of 1988 to generate cash flow predictions for the first quarter of 1989. The Friedman ANOVA ranks test was used to investigate the accuracy of the cash flow prediction.

Lorek and Willinger concluded that the multivariate time series regression model provided superior performance to both the undeflated CF and deflated CF prediction model to each of the quarters and years individually. The SAR model showed the worst performance. In

summary, their research implied that accrual accounting data enhance the accuracy of cash flow predictions.

According to Espahaodis (1988) research, other data derived from balance sheets and income statements based on the accrual accounting basis may be useful in predicting future cash flow. However, the research showed weak evidence on the predictive ability of those predictors.

Studies on the abilities of statistical techniques have provided evidence that predictive ability depends on the statistical techniques used. For example, Lorek and Willinger (1996) suggested that the multivariate time-series regression model provided superior performance,

77

whereas the SAR model provided the worst performance. Furthermore, Lorek, Schaefer and Willinger (1993) concluded that the SAR ARIMA model was suitable as a prediction model for the CF series and both the ASA and GW ARIMA models perform well on the WCF series. Those may be an indication that the predictive ability of a predictor is based on a statistical technique.

3.6 Research Gaps, Research Issues and Model of Conceptual Framework

3.6.1 Research gaps

Based on the review of the relevant literature, the aim of this section is to provide an overview of the findings on how cash flow and accrual accounting data predict future cash flows. According to the assertion of FASB, which is similar to ISAC, cash flow data on the statement of cash flows can be used with accrual accounting data on the balance sheet and income statement can help users to assess and estimate future cash flow. As noted before, FASB (1978) argued that earnings provide a better predictor of future cash flow than cash flows. Many researchers have investigated this issue and tried to construct cash flow prediction models, as reviewed in Section 3. In summary, it can be seen that their results are mixed. Some research has consistent results with FASB that earnings are a better predictor of future cash flow than cash flows (Greenberg, Johnson and Ramesh 1986). In contrast, some findings disagreed with the FASBs argument (Finger 1994, Bowen, Burgstahler & Daley 1986, Percy and Stokes 1992). However, McBeth (1993) provided evidence that neither cash flows nor earnings are a good predictor of future cash flows. Overall, the limitations of prior studies were observed as follows:

1) Cash flow measurement. Previous research has measured cash flow in different ways. Some research had been undertaken before FASB and ISAC issued the standard of reporting cash flow statements. Operating cash flow was measured as net income adjusted by non-cash items and change in current assets and current liabilities (Greenberg, Johnson & Ramesh 1986, 1986, Percy & Stokes 1992, Murdoch and Krause 1990). In contrast, some research employed cash flows from operations directly derived from the statement of cash flows

78

(McBeth 1993). Nowadays, most companies are required to release a statement of cash flows. To investigate the usefulness of the cash flow statement, cash flow data used in the research should be derived directly from the statements themselves, rather than derived from accrualbased statements. 2) Research variety. Most of the research investigating the predictive content of cash flows has focused too narrowly on operating cash flow, earnings and accrual components of earnings. Those previous studies have ignored the potential of other cash flow variables, in particular cash flow ratios. 3) The lack of research evidence. It appears that most future cash flow prediction research has been undertaken in the United States. No research has been undertaken on cash flow prediction in Thailand. In addition, companies listed on the Thai stock market have been regulated by TAS No.25 to disclose the statements of cash flows since 1994. TAS No.25 is adapted from IAS No.7. As mentioned in Chapter 2, US and Thai accounting standards have taken different approaches in the presentation of the cash flow statement. Therefore, it is possible that the results of research in the case of Thai companies could be different from the USA and other countries.

3.6.2 Research problem

This research intends to replicate the previous studies in order to fill gaps from previous research, by using cash flow information directly from cash flow statements. The following research problem is proposed:

How can cash flow and accrual accounting data be used to predict future cash flows of Thai listed companies?

3.6.3 Research questions

Several research questions established from the research problem are defined below.

79

Research Question 1. Are past earnings significant predictors of future cash flows of Thai listed companies?

Research Question 2. Are past cash flows significant predictors of future cash flows of Thai listed companies?

Research Question 3. Are past cash flows and accrual components of earnings significant predictors of future cash flows of Thai listed companies?

Research Question 4. Are there different predictive powers between three prediction models, earnings, cash flows and cash flows and accrual components of earnings models?

Research Question 5. Are past cash flow ratios significant predictors of future cash flows of Thai listed companies?

3.6.4 Research hypothesis

The following research hypotheses have been formulated to predict the answer to the related research question.

Research Hypothesis 1: Past earnings have significant predictive power in predicting future cash flows of Thai listed companies.

Research Hypothesis 2: Past cash flows have significant predictive power in predicting future cash flows of Thai listed companies.

Research Hypothesis 3: Past cash flows and accrual components of earnings have significant predictive power in predicting future cash flows of Thai listed companies.

80

Research Hypothesis 4: There are different predictive powers between three prediction models: earnings, cash flows and cash flows and accrual components of earnings models.

Research Hypothesis 5: Past cash flow ratios are significant predictors of future cash flows of Thai listed companies.

3.6.5 Conceptual framework

The following models are constructed to develop the research issues. These models will consider relationships between concepts and be used to guide data collection.

Model 1: Model of predicting future cash flows using earnings

This model seeks to depict the relationship between earnings and future cash flow as the first hypothesis. The test should show that earnings are significantly related to future cash flows of Thai listed companies. This model was tested using regression analysis.

Figure 3.2 Model of Predicting Future Cash Flows using Earnings Earnings

Future cash flows

81

Model 2: Model of predicting future cash flows using cash flows

This model depicts the relationship between cash flows and future cash flow as the second hypothesis. The test should show that cash flows are significant related to future cash flows in the case of Thai listed companies. This model was tested using regression technique.

Figure 3.3 Model of Predicting Future Cash Flows using Cash Flows Cash flows

Future cash flows

Model 3: Model of predicting future cash flows using cash flows and accrual components of earnings

This model deals with the third hypothesis. The third hypothesis expects the answer that cash flow and accrual components of earnings have a significant capacity in predicting future cash flows. This model was analysed by using a technique of multiple regression. Figure 3.4 Model of Predicting Future Cash Flows using Cash Flows and Accrual Components of Earnings Accrual component of earnings Cash flows

Future cash flows

82

Model 4: Model of predicting future cash flows using cash flow ratios This model is related to the final research question. The results are expected as the fifth hypothesis, that cash flow ratios provide a significant predictor of future cash flow. This model was tested by using technique of stepwise regression.

Figure 3.5 Model of Predicting Future Cash Flows using Cash Flow Ratios

Cash flow Ratios

Future cash flows

3.7 Conclusion

The literature review provides a discussion of the concepts involved in the use of cash flow and accrual accounting data in predicting future cash flow. In the review of cash flow prediction there were two parent disciplines and an immediate discipline in the application of cash flows and accrual accounting data in predicting future cash flow. These helped to define the gaps prior to constructing the research questions and hypotheses. Models of the theoretical framework were developed to justify the variables and their relationship and to develop data collection and analysis methodologies. The next Chapter develops methodology to build and test models of cash flow prediction for companies in the Thai stock market.

83

Chapter 4
Methodology
4.1 Introduction

In the previous chapter, the relevant literature, two parent disciplines, accrual and cash flow accounting, and empirical research in cash flow prediction were reviewed. In addition, previous research gaps were discussed and models of predicting future cash flow using cash flow and accrual accounting data were developed, in relation to solving the research problem.

This chapter explains the methodology used to collect the data and test the research hypotheses. It is comprised of eleven sections as shown in Figure 4.1. Section 4.1 introduces the outline of the chapter. Section 4.2 justifies the paradigm chosen for this research. Section 4.3 describes the research design and the method employed to achieve the outcome of the research. Section 4.4 provides the definition of variables and measurements in the data analysis. Section 4.5 describes the model building, including earnings, cash flows, cash flows and accrual components of earnings, and cash flow ratios model. Section 4.6 deals with measuring the explanatory power of models. Section 4.7 defines the predictive ability of models evaluated by mean of absolute percentage error. Section 4.8 describes the tests of the hypotheses. Section 4.9 identifies the data specification including the period of testing, sampling selection and data source. Section 4.10 explains the validity of the research. Finally, Section 4.11 presents the conclusion of the chapter.

84

Figure 4.1 Structure of Chapter 4: Methodology

4.1 Introduction

4.2 Justification of the Research Paradigm

4.3 Research Design

4.4 Variable and Measurement

4.5 Model Building

4.6 The explanatory Ability of Models

4.7 Evaluation of the Predictive Ability of Models

4.8 Hypothesis Testing

4.9 Data Specification

4.10 The Validity of Research

4.11 Conclusion

85

4.2 Justification of the Research Paradigm

A paradigm, a set of beliefs, deals with ultimates or first principles (Guba & Lincoln 1994). Researchers need to identify which paradigm they have chosen because a paradigm describes clearly for researchers what the research is about and what falls within and outside the limits of legitimate inquiry (Guba & Lincoln 1994). According to Guba and Lincoln (1994) there are four inquiry paradigms including postpositivism, constructivism, critical theory and positivism. Sometimes, constructivism and critical theory paradigms are grouped as one paradigm and called the phenomenological paradigm. The fundamental ideas of each paradigm can be defined by three basic questions, the ontological, epistemological and methodological questions. The answer to the ontological question provides the form and nature of reality. Epistemology questions the nature of the relationship between the researcher and what can be found. Methodology question guides how the researcher can find out answers to the research problem (Guba & Lincoln 1994).

The realism or post positivist paradigm supports the existence of a reality which can only be known imperfectly. The researchers focus on meanings and try to understand what is happening (Easterby, Thorpe & Lowe 1991). Furthermore, the triangulation of several perceptions of that reality helps to find a better description of events. The qualitative method is a key instrument in searching for answers (Easterby, Thorpe & Lowe 1991; Guba & Lincoln 1994; Perry, Riege & Brown 1999).

The realism paradigm is inappropriate for this study as the purpose of this research is to test hypotheses. This research uses quantitative method involving numbers, whereas the realism paradigm deals with the qualitative method which focuses on peoples attitudes and the meanings they attribute to people and events (Ticehurst & Veal 2000). Moreover, this research employs secondary data to answer the research questions. The researcher does not participate in what is being researched.

The critical theory paradigm answers the ontological question that reality seems to be understandable, called historical realism, but in fact reality is not factual, being creatively

86

formed by social, political, cultural, economic, ethnic and gender values (Guba & Lincoln 1994). Critical theory emphasises interpretation rather than prediction or explanation (Ryan, Scapens & Theobald 1992). Usually, an investigation using this paradigm is a long-term ethnographic and historical study of organisational processes and structure. Knowledge is grounded in social and historical routines (Perry, Riege & Brown 1999).

The critical theory paradigm is unsuitable for this research. This research aims at testing hypotheses, while research under the critical theory paradigm aims at transforming actions or approaches to strategy formulation.

The constructivism paradigm assumes that realities are understandable in the form of multiple, intangible rational constructions based on society and the experience of an individual. Constructivist research depends on the interaction between interviewer and interviewee, and the researcher needs to participate in the investigation. The research aims at finding meanings rather than values (Guba & Lincoln 1994; Perry, Riege & Brown 1999).

This research does not rely on the constructivism paradigm as it does not involve people. This research aims to utilise data from secondary data sources, so the researcher is separate from people. Statistical techniques are essentially used to analyse the data and the findings of this research depend on results of statistical analysis. Conversely, constructivism research findings rely on peoples perceptions.

The key idea of positivism is that the social world exists externally and understanding is compelled by incontrovertible natural laws and instruments (Easterby, Thorpe & Lowe 1991; Guba & Lincoln 1994). Positivists isolate themselves from what they are studying (Perry, Riege & Brown 1999), that is, the researcher works independently on the research without influencing objects or being influenced by them. This method is experimental or

manipulative. The positivist paradigm highlights chiefly quantitative methods and verification of hypotheses (Guba & Lincoln 1994). Research is carried out by testing and verifying hypotheses or models developed for the research. Some confounding conditions may be

87

carefully controlled or manipulated to avoid biases (Guba & Lincoln 1994; Ticehurst & Veal 2000).

This research is based on the positivism paradigm as testing hypotheses is its main purpose. This research intends to test whether cash flow and accrual accounting data can be used to predict future cash flows of firms by using secondary data. Three research models are analysed to test the hypotheses and obtain results. Secondary data are the existing data gathered and recorded by someone else prior to the current research, that is, they are usually historical, already assembled and do not require access to respondents or subjects (Zikmund 2000). So, the researcher studies independently to uncover the solution, without influencing or being influenced by what is observed. Furthermore, this research uses a quantitative approach involving numbers. Therefore, this approach is representative of the positivism paradigm.

Overall, the three paradigms: realism, critical theory and constructivism, mainly deal with qualitative methods. On the other hand, positivist research deals chiefly with quantitative methods. This research involves quantitative methods and therefore it supports the positivist paradigm. Moreover, the main purpose of this research is to test theory based on positivism, rather than building theory, which underpins the other three paradigms. The main tenets of the four paradigms are summarised in Table 4.1

88

Table 4.1 Summary of Four Research Paradigms Item Ontology Realism Critical realismReal reality but imperfectly understandable. Critical Theory Historical realismnearly factual formed by social, political, cultural, economic, ethnic, and gender values. Epistemology Researcher is part of research Interactive link between investigator and object. Findings are created. Methodology Case studies, convergent interviewing. Quantitative and qualitative methods. Based on a dialogue between the investigator and the subject of investigation, Focus groups. Qualitative method. Based on the interaction between interviewer and respondent. Qualitative method. Formulate and test hypotheses, quantitative method, taking large samples. Experiment and survey. Researcher participates in field work. Findings are arbitrated. Researcher is independent. Constructivism Relativism. Multiple, intangible constructions based on social and experience. Positivism Nave realism-Real reality and testable.

Source: adapted from Perry, Riege and Brown (1999); Guba and Lincoln (1994); and Easterby, Thorpe, and Lowe (1991)

4.3 Research Design A research design is a plan developed to achieve the research purpose. It aims to ensure the research can clearly answer the research problem, and involves systematising the research activity, involving the collection of data and analysing the data (Easterby, Thorpe & Lowe 1991). A good research design can provide valid conclusions and suggestions from the research (Ryan, Scapens & Theobald 1992).

4.3.1 The purpose of the study

The purpose of the research should be clearly identified whether it is exploratory, descriptive or testing hypotheses. An exploratory study intends to build theory and is conducted in order to understand the phenomena in the situation of interest and to advance knowledge (Sekaran 89

1992). A descriptive study describes the characteristic of variables in the phenomena of interest to the researcher from an individual, organisation, industry or other perspective (Sekaran 1992). In contrast, studies that aim at testing hypotheses aim to explain the nature of certain relationships, or establish the difference among groups, or the independence of two or more factors in the situation (Sekaran 1992).

As mentioned in Chapter 3, three hypotheses were constructed for testing by this research. That is, this research deals mainly with testing hypotheses to investigate the relationship between cash flow and accrual accounting variables and future cash flows of firms. Further, a descriptive study was initially undertaken in order to describe the variables of this research.

4.3.2 The basic research method

Many research methods can be used to collect data such as survey, experimental and using secondary data (Ticehurst & Veal 2000). However, the use of secondary data in which data already exists, is the most appropriate for this research. The major purpose of this research is to test hypotheses. Three hypotheses have been constructed to answer the research questions, in which the relationship between accounting data including cash flow and accrual accounting data, and future cash flow are investigated. Thus, the information needs for this research are accounting data. Accounting data from secondary data sources exist in data recorded and reported by accountants for each period. Using secondary data has the advantage of saving time and costs involved data collection (Ticehurst & Veal 2000; Zikmund 2000). In addition, the secondary data technique has been used in the cash flow prediction area by the majority of previous researchers (Barth, Cram & Nelson 2001; Bowen, Burgstahler & Daley 1986; Finger 1994; McBeth 1993; Quirin et al. 1999; Stammerjohan & Nassiripour 2000/2001). Those studies employed data from financial statements of sample firms.

4.4 Variables and Measurement

In basic terms, a variable represents a property of an event or phenomenon associated with a particular object (Ryan, Scapens & Theobald 1992). Variables can be dichotomised as

90

dependent and independent (Hair 1995). The impact of changes in the independent variable upon the dependent variable is considered in data analysis. That is, a dependent variable is predicted to change when independent variables change.

Based on the research hypotheses, this study focuses on five major variables: future cash flow, earnings, cash flows, accrual components of earnings and cash flow ratios. Future cash flows are investigated as the dependent or criterion variable caused by independent variables. Independent or explanatory variables are earnings, cash flows, accrual components of earnings and cash flow ratios. The independent variables are expected to provide predictive power to predict a dependent variable in the prediction model. These variables are defined and measured for this research as follows. 4.4.1 Dependent variable: future cash flows

Previous research has measured cash flows in different ways. Some researchers used proxy cash flow calculated from data on income statements and balance sheets. For example, Bowen, Burgstahler and Daley (1986) defined five measures of cash flow including net income before depreciation, working capital, cash flows from operations, cash flows after investment and change in cash. Murdoch and Krause (1990) described two cash flow measures, cash flows from operations and working capital from operations. On the other hand, some defined cash flows as actual cash flows from operations derived directly from cash flow statements, shown by Barth, Cram and Nelson (2001), McBeth (1993) and Stammerjohan and Nassiripour (2000/2001). Since this research intends to investigate the usefulness of cash flow statements based on Thai Accounting Standard No.25 (TAS 25), in the analysis, future cash flows of firms are defined as net cash flows from operations (CFO) reported in cash flow statements for the yeart, represented by the symbol of CFOt. Based on TAS 25, CFO are presented under the indirect method form and reported by adjusting earnings for non-cash items and for changes in current assets and current liabilities.

91

4.4.2 Independent variables

Year lags of earnings

In Model 1, year lags of earnings are the dependent variables. This study measures year lags of earnings as net income before extraordinary items and discontinued operations reported on income statements for year t-i symbolized by EARNt-i. This measure was used in various previous studies (Barth, Cram & Nelson 2001; Quirin et al. 1999; Stammerjohan & Nassiripour 2000/2001).

Based on prior research such as Cheng, Liu and Schaefer (1997a), accounting earnings are comprised of both accruals (ACR) and cash flows from operations (CFO). The relationship of the composition is as follow.

EARN

= CFO + ACR

(4.4.1)

In the prediction model, current cash flows and accrual components of earnings are independent variables. They are predicted to be indicators of future cash flows.

Year lags of cash flows

This study aims at finding the usefulness of cash flow data presented in cash flow statements of Thai listed firms, therefore year lags of cash flow is defined as net cash flows from operations reported on the cash flow statements for year t-i, denoted by CFOt-i. This definition of cash flow variables was measured by Bath, Cram and Nelson (2001) and Stammerjohan and Nassiripour (2000/2001).

Year lags of accrual components of earnings

Under the requirement of TAS 25 cash flows from operations are presented by the indirect method form. The key point of this method is to adjust income based on an accrual method to

92

be income based on a cash method. This concept is derived from the concept of earnings components mentioned above (see Equation 4.4.1). The indirect method format begins with net income (or loss) adjusted by non-cash items such as depreciation and amortization, and changes in current assets and current liabilities. The adjusting values of these items are obtained from income statements and balance sheets, and can be described by the following relationship:

CFO

= EARN-ACR

(4.4.2)

ACR

= EARN-CFO

(4.4.3)

or ACR

= AR + INV + AP+ DEP + OTH

(4.4.4)

AR AP DEP

Whereas, = Change in accounts receivable during the period = Change in inventories during the period = Change in accounts payable during the period = Depreciation and amortisation during the period = Change in other current assets and liabilities during the period

INV

OTH

A few previous studies have investigated the ability of year lags of accrual components of earnings to predict future cash flows such as Bath, Cram and Nelson (2001) and Stammerjohan and Nassiripour (2000/2001). The accrual components of earnings in those studies were calculated by using data from income statements and balance sheets. In the Stammerjohan and Nassiripour study (2000/2001), accrual components of earnings were investigated in two definitions, the aggregated accrual component and the disaggregated accrual component in which the accrual component were measured as the difference between earnings and cash flows from operations (see equation 4.4.3), and the disaggregated accrual components of earnings, were divided into its seven main components, including changes in accounts receivable, changes in inventories, change in accounts payable, change in taxes

93

payable, depreciation and amortisation, deferred tax expense, and changes in other current assets and liabilities. This research also investigated the predictive ability of the two definitions of the accrual component of earnings. The aggregated accrual components of earnings for the year lags (t-i), was denoted as ACR,t-i. But the disaggregated accrual component was divided into only five components, changes in accounts receivable, changes in inventories, change in accounts payable, depreciation and amortisation and changes in other current assets and liabilities and a one year lag of these main components was denoted as AR
t-1,

INV t-1, AP t-1, DEP t-1, and OTH t-1 respectively (as equation 4.4.4) because the

data of change in taxes payable and deferred tax expense was unavailable.

Cash flow ratios

Cash flow ratios have been developed to evaluate firms financial performance (Carslaw & Mills 1993; Plewa & Friedlob 1995). Cash flow ratios can be classified as sufficiency and efficiency ratios (Giacomino & Mielke 1993; Hoggett, Edwards & Medlin 2003). Cash flow sufficiency ratios include cash flow adequacy ratio, debt coverage ratio and dividend payout. Efficiency ratios include cash flow on revenues, operations index and cash flow return on assets. These cash flows can be used to evaluate the performance of a firm (Figlewicz & Zeller 1991; Sylvestre & Urbancic 1994). In studies of bankruptcy, cash flow ratios have been investigated as a signal of bankrupt firms (Mossman et al. 1998; Neill et al. 1991; Sharma 2001). Therefore, in the same way, cash flow ratios may be a signal of future cash flows.

Sufficiency ratios imply the ability of a company to generate sufficient cash to meet its cash flow needs. Cash flow adequacy is an attempt to assess the entitys ability to produce sufficient operating cash flows to cover its main cash requirements, specifically, the payment of debt, the acquisition of assets and the payment of dividends. They may also be useful to predict future cash flows of a firm. Although there is no empirical evidence supporting these ratios, this study initiated these ratios as a sign of future cash flow.

94

Efficiency ratios show the ability of a firm to generate cash (Giacomino & Mielke 1993). Cash flow on revenue ratio and cash flow on asset ratio were studied in a retail firms ability to generate sufficient operating cash flow (Gombola & Ketz 1983a). Cash flow on revenues ratio is aimed at showing the ability of the company to turn revenue into cash, the higher the ratio, the greater the ability, and may be useful in predicting future cash flow because it shows the ability of a firm to generate cash. So this research aims to investigate the abilities of these cash flow ratios in predicting future cash flow.

These cash flow ratios were described in Section 3.3.5. They are possible predictor variables in Model 4. The information contained in each cash flow ratio was expected to provide a sign or a predictive power of future cash flows. Each cash flow ratio is represented and calculated as shown in Table 4.2

Table 4.2 Cash Flow Ratios, Symbols and Calculation

Cash flow ratios Cash flow adequacy ratio Debt coverage ratio Repayment of borrowings ratio Dividend payment ratio Reinvestment ratio Cash flow on revenues ratio Cash flow to net income ratio Cash flow return on assets ratio Cash flow return on stockholders equity ratio Cash flow per share ratio Symbols CFR1 i,t CFR2i,t CFR3i,t CFR4i,t CFR5i,t CFR6i,t CFR7i,t CFR8i,t CFR9i,t CFR10i,t Calculation CFO /(Long-term debt payment + Payment for property, plant and equipment + Dividends paid) Total debt/ CFO Long-term debt payment / CFO Dividends Paid/ CFO Payment for property, plant and equipment/ CFO CFO/ Revenues CFO/Net income CFO/ Average total assets CFO/ Average stockholders equity (CFO-Preferred Dividends) / Ave no of shares of common stock *

Source: Developed for this research Note: * Ave no of shares of common stock : Average number of shares of common stock

4.5 Model Building

Previous researchers have used different statistical approaches to determine the prediction of future cash flow. For instance, the random walk model, a simple prediction model which

95

generates a prediction of future cash flows that depends solely on the most recent cash flows observation (Percy & Stokes 1992). That is, the random walk model predicts that next years cash flows are the same as the current years cash flows.

Some researchers used the Box-Jenkins methodology for determining the time-series properties and prediction of future cash flows (Khumawala, Polhemus & Liao 1981; Lorek & Willinger 1996). The Box-Jenkins model tries to identify patterns in the historical values of time-series and extrapolate those patterns into the future (Box & Jenkins 1976). In the BoxJenkins method, the analysis is divided into three stages; identification, estimation and forecasting (Sincich 1996). This method requires at least 50 observations (Box & Jenkins 1976). Additionally, quarterly data is needed in the time-series analysis, which discount this method as useful for the present research, because Thai listed companies have not reported quarterly cash flow statements.

The other statistical approach is the regression model which uses the method of least-squares to fit a line to the sample data (Hair 1995). The purpose of a model formulation in regression analysis is to allow for the study of properties of least-squares estimators of statistical theory (Myers 1986). Regression models have been used by a number of researchers such as Barth, Cram and Nelson (2001), Finger (1994), Greenberg, Johnson and Ramesh (1986), McBeth (1993), Murdoch and Krause (1989), and Stammerjohan and Nassiripour (2000/2001). The technique of regression is appropriate for this research. In order to answer the research questions and test the hypotheses, four prediction models were constructed to present the relationship between dependent and independent variables. The relationship was described by an estimated equation and analysed by using the regression techniques. The simple regression technique was used when a dependent variable was predicted by the knowledge of a single independent variable while the multiple regression technique was employed when the dependent variable was predicted by the knowledge of several independent variables (Berry & Feldman 1985; Hair et al. 1998). The statistical results of the regression analysis such as adjusted R2, F-statistic, and the level of significance were used to interpret the model and test the hypotheses. The regression models were constructed as described in Sections 4.5.1 to 4.5.4.

96

4.5.1 Earnings model

The first model, earnings model, was constructed to investigate the predictive ability of past earnings in predicting future cash flows of Thai listed companies. Previous research has found that past earnings had significant power to predict future cash flows (for example: Finger 1994; Krause & Murdoch 1990; Stammerjohan & Nassiripour 2000/2001). In this study the relationship between past earnings (EARN) and future cash flow from operations (CFO) was analysed. This test provides a solution of the first research question as described in Chapter 3. The research question is restated below.

Research Question 1. Are past earnings significant predictors of future cash flows of Thai listed companies? Some previous researchers such as Greenberg, Johnson and Ramesh (1986), McBeth (1993) and Stammerjohan & Nassiripour (2000/2001) have incrementally investigated past earnings by containing more than one year lags of earnings in the prediction model and some suggested that additional lags of earnings can improve the predictive power of a prediction model. This research has also tested this argument. The model was tested in three ways of analysis with one, two and three-year lags of earnings, EARNt-1, EARNt-2, and EARNt-3 respectively. Estimated equations for one, two and three-year lag analysis were below.

Model 1.1: Simple regression model CFOt =

+ 1EARNt-1 +

(4.5.1.1)

Model 1.2: Multiple regression model CFOt =

+ 1EARNt- 1 + 2EARNt- 2 +

(4.5.1.2)

Model 1.3: Multiple regression model CFOt =

+ 1EARNt- 1 + 2EARNt- 2 + 3EARNt- 3 +

(4.5.1.3)

97

Where, CFOt EARNt-1 EARNt-2 EARNt-3 = = = = cash flows from operations for year t, earnings for year t-1, earnings for year t-2, earnings for year t-3, the unknown parameters, the error term.

,1, 2, 3 = =

As mentioned in Subsection 4.4.2, past earnings are expected to provide a positive relationship to the future cash flows. Thus, if they have an explanatory ability, a significant and positive sign of them should be found in Model 1.

4.5.2 Cash flows model

The second model, the cash flows model, was examined to answer Research question 2. Prior research suggested that year lags of cash flows themselves provide a good predictor of future cash flows (Barth, Cram & Nelson 2001; Quirin et al. 1999; Stammerjohan & Nassiripour 2000/2001). So this research also tests the predictive ability of year lags of cash flows in predicting future cash flows in Thai listed companies. The research question is restated below:

Research Question 2. Are past cash flows significant predictors of future cash flows of Thai listed companies?

In the same manner as earnings, cash flows from operations were investigated in three ways, one, two, and three-year lags. Three equations were estimated as below:

Model 2.1: Simple regression model + 1CFOt-1 +

CFOt

(4.5.2.1)

98

Model 2.2: Multiple regression model CFOt = + 1CFOt- 1 + 2 CFO t- 2 + Model 2.3: Multiple regression model CFOt = + 1CFOt- 1 + 2 CFOt- 2 + 3 CFOt- 3 + where, CFOt CFOt-1, CFOt-2, CFOt-3 , 1, 2, 3 = = = = = = cash flows from operations for year t, cash flows from operations for year t-1 cash flows from operations for year t-2 cash flows from operations for year t-3 the unknown parameters, the error term.

(4.5.2.2)

(4.5.2.3)

Year lags of cash flows are expected to express a positive relation to the future cash flows and the additional lags of cash flows were expected to provide an incremental predictive power. Thus, if they have predictive ability, a significant and positive sign of them should be found in model 2.

4.5.3 Cash flows and accrual components of earning model

The third model, cash flows and accrual component of earnings model, was developed to find the solution of Research question 3. This model was derived from the extension of the second model. The accrual component of earnings was combined into the second model. Therefore this model tests whether cash flows and accrual components of earnings have significant capacity over either earnings or cash flows for predicting future cash flows. Research question 3 is repeated as below:

Research Question 3. Are past cash flows and accrual components of earnings significant predictors of future cash flows of Thai listed companies?

99

Similar to the two previous models, one, two and three-year lags of cash flows and accrual components of earnings were examined. This model was analysed by using a technique of multiple regression.

Model 3.1: CFOn,t Model 3.2: CFOn,t

= + 1CFOt-1 + 2 ACR t-1 +

(4.5.2.1)

= + 1CFOt-1+ 2CFO t-2 + 3ACRt-1+ 4ACRt-2 + (4.5.2.2)

Model 3.3: CFOn,t

= + 1CFOt- 1 + 2 CFOt- 2 + 3 CFO t- 3 + 4ACRt-1 + 5 ACRt-2 + 6ACRt-3 + (4.5.2.3)

where CFOt ACRt-1 ACRt-2 ACRt-3 = cash flows from operations in year t, = accrual components of earnings for year t-1, = accrual components of earnings for year t-2, = accrual components of earnings for year t-3,

1, 2, 3, 4, 5, 6 = the unknown parameters, = the error term.

Past cash flows and accrual components of earnings are expected to express a relationship to the future cash flows and the additional year-lags were predicted to provide incremental predictive power. Additionally, this model was expected to provide a better predictor than the two previous models. Thus, if they have predictive ability, a significant sign of past cash flows and accrual components of earnings should be found in Model 3.

100

4.5.4 Cash flow ratios model

The final model consists of cash flow ratios. This model was constructed to answer Research question 4 as below: Research Question 4 . Are cash flow ratios significant predictors of future cash flows of Thai listed companies?

This aims to investigate whether cash flow ratios provide a good predictor of future cash flow. The model was tested by a stepwise regression technique. = + iCFRt-1 +
i=1 10

CFOit

(4.5.3)

where ,1,2,, n CFRt-1 = the unknown parameters, = Cash flow ratio1-10, for year t-1, = error term.

Since there were many cash flow ratios considered to be predictors in the model, variable selection technique was used to choose which cash flow ratios are important. The objective of choosing is to ensure that important variables are included in the model and the unimportant variables are eliminated from the model. Also, the order of entry of variables is assigned by the statistical technique (Dielman 1991; Hair 1995; Tabachnick, Fidell & Osterlind 2001). Stepwise regression was selected to examine each cash flow ratio. Each predictor variable is considered for addition prior to developing the regression equation. In the procedure, each possible predictor variable in simple regression was examined (Dielman 1991). Then the explanatory variable providing the largest partial F statistic was chosen to add to the model. Finally, the stepwise procedure generated suitable equations for the model.

101

4.6 The Explanatory Ability of Models

After analysing the relationship between predictor variables and future cash flows in each model, the explanatory ability of each model was measured. The adjusted R2 value or coefficient of determination was used to determine how well a set of predictors statistically explains an independent variable (Hair, Anderson, Tatham & Black 1992, Neuman 2003). It is an important indicator of the predictive accuracy of the model (Cooper & Emory 1995). The adjusted R2 value has been employed by previous prediction research to evaluate the explanatory power between models, such as Barth, Cram and Nelson (2001), Quirin et al (1999), Greenberg, Johnson and Ramesh (1986), and McBeth (1993). The model producing a high adjusted R2 value is a good explanatory model and it can be an important predictive model.

4.7 Evaluation of the Predictive Ability of the Models in Out-of-sample In evaluating the predictive ability between models with R2, it has been argued that it is not a good indicator because the model may have a large prediction error (Jordan & Waldron 2001). Some previous research utilised the analysis of residuals to measure the predictive ability of models (Jordan & Waldron 2001; Stammerjohan & Nassiripour 2000/2001). Furthermore, some researchers partitioned data into two subsamples: an estimated sample for creating the regression model and a validation sample or out-of-sample for testing the model (Stammerjohan & Nassiripour 2000/2001). This method can be used to validate a model (Myers 1990). So, in addition to adjusted R2, this research employed the analysis of residuals involving the mean absolute percentage error to evaluate the predictive abilities of the prediction models. The mean absolute errors were generated from the out-of sample period. That is, the sample was split into two parts. The first sample, the estimated sample, was for creating regression equations of each prediction model and the second sample, the-out-of sample, was for testing the estimated equation of the three models; earnings, cash flows and cash flows and accrual components of earnings models. The regression equations of each model generated from the validation sample were used to calculate predicted cash flows.

102

After calculating the predicted values of cash flows from the out-of sample, the r square (r2) and mean absolute percentage errors of prediction (MAPE) were calculated for the out-ofsample period. A model producing relatively low MAPE would be considered to be a better predictor than model yielding higher MAPE. The MAPE is calculated as below: MAPE = mean|(predicted cash flows- actual cash flows)/actual cash flows|

4.8 Hypothesis Testing

An hypothesis is rejected or accepted depending upon the statistical results from the data analysis (Cooper & Emory 1995). In order to test hypotheses, an appropriate statistical test must be chosen. The research hypotheses constructed to answer the research questions and the statistical tests suitable for testing the hypotheses are discussed below.

Research Hypothesis 1: Past earnings have significant predictive power in predicting future cash flows of Thai listed companies.

Hypothesis 1 is concerned with the first research question.

The statistical results of the

earnings model (Model 1) were used to test the hypothesis. The null hypothesis states that there is no relationship between year-lags of earnings and future cash flows, while the alternate hypothesis expresses a significantly positive relationship between year-lags of earnings and future cash flows. The results of statistical tests would indicate whether or not support has been found for the alternate hypothesis. The significant level was established at be accepted if the F-statistic shows the significant level equal or less than 0.05 (p 0.05). If the alternate hypothesis is accepted, Research hypothesis 1 is substantiated. the 0.05 level, that is, the null hypothesis would be rejected and the alternate hypothesis would

Research Hypothesis 2: Past cash flows have significant predictive power in predicting future cash flows of Thai listed companies.

The hypothesis 2 is concerned with the second research question. The null hypothesis states that there is no relationship between year-lags of cash flows and future cash flows while, the

103

alternate hypothesis expresses a significantly positive relationship between year-lags of cash flows and future cash flows. Similarly to the earnings model, the significant level was established at the 0.05 level. The null hypothesis would be rejected and the alternate hypothesis would be accepted if the F test of the cash flows model (Model 2) show that the the significant level equal or less than 0.05 (p 0.05). If the alternate hypothesis is accepted, Research hypothesis 2 is substantiated. variance in future cash flows has been significantly explained by the year-lags of cash flows at

Research Hypothesis 3: Past cash flows and accrual components of earnings have significant predictive power in predicting future cash flows of Thai listed companies.

Hypothesis 3 concerned with the third research question. The null hypothesis states that there is no relationship between cash flows and accrual components of earnings and future cash flows, while the alternate hypothesis expresses a significant relationship between year-lags of cash flows and accrual components of earnings and future cash flows. The null hypothesis would be rejected and the alternate hypothesis would be accepted if the F-statistic of cash or less than 0.05 (p 0.05). Research hypothesis three is substantiated when the alternate hypothesis is accepted. flows and accrual components of earnings model (Model 3) shows the significant level equal

Research Hypothesis 4: There are different predictive powers between three prediction models: earnings, cash flows and cash flows and accrual components of earnings models. Hypothesis 4 intends to answer Research question 4. Adjusted R2 values, correlations between actual and predicted cash flows, and mean absolute percentage errors of each model were used to evaluate the predictive ability of each model. The comparisons of these values were used to test the hypotheses.

Research Hypothesis 5: Past cash flow ratios are significant predictors of future cash flows of Thai listed companies.

104

This hypothesis involves testing the predictive power of cash flows ratios in predicting future cash flows. The null hypothesis states that the cash flow ratios do not provide a good predictor for predicting future cash flows of Thai listed companies. In other words, the alternate hypothesis states that cash flow ratios provide a good predictor for predicting future cash flows of Thai listed companies. The results of the stepwise analysis were used to test the hypothesis.

4.9 Data Specification

4.9.1 Testing periods The research sample was comprised of companies listed on the Stock Exchange of Thailand from 1994 to 2002. These periods were selected because Thai listed firms started presenting cash flow statements in 1994 under the requirement of the Stock Exchange of Thailand. In addition, the data collection process was conducted in the end of year 2003. The information of annual financial statements available at that time was the annual report of the year ended 2002. Consequently, the year 2002 was chosen as the final year for the testing data of this study. These data also cover the economic crisis period of 1997 and 1998 to compare the result of analysis between before and after crisis.

According to the models mentioned in Section 4.5, annual data for the values of the dependent variables (future cash flows) in year t were collected from the end of year 1996 to the end of year 2002 and annual data for year-lagged values of independent variables in year t-i were gathered from the end of year 1994 to 2001. A data matching between future cash flows of year t and independent variables of year-lags one, two and three are shown in Table 4.3

105

Table 4.3 Annual Data-matching in Different Periods between the Predictors and Future Cash Flows Year of prediction One-year lag (t-1) 1996 1997 1998 1999 2000 2001 2002 Source: developed for this thesis 1995 1996 1997 1998 1999 2000 2001

Annual data used in model

Two-year lags (t-1+t-2) 1994-1995 1995-1996 1996-1997 1997-1998 1998-1999 1999-2000 2000-2001 Three-year lags(t-1+t-2+t-3) 1994-1996 1995-1997 1996-1998 1997-1999 1998-2000 1999-2001

4.9.2 Sampling selection

The sample of this study included all non-financial firms listed on the Stock Exchange of Thailand. Financial sector firms including commercial banks, financial companies, finance and securities companies, securities companies and insurance companies were excluded from the sample, because their financial statements contain some different items from non-financial firms. Additionally, companies under rehabilitation were excluded because they were subject to being delisted due to their financial and operational problems and their financial statements reported stockholders equity values less than zero. Initially, there were between 381 and 454 companies of 31 industry groups from 1994 to 2002 (as reported by SET). The samples were selected according to the following criteria: The companies must have completed data for all variables such as cash flow statements, income statements and balance sheets. The companies must have been operating during the fiscal year ended December 31. The reason for excluding non-December fiscal year ended firms is to ensure that data for all firms can be compared.

After screening data based on the sample selection criteria, the number of listed firms included in this study was between 249 and 291. The number of companies in each year from 1994 to 2002 is shown in Table 4.4.

106

Table 4.4 Number of Companies in Each Year

1994 389 65 324 1995 416 68 348 1996 454 77 377 Number of companies 1997 1998 1999 2000 431 418 392 381 77 354 66 352 58 334 51 330 2001 382 48 334 2002 389 47 342

Number of companies listed on SET Companies under rehabilitation Number of companies in regular sector Number of companies in financial sector: Commercial banks Finance and securities Insurance Number of non-financial companies

15 40 20 249

15 44 21 268

16 51 22 288

15 26 22 291

14 26 22 290

15 20 22 277

15 20 22 273

15 22 22 275

14 27 22 279

Source: Stock Exchange of Thailand

4.9.3 Sources of data

Data were collected from financial statements of companies listed on the Stock Exchange of Thailand (SET) obtained on the SETs Listed Company Info CD-ROM for annual reports 1994 to 2002. To focus on the statements of cash flows, cash flows from operating activities were selected directly from the cash flow statements. Earnings were derived from income statements. Accrual accounting data came from cash flow statements. Total assets, sales and other interested variables to compute cash flow ratios as described in Subsection 4.4.2 were selected from balance sheets and income statements.

4.10 The Validity of the Research

External validity refers to the extent to which the result of a study may be generalised to other samples. Accounting data used in this study are historical which may have caused a low level of external validity. That is, the result of a particular study at a point in time may not be generalised to other periods of time. In the analysis, this study attempted to eliminate this disadvantage by using data for many years which cover a period during the Asian economic crisis. Data of each year were analysed separately to observe whether the economic crisis caused a different result. In addition, the data sample was partitioned into subsamples for cross validation (Myers 1990).

107

4.11 Conclusion

This chapter provided the methodology used in this study including the research design, variable definition and measurement, model building, data specification, and the validity of the research. The nature of research paradigms was discussed to justify the chosen paradigm. The positivism paradigm is suitable for this research as the research deals with testing hypotheses and chiefly focuses on quantitative methods. The basic method used to collect data was the use of secondary data, which suited the data needs in this research. That is, accounting data reported on financial statements of firms were selected to measure variables. Additionally, the measurement of variables was defined to test the four prediction models of this research. The four prediction models consisted of earnings, cash flows, cash flows and accrual components of earnings and cash flow ratios models. These models were tested by regression analysis. The study was designed to use an adjusted R2 reflecting the explanatory power of the model in comparing regression models. Furthermore, the study planned to evaluate the predictive power of the models by using a mean of absolute percentage errors generated from the sample data. Finally, the validity issue of the research was discussed in support of the research design, which reduces the possible low level of external validity of the research. In the next chapter, the data analysis and results of the analysis are provided.

108

Chapter 5
Data analysis
5.1 Introduction

The previous chapter described the research methodology employed for this research. This chapter aims to report the empirical results of the data analysis and testing of the hypotheses. The chapter consists of 10 sections. Section 5.1 introduces the organisation of the chapter. Section 5.2 summarises the regression equation and model of this study. Section 5.3 discusses the data preparation before the analysis, including data arrangement, missing data, outliers and multicolinearity. Section 5.4 reports the results of preliminary analysis including descriptive statistics and correlations. Section 5.5 presents the results of regression analysis for each model: earnings, cash flows, cash flow and components of earnings, and cash flow ratios models. Section 5.6 reports the comparison of the explanatory power of the three models including earnings, cash flows and cash flows and accrual components of earnings models. Section 5.7 discusses the predictive ability of each model tested in the out-of-sample period and measured by two approaches, the correlation between actual and predicted cash flows and mean absolute percentage errors. Section 5.8 discusses hypothesis testing. The statistical

results of the regression models were used to determine whether hypotheses were accepted or rejected. Section 5.9 reports the results of the testing of the cash flows and components of earnings model by disaggregating accrual components of earnings into major components: change in accounts receivable, change in accounts payable, change in inventory, depreciation and amortisation, and other accrual components and the additional regression analysis in pooled-year data, excluding data of prediction year 1998, to eliminate the impact of the economic crisis. Finally, Section 5.10 concludes the chapter.

109

Figure 5.1 Structure of Chapter 5: Data Analysis

5.1 Introduction

5.2 Regression Equations and Models

5.3 Data Preparation

5.4 Preliminary Analysis

5.5 Regression Analysis Results

5.6 Comparison of Explanatory Ability among Models

5.7 Evaluation of the Predictive Ability of Models with Different Approaches

5.8 Test of Hypotheses

5.9 Additional Analysis

5.10 Conclusion

110

5.2 Regression Equations and Models

This section discusses regression equations and the prediction model, to provide ideas for data arrangement. This study developed three main cash flow prediction models: earnings, cash flow, and components of earnings model. Year-lags of earnings (EARN), cash flows (CFO) and the components of earnings, cash flow and accrual (CFO & ACR), are the independent variables for the first, second and third models respectively. The year-lags of the independent variables of each prediction model were tested in three ways. Firstly, the model using a oneyear lag (t-1) of independent variables. Secondly, model using two-year lags included lags one (t-1) and two (t-2) of independent variables. Thirdly, the model using three-year lags included lags one (t-1), two (t-2) and three (t-3) of independent variables. The analysis of each model was conducted by regression procedures using either simple or multiple regression. The regression equations and models are summarised in Table 5.1.
Table 5.1 Summary of Regression Equations for All Models in Data Analysis
No. Model Regression equation Type of regression 1 1.1 1.2 1.3 2 2.1 2.2 2.3 3 Earnings model : One-year lag Two-year lags Three-year lags Cash flows model: One-year lag Two-year lags Three-year lags Components of earnings model: 3.1 3.2 3.3 One-year lag Two-year lags Three-year lags CFOn,t= + 1CFO n,t-1 + 2 ACRi , t-1 + Multiple Multiple Multiple CFOn,t = + 1CFOn t 1+ 2CFO n t 2 + 3ACR i ,t-1+ 4ACR i ,t-2 + 4ACRn , t-1 + 5 ACRn , t-2 + 6ACR n , t-3 + CFO n,t = + 1CFOn, t- 1 + 2 CFOn, t- 2 + CFO n,t= + 1CFOn,t-1 + Simple Multiple Multiple CFO n,t= + 1 EARNn, t- 1 + 2 EARNi, t- 2 + CFO n,t = + 1EARN n,t-1 + Simple Multiple Multiple

CFO n,t= + 1 EARNn, t- 1 + 2EARNn,t-2 + 3 EARN n, t- 3 +

CFO n,t = + 1CFO n, t- 1 + 2 CFOn, t- 2 + 3 CFOn, t- 3 +

CFOn,t = + 1CFOn, t- 1 + 2 CFO n, t- 2 + 3 CFOn, t- 3 +

Source: summarised from section 4.5 in Chapter 4 of this study

111

5.3 Data Preparation

This study tested the predictive ability of accounting information for future cash flow by using a regression procedure. The accounting information collected was arranged and examined before use in the analysis. Data were analysed in two ways including yearly and pooled-year analysis. Additionally, before processing the regression, data were investigated to detect whether presence of missing data, outliers and multicollinearity existed. These unusual data can unduly influence the results of the analysis, and their presence may be a signal that the regression model has failed to capture important characteristics of the data. In this section, data arrangement, missing data, outliers and multicollinearity issues are discussed.

5.3.1 Data arrangement Data from year 1994 to 2002 were matched between dependent variable year t (or prediction year) and independent variables, year-lags one (t-1), two (t-2), and three (t-3) as shown in table 5.2. The model with one-year lag of the independent variables required two years of information for each prediction year including a prediction year (t) and a lagged explanatory year (t-1). The model with two-year lags of the independent variables required three years of information for each prediction year including a prediction year (t) and two lagged explanatory years; t-1 and t-2. The model with three-year lags of the independent variables required four years of information including a prediction year (t) and three lagged explanatory years; t-1, t-2, and t-3. Six prediction years (1996 to 2002) were arranged in the sample period. Due to the limitation of data, only lags one and two of independent variables were matched for prediction year 1996. Two data files were used in the analysis. The first file was for yearly analysis and the second file was for pooled-year analysis. In the yearly analysis, a regression equation was calculated separately for each set of data in each prediction year. So the data set for each prediction year from 1996 to 2002 was arranged separately and each prediction year (year t) contained three sets of data for one-, two- and three-year lag analysis. In the pooled-year analysis, the data of all prediction years from 1996 to 2001 were pooled together and analysed so that an organisation-year represented a case. Consequently, a set of

112

data comprised of data for all prediction years. Totally, three sets of data, for one-, two- and three-year lags were analysed for pooled year analysis.

Table 5.2 Matching Years in the Analysis

Year t 1996 1997 1998 1999 2000 2001 2002 Year t-1 1995 1996 1997 1998 1999 2000 2001 Year t-2 1994 1995 1996 1997 1998 1999 2000 Year t-3 1994 1995 1996 1997 1998 1999

Source: developed for this study

5.3.2 Missing data

All data used in this study were the items on financial statements of listed companies in nonfinancial sectors on the Stock Exchange of Thailand. At first, all items were scanned for missing data, then, if missing data was found in any item of the observation, the observation was eliminated from the sample. Most missing data items were found on the cash flow statements. For instance, for year 1994, some companies did not disclose this statement to SET and some did not report some items in the statements because the listed companies started to report cash flow statements in 1994. Therefore, for this year the number of observations included in the sample was only 182 from a total of 249. A summary of the number of observations with missing data and final number of observations for each year from 1994 to 2002 is presented in Table 5.3. The sample of this study included only non-financial firms listed on the Stock Exchange of Thailand during the year 1994 to 2002. Initially, there were 2,490 firm-year observations. After deleting firms with missing data, there was a total of 1,970 firm-year observations. In addition, the observations were excluded in the sample if they did not have data available for a pair of matching years (as shown in Table 5.4). This is because independent variables of year lags were matched with the dependent variable of the next year. Data for dependent variables was from the year 1996 to 2002. Data for independent variables was collected from year 1994 to 2001.

113

After matching years, the number of cases in data analysis decreased. Additionally, the number of cases decreased when the number of year lags increased. The smallest number of cases occurred for the prediction year 1997 with three-year lag analysis and the largest number of cases occurred for the prediction year 1998 with one-year lag analysis. The number of cases with all data available, number of outliers and number of final cases included in the analysis for each prediction year with the one-, two- and three-year lags are shown in table 5.4.

Table 5.3 Number of Companies in the Sample for Each Year

1994 249 67 182 1995 268 83 185 1996 288 61 227 Number of companies 1997 1998 1999 2000 291 290 277 273 50 241 49 241 35 242 39 234 2001 275 41 234 2002 279 95 184 Total 2490 520 1970

Number of non-financial companies No. of companies with missing data No. of companies with all data available

Source: data analysis for this study

Table 5.4 Number of Cases in Data Analysis for Each Prediction Year Prediction year No. of case with all data available 1996 1997 1998 1999 2000 2001 2002 177 221 241 213 226 234 182 4 5 1 7 3 5 3 One lag No. of outliers Final no. of case 173 216 240 206 223 229 179 No. of case 174 173 222 213 199 226 182 6 4 1 8 3 6 4 Two lags No. of Final no. case 168 169 221 205 196 220 178 170 173 195 199 199 181 4 1 7 2 3 4 166 172 188 195 196 177 of No. of case Three lags No. of Final no. case of

outliers

outliers

Source: Analysis secondary data collected for this research

114

5.3.3 Outliers

Outliers are cases with such extreme values on one variable or a combination of variables that they distort the statistics. It was necessary to identify outliers and remove them. Otherwise the multiple regression equations would reflect the unusual cases, rather than the usual one. That is, the values of the regression coefficient can be influenced by outlying cases (Tabachnick, Fidell & Osterlind 2001). The outliers can be classified by residuals, which can be calculated basically as the actual minus predicted values of the independent variable (Hair et al. 1998). In this study, following Barth, Cram and Nelson (2001; 2002), before conducting the regression procedures outliers were identified when standardised residuals of cases were greater than +3.0. A standardized residual of + 3 means that the point is 3 standard deviation from the regression line. Then the regression was run omitting the outlying cases. The numbers of cases in data analysis used for each year are presented in Table 5.4.

5.3.4 Multicollinearity

Multicollinearity deals with the assumption of impartiality of the independent variables (Tabachnick, Fidell & Osterlind 2001). It has an impact on explanation and estimation of the regression model (Hair et al. 1998). Basically, multicollinearity can be identified by measuring a correlation of all independent variables in each multivariate model. If any correlation coefficients between a pair of the variables was higher than 0.90, a multicollinearity problem existed (Hair et al. 1998). For this research, the Pearson correlation between independent variables in Section 5.3.2. was used to detect the multicollinearity problem.

115

5.4 Preliminary Analysis

5.4.1 Descriptive statistics

All variables of this research derived from financial statements of Thai listed companies were numerical. In the first step of the data analysis, the corrected data were summarised and categorised to make it easier to understand. Additionally, in order to consider the characteristics of the variables, some statistics including maximum, minimum, mean and standard deviation were produced by the SPSS program. The descriptive statistics of the variables over the period of study, 1994 to 2002, are shown in Table 5.5.

Table 5.5 shows that cash flow from operations (CFO) ranges from -4,825 million baht to 94,791 million baht. Overall, the average mean of CFO is approximately 485 million baht. CFO is the lowest in the year 1997 followed by 1999 and 2000 (minimum = -4,825, -4,431 and -3,573 million baht). These results reflect the impact of the Asian economic crisis during 1997 and 1998, in that cash flows of companies dramatically decreased in 1997. However CFO was higher again in 2002, with mean 832.71 million baht. (minimum = 0.17 and maximum = 22,037.95). This reflects a sign of Thailands economic improvement in 2002 after suffering from the economic crisis (as discussed in chapter 2). The range of earnings (EARN) is approximately from -52,551 million baht to 22,209 million baht, with the overall mean about 104 million baht. EARN is lowest in 1997. Also, mean of EARN is negative in 1997 and (mean = -792.85 million baht). In the same way, this shows the impact of the Asian economic crisis on Thai listed companies. That is, most of the Thai listed companies incurred losses in 1997. The range of accrual components of earnings (ACR) is 94,443 million baht to 15,254 million baht. The overall mean of accrual components of earnings is roughly -337 million baht. However, ACR is extremely low in 1997 and 1998 (minimum = -47,725.95 and 94,443.04 million baht).

In general, these results are consistent with prior research in the USA (Barth, Cram & Nelson 2001; 2002; Dechow, Kothari & Watts 1998; Murdoch & Krause 1989; Sloan 1996), in that the means of cash flows from operations and earnings are positive. The mean of earnings is

116

smaller than that of cash flows, which reflects the fact that earnings are reduced by non-cash expenses such as depreciation and amortisation (Dechow, Kothari & Watta 1998), whereas cash flows from operations based on Thai accounting standard No.25 are calculated by adding those expenses back onto earnings. Meanwhile, the mean of the aggregate accrual component of earnings is negative. This result is caused by the calculation of the accrual component of earnings, that is, the accrual component of earnings is the difference between earnings and cash flows (Dechow, Kothari & Watta 1998).

In conclusion, these results suggest that the characteristics of the accounting information of Thai listed companies used in this study were in line with the range of data from other countries. Additionally, these statistically descriptive results imply that the Asian economic crisis during 1997 and 1998 had an impact on accounting information of Thai listed companies. That is, when the economic crisis occurred, cash flows from operations and earnings decreased sharply. Additionally, cash flow and earnings can be signs of economic performance, as they increase when the economy grows.

117

Table 5.5 Descriptive Statistics of Cash Flow from Operations, Earnings and Accrual Component of Earnings.
Variables1 CFO Year 1994 1995 1996 19972 19982 1999 2000 2001 2002 Average EARN 1994 1995 1996 19972 19982 1999 2000 2001 Average 182 185 227 241 241 242 234 234 N 182 185 227 241 241 242 234 234 184 Minimum -3,187.04 -3,506.04 -2,505.40 -4,825.17 -2,294.24 -4,431.86 -3,573.51 -1,381.65 0.17 -4,825.17 -718.64 -476.60 -1,924.12 -52,551.13 -4,484.94 -7,506.59 -5,717.31 -3,425.13 -52,551.13 Maximum 3,431.90 6,026.92 19,803.15 9,502.11 94,791.55 6,741.52 14,476.96 20,588.48 22,037.95 94,791.55 2,114.28 6,093.38 6,787.86 6,531.86 22,209.35 10,104.73 6,598.95 10,894.80 22,209.35 Mean 176.92 196.63 325.38 405.85 929.23 375.22 471.35 655.45 832.71 485.41 200.11 295.11 238.88 -792.85 490.43 -21.19 116.51 309.49 104.56 Std. Dev. 631.77 873.71 1,568.14 1,324.98 6,322.55 909.35 1,567.89 1,877.93 2,217.63 1,921.55 349.74 711.78 716.55 4,199.25 2,235.55 1,136.51 1,106.11 1,186.07 1,455.19

118

Table 5.5 (continued)

Variables ACR

Year 1994 1995 1996 19972 19982 1999 2000 2001

N 182 185 227 241 241 242 234 234

Minimum -2,656.41 -3,032.58 -13,015.29 -47,725.95 -94,443.04 -8,860.03 -9,043.83 -9,693.68 -94,443.04

Maximum 5,301.32 5,597.87 6,547.38 5,102.07 15,254.04 8,972.49 5,505.18 7,619.92 15,254.04

Mean 23.19 98.48 -86.50 -1,199.00 -438.80 -396.41 -354.84 -345.96 -337.44

Std. Dev. 623.93 956.79 1,199.66 4,317.43 6,315.42 1,149.10 1,398.52 1,450.24 2,176.39

Average

Source: Data analysis for this study Note 1. The variables are defined as below. Cash Flow from operation (CFO) is cash flow from operations at the end of an accounting period in million baht. Earnings (EARN) is net income before extraordinary items and discontinued operation reported on income statements in million baht. Accrual component of earning (ACR) is total operating accruals in million baht, calculated as EARN minus CFO.

2. The year during Asian economic crisis.

5.4.2 Correlation analysis

Investigating a possible association between each dependent and independent variable is essential before entering the variables into regression (Collis & Hussey 2003). If there is no correlation among each dependent and independent variable, the regression should not be conducted. However, the high correlation among independent variables indicates the presence of multicollinearity. A measure of the correlation is represented by correlation coefficients. The coefficient provides both the direction and strength of the relationship between a pair of

119

variables. In this research, the strength of association between all pairs of variables was statistically measured by Pearsons correlation coefficient. The correlations between pairs of variables in the sample are presented in Tables 5.6 to 5.10.
Table 5.6 Pearson Correlation Coefficients of Cash Flow from Operations for Each Year During 19942002 Variable CFO94 CFO95 CFO96 CFO97 CFO98 CFO99 CFO00 CFO01 CFO02
* **

CFO94 1 .687** .397 .487 .017 .283 .310 .359 .491

** ** ** ** ** **

CFO95

CFO96

CFO97

CFO98

CFO99

CFO00

CFO01

CFO02

1 .420** .468 .044 .394 .425 .392 .587

** ** ** ** **

1 .047 .218** .270** .259 .248 .272

** ** **

1 .022 .426 .501 .526 .548

** ** ** **

1 .083 .040 .031 .056 1 .734** .723 .728

** **

1 .861** .887
**

1 .860** 1

Correlation is significant at the 0.05 level (2-tailed). Correlation is significant at the 0.01 level (2-tailed).

Table 5.6 shows the correlation between cash flow variables for different years. It can be seen that overall, there is a statistically positive and significant correlation between cash flow from operations (CFO) for each year at 0.01 level. However there is no statistically significant relationship between CFO 1998 and other years except CFO 1996, and there is no significant correlation between CFO 1996 and CFO 1997. This may explain the low explanatory power of the cash flow model in those years. On average, the correlations between CFO are lower than 0.9 so this indicates no multicollinearity problem in the cash flow model (model 2).

Table 5.7 shows the correlation between earnings (EARN) for each year. It can be seen that there are highly significant correlations between earnings and lags of earnings at the 0.01 level. However, there are negative and significant correlations in some pairs of each years earnings, for instance EARN 1998 and EARN 1997, EARN 1997 and EARN 1996, and EARN 1997 and 1995. Overall, the correlation between EARN is lower than 0.9. Therefore there is no multicollinearity between independent variables in earnings model (Model 1).

120

Table 5.7 Pearson Correlation Coefficients Among Earnings for Each Year During 1994- 2001 Variable EARN94 EARN95 EARN96 EARN97 EARN98 EARN99 EARN00 EARN01 EARN94 1 .526** .577** -.179* .369** .069 .095 .286 ** EARN95 1 .591** -.412** .587** -.061 .208** .228** 1 -.486** .588** .379** .312** .619 ** 1 -.737** .221** .141* -.305** 1 -.110 -.044 .447** 1 .477** .450** 1 .444** 1 EARN96 EARN97 EARN98 EARN99 EARN00 EARN01

* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).

Table 5.8 Pearson Correlation Coefficients Among Accrual Component of Earnings for Each Year During 1994- 2001 Variable ACR94 ACR95 ACR96 ACR97 ACR98 ACR99 ACR00 ACR01
* **

ACR94 1 .537** .205** .082 -.031 -.082 -.013 .204*

ACR95

ACR96

ACR97

ACR98

ACR99

ACR00

ACR01

1 .261** -.022 .019 .001 .057 .161*

1
.582** .072 -.003 .166* -.100 1 -.135* .297 .364 .051
** **

1 -.108 -.077 -.070 1 .527** .411** 1 .499** 1

Correlation is significant at the 0.05 level (2-tailed). Correlation is significant at the 0.01 level (2-tailed).

Table 5.8 shows the correlation between accrual components of earnings (ACR) for each year. There are some positive and significant correlations between the ACR of each year at the 0.01 level. However, ACR 1997 is negatively and significantly correlated with ACR 1998 (p>0.05). Nevertheless, ACR 1998 does not correlate with ACR of any year. The correlation between ACR itself is lower than 0.9. This indicates that there is no multicollinearity between independent variables in cash flow and the accrual components of the earnings model (Model 3).

121

Table 5.9 Pearson Correlation Coefficients Between Variables for Each Year During 1994-2002 Variable EARN94 EARN95 EARN96 EARN97 EARN98 EARN99 EARN00 EARN01 ACR94 ACR95 ACR96 ACR97 ACR98 ACR99 ACR00 ACR01
*

CFO94 .299** .213** .343** -.086 .221** .136 .244** .172* -.845** -.466** -.129 -.272** .046 -.142 -.184* -.374**

CFO95 .305** .285** .482** -.186* .336** .053 .318** .206* -.525** -.701** -.054 -.345** .054 -.323
**

CFO96 .289** .139 .682** -.674** .435** .255** .188** .543** -.240** -.281** -.900** -.672** -.066 .034 -.165* .140

CFO97 .473 .270** .073 .067 .128* -.018 .055 .131 -.226** -.224** -.018 -.241** .023 -.352
**

CFO98 .079 .080 .230** -.129* .180** .084 .048 .144* .027 .020 -.148* -.133* -.937** .017 -.010 .084

CFO99 .498** .447** .373** .033 .279** .386** .469** .589** -.009 -.086 -.152* -.102 -.007 -.409
**

CFO00 .359** .401** .291** -.075 .241** .171* .498** .547** -.114 -.160* -.187** -.231** .026 -.381
**

CFO01 .384** .375** .283** -.097 .254** .227** .524** .635** -.149 -.144 -.176* -.260** .038 -.318
**

CFO02 .390** .462** .336** -.051 .253** .200** .565** .581** -.287** -.280** -.180* -.221** .013 -.377** -.649** -.657**

-.265** -.393**

-.548** -.584**

-.456** -.451**

-.728** -.668**

-.551** -.775**

Source: data analysis for this thesis. Correlation is significant at the 0.05 level (2-tailed). Correlation is significant at the 0.01 level (2-tailed).
**

Table 5.9 reveals the correlation between cash flows from operations (CFO) and earnings (EARN). On average, there are considerable significant correlations between CFO and EARN for each year at the 0.01 and 0.05 levels. However, when considering the correlation between cash flow and each year-lag of earning (down to three lags), CFO 1996 has no significant correlation with EARN 1995. CFO 1997 has no significant correlation with EARN 1996 and 1994, and CFO 1999 has no significant correlation with EARN 1997. These may be the reason why the earnings model is not significant in explaining future cash flows for those years.

Regarding the correlation between cash flow from operations (CFO) and accrual components of earnings (ACR), on average, there are significantly negative correlations between CFO and ACR for each year at the 0.01 and 0.05 levels. Nevertheless, correlations between some pairs

122

of cash flow and accrual components have not been found, such as for CFO 1997 and ACR 1996, CFO 1998 and ACR 1997 and ACR 1996. This may cause the model to be insignificant in those years. Overall, the correlation between CFO and ACR is lower than 0.9 except for CFO and ACR in year 1996. This indicates that apart from year 1996, there is no evidence of a multicollinearity problem in cash flow and accrual components of the earnings model (Model 3).

Overall, the variables for year 1998 are not correlated with the variables for other years, therefore the results of the regression analysis comprised of variables of that year may not be significant. Moreover, the correlation coefficient of a pair of variables for year 1997 usually shows negative correlations. This may imply that the Asian economic crisis during 1997 and 1998 had an impact on the relationship between the variables of this study.

Table 5.10 presents the Pearson correlation coefficients of all pairs of variables for pooledyear data in which data of all prediction years from 1996 to 2002 was included and pooled together in a data file. It can be seen that there is a statistically significant correlation between cash flows year t (CFOt) and one-year lag of cash flow (CFOt-1), two-year lags of cash flows (CFOt-2), and that three-year lags of cash flows (CFOt-3). CFOt is significantly positively correlated with the two-year lags of earnings (EARNt-2) at the 0.01 level. However, it is negatively correlated with two- and three- year lags of accrual components of earnings (ACRt2

and ACRt-3) at the 0.05 and 0.01 level respectively. The correlation between variables is

lower than 0.90 so there is no evidence of multicollinearity for pooled year data. When compared with each-year data, on average, the correlation of a pair of variables for pooledyear data is lower than that for each-year data. Therefore, the explanatory power of each model for pooled-year analysis could be lower than that of model for yearly analysis.

123

Table 5.10 Pearson Correlation Coefficients of a Pair of Variables for Pooled Year During 1994- 2002
Variable CFOT-3 CFOT-2 CFOT-1 CFOT EARNT-3 EARNT-2 EARNT-1 ACRT-3 ACRT-2 ACRT-1 CFOt-3 1 .049 .099 .073 .127
** * ** **

CFOt-2

CFOt-1

CFOt

EARNt-3

EARNt-2

EARNt-1

ACR t-3

ACRt-2

ACR t-1

1 .090** .143
** **

1 .139 ** .079
** *

1 .047 .106
**

-.091 .144

1 -.325 ** .102 .532

* ** ** **

-.077 .080

**

-.057 .173

1 -.295 ** -.136 .521

** ** **

** ** **

-.047 -.098 -.769 -.109

** ** **

**

-.017 -.035

1 -.008 -.144 .500

** **

-.773 -.091 -.051

-.038 -.114 -767

** **

1 -.004 .032 1 .005 1

-.060 -134

-.133 -.021

**

-.143

* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).

In addition to the results for each year, the pooled-year data also provide evidence that overall, there is a significantly positive correlation between cash flow and lags of cash flow and earnings. In contrast, there is a significantly negative correlation between cash flows and the accrual component of earnings. These conclusions are consistent with prior research such as Barth, Beaver, Hand and Landsman (1999), Barth, Cram and Nelson (2001), Dechow, Kothari and Watts (1998), and Sloan (1996). Additionally, it can be concluded that the correlation of future cash flow from operations with earnings is lower than the correlation of future cash flows with year-lag cash flows. A low correlation between cash flows and earnings could cause the earnings model to have a lower explanatory power than the cash flows model. In conclusion, these results suggest that these interesting variables could be included in the regression model to predict future cash flows.

124

5.5 Regression Results

This section presents the results of four regression models. The models were constructed to investigate the ability of independent variables to predict future cash flow of Thai listed firms (as summarised in Table 5.1). Cash flows from operations were the dependent variable of each model. The predictive power of earnings, cash flows from operation, and components of earnings were investigated in models one, two and three respectively. Each model was tested first for yearly and then pooled-year analysis (see Section 5.2). Accordingly, the results of the statistical analysis of each model are presented in two subsections. The first subsection shows the results of data analysis for each prediction year 1996 to 2002 and the second subsection shows the results of data analysis for pooled prediction year 1996 to 2001.

5.5.1 Earnings model

Simple and multiple regression models were performed between cash flow from operations year t, with CFOt as the dependent variable and year lags of earnings, EARNt-i as the independent variable. EARNt-i, were expected to provide a powerful model for predicting future cash flows, CFOt. In order to test the incremental predictive power of year lags, yearlags of earnings were added systematically to the regression equations. The first equation contained only lag one, the second equation contained both lags one and two and the third equation contained lags one, two and three. The earnings model, with two- and three-year lags, was expected to have more predictive power than the one-year lag. Both sets of data were analysed for each prediction year spanning 1996 to 2002 and for pooled-year period 1996 to 2001.

Yearly analysis

For the yearly analysis, the regression model was processed to analyse each set of prediction years separately. Using this method the statistical results are presented for each prediction year. The analysis was performed for six prediction years spanning 1996 to 2002. In each prediction year, three sets of data were analysed including a set of the one-, two- and three-

125

year lags of earnings. The regression was run twenty times for the six prediction years 1996 to 2002. The statistical results are presented for each prediction year with one-, two- and threeyear lags of earnings, in Table 5.11.

Using one-year lag for prediction (Table 5.11), earnings in year t-1 (EARNt-1) were found to be a significant predictor (p < 0.005) of cash flow at year t (CFOt), in all years except for the prediction year 2000 (p > 0.5). Excluding the prediction year 2000, EARNt-1 was found to explain between 4.2% and 63.9% of the variation in CFOt, with an overall average of 27% (adjusted R2 = 0.042, 0.636, and 0.27, respectively). The explanatory power of the model was very low in prediction year 1996 and year 1997 (R2 = 0.04 and 0.06). In contrast, the explanatory power of the model was highest in prediction year 2002 in which the Thai economy grew after suffering from the economic crisis (adjusted R2 = 0.64). These results imply that the explanatory power of earnings is unstable. Moreover, EARN 1999 was insignificant for prediction year 2000 which was the year in which the Thai accounting standards were revised and issued (see Chapter 2). The majority of these results showed positive signs of coefficients (1), as expected. Nevertheless, the analysis for prediction year 1998 provided a negative coefficient. This implies that lag one of earnings had a negative sign to predict cash flows of year 1998 in which the economic crisis occurred.

In using two-year lags for prediction, significant relationships were found between the pair of independent variables (EARNt-1 and EARN t-2) and the dependent variable (CFOt) in all years (p < 0.005). The two independent variables were found to explain between 11% and 81% of the variation in CFOt, with an overall mean of 39% (adjusted R2 = 0.11, 0.81 and 0.39 respectively). The explanatory power of the model was lowest in prediction year 1997 followed by year 1996 (R2 = 0.11 and 0.19 respectively). The explanatory power of the model was highest in 1998 followed by 2002 (adjusted R2 =0.81 and 0.70 respectively). This result is consistent with the result of one-year lag analysis in that the explanatory power of the earnings model was low in the prediction years 1996 and 1997 and high in 2002. When the relative contributions to predictions of the two independent variables were evaluated, it was found that EARNt-1 did not contribute significantly to prediction of CFO 1996 (t = 1.16, p > .05) and CFO 1997 (t = .10, p > .05), and EARNt-2, EARN 1998 and 1999 did not contribute

126

significantly to prediction of CFO 2000 (t = -0.38, p > .05) and CFO 2001 (t = -1.28, p > .05) respectively. Corresponding with one-year lag analysis, earnings of the year during the economic crisis did not have explanatory power to predict future cash flows. Moreover, EARNt-1 (EARN 1997) had a negative coefficient for prediction of CFO 1998, the year of the economic crisis. This may imply that the economic crisis influenced the explanatory power of earnings to predict future cash flows.

In using three-year lags for prediction, significant relationships were found between the independent variables (EARNt-1, EARNt-2, and EARNt-3) and the dependent variable (CFOt) in all years (p < 0.005). The three independent variables were found to explain between 0.7 % and 72 % of the variation in CFOt, with an overall mean of 42% (adjusted R2 = 0.07, 0.72 and 0.42 respectively). The explanatory power was the lowest in the prediction year 2000 in which lags of earnings were from the years 1997 and 1998 which were during the economy crisis, and 1999, in which Thai accounting standards were revised and issued. Additionally, similar to the results of the one- and two-year lags, the explanatory power was highest in the year 2002. When the relative contributions to prediction of the three independent variables were evaluated, it was found that EARNt-1, EARN 1996 and EARN 1999 did not contribute significantly to prediction of CFO 1997 (t = 0.09, p > .05) and CFO 2000 (t = 1.74, p > .05), respectively. EARNt-2, EARN 1995 and EARN 1999, did not contribute significantly to prediction of CFO 1997 (t = .89, p > .05) and CFO 2001 (t = -1.25, p > .05) respectively, and EARNt-3, EARN 1995 and EARN 1997 did not contribute significantly to prediction of CFO 1998 (t = -0.68, p > .05) and CFO 2000 (t = -0.23, p > .05) respectively. These results correspond with the one- and two-year lag analyses; earnings lost their explanatory power to predict future cash flow when they were either from the years during the economic crisis or when the prediction year was during the period of the economic crisis.

In conclusion, the earnings model could explain future cash flows however the explanatory power of the model was unstable. On average, comparing each prediction year, the model had a high explanatory power in the year 2002 while it had a low explanatory power in the years 1996, 1997 and 2000. There may be many reasons for this as the period of this study encompassed many situations in Thailand. Firstly, the accounting standard for reporting cash

127

flow statements has been effective since 1994. The cash flow information of that year prepared under the new standard had a weak correlation with earnings suggested by the results of correlation analysis (in Section 5.3). Secondly, the Asian economic crisis occurred in the

period 1997 to 1998. It can be seen that the earnings model lost explanatory power when either earnings or predicted cash flow were in the years of the economic crisis. This implies that the economic crisis had an impact on the reported accounting information. Thirdly, the Thai accounting committee revised and issued many Thai accounting standards in 1999 (as discussed in Chapter 2). Thai listed companies had to follow the new accounting standards and changing accounting standards could have had an effect on reported earnings. This may have caused earnings information to lose its explanatory power in predicting future cash flows. As seen from the result of prediction year 2000 using one-year lag, the earnings from the year 1999 were not significant in explaining future cash flows. Comparison of model using different year lags In Table 5.11 the mean of adjusted R2 of one-, two- and three-year lags are 0.23, 0.39, and 0.42 respectively. A comparison of the mean of the adjusted R2 between the model with twoyear lags of earnings and the model with one-year lag of earnings reveals that the model using two-year lags of earnings provides a higher adjusted R2 value, with the difference of mean of adjusted R2 0.16 (0.39-0.23). The three-year lags produced the highest values of adjusted R2. This implies that on average, the additional lags of earnings can improve predictive power. However, when considering the adjusted R2 for each prediction year, adjusted R2 values of the three-year lags analysis were not higher than those of predictions with two-year lags in the prediction years 1998 and 2000. This implies that the Asian economic crisis during the year 1997 and 1998 had an impact on the explanatory power of the earnings model.

128

One lag Year Predictor n Beta t sig. R

2

Two lags F d.f. sig. n Beta t sig. R 0.00

2

Three lags F d.f. sig. n Beta t sig. R 2 F d.f. sig.

Source: data analysis for this thesis

Table 5.11 Summary Statistics from Regressions of Future Cash Flow on Earnings During 1994 to 2002

1996 EARN95 173 0.22 2.92 EARN94 1997 EARN96 216 0.25 3.74 EARN95 EARN94

0.00 0.04 8.55

1,171 0.00 168 0.09 1.16 0.39 4.86

0.30 0.19 20.28 2,165 0.00

0.00 0.06 13.96 1,214 0.00 169 0.16

1.61 0.10 0.11 10.99 2,166 0.00 166 0.40 0.09 0.93 0.20 14.90 3,162 0.00 0.00 0.09 0.01 0.89 0.37 4.54 0.00 3.17 0.00 0.57 76.61 3,168 0.00 8.93 0.00

0.21 2.16

1998 EARN97 240 -0.70 -15.00 0.00 0.48 225.22 1,238 0.00 221 -0.38 -11.4 0.00 0.81 472.8 2,218 0.00 172 0.23 EARN96 EARN95 1999 EARN98 206 0.41 6.49 EARN97 EARN96 2000 EARN99 223 -0.02 -0.33 0.74 0.00 0.11 EARN98 EARN97 2001 EARN00 229 0.47 7.93 EARN99 EARN98 0.00 0.21 62.95 1,227 0.00 220 0.62 8.94 0.00 0.17 42.17 1,204 0.00 205 0.93 7.95 0.74 6.29 0.65 19.4 0.00 0.59

-0.04 -0.68 0.49 0.00 0.23 32.04 2,202 0.00 188 0.72 0.00 0.79 1,221 0.74 196 0.61 10.69 0.00 0.37 57.21 2,193 0.00 197 0.21 -0.02 -0.38 0.70 0.43 8.85 0.00 0.44 50.41 3,184 0.00 7.80 0.00 1.74 0.08 0.07 3.58 0.00 5.72 3,193 0.00

1.22 10.69 0.00

-0.02 -0.23 0.82 0.00 0.32 52.03 2,217 0.00 196 0.49 0.59 9.46 0.00 0.50 66.70 3,192 0.00 9.99 0.00

-0.09 -1.28 0.20

-0.08 -1.25 0.21

2002 EARN01 179 0.80 17.67 0.00 0.64 312.14 1,177 0.00 178 0.70 15.36 0.00 0.70 205.60 2,175 0.00 177 -0.18 -3.94 0.00 0.72 148.21 3,173 0.00 EARN00 EARN99 mean 0.23 0.39 0.24 5.31 0.00 0.31 6.63 0.00 0.42 0.74 15.90 0.00

129

Pooled-year analysis

In pooled-year analysis, all data for dependent and independent variables of the six-year prediction period of 1996 to 2001 was pooled and analysed together. The number of regression equations totalled three: one-, two- and three-year lags. The statistical results of pooled-year analysis with one-, two-, and three-year lags are shown in Table 5.14.

In using the one-year lag of earnings, earnings in year t-1 (EARNt-1) were not found to be a significant predictor (p > .05) of cash flow at year t (CFOt). In using the two-year lags of earnings, a significant relationship was found between the pair of independent variables (EARNt-1 and EARNt-2) and CFOt (F [2,1024] = 13.73, p < .005). The two independent variables were found to explain only 2% of the variation in CFOt (R2 = 0.02). When the relative contributions to prediction of the two independent variables were evaluated, both EARNt-1 and EARNt-2 were found to be positive and significant in prediction of CFOt, (t = 3.49, p< 0.005 and t = 0.4.87, p < 0.005 respectively). In using three-year lags of earnings, a significant relationship was found between the three independent variables (EARNt-1, EARNt2,

and EARN

t-3)

and dependent variable (CFOt) (F (3,927) = 13.44, p < 0.005). The three

dependent variables were found to explain only 4% of the variation in CFOt (adjusted R2 = 0.04). When the relative contributions to prediction of the three independent variables were evaluated, EARNt-1, EARNt-2, and EARN
t-3

were found to be positive and significant in

prediction of CFOt, (t = 3.73, p < 0.005, t = 5.53, p < 0.005, and t = 4.16, p < 0.005 respectively). Overall, these results are consistent with the results of the yearly analysis in that the earnings model had significant explanatory power in predicting future cash flows. Comparison of model using different year lags Similar to the results of the yearly analysis (Table 5.14) comparing the adjusted R2 of models with different year-lags of earnings, the three-year lags of earnings model had the greatest power in explaining future cash flows. This implies that additional year-lags of earnings can improve the explanatory power of the model in explaining future cash flows. However the adjusted R2 values were very small (difference of adjusted R2 = 0.02). This indicates that the

130

earnings model generated from pooled-year data did not have good explanatory power to predict future cash flows. This may result from the mixture of prediction years in the period of the study.

5.5.2 Cash flows model Simple and multiple regression models were performed between cash flows from operations in year t, with CFOt as the dependent variable, and year-lags of cash flows from operations, with CFOt-i as the independent variable. CFOt-i were expected to provide a powerful model for predicting future cash flows, CFOt. In the same manner as the earnings model, a year-lag of cash flows was added systematically to the regression equations to test the incremental predictive power of year-lags. The first equation contained only lag one of cash flows, the second equation contained both lags one and two of cash flows and the third equation contained lags one, two and three of cash flows (summarised in Table 5.1). The model using two- and three-year lags were expected to have more predictive power than the model using only the one-year lag. The cash flow model was tested in both yearly data and pooled-year data.

Yearly analysis

For the yearly analysis, the regression model was processed to analyse each set of prediction years (year t) separately. The analysis was performed for six prediction years spanning 1996 to 2002. In each prediction year, three sets of data were analysed including a set of the one-, two- and three-year lags of the cash flows. The statistical results of the regression procedure for each year prediction with one-, two- and three-year lags of cash flows are presented in Table 5.12.

Using a one-year lag for prediction (Table 5.12), cash flows from operations in year t-1 (CFOt1)

were found to be a significant predictor (p < 0.005) of cash flows from operations at year t

(CFOt), in all years except for the prediction year 1999 (p > 0.5). Excluding the prediction year 1999, CFOt-1 was found to explain between 15% and 88% of the variation in CFOt, with

131

an overall average of 40% (adjusted R2 = 0.15, 0.88, and 0.40, respectively). The explanatory power of the model was low in prediction years 1996, 1997 and 1998 (R2 = 0.19, 0.15 and 0.15). In contrast, the explanatory power of the model was high in prediction years 2000, 2001 and 2002 in which the Thai economy grew after suffering from the economic crisis (adjusted R2 = 0.67, 0.74, and 0.88). These results indicate that the explanatory power of the model in predicting the years before and after 1999 were different. Before 1999 the explanatory power of the model was low, but after 1999 the explanatory power of the cash flows model increased.

In using two-year lags for prediction, significant relationships were found between the pair of independent variables (CFOt-1 and CFOt-2) and the dependent variable (CFOt) in all years (p < 0.005). The two independent variables were found to explain between 22% and 88% of the variation in CFOt, with an overall mean of 54% (adjusted R2 = 0.22, 0.88, and 0.54, respectively). The explanatory power of the model was low in prediction years 1996, 1997 and 1999 (R2 = 0.22, 0.38 and 0.24). In contrast, the explanatory power of the model was high in prediction years 1998, 2000, 2001 and 2002 (R2 = 0.63, 0.66, 0.77 and 0.88). These results are consistent with the results of the model with a one-year lag in that the cash flows model had a higher explanatory power in the prediction years after 1999. When the relative predictive contributions of the two independent variables were evaluated, it was found that CFOt-1, CFO 1997 and CFO 1998 did not contribute significantly to the prediction of CFO 1998 (t = -0.02, p > .05) and CFO 1999 (t = 1.61, p > .05), and CFOt-2, CFO 1998, did not contribute significantly to the prediction of CFO 2000 (t = -0.80, p > .05). These results imply that the cash flow lags one and two, of the years 1997 and 1998 which were during the economic crisis, could not have explanatory power to predict future cash flows.

In using three-year lags for prediction, significant relationships were found between the independent variables (CFOt-1, CFOt-2 and CFOt-3) and the dependent variable (CFOt) in all years (p < 0.005). The three independent variables were found to explain between 19% and 92% of the variation in CFOt, with an overall mean of 54% (adjusted R2 = 0.19, 0.92, and 0.54, respectively). The explanatory power of the model was low in prediction years 1997, 1998 and 1999 (R2 = 0.47, 0.21 and 0.19). In contrast, the explanatory power of the model was

132

high in prediction years 2000, 2001 and 2002 (R2 = 0.66, 0.78 and 0.92). This indicates that the explanatory power of the three-year lags of cash flows model changed over the years. This is consistent with the results of the models with one- and two-year lags, as the explanatory power of the three-year lags model was higher for the prediction years after 1999. When the relative predictive contributions of the three independent variables were evaluated, it was found that CFOt-1, CFO 1998, did not contribute significantly to the prediction of CFO 1999 (t = 0.35, p > .05). CFOt-2, CFO 1995, CFO1996 and CFO 1998 did not contribute significantly to the prediction of CFO 1997 (t = -0.83, p > .05), CFO 1998 (t = -1.11, p > .05) and CFO 2000 (t = -0.78, p > .05), respectively. CFOt-3, CFO 1997 and CFO 1998 did not contribute significantly to the prediction of CFO 2000 (t = 0.92, p > .05) and CFO 2001 (t = -0.07, p > .05) respectively. These results correspond with the results of the two-year lag analysis, in that the cash flow lags one and two, of the years 1997 and 1998 during the economic crisis, did not have explanatory power to predict future cash flows.

Comparison of model using different year lags In Table 5.12 the mean of adjusted R2 of model using one-, two- and three- year lags are 0.40, 0.54, and 0.54 respectively. Comparing the adjusted R2 between the model with one- and twoyear lags, the model with two-year lags of cash flows provided a higher adjusted R2, with the difference in mean of adjusted R2 0.14 (0.54-0.40). However, the adjusted R2 of the model using three-year lags and the model using two-year lags are not different. This implies that on average, the additional year lags of cash flows can improve the predictive power of the model. When considering adjusted R2 of each prediction year, adjusted R2 of the model with threeyear lags were not higher than the model with two-year lags for the prediction years 1998 and 1999. This implies that the Asian economic crisis during the 1997 and 1998 had an impact on the explanatory power of the cash flow model.

133

Pooled-year analysis

In pooled year analysis, all data for dependent and independent variables of the six-year prediction period of 1996 to 2001 was pooled and analysed together. The statistical results of pooled-year analysis with one-, two- and three-year lags of cash flows are shown in Table 5.14.

In using the one-year lag of cash flow, cash flows in year t-1 (CFOt-1) was found to be a significant predictor (F [1,1128] = 21.19, p < .005) of cash flow at year t (CFOt). CFOt-1 was found to explain only 2% of the variation in CFOt (adjusted R2 =0.02).

Similarly, in using the two-year lags of cash flows, a significant relationship was found between the pair of independent variables (CFOt-1 and CFOt-2) and CFOt (F [2,1024] = 21.35, p < .005). The two independent variables were found to explain only 4% of the variation in CFOt (R2 = 0.04). When the relative contributions to prediction of the two independent variables were evaluated, both CFOt-1 and CFOt-2 were found to be positive and significant in prediction of CFOt, (t = 5.15, p< 0.005 and t = 3.84, p < 0.005 respectively).

In using three-year lags of cash flows, a significant relationship was found between the independent variables (CFOt-1, CFOt-2, and CFOt-3) and dependent variable (CFOt) (F [3,927] = 13.51, p< 0.005). The three dependent variables were found to explain 4% of the variation in CFOt (adjusted R2 = 0.04). When the relative predictive contributions of the three independent variables were evaluated, CFOt-1, CFOt-2, and CFOt-3 were found to be positive and significant in prediction of CFOt, (t = 4.51, p < 0.005, t = 3.21, p < 0.005, and t = 2.61, p < 0.05 respectively). These results are consistent with the results of the yearly analysis in that the cash flows model from pooled-year data had significant explanatory power in predicting future cash flows. Comparing the adjusted R2 of the models with different year-lags of cash flows, the explanatory power of the model with two-year lags was slightly higher than that of the model with a one-year lag, in explaining future cash flow (difference of adjusted R2 = 0.02).

134

However, the explanatory power of the model with two-year lags was not different from the model with three-year lags. These results are consistent with the results of yearly analysis in that the additional lags of cash flows can improve the explanatory power of the cash flows model. However, the results of the pooled-year analysis provide weak evidence because the model had a very low level of explanatory power. This may result from pooling the data of all the prediction years together. This suggests that the model generated from the pooled-year data has a lower explanatory power than the model generated from the yearly data.

In conclusion, these results indicate that year lags of cash flows could explain future cash flows of Thai listed companies. However, the explanatory power of yearly analysis changed over the years 1996 to 2002. The explanatory power of the model was high for the prediction years after 1999 and low for the prediction years before 1999. Moreover, the year-lags of cash flows could not explain future cash flows when they were obtained from the years 1997 and 1998, which were during the Asian economic crisis. It is possible to conclude that the Asian economic crisis had an impact on the explanatory power of the cash flows model.

135

Year Predictor

One lag n Beta t sig. R

2

Two lags F d.f. sig. n Beta t sig. R

2

Three lags F d.f. sig. n Beta t sig. R2 F d.f. sig.

Source : data analysis for this thesis

Table 5.12 Summary Statistics from Regressions of Future Cash Flows on Cash Flows During 1994 to 2002

1996 CFO95 171 0.45 6.47 0.00 0.19 41.79 1,169 0.00 168 0.34 4.24 0.00 0.22 24.18 2,165 0.00 CFO94 0.2 2.56 0.01 6.64 0.00 0.47 48.92 3,162 0.00 -0.83 0.41 5.39 0.00 3.83 0.00 0.21 16.35 3,168 0.00

1997 CFO96 216 0.39 6.23 0.00 0.15 38.76 1,214 0.00 167 0.40 5.92 0.00 0.38 51.61 2,164 0.00 166 0.43 CFO95 CFO94 0.33 4.80 0.00 -0.1 0.44

1998 CFO97 239 0.43 7.39 0.00 0.18 54.64 1,237 0.00 221 0.00 -0.02 0.99 0.63 188.22 2,218 0.00 172 0.31 CFO96 CFO95 1999 CFO98 206 0.09 1.30 0.20 0.00 1.68 CFO97 CFO96 0.80 19.38 0.00

-0.09 -1.11 0.27 0.29 3.63 0.00 0.35 0.73 0.19 15.87 3,185 0.00 5.99 0.00 3.42 0.00

1,204 0.20 205 0.10 1.61 0.11 0.24 32.26 2,202 0.00 189 0.02 0.48 7.90 0.00 0.39 0.23

2000 CFO99 222 0.82 21.04 0.00 0.67 442.52 1,220 0.00 195 0.82 19.46 0.00 0.66 189.42 2,192 0.00 195 0.80 18.32 0.00 0.66 126.50 3,191 0.00 CFO98 CFO97 -0.03 -0.80 0.42 -0.03 -0.78 0.44 0.04 0.92 0.36

2001 CFO00 227 0.86 25.54 0.00 0.74 652.03 1,225 0.00 220 0.66 13.36 0.00 0.77 367.24 2,217 0.00 194 0.67 12.89 0.00 0.78 223.62 3,190 0.00 CFO99 CFO98 0.26 2.34 0.00 0.26 0.00 4.89 0.00 -0.07 0.94 7.68 0.00 0.92 711.50 3,172 0.00 7.32 0.00 8.67 0.00 0.54

2002 CFO01 179 0.94 36.57 0.00 0.88 1337.0 1,177 0.00 178 0.55 8.72 0.00 0.88 663.22 2,175 0.00 176 0.42 CFO00 CFO99 Mean 0.40 0.54 0.41 6.47 0.00 0.38 0.24

136

5.5.3 Cash flows and accrual components of the earnings model

Simple and multiple regression models were performed between cash flows from operations year t, with CFOt as the dependent variable and year-lags of cash flows and accrual components of earnings (CFOt-1 and ACRt-1), as independent variables. CFOt-i and ACRt-i, were expected to provide a powerful model for predicting future cash flows, CFOt. In order to test the incremental predictive power of year-lags, year-lags of cash flows and accrual components of earnings were added systematically to the regression equations. The first equation contained only lag one, the second equation contained both lags one and two and the third equation contained lags one, two and three. The cash flows and accrual components of the earnings model with two- and three-year lags were expected to provide more predictive power than the model with the one-year lag. Both sets of data were analysed, namely, for each prediction year spanning 1996 to 2002 and for pooled-year period 1996 to 2001.

Yearly analysis

For the yearly analysis, the regression model was processed to analyse each set of prediction years (year t) separately. Using this method the statistical results are presented for each prediction year. The analysis was performed for six prediction years spanning 1996 to 2002. Each prediction year included three sets of data: a set of the one-, two- and three-year lags of cash flows and accrual components of earnings. The statistical results of the regression procedure for each year prediction with one, two and three year lags of cash flows and accrual components of earnings are presented in Table 5.13.

Table 5.13 reveals the results of regression analysis from the cash flows and accrual components of earnings models with one-, two- and three-year lags for each prediction year from 1996 to 2002. In using a one-year lag for prediction, significant relationships were found between a pair of independent variables (CFOt-1 and ACRt-1) and the dependent variable (CFOt) in all years (p < 0.005). The two independent variables were found to explain between 16% and 88% of the variation in CFOt, with an overall mean of 41% (adjusted R2 = 0.16, 0.88 and 0.48 respectively). The explanatory power of the model was low in prediction years

137

1996, 1997, and 1999 (R2 = 0.20, 0.16 and 0.17). In contrast, the explanatory power of the model was high in prediction years 2000, 2001, and 2002 (R2 = 0.68, 0.74, and 0.88 respectively). When the relative predictive contributions of the two independent variables were evaluated it was found that ACRt-1, ACR 1995 and ACR 1996 did not contribute significantly to prediction of CFO 1996 (t = 1.65, p > .05), CFO 1997 (t = 1.67, p > .05), CFO 2001 (t = 1.30, p > .05), and CFO 2002 (t = 0.90, p > .05). Moreover, the coefficient sign of ACRt-1 was significantly negative in prediction of CFO 1998 (t = -15.18, p < .005), and CFO 2000 (t = -3.66, p < .005) and the coefficient sign of CFOt-1 was significantly negative in prediction of CFO 1998 (t = -2.83, p < .05) which was the year during the economic crisis. This means that CFOt-1and ACRt-1 explained the cash flow of those years in opposite directions.

In using two-year lags for prediction, significant relationships were found between the independent variables (CFOt-1, CFOt-2, ACRt-1, and ACRt-2) and dependent variable (CFOt) in all years (p < 0.005). The four independent variables were found to explain between 30% and 90% of the variation in CFOt, with an overall mean of 67% (adjusted R2 = 0.30, 0.90 and 0.67 respectively). Similar to the model with a one-year lag, the explanatory power of prediction year 2002 was highest, followed by years 2001 and 2000 respectively (R2 = 0.88, 0.74 and 0.68 respectively). These results may reflect the improvement of the quality of accounting information, the preparation of which is based on the new Thai accounting standard effective in 1999. In contrast, the explanatory power was low in prediction year 1997, which was during the economic crisis and 1999, in which the new Thai accounting standards became effective (R2 = 0.16 and 0.17 respectively). When the relative predictive contributions of the four independent variables were evaluated it was found that CFOt-1, CFO 1996, did not contribute significantly to prediction of CFO 1997 (t = 0.25, p > .05). ACRt-1, ACR 1995 and ACR 2001, did not contribute significantly to prediction of CFO 1996 (t = 1.00, p > .05) and CFO 2002 (t = 0.81, p >.05), respectively and ACRt-2, ACR 1999 did not contribute significantly to prediction of CFO 2001 (t = -1.56, p > .05). These results indicate that the independent variables cannot significantly contribute to the explanatory power when the variables included in the prediction model were from either the years during the economic crisis or the year 1999 in which the new Thai accounting standards became effective.

138

Moreover, the coefficient sign of ACRt-1, ACR 1996, ACR 1997 and ACR 1999 was significantly negative in the prediction of CFO 1997, CFO 1998 and CFO 2000 respectively. And the coefficient sign of ACRt-2, ACR 2000, was significantly negative in the prediction of CFO 2002 and the coefficient sign of CFOt-1 was negative in the prediction year 1998. This shows that CFOt-1, ACRt-1 and ACRt-2 explained the cash flows of those years in the opposite directions.

In using three-year lags for prediction, significant relationships were found between the independent variables (CFOt-1, CFOt-2, CFOt-3, ACRt -1, ACRt-2 and ACRt-3) and the dependent variable (CFOt) in all years (p < 0.005). The six independent variables were found to explain between 51% and 93% of the variation in CFOt, with an overall mean of 70% (adjusted R2 = 0.51, 0.93 and 0.70 respectively). Similar to the models with one and two-year lags, the explanatory power of prediction year 2002 was highest, followed by 2001 and 2000 respectively (adjusted R2 = 0.93, 0.80 and 0.69 respectively), while the explanatory power of prediction year 1997 was lowest (adjusted R2 = 0.51). Overall, the explanatory power of the model increased from prediction year 1997 to 2002. When the relative predictive contributions of the six independent variables were evaluated it was found that CFOt-2, CFO 1995 and CFO 1998 did not contribute significantly to prediction of CFO 1997 (t = 0.45, p > .05) and CFO 1999 (t = 1.44, p > .05), and CFOt-3, CFO 1997 did not contribute significantly to prediction of CFO 2000 (t = 0.25, p > .05). ACRt-1, ACR 1996, ACR 1997, and ACR 2001 did not contribute significantly to prediction of CFO 1997 (t = 0.25, p > .05), CFO 1998 (t = 0.30, p > .05) and CFO 2002 (t = 0.08, p > .05), respectively. ACRt-2, ACR 1995, ACR 1998, ACR 1999, and ACR 2000 did not contribute significantly to prediction of CFO 1997 (t = 1.90, p > .05), CFO 2000 (t = 1.57, p > .05), CFO 2001 (t = -1.10, p > .05), and CFO 2002 (t = -1.72, p > .05), respectively. And ACRt-3, ACR 1994, ACR 1997 and ACR 1999 did not contribute significantly to prediction of CFO 1997 (t = 1.07, p > .05), CFO 2000 (t = 0.24, p > .05) and CFO 2002 (t = -0.36, p > .05) respectively. This indicates that the majority of independent variables could not contribute significantly to prediction when they were from the years 1997, 1998 and 1999, and when the prediction years were 1997 and 1998. Moreover, the coefficient sign of ACRt-1, ACR 1999, was negative in prediction year 2000.

139

The mean of adjusted R2 values between models with one-, two-, and three-year lags, threeyear lags of cash flows and accrual components of the earnings model contained the highest explanatory power for predicting future cash flow (mean adjusted R2 = 0.41, 0.67 and 0.70 respectively). These results are consistent with previous research such as Barth, Cram and Nelson (2002): using more than a one-year lag of cash flow and accrual component of earnings improves the accuracy of the cash flow prediction. However, the adjusted R2 value of the model with three-year lags did not improve to a large extent.

Pooled-year analysis

In pooled-year analysis, all data for cash flows of prediction years (CFOt) and cash flows and accrual components of earnings for lags one, two and three (CFOt-1, CFOt-2, CFOt-3, ACCt-1, ACCt-2, and ACCt-3) of the six-year prediction period of 1996 to 2001 was pooled and analysed together. The statistical results of pooled-year analysis with one, two and three-year lags are shown in Table 5.14.

In using the one-year lag of cash flows and accrual components of earnings, a significant relationship was found between the independent variables (CFOt-1 and ACRt-1) and CFOt (F [2,1126] = 59.99, p < .005). The two independent variables were found to explain 10% of the variation in CFOt (adjusted R2 = 0.10). When the relative predictive contributions of the two independent variables were evaluated, ACRt-1was not found to be significant in prediction of CFOt, (t = -0.84, p > 0.05).

Similarly, in using the two-year lags of cash flows and accrual components of earnings, a significant relationship was found between the independent variables (CFOt-1, CFOt-2, ACRt-1, and ACRt-2) and CFOt (F [2,1126] = 60.00, p < .005). The four independent variables were found to explain 10% of the variation in CFOt (R2 = 0.10). When the relative contributions to prediction of the four independent variables were evaluated, all independent variables were found to be positive and significant in prediction of CFOt (p< 0.05).

140

In using three-year lags of cash flows and accrual component of earnings, a significant relationship was found between independent variables (CFOt-1, CFOt-2, CFOt-3, ACCt-1, ACCt2,

and ACCt-3) and dependent variable (CFOt) (F (3,927) = 13.51, p< 0.005). The six

independent variables were found to explain 26% of the variation in CFOt (adjusted R2 = 0.26). When the relative contributions to prediction of the six independent variables were evaluated, all independent variables were found to be positive and significant in prediction of CFOt (p < 0.05). Comparing the adjusted R2 of the models with one, two-, and three-year lags, the model with three-year lags had the highest power in explaining future cash flow (adjusted R2 = 0.26). However, the explanatory power of the models with one- and two-year lags was not different (adjusted R2 = 0.10). Therefore, no conclusion could be drawn, as to whether the two-year lags of cash flows and accrual components of earnings had a higher explanatory power than the one-year lag of cash flows and accrual components of earnings model.

Overall, the results of both yearly and pooled-year analysis provide evidence that the cash flows and accrual components of the earnings model were significant in explaining future cash flows. However the explanatory power was unstable for each year of prediction. The model had a high explanatory power for the prediction years after 1999 in which the economy grew. This may imply that explanatory power depends on economic conditions.

141

Table 5.13 Summary Statistics from Regressions of Future Cash Flows on Cash Flows and Accrual

Year Predictor

One lag n Beta t sig. R

2

Two lags F d.f. sig. n Beta t sig. R

2

Three lags F d.f. sig. n Beta t sig. R 2 F d.f. sig.

1996 CFO95 171 0.57 ACR95 CFO94 ACR94 0.17

5.55 0.00 0.20 22.47 2168 0.00 167 0.46 1.65 0.10 0.11 0.55 0.43

3.86 0.00 0.30 18.57 4,162 0.00 1.00 0.32 3.63 0.00 3.04 0.00

Components of Earnings

1997 CFO96 216 0.52 ACR96 CFO95 ACR95 CFO94 ACR94 0.17

5.19 0.00 0.16 20.94 2,213 0.00 168 0.03 1.67 0.10

0.25 0.80 0.49 40.91 4,163 0.00 166 0.43 2.85 0.00 0.51 29.34 6,159 0.00 0.03 0.25 0.81 0.06 0.45 0.65 0.21 1.90 0.06 0.59 3.60 0.00 0.17 1.07 0.28

-0.43 -3.96 0.00 0.82 0.52 7.85 0.00 5.50 0.00

1998 CFO97 240 -0.14 -2.83 0.01 0.49 115.62 2,237 0.00 221 -0.08 -2.77 0.01 0.83 274.16 4,216 0.00 172 0.30 4.34 0.00 0.62 47.86 6,165 0.00 ACR97 CFO96 ACR96 CFO95 ACR95 -0.72 -15.18 0.00 -0.28 -7.22 0.00 1.42 20.36 0.00 0.90 14.17 0.00 0.02 0.30 0.77 0.64 5.97 0.00 0.84 8.27 0.00 0.24 2.09 0.04 0.28 2.97 0.00

1999 CFO98 206 2.28 ACR98 CFO97 ACR97 CFO96 ACR96 2.22

6.65 0.00 0.17 22.14 2,203 0.00 206 2.99 6.47 0.00 2.93 0.56 0.61

7.36 0.00 0.37 30.69 4,201 0.00 189 3.03 8.73 0.00 0.64 57.68 6,182 0.00 7.24 0.00 9.35 0.00 5.50 0.00 3.03 8.77 0.00 0.71 15.13 0.00 0.86 7.91 0.00 1.14 8.35 0.00 0.85 5.83 0.00

142

Year Predictor

Beta

One lag sig. R2 F

d.f.

sig.

Beta

Two lags sig. R2 F

d.f.

sig.

Beta

Three lags sig. R 2 F

Source: data analysis for this thesis

Table 5.13 (continued)

d.f.

sig.

2000 CFO99 222 0.77 19.15 0.00 0.68 240.43 2,219 0.00 222 0.75 17.47 0.00 0.69 109.53 4,190 0.00 195 0.75 16.62 0.00 0.69 72.31 6,188 0.00 ACR99 CFO98 ACR98 CFO97 ACR97 -0.15 -3.66 0.00 -0.13 -3.00 0.00 0.38 0.42 2.35 0.02 2.58 0.01 -0.13 -2.88 0.00 0.44 0.48 0.01 0.02 1.44 0.15 1.57 0.12 0.25 0.80 0.24 0.81

2001 CFO00 227 0.90 20.98 0.00 0.74 327.88 2,224 0.00 220 0.66 13.68 0.00 0.78 200.25 4,215 0.00 193 0.76 13.79 0.00 0.80 128.90 6,186 0.00 ACR00 CFO99 ACR99 CFO98 ACR98 2002 CFO01 179 0.96 26.14 0.00 0.88 668.16 2,176 0.00 177 0.49 ACR01 CFO00 ACR00 CFO99 ACR99 0.41 0.67 0.03 0.90 0.37 0.03 0.40 0.06 1.30 0.19 0.14 0.38 3.21 0.00 9.50 0.00 0.16 0.24 0.32 0.34 6.82 0.00 0.90 390.62 4,172 0.00 176 0.46 0.81 0.42 5.31 0.00 0.00 0.30 0.24 3.36 0.00 4.89 0.00 2.45 0.02 2.56 0.01 7.29 0.00 0.93 361.13 6,169 0.00 0.08 0.94 4.51 0.00 7.88 0.00

-0.06 -1.56 0.12

-0.04 -1.10 0.27

-0.13 -3.61 0.00

-0.07 -1.72 0.09 -0.01 -0.36 0.72 0.70

143

Source: data analysis for this thesis

Table 5.14 Summary Statistics from Regressions of All Models (Pooled-year Data)

One lag Predictor n Beta t sig. EARNt-1 1130 .05 1.66 .10 EARNt-2 EARNt-3 R2 F d.f. sig. n
Beta

Two lags t sig. R 2 F d.f. sig. n 931 Beta t

Three lags sig. R2 F d.f. sig.

.00 2.77 1,1128 .10 1027 .11 3.49 .00 .02 13.73 2,1024 .00 .16 4.87 .00

.13 3.73 .00 .04 .20 5.53 .00 .14 4.16 .00

13.44 3,927 .00

CFOt-1 1130 .14 4.60 .00 CFOt-2 CFOt-3

.02 21.19 1,1128 .00 1027 .16 5.15 .00 .04 21.35 2,1024 .00 .12 3.84 .00

931

.15 4.51 .00 .04 .10 3.21 .00 .08 2.61 .01

13.51 3,927 .00

CFOt-1 1129 .30 8.97 .00 ACCt-1 CFOt-2 ACCt-2 CFOt-3 ACCt-3 -.03 -.84 .40

.10 59.99 2,1126 .00 1026 .57 16.43 .00 .10 60.00 2,1126 .00 .07 2.33 .02 .50 10.66 .00 .54 10.75 .00

930

.56 15.09 .00 .26 .09 2.57 .01 .57 10.73 .00 .61 10.90 .00 .23 4.44 .00 .23 4.58 .00

55.12 6,623 .00

144

5.5.4 Cash flow ratios model

This section reports the statistical results of the regression model processed by stepwise regression analysis in which ten cash flow ratios were investigated in predicting future cash flows (CFOt). Since there were ten cash flow ratios (CFR) as independent variables expected to be included in the prediction model, a stepwise regression procedure was required to determine which cash flow ratio significantly contributes to explanations of future cash flows. The ratios providing the largest partial F statistic were added to the model. In the final step, the stepwise procedure generated suitable equations for the model. Stepwise regression produced models for each year of the analysis during the 1994 to 2002 period. A cash flow ratio was entered into the model if it met the statistical criteria and the order of inclusion was determined by the contribution of its cash flow prediction.

The ten cash flow ratios (see Section 4.4.2.5 in Chapter 4) were cash flow adequacy (CFR1), debt coverage (CFR2), repayment of borrowings (CFR3), dividend payment (CFR4), reinvestment (CFR5), cash flow on revenues (CFR6), cash flow to net income (CFR7), cash flow return on assets (CFR8), cash flow return on stockholders equity (CFR9), and cash flow per share (CFR10). The relationships of these cash flow ratios, as independent variables, and future cash flows (CFOt) were examined to build the prediction model for each prediction year. In order to test the incremental predictive power of year-lags of cash flow ratios, the models were tested in two ways; firstly, a model containing one-year lag of cash flow ratios (CFRt-1) and secondly a model containing two-year lags of cash flow ratios (CFRt-1 and CFRt-2). The statistical results of each year analysis are presented in Table 5.15.

The stepwise regression selected and placed cash flow ratios into models and generated suitable models for each prediction year 1996 to 2002. In using the one-year lag of cash flow ratios, for the prediction of cash flow year 1996, CFR6 was included in the model (F [1,175] = 12.37, p < 0.005). CFR6t-1 was found to explain 6% of the variation in CFOt (adjusted R2 = 0.06). For prediction year 1997, CFR9t-1, CFR2t-1 and CFR5t-1 were included in the model (F

145

[3,210] = 8.14, p < 0.005). These three independent variables were found to explain 9.1% of variation in CFOt (R2 = 0.091). For prediction year 1998 CFR8t-1 was included in the model (F [1,232] = 9.18, p < 0.005). It was found to explain only 3% of the variation in CFOt (R2 = 0.03). For prediction year 1999, CFR2t-1 and CFR3t-1 were included in the model (F (2,202) = 13, p < 0.005). These independent variables were found to explain 11% of the variation in CFOt (R2 =0.11). For prediction year 2000, CFR8t-1 was included in the model (F [1,221] = 47, p < 0.005) and was found to explain 17% of the variation in CFOt (R2 = 0.17). For prediction year 2001, CFR6t-1 and CFR10t-1 were included in the model (F [2,226] = 9.7, p < 0.005). These independent variables were found to explain only 7% of the variation in CFOt (adjusted R2 = 0.07). For prediction year 2002, CFR6t-1 and CFR5t-1 were selected into the model (F [2,176] = 32.7, p < 0.005). These independent variables were found to explain 26% of the variation in CFOt (adjusted R2 = 0.26).

In using two-year lags of cash flow ratios for prediction year 1996, CFR3t-2, CFR6t-1, CFR5t-2 and CFR6t-2 were included in the model (F [4,158] = 7.6, p < 0.005). These independent variables were found to explain 14% of the variation in CFOt (adjusted R2 = 0.14). For prediction year 1997, CFR10t-1, CFR2t-1 and CFR5t-1 were included in the model (F [4,158] = 7.6, p < 0.005). These independent variables were found to explain 50% of the variation in CFOt (adjusted R2 =0.50). For prediction year 1998, CFR10t-2, CFR7t-1, CFR8t-2, CFR8t-1, CFR10t-1 and CFR9t-2 were included in the model (F [6,208] = 44.1, p < 0.005). These independent variables were found to explain 54% of the variation in CFOt (adjusted R2 = 0.54). For prediction year 1999, CFR2t-1, CFR10t-2, CFR3t-1, CFR1t-2 and CFR8
t-2

were

entered into the model (F [5,194] = 10.2, p < 0.005). These independent variables were found to explain 19% of the variation in CFOt (adjusted R2 = 0.19). For prediction year 2000, only CFR6t-1 was included in the model (F [1,192] = 46.6, p < 0.005). It was found to explain 19% of the variation in CFOt (adjusted R2 = 0.19). For prediction year 2001, CFR8t-2 and CFR6t-1 were included in the regression model (F [2,217] =41.3, p < 0.005). These independent variables were found to explain 27% of the variation in CFOt (adjusted R2 =0.27). For prediction year 2002, CFR6t-1 and CFR5t-1 were included in the model (F [2,175] = 32.5, p < 0.005). They were found to explain 26% of the variation in CFOt (R2 = 0.26).

146

Source: data analysis for this thesis

Table 5.15 Summary Statistics from Stepwise Regressions of Future Cash Flows on Cash Flow Ratios

One lag Beta t sig. R2 F d.f. sig. Variable n 1996 CFR6_95 177 0.26 3.52 0.00 0.06 12.4 1,175 0.00 Year

n Beta Variable CFR3_94 163 0.24 CFR6_95 CFR5_94 CFR6_94 0.28 0.19

t sig. R2 3.32 0.00 0.14 3.52 0.00 2.61 0.01

Two lag F d.f. sig. 7.6 4,158 0.00

-0.2 -2.22 0.03 12.7 0.00 0.50 3.32 0.00 57 3,164 0.00

1997 CFR9_96 214 CFR2_96 CFR5_96

0.25 3.77 0.00 0.09 8.14 3,210 0.00 CFR10_96 168 0.69 0.21 3.12 0.00 -0.2 -2.3 0.02 CFR2_96 CFR5_96 0.19

-0.2 -2.68 0.01

1998 CFR10_96 234

0.20 3.03 0.00 0.03 9.18 1,232 0.00 CFR10_96 215 0.74 15 0.00 0.55 44.1 6,208 0.00 CFR7_97 0.1 2.08 0.04 CFR8_96 -0.3 -3.18 0.00 CFR8_97 0.16 2.95 0.00 CFR10_97 -0.2 -2.72 0.01 CFR9_96 0.23 2.28 0.02 13 2,202 0.00 CFR2_98 200 0.33 CFR10_97 CFR3_98 CFR1_97 CFR8_97 47 1,221 0.00 4.86 0.00 0.19 10.2 5,194 0.00

1999 CFR2_98 205 0.35 5.03 0.00 0.11 0.16 -2.4 0.02

0.35 4.00 0.00 -0.2 -2.68 0.01 0.14 2.21 0.03 -0.2 -1.99 0.05 6.83 0.00 0.19 46.6 1,192 0.00 7.53 0.00 0.27 41.3 2,217 0.00 3.36 0.00

2000 CFR8_99 194 0.42 6.86 0.00 0.17

CFR6_99 194 0.44 CFR8_99 200 0.45 CFR6_00 0.2

2001 CFR6_00 205 0.22 3.37 0.00 0.07 9.7 2,226 0.00 CFR10_00 0.13 2.04 0.04 2002 CFR6_01 179 0.49 7.64 0.00 0.26 32.7 2,176 0.00 CFR5_01 -0.2 -2.6 0.01

CFR6_01 178 0.49 7.63 0.00 0.26 32.5 2,175 0.00 CFR5_01 -0.2 -2.61 0.01

147

In summary, the results of each prediction year were not consistent after using one- and twoyear lags of cash flow ratios as the stepwise procedure generated different models for each prediction year. This implies that some cash flow ratios are significant in the prediction of future cash flows such as CFR3, CFR5, CFR6, CFR7, CFR8, CFR9 and CFR10. However, it cannot be concluded that cash flow ratios have good predictive power, because the explanatory power of each model built for each prediction year was low in explaining future cash flows; only 6% to 26% of the variation in future cash flows can be explained by the models. These results show that the cash flow ratios model built from stepwise regression is not stationary and may be useless in practice.

5.6 Comparison of Explanatory Ability among Models

This section reports the predictive ability among the prediction models, earnings, cash flows and cash flow and accrual components of earnings models, by comparing the adjusted R2 value of each model summarised in Table 5.16. The adjusted R2 value or coefficient of determination can be used to determine how well a set of predictors statistically explains a dependent variable (Hair, Anderson, Tatham & Black 1992; Neuman 2003). The adjusted R2 value has been employed in previous research to evaluate the explanatory power among models, such as Barth, Cram and Nelson (2001), Quirin et al (1999), Greenberg, Johnson and Ramesh (1986) and McBeth (1993).

Yearly analysis

For each year analysis, among the three models, the cash flows and accrual components of earnings model (CFO and CAR model) within one-, two- and three-year lags, provided the highest adjusted value for all prediction years from 1996 to 2002 (mean of adjusted R2 = 0.41, 0.67 and 0.70 respectively). Comparing the cash flows model (CFO model) and the earnings model (EARN model), the cash flows model within all year-lag models, provided a higher adjusted R2 value than the earnings model in five of the seven years. The earnings model with all year lags, provided a higher adjusted R2 than the cash flows model for prediction years 1998 and 1999. This indicates that the cash flow and accrual component of the earnings model

148

can explain future cash flows more accurately than the other models; and, the earnings model can explain future cash flows less accurately than the other models. However from prediction year 2000 to 2002 the difference of adjusted R2 values between the cash flows model and the CFO and ACR model is very small.

Pooled-year analysis

Regarding the one-year lag model, it can be seen that the cash flow model (CFO model) provides higher adjusted R2 values than the one-year lag of the earnings model (EARN model) roughly 0.02 (0.02-0.00). The cash flows and accrual components of the earnings model (CFO and ACR model) provided the highest adjusted R2 values (adjusted R2 = .10). This means that the cash flows and accrual components of the earnings model provided the highest explanatory power for future cash flows followed by the cash flows model.

Regarding the two-year lag model, the cash flow and accrual component of the earnings model (CFO and ACR model) has the highest adjusted R2 values (adjusted R2 = 0.10). That is, its adjusted R2 value is higher than that of the earnings model by about 0.10 (0.10-0.00) and it is higher than that of the cash flow model by about 0.08 (0.10-0.02). Comparing the earnings model and the cash flows model, the cash flows model provided higher adjusted R2 values. This means that the cash flows and accrual components of the earnings model had the highest explanatory power, followed by the cash flows model.

Regarding the three-year lag model, the cash flows and accrual components of earnings provide the highest adjusted R2 values (adjusted R2 = 0.26). The adjusted R2 value of the cash flows model was equal to that of the earnings model. This means that the cash flows and accrual components of the earnings model provided the highest explanatory power for future cash flows. However, it cannot be concluded that the cash flows model had a higher explanatory power than the earnings model.

Overall, for both yearly and pooled-year analysis, the cash flows and accrual components of earnings has the highest explanatory power and the cash flow model has better explanatory

149

power of future cash flow than the earnings model. This may indicate that adding the accrual component of earnings into the cash flow model improves the explanatory ability of the model.
Table 5.16 Summary of Coefficient of Determination (Adjusted R2) for Each Model
Prediction year One lag EARN model Two lags 1996 1997 1998 1999 2000 2001 2002 Mean Pooled year 0.04 0.06 0.48 0.17 0.00 0.21 0.64 0.23 0.00 0.19 0.11 0.81 0.23 0.37 0.32 0.70 0.39 0.02 0.20 0.57 0.44 0.07 0.50 0.72 0.42 0.04 Three lags 0.19 0.15 0.18 0.00 0.67 0.74 0.88 0.40 0.02 One lag CFO model Two lags 0.22 0.38 0.63 0.24 0.66 0.77 0.88 0.54 0.04 0.47 0.21 0.19 0.66 0.78 0.92 0.54 0.04 Three lags 0.20 0.16 0.49 0.17 0.68 0.74 0.88 0.41 0.10 CFO & ACR model One lag Two lags 0.30 0.49 0.83 0.37 0.69 0.78 0.90 0.67 0.10 0.51 0.62 0.64 0.69 0.80 0.93 0.70 0.26 Three lags

Source: data analysis for this thesis

5.7 Evaluation of the Predictive Ability of the Models with Different Approaches

This section provides additional analysis to support the validity of the three models: earnings, cash flows, and cash flows and accrual components of earnings, and to evaluate which model provides the best predictor for future cash flows, by considering different approaches. The sample data was divided into two groups, the first group for building regression equations of the three models and the second group for generating predicted cash flows. The three models were evaluated by both yearly and pooled-year analysis.

The predictive power of the three models is measured by the correlation between predicted and actual cash flows (r), the square of the correlation (r2) and mean absolute percentage errors (MAPE). The r between predicted and actual cash flows was estimated by Pearsons correlation coefficient. The r and r2 show how much a regression equation can predict future cash flows. A high correlation indicates that a model has high predictive power. MAPE shows the mean values of percentage of errors. A low value of MAPE means that a model has a high

150

predictive power. The errors (E), and mean absolute percentage errors (MAPE) of each model were calculated as below.

E MAPE

= predicted cash flow actual cash flows = mean|(predicted cash flows- actual cash flows)/actual cash flow|

Yearly analysis

For the yearly analysis, the sample data was divided into two groups. The first group was comprised of data for prediction years 1996 to 1999 and the second group was comprised of data for prediction years 2000 to 2002. The first data group was employed to build regression equations of each model with one, two and three-year lags of independent variables. The second data group was utilised to generate predicted cash flows for years 2000, 2001 and 2002, calculated by using the regression equations of the three prediction models formulated from the first data group. That is, the regression equations generated from prediction years 1997, 1998, and 1999 were employed to predict cash flows for years 2000, 2001 and 2002 respectively. Then the correlation (r) between predicted and actual cash flows of each year, 2000, 2001 and 2002 and the associated r2, was calculated. In addition, the mean absolute percentage errors (MAPE) of each model were calculated. The r, r2 and MAPE of each model are presented in Table 5.17.
Table 5.17 r2 (r) and MAPE for Each Model
Prediction year 2000 Approach r2 (r)
MAPE

N One lag .02 (.17*) 11.98 .27 (-52 **) 6.29 .33 (.58 **) 15.35

181

2001

r2 (r)
MAPE

181

2002

r2 (r)
MAPE

181

EARN model Two Three lags lags 0.09 .00 (.30 **) (-.06) 12.09 29.59 .03 .08 (.18 **) (.29**) 7.16 8.63 .41 .18 (.64 **) (.42**) 18.22 9.16

One lag 1 (1 **) 10.06 .73 (.86**) 6.30 .74 (.86**) 14.17

CFO model Two Three lags lags .03 .38 (.17*) (.62 **) 13.90 11.34 .52 .02 (.72**) (.14 *) 7.02 10.67 .79 .86 (.89**) (.93 **) 12.25 12.70

CFO and ACR model One Two Three lags lags lags .37 .04 .18 (.61**) (.20**) (.43**) 10.83 16.48 12.19 .00 .09 .18 (-.05) (.30**) (.43**) 5.38 6.27 8.02 .37 .69 .44 (.61**) (.83**) (.66**) 15.18 16.51 8.58

* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).

151

As seen in Table 5.17, there was a significant relationship between actual and predicted cash flows in all prediction years at the 0.05 and 0.01 levels, apart from prediction year 2000, with the three-year lags of earnings model. Considering each prediction year, for prediction year 2000, using one-year lag of predictors, the cash flows model provided the highest value of r2 followed by cash flows and accrual components of earnings model (CFO and ACR model) (r2 = 1 and 0.37 respectively). Using two-year lags of predictors, the earnings model (EARN model) provided the highest values of r2 followed by CFO and ACR model (r2 = 0.09 and 0.04 respectively). Using three-year lags of predictors, the cash flows model (CF model) provided the highest value of r2, followed by the CFO and ACR model (r2 = 0.38 and 0.18 respectively), while the r of the earnings model (EARN model) was not significant. For prediction year 2001, using a one-year lag of predictors, the CF model provided the highest value of r2 followed by the EARN model (r2 = 0.73 and 0.72) while the r of the CFO and ACR model was not significant. Using two-year lags, similar to using one-year lag, the CFO model provided the highest value of r2 followed by the CF and ACR model (r2 = 0.52 and 0.09). Using threeyear lags, the CFO and ACR model provided the highest r2 value, followed by the EARN model (r2 = 0.18 and0.08) For prediction year 2002, the CFO model provided the highest r2 value, followed by the CFO and ACR model in all year-lags of predictors. This analysis indicates that in estimating the predictive power of the model by r2, the vast majority of results shows that the CFO model had a higher predictive power than the other models and the CFO and ACR model had a higher predictive power than the earnings model.

This analysis also shows the mean absolute percentage errors (MAPE) of each model (Table 5.17), where the smaller the MAPE, the better the predictive model. For prediction year 2000, using one-year lag of predictors, the CFO model provided the smallest MAPE, while the EARN model provided the largest MAPE (MAPE = 10.06 and 11.98); using two-year lags of predictors, the EARN model provided the smallest MAPE, while the CFO and ACR model provided the largest MAPE (MAPE = 12.08 and 16.48); using three-year lags of predictors, similar to using one-year lag of predictors, the CFO model provided the smallest MAPE, while the earnings model provided the largest MAPE (MAPE = 11.34 and 29.59). For prediction year 2000, the vast majority of results shows that the CFO model produced the

152

smallest MAPE, therefore it is possible to conclude that the CFO model provided a higher accuracy of prediction than the others. For prediction year 2001, the CFO and ACR model provided the smallest MAPE in using all year-lags of predictors. So this shows that the CFO and ACR model produced a higher accuracy of prediction than the other models. For prediction year 2002, the CFO model provided the smallest MAPE, in using one and two year lags of predictors (MAPE = 14.17 and 12.25), while the EARN model provided the highest MAPE. Therefore, the CFO model provided a higher accuracy of prediction than the other models. Using three-year lags, the CF and ACR model provided the smallest MAPE (MAPE = 8.58) while the CF model provided the largest MAPE, (MAPE = 12.70). Therefore, the CFO and ACR model was a better predictor than the other models Overall, the results of r2 show that the cash flows model had a higher predictive power than the other models and the earnings model had the lowest predictive power, while the results of the MAPE values provide weak evidence that the cash flows model had higher predictive power than the CF and ACR models. When considering both approaches, these results could not provide strong evidence as to whether the cash flows and accrual components of the earnings model had higher predictive ability than the cash flow model. However, both sets of results show evidence that the earnings model had the lowest predictive power. In conclusion, these two approaches provide inconsistent results, particularly in prediction year 2001, which the prediction models derived from the prediction year 1998. This may result from the impact of the economic crisis on the prediction found in the results of the original analysis in Section 5.5. The results of the two approaches are consistent for the prediction year 2000 and 2002. This may indicate that the results of prediction years 2000 and 2002 are more reliable because they do not contain data from the year 1998 in any year-lag. Therefore, if considering only prediction years 2000 and 2002, the results show that the CFO model had a high predictive value than the CF and ACR model. In addition, when comparing these results with the adjusted R2 from the original yearly analysis in Section 5.5 (Table 5.16), the results of these two approaches are inconsistent with the adjusted R2 from the original analysis. This indicates that using the adjusted R2 may not provide an accurate measurement. The results of the adjusted R2 from the original analysis

153

show that the cash flows and accrual components of the earnings model is a better predictor than the other models, however in some prediction years, the cash flows and accrual components of the earnings model provide slightly better explanatory power than the cash flows model. It is possible that using these two approaches can assist in deciding that the cash flows model is a better predictor than the cash flows and accrual components of earnings.

Pooled-year analysis

In the same manner as yearly analysis, the sample data of the pooled-year analysis was separated into two groups. The first group, containing data for prediction years 1996 to 2001, was utilised to formulate a regression equation of each model with one-, two- and three-year lags of independent variables; and the second group, containing data for prediction year 2002, was employed to generate predicted cash flows which were calculated by using the regression equations formulated from the first data group. The correlation between predicted and actual cash flows (r), the associated r2, and the mean of absolute percentage errors (MAPE) were then calculated. The values of these indicators are reported in Table 5.18
Table 5.18 Summary of the Values of r2 (r) and MAPE for Each Model (pooled year)
Predictors r2 (r) MAPE N 181 181 EARN model Two Three lags lags .34 0.45 0.32 (0.58**) (0.67**) (0.57**) 21.93 23.69 22.79 One lag One lag 0.74 (0.86**) 21.23 CF model Two lags 0.81 (0.90**) 21.09 Three lags 0.83 (0.91**) 20.49 CF and ACR model One lag Two Three lags lags 0.72 0.74 0.73 (0.85**) (0.86**) (0.85**) 21.06 23.42 22.28

Source: data analysis for this thesis Note ** Correlation is significant at the 0.01 level (2-tailed).

Table 5.18 shows the values of each indicator, the correlation (r), square of r (r2), and mean of absolute percentage of errors (MAPE). There was a significant correlation between actual and predicted cash flows for all models at the 0.01 level. Regarding the value of r2, the cash flows model (CFO model) provided the highest value of r2 for using all year lags of predictors, (r2 = 0.86, 0.90 and 0.91) while the earnings model (EARN model) provided the lowest value of r2 (r2 = 0.34, 0.45 and 0.32). This means that the CF model had the highest predictive power and the EARN model had the lowest predictive power. Regarding the mean of absolute percentage errors (MAPE), the CF model provided the lowest value of MAPE, for using one- and two-

154

year lags of predictors; the CF and ACR model provided the largest MAPE for using threeyear lags of predictors; and the EARN model provided the highest value of MAPE. This means that the CFO model had the highest predictive power while the EARN model had the lowest predictive power.

Comparing the two approaches, overall the results of both approaches are consistent. The majority of the results show that the cash flows model had higher predictive power than other models because it outperformed others when evaluated by r2 and MAPE, whereas the earnings model provided the lowest predictive power.

In summary, the results of the pooled-year analysis support the cash flows model as producing a higher predictive power than the other models and the earning model as having the lowest predictive power in predicting future cash flows of Thai listed companies. When considering the adjusted R2 from the original pooled-year analysis in Table 5.15, although those results show that adjusted R2 values of the cash flows and accrual components of the earnings model are the highest, the model provides low adjusted R2 values (adjusted R2 = 0.1, 0.1 and 0.26 for models with one-, two- and three-year lags), which means that the cash flows and accrual components of the earnings model from pooled-year data provide low predictive power. The results of the r2 approaches support the model from pooled-year data also as providing predictive power in predicting future cash flows because the model provides a high r2 value (r2 = 0.86, 0.90 and 0.91 for one-, two- and three-year lag models) showing a high relationship between actual and predicted cash flows. 5.8 Test of Hypotheses

This section discusses the testing of the five hypotheses constructed in Chapter 3. These five hypotheses were developed from the research problem, How can cash flow and accrual accounting data be used to predict future cash flows of Thai listed companies? The testing of the three hypotheses is discussed below.

155

Hypothesis 1: Past earnings have significant predictive power in predicting future cash flows of Thai listed companies.

The first hypothesis deals with testing the ability of earnings in predicting future cash flows (Model 1). One-, two- and three-year lags of earnings were investigated with the simple and multiple regression models as mentioned in Chapter 4 (Section 4.5.1).

According to the statistical results in Table 5.11, for yearly analysis, on average, the F-statistic is significantly different from zero (p < 0.005) for all prediction years except prediction year 2000. Similarly, the statistical result of the pooled-year analysis in Table 5.14 showed that the F-statistic is significantly different from zero (p < 0.005). The results lead to accepting the Hypothesis 1. Additionally, the majority of the coefficients showed positive signs. This supports earnings as positive and significant in predicting future cash flows. Therefore, the first hypothesis is valid. Earnings have a significant predictive power for predicting future cash flows of Thai listed companies.

Hypothesis 2 : Past cash flows have significant predictive power in predicting future cash flows of Thai listed companies.

The second hypothesis involves testing the ability of past cash flows in predicting future cash flows (Model 2). As seen in Table 5.12 for yearly analysis, the level of R was found to be significantly different from zero (p < 0.005). Table 5.14 for pooled-year analysis also provides the evidence that the cash flow model was significantly different from zero in predicting future cash flows (p <0.005). Additionally, the majority of cash flow variables showed positive signs. These results support the notion that cash flows are positive and significant in predicting future cash flows. The results lead to accepting the Hypothesis 2. Cash flows are positive and significant for predicting future cash flows of Thai listed companies.

156

Hypothesis 3: Past cash flows and accrual components of earnings have significant predictive power in predicting future cash flow of Thai listed companies.

The third hypothesis involves testing the ability of cash flows and accrual components of earnings in predicting future cash flows (Model 3). The statistical results shown in Table 5.13, indicate that the third model is significantly different from zero with p value< 0.005. In Table 5.14 the results also show that the F-statistic is significantly different from zero at p value< 0.005. These results provide evidence that cash flows and accrual components of earnings are significant in predicting future cash flows. Therefore Hypothesis 3 is accepted. That is cash flows and accrual components of earnings have a significant predictive power for predicting future cash flows of Thai listed companies.

Hypothesis 4: There are different predictive powers between three prediction models, earnings, cash flows and cash flows and accrual components of earnings models.

The fourth hypothesis was constructed in order to answer the question of whether there are different predictive powers between the predictive models, the earnings, cash flows and cash flows and accrual components of earnings models. Regarding the adjusted R2 between the three models in Table 5.15, the adjusted R2 between the three models are different. Moreover, when evaluating the predictive power by two approaches, the r2 and MAPE (in Tables 5.17 and 5.18) the results show that the three models have different predictive powers. Therefore the null hypothesis is rejected. The results lead to accepting Hypothesis 4, that there are different predictive powers between the three prediction models, the earnings, cash flows and cash flow and accrual components of earnings models Thai listed companies. In addition, when comparing the adjusted R2 between the three models (Table 5.15), the cash flows and accrual components of the earnings model provided a higher value of adjusted R2 than the earnings model and the cash flow model, and the earnings model provided a lower value of adjusted R2 than the others. Nevertheless when comparing r2 and MAPE (in Tables 5.17 and 5.18) from the out-of-sample period for both yearly and pooled-year analysis, the majority of results shows that the cash flows model has higher predictive power than the other

157

models. Therefore this indicates that although the cash flow and accrual components of the earnings model provides the high adjusted R2 which means that they can explain variation in future cash flows better than the earnings model and the cash flows model, when using the three models to predict future cash flow in the other data periods, the cash flows model produces greater accuracy than the other models.

Hypothesis 5: Past cash flow ratios are significant predictors of future cash flows of Thai listed companies.

This hypothesis involves testing the predictive power of cash flows ratios in predicting future cash flows (Model 4). The results of the stepwise analysis in Table 5.16 show that the suitable cash flow ratios model built for each prediction year contains different ratios. Consequently, this indicates that the cash flow model is not stationary. Moreover, the explanatory power of the model for each prediction year was low. This lead to rejecting Hypothesis 5, inferring that cash flow ratios are not good predictors for predicting future cash flows of Thai listed companies.

5.9 Additional Analysis

5.9.1 Test for the predictive ability of disaggregated accrual components of earnings

The analysis aims to test the ability of the components of earnings, cash flow and accrual components for predicting future cash flows in which the accrual components were (DEP), change in accounts receivable (AR), changes in inventories (INV), change in accounts payable (AP) and change in other current assets and liabilities (OTH). These major components were expected to be useful for predicting future cash flows and to have higher explanatory ability than the cash flows and aggregated accrual components model (model 3 in section 5.5). disaggregated into their major components including depreciation and amortisation expenses

158

In the analysis, a multiple regression model was performed between the cash flows from operations year t, CFOt, as the dependent variable and a one-year lag of the independent variables; cash flows from operations year t-1 (CFOt-1), and disaggregated accrual components receivable (ARt-1) and changes in inventories (INt-1), change in accounts payable (APt-1), (discussed in section 5.5), this analysis was tested in yearly and pooled-year data. However, due to the limitation of the number of independent variables in the regression model, only one-year lag of cash flows and disaggregated accrual components of earnings were included in the model. For yearly analysis, data of each prediction year from 1995 to 2002 was examined and for pooled-year analysis, to be comparable to other models, data for all prediction years from 1996 to 2002 was pooled and analysed together. The statistical results of the regression procedure for yearly and pooled-year analyses are summarised in Tables 5.19 and 5.20 respectively. of earnings, including depreciation and amortisation expenses (DEPt-1), change in accounts

change in other current assets and liabilities (OTHt-1). Similar to the three main models

Yearly analysis

In the yearly analysis, the regression model was processed to analyse data for each set of a prediction year (year t) separately. The analysis was performed for eight prediction years spanning 1995 to 2002. The statistical results of the regression procedure for each year prediction with a one-year lag of cash flows and disaggregated accrual components of earnings are presented in Table 5.19.

Table 5.19 reveals the results of the regression analysis from the cash flows and disaggregated accrual components of the earnings model for each prediction year from 1995 to 2002. It can be seen that significant relationships were found between the six independent variables; cash flows from operations year t-1 (CFOt-1), depreciation and amortisation expenses (DEPt-1), payable (APt-1), change in other current assets and liabilities (OTHt-1) and the dependent change in accounts receivable (ARt-1), changes in inventories (INt-1), change in accounts variable (CFOt) in all years (p < 0.005). The six independent variables were found to explain between 16% and 97% of the variation in CFOt, with an overall mean of 61% (adjusted R2 =

159

0.16, 0.97 and 0.61 respectively).

2

The explanatory power of the model was lowest in

prediction year 1999 (R = 0.16) and highest in prediction year 2002 (R2 = 0.16). The explanatory power of the model increased for the prediction years after the year 1999. When the relative contributions to prediction of the four independent variables were evaluated it was found that CFOt-1, did not contribute significantly to the prediction of CFO 1998 (t = 1.70, p > CFO 2001 (t = 0.20, p >.05), INt-1 did not contribute significantly to prediction of CFO 1996 (t = 0.96, p > .05), APt-1 did not contribute significantly to prediction of CFO 1995 (t = 0.15, p > .05) and CFO 1999 (t = 0.51, p > .05), OTHt-1 did not contribute significantly to prediction of CFO 1996 (t = 0.49, p > .05), CFO 1997 (t = 0.55, p > .05), CFO 1998 (t = -1.71, p > .05). .05). DEPt-1 did not contribute significantly to prediction of CFO 1999 (t = -0.57, p > .05) and

Overall, these results indicate that the cash flows and disaggregated accrual components of the earnings model had explanatory power to predict future cash flows and the explanatory power increased from the prediction year 2000 onwards. However some independent variables could not explain individually the cash flows of prediction years before the year 1999.

Pooled-year analysis

In the pooled-year analysis, all data for dependent and independent variables of the six-year prediction period of 1996 to 2001 was pooled and analysed together. The statistical results of pooled year analysis with one-year lag of disaggregated accrual components of earnings are showed in Table 5.20.

In using the one-year lag of cash flow and disaggregated accrual components of earnings, a significant relationship was found between the six independent variables, CFOt-1, DEPt-1, ARt-1, INt-1, APt-1, OTHt-1, and the dependent variable, CFOt, (F [6,1083] = 53.76, p <

.005). The six independent variables were found to explain 23% of the variation in CFOt (adjusted R2 = 0.23). When the relative contributions to prediction of the six independent variables were evaluated, they were found to be significant in predictions of CFOt, (p > 0.05).

160

Table 5.19 Summary of Statistical Results of Regression Analysis from the Cash Flows and Disaggregated Accrual Components of Earnings Model for Each Prediction Year from 1995 to 2002.
Prediction 1995 Predictor CFO 94 DEP94 AR94 IN94 AP94 OTH94 CFO 95 DEP95 AR95 IN95 AP95 OTH95 CFO 96 DEP96 AR96 IN96 AP96 OTH96 CFO 97 DEP97 AR97 IN97 AP97 OTH97 CFO 98 DEP98 AR98 IN98 AP98 OTH98 CFO 99 DEP99 AR99 IN99 AP99 OTH99 CFO 00 DEP00 AR00 IN00 AP00 OTH00 CFO 01 DEP01 AR01 IN01 AP01 OTH01 n 174 Beta 1.03 -.15 .31 .72 .01 .99 .41 -.26 -.26 .12 .41 .09 .37 .30 -.20 -.04 -.08 .10 .09 .77 -.08 -.15 -.23 -.09 1.60 -.12 .23 -.26 .04 1.57 .67 .12 -.24 -.12 -.34 -.07 .95 .05 -.13 -.13 -.23 .21 .66 .23 -.27 -.17 -.14 .01 t 11.41 -2.19 4.00 4.94 0.15 4.60 3.93 -2.56 -2.81 .96 4.84 .49 3.19 3.49 -.22 -.35 -.95 .55 1.70 11.86 -2.02 -3.68 -3.78 -1.71 3.29 -.57 2.40 -2.07 0.51 3.23 14.96 2.52 -6.95 -3.27 -10.54 -2.14 21.22 1.30 -3.09 -3.30 -6.32 4.07 23.46 7.24 -13.08 -9.83 -8.50 .42 sig. .00 .03 .00 .00 .89 .00 .00 .01 .01 .34 .00 .62 .00 .00 .04 .00 .00 .09 .09 .00 .04 .00 .00 .09 .00 .57 .02 .04 .61 .00 .00 .01 .00 .00 .00 .03 .00 .20 .00 .00 .00 .00 .00 .00 .00 .00 .00 .68 adj. R2 0.71 F 72.29 d.f. 6,167 sig. .00

1996

169

0.37

17.12

6,162

.00

1997

215

0.27

14.43

6,208

.00

1998

240

0.66

76.69

6,233

.00

1999

204

0.16

7.63

6,197

.00

2000

221

0.87

256.11

6,214

.00

2001

227

0.81

157.10

6,220

.00

2002

177

0.97

1005.47

6,170

.00

Mean

0.61

Source: data analysis for this thesis

161

Table 5.20 Summary of the Statistical Results from Pooled Year Analysis of Cash Flows and Disaggregated Accrual Components of Earnings Model
Prediction 1996-2002 Predictor CFOt-1 DEPt-1 ARt-1 INt-1 APt-1 OTHt-1 n 1090 Beta 0.28 0.44 -0.12 -0.15 0.09 0.20 t 4.93 14.72 -3.94 -4.90 2.70 3.15 sig. 0.00 0.00 0.00 0.00 0.00 0.00 adj. R2 0.23 F 53.76 d.f. 6,1083 sig. 0.00

Source: data analysis for this thesis Table 5.21 Comparison of Adjusted R2 between Cash Flows and Aggregated Accrual Components of Earnings Model and Cash flows and Disaggregated Accrual Components of Earnings Model
Prediction year 1996 1997 1998 1999 2000 2001 2002 Mean Pooled year CFO and ACR model 0.20 0.16 0.49 0.17 0.68 0.74 0.88 0.41 0.10 CFO and DACR model 0.37 0.27 0.66 0.16 0.87 0.81 0.97 0.61 0.23

Source: data analysis for this thesis

The results of the pooled-year analysis are consistent with the yearly analysis, in that the cash flows and disaggregated accrual components of earnings model provided explanatory power for future cash flows. In addition, in comparison with the cash flows and aggregated accrual components models (as shown in Table 5.21), the cash flows and disaggregated accrual components of earnings model provided a higher adjusted R2 than the cash flows and aggregated accrual components of earnings model for either yearly analysis except for the prediction year 1999, or pooled-year analysis (mean of adjusted R2 = 0.61 for yearly analysis and adjusted R2 = 0.23 for pooled-year analysis). These results are consistent with the finding of Barth, Cram and Nelson (2001), in that the cash flows and disaggregated accrual components of earnings model provided higher explanatory power than the cash flows and aggregated accrual components of earnings model.

162

5.9.2 Pooled-year analysis excluding prediction year 1998 (economic crisis)

This section reports the statistical result of pooled year analysis excluding prediction year 1998. This analysis is based on the assumption that the Asian economic crisis in 1998 had an impact on the predictive ability of the prediction models. Referring to the results of each years analysis, it can be seen that the predictive power of each model dropped in prediction year 1998. This corresponds to the results of the correlation analysis (Section 5.3.2). That is, the cash flow variable for years 1997 and 1998 did not relate significantly to the independent variables. Furthermore, the results of the pooled data, including all prediction years, show that the prediction models from pooled-year data have low predictive powers. It appears that the results of the pooled year analysis were affected by the prediction year 1998. In answer to this question, the regression analysis of each model, earnings, cash flows, and cash flows and accrual components of earnings models, was processed for pooled-year data in order to predict the cash flow of years 1996 to 2002, omitting a prediction for cash flow year 1998. The prediction models from pooled-year data, without prediction year 1998, were expected to have better predictive powers. The results of the analysis for each model are reported in Table 5.22 and the comparisons of adjusted R2 between models for all prediction years and the models for prediction years, excluding prediction year 1998, are presented in Table 5.23.

Table 5.22 presents the results of the pooled-year data analysis, excluding prediction year 1998, and Table 5.23 shows the difference of adjusted R2 shown in analysis of all prediction years, excluding prediction year 1998. For the earnings model, in using one-year lag of earnings, the relationship between the correlation coefficients was significantly different from zero, F (1,1227) = 108.20, p < .005. EARNt-1 was found to explain 8% of the variation in CFOt. When compared with the analysis which includes all prediction years, the explanatory power increased by about 5% (difference of adjusted R2 = 0.05). Similarly, in using a two-year lag for prediction, it was shown that the correlation coefficient was significantly different from zero, F (2,1144) = 150.84, p < .0005. EARNt-1 and EARNt-2 were found to explain 21% of the variation in CFOt. The explanatory power increased by about 18% (difference of adjusted R2 = 0.18). Additionally, both EARNt-1 and EARNt-2 were found to contribute positively and significantly to prediction of CFOt, with coefficients of 0.50 and 0.36

163

respectively. In using three-year lags, the F-statistic showed that the model is significantly different from zero, F (3,922) = 68.91, p< 0.005. EARNt-1, EARNt-2 and EARN t-3 were found to explain 18% of the variation in CFOt and the explanatory power increased by about 9% (difference of adjusted R2 = 0.09). EARNt-1, (t = 3.83, p < 0.005), EARNt-2, (t = 10.66, p < 0.005) and EARN
t-3,

(t = 13.44, p < 0.005) were found to contribute positively and

significantly to predictions of CFOt with coefficient of 0.12, 0.38 and 0.47 respectively.

For the cash flow model, in using one-year lag of cash flows, the correlation coefficient was significantly different from zero: F (1,1231) = 246.385, p < .0005. CFOt-1 was found to explain 16% of the variation in CFOt (adjusted R2 = 0.16). The explanatory power slightly increased to 2% (difference of adjusted R2 = 0.02). In using two-year lags for prediction, the model is statistically significant different from zero (F (2,1144) = 161.23, p < .0005). CFOt-1 and CFOt-2 were found to explain 22% of the variation in CFOt. However, the explanatory power decreased by 10% (difference of adjusted R2 = -0.10). Therefore, CFOt-1 and CFOt-2 are positive and significant to predictions of CFOt, with a coefficient of 0.36 and 0.24 respectively. In using three-year lags for prediction, the model is significantly different from zero: F (3,923) = 124.05, p< 0.005. CFOt-1, CFOt-2 and CFOt-3 were found to explain 29% in the variation in CFOt. The explanatory power slightly increased by 3% (difference of adjusted R2 = 0.03). CFOt-1 (t = 2.62, p < 0.005), CFOt-2 (t = 10.14, p < 0.005) and CFOt-3 (t = 10.35, p < 0.005) were found to contribute positively and significantly to predictions of CFOt.

For the cash flow and accrual components of earnings model, in using a one-year lag of cash flow and accrual components of earnings, the multiple correlation coefficients were found to be significant predictors of CFOt, F (2,1227) = 272.57, p < .005. CFOt-1 and ACCt-1 were found to explain 31% in the variation in CFOt. The explanatory power increased about two times (difference of adjusted R2 = 0.17). Both CFOt-1 (t = 22.18, p < 0.005), and ACC t-1 (t = 6.72, p < 0.005) were found to contribute positively and significantly to predictions of CFOt. In using two-year lags for prediction, a significant relationship was found between the pair of independent and dependent variables (F [4,1140] = 243.77, p < .0005). CFOt-1, CFOt-2, ACCt-1 and ACCt-2 were found to explain 46% in the variation in CFOt. The explanatory power increased by 10% (difference of adjusted R2 = 0.10). All year lags of CFO and ACC were

164

found to contribute positively and significantly to predictions of CFOt. In using three-year lags, a significant relationship was found between the pair of independent and dependent variables (F [6,918] = 137.68, p< 0.005). CFOt-1, CFOt-2, CFOt-3, ACCt-1, ACCt-2 and ACCt-3 were found to explain 46% in the variation in CFOt. The explanatory power slightly increased by about 1% (difference of adjusted R2 = 0.01). All independent variables were found to contribute positively and significantly to predictions of CFOt.

Overall, these results indicate that the predictive powers of earnings, cash flows and cash flows and accrual components of earnings models for analysis improved when excluding prediction year 1998. This may be evidence that the pooled-year analysis in Section 5.5 was influenced by the Asian economic crisis during the years 1997 and 1998.

165

Table 5.22 Summary Statistics from Regressions of Future Cash Flows on Predictors (Pooled-year Data

One lag Predictor n EARNt-1 EARNt-2 EARNt-3 Beta t adj. sig. R

2

Two lags adj. F d.f. sig. n Beta t sig. R

2

Three lags adj.

Excluding Prediction Year 1998)

d.f.

sig.

Beta

sig.

R2

d.f. Sig.

1229 0.29 10.40 0.00 0.08 108.20 1,1227 0.00 1147 0.50 16.78 0.00 0.21 150.84 2,1144 0.00 926 0.12 3.83 0.00 0.18 68.91 3,922 0.00 0.36 12.00 0.00 0.38 10.66 0.00 0.47 13.44 0.00

CFOt-1 CFOt-2 CFOt-3

1233 0.40 15.47 0.00 0.16 239.31 1,1231 0.00 1147 0.36 13.47 0.00 0.22 161.23 2,1144 0.00 927 0.09 2.62 0.00 0.29 124.05 3,923 0.00 0.24 8.84 0.00 0.29 10.14 0.00 0.34 10.35 0.00

CFOt-1 ACCRt-1 CFOt-2 ACCRt-2 CFOt-3 ACCRt-3

1230 0.64 22.18 0.00 0.31 272.57 2,1227 0.00 1145 0.63 19.56 0.00 0.46 243.77 4,1140 0.00 925 0.11 3.84 0.00 0.47 137.68 6,918 0.00 0.20 6.72 0.00 0.15 5.42 0.00 0.25 8.48 0.00 0.51 16.8 0.00 0.36 11.12 0.00 0.63 16.3 0.00 0.12 3.92 0.00

0.44 18.59 0.00 0.36 12.88 0.00

166

Table 5.23 Comparison of Adjusted R2 between Analysis of All Prediction Years and Analysis Excluding Prediction Year 1998 Model Adjusted R2 all prediction years 0ne Two Three 0.03 0.03 0.09 0.14 0.32 0.26 0.14 0.36 0.45 Adjusted R2 excluding prediction year 1998 One Two Three 0.08 0.21 0.18 0.16 0.22 0.29 0.31 0.46 0.46 Difference One 0.05 0.02 0.17 Two 0.18 -0.10 0.10 Three 0.09 0.03 0.01

EARN CFO CF and ACR

5.10 Conclusion

This chapter discussed the data analysis and interpreted the results of the analysis to solve the research questions and hypotheses tests of this study. Three main regression models were built for predicting future cash flows. The first model was the earnings model, the second was the cash flows model and the third model was the cash flows and accrual components of earnings model. At first, the description of the variables and the correlation of a pair of variables in the models were examined by statistical techniques. The descriptive statistics show that the values of the variables ranged by year, and the value of earnings and cash flows, sharply decreased in year 1997 which was during the Asian economic crisis. Overall, the significant correlation between the dependent variable and independent variables was high. This suggested that the regression procedure could be used for testing these models. However there was no correlation of a pair of some variables for the years 1997 and 1998. This could make the model insignificant for predicting future cash flows.

These three main models: earnings, cash flow, and cash flows and accrual components of earnings, were tested by both yearly and pooled-year analysis and by the different year lags of independent variables, one-, two-, and three-year lags. The results of the analysis led to the acceptance of the first three hypotheses. Overall, the statistical results of the data analysis suggest that the three models are significant for predicting future cash flows of Thai listed companies. Moreover, it has been seen that the additional lags of independent variables in each model improved their explanatory power.

167

The explanatory power of each model changed according to the year to which the model was applied. The explanatory power of the three models was highest in prediction year 2002. Additionally, when comparing the adjusted R2 among the three models, the cash flows and accrual components of earnings model had the highest explanatory power in predicting future cash flows.

When the predictive ability of each model was measured by the two approaches: the correlation and mean absolute percentage errors in the out-of-sample period, the consistent results of both yearly and pooled-year analysis provide evidence that the earnings model had the lowest predictive power for predicting future cash flows and the cash flows model is a better predictor than the other models.

In the stepwise analysis for the cash flow ratios model, the stepwise procedure produces different cash flow ratio models for each prediction year and the adjusted R2 value of the model for each prediction year is low. This indicates that the cash flow ratios model is not a good predictor in predicting future cash flows of Thai listed companies.

Under additional analysis, the results indicate that in regard to the cash flows and accrual components of earnings model, the model which used disaggregated accrual components of earnings gave greater explanatory power than the model which aggregated the accrual component of earnings. Moreover, the results of the three models in the pooled-year analysis which excluded data of prediction year 1998, suggested that the pooled-year analysis of the three models without prediction year 1998 had a higher explanatory power than the pooledyear analysis which included all the prediction years. This may be evidence that the prediction year 1998, during the economic crisis, was the reason for the low explanatory power of the models generated from pooled-year data.

The next chapter will conclude the research findings and discuss the implications for accounting standards, policy and users of financial statements.

168

Chapter 6
Conclusions and Implications
6.1 Introduction The purpose of this research is to investigate the ability of cash flow and accrual accounting data in predicting future cash flows in Thai listed companies. The research problem is, How can cash flows and accrual accounting data be used to predict future cash flows of Thai listed companies? The problem was justified by the identified research gaps and the limitations of previous research into the usefulness of cash flows statements of Thai firms.

Chapter 1 provided the background to the research which determined the research problem and briefly explained the research design. Chapter 2 reviewed the literature of Thailand relevant to the sample of the research, including a background to the country, its economic condition, the Thai stock market and Thai accounting standards. Chapter 3 reviewed the literature relevant to the research problem. Two parent disciplines, cash flows and accrual accounting bases, were discussed to demonstrate the concept of reporting accounting data under the two assumptions, accrual and cash flow bases. Prior empirical research in predicting future cash flows was reviewed in order to find the research gaps. In the final section of Chapter 3, the research framework adopted for this thesis was developed. Chapter 4 presented the justification for the adoption of the positivist paradigm. The research design, including the measurement of variables, model building and sample selection, was explained in Chapter 4. Furthermore, the statistical techniques used to analyse the data and test the hypotheses were introduced. Chapter 5 provided the results of the data analysis and the statistical results which led to accepting or rejecting the research hypotheses. Chapter 6 provides the conclusions and implications of the research findings. The objectives of the final chapter are to summarise and discuss the research findings regarding the research hypotheses, to suggest how the outcomes of the research study can be implemented by users of financial statements and the further research in accounting and financial areas which should be conducted in order to extend this study. This final chapter comprises of six sections. Section 6.1 introduces the structure of the chapter. Section 6.2 summarises the research 169

findings related to the research hypotheses and the model testing. Sections 6.3 and 6.4 indicate the implications of the research outcomes for theory, policy and practice, involving all parties which use financial statements. Section 6.5 discusses the limitations of this research. Section 6.6 suggests further research to extend this research in the financial and accounting areas. Section 6.7 presents the conclusions of the thesis. The outline of this chapter is depicted in Figure 6.1.

Figure 6.1 Structure of Chapter 6: Conclusions and Implications

6.1 Introduction

6.2 Conclusion of Research Finding

6.3 Implications for Theory

6.4 Implications for Policy and Practice

6.5 Limitations of the Research Study

6.6 Suggestions for Future Research

6.7 Conclusion

170

6.2 Conclusion of the Research Findings

This research constructed five research questions and five research hypotheses with respect to the research problem. Four main models for predicting future cash flows were formulated using regression analysis. The conclusions of the research findings relating to the research hypotheses and questions are discussed below.

Research hypothesis 1 which states the relationship between past earnings and future cash flows of Thai listed companies is stated below. Hypothesis 1: Past earnings have significant predictive power in predicting future cash flows of Thai listed companies.

The earnings model (Model 1) was constructed to test this hypothesis, and regression analysis was conducted to examine the model. The statistical results of the earnings model were used to test the hypotheses. The regression analysis investigated the ability of year-lags of earnings in explaining future cash flows. The earnings model was examined in three ways. Firstly, the model contains one-year lag of earnings, secondly, the model contains two-year lags of earnings, and third, the model contains three-year lags of earnings. In addition, the model was tested in yearly and pooled-year data. The results of the regression analysis for the earnings model are as follows.

Significance of the model

The results of the yearly analysis (Table 5.11) show that the model using one-year lag of earnings was found to be a significant predictor of future cash flows in all prediction years from 1997 to 2002 except for prediction year 2000 (p < 0.005). Additionally, both two- and three-year lags of the earnings models were found to be significant predictors of future cash flows (p < 0.005). Moreover, the results of the pooled-year analysis (Table 5.14) indicate that the earnings model using two- and three-year lags of earnings is a significant predictor of

171

future cash flows (p < 0.005). Overall, this finding suggests that earnings have a significant predictive power for future cash flows of Thai listed companies. Goodness of fit of the model (adjusted R2) When considering the adjusted R2 of the earnings model (Table 5.11), for yearly analysis, the adjusted R2 for each prediction year was found to be unstable. The adjusted R2 was low in the prediction years before 1998 and high in the prediction years after 2000. This means that the explanatory power of the earnings model in predicting future cash flow was variable over the prediction years. In addition, when comparing the adjusted R2 among one-, two- and threeyear lags of earnings, two- and three-year lags of earnings have higher explanatory power than one-year lag of earnings and the three-year lags produce the highest explanatory power. For pooled-year analysis (Table 5.14), the results show that the adjusted R2 was very small. This means that the explanatory power of the model derived from pooled-year data is very low; lower than that of the model generated from yearly-data. Moreover, in comparison with the explanatory power among the different year-lags of earnings, the two- and three-year lags of earnings provided a higher adjusted R2 value than the one-year lag and the three-year lags provide the highest adjusted R2 value. These results are similar to the results of the yearly analysis. This indicates that the addition of year-lags of earnings can improve the explanatory power of the model.

Overall, apart from earnings in years 1997, 1998 and 1999, the year-lags of earnings were found to contribute significantly to explanations of future cash flows. However, for prediction years 1997 and 1998 some year-lags of earnings were not found to contribute significantly to future cash flows. This may show that normally, a year-lag of earnings can explain future cash flows, but it may lose explanatory ability when the year-lag was in the year during the economic crisis and also when the prediction year was in the economic crisis. This finding provides a reason why analysts could not use earnings as a sign of companies financial status before the economic crisis occurred.

172

In summary, the earnings model can be used to predict future cash flows of Thai listed companies and the explanatory ability of the model in predicting future cash flows differs across years. Moreover, the additional year-lags of earnings can improve models explanatory power. However, the model may lose its explanatory power when the prediction contains a year-lag of earnings in the year during the Asian economic crisis or the prediction is of the year during the economic crisis. As a result, hypothesis 1 is accepted. That is, earnings have significant predictive power in predicting future cash flows of Thai listed companies.

This research finding is consistent with previous research conducted in developed countries. Many previous researchers investigated the ability of earnings to predict future cash flows and found that earnings are a significant predictor of future cash flows (Barth, Cram & Nelson 2002; Bowen, Burgstahler & Daley 1986; Dechow, Kothari & Watts 1998; Espahbodi 1988; Finger 1994; Greenberg, Johnson & Ramesh 1986; Jordan & Waldron 2001; Murdoch & Krause 1990; Percy & Stokes 1992; Stammerjohan & Nassiripour 2000/2001). Hypothesis 2: Cash flows have significant predictive power in predicting future cash flows of Thai listed companies.

The cash flows model (Model 2) was constructed to test the second hypothesis, and regression analysis was conducted to examine the model. The statistical results of the cash flows model (Model 2) were used to test the hypothesis. The regression analysis investigated the ability of year-lags of cash flows to explain future cash flows. Using the same method as for the earnings model, the cash flows model was examined in three ways with one-, two- and threeyear lags of cash flows. Additionally, the model was tested with yearly and pooled-year data. The results of the regression analysis for the cash flows model are summarised as follows.

173

Significance of the model

The statistical results of the cash flows model (Table 5.12) indicate that when using one- year lag of cash flows in predicting future cash flows, the year-lag was found to be significant in explaining future cash flows for all prediction years (p < .005) except for prediction year 1999, for which the predictor was the cash flows of year 1998. For using two- and three-year lags of cash flows, the models were found to be significant in explaining future cash flows for all prediction years (p < .005). Additionally, for pooled-year analysis, the statistical results (in table 5.14) also show that the cash flows model with all year-lags was found to be significant in explaining future cash flows (p <.005). This shows that cash flows are a significant predictor in predicting future cash flows. However, the model may be insignificant when containing a year-lag of cash flows derived from the year of the Asian economic crisis. Goodness of fit of the model (adjusted R2) For the yearly analysis, the adjusted R2 of the model was unstable. Excluding the prediction year 1999, the adjusted R2 was low in the prediction years before 1999 and high in the prediction years after 1999. This shows that the explanatory power of the cash flows model was changeable over the prediction years. The explanatory power of the model for the prediction years after year 1999 was higher than the model for the prediction years before 1999. This may imply that the explanatory power of the model improves after 1999 which is the year after the amendment and issuing of new Thai accounting standards. For the pooledyear analysis, the adjusted R2 of the model was lower than that of the model for the yearly analysis. Additionally, for both the yearly and pooled-year analyses, the model containing two- and three-year lags of cash flows provided a higher adjusted R2, than the model containing only one-year lag of cash flows, and the model containing three-year lags of cash flows provided the highest adjusted R2 value. This shows that the additional lags of cash flows can improve the explanatory power of the model.

When the relative contribution to prediction of the year lags of cash flows was evaluated (Table 5.12), it was found that overall the year-lags one, two and three of cash flows

174

contribute significantly to the prediction. However, the year-lags of cash flows from the year 1997 and 1998 did not contribute significantly to the prediction. Moreover, the year lags of cash flows did not contribute significantly to the prediction of cash flow years 1997 and 1998. This implies that a year-lag of cash flows cannot contribute significantly to the prediction when it was from the year during the Asian economic crisis, or, it was used to predict cash flows of the year during the Asian economic crisis. It is possible to conclude that the Asian economic crisis had an impact on the cash flow prediction. In addition, for the results of pooled-year analysis (Table 5.14), all year lags of cash flows contribute significantly to the prediction (p < .05) in all year-lag models. These results indicate that when considering individual contribution of each year-lag of cash flows to the prediction, usually each year lag of cash flows contributes significantly to the prediction.

These findings are consistent with previous research which investigated the ability of cash flows in predicting future cash flows in developed countries. Many researchers found that cash flows are an important predictor in predicting future cash flow. (Barth, Cram, and Nelson (2002), Bowen, Burgstahler, and Daley (1986), Dechow, Kothari, and Watts (1998), Espahbodi (1988), Finger (1994), Greenberg, Johnson and Ramesh (1986), Jordan and Waldron (2001), Murdoch and Krause (1990), Percy and Stokes (1992), Stammerjohan and Nassiripour (2000/2001)).

Hypothesis 3: Past cash flows and accrual components of earnings have significant predictive power in predicting future cash flows of Thai listed companies.

The cash flows and accrual components of earnings model was constructed to test the third hypothesis and multiple regression analysis was conducted to examine the model. The

regression analysis investigated the ability of year-lags of cash flows and accrual components of earnings in explaining future cash flows. Using the same process as was used for the earnings and cash flows models, the cash flows and accrual components of earnings model was examined in three ways. That is, in the first analysis, the model contains a one-year lag of cash flows and accrual components of earnings; in the second analysis, the model contains two-year lags of cash flows and accrual components of earnings, and the in third analysis, the

175

model contains three-year lags of cash flows and accrual components of earnings. The results of the regression analysis for the cash flows and accrual components of earnings model are summarised as follows.

Significance of the model

The statistical results of the cash flows and accrual components of earnings model in the yearly analysis (in Table 5.13) indicate that when using one-, two- and three-year lags of cash flows in predicting future cash flows, the model was found to be significant in explaining future cash flows for all prediction years (p< .005). Additionally, the results of the pooledyear analysis show that using one-, two- and three-year lags of cash flows and accrual components of earnings, the model was also found to be significant in explaining future cash flows (p < .005). Goodness of fit of the model (adjusted R2) For the yearly analysis (Table 5.13), the adjusted R2 of the model was unstable for each prediction year. The adjusted R2 was low in the prediction year before 1999 and high in prediction year after 1999. This shows that the explanatory power of the cash flows and accrual components of earning model differs across years. The explanatory power of the model in the prediction years after year 1999 was higher than the model in the prediction years before 1999. For pooled-year analysis, the adjusted R2 of cash flows and accrual components of earnings model was low, and lower than the model of yearly analysis. This implies that the model from pooled-year analysis has lower explanatory power than the model from yearly analysis. Moreover, when comparing the adjusted R2 among models using different year-lags of cash flows and accrual components for yearly analysis, overall, the twoand three-year lags of cash flows and accrual provided higher adjusted R2 than the one-year lag, and the three-year lag provided the highest adjusted R2 value. Similar to the yearly analysis, the three-year lag provided the highest adjusted R2 value, but the adjusted R2 between one- and two-year lags were not different. This implies that in general, the incremental year

176

lags of cash flows and accrual components of earnings improve the explanatory power of the model.

When the relative contribution to prediction of the year-lags of cash flows was evaluated (Table 5.13), it was found that overall the year-lags one and two of cash flows and accrual components of earnings contributed significantly to the prediction. However the year-lags of cash flows and accrual components of earnings from the year 1997, 1998, and 1999 did not contribute significantly to the prediction. Moreover, on average, the year lags of cash flows and accrual components of earnings did not contribute significantly to the prediction of cash flow years 1997 and 1998. This implies that a year-lag of cash flows and accrual components of earnings cannot contribute significantly to the prediction when it was from the year during the Asian economic crisis, or, it was used to predict cash flows of the year during the Asian economic crisis. It is possible to conclude that the Asian economic crisis had an impact on the cash flow prediction. In addition, a year-lag of accrual components of earnings did not contribute significantly to the prediction of cash flows in many years. Perhaps this means that the year-lag of accrual components of earnings did not contribute significantly to the prediction of cash flows. In addition, for the results of the pooled-year analysis (Table 5.14), all year-lags of cash flows contribute significantly to the prediction (p < .05) in models with all year lags, whereas, for the one-year lag model, the year-lag one of accrual components of earnings was not found to contribute significantly to prediction (p >.05) and the year lags two and three in two- and three-year lag analysis were found to contribute significantly to prediction. These results indicate that in general, each year-lag of cash flows contributes significantly to the prediction, but the year-lag of accrual components of earnings seem to not contribute significantly to prediction.

When compared with related prior research, these findings are found to be consistent with the results of prior studies. Barth, Cram, and Nelson (2001; 2002) and Stammerjohan and Nassiripour (2000/2001) examined the ability of cash flows and accrual components of earnings in predicting future cash flows. They found that the cash flows and accrual components of earnings models provided a good predictive power in predicting future cash

177

flows and found that adding year-lags of cash flows and accrual components of earnings into the model can improve the predictive ability of the model.

Hypothesis 4 : There are different predictive powers between three prediction models: earnings, cash flows and cash flows and accrual components of earnings models.

To test this hypothesis a comparison of predictive power among the three models, earnings, cash flows, and cash flows and accrual components of earnings models, was conducted, based on the following statistical results. Comparison of adjusted R2 between models When comparing the adjusted R2 between models within all year-lags for yearly and pooledyear analysis (Table 5.15), overall the results show that the cash flows and accrual components of earnings provided a higher adjusted R2 than the earnings and cash flows models. However the difference of adjusted R2 between the cash flows model and cash flows and accrual components of earnings model is small. These results indicate that the cash flows and cash flows and accrual components of earnings models provide better predictors than the earnings model, but the results cannot provide strong evidence that the cash flows and accrual components of earnings model provides a better predictor than the cash flows model.

Comparison of correlation between actual and predicted cash flows and of absolute percentage errors

This study also tested the ability of each model in an out-of-sample period. The correlation between actual and predicted cash flows and the mean absolute percentage errors (MAPE) were calculated by using out-of-sample data with the model providing a higher correlation and a smaller MAPE being a better model. Based on the consistent results of the correlation and MAPE (table 5.17 and 5.18), the majority of results from yearly and pooled- year analysis, show that the cash flows model outperforms other models and the earnings model has the lowest predictive power in predicting future cash flows. The results of the correlation and

178

mean absolute percentage errors can assist in deciding which model, between the cash flows and cash flows and accrual components of earnings model, is a better predictor in predicting future cash flows of Thai listed companies.

Consequently, the results led to accepting the fourth hypothesis. There are different predictive powers among the earnings, cash flows, and cash flows and accrual components of earnings models. In addition, the results provide evidence that the cash flows model is a better predictor than the other models, while the earnings model has the lowest predictive power in predicting future cash flows.

This finding is partly consistent with prior research. Some research compared the predictive ability between earnings and cash flows (Finger 1994; Krause & Murdoch 1990; Stammerjohan & Nassiripour 2000/2001). Finger (1994) and Stammerjohan & Nassiripour (2000/2001) concluded that cash flows are a better predictor of future cash flows than earnings, whereas Krause & Murdoch (1990) indicated that earnings are a better predictor than cash flows. The current research findings support the conclusion that cash flows are a better predictor than earnings. Some research has compared the predictive ability of earnings and cash flows and components of earnings models and found that a cash flows and accrual components of earnings model was a better predictor than an earnings model (Barth, Cram & Nelson 2002; Stammerjohan & Nassiripour 2000/2001). Additionally, Stammerjohan and Nassiripour (2000/2001) compared the ability of the cash flows model and the cash flows and accrual component of earnings model but did not provide strong evidence that the cash flows and accrual components of earnings outperform the cash flows model. This research, using the the two approaches, the correlation and mean absolute percentage errors, concludes that the cash flows model provides a better predictor than the cash flows and accrual components of earnings model.

179

Hypothesis 5 : Past cash flow ratios are significant predictors of future cash flows of Thai listed companies.

The cash flow ratios model was constructed to test this hypothesis by using the technique of stepwise regression. The results of the stepwise regression show that a suitable cash flow ratio model for each year contains different cash flow ratios. Additionally, the adjusted R2 value of the cash flow ratio model is low (Adjusted R2 < .50) and varies across prediction years. As a result, it indicates that the cash flow ratios model built from stepwise regression is not stationary and may be useless in practice. In summary, Research Hypothesis 5 is rejected. The cash flow ratio model cannot provide good predictive power for predicting future cash flows of Thai listed companies.

There is no previous study using cash flow ratios to predicting future cash flows. This study fills this gap in the literature by testing the use of cash flow ratios in predicting future cash flows. However, this study was unsuccessful in finding any new alternative predictor in predicting future cash flows.

In summary, to answer the research problem, this research found that cash flows and accrual accounting data can be used to predict future cash flows of Thai listed companies. A comparison of the predictive ability indicates that the cash flows model based on a cash basis from the cash flow statements under Thai accounting standard No. 25 was a better predictor than the earnings models based on accrual accounting, and cash flows and accrual components of earnings based on both cash and accrual accounting bases. Nevertheless, additional yearlags of predictors can improve the predictive ability of the model.

In addition, the results of the analysis suggest that the Asian economic crisis had an impact on the predictive power of the models. The prediction models may lose their explanatory power for the prediction year during the economic crisis or when the model contains year-lags of predictors in the years during the economic crisis. The results of pooled-year analysis excluded the prediction year 1998 of each prediction model (Section 5.9.2) support this

180

conclusion. It was found that the explanatory power of each model in predicting future cash flows improves when the model generated from pooled-year data excluded the prediction year 1998 in which the main impact of the Asian economic crisis was felt. In other words, eliminating the effect of the economic crisis can improve the predictive power of the models for future cash flows.

This conclusion corresponds with previous research conducted in Thailand which showed that the Asian economic crisis influenced the results of the research. Narktabtee (2000), for instance, studied the implications of accounting data in the Stock Exchange of Thailand and suggested that the accounting information had more information content when the economic situation was normal and lost information content when the Asian economic crisis occurred. Another sample, the study of the value relevance of accounting information during the financial crisis in Thailand (Graham, King & Bailes 2000) suggested that after the devaluation of the baht, the value relevance of accounting information declined.

6.3 Implications for Theory

The results of this research confirm that accounting information, earnings and cash flows is useful in predicting future cash flows of Thai listed companies. In addition, cash flows from operations on a cash basis are better predictors of future cash flows than earnings based on accrual accounting. The findings of this research can be used as evidence to support theory in financial analysis.

Usefulness of cash flows statement under Thai accounting standard No. 25. Thai Accounting Standard No. 25 (TAS 25), adopted from International Accounting Standard No. 7 (IAS 7), requiring a cash flow statement, suggests that the cash flow statement can provide information for predicting future cash flow (IASC 1995, p. 115). The findings of this research support this statement: that cash flow data on cash flow statements prepared under the TAS 25 can be used to predict a companys future cash flows.

181

Quality of accounting information of Thai listed companies. As has been found in the analysis in the holdout sample, the research findings suggest that the cash flow prediction model built from the original sample data can be used to predict future cash flows of the holdout sample data. This confirms the validity of models which use accounting information prepared by Thai listed companies. In addition, considering the explanatory power of the models, the prediction models provided higher predictive power for the prediction year after 1999. This may imply that the quality of accounting information under the new accounting standards revised and issued since 1999 have improved.

6.4 Implications for Policy and Practice

In addition, knowing that past cash flow can be used to predict future cash flows of Thai listed companies provides some implications for practice and policy relating to Thai accounting standards and the capital market.

Stock Exchange of Thailand (SET) policy setters have a role in announcing policy or rules to benefit investors and all market participants. The SET policy setters can use the findings of this research to regulate policy in financial information disclosures of Thai listed companies through the information management system. That is, the SET should support the consideration of other performance measures, such as cash flows from operations, as a complementary indicator to manage investment. At the same time, Accounting Standard setters or the Institute of Certified Accountants and Auditors of Thailand (ICAAT) should continually develop Thai accounting standards to promote quality and reliable accounting information.

Moreover, the research findings are meaningful to financial analysts. Creditors can employ the prediction model to determine their customers ability to pay interest and repay amounts borrowed. Investors can access future cash flows of companies they are interested in for estimating future return, by using the prediction model. Other related parties, such as company managers can apply the findings of this research to decision making. For example, the prediction model can be applied to forecast future cash flows of a reinvestment project.

182

6.5 Limitations of the Research Study

The results of this research may be partially influenced by some limitations. With time constraints and nonavailability of data sources, the current research is limited by the following issues.

Sample group. The findings arising out of the research are confined to Thai listed companies in the nonfinancial sector. The sample of this research excluded Thai listed companies in the financial sector such as commercial banks, finance and insurance companies. Consequently, the findings may not apply to the companies in the financial sector. Moreover, companies under rehabilitation were excluded from the sample, therefore the outcomes of this research cannot be used for companies under rehabilitation.

Time period. The time period of this study from the year 1994 to 2002, involves many situations. Firstly, Thai listed companies have been regulated to disclose cash flow statements since 1994. Secondly, the Asian economic crisis occurred during 1997 and 1998, and thirdly, a number of Thai accounting standards were revised and issued in 1999. In addition, according to the economic index, the economy of Thailand has recovered since 2000 after suffering from the crisis. It is possible that these situations caused the instability of the predictive power of the models for each prediction year. As a result, this research cannot produce efficient estimated equations suitable for every prediction year in relation to Thai listed companies.

Data analysis. This study assumed that the cross-sectional regression model fits across firms from different industries. That is, an explanatory variable in each prediction model would have the same effect on future cash flows across industries. It is possible that the relationship between an explanatory variable and future cash flows is not stable across industries, which may have caused errors in the prediction.

183

6.6 Suggestions for Future Research

Further research can extend this study by replicating the methodology to investigate data of listed companies in the financial sector. Moreover, the data sample can be separated and analysed by industry. In this way, a contribution could be made to developing a more industry-specific theory.

In addition, this research generates prediction models by using a cross-sectional regression model which can be applied to general firms in the Thai stock market. Perhaps the prediction model would prove more accurate if a firm-specific model is used, with data of each firm tested separately. Nevertheless, future research has an opportunity to test longer time periods for the data, which could provide an appropriate prediction model for Thai listed companies.

The results of this research rely solely on a secondary data method. The findings of this research would be given further credibility by conducting survey research to collect data directly from the users of financial statements or related parties. Further research may provide evidence in practice, and users of financial statements could use cash flow predictors in predicting future cash flows.

This research used the technique of regression analysis to estimate the equation for predicting future cash flows. It may be possible that the estimation of the regression model did not match the cash flow data. In other words, a better model may be derived from a different method, such as the Box-Jenkins time-series model (Box & Jenkins 1976). However, that model needs many time-series of data. Future research could be conducted when cash flow data is adequate. Future studies may use the technique of the Box-Jenkins methodology to generate a prediction model and then compare it with the regression model.

It can be seen that the economic condition had an impact on the results of the current research. Extended research could investigate the effect of the economic condition by considering economic indices such as the exchange rate, inflation rate and growth rate, thereby providing a clearer picture of how different economic conditions affect the accounting information to

184

predict future cash flows. Moreover, it is possible that non-financial information is significant in predicting future cash flows. Further research may include non-financial data such as a companys financial policy, government aids and other economic factors in prediction models.

6.7 Conclusion

This final chapter concludes the research findings and implications. In summary, earnings and cash flows can be used as predictors of future cash flows of Thai listed companies and cash flows are a better predictor than earnings. However, the findings suggest that accrual components of earnings added to the cash flows model cannot provide an incremental predictive power of the prediction model. The findings are meaningful to many parties including policy and practice. However, the current research has some limitations and needs to be extended by future research.

185

References
Ali, A. 1994, 'The incremental information content of earnings, working capital from operations, and cash flows', Journal of Accounting Research, vol. 32, no. 1, pp. 61-74. Arnold, W. 2002, 'World Business Briefing Asia: Thailand: Economic Growth', The New York Times, 19 March, p. 1. Ashton, R.H. 1974, 'The predictive-ability criterion and user prediction models', The Accounting Review, vol. 49, no. 4, pp. 719-32. Ball, R. & Brown, P. 1968, 'An empirical evaluation of accounting income numbers', Journal of Accounting Research , vol. 6, no. 2, pp. 159-77. Ball, R., Brown, P., Finn, F.J. & Officer, R.R. 1989, Share Markets and Portfolio Theory: Readings and Australian Evidence, 2nd edn, University of Queensland Press, St. Lucia, Qld. Bamber, L.S. & Cheon, Y.S. 1995, 'Differential price and volume reactions to accounting earnings announcements', The Accounting Review, vol. 70, no. 3, pp. 417-41. Bank of Thailand 2002, Thailand's Economic and Monetary Conditions in 2001, Bank of Thailand, viewed 26 July 2003, <http://www.bot.or.th/BOTHomepage/DataBank/Econcond/econreport/annual/5-16-2002Eng-i-1/Total_2001_eng.pdf>. ---- 2003a, Thailand's Economic and Monetary Conditions in 2002 , Bank of Thailand, viewed 26 November 2003, <http://www.bot.or.th/BOTHomepage/DataBank/Econcond/econreport/annual/5-8-2003-Engi/Full.pdf>. ---- 2003b, Thailand's Economic and Monetary Conditions in 2000 , Bank of Thailand, viewed 26 July 2003, <http://www.bot.or.th/BOTHomepage/DataBank/Econcond/econreport/annual/3-26-2001Eng-i/total_2.pdf>. ---- 2003c, Annual Economic Report 1999, Bank of Thailand, viewed 26 July 2003, <http://www.bot.or.th/BOTHomepage/DataBank/Econcond/econreport/annual/8-11-2000-Thi/annual_eng.pdf>. ---- 2003d, Annual Economic Report 1998, Bank of Thailand, viewed 26 July 2003, <http://www.bot.or.th/BOTHomepage/DataBank/Econcond/econreport/annual/8-4-2003-Engi/Annual-98.pdf>. Barth, M.E., Beaver, W.H., Hand, J.R.M. & Landsman, W.R. 1999, 'Accruals, cash flows, and equity values', Review of Accounting Studies, vol. 3, no. 3-4, pp. 205-29.

186

Barth, M.E., Cram, D.P. & Nelson, K.K. 2001, 'Accruals and prediction of future cash flows', The Accounting Review, vol. 76, no. 1, pp. 27-58. ---- 2002, 'Accruals and the prediction of future cash flows', Finance India, vol. 16, no. 2, pp. 627-55. Bartov, E., Goldberg, S.R. & Kim, M. 2001, 'The valuation-relevance of earnings and cash flows: an international perspective', Journal of International Financial Management and Accounting, vol. 12, no. 2, pp. 103-32. Bernard, V.L. & Stober, T.L. 1989, 'The nature and amount of information in cash flows and accruals', The Accounting Review, vol. LXIV, no. 4, pp. 624-52. Bernstein, L.A. 1993, Financial Statement Analysis: Theory, Application, and Interpretation, 5th edn, Richard D. Irwin, Inc., USA. Berry, W.D. & Feldman, S. 1985, Multiple Regression in Practice, Quantitative Applications in the Social Sciences, SAGE Publications, Newbury Park. Bierman, H. 1988, 'Extending the usefulness of accrual accounting', Accounting Horizons, vol. 2, no. 3, pp. 10-4. ---- 1992, 'In defense of accounting information', Financial Analysts Journal, vol. 48, no. 3, pp. 81-3. ---- 1994, 'Accounting for valuation and evaluation', Journal of Portfolio Management, vol. 20, no. 3, pp. 64-7. Blake, J., Bond, R., Amat, O. & Oliveras, E. 2000, 'The ethics of creative accounting some Spanish evidence', Business Ethics, vol. 9, no. 3, pp. 136-42. Board, J.L.G. & Day, J.F.S. 1989, 'The information content of cash flow figures', Accounting and Business Research, vol. 20, no. 77, pp. 3-11. Bowen, R.M., Burgstahler, D. & Daley, L.A. 1986, 'Evidence on the relationships between earnings and various measures of cash flow', The Accounting Review, vol. LXI, no. 4, pp. 713-25. Box, G.E.P. & Jenkins, G.M. 1976, Time Series Analysis: Forecasting and Control, Rev. edn, Holden-Day, San Francisco. Boyd, T. & Cortese-Danile, T.M. 2000/2001, 'Using the cash flow statement to improve credit analysis', Commercial Lending Review, vol. 16, no. 1, pp. 55-9. Brigham, E.F. & Gapenski, L.C. 1999, Financial Management: Theory and Practice, 9th edn, The Dryden Press, Fort Worth, TX.

187

Brown, P. 1994, Capital Markets- Based Research in Accounting: An Introduction, Coopers & Lybrand, Melbourne. Calabrese, J. & Rafferty, B.A. 2003, 'EBITDA: What your borrower is measuring and how it affects cash', Commercial Lending Review, vol. 18, no. 3, pp. 41-4. Carslaw, C. A. & Mills, J. R. 1991, 'Developing ratios for effective cash flow statement analysis', Journal of Accountancy, vol. 172, no. 5, pp. 63-8. ---- 1993, 'Including the cash flow statement in financial analysis', Executive Accountant, vol. 20, no. 3, pp. 14-6. Chang, J. 2002, 'Investors sharpen focus on free cash flow', Chemical Market Reporter, July 22-July 29. Charitou, A. 2000, 'The impact of losses and cash flows on dividends: Evidence for Japan', ABACUS, vol. 36, no. 2, pp. 198-225. Charitou, A. & Ketz, E. 1991, 'An empirical examination of cash flow measures', Abacus, vol. 27, no. 1, pp. 51-64. Cheng, A., Liu, C.S. & Schaefer, T.F. 1997a, 'The value-relevance of SFAS No. 95 cash flows from operations as assessed by security market effects', Accounting Horizons, vol. 11, no. 3, pp. 1-15. ---- 1997b, 'Accounting accruals and the incremental information content of earnings and cash flows from operations', Advances in Accounting, vol. 15, pp. 101-23. Cheng, A., Yang, S. & Clubb, C. 2003, 'The incremental information content of earnings and cash flows from operations affected by their extremity', Journal of Business Finance & Accounting , vol. 30, no. 1/2, pp. 73-110. Cheung, J.K., Krishnan, G.V. & Min, C. 1997, 'Does interperiod income tax allocation enhance prediction of cash flows?' Accounting Horizons, vol. 11, no. 4, pp. 1-15. Climo, T.A. 1976, 'Cash flow statements for investors', Journal of Business Finance & Accounting, vol. 3, no. 3, pp. 3-16. Collis, J. & Hussey, R. 2003, Business Research, 2nd edn, Palgrave Macmillan, New York. Cooper, D.R. & Emory, W. 1995, Business Research Methods, 5th edn, The Irwin series in statistics, Irwin, Chicago. Dambolena, I.G. & Shulman, J.M. 1988, 'A primary rule for detecting bankruptcy: Watch the cash', Financial Analysts Journal, vol. 44, no. 5, pp. 74-8.

188

Dechow, P.M. 1994, 'Accounting earning and cash flows as measures of firm performance: The role of accounting accruals', Journal of Accounting and Economics, vol. 18, no. 1, pp. 3-42. Dechow, P.M., Kothari, S.P. & Watts, R.L. 1998, 'The relation between earning and cash flows', Journal of Accounting and Economics, vol. 25, no. 2, pp. 133-68. Dennis, M.C. 1994, 'Understanding cash flow statements', Business Credit, vol. 96, no. 1, pp. 40-2. Dielman, T.E. 1991, Applied Regression Analysis for Business and Economics, PWS-KENT Publishing Company, Boston. Easterby, M., Thorpe, R. & Lowe, A. 1991, 'The Philosophy of Research Design', in Management Research an Introduction , Sage Publications, London. EIU 2004, Country Profile 2004: Thailand, The Economist Intelligence Unit Limited (EIU), London. Epstein, M.J. & Pava, M.L. 1992, 'How useful is the statement of cash flows?' Management Accounting, vol. 74, no. 1, pp. 52-5. Espahbodi, H. 1988, 'Predictors of cash flows', The Journal of Business Forecasting Methods & Systems, vol. 7, no. 3, pp. 8-10. FASB 1978, Statement of Financial Accounting Concepts, Financial Accounting Standards Board (FASB), Stamford. Figlewicz, R.E. & Zeller, T.L. 1991, 'An analysis of performance, liquidity, coverage, and capital ratios from the statement of cash flows', Akron Business and Economic Review, vol. 22, no. 1, pp. 64-81. Finger, C.A. 1994, 'The ability of earnings to predict future earnings and cash flow', Journal of Accounting Research, vol. 32, no. 2, pp. 210-23. Fleming, J.M. 2004, 'Analytical procedures are a helpful tool', Pennsylvania CPA Journal, vol. 74, no. 4, pp. 3-4. Foster, G. 1986, Financial Statement Analysis, Prentice-Hall, New Jersey. Frigo, M.L. & Graziano, R. 2003, 'Strategic decisions and cash flow', Strategic Finance, vol. 85, no. 1, pp. 8-11. Gallinger, G.W. 2000, 'A framework for financial statement analysis part 4: Statement of cash flows', Business Credit, vol. 102, no. 5, pp. 32-7.

189

Giaccotto, C. 1990, 'Cash flow modeling and forecasting in capital budgeting under uncertainty', Decision Sciences, vol. 21, no. 4, pp. 825-41. Giacomino, D.E. & Mielke, D.E. 1987, 'Preparation and use of cash flow statements [2]', The CPA Journal, vol. 57, no. 3, pp. 30-5. ---- 1988, 'Using the statement of cash flows to analyze corporate performance', Management Accounting, vol. 69, no. 11, pp. 54-7. ---- 1993, 'Cash flows: another approach to ratio analysis', Journal of Accountancy, vol. 175, no. 3, pp. 55-8. Glassman, J. 2001, 'Economic crisis in Asia: The case of Thailand', Economic Geography, vol. 77, no. 2, pp. 122-47. Godfrey, J.M., Hodgson, A. & Holmes, S 2003, Accounting Theory, 5th edn, John Wiley & Sons Australia, Milton, Queensland. Gombola, M.J. & Ketz, J.E. 1983a, 'A note on cash flow and classification patterns of financial ratios', The Accounting Review, vol. 58, no. 1, pp. 105-14. ---- 1983b, 'Financial ratio patterns in retail and manufacturing organizations', Financial Management, vol. 12, no. 2, pp. 45-56. Gombola, M.J., Haskins, M.E., Ketz, J.E. & Williams, D.D. 1987, 'Cash flow in bankruptcy prediction', Financial Management, vol. 16, no. 4, pp. 55-65. Graham, R., King, R. & Bailes, J. 2000, 'The value relevance of accounting information during a financial crisis: Thailand and the 1997 decline in the value of the Baht', Journal of International Financial Management and Accounting, vol. 11, no. 2, pp. 84-107. Greenberg, R.R., Johnson, G.L. & Ramesh, K. 1986, 'Earnings versus cash flows as a predictor of future cash flow measures', Journal of Accounting Auditing and Finance, vol. 1, no. 4, pp. 266-77. Grossman, S.D. & Pearl, D. 1988, 'Financial analysis of the statement of cash flows', Ohi CPA Journal, vol. 47, no. 3, pp. 11-4. Guba, E.G. & Lincoln, Y.S. 1994, 'Competing paradigms in qualitative research', in NK Denzin & YS Lincoln (eds), Handbook of Qualitative Research, SAGE Publications, Thousand Oaks, pp. 105-17. Hackel, K.S., Livnat, J. & Rai, A. 1994, 'The free cash flow/small-cap anomaly', Financial Analysts Journal, vol. 50, no. 5, pp. 33-42.

190

Hair, J.F. 1995, Multivariate Data Analysis with Readings, 4th edn, Prentice Hall, Englewood Cliffs, New Jersey. Hair, J.F., Anderson, R.E., Tatham, R.L. & Black, W.C. 1998, Multivariate Data Analysis, 5th edn, Prentice Hall, New Jersey. Henderson, S. & Peirson, G. 1994, Issues in Financial Accounting, Longman, Australia. ---- 2000, Issues in Financial Accounting , 9th edn, Pearson Education Australia, NSW. Hill, R., Benston, G.J. & Hartgraves, A. 2003, 'Measuring corporate performance: The perils of turning to cash flow', Public Utilities Fortnightly, vol. 141, no. 10, pp. 20-4. Hodgson, A. & Stevenson-Clarke, P. 2000, 'Accounting variables and stock returns: The impact of leverage', Pacific Accounting Review, vol. 12, no. 2, pp. 37-64. Hoggett, J., Edwards, L. & Medlin, J. 2003, Accounting in Australia, 5th edn, John Wiley& Suns Australia Ltd., Queensland. IASC 1995, International Accounting Standards 1995 , International Accounting Standards Committee, London. ---- 2000, International Accounting Standards Explained, John Wiley & Sons, New York. ICAAT 1999a, Institute of Certified Accountants and Auditors of Thailand , Institute of Certified Accountants and Auditors of Thailand (ICAAT), viewed 23 June 2003, <http://www.icaat.or.th/about/main_sub1.htm>. ---- 1999b, The Progress in Developing Accounting in Thailand, Institute of Certified Accountants and Auditors of Thailand, viewed 25 June 2003, <http://www.icaat.or.th/about/index.html>. Ingram, R.W. & Lee, T.A. 1997, 'Information provided by accrual and cash-flow measures of operating activities', Abacus, vol. 33, no. 2, pp. 168-85. Jones, S. 1998, 'An evaluation of user ratings of cash vs accrual based financial reports in Australia', Managerial Finance, vol. 24, no. 11, pp. 16-28. Jones, S. & Widjaja, L. 1998, 'The decision relevance of cash-flow information: A note', Abacus, vol. 34, no. 2, pp. 204-19. Jordan, C.E. & Waldron, M.A. 2001, 'Predicting cash flow from operations: Evidence on the comparative abilities for a continuum of measures', Journal of Applied Business Research, vol. 17, no. 3, pp. 87-94.

191

Kelly, J. & O'Connor, J. 1997, 'Is profit more important than cash flow?', Management Accounting, vol. 75, no. 6, pp. 28-9. Khumawala, S.B., Polhemus, N.W. & Liao, W.M. 1981, 'The predictability of quarterly cash flows', Journal of Business Finance & Accounting, vol. 8, no. 4, pp. 493-510. Kimmel, P.D., Weygandt, J.J. & Kieso, D.E. 2000, Financial Accounting: Tools for Business Decision Making , John Wiley& Sons, Inc., New York. Kremer, C. & Rizzuto, R. 2000, 'Three bottom lines', Executive Excellence, vol. 17, no. 9, pp. 6-7. Krishnan, G.V. & Largay, J.A. 2000, 'The predictive ability of direct method cash flow information', Journal of Business Finance & Accounting , vol. 27, no. 1/2, pp. 215-45. Lancaster, C., Stevens, J.L. & Jennings, J.A. 1998, 'Corporate liquidity and the significance of earnings versus cash flow', Journal of Applied Business Research, vol. 14, no. 4, pp. 27-38. Lawson, G.H. 1992, Studies in Cash Flow Accounting and Analysis, Garland Publishing, Inc., New York and London. Lee, T.A. 1981, 'Cash flow accounting and corporate financial reporting', in M Bromwich & A Hopwood (eds), Essays in British Accounting Research, Pitman Publishing, London, pp. 63-78. ---- 1986, Towards a Theory and Practice of Cash Flow Accounting , Accounting Thought and Practice through the Years, Garland Publishing, New York. ---- 1992, Cash Flow Reportings Recent History: a Kuhnian Interpretation of a Changing Accounting Emphasis, Working Paper, University of Alabama. ---- 1993, Cash Flow Reporting: A Recent History of an Accounting Practice, New works in Accounting History, Garland Pub, New York. Lee, T.A., Ingram, R.W. & Howard, T.P. 1999, 'The difference between earnings and operating cash flow as an indicator of financial reporting fraud', Contemporary Accounting Research, vol. 16, no. 4, pp. 749-86. Lees, A.T. & Leibman, P.R. 2000, 'Free cash flow: back to basics', Oil & Gas Investor, vol. 20, no. 4, pp. 67-70. Lipe, R. 1990, 'The relation between stock returns and accounting earnings given alternative information', The Accounting Review, vol. 65, no. 1, pp. 49-71. Lorek, K. S., Schaefer, T. F. & Willinger, G. L. 1993, 'Time-series properties and predictive ability of funds flow variables', The Accounting Review, vol. 68, no. 1, pp. 151-63.

192

Lorek, K.S. & Willinger, G.L. 1996, 'A multivariate time-series prediction model for cashflow data', The Accounting Review, vol. 71, no. 1, pp. 81-101. Lundholm, R.J. 1999, 'Reporting on the past: A new approach to improving accounting today', Accounting Horizons, vol. 13, no. 4, pp. 315-22. Mahidol University 1997, The Character and Structure of The Thai Economy, Mahidol University, viewed 7 July 2003, <http://www.mahidol.ac.th/Thailand/economy/charac-econ.html>. McBeth, K.H. 1993, 'Forecasting operating cash flow: Evidence on the comparative predictive abilities of net income and operating cash flow from actual cash flow data', The Mid-Atlantic Journal of Business, vol. 29, no. 2, pp. 173-87. Mills, J. & Yamamura, J.H. 1998, 'The power of cash flow ratios', Journal of Accountancy, vol.186, no. 4, p. 53-8. Moorehead, C. 2001, 'The working of the statement of cash flows', Business Credit, vol. 103, no. 8, pp. 22-3. Mossman, C.E., Bell, G.G., Swartz, L.M. & Turtle, H. 1998, 'An empirical comparison of bankruptcy models', The Financial Review, vol. 33, no. 2, pp. 35-54. Murdoch, B. & Krause, P. 1989, 'An empirical investigation of the predictive power of accrual and cash flow data in forecasting operating cash flow', Akron Business and Economic Review, vol. 20, no. 3, pp. 100-13. ---- 1990, 'Further evidence on the comparative ability of accounting data to predict operating cash flows', The Mid - Atlantic Journal of Business, vol. 26, no. 2, pp. 1-14. Myers, R.H. 1990, Classical and Modern Regression with Applications, 2nd edn, Duxbury Press, Belmont, Calif. Narktabtee, K. 2000, 'The implications of accounting information in the Thai capital market', PhD thesis, University of Arkansas. Neill, J.D., Schaefer, T.F., Bahnson, P.R. & Bradbury, M.E. 1991, 'The usefulness of cash flow data: A review and synthesis', Journal of Accounting Literature, vol. 10, pp. 117-50. Nurnberg, H. 1993, 'Inconsistencies and ambiguities in cash flow statements under FASB Statement No. 95', Accounting Horizons, vol. 7, no. 2, pp. 60-75. O'Bryan, D., Quirin, J.J. & Berry, K.T. 1999, 'The role of accruals and cash flows in the corporate bond market', The Mid - Atlantic Journal of Business, vol. 35, no. 4, pp. 189-202.

193

Penman, S.H. 2001, 'On comparing cash flow and accrual accounting models for use in equity valuation: A response to Lundholm and O'Keefe', Contemporary Accounting Research , vol. 18, no. 4, pp. 681-92. Percy, M. & Stokes, D.J. 1992, 'Further evidence on empirical relationships between earnings and cash flows', Accounting and Finance, vol. 32, no. 1, pp. 27-49. Perry, C., Riege, A. & Brown, L. 1999, 'Realism's role among scientific paradigms in marketing research', Irish Marketing Review, vol. 12, no. 2, pp. 16-23. Pizzey, A. 1991, 'Cashflow statements: A step in the right direction or a step in the dark?' Management Accounting, vol. 69, no. 5, pp. 20-2. Plewa, F.J. & Friedlob, G.T. 1995, Understanding Cash Flow, Wiley, New York. ---- 2002, 'New ways to analyze cash flows', National Public Accountant, vol. 47, no. 1, pp. 25-43. Quirin, J.J., O'Bryan, D. & Berry, K.T. 2000, 'Forecasting cash flow from operations: evidence from the oil and gas industry', Oil, Gas & Energy Quarterly, vol. 48, no. 3, pp. 517-28. Quirin, J.J., O'Bryan, D., Wilcox, W.E. & Berry, K.T. 1999, 'Forecasting cash flow from operations: Additional evidence', The Mid - Atlantic Journal of Business, vol. 35, no. 2/3, pp. 135-42. Reilly, F.K. 1994, Investment Analysis and Portfolio Management, The Dryden Press, Florida. Riahi-Belkaoui, A. & Jones, S. 2002, Accounting Theory, 2nd Australian edn, Nelson Thomson Learning, South Melbourne. Ryan, B., Scapens, R.W. & Theobald, M. 1992, Research Method and Methodology in Finance and Accounting , Academic Press Inc., San Diego. Salmi, T., Virtanen, I. & Yli-Olli, P. 1990, On the classification on financial ratios: A factor and transformation analysis of accrual, cash flow, and market-based ratios, Acta Wasaensia, no. 25, University of Vaasa, Finland. Schaeffer, H.A. 2002, 'Cash flow: the lifeblood of a business', Controllers Update, no. 200, pp. 1,3. Sekaran, U. 1992, Research Methods for Business, John Wiley & Sons, New York.

194

SET 2000a, Information disclosure& dissemination, The Stock Exchange of Thailand, viewed 12 August 2003, <http://www.set.or.th/en/about/how/disclosure/discloure_p3.html>. ---- 2000b, Overview, The Stock Exchange of Thailand, viewed 20 June 2003, <www.set.or.th/en/about/set/overview_pl.html>. ---- 2000c, History, The Stock Exchange of Thailand, viewed 20 June 2003, <www.set.or.th/en/about/set/history_p1.html>. SET 2000d, Broad, The Stock Exchange of Thailand, viewed 20 June 2003, <www.set.or.th/en/about/set/board_p1.html>. ---- 2000e, Information disclosure& dissemination , The Stock Exchange of Thailand, viewed 12 Aug 2003, <http://www.set.or.th/en/about/how/disclosure_p1html>. ---- 2000f, Information Disclosure and Dissemination, The Stock Exchange of Thailand, viewed 12 Aug 2003, <http://www.set.or.th/th/about/how/disclosure/disclosure_p2.html>. Sharma, D.S. 2001, 'The role of cash flow information in predicting corporate failure: The state of the literature', Managerial Finance, vol. 27, no. 4, pp. 3-28. Sharma, D.S. & Iselin, E.R. 2003, 'The relative relevance of cash flow and accrual information for solvency assessments: A multi-method approach', Journal of Business Finance & Accounting , vol. 30, no. 7/8, pp. 1115-40. Sharma, R. & Jones, S. 2000, 'Managing cash flow: Taking a strategic view', Australian CPA, vol. 70, no. 1, pp. 32-3. Sincich, T. 1996, Business Statistics by Example, 5th edn, Prentice Hall, Upper Saddle River, NJ. Sloan, R.G. 1996, 'Do stock prices fully reflect information in accruals and cash flows about future earnings?' The Accounting Review, vol. 71, no. 3, pp. 289-315. Sondhi, A.C., Sorter, G.H. & White, G.I. 1988, 'Cash flow redefined: FAS 95 and security analysis', Financial Analysts Journal, vol. 44, no. 6, pp. 19-20. Sorter, G.H., Ingberman, M.J. & Maximon, H.M. 1990, Financial Accounting: an Events and Cash Flow Approach , McGraw-Hill, New York. Stammerjohan, W.W. & Nassiripour, S. 2000/2001, 'Predicting SFAS 95 cash flows: the relative importance of prior earnings, cash flows, and accruals', Accounting Enquires, vol. 10, no. 1, pp. 87-146.

195

Staubus, G.J. 1989, 'Cash flow accounting and liquidity: cash flow potential and wealth', Accounting and Business Research, vol. 19, no. 74, pp. 161-9. ---- 2004, 'Two views of accounting measurement', Abacus, vol. 40, no. 3, pp. 265-79. Supriyadi 1998, 'The association between accounting information and future cash flows: an Indonesian case study', PhD thesis, University of Kentucky. Sylvestre, J. & Urbancic, F.R. 1994, 'Effective methods for cash flow analysis', Healthcare Financial Management, vol. 48, no. 7, pp. 62-70. Tabachnick, B.G., Fidell, L.S. & Osterlind, S.J. 2001, Using Multivariate Statistics, 4th edn, Allyn and Bacon, Boston; London. Ticehurst, G.W. & Veal, A.J. 2000, Business Research Methods, Pearson Education Australia, NSW. Tourism Authority of Thailand 2003, About Thailand , Tourism Authority of Thailand, viewed 26 Jul 2003, <http://www.tourismthailand.org/about_thailand.php?module=plan&file=plan01>. Trotman, K. & Gibbins, M. 1998, Financial Accounting: Integrated Approach , Nelson ITP, Canada. Wallace, R.S.O., Choudhury, M.S.I. & Pendlebury, M. 1997, 'Cash flow statements: An international comparison of regulatory positions', The International Journal of Accounting , vol. 32, no. 1, pp. 1-22. Wang, Z. & Eichenseher, J. 1998, 'Informativeness and predictability of cash flows', Journal of Applied Business Research, vol. 14, no. 2, pp. 21-32. Watts, R.L. & Zimmerman, J.L. 1986, Positive Accounting Theory, Prentice-Hall, Englewood Cliffs, N.J. Weaver, W. & Michelson, S. 2003, 'A practical tool to assist in analyzing risk associated with income capitalization approach valuation or investment analysis', The Appraisal Journal, vol. 71, no. 4, pp. 335-44. Wertheim, P. 1989, 'Forecasting cash flow by industry', The Journal of Business Forecasting Methods & Systems, vol. 8, no. 4, pp. 9-12, 27. Wertheim, P. & Robinson, M.A. 1993, 'Issues in forecasting company liquidity', Business Credit, vol. 95, no. 9, pp. 14-7. ---- 1993, 'Earnings versus cash flow: The information provided about changes in company liquidity', Journal of Applied Business Research, vol. 9, no. 4, pp. 65-75.

196

Williamson, D. 2003, 'Free cash flow', Teaching Business & Economics, vol. 7, no. 1, pp. 13-7. Yli-Olli, P. & Virtanen, I. 1989, 'On the long-term stability and cross-country invariance of financial ratio patterns.' Journal of Operational Research, vol. 39, no. 1, pp. 40-53. Zega, C.A. 1988, 'The new statement of cash flows', Management Accounting, vol. 70, no. 3, pp. 54-9. Zikmund, W.G. 2000, Business Research Method, The Dryden press, Florida. Zwaig, M.C. & Pickett, M.M. 2001, 'Early warning signs of a bankruptcy', Business Credit, vol. 103, no. 1, p. 65.

197