The Determinants of Household Indebtedness: Household-Level Evidence from Thailand

The determinants of household indebtedness: household-level evidence from Thailand

Pasit Chotewattanakul

A thesis in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Business UNSW Canberra

August 2019 Thesis/Dissertation Sheet

Surname/Family Name : Chotewattanakul Given Name/s : Pasit Abbreviation for degree as give in the University calendar : PhD Faculty : UNSW Canberra School : School of Business The determinants of household indebtedness: household-level Thesis Title : evidence from Thailand

Abstract 350 words maximum: (PLEASE TYPE)

Rapid increase in household debt around the world has been an issue of concern amongst central bankers. This thesis explores the nature of and the dangers posed by household indebtedness, using Thailand as a case study. Data collected through Thailand's official household surveys, together with quantitative static and dynamic analysis have been used to determine the factors underscoring household indebtedness. The conceptual framework for the foregoing analysis is based on three strands of the literature: the Life-Cycle and Permanent- Income Hypothesis, credit rationing, and behavioural finance. The analysis suggests that secure income and sufficient savings allow households to improve their debt performance and reduce their chances of being over-indebted. Conversely, higher level of dependency on finance (more household members with no income) leads to poorer debt performance over time. Credit constraints can cause more unsound debt performance as well, whereas higher financial literacy is correlated with superior debt performance.

This thesis also investigates the correlation between perceptions of being over-indebted by the household head (subjective over- indebtedness) and quantitative measures of over-indebtedness using data on income and debt repayments (objective over-indebtedness). The results from the logit regressions show both positive and negative correlations between these two types of over-indebtedness assessments. Households with regular income and those who keep doing income-and-expenditure accounts have consistent assessments of over-indebtedness. In contrast, households with an illiterate head tend to not have perceptions of being over-indebted when the objective metrics suggest otherwise. Additionally, education loans bias subjective assessment of over-indebtedness vis-a-vis the objective measurement.

The finding that a secure income can improve debt performance suggests that access to an income safety net may help those with volatile incomes (such as farmers) to have better economic wellbeing. Furthermore, the finding of financial illiteracy correlated with incorrect over- indebtedness assessments suggests that financial literacy might encourage households to obtain a proper amount of debt to smooth their intertemporal consumption and grow their permanent income. Finally, the conservative stance of households regarding education loans suggests that income-contingent education loans have the potential to raise educational investment.

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

………………………………………………… ……………………………………………..………… …….……………………...…….…

Signature Witness Signature Date

The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for award:

ORIGINALITY STATEMENT

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed ……………………………………………......

Date ……………………………………………......

INCLUSION OF PUBLICATIONS STATEMENT UNSW is supportive of candidates publishing their research results during their candidature as detailed in the UNSW Thesis Examination Procedure.

Publications can be used in their thesis in lieu of a Chapter if:  The student contributed greater than 50% of the content in the publication and is the “primary author”, ie. the student was responsible primarily for the planning, execution and preparation of the work for publication  The student has approval to include the publication in their thesis in lieu of a Chapter from their supervisor and Postgraduate Coordinator.  The publication is not subject to any obligations or contractual agreements with a third party that would constrain its inclusion in the thesis

Please indicate whether this thesis contains published material or not. ☐ This thesis contains no publications, either published or submitted for publication

Some of the work described in this thesis has been published and it has been ☒ documented in the relevant Chapters with acknowledgement

This thesis has publications (either published or submitted for publication) ☐ incorporated into it in lieu of a chapter and the details are presented below

CANDIDATE’S DECLARATION I declare that:  I have complied with the Thesis Examination Procedure  where I have used a publication in lieu of a Chapter, the listed publication(s) below meet(s) the requirements to be included in the thesis. Name Signature Date (dd/mm/yy)

iii

‘I hereby grant the University of New South Wales or its agents a non-exclusive licence to archive and to make available (including to members of the public) my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known. I acknowledge that I retain all intellectual property rights which subsist in my thesis or dissertation, such as copyright and patent rights, subject to applicable law. I also retain the right to use all or part of my thesis or dissertation in future works (such as articles or books).’

‘For any substantial portions of copyright material used in this thesis, written permission for use has been obtained, or the copyright material is removed from the final public version of the thesis.’

Signed ……………………………………………......

Date ……………………………………………......

AUTHENTICITY STATEMENT

‘I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis.’

Signed ……………………………………………......

Date ……………………………………………......

Acknowledgements

I would like to acknowledge all the support and assistance I have received during my PhD candidature at the School of Business, the University of New South Wales at the Australian Defence Force Academy (UNSW Canberra). First of all, my great appreciation is to my supervisors: Professor Satish Chand and Dr Keiran Sharpe. Both of them were extremely supportive and always provided me with invaluable feedback. I would like to say thank you very much to them for their time and their help. With their guidance, I have improved my research skills (particularly quantitative research skills), including English academic writing skills. Next, I would like to express my thanks to my panel committee: Professor Michael O’Donnell, and Associate Professor Twan Huybers. In every annual progress review meeting, Professor O’Donnell and Professor Huybers were both very kind and supportive. After each meeting, I felt encouraged to continue on this long research journey, sometimes without knowing what was in store. Additionally, I extend my appreciation to the administrative staff of the School of Business (especially Mrs Valerie-Ann Verdin and Ms Jessica Campbell) who were always helpful with prompt responses to all enquiries.

I also would like to thank Xinghua Liu, Fei Meng, Churaporn Charoenporn, Trisukon Sawatrukkiat, and Phitawat Poonpolkul for constructive comments on earlier drafts of the study of the drivers of household indebtedness, the first published paper during my PhD candidature. Moreover, friendships from all of my friends (UNSW friends, friends in Canberra, and friends in Thailand), and love from both family and my girlfriend helped me throughout this journey. Their encouragement was invaluable in keeping my spirits up throughout the candidature.

Lastly, this study could not have been completed without financial support from the Australian Government through a Research Training Program Scholarship. I would like to sincerely thank the UNSW, Korean Studies Association of South East Asia (KoSASA), and Chulalongkorn University (a KoSASA member university) for offering me the opportunity to study towards a Doctorate at UNSW Canberra through their prestigious scholarship. In addition, I also would like to thank my bosses at the Bank of Thailand for allowing me to undertake this study.

Abstract

Rapid increase in household debt around the world has been an issue of concern amongst central bankers. This thesis explores the nature of and the dangers posed by household indebtedness, using Thailand as a case study. Data collected through Thailand's official household surveys, together with quantitative static and dynamic analysis have been used to determine the factors underscoring household indebtedness. The conceptual framework for the foregoing analysis is based on three strands of the literature: the Life-Cycle and Permanent-Income Hypothesis, credit rationing, and behavioural finance. The analysis suggests that secure income and sufficient savings allow households to improve their debt performance and reduce their chances of being over-indebted. Conversely, higher level of dependency on finance (more household members with no income) leads to poorer debt performance over time. Credit constraints can cause more unsound debt performance as well, whereas higher financial literacy is correlated with superior debt performance.

This thesis also investigates the correlation between perceptions of being over- indebted by the household head (subjective over-indebtedness) and quantitative measures of over-indebtedness using data on income and debt repayments (objective over-indebtedness). The results from the logit regressions show both positive and negative correlations between these two types of over-indebtedness assessments. Households with regular income and those who keep doing income-and-expenditure accounts have consistent assessments of over-indebtedness. In contrast, households with an illiterate head tend to not have perceptions of being over-indebted when the objective metrics suggest otherwise. Additionally, education loans bias subjective assessment of over-indebtedness vis-a-vis the objective measurement.

The finding that a secure income can improve debt performance suggests that access to an income safety net may help those with volatile incomes (such as farmers) to have better economic wellbeing. Furthermore, the finding of financial illiteracy correlated with incorrect over-indebtedness assessments suggests that financial literacy might encourage households to obtain a proper amount of debt to smooth their intertemporal consumption and grow their permanent income. Finally, the conservative stance of households regarding education loans suggests that income-contingent education loans have the potential to raise educational investment.

Author publications derived from this thesis

Chotewattanakul, P., Sharpe, K., & Chand, S. (2018). Differences between subjective over-indebtedness and objective over-indebtedness: household-level evidence from Thailand. Asia-Pacific Conference on Economics and Finance e-proceeding, Singapore, 2018. ISBN: 978-981-11-6602-0.

Chotewattanakul, P., Sharpe, K., & Chand, S. (2019). The drivers of household indebtedness: evidence from Thailand. Southeast Asian Journal of Economics, 7(1): 1-40.

vii

Table of Contents

Acknowledgements ...... v

Abstract ...... vi

Author publications derived from this thesis ...... vii

List of Figures ...... xi

List of Tables ...... xiv

Chapter 1: Introduction ...... 1

1.1 Background ...... 1 1.2 Research questions and hypotheses ...... 12 1.3 Organisation of the study ...... 13 Chapter 2: Literature Review ...... 15

2.1 Development of global household debt ...... 15 2.2 Impacts of household debt on household’s financial stability and the macroeconomy ...... 19 2.2.1 Positive impacts of household debt on household’s financial stability and the macroeconomy…………………………………………………………………….19 2.2.2 Negative impacts of household debt on household’s financial stability and the macroeconomy…………………………………………………………………….20 2.2.3 Policy responses to rising household debt in the region of Southeast Asia .... 23 2.3 Alternative explanations of household indebtedness ...... 24 2.4 Household indebtedness measurement ...... 30 2.5 Research on the drivers of household indebtedness ...... 33 2.6 Research on household indebtedness in Thailand ...... 38 2.7 Research gaps ...... 41 Chapter 3: Context Chapter ...... 44

3.1 Geographical and economic background of Thailand ...... 44 3.2 Thai economy over the past decade (2008 to 2017) ...... 46 3.2.1 Aggregate economic performance ...... 46 3.2.2 Contribution to economic growth from the demand side ...... 51 3.2.3 Contribution to economic growth from the supply side (Sectoral economic performance) ...... 54

viii

3.3 Financial system and household finance in Thailand ...... 56 3.3.1 Structure of Thailand’s financial system ...... 56 3.3.2 Household access to finance ...... 59 3.3.3 Household financial inequality ...... 61 3.3.4 Household debt development...... 65 3.4 Conclusion ...... 71 Chapter 4: The Drivers of Household Indebtedness: Evidence from Thailand ..... 72

4.1 Introduction ...... 72 4.2 Data and overview of household indebtedness in Thailand ...... 75 4.2.1 Data ...... 75 4.2.2 Overview descriptive statistics ...... 79 Household’s indebtedness by income group ...... 80 Proportion of indebted households ...... 84 Household debt performance in Thailand ...... 84 Proportion of over-indebted households ...... 86 4.3 Methodology ...... 87 4.4 Empirical results ...... 90 4.5 Conclusion ...... 98 Chapter 5: Differences between Subjective Over-indebtedness and Objective Over- indebtedness: Household-Level Evidence from Thailand ...... 101

5.1 Introduction ...... 101 5.2 Data and overview of the differences between subjective over-indebtedness and objective over-indebtedness in Thailand ...... 103 5.2.1 Data ...... 103 5.2.2 Overview descriptive statistics ...... 104 The proportion of indebted households by indebtedness status ...... 105 Households with consistent debt perceptions by social class, region, and household head’s education level ...... 106 Households with inconsistent debt perceptions by social class, region, and household head’s education level ...... 108 5.3 Methodology ...... 109 5.4 Empirical results ...... 111 5.5 Conclusion ...... 120

Chapter 6: Dynamic Analysis of Household Indebtedness: Evidence from Thailand ...... 124

6.1 Introduction ...... 124 6.2 Literature review ...... 126 6.2.1 Dynamic analysis of household indebtedness ...... 126 6.2.2 Pseudo-panel analysis using household surveys in the literature ...... 128 6.3 Data and development of household debt in Thailand from 2009 to 2015 ...... 130 6.3.1 Data ...... 130 6.3.2 Overview descriptive statistics ...... 132 Development of indebted households’ financial status in Thailand during 2009 to 2015 ...... 132 Households with the most unsound financial status ...... 135 6.4 Methodology ...... 149 6.5 Empirical results ...... 151 6.6 Conclusion ...... 156 Chapter 7: Conclusions and Policy Implications ...... 159

7.1 Conclusions ...... 159 7.1.1 Thailand’s household debt situation from 2009 to 2015 ...... 159 7.1.2 Empirical results ...... 161 Role of the life cycle hypothesis and the permanent income hypotheses ...... 162 Role of credit rationing ...... 164 Role of behavioural finance ...... 166 7.2 Policy implications ...... 168 7.3 Main contributions of the study ...... 174 7.4 Areas for further research ...... 175 References ...... 176

Appendix A ...... 188

Appendix B ...... 190

List of Figures

Figure 1.1 ODCs’ Average Household Loan Growth by Country from 2009 – 2013…..3

Figure 1.2 Household Debt and Consumption Loss.…………………………….……….6

Figure 1.3 Household Debt and Consumption in Thailand.….…………………..……....7

Figure 2.1 Household Debt-to-GDP ratio in Advanced and Emerging Economies (Per cent)…...………………………………………………………………………………..18

Figure 2.2 Life-Cycle Hypothesis…………………………………………………...... 26

Figure 2.3 Intertemporal Consumption ………………………...……………………...27

Figure 2.4 Intertemporal Consumption with a Kinked Budget Line …………………..27

Figure 2.5 Intertemporal Consumption with Liquidity Constraint ..…………………...28

Figure 3.1 Map of Thailand ……………………………………………………………45

Figure 3.2 Thailand’s Real GDP and Real GDP Growth ……………………………...47

Figure 3.3 Growth in Private Consumption, Private Investment, and Exports of Goods and Services ……………………………………………………………………………48

Figure 3.4 Real GDP Growth, Growth in Fiscal Spending, and Policy Interest Rate.....49

Figure 3.5 ASEAN-5’s Real GDP Growth …………………………………………….51

Figure 3.6 Shares of GDP Components (Demand Side)…………………………….…52

Figure 3.7 Contribution to GDP Growth (Demand Side) ……………………………...53

Figure 3.8 Shares of GDP Components (Supply Side) ………………………………..55

Figure 3.9 Contribution to GDP Growth (Supply Side) ……………………………….55

Figure 3.10 Share of Household Income and Expenditure by Income Group ....……...63

Figure 3.11 Monthly Expenditure to Income Ratio by Income Group ………………..63

Figure 3.12 Share of Household Debt and Debt-Service-Payment Burden by Income Group …………………………………………………………………………………...64

Figure 3.13 Household Debt Growth, GDP Growth and Consumption Growth in Thailand.…………………………………………………………………………….…..66

Figure 3.14 Consumer Loan Growth…...……………………………………………….67 xi

Figure 3.15 Non-Performing Consumer Loans…...…………………………………….67

Figure 3.16 ODCs’ Household Loans by Country (Average Growth)……..….…..…....68

Figure 3.17 ODCs’ Household Loans by Country (Outstanding: Local Currency)...... 69

Figure 4.1 Debt Service Ratio by Income Group.....…………………………………....82

Figure 4.2 Debt to Annual Income Ratio by Income Group.………………………...…83

Figure 4.3 Average Composition of Debt by Income Group.……………………….....83

Figure 4.4 Debt Performance by Household Head’s Age..…………………………..…86

Figure 4.5 Proportion of Households With Expectation of Better Households’ Economic Situation (by Household Head’s age group)……………………………………….……93

Figure 4.6 Effective Interest Rate by Type of Loan..…………………..………….……93

Figure 4.7 Debt to Annual Income Ratio by Household Head’s Age.…………...……..97

Figure 5.1 Proportion of Indebted Households………………………………………...105

Figure 5.2 Homeownership Rate by Type of Indebted Households....………….….….115

Figure 6.1 Indebted Households’ Income, Expenditure, and Debt for the Period 2009 to 2015.…..……………………………………………………………………………….134

Figure 6.2 Household Saving-to-Disposable-Income Ratio from National Income Accounts.………………………………………………………………………………135

Figure 6.3 Household Income by Income Group…..………………………..……...... 139

Figure 6.4 Household Debt by Income Group……………………………….………...139

Figure 6.5 Household Debt Service Payment by Income Group….………...………..140

Figure 6.6 Household Income by Region...………………………………….………...142

Figure 6.7 Household Debt by Region..……………………………………..………...143

Figure 6.8 Household Debt Service Payment by Region.…..………………..………..143

Figure 6.9 Rubber Price……………………………………………………….……….144

Figure 6.10 Household Income by Social Class..……………………………..……….147

Figure 6.11 Household Debt by Social Class..………………………………..……….147

Figure 6.12 Household Debt Service Payment by Social Class….…………...……….148 xii

Figure 6.13 Average Effective Interest Rate by Source of Credit.…………...... ……154

Figure 6.14 Average Borrowers’ Debt Performance by Source of Credit…..……...….154

xiii

List of Tables

Table 3.1 Contribution to GDP Growth (Demand Side)..…………………….……...... 53

Table 3.2 Contribution to GDP Growth (Supply Side).………………………..……….56

Table 3.3 Overview Statistics of the Average Household’s Financial Status…....…...…70

Table 3.4 Proportion of Indebted Households (Households with debt and debt service burden).…..……………………………………………………………………..…...…..70

Table 3.5 Overview Statistics of the Average Indebted Households’ Financial Status....70

Table 4.1 Number of Observations……..…………………………………..…………...78

Table 4.2 Poverty Line (THB/ Person/ Month)…..…………..……………..……….….78

Table 4.3 Household’s Income in 2013 by Income Group (THB/ Month)....…...... 82

Table 4.4 Proportion of Indebted Households..…………………………….……………84

Table 4.5 Household Debt Performance..………………………………….……………85

Table 4.6 Proportion of Over-indebted Households..……………………….…………..87

Table 4.7 List of Explanatory Variables..…………………………….……………..…..89

Table 4.8 List of Control Variables.…………..…………………………..………….....90

Table 4.9 Debt Performance Models (OLS)..…………………………..….………..…..92

Table 4.10 Marginal Effects on Probability of Being Subjectively Over-indebted...…...95

Table 4.11 Marginal Effects on Probability of Being Objectively Over- indebted....………………………………………………………………………………96

Table 4.12 Proportion of Over-indebted Households by Household Head’s Age..…...... 97

Table 5.1 Proportion of Indebted Households by Indebtedness Status..….…….……..106

Table 5.2 Proportion of Indebted Households by Indebtedness Status and Social Class …………………………………………………………………………………………107

Table 5.3 Proportion of Indebted Households by Indebtedness Status and Region.…. 107

Table 5.4 Proportion of Indebted Households by Indebtedness Status and Household Head’s Education Level………………………………...……………………………..108

Table 5.5 List of Explanatory Variables……………………………………………….110 xiv

Table 5.6 List of Control Variables……………………………………………………111

Table 5.7 Marginal Effects on a Probability of Being Over-indebted…………………112

Table 5.8 Proportion of Indebted Households by Debt Repayment Status….…….…..116

Table 5.9 Proportion of Indebted Households by Indebtedness Status and Household Head’s Gender.……………………………………………………………….……..…117

Table 5.10 Average Financial literacy Score by type of indebtedness.……….....…….118

Table 5.11 Proportion of Indebted Households by Indebtedness Status and Dependency Ratio.………………………………………………………………………….……….119

Table 5.12 Households by Social class and Government Debt Relief…..….…...…….120

Table 6.1 Poverty Line (THB/ Person/ Month)….……………………….…..……….131

Table 6.2 Overview Statistics of Indebted Households’ Financial Status..…...……….134

Table 6.3 Household Financial Status by Income Group…….…………...……..…….137

Table 6.4 Household Financial Status by Region..…….……………..………………..141

Table 6.5 Proportion of Households with the Perception of Having Secure Income by Region………………………………………………..………………………….…….144

Table 6.6 Household Financial Status by Social Class..………………………..……..146

Table 6.7 Proportion of Households with the Perception of Having Secure Income by Region..………………………………………………………………………….…….148

Table 6.8 Household Head's Characteristic Groups..…………………………..……..149

Table 6.9 Region Groups……………………………………………………….……..149

Table 6.10 List of Variables...………………………………………………….……..151

Table 6.11 Debt Performance Models (with Fixed Effects/ Random Effects)….……..153

Table 6.12 Homeownership (House and Land) and Credit Access..……………..……155

Table 6.13 Financial Status of Households with Degrees and with No Degrees.....…..156

Table 6.14 Average Financial Literacy Scores and Education levels..…………..……156

Table A1 Model Specifications for Subjective Over-indebtedness Model (Logit Model) …………………………………………………………………………………………188

Table A2 Model Specifications for Objective Over-indebtedness Model (Logit Model) …………………………………………………………………………………….…...189

Table B1 Model Specifications for Over-indebtedness Model (Logit Model) …………………………………………………………………………………….…...190

xvi

Chapter 1: Introduction

1.1 Background

Over the past three decades, household debt in most countries around the world has been increasing. In the case of both developed and developing countries, many researchers and organisations have reported a sharply upward trend in the levels of household debt (Debelle, 2004; Barba & Pivetti, 2009; Bingxi & Lijuan, 2009; Endut & Hua, 2009; Davies, 2009; IMF, 2017; Bank of Thailand, 2018a). They provide quantitative evidence of high rates of the growth of household debt over the past three decades, with some noting even double-digit growth spurts for several developing countries (Bingxi & Lijuan, 2009; Endut & Hua, 2009; IMF, 2017; Bank of Thailand, 2018a). For example, in the case of China, the average annual growth of consumer loans from 2000 to 2007 was greater than 20% per annum. Especially, at the end of 2007, Bingxi and Lijuan (2009) report that the growth of China’s consumer loans reached 30% per annum. In the case of Malaysia, over the same period (2000 to 2007), there was a double-digit annual growth rate of consumer loans. Endut and Hua (2009) stated that the average growth of Malaysia’s consumer loans from 2001 to 2007 was around 15% per annum, with the highest growth rate of 27% reached in 2001. In the case of Thailand, the Bank of Thailand (2018a) also reported double-digit growth rates of financial institutions’ loans to households from 2000 to 2013.

The rapid increase in household debt has often exceeded the rate of growth of household income and the economy as a whole (i.e. Gross Domestic Product: GDP), meaning that both household-debt-to-income ratio and household-debt-to-GDP ratio have had an upward trajectory. Many researchers (Debelle, 2004; Barba & Pivetti, 2009; Bingxi & Lijuan, 2009; Davies, 2009; IMF, 2017) have reported rising ratios of household-debt-to-income and household-debt-to-GDP for both developed countries (e.g. Australia, Canada, France, Germany, Japan, the United Kingdom (UK), and the United States (US)) and developing countries (e.g. China, Malaysia, Thailand, and The Philippines). For instance, in the case of Germany, the household-debt-to-income ratio rose from 75% in the early 1990s to more than 100% in the early 2000s. Whereas, in The Netherlands, the household-debt-to-income ratio increased from 100% in the 1980s to 200% in 2002 (Debelle, 2004).

With regards to the comparison between household debt growth and GDP growth, some international organisations, including the International Monetary Fund (IMF), report that global household debt has been increasing faster than the size of respective economies. This circumstance causes the ratio of household debt to GDP to rise over the same period. IMF (2017) used data from 2008 to 2016 to reveal that the average value of the household debt as a proportion of GDP amongst developing countries rose from 15% to 21%, while the corresponding figure for developed countries rose from 52% to 63%. There will be further detailed discussion about the development of household debt around the world in Chapter 2.

Thailand is one of the selected countries in the analysis of household debt by the IMF (IMF, 2017). Thailand was noted by the organisation as an outlier of the group of developing countries. Thailand’s household debt-to-GDP increased from 52.4% at the end of 2008 to 76.6% by the end of 2013. The value of household loans outstanding jumped from THB 5.1 trillion at the end of 2008 to THB 9.9 trillion at the end of 2013 (Bank of Thailand, 2018a). Thus, while the size of the economy increased by a third between 2008 and 2013, household debt nearly doubled. The size of the growth in household debt in Thailand can be contextualised by comparing it with the rates of growth of household loans of other countries in the region in which Thailand is located, namely in Asia and Oceania (see Figure 1.1). Evidently, Thailand has a rate of double- digit growth in household debt. More of the contextual information on Thailand’s household debt will be provided in Chapter 3.

Figure 1.1 ODCs’1/ Average Household Loan Growth2/ by Country from 2009 - 2013

% 30 28.1

25 21.1 20

15 14.4 14.5 11.3 11.9 12.0 10 8.9 7.5 7.6 5.8 5 2.9

0 -1.0 -5

Notes: 1/ Other depository corporations. 2/ Average annual growth (geometric mean). Source: The Long series on total credit and domestic bank credit to the private non-financial sector, Bank for International Settlements.

The implications of rising household debt are twofold. On the one hand, increasing household debt may reflect better access to credit by the households. This has the beneficial effect of allowing households to smooth their consumption (Debelle, 2004; Barba & Pivetti, 2009; Kang & Ma, 2009). Debelle (2004) and Barba and Pivetti (2009) claimed that, in a low-interest-rate environment, households could have access to loans at low costs. Consequently, they could manage their funds for a desirable path of consumption, including productive investment.

On the other hand, excessive household debt could cause harm by raising the vulnerability of households to insolvency (Debelle, 2004; Barba & Pivetti, 2009; Endut & Hua, 2009; IMF, 2017). The IMF (2017) pointed out that, in case of an excessive increase in household debt together with adverse shocks (e.g. income shocks, and interest rate shocks), the high level of household indebtedness could lead to household financial instability. To illustrate this point, the higher level of household-debt-to- income ratio and household-debt-to-GDP ratio, coupled with a lower saving rate, could make households more sensitive to unemployment and interest-rate hikes. Debelle (2004) and Endut and Hua (2009) opined that, households with high levels of debt and debt service burdens, coupled with low savings, lacked the financial cushions needed to

3 cope with unanticipated adverse financial shocks (e.g. unemployment, and interest rate rises). Conversely, households with higher savings were better cushioned. Moreover, when households needed more funds, indebted households with high indebtedness but low savings tended to have a lower chance of securing more credit. This vulnerability could lead to financial stability risks and macroeconomic risks.

Regarding the impact of household debt on overall financial stability, a high level of household debt is one of the most concerning issues for financial systems (IMF, 2017). Many authorities in both developed and developing countries have raised default risk as one of the negative impacts of the high household debt level, along with slow economic development such as low income growth and a low inflation rate (Bank of Canada, 2017; Bank of England, 2017; Bank of Thailand, 2017a; Reserve Bank of Australia, 2017; South African Reserve Bank, 2017). This is a potential risk to the financial system via credit markets. Central Banks have been concerned about the ability of households to service their debts due to the rapid increase in household debt over recent years. High levels of debt and the resulting debt service burden could lead to debt service pressures for some households. Households with heavy debt service burdens tended to have poor debt serviceability and then face debt difficulties of repaying their debts (D’Alessio & Iezzi, 2013).

There is considerable research that attempts to find linkages between household indebtedness and insolvency within the financial system (Martin, 2011; Mian & Sufi, 2011; Jorda et al., 2016). Martin (2011) and Mian and Sufi (2011), in particular, sought to explain the impact on household indebtedness of the credit crunch in the US in 2008. They claimed that the excessive ‘household leverage’ from more easily available mortgage loans during the period 2002 to 2007 led to a credit crunch via an asset-price channel, where household leverage is defined as the ratio of household debt to income (Mian & Sufi, 2011). Martin (2011) pointed out that the boom in sub-prime mortgages (which were the underlying assets of the sub-prime asset-backed collateralised debt obligations (CDOs)) before the crisis caused a house-price bubble and economic boom. Once the bubble burst, the value of households’ assets (i.e. the values of houses) depreciated dramatically, and lower economic activity led to an economic downturn along with a higher unemployment rate. This situation caused households to be unable to service their debts (Martin, 2011). In addition, Jorda, Schularick, and Taylor (2016), who have done long-term credit analyses in the case of developed countries using the 4

Bank of International Settlements’ credit data, supported this point of view. They believed that credit booms, especially mortgage booms, were an essential source of financial instability. High levels of household leverage (i.e. household-debt-to-income ratios), along with the sharp depreciation in house prices from the bursting of the housing bubble, leads to households’ poor debt serviceability and a higher debt default rate. The higher default rate causes the deterioration in financial institutions’ loan quality (i.e. more overdue loans), resulting in higher credit standards followed by a credit crunch (Martin, 2011).

Concerning the effects of household debt on the macroeconomy, some scholars have pointed out some circumstances where a high level of household debt is another source of macroeconomic vulnerability (Debelle, 2004; Mian & Sufi, 2011; IMF, 2012; Muthitacharoen et al., 2015; Jorda et al., 2016). Firstly, although substantial growth in household debt can boost economic growth in the short term, this kind of growth is debt-driven growth, which is not sustainable. A high level of household debt, together with income shocks can cause financial vulnerability to the household sector. When indebted households with low financial cushioning (i.e. low savings) face negative income shocks (e.g. unemployment, and illness) and cannot obtain more credit, they have difficulties in financing themselves (Debelle, 2004; Mian & Sufi, 2011). Eventually, some need to sacrifice their consumption, and this situation causes a drag in consumption. Secondly, when a country unexpectedly experiences extremely negative shocks, a country with a higher level of household debt tends to face a more prolonged recession than a country with a lower level of household debt faces (IMF, 2012).

The impact that excessive household leverage can have on household consumption over time is discussed by the IMF (2012) and Muthitacharoen, Nuntramas, and Chotewattanakul (2015), both of whom believe that higher household debt increased the probability of lower household consumption in the future. In the case of developed countries, the IMF (2012) selected 37 advanced economies around the world and calculated the changes in household-debt-to-income ratio during 2002 to 2006 (prior to the global financial crisis in 2008) and consumption loss in 2010. Figure 1.2 shows that a greater change in the household-debt-to-income ratio leads to more consumption loss in the next period.

Figure 1.2 Household Debt and Consumption Loss

Notes: The consumption loss in 2010 is the gap between the (log) level of real household consumption in 2010 and the projection of where real household consumption would have been that year based on the pre-crisis trend. The pre-crisis trend is defined as the extrapolation of the (log) level of real household consumption based on a linear trend estimated from 1996 to 2004. AUS: Australia; AUT: Austria; BEL: Belgium; CAN: Canada; CHE: Switzerland; CYP: Cyprus; CZE: Czech Republic; DEU: Germany; DNK: Denmark; ESP: Spain; EST: Estonia; FIN: Finland; FRA: France; GBR: UK; GRC: Greece; HRV: Croatia; HUN: Hungary; IRL: Ireland; ISL: Iceland; ISR: Israel; ITA: Italy; JPN: Japan; KOR: Korea; LTU: Lithuania; LVA: Latvia; NLD: Netherlands; NOR: Norway; NZL: New Zealand; POL: Poland; PRT: Portugal; ROM: Romania; SVK: Slovak Republic; SVN: Slovenia; SWE: Sweden; TWN: Taiwan Province of China; USA: US. Source: IMF (2012). Dealing with Household Debt.

In the case of Thailand, Figure 1.3 shows Thailand’s household debt to GDP ratio, household debt growth, and private consumption growth from 2008 to 2015. It is evident that high household debt growth from 2009 to 2013 (which was more than 10% per annum) led to a low rate of consumption growth over the next period, which was less than 5% per annum.

Figure 1.3 Household Debt1/ and Consumption in Thailand

% HH debt to GDP HH debt growth (RHS) Real consumption growth (RHS) % 90 20

81.1 15 79.7 80 76.6

10 71.5 70 66.2 5

59.3 60 57.9 0

52.4

50 -5 2008 2009 2010 2011 2012 2013 2014 2015 Note: 1/ Household loans from all financial institutions. Source: The Office of the National Economic and Social Development Board and the Bank of Thailand. Author’s calculations.

Furthermore, the accumulation of debt by an increasing share of the population had a bearing on the risks of financial crisis. Many researchers assumed that persistent rates of rapid growth in household debt could lead to financial instability in the household sector, and then on the macroeconomy as well (Barnes & Young, 2003; Debelle, 2004; Endut & Hua, 2009; Cecchetti et al., 2011; Martin, 2011). In the case where the rapid accumulation of debt amongst the households was due to insufficient savings and/or low income, then an interest rate hike could put households at risk of bankruptcies which, if widespread, could lead to a financial crisis of the nature observed during the global financial crisis in 2008. The global financial crisis of 2008 was a timely reminder of the effects of an unwinding of high household debt on the macroeconomy. When dealing with household debt, the IMF (2017) observed that developing countries and developed countries responded differently to rising household debt. The effects of the high debt-level were different depending on the stage of financial development and institutional characteristics. Developing countries may deal less effectively than developed countries with the de-leveraging process after a high- indebtedness period, owing to a lack of pre-emptive plans for debt restructuring, including personal bankruptcy. The de-leveraging process in this circumstance refers to

7 a stage where indebtedness declines due to tightened credit standards by financial institutions (Geanakoplos, 2010). For example, when there is a sharp and persistent increase in household debt, banks will expect higher debt defaults and will require a higher level of collateral or ‘down payment’. Moreover, there may be less access to credit in emerging economies than in advanced economies. As a result of a credit crunch, people in developing countries may have lower access to loans than people in developed countries.

In light of the discussion above – which articulates the nature and risks to households and the financial system of excessive household debt – it is desirable to determine the drivers of household indebtedness and ‘over-indebtedness’. A number of researchers have explored this issue (Haas, 2006; Vante, 2006; Betti et al., 2007; Gumy, 2007; Anderloni & Vandone, 2008; Keese, 2009; Lusardi & Tufano, 2009; European Commission, 2010; Gathergood, 2012; D’Alessio & Iezzi, 2013; Liv, 2013; Xiao & Yao, 2014; Muthitacharoen et al., 2015; Chantarat et al., 2017). The research in this area helps us to understand household debt behaviours and debt performance, and the implications from the empirical results can be linked to the issues of household debt sustainability1. The remainder of this thesis explores the drivers of household indebtedness in Thailand.

Regarding the research on the drivers of household indebtedness, most studies are based on three main economic theories. The first is from the neoclassical theory, namely the Permanent Income Hypothesis and the Life-Cycle Hypothesis. The second is credit rationing, and the third is from behavioural finance. Firstly, the Permanent Income Hypothesis (Friedman, 1957) and the Life-Cycle Hypothesis (Modigliani, 1966) both provide reasons why a household may take on debt at any time. Friedman (1957) and Modigliani (1966) explained that economic agents have to manage their cash flows over a lifetime in order to smooth consumption levels. Once the current income level is less than the expenditure level, the agents may be borrowers because they need some credit to smooth their consumption. Otherwise, if the current income exceeds the current expenditure, the agents can save and then accumulate wealth. The financial sector provides the means to both save and to borrow to allow households to smooth their

1 Betti, Dourmashkin, Rossi, and Yin (2007) define household debt sustainability as the situation that households can manage and repay their debt without cutting their standard of living. 8 consumption in the face of volatile income (and sometimes one-off large purchases, such as the family home, for example).

However, the above assumes efficient credit markets (i.e. perfectly competitive markets), and the assumption is far from the norm in most countries, particularly in developing nations. The demand for credit often exceeds its supply with given interest rates, and this leads to credit rationing by the formal financial sector. Flemming (1973) and Stiglitz and Weiss (1981) were among the first to note that real-world capital markets were imperfect because of asymmetric information between agents (i.e. borrowers) and financial institutions (i.e. the suppliers of credit). Consequently, in the real world, access to credit for consumption smoothing is assumed in the life cycle and a permanent income hypothesis is not always possible. In practice, working status and credit profiles are used to establish creditworthiness.

Thirdly, more recent research in behavioural finance has identified the role of financial literacy and financial management in saving-and-borrowing behaviours of households (Anderloni & Vandone, 2010; Disney & Gathergood, 2012). The Permanent Income and the Life-Cycle Hypotheses both assume rationality and perfect foresight on the part of households - whilst behavioural finance disputes both assumptions. This strand of the literature has identified that a low-level of financial literacy can lead to poor financial management, which can result in over-indebtedness. Moreover, some households consistently “over-discount” the future cost of borrowing. These households are “hyperbolic discounters”. Such borrowers are then “surprised” by the size of the debt owed when it is due. This status can leave them in a situation of over-indebtedness.

In addition to the drivers of household indebtedness, many researchers have attempted to construct a threshold to indicate ‘over-indebtedness status’. Although indicators of over-indebtedness are not an exact science, progress has been made in constructing two separate measures of this phenomenon. The first is the subjective measurement, which is about households’ self-reporting their ability to repay debt (Anderloni & Vandone, 2008; D’Alessio & Iezzi, 2013). The second is the objective measurement which is a quantitative measure calculating households’ financial ratios to determine their financial vulnerability (Haas, 2006; Betti et al., 2007; Keese, 2009; D’Alessio & Iezzi, 2013).

A number of measures are used to assess objective over-indebtedness of households. This thesis will focus on two separate financial ratios: (i) debt service ratio, and (ii) debt performance. This research defines the debt service ratio (DSR) as the ratio of monthly debt service payments to monthly income (Equation 1.1), while debt performance (DP) is defined here as the ratio of monthly income after monthly debt service payments to the minimum subsistence income level (Equation 1.2).

퐷푆푃 DSR = (1.1) 퐼 where 퐷푆푅 = Debt service ratio

퐷푆푃 = Monthly debt service payments

퐼 = Monthly income

퐼 − 퐷푆푃 DP = (1.2) 푆퐼 where 퐷푃 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

푆퐼 = Minimum subsistence income level

Some researchers define objectively over-indebted households as households with the debt service ratio greater than 30% (Equation 1.1 when DSR greater than 0.3) (D’Alessio & Iezzi, 2013) or households with debt performance ratio less than 1 (Equation 1.2 when DP less than 1) (Keese, 2009, D’Alessio & Iezzi, 2013). Keese (2009) used non-discretionary income as the minimum subsistence income level, whereas D’Alessio & Iezzi (2013) used the poverty line.

The conclusions of the literature survey of household indebtedness have uncovered three research gaps. The first is the empirical findings of the relationship between households’ financial management and household indebtedness in the case of developing countries. Based on few preliminary findings of the correlation between the financial management and the probability of being over-indebted, this thesis will employ variables related to household financial management in the analysis to confirm the previous findings. The added variables are (i) a proxy of keeping doing income-and- expenditure accounts, (ii) the dummy variable of having problems with other financial 10 commitments (rent and utility bills) over the past twelve months; and (iii) a proxy of hyperbolic discounting behaviour. The literature suggests three groups of variables, which can be used to determine household indebtedness. These three groups are motivated by the theories previously stated: neoclassical economics (the Life-Cycle Hypothesis and the Permanent Income Hypothesis), credit rationing, and behavioural finance.

All three approaches will be explored in Chapter 4. The analysis of the drivers of household indebtedness in Chapter 4 will support the examination of the role of each theory, in this instance in Thailand as one of the developing countries with a high-level of household debt. Household-level data collected from Thailand’s official household surveys will be employed for the analysis, and the previously stated three added variables will be used to prove the role of behaviour finance. Then the empirical findings of the analysis will be compared with the previous studies.

Secondly, most of the extant studies of household over-indebtedness rely on a single measure (i.e. either subjective or objective measurement). Therefore, a study of the drivers of household indebtedness using both subjective and objective measures of household indebtedness can be done as a kind of robustness check to confirm the empirical findings (Chapter 4). Moreover, from such analysis, the differences between these two types of measurements can be investigated. Chapter 5 in this thesis will investigate the correlation between these two measurements of household over- indebtedness. This will be based on the three main economic theories to identify the underlying factors for inconsistencies between the two measures.

The empirical results of this study can lead to significant policy implications for lending practices and household financial sustainability. Considering the implications for lending practice, the findings from this study could provide some useful indicators, which could be added into financial institutions’ credit scoring applications. More useful information will allow financial institutions to allocate their assets better, and the gap of asymmetric information in credit markets may be filled. Considering the implications for household financial sustainability, the results from this analysis suggest ways by which households can have a sound credit status, including a consistent debt assessment from both subjective and objective perspectives. Additionally, a study of household indebtedness by using micro-data clarifies the different characteristics

11 between each household, especially in developing countries where there are more heterogeneities in each group of households.

The third research gap is the absence of dynamic analysis in terms of the factors driving household debt over time in the case of Thailand. This investigation will enable the identification of the attributes that underscore household debt status in details and their persistence over time. The data for such analysis are available for Thailand through the Household Socio-Economic Surveys, which have rich household level data (see Chapter 2). However, a challenge in the use of these data for dynamic analysis is that the surveys do not track the same households over time. Consequently, this thesis constructs pseudo-panel data for a time-series analysis.

The final analytical chapter (Chapter 6) consists of a dynamic analysis of household indebtedness using pseudo-panel data. Such analysis permits discovery of relationships between the indebtedness of Thai households and a set of explanatory variables over time.

To conclude, this thesis explores household debt situation in Thailand, examines the determinants of household indebtedness using both static and dynamic analysis, and clarifies the differences between subjective over-indebtedness and objective over- indebtedness. Information from the official household-level surveys from Thailand is used for this study, and quantitative methods applied for the analysis. As mentioned earlier, Thailand is an interesting case study on the determinants of indebtedness for at least three reasons: (i) Thailand has experienced a high-level of household debt over the past ten years; (ii) there is little research on this issue in the case of developing countries; and (iii) micro-data from official household surveys are available.

1.2 Research questions and hypotheses

Central research question

The central research question for this thesis is: “Who is over-indebted, and why?”. The answer to this question leads to the search for the characteristics of households which are likely to be over-indebted.

Sub-research questions

1. What are debt characteristics for different groups of households in terms of the kind of debt they hold and the amount of debt they have? And do these 12

characteristics relate to other characteristics of households (e.g. household size, age profile, income, occupation, region, etc.), including households’ financial behaviours? 2. Are there any differences between subjective (i.e. perception-based) and objective (i.e. quantitative) assessments of over-indebtedness? Moreover, if so, then what explains such differences? 3. Are these debt characteristics persistent or ephemeral? And what can explain household indebtedness over time? Hypotheses

1. The Permanent Income Hypothesis and the Life-Cycle Hypothesis hold: income and head of household’s age play the important roles in explaining household over-indebtedness. 2. Credit constraints (lack of formal credit access), especially in the case of doing business, can be a driver of household indebtedness. 3. Poor financial literacy and management can cause poor credit status. 4. The expectation of household’s economic status (e.g. employment status, occupation, and income level) can make the difference between subjective and objective over-indebtedness. 5. Households with riskier profiles (i.e. households with a higher share of variable income, households with higher dependency ratio, and households who borrow from the informal source of credit) tend to have a greater chance of facing debt unsustainability.

1.3 Organisation of the study

This thesis is divided into seven chapters. The first chapter introduces the household debt issue as one of the recent public concerns, including the research questions and the contributions of this study. The second chapter is a literature review, pointing to what has been done and the gaps that this thesis addresses. Related hypotheses and the previous studies of household indebtedness are reviewed, and research gaps are pointed out in this area of study. The third chapter is a context chapter. Thailand’s economic background and the structure of the financial system in Thailand are explained, including household financial status over the past ten years. The study of the drivers of household indebtedness using Thailand as a case study is presented in the

13 fourth chapter. The fifth chapter is the analysis of differences between subjective over- indebtedness and objective over-indebtedness. A dynamic analysis of household indebtedness is presented in Chapter 6. Lastly, Chapter 7 concludes the empirical results from the previous three analysis chapters and discusses relevant policy implications.

Chapter 2: Literature Review

This literature review focuses on the study of household indebtedness. First it will review the development of household debt around the world. Second, it will discuss the impacts of household debt on a household’s financial stability and the macroeconomy, in both positive and negative ways. Then, it will look at the theoretical hypotheses of household indebtedness and the empirical analysis of household indebtedness. Finally, the discussion will uncover certain some research gaps in the literature.

There are seven sections in this literature review. Firstly, the development of global household debt is described. Secondly, the impacts of household debt on a household’s financial stability and the macro-economy are debated. Thirdly, the hypotheses of household indebtedness, based on economic theories, are reviewed. The measures of household indebtedness are identified in the fourth section. In the fifth section, the extant research on the drivers of household indebtedness is reviewed. After that, the existing research on household indebtedness is reviewed in the case of Thailand. Finally, the research gaps are discussed in this area of research.

2.1 Development of global household debt

The first part of this literature review narrates the development of household debt in both developed and developing countries over the past three decades. This section provides the trend in the levels of household debt across nations where there are available data. The level of household debt has grown rapidly in most countries over the past thirty years. This rising trend is attributed to the historically low interest rates across the world and injections of liquidity by the major central banks in the financial market (Barba & Pivetti, 2009; Davies, 2009; IMF, 2017). The timeline has been divided into two periods: 1) before the global financial crisis in 2008, and 2) after the global financial crisis.

Prior to the global financial crisis, in the case of advanced economies, Debelle (2004) stated that the increasing household debt in developed countries was both level and relative terms to household income. Household debt levels in France, Japan, and the UK rose rapidly from 1980 to 1990, while debt levels in Australia and The Netherlands started increasing dramatically from 1990 to 2000. In addition, in the case of the US, the

15 household debt-to-income ratio increased substantially from the early 1980s to the late 2000s when the global financial crisis occurred.

In the case of Australia and the US, the main contribution to the sharp increase in household debt was the increase in mortgage loans (Barba & Pivetti, 2009; Davies, 2009). The mortgage-loan-to-disposable-income ratio in Australia increased from 31% in 1990 to more than 120% by 2007, whereas the mortgage-loan-to-disposable-income ratio in the US rose from 46% in 1980 to more than 100% in 2007. Regarding overall household debt levels during 1980 to 2007, household-debt-to-income ratios in many developed countries also increased sharply (Debelle, 2004). For example, the household-debt-to-income ratio in The Netherlands jumped from around 100% in the 1980s to 200% in 2002, and the household-debt-to-income ratio in Germany increased from 75% in the early 1990s to more than 100% in the early 2000s.

The IMF (2012) confirmed the same point of view by stating that the average value of household-debt-to-income ratio among advanced economies increased significantly; from an average of 39% in 2003 to 138% in 2007. Considering the household-debt-to-income ratio by income group, in the case of developed countries, indebted households with lower incomes tended to have a higher debt-to-income ratio than indebted households with higher incomes (Barba & Pivetti, 2009). However, Debelle (2004) mentioned that, even though the household-debt-to-income ratio could be examined to reveal the trend of household indebtedness over time, there was no threshold in terms of the debt-to-income ratio to determine the level of ‘over- indebtedness’ of each household. Debelle (2004) suggested that this ratio could be used for a comparison between different cross sections (countries or groups of households) and different times only because it was comparing a stock (debt) with a flow (income).

Regarding household debt development in emerging economies, this thesis focuses on development in Asian developing countries. Before the global financial crisis, in the case of China in the 2000s, consumer loans increased rapidly along with high economic growth (Bingxi & Lijuan, 2009). During 2003 to 2007, China’s GDP grew at an average annual rate of 10%, while consumer credit at the end of 2007 increased with a growth rate of 30% per annum. The share of consumer loans to total loans in China rose from 1.5% in 1999 to 12.5% at the end of 2007. The potential growth in consumer loans in China from 2000 to 2007 was from mortgage loans, which had a growth rate of 34% in 2007. Bingxi and Lijuan (2009) believed that the 16 development of the housing market in China since 1998; which included the changes in the residential real estate taxes, mortgage financing, and the use of land, had been the primary drivers of the rapid increase in mortgage loans. In the case of Malaysia, Endut and Hua (2009) showed there was a double-digit growth rate (around 15% per annum) of consumer loans during the period of 2001 to 2007 with a peak of 27% in 2001. They also claimed that the low inflation rate and the low-interest-rate environment over the past two decades had encouraged people to obtain more loans at a low cost of borrowing.

Similarly for The Philippines, before 2007, the retail credit market was at an early stage of development (Tan, 2009). Higher growth in consumer loans accompanied rising delinquency on credit repayments (measured as being late by one-to-three-month on overdue loans). Part of the rise in this delinquency was attributed to unfamiliarity with consumer credit and the absence of a sound credit culture (e.g. the over-extension of unsecured consumer loans to low-income borrowers). At the end of 2007, credit card loans in The Philippines grew at 16.5% year-on-year (Tan, 2009).

After the global financial crisis, the IMF (2017) stated that, on average, the level of household debt had kept increasing in both developed and developing countries. It estimated the household-debt-to-GDP ratio of each country around the world2 and calculated the median value of these two groups of countries. Figure 2.1 shows the trends of the median household debt-to-GDP ratio of developed countries (i.e. advanced economies) and developing countries (i.e. emerging market economies) calculated by IMF (2017), including 10th-to-90th-percentile ranges and 25th-to-75th-percentile ranges. In the case of advanced economies, the IMF (2017) reported that the median value increased from 52% in 2008 to 63% in 2016 as a result of the continually increasing household debt in many developed countries such as Australia, and Canada. Moreover, the IMF (2017) said that although the household-debt-to-GDP ratios in the UK, the US, and some European countries (e.g. Iceland, Ireland, Portugal, and Spain) have declined

2 List of developed countries: Australia, Austria, Belgium, Canada, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hong Kong SAR, Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Malta, Netherlands, New Zealand, Norway, Portugal, Singapore, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, UK, and US. List of developing countries: Argentina, Bangladesh, Bolivia, Botswana, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Egypt, FYR Macedonia, Georgia, Ghana, Hungary, India, Indonesia, Jordan, Kazakhstan, Kenya, Kuwait, Malaysia, Mauritius, Mexico, Mongolia, Montenegro, Morocco, Namibia, Nigeria, Pakistan, Panama, Paraguay, Philippines, Poland, Romania, Russia, Saudi Arabia, Serbia, South Africa, Thailand, Turkey, Ukraine, Uruguay, and Venezuela. 17 since 2008, the levels of household debt in this group of countries was still high compared with the long-term average values.

Figure 2.1 Household Debt-to-GDP ratio in Advanced and Emerging Economies (Per cent)

Note: Panels show the cross-country dispersion of household debt-to-GDP ratios. Source: IMF (2017). Global financial stability report: Is growth at risk? IMF’s calculations.

Rising household debt has also been evident in emerging market economies where the median household-debt–to GDP ratio increased from 15% in 2008 to 21% by 2016 (IMF, 2017). The IMF (2017) reported that household-debt-to-GDP ratios in most developing countries increased rapidly over the preceding decade. It stressed that the sharp increase in household debt in Malaysia, South Africa, and Thailand led to household-debt-to-GDP ratios greater than 50% by 2016. In the case of Thailand, the Bank of Thailand (2018a) reported that, in five years from 2008 to 2013, the nominal value of household debt outstanding at THB 9.9 trillion at the end of 2013 was nearly double to THB 5.1 trillion at the end of 2008. The sharp increase in household debt caused Thailand’s household-debt-to-GDP ratio to increase by 25% (i.e. the ratio jumped from 52% in 2008 to 77% in 2013).

Over the past three decades, the higher household debt level, along with a lower level of household saving rate, caused households to have fewer cushions to income shocks (Barba & Pivetti, 2009; IMF, 2017). This household debt situation has led to considerable attention by researchers and regulators who have voiced concerns regarding the risks on the economy and the financial sector from the high and rapidly rising levels of household debt under the environment of low inflation and low wage growth (IMF, 2017). The next section will discuss both the positive and negative impacts of household debt on a household’s financial stability and the macro-economy.

2.2 Impacts of household debt on household’s financial stability and the macroeconomy

The implications of greater household indebtedness are debated in this section. As explained next, household debt can cause either positive or negative effects on a household’s financial stability and the macroeconomy.

2.2.1 Positive impacts of household debt on household’s financial stability and the macroeconomy

Debt can be obtained by households to smooth their consumption or invest for the purposes of wealth accumulation. Households can use debts to finance their expenditure and repay their debts during the period of higher income (Friedman, 1957, pp. 20-37; Modigliani, 1966, pp. 160-217). The household life-cycle and permanent income theories predict that households will take on new debts when they face a temporary drop in income or when their anticipated income is larger than the current income. Therefore, debt can help households smooth consumption and in the process stimulate economic growth through higher consumption spending than would be possible otherwise (Barba & Pivetti, 2009). Moreover, in the case where households foresee an increase in their future income, they may borrow to use the proceeds for investments (e.g. buying houses, obtaining higher levels of education).

Additionally, under the assumption of households with rational expectations, a low interest rate environment allows households to allocate their credit to achieve a better path of consumption over their lifetime (Debelle, 2004). People can secure greater funds at low costs of borrowing and also use the funds in productive ways. In theory, households with rational expectations make decisions based on all available information

19 and learn from their experiences (Lucas & Sargent, 1981). They also have perfect foresight (unbiased forecasts) and understand economic theories.

Following this perspective of the effect of household debt, access to private credit leads to consumption and investment growth, which can increase demand and then stimulate the economy. Some researchers affirm that there is a positive relationship between the private sector’s credit growth and economic growth (Beck et al., 2000; Beck et al., 2012). Beck, Levine, and Loayza (2000) claimed that credit to the private sector was a productive form of resource allocation, which encouraged total factor productivity growth and long-run economic growth. However, Beck, Büyükkarabacak, Rioja, and Valev (2012), who classified private credit into corporate and household credit, suggested that although household credit could boost economic growth in the short-run through consumption growth, most types of consumer loans were short-term, and the role of consumer credit tended to be small in the long-run. They found that the positive relationship between household credit and long-term economic growth was not significant, while the relationship between corporate credit and long-term economic growth was robust (from productivity growth and the improvement in income distribution).

Furthermore, an increase in household debt could reflect greater access to finance of the household sector. In this regard, the growth in household loans could represent greater access to consumer credit of household sector (Debelle, 2004; Kang & Ma, 2009). When households had more access to credit, it meant they had greater ability to obtain more funds from financial institutions to finance themselves (Debelle, 2004). Moreover, Kang and Ma (2009) pointed out that credit growth could also reflect financial liberalisation, including financial institutions’ portfolio diversification. This portfolio diversification implies that financial institutions are able to allocate their assets with the highest expected returns (Kang & Ma, 2009). These benefits could lead to more integration in the credit markets and banking system. Next, risks will be discussed concerning a high-level of household debt.

2.2.2 Negative impacts of household debt on household’s financial stability and the macroeconomy

High household debt, however, can cause harm to both the welfare of individual and the economy at large. Focusing on the negative impacts on the welfare of individual

20 household first, in many countries, high growth in household debt usually associates with a low rate of savings (Debelle, 2004; Barba & Pivetti, 2009). For example, in Finland, Norway, Sweden, and the UK, Debelle (2004) stated that there was a sharp increase in household debt along with a decline in household saving rates during the 1980s. In the case of the US, Barba and Povetti (2009) outlined the statistics of the saving rate of the household sector before the global financial crisis in 2008. They showed that the household-saving-to-disposable-income ratio in the US dropped from 10% in 1980 to less than 1% by 2007. While, during the same period, the household- debt-to-disposable-income ratio increased from 72% in 1980 to more than 130% by 2007.

High indebtedness, coupled with a low saving rate, made the household sector more vulnerable to income shocks and interest-rate shocks (Debelle, 2004; Endut & Hua, 2009). Debelle (2004) claimed that households with higher exposure to debt but low financial cushions from savings tended to have difficulties in repaying their debts when they become unemployed, or, face an unanticipated increase in interest rate. Therefore, making decisions based on interest rates was a challenging issue for a central bank (Endut & Hua, 2009). When the central bank announces a ‘hike’ in policy interest rates, the results are higher market interest rates and a concomitant increase in households’ debt service burden. Furthermore, countries with more variable interest rates would be more sensitive than countries with more stable rates, especially in the case of mortgage loans, which are the largest proportion of loans (Debelle, 2004).

Next, regarding the adverse effects on the economy, the fragility of the financial sector, and the weakness of the economy from excessive household debt more broadly remains to be rigorously assessed by many researchers (Barnes & Young, 2003; Cecchetti et al., 2011). They have suggested that a high level of household indebtedness, along with a rise in the unemployment rate, could lead to a collapse in demand through unwinding in both consumption and investment. The assumption behind this is that debt repayments were missed only as a last resort: that is, a fall in income led to a drop in investment first, then in consumption, and finally in debt repayments.

For instance, in the case of the UK, Debelle (2004) explained that although the rising household debt promoted economic growth and house prices in the late 1980s, after that the high level of household debt dragged-down the UK economy during the 21 economic downturn in the early 1990s. The boom in the housing market in the late 1980s caused a sharp increase in mortgage credit and speculation in the housing sector. This speculation caused house prices to soar up exceed the fundamental values. Once people in the UK faced income shock from unemployment under the high interest rate environment, they were not able to pay-back their mortgage loans. Some of them needed to sell their houses and could not smooth their consumption. This situation led to a considerable drop in house prices and economic recession in the next period.

In the case of the US, Martin (2011) also claimed that the sub-prime mortgage boom during 2001 to 2006 led to the bubble bursts in house prices and the global financial crisis in the next two years. The financial crisis caused broad adverse effects on the real economy and the financial markets (Claessens et al., 2010; Longstaff, 2010; Aloui et al., 2011). Claessens, Dell’Ariccia, Igan, and Laeven (2010) and Martin (2011) stated that the crisis caused an economic downturn in both advanced and emerging countries which led to less economic activities resulted in higher unemployment rates. Debelle (2004) supported the argument that the economic downturn caused households to experience negative income shocks, and it could lead to a drag in consumption, if households could not get more funds to smooth consumption. In addition, the IMF (2012) investigated the relationship between household indebtedness and the period of recession and identified that countries with higher household debt-to-income ratio were more likely to face greater loss in consumption3 and a longer period of recession. Moreover, regarding the effects on financial markets, the global financial crisis in 2008 caused more volatility in financial markets (Longstaff, 2010). Longstaff (2010) claimed that the volatility led to a widespread vulnerability in the financial sector and prevented financial intermediaries from operating efficiently.

In addition, according to the negative impacts of excessive household debt on the macro-economy and financial stability mentioned above, some scholars have tried to find the certain level of household debt-to-GDP that can cause the adverse effects on economic growth (Terrones & Mendoza, 2004; Barajas et al., 2007; Cecchetti et al., 2011). Although there is no certain threshold, Cecchetti, Mohanty, and Zampolli (2011) suggested the level of household debt to GDP at 85% could flip an economy into

3 ”The consumption loss in 2010 is the gap between the (log) level of real household consumption in 2010 and the projection of where real household consumption would have been that year based on the pre-crisis trend. The pre-crisis trend is defined as the extrapolation of the (log) level of real household consumption based on a linear trend estimated from 1996 to 2004” (IMF 2012). 22 financial fragility. Furthermore, Terrones and Mendoza (2004) and Barajas, Dell’Ariccia, and Levchenko (2007) defined a credit boom situation as a period when the real private credit or the credit-to-GDP ratio exceeded its trend ‘significantly’.

Some researchers also raised the issue of household ‘debt sustainability’ (Betti et al., 2007; Barba & Pivetti, 2009; Cecchetti et al., 2011). Betti, Dourmashkin, Rossi, and Yin (2007) defined household debt sustainability as the circumstance where households could repay their debt without hurting the necessary expenditure for their livelihoods. Many researchers supported the idea that household debt unsustainability could hurt the macro-economy and overall financial stability (Rinaldi & Sanchis-Arellano, 2006; Barba & Pivetti, 2009; Davies, 2009; Dynan & Kohn, 2007; Kang & Ma, 2009; Subhanij, 2009).

2.2.3 Policy responses to rising household debt in the region of Southeast Asia

This research will not consider policy responses to rising levels of household debt, drawing on the experience of several nations from Southeast Asia, of which Thailand is the focus. In Malaysia, for example, financial institutions’ loans have been increasingly concentrated in the household sector after the Asian financial crisis in 1997, and Bank Negara Malaysia has taken into account the impact of an increase in the policy interest rate on household debt servicing ability and consumption (Endut & Hua, 2009). Endut and Hua (2009) suggested that, as higher household indebtedness makes households more sensitive to interest-rate hikes, the central bank should impose forward-looking monetary policy. They believed that the smooth and steady changes in monetary policy (i.e. the changes in the policy interest rate) based on timely and frequent data would let households have accurate expectations and have enough time to prepare for the future debt repayments. Endut and Hua (2009) also believed that such a forward-looking policy would lead to less economic fluctuation.

In the case of The Philippines, a higher amount of non-performing loans for consumer loans from 2005 to 2007 led to the tightening of credit standards by the authorities (Tan, 2009). Regarding this case, Tan (2009) said that the Central Bank of the Philippines had tried to set a central credit information system by establishing a national credit bureau. The credit bureau, then, can provide reliable information on

23 borrowers’ credit profiles to banks. This set of information allows banks to review the credit profiles and to evaluate the credit risk of each borrower (Tan, 2009).

In the case of Thailand, more mortgage loans with both higher loan-to-value ratio and loan-to-income ratio in recent years have raised concerns with the authorities about ‘search-for-yield’ behaviour (Bank of Thailand, 2018b). A low interest rate environment, as a result of a crisis, could lead to more risk-taking behaviours and macro-prudential policies e.g. loan-to-value ratio are considered to be the complement instruments for authorities (Claessens et al., 2010). The Bank of Thailand (2018b) claimed that the growth in mortgage loans during recent years could facilitate the behaviour of ‘search for yields’ from rental yields and capital gains, especially in cases where people have bought second or third houses. Eventually, the Bank of Thailand has imposed new rules for mortgage loan practices (effective 1 April 2019) stating that, for houses valued less than THB 10 million, the loan-to-value of second-mortgage contracts was limited to 90% if the first-mortgage contract had been paid for three years longer, and the limit was 80% if the first-mortgage contract had been paid for less than three years. The loan-to-value limit for all third-and–subsequent-mortgage contracts was limited to 70%. These current limits are less than the previous one, which was set at 95% for all mortgage contracts of houses valued less than THB 10 million.

From all the evidence provided here, the analysis of the over-indebtedness phenomenon and household debt performance is, therefore, a key issue in the promotion of household debt sustainability and economic sustainability (DeVaney, 1995). Moreover, a good understanding of the development of household debt, including debt characteristics, will help authorities to deal with household debt situation efficiently (Barba & Pivetti, 2009; Endut & Hua, 2009).

Next, this thesis will examine the theoretical hypotheses of household indebtedness that explain why, and how, households become indebted. This relates to the analysis undertaken in Chapter 4.

2.3 Alternative explanations of household indebtedness

There are four main related theories that explain household indebtedness: the Permanent Income Hypothesis, the Life-Cycle hypothesis, the theory of credit rationing, and behavioural finance theory. Regarding the first two hypotheses (the Permanent Income Hypothesis and the Life-Cycle Hypothesis), which are based on neoclassical 24 economic theories, Friedman (1957) and Modigliani (1966) explained that individuals were savers, who could build their wealth when their spending was less than earnings. Then, they are able to invest the savings for growth over the lifetime. Whereas, individuals were borrowers, who needed to get some loans, when their income was less than expenditure.

According to the Permanent Income Hypothesis (Friedman, 1957, pp. 20-37), at any given point of time, individuals’ consumption levels are not only determined by their current income, but also by the expected levels of their income over their lifetime, which is called the permanent income. This theory posits that individuals borrow when transient income falls below the permanent income, save when the opposite takes place, and so smooth their consumption over time. Two results follow from the above: (i) annual income is more volatile than annual consumption; and (ii) savings and borrowings net out over the lifetime. The first result raises an issue about welfare. Friedman (1957) claimed that individuals preferred stable consumption to volatile consumption. The second result was linked to an inter-temporal budget constraint where lifetime income equalled lifetime consumption.

Regarding the second hypothesis, which is related to the first, Modigliani (1966, pp. 160-217) (who initiated the Life-Cycle Hypothesis) explained that agents attempted to plan their consumption and savings over their life cycle. Considering a given level of consumption, households are borrowers early and late in life, when their income is less than their expenditure. Otherwise, during their working years, they are savers. Individuals can accumulate their wealth for the period of dis-saving (Modigliani, 1966). Figure 2.2 provides three stages of life: young age, working age and old age (retirement). In the model, individuals cannot earn income prior to reaching working age, so they dis-save during childhood to fund their consumption. Once they start working, they are able to start to save for their retirement.

Figure 2.2 Life-Cycle Hypothesis

Income (Y)/ expenditure (C)

Saving Average LT income

Expenditure level

Borrowing Use of saving

Age Young age Working age Old age (Y < C) (Y > C) (Y < C)

In addition to the previous hypotheses (the Permanent Income Hypothesis and the Life-Cycle Hypothesis), many scholars have investigated the consumption choices of consumers who are also liquidity constrained. All these investigations show some empirical support for the Permanent Income Hypothesis and the Life-Cycle Hypothesis, but they have argued that liquidity constraints were also needed to account for certain aspects of consumers’ behaviour (Hall, 1978; Campbell & Mankiw, 1989; Deaton, 1989; Attanasio, 1999). In particular, consumers seemed unable to smooth consumption as much they would like because of the liquidity constraints under which they operated. In part, this constraint manifests as a difference between the interest rate of borrowing and the rate of lending (saving). In a perfectly competitive community with unlimited access to credit, the inter-temporal budget constraint is a straight line, where the saving rate is equal to the borrowing rate (Figure 2.3). However, in reality, the borrowing rate is normally higher than the saving rate. Figure 2.4 shows that inter-temporal consumption, with a higher borrowing rate, will lead to a kinked budget line, where the absolute terms of the slopes of the kinked budget line represent the interest rates in the market (Stiglitz & Weiss, 1981; Attanasio, 1999).

Figure 2.3 Intertemporal Consumption

Consumption in period 2 (C2) Budget line when consumer is a saver (C < M ) 1 1

Budget line when Income in consumer is period 2 (M ) 2 a borrower

(C1 > M1)

Consumption Income in in period 1 (C1) period 1 (M1)

Figure 2.4 Intertemporal Consumption with a Kinked Budget Line

Consumption in period 2 (C2) Budget line when consumer is a saver

(C1 < M1)

Budget line when Income in consumer is period 2 (M ) 2 a borrower

(C1 > M1)

Consumption Income in in period 1 (C1) period 1 (M1)

Additionally, in case of credit rationing, Flemming (1973) claimed that the imperfection in financial markets could affect the borrower’s debt service burden. According to the literature on asymmetric information in credit markets, such as Stiglitz and Weiss (1981), the imperfect information in markets prevents lenders from having

27 complete knowledge of the borrowers’ capacity and willingness to pay debts. Hence, the lenders will set the lending rates taking this asymmetry of information into account. The interest rates that issue from this information-constrained optimisation program aim to sort potential borrowers. They also affect the borrowers’ actions. One implication of the credit-constrained literature is that the interest rate with imperfect information does not clear the market for loans (where there is excess demand). Linking this issue to intertemporal consumption, if there are credit constraints (Figure 2.5), Deaton (1989) and Attanasio (1999) stated that consumers would be not able to secure loans from the formal credit market. The lack of formal credit access would affect consumers’ intertemporal consumption choices and their capacity to save.

Figure 2.5 Intertemporal Consumption with Liquidity Constraints

Consumption in period 2 (C2) Budget line when consumer is a saver

(C1 < M1)

Income in

period 2 (M2)

No access to formal sources of credit

Consumption Income in in period 1 (C1) period 1 (M1)

In a world of imperfect information, the interest rate is not the sole instrument used to allocate credit. The lender imposes several other conditions with regards to the issuing of loans. This is because higher interest rates may encourage adverse selection and/or moral hazard, and so may lower, rather than increase, loan margins. In the current context, adverse selection arises when the principal - the lender - lacks all the information needed to fully distinguish high-risk from low-risk agents - i.e. borrowers; and moral hazard arises when agents have an incentive to underperform - i.e. not repay the loans - when subject only to a budget constraint. In order to abate both adverse selection and moral hazard, non-price instruments are needed to allocate loans 28 optimally. Consequently, banks (lenders) utilise such measures as credit histories and collateral as instruments in rationing credit.

Lastly, this study will consider the assumptions of rationality and perfectly foresight on the part of borrowers, who are presumed to know their permanent income. These assumptions are often made to simplify the analysis, but research on behavioural finance puts doubts on the validity of these suppositions. Many researchers have presented evidence in support of the proposition that financial imprudence is one of the main drivers of household indebtedness (Anderloni & Vandone, 2010; Disney & Gathergood, 2012). Anderloni and Vandone (2010) claimed that financial imprudence resulted from low financial literacy and leads to poor financial management. For example, some households cannot perceive the actual cost of borrowing, and they obtain loans regardless of the debt burden.

In addition, Disney and Gathergood (2012) have raised the point that people with low financial literacy were more prone to lack self-control, thus they consistently and persistently underestimated their debt burdens: referred to as the concept of ‘hyperbolic discounting’ behaviour in the behavioural finance literature. Laibson (1997) defined hyperbolic discounting behaviour as the behaviour of consumers who had a high preference for consumption and consequently discount future consumption heavily. Moreover, Gathergood (2012) also supported the argument that hyperbolic discounters tended to persistently underestimate the cost of debt repayment in terms of foregone future consumption. Consequently, hyperbolic discounters, who bring their consumption forward by borrowing, usually have less savings and face debt repayment problems because they are ‘caught by surprise’ in the future by the amount they have to repay (Laibson, 1997; Gathergood, 2012).

In summary, the fundamental model from the literature is the model of the Permanent Income Hypothesis and the Life-Cycle Hypothesis. Although it has some empirical support, it is known to be problematic. The model assumes perfect capital markets and perfect rationality (i.e. rational expectations). However, it has been known for some time that liquidity constraints apply to households. This means that their ability to borrow and the interest rate they face depends on a number of factors. Therefore, this research has tried to ascertain these in Thai households. Finally, recent literature has confirmed the use of heuristics and other sub-optimal decision rules. This

29 dissertation examines which households are particularly prone to these failures of rationality. Next, the measures of household indebtedness are explained.

2.4 Household indebtedness measurement

This section clarifies how researchers measure the level of household indebtedness in both quantitative and qualitative ways. There are both continuous and dichotomous indicators of household indebtedness (Keese, 2009). Focusing on the continuous indicators, Keese (2009) examined the measure of household debt performance that could be a gauge of financial fragility. The ratio of debt to asset (in terms of stocks: Equation 2.1), and the ratio of income after debt service payments to non-discretionary income (in terms of flows: Equation 2.2) could be applied in this sense (Betti et al., 2007; Gumy, 2007; Keese, 2009). DeVaney (1994) suggests the term of household insolvency as “the failure to submit the timely repayment of debts as they mature” which can be reflected from the negative net worth by using the debt to asset ratio (Equation 2.1 when DAR greater than 1). Whereas, for the ratio of income after debt service payments to non-discretionary income, Keese (2009) used non- discretionary income as the minimum subsistence level of income, and the ratio could imply the distance between income after debt service payments and the subsistence income. Keese (Ibid.) claimed that this ratio was a proxy of income cushion after the debt service payments.

퐷 DAR = (2.1) 퐴 where 퐷퐴푅 = Debt-to-asset ratio

퐷 = Stock of debt

퐴 = Stock of assets

퐼 − 퐷푆푃 DP1 = (2.2) 푁퐷퐼 where 퐷푃1 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

푁퐷퐼 = Non-discretionary income

This study will next review the dichotomous indicators. Many researchers have attempted to pick a threshold to determine over-indebtedness, but there is no consensus on how this threshold is constructed (Anderloni & Vandone, 2008; Keese, 2009; European Commission, 2010). They have tried to clarify the definition of over- indebtedness, unfortunately, there is no consensus on the meaning of the term. When is a household over-indebted? The answer to this question is not always clear. Measures used to signal over-indebtedness are related to subjective and objective indicators (Keese, 2009). Subjective over-indebtedness is often reported by the borrowers’ perceptions of difficulties in their debt service repayments (Anderloni & Vandone, 2008; D’Alessio & Iezzi, 2013). In contrast, objective over-indebtedness is measured using data drawn from the households’ financial statements and balance sheets (Haas, 2006; Betti et al., 2007; Keese 2009; European Commission, 2010; Gathergood, 2012; D’Alessio & Iezzi, 2013).

The measure of subjective over-indebtedness is drawn from household level surveys. One of the most popular primary data sources for this approach is household surveys (European Commission, 2010; Liv, 2013) which can be collected from the selected sample of households or the representatives of the whole household population (Anderloni & Vandone, 2008). Anderloni and Vandone (2008) denoted subjectively over-indebted households as the selected households who declared that they have difficulties on payment obligations or commitments. They proposed three criteria of these difficulties: 1) only difficulties on debt obligations or commitments, 2) difficulties on all financial obligations or commitments, including utility bills, insurance bills, tax and rent; and 3) difficulties on all financial obligations or commitments which are related to household’s debt. Whereas, according to the European Commission (2010), subjective over-indebtedness could, generally, be identified as the perception of having difficulties on all aspects of household’s financial commitments that were recurrent expenditure. Moreover, the European Commission (2010) claimed that, for subjectively over-indebted households, there might be no possible way to meet the commitments by getting more loans. Additionally, D’Alessio and Lezzi (2013) also labelled the perception of having heavy debt service burden as another indicator of subjective over- indebtedness.

Regarding the challenges of collecting reliable responses by using official household surveys, the subjective responses were individual judgements on the 31 questions asked in the survey. This raised questions on the reliability of the collected information (European Commission, 2010). However, Betti, Dourmashkin, Rossi, and Yin (2007) argued that although responses to questions about debt perception were subjective assessments of the respondents, there was no evidence that these responses untruthful.

On the other hand, for the objective over-indebtedness measures, macro-data (Haas, 2006; European Commission, 2010) and micro-data (Betti et al., 2007; Keese, 2009; Gathergood, 2012; D’Alessio & Lezzi, 2013) could be collected with accounting statistics. By using quantitative models, objectively over-indebted households were measured as those with the inability to repay their debt corresponding to certain levels of indebtedness (Betti et al., 2007). For example, focusing on the dichotomous indicators of objective over-indebtedness, in the case of Germany, households with income after debt service payment less than non-discretionary income (Equation 2.2 when DP1 less than 1) (Keese, 2009) or living standard (Equation 2.3 when DP2 less than 1) (Haas, 2006) were deemed to be over-indebted. Non-discretionary income for example, child allowances and old age pensions could be referred to as the minimum subsistence income level for each household.

퐼 − 퐷푆푃 DP2 = (2.3) 퐿푆 where 퐷푃2 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

퐿푆 = Living standard

Additionally, D’Alessio and Iezzi (2013) selected over-indebted households by using two criteria: 1) households with a ratio of monthly debt service payments to monthly income of more than 30% (Equation 1.1 when DSR greater than 0.3) or 2) households with monthly income after debt service payments less than the respective poverty line (Equation 2.4 when DP3 less than 1). Commonly, the poverty line reflected the minimum expenditure for essential goods and services for life (Haas, 2006; D’Alessio & Lezzi, 2013). Furthermore, the arrears on financial commitment could be used to identify over-indebted households. In the case of the UK, one-month and three-

32 month delinquency on debt was claimed as another proxy of objective over- indebtedness (Gathergood, 2012).

퐼 − 퐷푆푃 DP3 = (2.4) 퐻푃퐿 where 퐷푃3 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

퐻푃퐿 = Household’s poverty line

HPL = PL x HS (2.5) where 퐻푃퐿 = Household’s poverty line

푃퐿 = Poverty line

퐻푆 = Household size

Next, the previous studies of the drivers of household indebtedness will be reviewed to gain insight into a common set of factors.

2.5 Research on the drivers of household indebtedness

As indicated, this section will review the extant research on the drivers of household indebtedness and identify common variables used by scholars. These will be linked with the variables of the related hypotheses, which are the Permanent Income Hypothesis, the Life-Cycle hypothesis, the theory of credit rationing, and behavioural finance theory.

First, many empirical studies investigating the drivers of household debt performance and over-indebtedness have been done based on the Permanent Income Hypothesis and the Life-Cycle Hypothesis, and many scholars have provided evidence in support for these hypotheses (Vante, 2006; Betti et al., 2007; Anderloni & Vandone, 2008; Keese, 2009). For example, some scholars reported that young individuals with low income tended to be over-indebted (Vante, 2006; Betti et al., 2007). In the case of European countries, Betti, Dourmashkin, Rossi, and Yin (2007) applied the measure of subjective over-indebtedness by using European household survey data. They reported

33 that young individuals on low incomes had a higher probability of being subjectively over-indebted. Similarly, for Norway, after controlling for household’s characteristics, the main difference between households with sound and poor financial status were their average income level (Vante, 2006). Moreover, Anderloni and Vandone (2008), who integrated an analysis of the literature on household over-indebtedness, concluded that heads of households aged between 30-39 years seemed to have higher chances of being over-indebted.

Notwithstanding this, in the case of Germany, Keese (2009) argued that income levels might not be the only driver of household debt difficulties. Expenditure and income shocks, such as those arising from childbirth and family breakdown, respectively, could also lead to severe indebtedness. Keese (Ibid.) employed Germany’s panel household survey (the German Socio-Economic Panel Surveys of 2002 to 2007) and conducted dynamic analysis of the determinants of severe household indebtedness. This latter finding from Germany leads to other significant factors such as the number of children in the family and head of household’s marital status that can affect the level of household debt. Betti, Dourmashkin, Rossi, and Yin (2007), and Xiao and Yao (2014) said that single-adult households and young heads of households with many children were also prone to over-indebtedness.

Second, with regards to the effects of liquidity constraints, Betti, Dourmashkin, Rossi, and Yin (2007) suggested that lack of access to a formal source of credit could result in being subjectively over-indebted. They claimed that, in the case of European countries, people with less access to formal loans tended to express more difficulties in making debt repayments. In particular, if these people had myopic consumption behaviour (i.e. hand-to-mouth consumers with high expenditure-to-income ratios), they would struggle to finance themselves and deal with negative shocks (e.g. unemployment). Thus, the proxies of credit constraints, coupled with consumption behaviours, were mentioned as significant determinants of consumer over-indebtedness (Anderloni & Vandone, 2008).

Furthermore, Anderloni and Vandone (2008) raised the significance of a household’s economic status, which includes employment status, occupation, and income level, as determinants of credit scoring. There was a close correlation between a chance of being over-indebted, as indicated by the quantitative measures, and credit scores. The latter (i.e. credit score) is based on a credit database, which is collected from 34 borrowers’ credit profiles. Thus, a borrowers’ repayment history can be indicated high or low credit risk (default risk). Anderloni and Vandone (2008) explained that credit- scoring models commonly utilised socio-demographic characters (e.g. age, gender, education level, and marital status), economic status of borrowers (e.g. employment status, occupation, and income level), and historical credit data of borrowers (i.e. the record of credit repayments). The models then calculated the predicted probability of credit default of each borrower.

Regarding the relationship between employment status and the probability of obtaining loans, according to lending practice, full-time employees with secure income will have a greater chance of obtaining credit from formal sources of credit with cheaper interest rates compared with those who are jobless (Anderloni & Vandone, 2008). Moreover, some researchers believe that persistent unemployment can amplify the effect of income shocks on household debt performance because of the interaction of the lack of income and less chance of getting credit from formal sources of credit (Anderloni & Vandone, 2008; Keese, 2009).

In addition, some researchers used types of loans, e.g. mortgage loans (Keese, 2009), credit card, and education loans (Xiao & Yao, 2014) as the other factors to explain household debt performance. These types of loans were claimed as the causes of having poor debt performance. Focusing on the role of mortgage loans, Keese (2009) argued that although homeowners could benefit from asset accumulation, they might be unable to manage their housing service burden, which was “lumpy”, compared with other types of loans.

Finally, regarding behavioural finance, some researchers have tried to find the relationship between households’ financial literacy and their credit status (Lusardi & Tufano, 2009; Disney & Gathergood, 2012). In the case of the UK, Disney and Gathergood (2012) employed Ordinary Least Squared regression by using UK household surveys, including financial literacy questions4, and pointed out that households with poor financial literacy tended to have a higher cost of credit financing. Moreover, in the case of the US, Lusardi and Tufano (2009) undertook an empirical study with a national sample of Americans and found that a household’s debt literacy was strongly related to the probability of becoming objectively over-indebted. They also

4 Financial literacy questions include simple interest question, interest compounding question and minimum payment (in the case of using credit card) question 35 asserted that households with low debt literacy did not perceive the real cost of borrowing and were also unable to assess their financial positions.

In addition to the issues of financial literacy, there were some issues of financial management. However, according to the extant literature, there are only preliminary results of the positive relationship between poor financial management and the likelihood of being over-indebted (Anderloni & Vandone, 2008; Lusardi & Tufano, 2009). Anderloni and Vandone (2008) and Gathergood (2012) found that poor financial management could make households suffer from financing themselves and then become objectively over-indebted. To identify poor financial management, they suggested that financial management depended on the household’s attitude, which was related to psychological aspects. Gatherhood (2012) identified households with impulsive spending or hyperbolic discounting behaviour as households with poor financial management (i.e. people with self-control problems). He claimed that, in the case of the UK, impulsive spenders statistically tend to be over-indebted. The question for impulsive spending behaviour is: “are you impulsive and tend to buy things even if you cannot really afford them?”. However, Gatherhood (2012) could not find a significant relationship between hyperbolic discounting behaviour and over-indebtedness. The question for hyperbolic discounting behaviour is: “are you prepared to spend now and let the future take care of itself?”. Based on these preliminary findings, it is appropriate to test the significance of the role of financial management in examining household indebtedness. The proxies of a household’s financial management would be added to the studies in this thesis. More details of this issue will be explained in the subsequent section on research gaps.

According to the existing literature, most studies on household over- indebtedness focus on only one measurement. Many researchers focus on only the objective over-indebtedness (DeVaney, 1994; Haas, 2006; Gumy, 2007; Gathergood, 2012; Disney & Gathergood, 2012), whereas others focus only on subjective over- indebtedness (Betti et al., 2007; Liv, 2013; Muthitacharoen et al., 2015).

There are a few researchers who have tried to clarify the differences between subjective and objective over-indebtedness (Keese, 2010). According to Keese (2010), the analysis of subjective over-indebtedness covers issues beyond household financial status. It is related to non-financial factors such as an expectation of the economic situation, gender, social status, and type of loans. In other words, Keese (2010) believed 36 that the analysis of only objective over-indebtedness might not cover the latent debt burden of each family. To the best of this research’s knowledge, the study by Keese (2010) was the only study that investigated the differences between these two types of over-indebtedness. A household’s economic situation is a factor that can cause differences in perceptions of debt difficulty. In the case of Germany, Keese (2010) used data from the German Socio-Economic Panel and pointed out that subjective over-indebtedness was not only related to the current financial status but also related to the expectations of personal and general economic conditions. He pointed out that households would have more concerns about their debt commitments if they expected worse economic conditions, such as higher future unemployment. A similar focus on European countries by Gerogarakos, Lojschova, and Ward-Warmedinger (2010) looked at debt distress from mortgage loans and their research demonstrated that households’ perceptions of debt distress could be affected by working status and marital status. Moreover, a higher dependency ratio (i.e. the number of children and non-working adults in each household) could affect household financial decisions and their sense of debt distress (Jianakoplos & Bernasek, 1998; Lenton & Mosley, 2008; Georgarakos et al., 2010). Gender was another factor that raised the concern of individuals regarding their ability to service debt. After controlling the level of the debt service ratio, Keese (2010) presented evidence in support of the proposition that women were more (fiscally) conservative than men. Jianakoplos and Bernasek (1998) and Lenton and Mosley (2008), who determined the differences in the household financial decision and debt concern respectively, affirmed that women tended to be more risk-averse than men when they come to taking on debt. However, there was no general consensus on this proposition. Del Rio and Young (2005), for example, used the British Household Panel Survey to examine self-assessed debt burden and argued that gender did not affect the perception of debt burden. They claimed that the type of loans mattered on the perception of the debt burden and, in particular, those of unsecured loans was important. According to these findings, it is useful to study the differences between subjective and objective over-indebtedness in developing countries where there are more heterogeneities in each household. The analysis in the case of Thailand will be carried out in Chapter 5.

Concluding the review of research on the drivers of household indebtedness, this thesis will now survey the research on household indebtedness in Thailand to clarify the extant studies in this area.

2.6 Research on household indebtedness in Thailand

This part will review the previous research on household indebtedness in Thailand, focusing on two issues: (i) household credit prudence, and (ii) household debt sustainability. Firstly, focusing on the credit prudence of the household sector in Thailand, the Kenan Institute Asia has used the official household surveys of 2013, together with macro data of household financial status. Kenan Institute Asia (2015) concluded that household debt in Thailand had increased significantly, whereas household saving tended to decline over the previous ten years. It also claimed that poor financial literacy and management, including a lack of adequate financial access, could make households more vulnerable by weakening their financial standing. The most vulnerable household groups who had the greatest chance of facing financial difficulties were agricultural households and blue-collar workers (Kenan Institute Asia, 2015). These groups of households were also noted as households with low debt tolerance by Muthitacharoen (2016), who estimated the debt tolerance levels of households with different social classes (i.e. different occupations) using micro data from household surveys and ranked the debt tolerance level for each social class. He defined household debt tolerance as the households’ ability to obtain and manage debt with no concerns, and pointed out that farmers and general workers with insecure income had the lowest level of debt tolerance. Conversely, professional households had the highest debt tolerance. Apart from identifying households with financial vulnerability by occupation, household income group is another factor for examining households with credit risk. Subhanij (2009) examined the credit condition of each income group and provided evidence to suggest that households with lower income had a higher rate of delinquency compared with households with higher income. Moreover, Subhanji (Ibid.) also claimed that poor households tended to face credit constraints which impeded access to the formal credit market.

Additionally, some researchers have tried to examine household credit prudence using the data from the National Credit Bureau of Thailand (Chantarat et al., 2017). The National Credit Bureau collected data on credit by households from most of its member

38 financial institutions. The National Credit Bureau provided other micro data from its database which stated borrowers’ ages and addresses. Chantarat, Lamsam, Samphantharak, and Tangsawasdirat (2017) reported that, as of the mid-year of 2016, the National Credit Bureau’s database (i.e. the credit data from the members of the National Credit Bureau) covered 87% of total household loans from the formal sector. Using this database, they found that the median value of debt per retail borrower was around THB 147 000 by mid-2016. This value was double that of 2009, while debt headcount ratio (defined as the ratio of borrowers who had a record with the National Credit Bureau membership to the total population) remained stable at around 20%. On the spatial distribution of this debt by region from 2009 to mid-2016, Chantarat, Lamsam, Samphantharak, and Tangsawasdirat (2017) reported that households in the North and the North-east of the country had a higher amount of debt per borrower than households in other regions. By age group, they suggested that people aged 36-60 years had the highest amount of debt per borrower, whereas those of working age but 25-35 years-old had the sharpest increase in the average debt per borrower. This suggested that people entering the workforce quickly accumulated debt, but that this stabilised once they had settled in their careers - a finding that was consistent with the predictions of life-cycle and permanent income hypotheses.

The debt-picture presented by Chantarat, Lamsam, Samphantharak, and Tangsawasdirat (2017) is only partial. This is because there are several limitations of the credit data from the National Credit Bureau. Firstly, the loan information from the National Credit Bureau did not cover all financial institutions and did not include informal household loans. This is in contrast to the information from Thailand’s Household Socio-Economic Surveys (used in this thesis) which is administered by the National Statistical Office and covered all types of loans. Secondly, the number of the National Credit Bureau’s members (i.e. financial institutions that reported credit data to the National Credit Bureau) had not been constant yet. Therefore, the changing number of the National Credit Bureau’s members makes researchers more careful when doing dynamic analysis using this data set. Thirdly, there was no information on borrowers’ income and debt service burden from the National Credit Bureau’s database, so the debt-to-income ratio and debt service ratio, which are common indicators used by other researchers, could not be calculated (Keese, 2009; D’Alessio & Iezzi, 2013; Muthitacharoen et al., 2015)

On thresholds regarding household debt sustainability, some researchers have investigated the effect of high debt service burdens on household debt sustainability (Muthitacharoen et al., 2015). Household debt sustainability is the circumstance that households can repay their debt without falling below the poverty line (Betti et al., 2007). Muthitacharoen, Nuntramas, and Chotewattanakul (2015) applied a subjective over-indebtedness measurement to indicate the threshold for a household’s debt service ratio. Thailand’s Household Socio-Economic Surveys (conducted by the National Statistical Office) and the Bank of Thailand’s supplement household survey of 2013 were used as Thailand’s official household surveys. They estimated probit regression by using the perception of having a concern about the next debt repayment as the measure of subjective over-indebtedness. After controlling for the head of household’s age, household’s social class and region, they reached the conclusion that the debt service ratio at around 40% presented a significantly higher possibility of having difficulties on debt repayment. Muthitacharoen, Nuntramas, and Chotewattanakul (2015) suggested that the debt service ratio at 40% could be defined as heavy debt service burden.

In addition, according to Muthitacharoen, Nuntramas, and Chotewattanakul (2015), a higher household debt service burden not only corresponded to a higher probability of financial difficulties but also taxed future consumption growth. Using Thailand’s Household Socio-Economic Surveys in 2009 to 2013 (which were not panel surveys), Muthitacharoen, Nuntramas, and Chotewattanakul (2015) grouped households into 776 districts and run a linear regression of a change in consumption under the assumption of 10-percentage-point changes in income. They found that households in the top third of a change in the debt service ratio during 2009 to 2011 (27%) had the lowest consumption growth in 2013 at 5.7%. By using micro-simulation, they also identified that households in the bottom income quintile were more sensitive to interest rate shocks than the others because of a higher proportion of variable-rate debt. According to the extant studies of household indebtedness in Thailand to date, empirical dynamic analysis of debt performance and household over-indebtedness has not yet been done.

In conclusion, most previous research on household indebtedness is based on neoclassical theories and has been done using the case of developed countries. However, there are just a few studies on the differences between subjective and objective over-indebtedness. Therefore, Thailand can be a good case study regarding 40 household debt in developing countries. Next, this dissertation will summarise the research gaps that remain in this area.

2.7 Research gaps

Regarding research gaps in this research area, from the literature survey, this thesis has found three main gaps:

The relationship between financial management and household indebtedness

To investigate the role of behavioural finance in determining household indebtedness, based on existing literature on the drivers of household indebtedness, there are few studies of the correlation between household financial management and indebtedness. Moreover, the results from the previous studies were preliminary findings (Anderloni & Vandone, 2008; Lusardi & Tufano, 2009). Therefore, this thesis attempts to construct and add three related variables to confirm the role of household financial management in explaining household borrowing behaviour. The first added indicator is a proxy of keeping income-and-expenditure accounts; a dummy variable, which is created based on households’ income-and-expenditure account status (0 = N, 1 = Y), will be used to capture the effect of households’ income-and-expenditure management (i.e. regular doing income-and-expenditure accounts) on debt status. Secondly, the dummy variable based on having problems with other financial commitments (rent and utility bills) over the past twelve months (0 = N, 1 = Y) will be generated. If there is a robust finding of the impact of this type of financial default, this factor can be used as a leading indicator of default risk. Thirdly, a proxy of hyperbolic discounting behaviour will be created relying on the statement: “I am happy with spending now more than saving for the future”.

The above added variables will be used to statistically prove the significance of behavioural finance which is a complement to other two economic theories, namely neoclassical economic theory (i.e. the Permanent Income Hypothesis and the Life-Cycle hypothesis) and the theory of credit rationing and liquidity constraints. The behavioural variables are employed to more fully characterise household indebtedness in the case of Thailand. As this has not been done before, this is a contribution to the literature.

The empirical study of both subjective over-indebtedness and objective over- indebtedness

According to the previous studies of household over-indebtedness, it seems that relying on only subjective over-indebtedness or objective indebtedness alone is not enough in order to paint a complete picture of the household debt phenomenon. However, most studies of household over-indebtedness focus on either subjective or objective measures of over-indebtedness. Therefore, this gap in the literature has been pursued using household data on Thailand, as will be explained next. In this thesis, both subjective over-indebtedness and objective over-indebtedness measurements have been used to test the correlation between the two and investigate reasons for any observed divergences. This analysis can make the clear distinction between subjectively over- indebted households and objectively over-indebted households. Moreover, the analysis of the differences between these two indebtedness measurements is significant in the sense of credit policy implications.

The panel analysis of household indebtedness in Thailand

In the case of Thailand, the current official household surveys are not panel surveys (i.e. the surveys are repeated cross-sectional surveys), so existing studies of household indebtedness using these surveys are based on static analysis, which cannot completely control for household heterogeneity (Kenan Institute Asia, 2015; Muthitacharoen et al., 2015; Muthitacharoen, 2016). This is a limitation in identifying the true determinants of household indebtedness. According to Verbeek (2008), repeated cross-sectional surveys lack data from the same respondent over time. Verbeek (2008) pointed out the limitations from such surveys where individual historical data is not available for dynamic analysis (i.e. panel analysis). Moreover, based on the original data from this type of survey, it is impossible to transform models to first-difference forms or deviations from individual means (Verbeek, 2008). These limitations lead to the challenges in doing pseudo-panel analysis from repeated cross-sectional surveys, and the use of cohorts is one of the most popular solutions (Deaton, 1985; Moffitt, 1993). As genuine-panel data at the level of individual households are not available in Thailand, this research has constructed pseudo-panels for analysis. Deaton (1985) constructed pseudo-panel data by grouping respondents with the same birth year into the same cohort. Many researchers (e.g. Deaton, 1985; Moffitt, 1993; Verbeek &

Nijman, 1992) have attempted to construct pseudo-panel data using repeated cross- section data because there are at least two advantages of doing panel analysis (Wooldridge, 2010). First, dynamic relationships between a dependent variable and a set of independent variables can be examined using such analysis, but this type of relationships cannot be found from cross-section analysis. Second, unobserved cross- section heterogeneity (i.e. different characteristics between each group of households) can be controlled by using panel data. Dynamic analysis of household indebtedness, and pseudo-panel analysis using household surveys in literature will be reviewed in Chapter 6.

The next chapter will narrate the context information of Thailand in terms of its economy, financial system structure, households’ financial access, inequality issues, and household debt situation.

Chapter 3: Context Chapter

This chapter will provide contextual information on Thailand for subsequent chapters. It begins from general details, which are geographical background, economic size and its main contribution (including the economic development over the past decade), and the structure of the financial system. Then, it focuses on the household sector and specifically on access to finance by households, financial inequality across households, and the development in household debt over the past ten years.

There are four main sections in this chapter. Firstly, the geographical and economic background of Thailand is explained. Secondly, Thailand’s economic performance over the past ten years is depicted, including aggregate economic performance, and the contribution to economic growth from both the demand and supply side. The third section deals with the financial system and household finance in Thailand, including the structure of the financial system, household financial access, household financial inequality, and household debt development over the past ten years. The final section will then furnish a summary conclusion.

3.1 Geographical and economic background of Thailand

In this section, the location of Thailand is illustrated, and then Thailand’s size of the economy and its main components are clarified. In addition, a comparison of economic size between Thailand and other countries in the same region is provided at the end of this section.

On geography and population, Thailand lies in the middle of mainland Southeast Asia and covers around 514 000 square kilometres. Its neighbours are Myanmar (in the North and the West), Laos (also in the North and North-east), Cambodia (in the East) and Malaysia (to the South) (Figure 3.1). The total population as of 2017 Census was approximately 69 million. The nation is divided into four main geographical regions: the North, the North-east, the South and Central (which includes Bangkok, the capital city of Thailand). The country’s principal river systems are the Chao Phraya River and Mekong River, which collectively support the nation’s agricultural economy, especially wet-rice cultivation (also known as Thai Jasmine Rice). Due to high productivity in the rice industry, along with high quality of the product, Thailand is one of the major countries of rice cultivation and foreign trade in South-east Asia.

Figure 3.1 Map of Thailand

Source: Google Map.

Thailand has a diversified economy with agriculture contributing 8.3% of the nominal Gross Domestic Product (nominal GDP), industry5 35.3%, and services 56.4% (data for 2017 from the Office of the National Economic and Social Development Board). The total nominal GDP for 2017 was THB 15.5 trillion (USD 455.3 billion) with nominal GDP per capita of USD 6595. Thailand is a newly industrialised country, and the economy is mainly based on exports and domestic private consumption. From 2010 to 2017, Thailand’s export value and private consumption accounted for approximately 69% and 52% of the nominal GDP, respectively (Thailand’s net export value accounted for 7%, on average, from 2010 to 2017). According to the data from the

5 Industry includes mining and quarrying, manufacturing, electricity, gas and water supply, and construction. 45

Ministry of Commerce, Thailand’s top three goods and services exports were cars, computers and jewellery, while the top three goods and services imports were machinery, integrated circuits and crude oil. Thailand’s main trading partners were the US, China and Japan, while the top three sources of imports are China, Japan and the US.

Regarding the comparison of economic size between Thailand and other countries in the same region, Thailand is one of the founding members of the Association of Southeast Asian Nations (also known as ASEAN). The other four founding members are Indonesia, Malaysia, The Philippines and Singapore. In total, there are 10 current member countries, including Brunei Darussalam, Cambodia, the Lao People’s Democratic Republic, Myanmar, and Vietnam. Focusing on economic size in terms of nominal GDP and nominal GDP per capita as of 2017, compared to other ASEAN countries, Thailand is ranked number 2 (number 1: Indonesia [USD 1 trillion]) and number 4 (number 1: Singapore [USD 57 714.3]; number 2: Brunei Darussalam [USD 28 290.6]; number 3: Malaysia [USD 9 951.5]), respectively.

Next, Thailand’s economic development will be analysed over the past 10 years.

3.2 Thai economy over the past decade (2008 to 2017)

This section explains the economic development of Thailand over the past ten years based on the Bank of Thailand’s annual reports (Thailand’s Economic Conditions in 2008-2016) and the Bank of Thailand’s Monetary Policy Report of 2016-2018. This discussion will take as its starting point 2008, the year of the Global Financial Crisis and end in 2017. It begins with the aggregate economic performance by looking at overall GDP growth. Then, it will move to the contribution to economic growth from the demand side. Lastly, the analysis will focus on the contribution to economic growth from the supply side (i.e. sectoral economic performance).

3.2.1 Aggregate economic performance

Over the past 10 years, Figure 3.2 shows that Thailand experienced a fluctuation in economic growth, although its real GDP (reference year is 2002) increased from THB 7.7 trillion in 2008 to THB 10.2 trillion in 2017. During 2008 to 2009, the Thai economy was affected by the global economic downturn, which resulted from the global financial crisis. The export sector suffered severely from lower demand from its major

46 trading-partner countries (e.g. the US). Consequently, there was less economic activity, along with less manufacturing and low business confidence. In addition, the internal political unrest during this period led to a lower number of tourists and domestic demand (i.e. private consumption and private investment). The global economic downturn and the political unrest lowered Thailand’s overall economic growth to 1.7 and -0.7% in 2008 and 2009, respectively (Figure 3.2). The growth rates of private consumption, private investment, and exports of goods and services in 2008 were 2.8%, 6.0% and 6.3%. While the growth values in 2009 were -0.9%, -16.7% and -12.1%, respectively (Figure 3.3).

Figure 3.2 Thailand’s Real GDP and Real GDP Growth

Trillion Baht % Real GDP Real GDP growth (RHS) 11 9

7.5 7.2 10.2 9.8 9.5 6 9.1 9.2 9 8.9 4.0 8.2 8.3 3.4 3.1 3 7.7 7.7 2.7

7 1.7 0.8 1.0 0

-0.7

5 -3 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Figure 3.3 Growth in Private Consumption, Private Investment, and Exports of Goods and Services

% 20 Private consumption Private investment Exports of goods and services

-10

-20 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

In 2010, the economy recovered with strong economic fundamentals, and accommodative fiscal and monetary policies. The expansion in fiscal spending, and policy rate cuts in 2009 stimulated the economy to reach a growth of 7.5% in 2010 (Figure 3.4). The supporting policies raised business confidence and encouraged the expansion in private investment to hit a growth of 16.4% in 2010 (Figure 3.3). The recovery continued over the first three quarters of 2011 until there were huge floods in the fourth quarter of 2011. The worst floods in 70 years brought down the annual growth of 2011 to 0.8%. It caused the disruption in supply chains and transportation system resulting in less production capacity and merchandise exports.

Figure 3.4 Real GDP Growth, Growth in Fiscal Spending, and Policy Interest Rate

% Real GDP growth Fiscal spending growth Policy rate (RHS) % 4 10 9.9 8.7

8 7.5 7.27.1 3

6.0 6

4.3 4.0 2 4 3.4 3.1 2.7 2 1.7 1.3 0.8 1.0 1 0.5 0.2 0.4 0 -0.3 -0.7 -2 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

The economy was able to expand well again in 2012 from recovered domestic demand. Due to the great damage from the floods, the government imposed a set of stimulus measures (e.g. the government’s first-car tax rebate scheme), which improved the confidence of the private sector. The higher confidence led to the expansion in both private consumption and private investment with a growth of 6.7% and 11.7%, respectively (Figure 3.3). Private investment expanded in the sense of 1) production capacity expansion to support domestic demand as reflected by manufacturing production for domestic consumption, which expanded well in line with strong domestic spending, 2) restructuring of the production process to reduce dependence on labour, and 3) repairing the damaged infrastructure from the huge flood.

However, during 2013 to 2014, the economy slowed down again because of slower-than-expected recovery in trading-partner economies, and political uncertainties. The slow recovery in trading-partner economies led to sluggish exports of goods and services with a growth of 2.7% and 0.3% in 2013 and 2014, respectively (Figure 3.3). At this time, the internal political unrest hindered certain government operations and undermined the confidence of the private sector. Additionally, a sharp increase in household debt in the previous period adversely affected household consumption in this period (more details about the development of household debt and its effects will be discussed later). These internal factors led to more cautious spending by households and 49 postponement of investment projects by the business once the government’s stimulus measures expired. The private consumption growth during 2013 to 2014 was less than 1%, whereas private investment growth was negative (Figure 3.3). These negative factors caused the economy to slow from a growth of 7.2% in 2012 to a growth of 2.7% in 2013, and 1.0% in 2014 (Figure 3.2).

From 2015 to 2017, the economy recovered gradually from continued expansion in domestic demand, especially in private consumption. A series of government stimulus measures were imposed with the purpose of encouraging economic growth. For example, in the last week of 2015, the government introduced a tax-deduction scheme for up to THB 15 000 on the purchases of goods and services. Additionally, the government’s mega projects, which were launched in 2016, increased potentially the employment rate. These measures made the household sector more confident and led to higher private consumption growth. Private consumption growth increased from 0.8% in 2014 to 3.0% in 2017 (Figure 3.3). In addition, the expansion of merchandise exports to the major trading partners during this period supported economic growth which continued increasing to reach 4.0% in 2017 (Figure 3.2).

To compare Thailand’s economic performance with other countries in the same region, Figure 3.5 represents economic growth in five founding members of ASEAN (also known as ASEAN-5), which included Indonesia, Malaysia, The Philippines, Singapore, and Thailand, during 2008 to 2017. It showed that, since 2010, Thailand has experienced more fluctuation in economic growth than the other four countries. Moreover, after the period of the global financial crisis (2008 to 2009), Thailand has the lowest average economic growth (from 2010 to 2017) in ASEAN-5 at 3.7%. The average growth for Indonesia, Malaysia, The Philippines, and Singapore were 5.5%, 6.4%, 5.5%, and 5.4%, respectively.

Figure 3.5 ASEAN-5’s Real GDP Growth

% Indonesia Philippines Malaysia Singapore Thailand 16

-2 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: World Bank. Author’s calculations.

Next, this thesis will clarify the contribution to Thailand’s economic growth over the past 10 years from both the demand and supply sides.

3.2.2 Contribution to economic growth from the demand side

This section will focus on the contribution to Thailand’s economic growth during 2008 to 2017. First, it will look at the contribution to the economic performance from the demand side. Regarding GDP from the demand side, there are two main parts: 1)domestic demand (which includes private consumption, private investment, and fiscal spending), and 2) external demand (which includes net exports of goods and services).

Focusing on Thailand’s domestic demand, Figure 3.6 displays the share of each GDP component from 2008 to 2017. It shows that, over the past 10 years, private consumption has played an important role in the economy with a share of around 50% of GDP. While fiscal spending and private investment have accounted for around 22% and 19% GDP, respectively. Although the share of private consumption has decreased since 2010, the contribution of private consumption to economic growth has continued to increase over the past four years (Figure 3.7). Figure 3.7 and Table 3.1 show the contribution of each component to GDP growth, and the contribution is calculated from the growth of each component weighted by its share to GDP. The higher contribution of

51 private consumption is from a higher level of consumer confidence and non-farm income (Bank of Thailand, 2017b). Apart from the role of private consumption, fiscal spending also has made a great contribution to economic growth, especially when the country has experienced an economic downturn. For instance, when the economy slowed down in 2009 and 2014, government consumption-and-investment expenditure made a large contribution to the economic growth a year later (Figure 3.7).

Figure 3.6 Shares of GDP Components (Demand Side) (Per cent of Nominal GDP)

Private consumption Private investment % Fiscal spending Exports of goods and services Imports of goods and services 80 71 71 70 68 69 69 68 68 64 66

60 54 53 52 53 53 52 52 51 50 49 40 22 22 21 23 22 22 23 23 22 20 22 20 21 20 19 20 16 18 18 17 17

-20

-40

-54 -54 -60 -55 -57 -61 -63 -65 -69 -69 -69 -80 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Figure 3.7 Contribution to GDP Growth (Demand Side)

% Private consumption Private investment 15 Fiscal spending Net exports of goods and services Change in inventories, and residuals Growth

10 7.5 7.2

5 4.0 3.4 2.7 3.1 1.7 1.0

0 0.8 -0.7

-5

-10 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Table 3.1 Contribution to GDP Growth (Demand Side) Net exports Change in GDP Private Private Fiscal Year of goods and inventories, growth consumption investment spending services and residuals (%) 2008 1.5 1.2 0.1 -3.1 2.1 1.7 2009 -0.5 -3.4 1.9 6.1 -4.9 -0.7 2010 2.9 2.8 1.3 -3.9 4.5 7.5 2011 1.0 1.7 0.0 -1.5 -0.3 0.8 2012 3.6 2.3 1.5 -0.4 0.3 7.2 2013 0.5 -0.3 0.3 0.9 1.4 2.7 2014 0.4 -0.2 0.1 4.2 -3.5 1.0 2015 1.2 -0.3 1.8 1.2 -0.7 3.1

2016 1.5 0.1 0.9 2.8 -2.0 3.4 2017 1.5 0.5 -0.1 0.1 2.0 4.0 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Moving on to the external demand, Thailand has an open economy, which has an export sector that constitutes a large share of GDP. Over the past 10 years, exports of goods and services have accounted for more than 60% of GDP (Figure 3.6). However, the contribution of net exports (exports minus imports) to economic growth has fluctuated because the export sector has faced underlying structural issues and external risks Bank of Thailand (2014b). Concerning structural issues, the Bank of Thailand (2014) claimed that a lack of investment in new high-technology products has led to 53 limited production capability to gain benefits from rising global demand for such products. For the external risks, as I mentioned before, the economic downturn in major trading-partner countries (as a result of the recent global financial crisis), along with slower-than-expected economic recovery in these countries, has caused Thai exporters to get struggled from estimating their sales.

3.2.3 Contribution to economic growth from the supply side (Sectoral economic performance)

This discussion will now look at the contribution to economic growth from the supply side (i.e. sectoral economic performance). There are three main components of GDP on the supply side: 1) the agricultural sector; 2) the industry sector (includes mining and quarrying, manufacturing, electricity, gas and water supply, and construction); and 3) the service sector (Warr, 2007). Figure 3.8 shows that, over the past 10 years, the service sector has played an important role in the economy with the largest share at around 53% of GDP, whereas industry sector and agricultural sector have accounted for around 37% and 10%, respectively. In addition, the share of the service sector has kept increasing for more than five years because of the expansion in the tourism. During 2010 to 2017, the number of foreign tourists, especially the number of tourists from East Asia has increased substantially. Such expansion in the tourism sector has led to great contribution to economic growth since 2010 (Figure 3.9 and Table 3.2).

Considering the role of the industry sector, the contribution of this sector to economic growth has been volatile. In the case of export-oriented industries, these industries have experienced challenges in maintaining constant performance because of the slow pace of economic recovery in trading-partner countries. However, this sector has flexibility in production capability. For example, even though this sector (especially the manufacturing sector) experienced great damage from the huge floods in 2011, it recovered quickly in the following year with support from the government. The fast recovery is reflected from the large contribution of the industry sector to economic growth in 2012 (Figure 3.9 and Table 3.2).

Regarding the role of the agricultural sector, although Thailand is partly an agricultural economy, its share has been declining over the past 10 years. The value of agricultural products has dropped due to declining agricultural-product prices (e.g. rice,

54 rubber, and sugar cane). Such a fall in the prices of these productions is from excess supply from major competitors. Therefore, the agricultural sector contributes less to economic growth than the other two sectors (Figure 3.9).

Figure 3.8 Shares of GDP Components (Supply Side) (Per cent of Nominal GDP)

% Agriculture Industry Services 110

50.3 51.5 49.5 50.3 51.1 51.7 53.1 70 54.9 55.8 56.4

40.0 38.1 30 39.6 38.7 37.4 37.0 36.8 36.3 35.7 35.3

10 10.1 9.8 10.5 11.6 11.5 11.3 10.1 8.9 8.5 8.3

-10 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Figure 3.9 Contribution to GDP Growth (Supply Side)

% 10 Agriculture Industry Services Residuals Growth

7.5 7.2

5 4.0 3.1 3.4 2.7 1.7 1.0 0.8 0

-0.7

-5 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Table 3.2 Contribution to GDP Growth (Supply Side)

GDP growth Year Agriculture Industry Services Residuals (%) 2008 0.2 0.9 0.6 0 1.7 2009 0 -0.8 -0.2 0.3 -0.7 2010 0 4 3.7 -0.2 7.5 2011 0.5 -1.6 2 0 0.8 2012 0.2 2.7 4.6 -0.3 7.2 2013 0.1 0.6 2.2 -0.2 2.7 2014 0 0 1.1 -0.1 1.0 2015 -0.5 1.1 2.9 -0.4 3.1

2016 -0.1 1 2.7 -0.2 3.4 2017 0.2 0.7 3.4 -0.2 4.0 Source: The Office of the National Economic and Social Development Board. Author’s calculations.

Following from this background on the economic development of Thailand, this thesis will move to the structure of the financial system and household finance in Thailand.

3.3 Financial system and household finance in Thailand

3.3.1 Structure of Thailand’s financial system

This section begins with the definition of financial system, then follows an enumeration of the components of the financial system in Thailand. This section will also clarify the role and the size of each component.

The financial system is an intermediary for the allocation of economic resources by providing and supporting the payment and settlement of services. Sustainable economic performance can be encouraged by a well-organised, efficient and stable financial system (Bernanke & Gertler, 1990; Diamond & Rajan, 2001; Loayza & Ranciere, 2006). Bernanke and Gertler (1990) and Loayza and Ranciere (2006) defined sustainable economic performance as economic development with consistent growth over time (less fluctuation). Generally, a financial system consists of financial institutions and financial markets (Bank of Thailand, 2013b).

Financial institutions

According to the National Account System, there are three types of financial institutions in Thailand; (i) the Bank of Thailand (BOT) as the central bank, (ii) other depository corporations (ODCs), and (iii) non-depository corporations. The ODCs consists of seven institutions:

1. Commercial banks 2. Specialised financial institutions (i.e. state-owned financial institutions)) 3. Financial companies 4. Credit fonciers (A public limited company licensed to undertake credit foncier business by accepting deposit from public, granting mortgage loan, and purchasing immovable property under contract of sale with right of redemption (Bank of Thailand).) 5. Saving cooperatives 6. Credit unions 7. Money market mutual funds Meanwhile, the non-depository corporations consist of seven other institutions:

1. Mutual funds 2. Insurance companies 3. Provident funds 4. Credit companies 5. Asset management companies 6. Security companies 7. Foreign exchange companies According to Bank of Thailand (2018b), at the end of the third quarter of 2018, there was approximately THB 41 trillion total assets of financial institutions (excluding the Bank of Thailand’s assets) with 68.2% held by other depository corporations, whereas the non-depository corporations’ assets accounted for 31.8% of total assets. In the Thai banking system, commercial banks, which are fully regulated by the Bank of Thailand, are the biggest player (Bank of Thailand, 2018b), while specialised financial institutions (fully regulated by the Ministry of Finance) have increased their role in promoting both economic and financial activities since the Asian financial crisis in 1997 (Bank of Thailand, 2013b).

In the case of Thailand, commercial banks and specialised financial institutions play an important role in the financial system as major deposit-takers and credit- providers. Regarding the role of deposit-takers, at the end of the third quarter of 2018, there was THB 13.1 trillion of deposits at commercial banks and around THB 5 trillion on deposit at specialised financial institutions. The main depository corporations’ creditor is the household sector with a share of more than 80% of total deposits (Bank of Thailand, 2013b; Bank of Thailand, 2018b).

As to their role as credit providers and regarding corporate loans, commercial banks and specialised financial institutions provided 72% and 17% of total corporate loans (as of the third quarter of 2018), respectively. While focusing on consumer loans, commercial banks and specialised financial institutions provided 42% and 24% of total consumer loans, respectively. Focusing on the loan structure of the commercial banks, at the end of 2018, 65.9% of total commercial banks’ private credit were corporate loans, while 34.1% were consumer loans (Bank of Thailand, 2019c). In terms of consumer loans, mortgage loans have the highest share at 49.7%, followed by car loans (auto leasing), personal loans, and credit-card loans with a share of 23.7%, 21%, and 5.6%, respectively (Bank of Thailand, 2019c).

In addition, commercial banks and specialised financial institutions have extensive branch networks, which allow them to provide broader payment-and- settlement services than the others. At the end of 2013, there were 8200 branches of commercial banks and specialised financial institutions with 55 400 Automatic Teller Machines (ATMs) nationwide.

Financial markets

There are four main elements in Thailand’s financial markets; (i) the money market, (ii) capital markets, (iii) the foreign exchange market, and (iv) derivatives markets. Financial institutions’ short-term liquidity is provided by the money market, while capital markets, through bond and stock markets, provide medium-to-long-term liquidity. Capital markets are the most important source of finance for both the public and the private sectors. According to the data from the Stock Exchange of Thailand (2019b), at the end of 2018, there was around THB 16 trillion for stock-market capitalisation, and the major players were corporates. Meanwhile, there was THB 12.6 trillion bonds outstanding, including short-term (less than one year) and long-term

(more than one year) bonds (Thai Bond Market Association, 2019). At the same time, the average monthly value of the foreign exchange turnover of commercial banks in 2018 was USD 260 billion (Bank of Thailand, 2019a), while money market size (monthly) was approximately THB 2 trillion with most transactions by the financial institutions (Bank of Thailand, 2019b). However, the derivative market has not been popular, and most transactions have been used for hedging by counter parties in the over-the-counter market. Examples of the current derivatives in Thailand are stock futures and options, gold futures, US-dollar futures, and interest-rate futures.

After understanding the structure of the financial system in Thailand, it is useful to consider the financial access level of Thai households.

3.3.2 Household access to finance

This section will summarise the statistics of household financial access in Thailand provided by the Bank of Thailand (Bank of Thailand, 2013a; Bank of Thailand, 2016a). The Bank of Thailand aims to promote financial inclusion in Thailand by monitoring the level of financial access for Thai households. Therefore, the Bank of Thailand has conducted a household survey on financial access from both a savings and borrowings perspectives since 2003. This survey was collected in 2003, 2006, 2010, 2013, and 2016. Since 2006, the survey has been a joint project between the Bank of Thailand and the National Statistic Office. The sample size is around 10 000 households and covers all provinces in Thailand. According to the Bank of Thailand (2016a) financial services are divided into 11 products as the followings:

1. Deposits/ savings 2. Loans (excluding credit cards) 3. Money transfers 4. Payments 5. Credit cards 6. Life insurance 7. Non-life insurance 8. Mutual funds 9. Government and central bank debt securities 10. Private securities 11. Rotating savings groups (one type of informal funding by groups of people)

According to Bank of Thailand’s Financial Access Surveys of Thai households of 2013 and 2016, 88% and 86.3% of Thai households, respectivly, had access to at least one financial service from all the above financial services compared to 84.6% in 2010. The main financial service providers were still commercial banks and specialised financial institutions, which were formal fully-regulated financial institutions, as mentioned in the previous section. Furthermore, the top-three popular financial services during 2013 to 2016 were deposit or saving services, money-transfer and payment services, and loan services. For deposit and money-transfer services, commercial banks had played an important role, whereas specialised financial institutions had an increasing role in extending credit.

Considering financial service channels, branches of commercial banks and specialised financial institutions, affiliated with Automatic Teller Machines, were the most well-known channels for deposit and money transfer services. Therefore, households asked for more Automatic Teller Machines or financial institutions’ branches in their neighbourhoods, particularly in rural regions. Additionally, in 2016, there was a slight increase in the usage of financial services through internet banking, coupled with mobile applications. This reflects the fact that Thai households had more willingness to use technology for their financial services and were adapting themselves to the cashless-society. Moreover, the statistics of the surveys of 2013 and 2016 also showed that Thai households had used more variety of financial services than previous surveys.

Focusing on around 12-14% of Thai households who did not use any financial service rather than cash in their hands, there were only 4.2% and 2.7% of total households for 2013 and 2016, respectively, who were truly unable to have access to at least one financial service (Bank of Thailand’s Financial Access Surveys of Thai households of 2013 and 2016). In addtition, the surveys showed that there are 7.8% of households in 2013 and 11% in 2016 who had voluntarily excluded themselves from using financial services (i.e. voluntary self-exclusion). The group of households who did not have access to any financial services, was concentrated in low-income households and those who lived in the North-eastern or southern part of Thailand.

Moving on to the limitations that obstruct some households from having access to formal finance (i.e. financial services provided by commercial banks and specialised financial institutions), the statistics showed that poor financial status (e.g. having 60

insufficient income) was one of the main obstacles to their financial access in terms of formal deposit and loan services. The other limitations were the complicated procedures of financial services and the complex conditions on loan applications. In addition, long distances between households’ locations and the nearest financial service providers was another factor obstructing households from having access to formal financial services.

The survey above suggests that, in the case of Thailand, access to financial services provided by formal institutions (e.g. commercial banks and specialised financial institutions) is far from being universal across different groups of households (e.g. income groups), limited mainly by households’ poor financial status. Therefore, this thesis will next examine household financial inequality by income group.

3.3.3 Household financial inequality

This section raises issues of inequality around the world based on previous studies and reports. It then examines financial inequality in the case of Thailand in terms of income, expenditure, debt, and debt-service burden.

Concerning the trend on global inequality, the World Bank (2018) and World Economic Forum (2019) said that global inequality had been improving over the past 20 years coupled with a declining global poverty rate. The World Bank (2018) estimated that, from 1990-2015, the global poverty rate declined by 1 percentage point every year on average. Moreover, the number of people who lived in extreme poverty6 dropped by approximately 70 million from 2013 to 2015. The World Bank (2018) also calculated the average income growth of the bottom 40% poorest people in 91 countries around the world from 2010-2015 and compared it with the average income growth of total population in each country. The statistics showed that, in East Asia and Latin America, the bottom 40% poorest people had a higher average level of income growth than the national level. This reflected the convergence of prosperity in these regions. However, in Sub-Saharan Africa and South Asia, the bottom 40% groups had a lower income growth rate than average for the population. Moreover, there was a negative income growth for the bottom 40% cohorts in some countries in Sub-Saharan Africa such as South Africa, Uganda, and Zambia. The World Bank (2018) claimed that the severe

6 The extreme poverty is defined as the consumption level less than USD 1.9 a day, using 2011 purchasing power parity. 61 inequality in parts of Africa was due to slow economic growth, conflict, weak institutions, and the failure of income redistribution by the public sector.

In addition, Oxfam (2017) argued that global inequality was still high and it provided some interesting examples. It mentioned that only the top eight richest people have held the same value of wealth as the bottom 50% of the poorest people, and the top 1% richest people have held more wealth than the rest of the world since 2015 (Credit Suisse, 2016). In the case of the US, over the past 30 years, Cohen (2016) pointed out that the average income growth of the top 1% richest American citizens has been 300%, whereas the income growth of the bottom 50% was zero. In Vietnam, Lam (2017) found that the richest person in the country could earn in one day more than the poorest person could earn in 10 years.

The issues of inequality are still causing economic concerns to international organisations who monitor and examine the economic conditions of developed and developing countries. The World Bank (2018) raises the need for shared prosperity to reach its aims of reducing poverty and inequality. The World Economic Forum (2019) supported this point of view and worried that the remaining high levels of inequality in countries around the world could be one of the biggest potential threats to social stability, even though there have been large reductions in the levels of absolute poverty since 2000.

In the case of Thailand, this research has examined household financial inequality in terms of income, expenditure, debt, and debt service burden using Thailand’s official household surveys (Household Socio-Economic Surveys of 2013). Thailand’s Household Socio-Economic Surveys are also the main database of this thesis; more details about these surveys will be provided in the next chapter.

To investigate financial inequality in all the previously listed perspectives, first of all, this research has divided sample households from all 77 provinces in Thailand into 10 groups equally by ordering households’ income from the poorest to the richest (income deciles). Then, the share of each financial variable held by each group of households was calculated. Figure 3.10 shows that, for the top three deciles of the richest households, their income accounted for 64.2% of total household income compared to 55% in terms of expenditure. In contrast, for the bottom three deciles of the poorest households, the share of their income was 8.7% of the total income compared to

62 a share of 12.9% in term of the expenditure value. Moreover, for these groups of households (i.e. the bottom three deciles of the poorest households), their average 1/ consumption-to-incomeShare ratio of was household more incomethan one and, expenditureand this means these households did (by income group)

% of total aggregate value 40 Income Expenditure 37.1

30 26.9

20 15.715.9

11.412.2 10.0 8.5 9.1 10 7.3 7.3 6.2 6.0 5.3 4.9 4.4 3.9 3.2 3.0 1.8 0 Group1 Group2 Group3 Group4 Group5 Group6 Group7 Group8 Group9 Group10 not have the ability to save and accumulate wealth (Figure 3.11).

Note: 1/ Population weight Figure 3.10 Share of Household Income and Expenditure by Income Group Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations. Figure 3.11 Expenditure-to-Income Ratio by Income Group

1/ Times Consumption-to-Income ratio 1.54 1.50

1.11 1.02 C-Y ratio = 1 1.00 0.95 0.91 0.87 0.83 0.80 0.76

0.63

0.50

0.00 Group1 Group2 Group3 Group4 Group5 Group6 Group7 Group8 Group9 Group10 Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.Note: 1/ Population weight 63

In terms of debt and debt-service burden distribution, it is useful to look at the distribution of debt outstanding as a term of stock, combined with the distribution of monthly debt service payment as a term of flow. Figure 3.12 shows that, for the top three deciles of the richest households, the share of their debt outstanding was 76.2% of total household debt outstanding compared to 65% in terms of the debt-service-payment value. In contrast, for the bottom three deciles of the poorest households, the share of their outstanding debt was 5.5% of the total outstanding debt compared to 9.3% in term of the debt-service-payment amount. For these groups of households (i.e. the bottom three deciles of the poorest households), these statistics reflected more concentration in their debt-service-payment burden (flow) than the concentration in their debt outstanding (stock). More details about household’s indebtedness by income group will be discussed in Chapter 4. Next, this dissertation will focus on the development of household debt over the past 10 years to provide a clearer picture of household debt situation in Thailand.

Share of household debt and debt service burden1/ Figure 3.12 Share of Household Debt(by and income Debt group)-Service -Payment Burden by Income Group

% of total aggregate value 50 Debt Debt service burden 44.9

40 35.4

20.1 20 18.0

11.211.6 8.6 10 7.1 7.3 5.4 3.9 4.7 4.8 3.3 2.7 3.5 1.3 2.1 1.9 2.3 0 Group1 Group2 Group3 Group4 Group5 Group6 Group7 Group8 Group9 Group10 Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.Note: 1/ Population weight

3.3.4 Household debt development

This section will illustrate the recent household debt situation in Thailand from both macro and micro points of view, which has led to a concern about negative effects on the macro-economy. From the macro-perspective of 2008-2017, Thailand experienced a significant increase in household loans. According to the Bank of Thailand (BOT)’s statistics of household loans from financial institutions, Figure 3.13 shows that the household-loan outstanding jumped from THB 5.1 trillion at the end of 2008 to THB 12.1 trillion at the end of 2017. Within 10 years, the rise was doubling the amount of household loans that led household-debt-to-GDP ratio to soar from 52% in 2008 to around 80% in 2017. According to the BOT’s annual reports, the expansion of consumer loans was partly from financial institutions’ willingness to lend, a low- interest-rate environment, and the government’s supporting policies (e.g. first-car tax rebate scheme). These factors encouraged Thai people to obtain more credit and spend more money on personal consumption during 2010-2012 (Bank of Thailand, 2012).

Given rising levels of household debt, policymakers are now questioning the implications for financial stability (Bank of Thailand, 2014b, pp. 33-34; Muthitacharoen et al., 2015). Some policymakers have mentioned the adverse effects of the resulting large debt-service-payment burden on the macro-economy, through the drag of consumption level (Muthitacharoen et al., 2015). According to Figure 3.13, Thailand has faced a low consumption growth rate for five years after the peak of household debt growth in 2012. This rate is less than 5% and below the historical average rate of growth (Bank of Thailand, 2014b, p. 33).

Figure 3.13 Household Debt Growth1/, GDP Growth and Consumption Growth in Thailand

Trillion %YoY Baht HH debt growth (RHS) Real GDP growth (RHS) 14 Real consumption growth (RHS) HH debt outstanding 20

12 15 10

10 8

6 5

4 0 2

0 -5 Q1/2008 Q1/2009 Q1/2010 Q1/2011 Q1/2012 Q1/2013 Q1/2014 Q1/2015 Q1/2016 Q1/2017 Note: 1/ Household loans from all financial institutions. Source: The Office of the National Economic and Social Development Board and the Bank of Thailand. Author’s calculations.

Moreover, the sharp increase in household debt, which had an average growth rate of 15.2 percentage points during 2010 to 2013, also affected financial institutions’ credit quality and liquidity through a higher default rate (Bank of Thailand, 2013b). The substantial rise in household debt led to a decline in credit quality (arrears on consumer- loan commitments) reflected by a rise in the ratio of non-performing consumer loans (more-than-3-month arrears on consumer-loan commitments) to total consumer loans (Figure 3.14 and 3.15).

However, according to the Bank of Thailand (2014a), the rising household debt caused limited risk exposure to financial institutions because most large financial institutions (i.e. commercial banks and specialised financial institutions) were supervised by the authorities. Commercial banks are under the supervision of the Bank of Thailand, while specialised financial institutions are under the supervision of the Ministry of Finance. Therefore, both commercial banks and specialised financial institutions must follow risk management processes.

Figure 3.14 Consumer Loan Growth

Source: Performance of the Thai banking system report of 2015, Bank of Thailand.

Figure 3.15 Non-Performing Consumer Loans

Source: Performance of the Thai banking system report of 2015, Bank of Thailand.

In order to put these Thailand figures in a global context, it is useful to compare the nation’s performance with the countries in the Asia-Oceania region. Figure 3.16 displays the average annual growth rate of household loans in Asian-Oceanian countries calculated from the Bank for International Settlements’ long time-series on total credit and domestic bank credit to the private non-financial sector. Figure 3.16 shows that Thailand had the average growth rate of household loans from other depository

67 corporations (ODCs) at 14% per annum during 2009 to 2013. This size of growth was smaller than only Indonesia and China, whose average annual growth rates were 21% and 28%, respectively. Moreover, considering the growth in household loans since the Asian financial crisis in 1997, Thailand had an upward average growth rate from 1999 to 2013 (Figure 3.16).

Figure 3.16 ODCs’1/ Household Loans by Country (Average Growth2/)

% 30 28 28 26 25 22 21

20 18 16 15 15 14 14 14 1414 12 111212 10 10 10 9 9 9 7 8 7 6 5 6 5 2 2 3 1 0 -1 -1 -1 -5 1992 - 1996 1999 - 2003 2004 - 2008 2009 - 2013 Japan Australia Korea Israel Hong Kong, SAR Saudi Arabia India Malaysia Thailand Singapore Indonesia China Notes: 1/ Other depository corporations include commercial banks, depository specialised financial institutions, saving cooperatives, finance companies and credit foncier companies. 2/ Average annual growth (geometric mean). Source: The Long series on total credit and domestic bank credit to the private non-financial sector, Bank for International Settlements.

Focusing on the long-term trend of household loans, Figure 3.17 displays the long-term trends of the household-loan index (the first quarter of 2013 = 100) from 1992 to 2015 using the Bank for International Settlements’ long time-series of household credit for Asian-Oceanian countries (in local currencies). We can see from Figure 3.17 that Thailand had been one Asian country that has the most sharply upward trend in household loans since 2008 with a more than ten-fold increase in the index (i.e. the index for Thailand rose from 17 in the first quarter of 1992 to more than 200 in 2015).

Figure 3.17 ODCs’ Household Loans by Country (Outstanding: Local Currency) Index (03/2008 = 100) 300 Australia Hong Kong, SAR Israel India 250 Japan Korea Malaysia Saudi Arabia 200 Singapore Thailand

150

100

03/1992 03/1997 03/1993 03/1994 03/1995 03/1996 03/1998 03/1999 03/2000 03/2001 03/2002 03/2003 03/2004 03/2005 03/2006 03/2007 03/2008 03/2009 03/2010 03/2011 03/2012 03/2013 03/2014 03/2015 Source: The Long series on total credit and domestic bank credit to the private non-financial sector, Bank for International Settlements.

Using household-level data, regarding household’s changes in financial status (average terms using population weights) using Thailand’s official household surveys of 2009 to 2015, Table 3.3 shows that average household’s monthly income increased by 28.8% or 4.8% per annum during 2009-2015, while the average household’s monthly expenditure grew faster at 30.6% or 5.1% per annum. With regards to the change in average household’s debt outstanding (for all households), the average debt per household increased by 16.4% or 2.7% per annum. For all households, Tables 3.3–3.4 show that the average debt outstanding per household increased at a lower rate than the average household’s monthly income along with declining proportions of indebted households (households with debt and debt service burden). However, in Thailand, households with no debt include households with no willingness to borrow (i.e. voluntary credit self-exclusion) and households with no credit access (i.e. households who have the willingness to borrow but cannot get loans). Therefore, next, the financial status of indebted households will be analysed to get a clearer picture of the changes in indebted households’ financial status.

Table 3.3 Overview Statistics for the Average Household’s Financial Status

Growth Unit: THB 2009 2011 2013 2015 (2009 – 2015)

Monthly income 20,903 23,236 25,194 26,915 28.8%

Monthly expenditure 16,205 17,403 19,061 21,157 30.6%

Debt 134,699 134,900 163,087 156,770 16.4% Note: Average values are calculated using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Table 3.4 Proportion of Indebted Households (Households with debt and debt service payments)

2009 2011 2013 2015

Proportion of indebted households (%) 55.2 51.9 50.5 46.5 Note: Using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Focusing on the changes in indebted households’ financial status, this thesis will analyse households with debt and debt service payments. From Table 3.5, during 2009- 2015, indebted households’ average monthly income increased by 35.8% or 6% per annum, while household’s average monthly expenditure increased more at 40% or 6.7% per annum. Regarding the change in indebted households’ average debt outstanding, the average debt per indebted household increased by 39.6% or 6.6% per annum. For indebted households, their average debt outstanding grew faster than average monthly income. This situation led to higher indebtedness among this group of households. More information on indebted households’ financial status over time will be provided in Chapter 6.

Table 3.5 Overview Statistics of the Average Indebted Households’ Financial Status

Growth Unit: THB 2009 2011 2013 2015 (2009 – 2015)

Monthly income 23,060 25,883 29,857 31,308 35.8%

Monthly expenditure 17,898 19,856 22,290 25,059 40.0%

Debt 234,003 250,823 314,894 326,543 39.6% Note: Average values are calculated sing population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations. 70

3.4 Conclusion

In conclusion, this chapter provides the context for the subsequent analysis. The key lessons of relevance for the next chapter are: (i) Thailand is an open developing country whose economy depends on both external and domestic demand: (ii) the tourism sector and private consumption play an important role in the contribution to economic growth: (iii) commercial banks, and specialised financial institutions are major formal financial service providers in terms of deposit takers and credit providers: (iv) although there is higher level of household access to finance, some Thai households are not able to have access to formal financial services because of poor financial status (e.g. insufficient income): (v) there is evidence of financial inequality in Thailand as poor households have imbalance between income and expenditure, and more concentration in debt-service-payment burden; and (vi) high growth in household debt during 2010-2013 hurt the Thai economy through a drag in consumption and financial institutions’ credit quality following later.

In the next chapter, I will turn to the first analysis part. In this thesis, there are three analysis sections: 1) the drivers of household indebtedness (covered in Chapter 4), 2) differences between subjective over-indebtedness and objective over-indebtedness (Chapter 5), and, 3) the determinants of household debt performance over time (Chapter 6).

Chapter 4: The Drivers of Household Indebtedness: Evidence from Thailand7

4.1 Introduction

In this chapter, the debt situation of different groups of Thai households will be outlined. Income, social class, region, and the household head’s marital status are commonly used by many researchers (Vante, 2006; Betti et al., 2007; Anderloni & Vandone, 2008; Keese, 2009; Lusardi & Tufano, 2009; Disney & Gathergood, 2012). Second, this research identifies the factors, which affect Thai households’ indebtedness by using household-level data from the official surveys. This is the first time that such data have been used to investigate the household characteristics that are the drivers of household indebtedness in Thailand.

Why does a household take on debt? This research draws on the theoretical literature to answer this question. There are three specific theories that provide reasons why individuals may take on debt. The first of these, the Permanent Income Hypothesis and the Life-Cycle Hypothesis, state that households facing a temporary adverse shock to their income will borrow to smooth out their current consumption with a view to repaying this debt when incomes recover (Friedman, 1957; Modigliani, 1966). Such smoothing of consumption raises the welfare of the household and has beneficial effects of stabilizing the aggregate economy (Barba & Pivetti, 2009). Friedman (1957), who introduced the Permanent Income Hypothesis, suggested that individuals’ current consumption levels were determined by their permanent income levels, and they attempted to smooth consumption over their lifetimes. Therefore, a drop in transient income could lead to an increase in debt which was used to finance current consumption. However, Modigliani (1966), who initiated the Life-Cycle Hypothesis and separated individual’s lifetime into three stages of life (i.e. young age, working age, and old age (retirement)), pointed out that the individual had different capacities to generate income at different stages of life. Taking into account the varying capacities of an individual to generate income, Modigliani (1966) believed that such capacities were minimal at the very early and late stages of life but were at a maximum point during the

7 An earlier version of this chapter was published as a journal article (See Chotewattanakul, P., Sharpe, K., & Chand, S., 2019, ‘The drivers of household indebtedness: evidence from Thailand’, Southeast Asian Journal of Economics, 7(1): 1-40). 72 working phase. Thus, the individual dis-saves during childhood and in retirement, and the consumption is funded by debts or savings accumulated during the working phase.

Both the permanent income and life-cycle hypothesis assume that credit is readily available, which may not be the case in most circumstances and not so in the context of developing countries where capital markets are under-developed. Liquidity constraints, therefore, may limit the extent of debt taken by households. Consequently, any improvements in access to credit, such as that arising from liberalisation financial markets can also raise the level of household debt (Debelle, 2004; Kang & Ma, 2009). Furthermore, considering the significance of credit rationing and liquidity constraints, the imperfection in financial markets can affect the borrower’s debt service-payment burden from different interest rates (Flemming, 1973; Stiglitz & Weiss, 1981). The borrower may have superior knowledge of their permanent income and the capacity to service debt compared to the lender, thus leading to asymmetries in information with regards to the supply of, and demand for, credit. The imperfect information in the markets makes banks (lenders) unable to know borrowers’ full credit profile. Therefore, the lenders may set their lending rates that do not equilibrate demand with supply, and in the process ration available supply (at the given interest rate).

The information asymmetry creates what is referred to in the literature as problems of moral hazard and adverse selection. Moral hazard arises when players in the markets provide misleading information before signing contracts or change their behaviours after obtaining the contracts (Prescott & Townsend, 1984). In the case of banking, moral hazard concerning borrowers creates a credit risk for lenders. Adverse selection arises when a market player has a characteristic hidden from the other players and which is adverse to their interests. In the case of either moral hazard or adverse selection, players with less information have a higher chance of being at a disadvantage to players with more information, and this results in a lack of efficiency in pricing and selecting goods and services (Prescott & Townsend, 1984).

High borrowing rates that leave some putative borrowers without loans results from the attempt by lenders to manage moral hazard and/or adverse selection. This results in the lenders using credit profiles to ‘sort’ the market. Therefore, while households with good credit profiles may obtain standard-interest-rate loans, households, with a poor credit profile are not allowed to have access to the formal credit

73 market; so it is necessary for them to seek some loans from the informal credit market, which incurs a higher interest rate than the formal market.

So far, it has been supposed that borrowers and lenders are rational economic agents - rationality is assumed both in the permanent income/life cycle hypotheses, and in the literature on credit rationing. However, recent work by behavioural economists, argues that this assumption is left wanting. In particular, so far as households are concerned, behavioural theories of economic behaviour add another layer of arguments explaining why households might end up in debt.

In addition to the above reasons for the growth of household debt, lastly, some researchers claim that low financial literacy and poor financial management can cause households to have a greater chance of being over-indebted (Anderloni & Vandone, 2008; Lusardi & Tufano, 2009; Disney & Gathergood, 2012). For example, some households cannot do correct interest calculations, so they are not able to perceive the actual cost of borrowing. In addition, some households such as hyperbolic discounters have financial self-control problems. Hyperbolic discounters (i.e. people with hyperbolic discounting behaviour) are consumers who have a high preference for the current consumption and tend to heavily discount their future consumption (Laibson, 1997).

Consequently, hyperbolic discounting behaviour causes people to excessively discount their future consumption. Moreover, people with hyperbolic discounting behaviour seem to underestimate the cost of debt repayment in terms of forgone future consumption. Eventually, households who are hyperbolic discounters tend to bring their consumption forward by borrowing and may face debt repayment problems due to insufficient savings. Moreover, they are more likely to be ‘caught by surprise’ in the future by the amount of debt they have to repay (Laibson, 1997; Gathergood, 2012).

Based on the theoretical literature and the previous studies (more details are given in Chapter 2), this thesis employs three separate models to characterise savings and consumption behaviour - first, the neoclassical models of the life cycle hypothesis and the permanent income hypothesis; second, the asymmetric information models of credit rationing; and, third, the behavioural finance models. All the previously described theories are important in explaining Thai households’ indebtedness. Firstly, this research has found that the neo-classical models are supported by the evidence from

Thailand. Secure income and sufficient savings allow households to improve their debt performance and reduce their chances of being over-indebted. Conversely, a higher level of dependency on finance can cause households to have poorer debt performance. Secondly, for some borrowers, credit constraints are binding in the formal market for loans, which lead those borrowers to enter the grey market. Moreover, their doing so is found to be correlated with higher indebtedness of households. Finally, this analysis has found that higher financial literacy is also correlated with superior debt performance.

The chapter is divided into five sections. The first section is the introduction. In Section 2, the primary data and descriptive statistics on household indebtedness in Thailand are described. Section 3 presents the econometric models, which will be used for the study of the drivers of household indebtedness, using official household survey data. The results of the econometric analysis are shown in Section 4. The last part of this chapter summarises the key findings from the study.

4.2 Data and overview of household indebtedness in Thailand

4.2.1 Data

In this chapter, quantitative analysis will be applied. Firstly, descriptive statistics will be generated for the empirical overview of household indebtedness in Thailand. Secondly, econometric models will be employed for the technical analysis. Thailand’s official household surveys (Household Socio-Economic Surveys) of the first quarter of 2013 will be used, along with the Bank of Thailand’s supplement survey for the completeness of the data set. Household Socio-Economic Surveys (SES) are administered by the National Statistical Office. For SES of the first quarter of 2013 (SES 2013Q1), 10 661 households were selected as being representative of the whole household population from all of 77 provinces in Thailand. These representative households were chosen using a stratified random sampling method where households from every district in all provinces were selected.

For the SES, households could be identified by the following criteria:

 Geography: province, region, municipal/ non-municipal area  Head of household’s characteristic: gender, age, marital status, education  Social class: household member’s occupation that generates the highest proportion of income of household.

Data on the surveyed household’s financial status were also collected in terms of flows and stocks.

Flows:

 Inflows: monthly income, including types of income (e.g. wage, salary, profit from business, pension, and transferred income)  Outflows: monthly expenditure, including types of expenditure (e.g. food, beverages, tobacco, other consumption, and non-consumption), and monthly debt service payment.

Stocks:

 Physical assets: dwelling, vehicles  Debt, including type of loans (mortgage, agricultural business, non- agricultural business, education, personal consumption).

For SES 2013Q1, additional indicators of financial difficulties were also collected, namely: the problems with other financial commitments (rent and utility bills) and credit access (in the case of emergency and for working purposes).

In addition to the data sourced from the SES, information from the Bank of Thailand (BOT)’s supplement household survey was used for the analysis on the drivers of debt at the level of the household. This BOT-survey was administered in the first quarter of 2013 (BOT 2013Q1) across the same 10 661 households as SES 2013Q1. Four parts of this survey have been used to ascertain the information of household’s perception of financial status, financial discipline, the extent of financial literacy, and credit constraints. These four parts are detailed as the following:

Household’s perception of financial situation:

 Concern about the next debt repayment (have/do not have the finances to meet the obligations)  Current situation and expectation of household’s economic status (worse, no change, better)  Current situation and expectation of household’s debt level (decrease, no change, increase)

 Current situation and expectation of household’s debt service burden (decrease, no change, increase)  Credit profile over the past twelve months (meet/cannot meet loan repayments by the due date)  Current situation and expectation on government help for credit default.

Financial discipline:

 Doing income-and-expenditure accounts  Saving behaviour (saving plan), including the period that saving can be used in the case of an income shock  Composition of financial assets.

Financial literacy (based on Organization for Economic Co-operation and Development (OECD)’s survey of financial literacy8):

 Knowledge score (the questions about interest rate and inflation)  Behaviour score (the questions about saving and spending behaviours)  Attitude score (the questions about hyperbolic discounting behaviour).

Financial constraints (credit rejection from the formal sector)

 The question was: have you ever been rejected from formal credit providers?

The survey provides data on a total of 10 661 individuals. This data incorporated values that are implausible such as individuals having negative or nil income, thus it had to be cleaned. The cleaning of the data entailed the following.

 Thirty-four households with income less than or equal to zero were dropped.  Four households with debt service ratio more than 1000% were also dropped.

Finally, 10 623 households were left for detailed analysis.

8 Measuring Financial Literacy: Questionnaire and Guidance Notes for Conducting and Internationally Comparable Survey of Financial Literacy 77

Table 4.1 provides the number of households by three main common categories: social class, region and head of household’s marital status. It shows that the majority (equal to 30% of the total population) were workers, 29% of these households resided in the Central region, and around 65% of the household heads were married.

Table 4.1 Number of Observations

No. of No. of Household head’s No. of Social Class Region households households marital status households Bangkok Agricultural 1930 (18%) 607 (6%) Single 1131 (11%) (BKK) Central Non-agricultural 2101 (20%) 3070 (29%) Married-couple 6990 (66%) (excl. BKK) Married-widowed/ Professional 1140 (11%) North 2598 (25%) 2497 (23%) separated/ divorced

Worker 3226 (30%) North-east 2814 (26%) N/A 5 (0%) Retired 2226 (21%) South 1534 (14%) Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office.

Lastly, a regional poverty line from National Statistical Office was used. The national poverty line is calculated by the National Economic and Social Development Board as the cost or expense of the individual in the acquisition of food and basic services essential to life. To calculate the poverty line, the National Economic and Social Development Board calculates the cost or expense of the individual in the acquisition of food and basic services essential to life for each household taking into account of sex, age, and region (including rural or urban area). Then, the National Economic and Social Development Board reports the national poverty line. Therefore, the official poverty line represents the average minimum cost of living per person (for everyone in Thailand). In this analysis, the poverty line has been employed for five regions in Thailand: Bangkok (capital city), Central (exclude Bangkok), North, North- east and South as Table 4.2 below.

Table 4.2 Poverty Line (THB/ Person/ Month) Region Bangkok Central Year North North-east South (BKK) (excl. BKK) 2013 3047 2775 2314 2273 2651 Source: National Statistical Office

4.2.2 Overview descriptive statistics

This section will use a number of measures discussed in Chapter 1 in order to characterise the situation of household debt in Thailand. The definitions are reprised for the readers convenience in the next paragraph. For overview statistics on household indebtedness in Thailand, firstly, the debt status of indebted households is presented by income group. Secondly, the proportion of indebted households is calculated by social class, region, and head of household’s marital status. Thirdly, household debt performance is analysed by the same categories of households.

The debt status of a household is defined in the extant literature (e.g., D’Alessio & Iezzi, 2013) as the ratio of monthly debt service payments to monthly income (i.e. debt service ratio) (Equation 4.1).

퐷푆푃 DSR = (4.1) 퐼 where 퐷푆푅 = Debt service ratio

퐷푆푃 = Monthly debt service payments

퐼 = Monthly income

Conversely, debt performance is defined as the ratio of household’s monthly income after debt service payments to minimum subsistence income level (e.g., Keese, 2009; D’Alessio & Iezzi, 2013) (Equation 4.2).

퐼 − 퐷푆푃 DP = (4.2) 푆퐼 where 퐷푃 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

푆퐼 = Minimum subsistence income level

In this chapter, this research used the poverty line as the minimum subsistence income level. Therefore, debt performance is defined as the ratio of household’s monthly income after debt service payments to the household’s poverty line (Equation 4.3 – 4.4).

퐼 − 퐷푆푃 DP = (4.3) 퐻푃퐿 where 퐷푃 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

퐻푃퐿 = Household’s poverty line

HPL = PL x HS (4.4) where 퐻푃퐿 = Household’s poverty line

푃퐿 = Poverty line

퐻푆 = Household size

Lastly, the proportion of both subjectively and objectively over-indebted households will be shown. Subjectively over-indebted households are defined as households with a concern about the next debt repayment, whereas objectively over- indebted households are defined as households with the ratio of income after debt service payment to poverty line less than one (Equation 4.3 when DP less than 1).

Household’s indebtedness by income group

For the overview of the indebtedness status of indebted households using Thailand’s official household surveys of 2009 to 2013, the average annual growth of monthly income, debt, and debt service payment are 6.7%, 7.7% and 5.7%, respectively. The rate of growth of household debt exceeding the rate of income growth has led to an increase in debt to annual income ratio, which rose from 78.7% in 2009 to 81.6% by 2013. However, there was a slower increase in the debt service costs; this being due to a decline in the average effective borrowing rate from 8.6 per annum in 2009 to 7.8 per annum in 2013. Therefore, the household’s debt service ratio (the ratio of monthly debt service payment to monthly income) decreased from 28.1% in 2009 to 27.4% in 2013.

Focusing on indebtedness status by income group using Thailand’s official household surveys of 2013, income deciles have been generated, and indebtedness indicators in terms of debt service ratio and debt-to-income ratio calculated for each income decile. Figure 4.1 shows the ratio of household’s debt service payment to 80 monthly income (i.e. debt service ratio) by income decile from the poorest to the richest. The value for each income decile is an average term using population weights. Regarding households with heavy debt service burdens, some researchers (D’Alessio & Iezzi, 2013) defined the heavy debt service burden as the debt service ratio greater than 30% (Equation 4.1 when DSR greater than 0.3). From Figure 4.1 it can be seen that households with heavy debt service burdens were concentrated in the lowest 30% of households by income. Furthermore, for the lowest income group, which has household’s monthly income less than THB 6000 (Table 4.3), households spend more than a half of their income on debt servicing (Figure 4.1).

Moving on to indebtedness in terms of debt-to-income ratio, Figure 4.2 shows the ratio of household’s debt to annual income by the same set of income deciles. The average debt-to-income ratio for the lowest income group is also the highest one at more than 120% (Figure 4.2). In addition, focusing on the average composition of debt, Figure 4.3 shows that the lowest income group’s agricultural business loans and personal consumption loans - which are unsecured loans - accounted for more than 80% of their total debt. On the other hand, the debt of the upper deciles is relatively more collateralised (i.e. mortgages) and fewer loans are for agricultural purposes, so the weighted average interest rates facing those households are lower. This explains why their debt-to-income ratios are relatively high (the top income decile has the second highest debt-to-income ratio), yet, at the same time, their debt-servicing-to-income ratios are relatively low (the top income decile has the lowest debt-service-payment--to- income ratio).

The facts presented above point to the heterogeneity of access to credit, and the debt-service commitments of households grouped by income deciles. The heavy debt service burden and the high proportion of unsecured loans among low-income households cause concern about financial inequality in Thailand. In the case of Thailand, as one of developing countries, apart from the income inequality issue (more details in Chapter 3), the high level of debt-to-income ratio in the group of low-income households, together with the high share of monthly debt service repayment to monthly income, causes them to trapped in poverty. Because these households (i.e. low-income households) need to spend a lot of their income on the debt repayments, coupled with their debts not being secured, they have no chance to accumulate wealth. This circumstance raises an issue on debt unsustainability in Thailand, which is referred to as 81 the situation when households have to cut their minimum standard of living due to their attempts to repay their debts (Betti et al., 2007). This then leads to questions about a proportion of over-indebted households and the drivers of indebtedness, which will be investigated later.

Figure 4.1 Debt Service Ratio by Income Group

% 62 60 60 2009 2013

45 40 40

31 29 30 27 27 25 24 25 24 25 23 23 23 23 22 22 22

0 Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10

Income group Poorest Richest

Note: Using population weights Source: Household Socio-economic Surveys, National Statistical Office. Author’s calculations

Table 4.3 Household’s Income in 2013 by Income Group (THB/ Month)

Min Max Group 1 (Poorest) 10 6,333 Group 2 6,335 8,772 Group 3 8,773 11,147 Group 4 11,148 13,730 Group 5 13,731 16,786 Group 6 16,788 20,715 Group 7 20,719 25,601 Group 8 25,604 33,116 Group 9 33,119 49,250 Group 10 (Richest) 49,257 8,820,684 Note: Using population weights Source: Household Socio-economic Surveys of 2013, National Statistical Office. Author’s calculations.

Figure 4.2 Debt to Annual Income Ratio by Income Group

% 160 2009 2013 140 137 128 120 106 102 103 100 97 85 79 80 75 74 76 69 64 66 65 59 59 59 60 60 58

0 Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10 Income group Poorest Richest

Note: Using population weights

Source: Household Socio-economic Surveys, National Statistical Office. Author’s calculations.

Figure 4.3 Average Composition of Debt by Income Group

% 120 Mortgage loans Agricultural loans Business loans Educational loans Personal consumption loans Other purpose loans

100 1.9 1.2 1.6 1.6 1.1 1.2 1.0 0.5 1.0 0.9

80 44.8 46.4 45.3 46.4 48.0 49.8 52.0 51.3 55.3 52.1 60 1.5 3.9 4.3 2.7 2.4 3.4 7.4 2.6 2.9 8.1 4.9 5.9 2.4 1.8 1.9 7.8 9.1 9.0 8.8 40 8.1 12.3 17.1 43.0 38.9 37.6 20 40.7 33.0 29.7 27.4 22.5 31.0 19.1 8.5 11.9 0 3.0 2.5 4.2 4.4 5.7 5.7 Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10 Income group Poorest Richest

Note: Using population weights Source: Household Socio-economic Surveys of 2013, National Statistical Office. Author’s calculations

Proportion of indebted households

Before moving on to the proportion of over-indebted households, in this part, first the number of indebted households (i.e. households with debt outstanding) will be counted to enumerate the proportion of indebted households by social class, region and head of household’s marital status. Table 4.4 shows that agricultural households, households in the north-eastern part and households with a married-couple household head having the highest proportion of indebted households at 72.9%, 67.9% and 63%, respectively. Whereas households who have a retired household head, households in Bangkok and households with a single household head have the lowest proportion of indebted households at 36.7%, 37% and 29.2%, respectively.

Table 4.4 Proportion of Indebted Households

Household head’s Social Class % Region % % marital status Bangkok Agricultural 72.9 37.0 Single 29.2 (BKK) Central Non-agricultural 61.4 48.3 Married-couple 63.0 (excl. BKK) Married-widowed/ Professional 59.1 North 57.8 43.8 separated/ divorced

Worker 51.8 North-east 67.9 Retired 36.7 South 48.7

Note: Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Household debt performance in Thailand9

In this part, this research has estimated average debt performance by social class, region and head of household’s marital status. As mentioned before, the household’s debt performance is the ratio of the household’s monthly income after debt service payments to the household’s poverty line (Equation 4.3 – 4.4). Table 4.5 shows that households with the poorest performance are agricultural households, households in north-eastern Thailand and single-adult households and widows, in particular. The intersection of the above-identified factors shows household most vulnerable to high

9 (Monthly income – Monthly debt service payment)/(Poverty line x Household size) 84 debt; namely, widows drawing the bulk of their income from agriculture and from the North-east.

Table 4.5 Household Debt Performance

Household head’s Social Class Times Region Times Times marital status Bangkok Agricultural 2.3 5.7 Single 4.9 (BKK) Central Non-agricultural 3.6 3.6 Married-couple 3.1 (excl. BKK) Married-widowed/ Professional 7.0 North 3.0 3.0 separated/ divorced

Worker 2.7 North-east 2.3 Retired 2.9 South 3.7

Note: Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

In addition, Figure 4.4 shows a scatter plot, which indicates a non-linear relationship between debt performance and head of households’ age. This observation has implications for both the theory and econometrics in explaining the drivers of household debt. In theory, the curvilinear relationship between the age of the household head and level of debt corroborates the Life-Cycle Hypothesis (as explained in Chapter 2). In econometrics, the curvilinear relationship of the level of household debt suggests the inclusion of a quadratic term in the regression model which has the level of household debt as the dependent variable.

Figure 4.4 Debt Performance by Household Head’s Age

Debt_Performance

0 20 40 60 80 100 HH_Head_Age

Note: Only households with positive debt performance and debt performance less than 50 times.

Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Proportion of over-indebted households

The proportion of subjectively indebted households and the proportion of objectively indebted households by social class, region and head of household’s marital status are explained next. Recall that, subjectively over-indebted households are households with a concern about the next debt repayment, while objectively over- indebted households are households with the ratio of income after debt service payment to poverty line less than one (Equation 4.3 when DP less than 1). Table 4.6 shows the similar orders of the proportion of subjectively and objectively over-indebted households. By social class, agricultural households have the highest proportion of subjectively and objectively over-indebted households at 21.5% and 26.8%, respectively. By region, households in North-eastern Thailand have the highest proportion (23–25%) of both subjectively and objectively over-indebted households. Lastly, by the head of household’s marital status, around 14-17% of households with both married-couple and married (widowed, separated, divorced) head of household are over-indebted.

Table 4.6 Proportion of Over-indebted Households

Sub. Obj. Sub. Obj. Household head’s Sub. Obj. Social Class Region (%) (%) (%) (%) marital status (%) (%) Bangkok Agricultural 21.5 26.8 7.0 2.6 Single 4.2 3.9 (BKK) Non- Central 17.1 10.6 10.9 9.4 Married-couple 17.1 16.8 agricultural (excl. BKK) Married-widowed/ Professional 5.6 3.0 North 15.2 14.4 14.5 14.4 separated/ divorced

Worker 15.8 11.4 North-east 22.8 25.0 Retired 11.7 18.2 South 12.1 11.6 Note: Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations. The bivariate correlations between the level of household debt and the extent of over-indebtedness provide evidence on the variables that may be candidates to be considered as the drivers of indebtedness. Next, this thesis will move to the methods used to investigate the drivers of household indebtedness. Firstly, the term household indebtedness will be clarified that is being used in the quantitative analysis. Secondly, a list of variables will be provided which will be used in the analysis.

4.3 Methodology

For the analysis of the drivers of household indebtedness, in this chapter, this research employs quantitative methods. This quantitative analysis uses both continuous and dichotomous terms of indebtedness. The analysis with continuous terms will be called ‘household debt performance analysis’, whereas the analysis with dichotomous terms will be called ‘household over-indebtedness analysis’.

Firstly, for the household debt performance analysis, ordinary least squares (OLS, using robust standard errors) will be estimated corresponding to the continuous dependent variable, which is the ratio of household’s monthly income after debt service payments to household’s poverty line (Equation 4.3-4.4). This ratio reflects the distance between income after debt service payments and subsistence level of income (Keese, 2009; D’Alessio & Lezzi, 2013). It is the income cushion after debt repayments (debt repayment is non-discretionary expense), which can be used as an early indicator for debt default. Moreover, this proxy of debt performance will be linked to the measure of objective over-indebtedness, which will be explained next. 87

Secondly, focusing on the household over-indebtedness analysis, logit models (using robust standard errors) will be estimated for both subjective and objective measurements. A dummy variable for households with a concern about the next debt repayment (0 = N, 1 = Y) will be used as a dependent variable for the subjective over- indebtedness estimation. In contrast, a dummy variable for households with the ratio of income after debt service payment to household’s poverty line less than one (Equation 4.3 when DP less than 1; 0 = N, 1 = Y) will be used as a dependent variable for the objective over-indebtedness estimation.

The list of explanatory variables of household indebtedness is provided in Table 4.7 below.

Table 4.7 List of Explanatory Variables

Variable Term

Perception of having secure income (regular income) Dummy (0 = N, 1 = Y) Expenditure to income ratio Level Insufficient savings in case of income shocks Dummy (0 = N, 1 = Y) (savings can fund monthly expenditure less than 3 months) Dependency ratio (number of household members with no income, excluding household members who help family businesses, to total number of Level household members) Age Level Main source of fund Dummy (0 = N, 1 = Y) (Bank for Agricultural and Agricultural Cooperatives: BAAC) Main source of fund Dummy (0 = N, 1 = Y) (Government Housing Bank and Government Saving Bank: GHB and GSB) Main source of fund (non-banks) Dummy (0 = N, 1 = Y) Main source of fund (cooperatives) Dummy (0 = N, 1 = Y) Main source of fund (informal source) Dummy (0 = N, 1 = Y) Credit constraint (in the case of emergency) Dummy (0 = N, 1 = Y) Credit constraint (in case of working purposes) Dummy (0 = N, 1 = Y) Credit constraint (rejection from formal sector) Dummy (0 = N, 1 = Y) Mortgage loans to annual income ratio Level Agricultural business loans to annual income ratio Level Non-agricultural business loans to annual income ratio Level Education loans to annual income ratio Level Personal consumption loans to annual income ratio Level Other purpose loans to annual income ratio Level Financial literacy score (total) Level Financial literacy score (knowledge) Level Financial literacy score (interest-rate calculations) Level Financial literacy score (behaviour) Level Hyperbolic discounting behaviour Dummy (0 = N, 1 = Y) (“I am happy with spending now more than saving for the future.”) Having problems with other financial commitments over the past twelve Dummy (0 = N, 1 = Y) months (rent or utility bills) Doing income-and-expenditure accounts Dummy (0 = N, 1 = Y)

Table 4.8 shows the list of control variables. This study employs social class, region, and the marital status of household heads as the control variables. This set of factors was selected to screen the effect of the heterogeneity of occupations, living areas, and marital statuses, respectively, in terms of debt performance and over- indebtedness status, as mentioned in the previous section.

Table 4.8 List of Control Variables

Dummy variable for social class (using professional as a benchmark)

Variable Term Agricultural business Dummy (0 = N, 1 = Y) Non-agricultural business Dummy (0 = N, 1 = Y) Worker Dummy (0 = N, 1 = Y) Retired Dummy (0 = N, 1 = Y) Dummy variable for region (using Bangkok as a benchmark) Variable Term Central (exclude Bangkok) Dummy (0 = N, 1 = Y) North Dummy (0 = N, 1 = Y) North-east Dummy (0 = N, 1 = Y) South Dummy (0 = N, 1 = Y) Dummy variable for household head’s marital status (using single as a benchmark) Variable Term Married (couple) Dummy (0 = N, 1 = Y) Married (widowed/ separated/ divorced) Dummy (0 = N, 1 = Y)

Next, follows a discussion of the empirical findings from the analysis of the drivers of household indebtedness and it will compare the findings with the previous studies in this area.

4.4 Empirical results

In this section, the empirical results have been divided into two parts: (i) the results from the household debt performance analysis; and (ii) the results from the household over-indebtedness analysis. Firstly, Table 4.9 provides the results from ordinary least squares (OLS) estimates on the determinants of household debt (i.e. the household debt performance analysis). Table 4.9 shows that the perception of having secure income has a positive relationship with household debt performance. It can be inferred that having secure income lets households attain a better debt status because

90 households with secure income seem to have better debt performance than households with no secure income.

However, insufficient savings (in case of income shocks) and the dependency ratio are negatively correlated with debt performance. Having insufficient savings reflects low ability to save that can adversely affect the performance of debt financing. Whereas a higher dependency ratio can lead to poorer debt performance because of higher financial burdens with dependent members (e.g. educational expenses, health expenses, and cost of living).

Moreover, in the case of Thailand, the results from OLS suggest that the turning points in age terms (level and squared term) is within the age-bracket of 45-55 years which corroborates the turning point shown in Figure 4.4. Moreover, Figure 4.5 shows that there is a higher proportion of households with the expectation of better households’ economic situation among household heads at the age between 21 and 60 years old (i.e. working age) compared with young or retired household heads. It can be explained that, on average, these households expect to earn more income during working age, which can lead to better debt performance. The above empirical findings support the Life-Cycle Hypothesis (LCH) and the Permanent Income Hypothesis (PIH).

Regarding credit constraints, the limited access to formal business loans plays an important role in deteriorating debt performance. For the effects of loans by type, agricultural loans have the largest effect on debt performance. The size of the effect of mortgage loans is smaller than the others, while the effect of non-agricultural business loans is not significant. To explain the effect of mortgage loans, the effect is small because mortgage loans are secured loans with collateral, and the effective interest rate (annual interest payment/debt outstanding) is lower than the others (Figure 4.6).

In addition, regarding the role of behavioural finance, households with poor financial literacy, and households with unsound financial management have weaker debt performance. Households with lower financial literacy, especially in terms of interest-rate calculations, seem to have a poorer debt status than households with higher financial literacy. Moreover, people who have problems with other financial commitments (i.e. rent and utility bills) over the past twelve months also have a poorer debt status. Whereas, keeping doing income-and-expenditure accounts can help people to improve their debt status.

Table 4.9 Debt Performance Models (OLS) Variable Model1 Model2 Model3 Model4 Perception of having secure income (dummy) 0.583*** 0.520*** 0.531*** 0.627*** Expenditure to income ratio (level) -0.24 -0.24 -0.209 -0.214 Expenditure to income ratio (squared term) 0.0004 0.0004 0.0003 0.0004 Insufficient savings (dummy) -0.660*** -0.763*** Dependency ratio (level) -2.521*** -2.618*** -2.690*** -2.637*** Age (level) 0.039** -0.004 0.041** 0.063*** Age (squared term) -0.0004** 5.15E-06 -0.0004** -0.001*** Main source of fund (BAAC) -1.099*** -1.031*** Main source of fund (GHB and GSB) -0.163 -0.032 Main source of fund (non-banks) -0.853*** -0.686*** Main source of fund (cooperatives) 1.207*** 1.309*** Main source of fund (village fund) -1.485*** -1.355*** Main source of fund (informal source) -0.905*** -0.904*** Credit constraint (in case of working purposes) -0.347*** -0.372*** Credit constraint (in case of emergency) -0.385*** -0.327*** Mortgage loans to annual income ratio -0.178** -0.146* Agricultural business loans to annual income ratio -0.746*** -0.726*** Non-agricultural business loans to annual income ratio -0.103 -0.095 Education loans to annual income ratio -0.498*** -0.555*** Personal consumption loans to annual income ratio -0.580*** -0.541*** Other purpose loans to annual income ratio -0.454** -0.507** Financial literacy score (total) 0.241*** 0.219*** Financial literacy score (knowledge) 0.326*** Financial literacy score (behaviour) 0.300*** Financial literacy score (attitude) -0.019 Financial literacy (interest-rate calculations) 0.503*** Hyperbolic discounting behaviour 0.021 Having problems with other financial commitments -0.690*** -0.570*** -0.576*** (dummy) Doing income-and-expenditure accounts 1.162*** 1.116*** 1.516*** Dummy variable for social class (using professional as a benchmark) Agricultural business -4.105*** -4.196*** -3.549*** -3.723*** Non-agricultural business -2.946*** -2.973*** -2.801*** -2.911*** Worker -3.756*** -3.739*** -3.625*** -3.772*** Retired -1.981*** -1.849*** -1.736*** -1.993*** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) -1.592*** -1.624*** -1.469*** -1.499*** North -1.811*** -1.839*** -1.548*** -1.510*** North-east -2.205*** -2.235*** -1.841*** -1.798*** South -1.209** -1.241** -1.293** -1.181** Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) -0.945*** -0.564*** Married (widowed/ separated/ divorced) -0.420** -0.216 Constant 5.693*** 6.692*** 5.448*** 7.817*** N 9,430 9,430 9,430 9,430 Adjusted R2 0.134 0.134 0.148 0.144 Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. *** Significant at the 1% level.

** Significant at the 5% level.

* Significant at the 10% level.

Figure 4.5 Proportion of Households with Expectation of Better Households’ Economic Situation (by Household Head’s age group)

% 20 18.2 17.0 16.5 16.5

12.0

10 9.4

0 <=20 21-30 31-40 41-50 51-60 >60

Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Figure 4.6 Effective Interest Rate by Type of Loans

% 35 31.9 32.4 30

25 24.4 22.1 20 15.1 15

10 8.4

0 Mortgage Agricultural Business Educational Personal Other loans loans loans loans consumption purpose loans loans Notes: Effective interest rate = (Annual interest payment/Debt outstanding)x100. Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Moving on to the analysis of household over-indebtedness, Tables 4.10 and 4.1110 show the results from logit models in terms of marginal effects. The results suggest that having secure income can lead to a lower probability of being both subjectively and objectively over-indebted. It can be claimed that households with less fluctuating income tend to have manageable debts (i.e., they have less concern about their debt repayments). While having insufficient savings can cause a higher chance to be over-indebted. Additionally, the dependency ratio has a significant effect on the likelihood of being objective over-indebtedness. The more dependent members that households have, the higher chance of being objective over-indebtedness that they get. This finding is consistent with the previous studies by Betti, Dourmashkin, Rossi, and Yin (2007), Vante (2006), Anderloni and Vandone (2008) and Keese (2009).

Regarding the role of household head’s age, the turning point of age terms is within the same age-bracket of 45-55 years. To explain this finding, although these households have better average debt performance, there are higher proportions of over- indebted households by head count because this age group has the ability to borrow more money compared to the other groups (Figure 4.7 and Table 4.12). The highest proportion is in the group of household heads with age between 51-60 years old.

10 For model specifications, see Appendix A. 94

Table 4.10 Marginal Effects on the Probability of Being Subjectively Over-indebted Variable Marginal Effect Perception of having secure income (dummy) -0.031*** Expenditure to income ratio (level) 0.019** Expenditure to income ratio (squared term) -0.002*** Insufficient savings (dummy) 0.062*** Dependency ratio (level) 0.019* Age (level) 0.004** Age (squared term) -4E-05** Main source of fund (BAAC) 0.283*** Main source of fund (GHB and GSB) 0.197*** Main source of fund (non-banks) 0.239*** Main source of fund (cooperatives) 0.212*** Main source of fund (village fund) 0.244*** Main source of fund (informal source) 0.331*** Credit constraint (in case of working purposes) 0.04*** Mortgage loans to annual income ratio 0.025*** Agricultural business loans to annual income ratio 0.028*** Non-agricultural business loans to annual income ratio 0.035*** Education loans to annual income ratio 0.071*** Personal consumption loans to annual income ratio 0.018*** Other purpose loans to annual income ratio 0.035 Financial literacy (interest-rate calculations) -0.004 Hyperbolic discounting behaviour 0.004 Having problems with other financial commitments (dummy) 0.104*** Doing income-and-expenditure accounts -0.043*** Dummy variable for social class (using professional as a benchmark) Agricultural business 0.055*** Non-agricultural business 0.081*** Worker 0.075*** Retired 0.044** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 0.006 North 0.008 North-east 0.023 South 0.024 Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 0.028 Married (widowed/ separated/ divorced) 0.043** Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level.

Table 4.11 Marginal Effects on the Probability of Being Objectively Over-indebted Variable Marginal Effect Perception of having secure income (dummy) -0.027*** Expenditure to income ratio (level) 0.18*** Expenditure to income ratio (squared term) 0*** Insufficient savings (dummy) 0.039*** Dependency ratio (level) 0.14*** Age (level) 0.004*** Age (squared term) -4E-05*** Main source of fund (BAAC) 0.07*** Main source of fund (GHB and GSB) 0.007 Main source of fund (non-banks) 0.019* Main source of fund (cooperatives) 0.025 Main source of fund (village fund) 0.066*** Main source of fund (informal source) 0.049*** Credit constraint (in case of working purposes) 0.019*** Mortgage loans to annual income ratio 0.01* Agricultural business loans to annual income ratio 0.058*** Non-agricultural business loans to annual income ratio 0.043*** Education loans to annual income ratio -0.06*** Personal consumption loans to annual income ratio 0.017*** Other purpose loans to annual income ratio 0.03 Financial literacy (interest-rate calculations) -0.024*** Hyperbolic discounting behaviour 0.004* Having problems with other financial commitments (dummy) 0.06*** Doing income-and-expenditure accounts -0.039** Dummy variable for social class (using professional as a benchmark) Agricultural business 0.12*** Non-agricultural business 0.038** Worker 0.053*** Retired 0.039** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 0.149*** North 0.176*** North-east 0.184*** South 0.157*** Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 0.112*** Married (widowed/ separated/ divorced) 0.073*** Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level.

Figure 4.7 Debt to Annual Income Ratio by Household Head’s Age

Debt_to_Income_Ratio

20 40 60 80 100 HH_Head_Age

Note: Only households with debt to annual income ratio less than 20 times. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Table 4.12 Proportion of Over-indebted Households by Household Head’s Age

Subjective over-indebted Objective over-indebted Age (years) households (%) households (%) <= 20 3.8 1.6 21 – 30 9.8 6.3 31 – 40 13.3 14.8

41 – 50 17.3 15.8 51 – 60 18.5 16.5 > 60 13.9 15.9 Note: Using population weights Source: Household Socio-economic Survey, National Statistical Office. Author’s calculations. Focusing on the effects of loans by type, according to the marginal effect summary from the logit models, education loans have the largest effect on the probability of being subjectively over-indebted, while agricultural loans have the largest effect on the probability of being objectively over-indebted. The policy implication of this particular finding will be discussed in detail in Chapter 7. The marginal effect of mortgage loans is still lower than the others.

Regarding behavioural finance issues, households with poor financial literacy and hyperbolic discounters have a higher chance to be objectively over-indebted. Households with low financial literacy in terms of the ability to do interest-rate calculations seem to have a greater probability of being objectively over-indebted. 97

Moreover, households who are hyperbolic discounters (i.e. households with over- discount on their future consumption) tend to be objectively over-indebted. However, these factors (financial literacy and hyperbolic discounting behaviour) do not have significant effects on subjective over-indebtedness.

4.5 Conclusion

To sum up, this chapter explores the nature of and the dangers posed by household over-indebtedness in Thailand. It has reviewed the three theories that provide reasons why households take debt; namely (i) neoclassical economic theory, (ii) the theory of credit rationing, and (iii) behavioural finance theory. Data is then compiled to allow identification of the drivers of the levels of the indebtedness of Thai households. It has found that all above theories play a role in accounting for indebtedness in Thailand. The key findings of the chapter may be summarised as follows.

First of all, the empirical results support the Life Cycle and Permanent Income Hypothesis, which confirms the findings of the literature (Betti et al., 2007; Vante, 2006; Anderloni & Vandone, 2008; Keese, 2009). In the case of Thailand, households that are ‘rational’ in the neoclassical sense - as evidenced by their doing income-and- expenditure accounts - and who perceive themselves to have secure income are more likely to have better debt performance and are less likely to be both subjectively and objectively over-indebted.

With regards to the household head’s age, although household heads with working age (21-60 years of age) have a better debt performance, these household groups have a greater percentage of both subjectively and objectively over-indebted households. In addition, the greater the number of children or retirees in these households (i.e. higher dependency ratio), the worse is their debt performance; which is to say that the dependency ratio has a significant impact on both debt performance for households whose head is in this age bracket. The overall conclusion is that households whose head is of working age perform better, on average, than other groups, but they have greater risks of falling into underperformance (i.e. there is a fatter tail to the distribution for this group of households).

Secondly, focusing on credit rationing, households who do not have access to the formal sources of credit have poorer debt performance and have a higher chance to be over-indebted. This result confirms the findings from the previous study by Betti, 98

Dourmashkin, Rossi, and Yin (2007), who suggested that the credit constraint was positively related to household over-indebtedness. Moreover, regarding the role of different types of loans, agricultural business loans have the largest impact on both debt performance and objective over-indebtedness. This finding implies the inherent poverty of Thai agricultural households coupled with the fact that this type of loan is not secured by collateral and has higher interest rate than secured loans such as mortgages.

Considering the role of education loans, which are another form of unsecured loan, these have the greatest positive effect on subjective over-indebtedness, but they have a negative effect on objective over-indebtedness. These results imply that indebted households, who expect themselves to be subjectively worse off as a result of taking out education loans, tend to overestimate the level of actual (i.e. objective) indebtedness. Consequently, some people collectively under-invest in their education. In light of the externality benefits of education on the Thai economy, this evidence suggests that a wedge is driven between subjectively-motivated educational choices of individuals and the level of education loans that are optimal for Thailand. This issue will be discussed more in Chapter 7.

Thirdly, moving on to the point of behavioural finance, financial literacy and hyperbolic discounting behaviour have impacts on debt performance and objective over- indebtedness. With regards to financial literacy, the results affirm that households with poor financial literacy have a higher likelihood of being over-indebted. Whereas the result about hyperbolic discounting behaviour shows the positive relationship between self-control problems (hyperbolic discounters) and a chance of being over-indebted.

Considering households more generally in terms of financial statements, this study has found that having insufficient savings to deal with income shocks and having problems with rent or utility bills are significantly correlated with debt performance and over-indebtedness. This finding suggests a useful indicator for banks and authorities in terms of credit applications. Although this finding is not surprising, it is still a useful implication of the analysis conducted. This will be discussed further in Chapter 7.

For further analysis, in this study, objective and subjective measures of household indebtedness have been employed, and some discrepancies between the two measures have been found. In particular, as mentioned above, it has been found that education loans have the highest marginal impact on the likelihood of being

99 subjectively over-indebted; while agricultural business loans have the highest marginal impact on the likelihood of being objectively over-indebted; while mortgages have a lower impact than the other two. Moreover, in the case of Thailand, financial literacy affects the probability of being objectively over-indebted and has no significant impact on subjective indebtedness. Therefore, it is appropriate to investigate the differences between these two types of household over-indebtedness (i.e. subjective over- indebtedness and objective over-indebtedness).

In the next chapter, the further analysis will be done to clarify the different characters between subjectively and objectively over-indebted households in Thailand. The study of this issue can help to extend the knowledge of household heterogeneity in developing countries.

100

Chapter 5: Differences between Subjective Over-indebtedness and Objective Over-indebtedness: Household-Level Evidence from Thailand11

5.1 Introduction

As mentioned in Chapter 2, an increase in household debt is not necessarily bad for the household or the macro-economy. The effects of a rise in household debt on the aggregate GDP and total employment may vary depending on the context. An increase in household debt can be due to an improvement in households’ credit access, possibly a reflection of the maturation of financial markets (Debelle, 2004; Kang & Ma, 2009). However, most of the extant literature argues that rising household debt can increase financial fragility leading to adverse economic consequences. In the case where household debt increases the fragility of the financial sector, the effects on the economy can be negative (Barnes & Young, 2003; Debelle, 2004; Cecchetti et al., 2011). A sharp rise in household debt can increase the vulnerability of households with debt to adverse interest-rate and income shocks (Debelle, 2004; Endut & Hua, 2009). Moreover, according to the UK’s and the US’ experiences, a high level of household leverage can lead to an economic downturn in subsequent periods through falls in household consumption and saving (Debelle, 2004; Barba & Pivetti, 2009). Martin (2011) also claimed that in the case of the US, the sub-prime mortgage boom during 2001 to 2006 led to the bubble in house prices that was followed by debt defaults leading to the global financial crisis in the next two years. This financial crisis caused broad-and-serious adverse effects on the real economy and financial markets (i.e. economic downturn and more volatility in financial markets) in both advanced and emerging countries (Claessens et al., 2010; Longstaff, 2010; Aloui et al., 2011). Therefore, persistent rates of rapid growth in household leverage can have an adverse effect on the financial fragility of the household, and if widespread within the population then on the macroeconomy as well (Barnes & Young, 2003; Debelle, 2004; Endut & Hua, 2009; Cecchetti et al., 2011; Martin, 2011). Moreover, in the case where the rapid growth of debt amongst the households was from insufficient savings and low income, then an interest

11 An earlier version of this chapter was published as a conference paper (See Chotewattanakul, P., Sharpe, K., & Chand, S., 2018, ‘Differences between subjective over-indebtedness and objective over- indebtedness: household-level evidence from Thailand’, Asia-Pacific Conference on Economics and Finance e-proceeding, Singapore, 2018. ISBN: 978-981-11-6602-0). 101 rate rise could lead to bankruptcies, which if widespread, could cause a recession. The Global Financial Crisis in 2008 is a good example of the negative effects of an unwinding of high levels of household debt on the macro-economy.

But how do policymakers decide if households are over-indebted? Many researchers have tried to define household over-indebtedness (Haas, 2006; Anderloni & Vandone, 2008; Keese, 2009; European Commission, 2010; D’Alessio & Iezzi, 2013), but there is no unique definition for household over-indebtedness. Two common measurements to identify household over-indebtedness, as explained in the previous chapter, are (i) subjective measurement and (ii) objective measurement. Subjective over-indebtedness is related to households’ self-reporting of their debt service difficulties (Anderloni & Vandone, 2008; D’Alessio & Iezzi, 2013), whereas objective over-indebtedness is calculated in terms of households’ financial ratios, which are used to define their financial vulnerability (Haas, 2006; Betti et al., 2007; Keese, 2009; D’Alessio & Iezzi, 2013).

This chapter will explore the relationship between subjective and objective over- indebtedness, and the factors underscoring the correlations (if any) between them. This study will show systemic and systematic factors within the data with respect to the observed correlations between these two types of over-indebtedness. As an example, both subjective and objective over-indebtedness are highly correlated for households with a secure income (regular income) and those who keep regular financial accounts. Conversely, households with illiterate heads tend to have an inconsistent relationship between subjective and objective over-indebtedness. Meanwhile, households with many dependants have a high chance of being objectively over-indebted, but they do not self- report as being over-indebted. Moreover, loans taken to fund education have a bigger impact on subjective assessment of over-indebtedness than what the objective measure reveals. Each of the above-mentioned cases has clear policy implications; namely, (i) improved literacy will improve assessment of debt levels by households; (ii) dependants are not costed fully in household accounts, thus supply of this information is likely to assist in debt assessments by households; and (iii) outlays on education may be less than what is socially desirable meaning that some forms of income contingent loans are likely to improve both the welfare of the household and society collectively. Finally, the findings point to the importance of using both subjective and objective measures for characterising household debt analysis and for implementing policies to address this 102 problem. In other words, relying on only subjective or objective indicators may lead to biased results (Keese, 2010). Thus, relying on single measures of over-indebtedness may only provide a partial picture of household debt.

The chapter is divided into five sections. The motivation of the study is narrated in Section 1. In the second section, the data is described, and summary statistics are presented. The quantitative techniques used to decipher the drivers of household debt are explained in the third section. Section 4 interprets the empirical results; the last section draws implications of these results for policy-making.

5.2 Data and overview of the differences between subjective over-indebtedness and objective over-indebtedness in Thailand

5.2.1 Data

The data for the subsequent analysis is sourced from Thailand’s Household Socio-Economic Surveys, administered by the National Statistic Office and the Bank of Thailand’s in the first quarter of 2013. A stratified random sample of 10 661 households was used to select participants from each of the district in all 77 provinces in Thailand. Therefore, the sample of selected households is representative of the whole household population. The sampled households can be disaggregated by geography (e.g. province, and region), household head’s characteristic (e.g. gender, age, marital status, and education level), and household’s social class (i.e. agricultural business, non- agricultural business, worker, professional, and retiree). In terms of households’ financial status, data is provided on the monthly income, monthly expenditure, assets and debt outstanding. Moreover, the Socio-Economic Surveys contain information on household’s access to the credit market and problems with overdue rent or utility bills.

The above data have been supplemented with information from the Bank of Thailand’s supplement to the household survey which has collected information on the following items:

103

(i) The level of households’ financial literacy (based on the Organization for Economic Co-operation and Development’s survey of financial literacy12),

(ii) The efficacy of financial management (e.g. doing income-and-expenditure accounts, and saving sufficiency), and

(iii) the perception of having debt difficulties.

Data on the last of the above was collected as response to the question: “does household have concerns about the next debt repayment?”

Information on the expectations of households concerning their economic circumstances is provided through indicators of both current and expected economic conditions (i.e. worse, no change, and better) in this supplement survey. Finally, the supplement survey contains households’ views of the likelihood of government help when the borrowers fall into credit arrears.

As explained in Chapter 4, this study dropped 34 households which had zero or negative income and another four households which had debt service ratio greater than 10 times, leaving a total of 10 623 for the analysis.

5.2.2 Overview descriptive statistics

This section provides some overview statistics of the proportion of indebted households by indebtedness status regarding two types of over-indebtedness, namely, subjective over-indebtedness and objective over-indebtedness. Firstly, it will focus on the proportion of indebted households by indebtedness status with regards to all indebted households. Then, it will clarify the proportion of indebted households with consistent debt perceptions for both subjective and objective perspectives (i.e. correct debt perceptions) and the proportion of indebted households with inconsistent debt perceptions (i.e. incorrect debt perceptions). Next, it focuses on the proportion of indebted households with consistent debt perceptions by social class, region, and household head’s education level. Lastly, the proportion of indebted households with inconsistent debt perceptions by the same set of characteristics is examined.

12 According to OECD INFE (2011), “Measuring Financial Literacy: Questionnaire and Guidance Notes for Conducting and Internationally Comparable Survey of Financial Literacy”, household financial literacy score is divided into three main categories: - Knowledge score (the questions about interest rate and inflation) - Behaviour score (the questions about saving and spending behaviours) - Attitude score (the questions about hyperbolic discounting behaviour) 104

The proportion of indebted households by indebtedness status

Concerning the definition of household over-indebtedness, it has been defined as subjectively over-indebted households as those expressing a concern about their next debt repayment (Anderloni & Vandone, 2008; Keese, 2010); whereas, objectively over- indebted households are defined as households whose ratio of income after debt service payment to the poverty line is less than one (Equation 4.1 when DP less than 1) (D’Alessio & Lezzi, 2013). Of the 10 623 households in my dataset, 55.4% had debt (Figure 5.1). Indebted households are divided into four groups: (i) indebted households that are not subjectively and objectively over-indebted; (ii) indebted households that are objectively over-indebted but not subjectively over-indebted; (iii) indebted households that are subjectively over-indebted but not objectively over-indebted; and (iv) indebted households that are both subjectively and objectively over-indebted (see Table 3.1). From Table 5.1, data from the surveys show that 60.6% of indebted households are not subjectively and objectively over-indebted, and 8.9% are both subjectively and objectively over-indebted. There is around 30% of indebted households that do not have the ‘correct’ perception of their degree of indebtedness (i.e. households with inconsistent debt perceptions). From this group of households, around 20% of indebted households are subjectively over-indebted but not objective over-indebted, while another 10% of the households have the opposite perception (are objectively over- indebted but not subjectively over-indebted).

Figure 5.1 Proportion of Indebted Households

44.6%

55.4%

HH with debt HH with no debt

Note: Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations. 105

Table 5.1 Proportion of Indebted Households by Indebtedness Status1/

% of indebted households Not objectively over-indebted Objectively over-indebted Not subjectively over-indebted 60.6% (type 1) 11.8% (type 2) Subjectively over-indebted 18.7% (type 3) 8.9% (type 4) Note: 1/ Using population weights Source: Household Socio-economic Survey, National Statistical Office. Author’s calculations.

Households with consistent debt perceptions by social class, region, and household head’s education level

This section investigates the attributes of households with consistent subjective and objective assessments of indebtedness (i.e. indebted households who are not subjectively and objectively indebted [household type 1], and indebted households who are both objectively and subjectively indebted [household type 4]) before focusing on those with inconsistent assessments (i.e. indebted household who are objectively over- indebted but not subjectively over-indebted [household type 2], and indebted household who are subjectively over-indebted but not objectively over-indebted [household type 3]). Social class, region, and household head’s education level are considered as the main attributes.

This part focuses on household type 1, first, because this group of households has correct debt perceptions without being over-indebtedness. Table 5.2 shows that 87% of households whose head is a professional occupy belong to this group (i.e. household type 1: neither subjectively nor objectively over-indebted), while Table 5.3 shows that 78% of households from Bangkok are also in Cell 1. This coincidence of the location of professional households and households in Bangkok may be due to the fact that most professional occupations are located in the capital city. Focusing on the level of education of the household head, Table 5.4 shows that Cell 1 comprises of household heads with high levels of education. This may be explained by the fact that higher levels of education are correlated with professional occupations and also with high levels of financial literacy.

Moreover, over a lifetime, professional households tend to earn more income and to have better debt status when they obtain higher levels of education. This evidence supports the Permanent Income Hypothesis since such household heads are expected to have high permanent income and thus the capacity to take on larger debts when young.

106

According to the above statistics, in the case of Thailand, this study can claim that professionals with a bachelor’s degree or higher in Bangkok have the highest degree of consistency between objective and subjective measures. This is not surprsing since the financial literacy of this type of person is bound to be greater than the other groups. Next, this analysis will move on to the attributes of households with the incorrect (inconsistent) debt perceptions.

Table 5.2 Proportion of Indebted Households by Indebtedness Status and Social Class

Social Class % of type 11/ 2/ % of type 21/ 2/ % of type 31/ 2/ % of type 41/ 2/ Total Agricultural 52.6 17.9 16.7 12.7 100 Non-agricultural 63.7 8.6 21.3 6.5 100 Professional 87.1 3.4 8.7 0.8 100

Worker 62.3 7.3 22.4 8.0 100 Retired 48.4 19.9 19.0 12.7 100 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Table 5.3 Proportion of Indebted Households by Indebtedness Status and Region

Region % of type 11/ 2/ % of type 21/ 2/ % of type 31/ 2/ % of type 41/ 2/ Total Bangkok (BKK) 78.4 2.6 17.7 1.2 100 Central (excl. BKK) 70.8 6.8 17.1 5.4 100 North 61.4 12.4 19.6 6.6 100

North-east 49.2 17.2 19.2 14.3 100 South 68.8 6.6 19.0 5.5 100 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations. 107

Table 5.4 Proportion of Indebted Households by Indebtedness Status and Household Head’s Education Level Education level % of type 11/ 2/ % of type 21/ 2/ % of type 31/ 2/ % of type 41/ 2/ Total Pre/ primary school 53.5 15.0 20.8 10.6 100 Secondary school 67.1 7.5 18.9 6.5 100 Post-secondary/ Diploma 80.2 6.8 10.5 2.5 100 Bachelor’s degree 85.8 2.2 9.7 2.3 100 Master’s/ Doctoral degree 97.1 0.6 2.1 0.2 100 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Households with inconsistent debt perceptions by social class, region, and household head’s education level

This section will focus on type 2 households (i.e. households who are objectively over-indebted but not subjectively so) and type 3 households (i.e. households who are subjectively over-indebted but not objectively over-indebted). From Table 5.2-5.4, therefore, households in Cell 2 and Cell 3 are the target groups. By social class, from Table 5.2, agricultural households and retired households have a high proportion of indebted households in Cell 2 and Cell 3 (i.e. for these groups of households, the proportion is greater than 15% for both cells). The inconsistency between subjective and objective assessment of debt by the households provides valuable information on the possible underlying causes for this. The lack of education (and by implication also limited financial literacy) may be a causal factor for those in the agricultural sector, which is located in the North and the North-east (Table 5.3-5.4). This issue is investigated in the next section. Regarding retirees, the inconsistency for them in believing themselves to be subjectively over-indebted whilst on objective measures they are not over-indebted, is difficult to rationalise. One explanation may be expectations where retirees believe that their expenses (e.g. health expenses) will rise as they are getting old and cannot work anymore. However, the evidence to corroborate this assertion is difficult to compile. Again, this issue is explored in further details next.

108

For workers and households with their own non-agricultural business, there are high proportions of households that are subjectively over-indebted, but not objectively over-indebted (households in Cell 3). These number 22.4% and 21.3%, respectively. Whereas, these groups of households have low proportions of households who are objectively over-indebted but not subjectively over-indebted (households in Cell 2). Again, the above may be rationalised by the fact that businesses face greater volatility in income than salaried professionals and therefore the subjective perceptions on a given level of debt between the two groups are likely to be very different. However, testing this proposition with the available data is not possible.

Next, this research investigates the causal chain behind the patterns of correlations noted above. It will explain the methods used for the analysis of the differences between household subjective over-indebtedness and objective over- indebtedness first and then present the variables used in the models.

5.3 Methodology

Quantitative methods will be employed for the study of the differences between household subjective over-indebtedness and objective over-indebtedness. Two logit models are estimated to identify the drivers of subjective over-indebtedness and objective over-indebtedness. Then, a comparison of the results from these logit estimations is made, and the differences between the parameters explained. A significant level of these differences will be calculated to identify that the effect of each factor on the chance of becoming subjectively over-indebted is more than, less than, or equal to the effect on a chance of becoming objectively over-indebted.

A proxy of subjective assessment of over-indebtedness in the form of a dummy variable for a household with a concern about the next debt repayment (0 = No, 1 = Yes) is used. While a dummy variable for a household with income after debt service payment less than the poverty line (0 = No, 1 = Y) is used as a proxy of objective over- indebtedness. According to the previous research on this issue (Keese, 2010), Tables 5.5 – 5.6 show the list of explanatory variables and control variables, respectively.

The list of explanatory variables is drawn from the literature. The explanatory variables detail the role of the life cycle hypothesis and the permanent income hypotheses, credit rationing, and behavioural finance, as have been used for the analysis of the drivers of household indebtedness (Chapter 4). Moreover, this analysis also uses 109 gender, the expectation of household’s economic status, and the expectation of government help for credit default as supplement factors that are mentioned in the previous study by Keese (2010).

The control variables, as listed in Table 5.6, for the regressions are social class, region, and the marital status of household heads as the control variables. In this analysis, the same set of control variables have been employed as in the analysis in Chapter 4. Again, the purpose of the use of these factors is to filter the impact of the heterogeneity of occupations, living areas, and marital statuses, respectively.

Table 5.5 List of Explanatory Variables

Variable Term Perception of having secure income Dummy (0 = N, 1 = Y) Expenditure to income ratio Level Insufficient savings in case of income shocks (less than 3 months) Dummy (0 = N, 1 = Y) Dependency ratio Level Age Level Main source of fund Dummy (0 = N, 1 = Y) (Bank for Agricultural and Agricultural Cooperatives: BAAC) Main source of fund Dummy (0 = N, 1 = Y) (Government Housing Bank and Government Saving Bank: GHB and GSB) Main source of fund (non-banks) Dummy (0 = N, 1 = Y) Main source of fund (cooperatives) Dummy (0 = N, 1 = Y) Main source of fund (village fund) Dummy (0 = N, 1 = Y) Main source of fund (informal source) Dummy (0 = N, 1 = Y) Credit constraint (in case of working purposes) Dummy (0 = N, 1 = Y) Mortgage loans to annual income ratio Level Agricultural business loans to annual income ratio Level Non-agricultural business loans to annual income ratio Level Education loans to annual income ratio Level Personal consumption loans to annual income ratio Level Other purpose loans to annual income ratio Level Financial literacy score (interest rate) Level Hyperbolic discounting behaviour Dummy (0 = N, 1 = Y) (“I am happy with spending now more than saving for the future.”) Having problems with other financial commitments over the past twelve Dummy (0 = N, 1 = Y) months (rent and utility bills) Doing income-and-expenditure accounts Dummy (0 = N, 1 = Y) Dummy (1 = Female, 0 Sex = Male) Dummy (1 = Worse, 0 Expectation of household’s economic status = otherwise) Expectation of government help for credit default Dummy (Y = 1, 0 = N) 110

Table 5.6 List of Control Variables

Dummy variable for social class (using professional as a benchmark) Variable Term Agricultural business Dummy (0 = N, 1 = Y) Non-agricultural business Dummy (0 = N, 1 = Y) Worker Dummy (0 = N, 1 = Y) Retired Dummy (0 = N, 1 = Y) Dummy variable for region (using Bangkok as a benchmark) Variable Term Central (exclude Bangkok) Dummy (0 = N, 1 = Y) North Dummy (0 = N, 1 = Y) North-east Dummy (0 = N, 1 = Y) South Dummy (0 = N, 1 = Y) Dummy variable for household head’s marital status (using single as a benchmark) Variable Term Married (couple) Dummy (0 = N, 1 = Y) Married (widowed/ separated/ divorced) Dummy (0 = N, 1 = Y)

Next, this study will show and discuss the empirical findings from the analysis.

5.4 Empirical results

After the two logit models are estimated for the determinants of subjective over- indebtedness and the determinants of objective over-indebtedness, Table 5.713 shows the empirical results from the models. The parameter estimates in Column 2 and Column 3 in Table 5.7 show the results from logit models in terms of marginal effects on the probability of being subjectively over-indebted the probability of being objectively over-indebted, respectively. Then Column 4 in Table 5.7 represents the differences between the parameters (absolute terms) from the subjective over- indebtedness model and the parameters (absolute terms) from the objective over- indebtedness model, including the significant level of each difference.

Regarding the comparison of these marginal effects using logit estimations, the differences between these parameter estimates are classified into four main categories as follows:

13 For model specifications, see Appendix B. 111

(i) An effect on subjective over-indebtedness is similar to an effect on objective over-indebtedness.

(ii) An effect on subjective over-indebtedness is more than an effect on objective over-indebtedness.

(iii) An effect on objective over-indebtedness is more than an effect on subjective over-indebtedness.

(iv) There is only an effect on objective over-indebtedness or subjective over- indebtedness.

Table 5.7 Marginal Effects on the Probability of Being Over-indebted Marginal Effect (ME) Variable Subjective Objective |ME_sub| - |ME_obj| Perception of having secure income -0.027*** -0.026*** 0.002 (dummy: Y = 1, N = 0) Expenditure to income ratio (level) 0.016** 0.18*** -0.163*** Expenditure to income ratio (squared term) -0.002*** -2.9E-04*** 0.002*** Insufficient savings (dummy) 0.056*** 0.037*** 0.019** Dependency ratio (level) 0.013 0.138*** - Age (level) 0.003** 0.004*** -0.001 Age (squared term) -3.2E-05** -3.9E-05*** -6.6E-06 Main source of fund (BAAC) 0.289*** 0.071*** 0.218*** Main source of fund (GHB and GSB) 0.206*** 0.01 - Main source of fund (non-banks) 0.244*** 0.02* 0.224*** Main source of fund (cooperatives) 0.219*** 0.027 - Main source of fund (village fund) 0.247*** 0.067*** 0.18*** Main source of fund (informal source) 0.328*** 0.046*** 0.282*** Credit constraint (in case of working purposes) 0.035*** 0.018** 0.017* Mortgage loans to annual income ratio 0.025*** 0.01* 0.015* Agricultural business loans to annual income ratio 0.025*** 0.058*** -0.032** Non-agricultural business loans to annual income ratio 0.035*** 0.043*** -0.008 Education loans to annual income ratio 0.072*** -0.059*** - Personal consumption loans to annual income ratio 0.016*** 0.016*** 3.6E-04 Other purpose loans to annual income ratio 0.041 0.032* - Financial literacy (interest rate) -0.005 -0.024*** - Hyperbolic discounting behaviour 0.004 0.004* - Having problems with other financial commitments 0.091*** 0.054*** 0.037*** (dummy: Y = 1, N = 0) Doing income-and-expenditure accounts -0.04*** -0.037** 0.003 Sex (dummy: female = 1, male = 0) 0.013* -0.011 - 112

Expectation of household’s economic status 0.082*** 0.032*** 0.05*** (dummy: worse = 1, otherwise = 0) Expectation of government help for credit default -0.011 -0.002 - (dummy: Y = 1, N = 0) Dummy variable for social class (using professional as a benchmark) Agricultural business 0.054*** 0.119*** -0.065*** Non-agricultural business 0.077*** 0.037** 0.041 Worker 0.073*** 0.051*** 0.022 Retired 0.043** 0.04** 0.003 Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 0.007 0.154*** - North 0.014 0.182*** - North-east 0.031 0.191*** - South 0.016 0.158*** - Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 0.03* 0.11*** -0.079*** Married (widowed/ separated/ divorced) 0.036** 0.076*** -0.039 Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. *** Significant at the 1% level. ** Significant at the 5% level. * Significant at the 10% level. Firstly, there are four determinants that have the same size of effects on both subjective and objective over-indebtedness. A secure income and maintaining income- and-expenditure accounts have negative partial correlations with the probability of being subjectively over-indebted and objectively over-indebted equally. One may interpret these partial correlations as suggesting that households with secure income and those with high financial literacy are less prone to experience over-indebtedness. At the same time, the effects of age and personal consumption loans on these two types of over-indebtedness are similar in size.

Secondly, there are six factors that have greater effects on the probability of being subjectively over-indebted than the probability of being objectively over- indebted. Having problems with other financial commitments (rent and utility bills), insufficient savings in case of income shock (of less than three months), a credit constraint (in case of working purposes), and the expectation of a worse household’s economic status (e.g. employment status, occupation, and income level) cause households to become subjectively over-indebted, whereas these factors have smaller effects on the likelihood of being objectively over-indebted. Moreover, different sources 113 of credit accessed, which are not commercial banks, can make households worried about their due dates of debt repayment on a larger scale than the impact on their objective debt status. These findings suggest that the source of the loan has an impact on the subjective assessment of over-indebtedness. Additionally, mortgage loans cause households to have more concerns about their next debt repayments, while this type of loans has a smaller impact on the objective measure of over-indebtedness. This may suggest that household heads are particularly concerned about losing their homes, which are the collateral for their mortgages.

To explore this hypothesis further, this study will next focus on the role of properties (as debt collateral) in households’ subjective and objective debt status. Four types of indebted households (with or without mortgage loans) are categorised as follows:

(i) Indebted households who are not subjectively and objectively over-indebted.

(ii) Indebted households who are objectively over-indebted but not subjectively over-indebted.

(iii) Indebted households who are subjectively over-indebted but not objectively over-indebted.

(iv) Indebted households who are both subjectively and objectively over- indebted.

Figure 5.2 shows that indebted households who are objectively over-indebted but not subjectively over-indebted have a higher homeownership rate (including land ownership) compared to those who are subjectively over-indebted but not objective over-indebted. An explanation for this difference could be that those with collateral are able to access credit at lower costs and therefore, can take on more debt and are comfortable in doing so. Moreover, according to the previous findings on the role of financial management (doing income-and-expenditure accounts, having insufficient savings), people with debt collateral who have good financial discipline and can perceive their total debt burden tend to have a lower likelihood of being both subjectively and objectively over-indebted.

114

Figure 5.2 Homeownership Rate by Type of Indebted Households1/2/

% 90

87.2

77.9

70 Type 2 Type 3 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Moving on to the factors that have larger effects on the likelihood of becoming objectively over-indebted as against the likelihood of becoming subjectively over- indebted, a higher level of expenditure-to-income ratio leads to a worse objective debt status on a greater scale than the effect on the concerns about debt serviceability. In addition, business loans, especially agricultural business loans, have a greater impact on objective over-indebtedness than subjective over-indebtedness. Table 5.8 shows that household group with agricultural business loans and non-agricultural business loans have a higher proportion of households with debt repayment problems (i.e. late repayment, debt restructure, borrowing from other sources to repay the current debt, and debt default). Therefore, farmers and households with their own businesses have to assess their debt status precisely and have to learn a way to manage their business loans.

115

Table 5.8 Proportion of Indebted Households by Debt Repayment Status

Having non- Debt repayment status Having agricultural agricultural business business loans % of households with % of households with loans no problem1/ 2/ problems1/ 2/ No No 86.3 13.7 Yes No 84.0 16.0 No Yes 83.0 17.0 Yes Yes 78.4 21.6 Note: 1/ Debt repayment problems include late repayment, debt restructure, borrowing from other sources to repay the current debt, and debt default 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Fourthly, there are four factors that have effects on subjective over-indebtedness or objective over-indebtedness only. This analysis will begin with the role of dependency. The effect of dependency ratio on objective over-indebtedness is significant compared to the insignificant effect on subjective over-indebtedness (Table 5.7). Thai households may not perceive that more household members with no income can increase the household burden in terms of expenses. They think that their cash flows are manageable even though they sometimes have expenditure exceeding income. Focusing on indebted households with dependency ratio between 0 and 0.8 (96% of total indebted households), Table 5.9 shows that household groups with a higher dependency ratio have a higher proportion in Cell 2 (that is, objectively over-indebted but not subjectively over-indebted).

116

Table 5.9 Proportion of Indebted Households by Indebtedness Status and Dependency Ratio

Dependency ratio % of type 11/ 2/ % of type 21/ 2/ % of type 31/ 2/ % of type 41/ 2/ Total Dependency ratio <= 0.2 70.2 7.2 18.0 4.6 100 0.2 < Dependency ratio <= 0.4 59.1 12.4 20.1 8.3 100 0.4 < Dependency ratio <= 0.6 55.9 13.3 19.8 11.1 100

0.6 < Dependency ratio <= 0.8 56.2 15.9 15.5 12.4 100 0.8 < Dependency ratio 50.2 13.9 20.5 15.4 100 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Additionally, according to the results from logit models (Table 5.7), education loans have an opposite effect on subjective and objective over-indebtedness. Education loans make households more concerned about their debt repayments, while it has a negative relationship with the probability of being objectively over-indebted. While education loans create some debt service burden, this type of loans can also generate future income that could lift a household out of poverty. Thus, the ‘error’ in expectations in this area is a concern for policymakers - which is a matter discussed in later chapters (especially in Chapter 7).

Moreover, Table 5.7 shows that financial literacy (interest-rate calculations) and hyperbolic discounting behaviour only affect the probability of being objectively over- indebted. Having a lower level of financial literacy can result in poorer financial status with no concern about debt repayment since households cannot perceive the real cost of borrowing and their debt service burden in the next period. If households cannot do the correct calculations on their repayments, they will not know the exact cost of borrowing and their true debt burden. Moreover, households that perform poorly in terms of financial literacy do not seem to know that they perform poorly in terms of financial management. In other words, financially illiterate people are unaware of their illiteracy and its consequences. This is an important discovery of the thesis – which can only be

117 uncovered by a consideration of objective and subjective measures jointly. The policy implications are discussed in the conclusion of this chapter and in Chapter 7.

According to the previous four types of indebted households, Table 5.10 shows that households who are both subjectively and objectively over-indebted (type 4) have the lowest financial literacy score. In addition, households with incorrect debt perceptions (type 2 and 3) have lower average financial literacy score than households who are not subjectively and objectively indebted (type1) in terms of financial knowledge and financial behaviours (spending and saving behaviours). For financial attitude score (the question is about hyperbolic discounting behaviour), there are slight differences between households in type 1, 2 and 3. Meanwhile, the score for households who are both subjectively and objectively indebted are significantly lower than the other three groups of households.

Table 5.10 Average Financial literacy Score by type of indebtedness Financial Financial Financial Financial Type of literacy score literacy score literacy score literacy score indebtedness1/ (Total) (Knowledge) (Behaviour) (Attitude) (Out of 22) 2/ (Out of 8) 2/ (Out of 9) 2/ (Out of 5) 2/ Type 1 13.8 4.2 6.3 3.3 Type 2 12.6 3.4 6.0 3.2 Type 3 12.4 3.8 5.4 3.2 Type 4 11.3 3.0 5.2 3.1 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

In addition, after controlling for their financial status, women seem to be more risk-averse than men. Table 5.11 shows that households with female heads have a higher proportion of household type 3 (subjectively over-indebted but not objectively over-indebted). Meanwhile, households with male heads have a higher proportion of household type 2 (objectively over-indebted but not subjectively over-indebted).

118

Table 5.11 Proportion of Indebted Households by Indebtedness Status and Household Head’s Gender

Gender % of type 11/ 2/ % of type 21/ 2/ % of type 31/ 2/ % of type 41/ 2/ Total Male 60.3 13.0 17.9 8.8 100 Female 61.1 9.3 20.5 9.2 100 Note: 1/ Type 1 is a group of indebted households who are not subjectively and objectively over-indebted. Type 2 is a group of indebted households who are objectively over-indebted but not subjectively over-indebted. Type 3 is a group of indebted households who are subjectively over-indebted but not objectively over-indebted. Type 4 is a group of indebted households who are both objectively and subjectively over- indebted. 2/ Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Finally, by social class, non-agricultural households have the highest chance of being subjectively over-indebted, while agricultural households have the highest chance of being objectively over-indebted. Over the past 10 years, agricultural households have been the main targets of the government’s debt-relief program (Table 5.12). For example, in 2001 and 2012, the government implemented a debt-relief program for Thai farmers. For the farmers who had outstanding loans less than THB 100 000 (2001) and THB 500 000 (2012), they were allowed to suspend repayments for three years or to receive a three-percentage-point reduction in the interest rate for three years (International Monetary Fund, 2002; Ministry of Finance, 2012). However, the policies were not applied consistently by the different politic parties in government, which may explain why the co-efficient on government help for credit default is not significant for both subjective and objective over-indebtedness.

119

Table 5.12 Households by Social class and Government Debt Relief Households with the expectation of debt relief Social Class Households with debt relief (%) (%) 25.2 24.9 Agricultural (42% of total households with (31.6% of total households with the debt relief) expectation of debt relief) 9.5 17.3 Non-agricultural (13.3% of total households with (18.3% of total households with the debt relief) expectation of debt relief) 8.5 13.6 Professional (6.8% of total households with (8.2% of total households with the debt relief) expectation of debt relief) 9.1 13.5 Worker (22.5% of total households with (25.4% of total households with the debt relief) expectation of debt relief) 9.3 13.0 Retired (15.6% of total households with (16.5% of total households with the debt relief) expectation of debt relief) Note: Using population weights Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

5.5 Conclusion

To conclude, this chapter corroborates measures of subjective and objective indicators of household over-indebtedness, and the factors underscoring the relationship between these two metrics. On definitions, households with a concern about the next debt repayment are defined as being subjectively over-indebted, while those with income after debt service payment less than the poverty line are defined as being objectively over-indebted. This study has drawn on data collected through the Household Socio-Economic Surveys and the Bank of Thailand’s supplement household survey of the first quarter of 2013. This analysis, using logit regressions, has shown both positive and negative correlations between subjective and objective over- indebtedness within subsets of the data.

For the differences between these two types of over-indebtedness, marginal effects from logit estimations are compared factor by factor. There are four categories of these differences: (i) an effect on subjective over-indebtedness is similar to the effect on objective over-indebtedness; (ii) an effect on subjective over-indebtedness is more than the effect on objective over-indebtedness; (iii) an effect on objective over-indebtedness is more than the effect on subjective over-indebtedness; and (iv) there is an effect only on subjective over-indebtedness or objective over-indebtedness.

120

First of all, having a secure income (regular income), similar age, and personal consumption loans have similar sizes of effects on both the chance of becoming subjectively over-indebted and objectively over-indebted. Moreover, doing income-and- expenditure accounts leads to a lower probability of being subjectively over-indebted and objectively over-indebted at the same size. Therefore, households who maintain income-and-expenditure accounts and those who have a secure source of income have consistent assessments of their exposure to debt. This point speak to there being rational thought in the neoclassical way (the Permanent Income Hypothesis and the Life-Cycle Hypothesis). Rational decision-makers do their accounts and manage their income flows, and hence they are unlikely to see a divergence between their subjective evaluation of their financial circumstances and the objective reality.

Secondly, having problems with rent or utility bills, having insufficient savings (i.e. in case of income shocks, savings can be used to finance expenditure for less than three months), credit constraints, and mortgage loans have larger effects on the probability of being subjectively over-indebted than on the probability of being objectively over-indebted. At the same time, the expectation of a deterioration in the household’s economic status causes households to have more concerns about their next debt repayment, whereas it has a less impact on households’ objective view of indebtedness; that is, expectations count in subjective assessments of over-indebtedness. This finding supports Keese (2010)’s study of the role of the household’s economic expectation on household’s debt status. The results of his analysis suggest that, if indebted households expect worse economic conditions (e.g. higher unemployment in the future), they tend to become concerned about their debt repayments.

Thirdly, a higher level of expenditure-to-income ratio leads to a higher chance to be objectively over-indebted than to be subjectively over-indebted. This is, perhaps, an unsurprising result - those spending a greater amount of their income on consumption are likely more complacent about the consequences of their doing so. That there are such complacent households speaks to the existence of ‘behavioural’ agency on the part of those households. This is to say, some households’ behaviour is likely best characterised as ‘behavioural’ rather than strictly ‘rational’.

Another category of borrowing which is more strongly correlated with worse objective debt performance than it is with subjective concerns about debt repayments is the category of business loans, especially in terms of agricultural business loans. This 121 result suggests that businesses (particularly agricultural businesses) are more confident than the objective data suggest ought to be the case. This is possibly a result of the inherent optimism that is needed to run a business. As John Maynard Keynes’s stated in the General Theory:

“Even apart from the instability due to speculation, there is the instability due to the characteristic of human nature that a large proportion of our positive activities depend on spontaneous optimism rather than mathematical expectations, whether moral or hedonistic or economic. Most, probably, of our decisions to do something positive, the full consequences of which will be drawn out over many days to come, can only be taken as the result of animal spirits—a spontaneous urge to action rather than inaction, and not as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities.” [Keynes, J.M. (1936). The General Theory of Employment, Interest, and Money London: Macmillan. Pp. 161-162]

According to Keynes (1936), businesses were more confident than the data suggested they ought to be about their likelihood of success, and this showed up in the data for Thailand’s business-running households. As in the other situations examined in this study, this can only really be exposed by the technique employed here, which compares objective and subjective measures of indebtedness, and validates the choice of technique. Whether, in the end, the divergence in debt measures for business households is a problem, will turn on how much it is believed that economic activity depends on the (unrealistic) optimism of such households and how costly it is to have businesses act on motives that are not purely data-driven. The more it is believed that optimism is a key ingredient to entrepreneurial activity, and the lower the damage of unrealised business expectations is estimated to be, the greater will be the tolerance of the expectational errors of business households.

Fourthly, with regards to the factors that have effects on subjective over- indebtedness or objective over-indebtedness only, women tend to be more conservative in taking debt compared to men even when there is no difference in their objective measures of debt status. This finding aligns with the previous research by Jianakoplos and Bernasek (1998), and Lenton and Mosley (2008), and Keese (2010). They also found that women are more risk-averse than men when seeking loans.

Contrarily, financial literacy and hyperbolic discounting behaviour only have impacts on objective over-indebtedness. The result about financial literacy affirms the

122 previous findings (Lusardi & Tufano, 2009; Disney & Gathergood, 2012) that claimed households with poor financial literacy had the higher cost of borrowing and then had a higher chance of being over-indebted. The result of hyperbolic discounting behaviour was also aligned with the literature (Gathergood, 2012), which pointed out the positive relationship between self-control problems (hyperbolic discounters and impulsive spenders) and over-indebtedness. In Thailand, people may be objectively over-indebted since they do not realise the real cost of borrowing. Thai households with low financial literacy and poor financial management do not seem to perceive their illiteracy and its impacts. This finding raises the importance of the comparison of subjective and objective over-indebtedness.

Additionally, although an effect of the dependency ratio on objective over- indebtedness is statistically significant, the effect on subjective over-indebtedness is not. This study’s results imply that households underestimate the cost of children and elderly dependants, which shows up as low subjective assessments of household over- indebtedness but as high objective assessments of household over-indebtedness. Access to financial literacy could rectify the misalignment between the subjective and objective assessments of household debt.

In addition, there is one factor that has an opposite effect on subjective and objective over-indebtedness. Education loans make households have a higher probability of being subjectively over-indebted, while this type of loans causes households to have a better objective debt status. In particular, the quantitative evidence shows that subjectively over-indebted poor households largely engaged in agriculture and residing in the North-east are highly conservative with regards to taking out loans for education of their young. The conservative stance could short-change the individual who has missed out on acquiring education, the household who may lose future (and permanent) income as a result, and the society at large through reduced output (i.e. GDP). Policies to address this particular concern are canvassed in Chapter 7.

In the next chapter, a series of Thailand’s official household surveys will be employed, and pseudo-panel data will be constructed for the dynamic analysis of household indebtedness.

123

Chapter 6: Dynamic Analysis of Household Indebtedness: Evidence from Thailand

6.1 Introduction

As explained in Chapter 2, the level of household debt in both developed and developing countries has grown rapidly over the past three decades. Scholars such as Debelle (2004), Barba and Pivetti (2009), and Davies (2009) provided reasons for the rapid rise in household debt, which included record low interest rates and abundant liquidity in global financial markets. The high level of household debt has become one of the most concerning issues in the financial system and macro-economies. Many central banks around the world have expressed their concerns regarding the potential negative impacts of the high household debt on the health of the financial sector and the performance of the economy (Bank of Canada, 2017; Bank of England, 2017; Bank of Thailand, 2017a; Reserve Bank of Australia, 2017; South African Reserve Bank, 2017).

Regarding the risks to the financial system, high household debt can lead to the deterioration in financial institutions’ loan quality, which in turn runs the risk of seeding a credit crunch. Along with slow economic development, such as low-income growth and a low inflation rate, high level of household debt is a source of vulnerability in the financial system, especially in credit markets. The sharp increase in household debt over the period makes the authorities concerned about household debt service ability. They argued that the high level of both debt and debt service burden might lead to a higher default rate. In other words, households with heavy debt service burdens were more likely to have a poor ability to service debts and consequently were more likely to default. To identify indebted households with heavy debt service burden, some researchers refer to households with a ratio of monthly debt service payments to monthly income of more than 30% (D’Alessio & Iezzi, 2013).

Turning to the risks to the macroeconomy, some authorities and researchers (Mian & Sufi, 2011; IMF, 2012; Muthitacharoen et al., 2015; Jorda et al., 2016) believed that a high level of household debt could be another source of macroeconomic vulnerability. Although substantial growth in household debt can boost economic growth in the short term, this kind of growth is debt-driven growth, which is not sustainable. A high level of household debt together with income shocks can result in household financial vulnerability, which causes some households to sacrifice their 124 consumption. This circumstance will lead to a drag in aggregate consumption. Moreover, when a country unexpectedly experiences extremely negative shocks, the country with a high level of household debt will face a prolonged recession.

This chapter will explain the development of household debt in Thailand using household-level data from Thailand’s official household surveys. These surveys have been conducted biennially from 2009 to 2015, and they have been used here to construct panel data to allow for the identification of the determinants of household debt, after controlling for household characteristics.

The analysis in this chapter builds on that of the last chapter by combining data from four separate surveys into one panel. This is done to allow for differences in the debt responses of individual households over time and to eliminate the issue of heterogeneity of the surveyed households. Ideally, the differences in responses of individual households can be purged with econometric analysis if data on the same households are collected over time. Unfortunately, this is not true of the household surveys used here. As the next best option, pseudo-panels are constructed by linking households with common attributes over time using repeated cross-sectional surveys. While pseudo-panels do not alleviate all of the problems of individual heterogeneity, they are better than the analysis of a single cross-sectional survey as undertaken in the last chapter. The aim of this chapter is to test for the robustness of the findings from the last chapter on the drivers of debt when problems of individual heterogeneity are mitigated, albeit not fully.

Based on literature, this research has found that five factors are the significant determinants of household debt status over time in term of household debt performance (ratio of income after debt service payment to poverty line). These are:

(i) Balance between monthly expenditure and monthly income as a proxy consumption behaviour,

(ii) The dependency ratio,

(iii) The source of credit as a proxy of credit access,

(iv) Home ownership, and

(v) Education level.

125

Higher expenditure-to-income ratio and having more dependent members (i.e. higher dependency ratio) are correlated with poorer debt performance over time. Moreover, people, who do not have access to the formal credit market, seem to have persistent weaker debt performance than people who have the access. Additionally, homeownership and a higher level of education can improve household debt status over time because of debt collateral and higher financial literacy, respectively.

There are six sections in this chapter. The first section is the introduction. Section 2 is the literature review on the dynamic analysis of household indebtedness and the arguments for pseudo-panel analysis using annual cross-sectional household surveys. Data description and descriptive statistics on the development of household financial status in Thailand are presented in the third section. In the fourth section, the quantitative methods used for the dynamic analysis of household indebtedness are explained. The empirical results of the study are interpreted and discussed in Section 5 and 6, respectively.

6.2 Literature review

The literature review is divided into two parts. The first part reviews the existing studies related to the dynamic analysis of household indebtedness. Many researchers conduct this type of analysis to explain the development of household debt over selected time periods (Vante, 2006; Anderloni & Vandone, 2008; Keese, 2009; Xiao & Yao, 2014). The second part explores pseudo-panel studies using household surveys. The subsection explains how researchers construct pseudo-panel data, including the method used in this chapter.

6.2.1 Dynamic analysis of household indebtedness

Based on the extant studies on the drivers of household debt, the analysis of the development in household debt can be done using data sets from either genuine-panel surveys or repeated cross-sectional surveys. On the one hand, with genuine-panel surveys, panel regressions can be employed to find the relationships between household indebtedness and explanatory variables. For example, Keese (2009) uses data from the German Socio-Economic Panel Surveys (SOEP: 2002-2007) to find the determinants of severe household indebtedness in Germany. Keese (2009) has found that child-birth and family breakdown can lead to changes in household’s expenditure and income,

126 respectively, which cause the household to have a higher chance of facing debt difficulties. This finding leads to significant factors such as the number of children in the family and head of household’s marital status that can affect the level of household debt. Moreover, he raises another point about the importance of having secure income and homeownership. When facing negative shocks, households with insecure income have a lower probability of getting more loans than those with secure income and homeownership.

In the absence of there being a natural panel, it is possible to use repeated cross- sectional surveys. There are two ways to do this analysis. First, the average effect of each explanatory variable on household indebtedness within the selected period can be determined as was done in the previous chapter. For example, Xiao and Yao (2014) explained consumer debt delinquency by family-life-cycle categories. To find the average behavioural patterns of consumers at different life-cycle stages, they ran multiple logit regressions (with year effects) using the data from US Surveys of Consumer Finances (SCF: 1992 – 2010). They found that households with more children tended to have a higher level of indebtedness, especially in case of single-adult households. Furthermore, Xiao and Yao (2014) pointed out that individuals at different ages had different levels of indebtedness. They claimed that, within the working-age group, younger households had a higher chance of being delinquent than older households.

The second method for modelling households over time in the absence of a panel is to utilise pseudo-panel data can be created for dynamic analysis. The literature on pseudo- panel analysis using household surveys is presented in the next part. Considering a common set of determinants, which is used in the study of household indebtedness, most researchers mentioned three main groups of theories as explained in Chapter 2. The first one is the Permanent Income Hypothesis and the Life-Cycle Hypothesis employed in both static and dynamic analysis (Vante, 2006; Betti et al., 2007; Anderloni & Vandone, 2008; Keese, 2009; Xiao & Yao, 2014), and common variables are income, age, and dependency level. The second is liquidity constraints and credit rationing mentioned in both static and dynamic analysis (Betti et al., 2007; Anderloni & Vandone, 2008; Keese, 2009). The third is behavioural finance, which is currently used in static analysis only (Lusardi & Tufano, 2009; Disney & Gathergood, 2012).

127

The dynamic analysis of household indebtedness has been reviewed and the common set of variables summarised, which are used to explain household indebtedness. Next, this research will review the previous studies related to pseudo-panel analysis using household surveys.

6.2.2 Pseudo-panel analysis using household surveys in the literature

This thesis now turns to the literature on pseudo-panel analysis using household surveys. In many countries, there are commonly a lack of genuine-panel household surveys, where the same set of households are tracked over time (e.g. the US Current Population Surveys and the UK’s Family Expenditure Surveys). Thailand household surveys are not an exception with regards to the above. This data gap leads to the challenging issues of doing panel analysis.

There are at least two benefits of using panel data (Wooldridge, 2010). Firstly, one purpose of doing panel analysis is finding dynamic relationships between an endogenous variable and a set of explanatory variables where cross-section analysis does not allow us to find such relationships. Secondly, unobserved cross-section heterogeneity can be controlled by using panel data.

To address the above-noted deficiency, some researchers have constructed pseudo-panel data from repeated cross sections (Deaton, 1985; Moffitt, 1993; Verbeek & Nijman, 1992). This is only possible when repeated household surveys are available. This type of survey is collected from a random household sample, which is taken from the population at two or more points in time. Such surveys allow the researcher to create household cohorts, which are household groups with the same characteristics. Deaton (1985) was the first to employ pseudo-panel analysis in consumer economics. Many researchers support the idea of constructing pseudo-panel data by creating household cohorts (Verbeek & Nijman, 1992; Moffit, 1993; Huang, 2007; Weis & Axhausen, 2009; Bernard et al., 2011; Tsai et al., 2014). Focusing on the set of household’s characteristics used for grouping households, researchers have commonly used a set of time-invariant variables (Deaton, 1985; Moffitt, 1993; Huang, 2007; Weis & Axhausen, 2009; Bernard et al., 2011). The most common characteristic for grouping households is the birth year (Deaton, 1985; Huang, 2007; Weis & Axhausen, 2009). The additional set of criteria for creating household cohorts can be used depending on the purpose of the study. For example, Weis and Axhausen (2009) use the year of birth, gender, and region

128 to differentiate travel demand of each household in different areas. Bernard, Bolduc, and Yameogo (2011) group households by household size and region to capture different electricity of consumption for households of various sizes residing in different locations.

Pseudo-panels have their own problems. When researchers construct a pseudo- panel data, they usually mention the issue of errors in variables: that is, when sample- cohort means are different from population-cohort means. The implication of the errors in variables is the endogeneity problem, which leads to biased estimates if standard regressions (i.e. ordinary least square regressions) are used (Verbeek & Nijman, 1992). To avoid this issue, some empirical studies show that the number of households in each cohort needs to be large (Browning et al., 1985; Blundell et al., 1989; Verbeek & Nijman, 1992). Verbeek and Nijman (1992) suggested that grouping households with the same characteristics needed a large enough number of observations, which is at least 100 observations per cohorts as the threshold, to get rid of the errors-in-variables problem. Moreover, many scholars accept this threshold as the minimum (Dargay & Vythoulkas, 1999; Tsai et al., 2014).

Additionally, according to the literature (Blundell et al., 1989; Verbeek & Nijman, 1992; Verbeek, 2008), there are two more conditions of creating usable household cohorts. One condition is the time variation in cohort means. It is necessary to make sure that the cohort’s mean of each variable in the models changes differentially over time (Verbeek & Nijman, 1992; Verbeek, 2008). Another condition is the variation between cohorts. Blundell, Browning, and Meghir (1989) suggested the use of instrumental variables, which were required to be uncorrelated to the unobserved variables, but relevant to the explanatory variables in the models. Verbeek and Nijman (1992) and Verbeek (2008) accepted and supported the use of instrumental variables for grouping households into cohorts (to create the variation between cohorts).

Taking into account the above three conditions (i.e. cohort size, time variation, and cohort variation), identical groups of households emerged that could be traced over time. After creating household cohorts, pseudo-panel analysis could be done using the sample cohort means, and each cohort treated as an individual observation in the data set. Then the pseudo-panel data could be treated as genuine-panel data.

129

Regarding the limitations and benefits of using this type of panel data (pseudo- panel data), on the one hand, using such panel data has some major limitations (Verbeek, 2008). First, this type of data does not allow us to transform models (e.g. first-differences and deviations from individual means) because there are no individual time-series data. Second, it is challenging to construct the robust instrumental variables for grouping households. Alternatively, pseudo-panel data constructed from repeated cross-section surveys have some advantages (Gardes et al., 2005; Verbeek, 2008). Firstly, repeated cross-section surveys usually have a higher response rate than genuine- panel surveys. This advantage leads to less non-response attribution bias resulted in less attrition when screening data. Furthermore, in some cases, repeated cross-section surveys have longer timespans than genuine-panels.

6.3 Data and development of household debt in Thailand from 2009 to 2015

6.3.1 Data

The data from the official household surveys from 2009 to 2015 is described next. Household Socio-Economic Survey (SES) is conducted by the National Statistic Office. The complete set of SES is collected every two years, and 52 000 households are selected as being representatives of the total household population of Thailand (comprising 77 provinces as of 2013). However, the actual numbers of representative households in 2009, 2011, 2013, and 2015 are 43 844, 42 083, 42 738, and 43 400, respectively. According to the National Statistic Office’s data manual, the households for the survey are selected through stratified random sampling where the stratums are based on the district in all provinces and household’s characteristics.

In the light of these biennial surveys, households can be identified by the following criteria:

130

Flows:

 Inflows: monthly income, including the types of income (wage, salary, profit from business, pension, transferred income)  Outflows: monthly expenditure, including the types of expenditure (food, beverages, tobacco, other consumption, non-consumption, and monthly debt service payments.

Stocks:

 Physical assets: dwelling, vehicles  Debt, including the types of loans (classified as mortgage, agricultural, business, educational, personal consumption).  This research uses the regional poverty line from the National Statistical Office. The National Economic and Social Development Board calculates this poverty line, and it is used to determine the cost to the individual of the acquisition of food and basic services essential to life. More details about the calculations in are given in Chapter 4. The poverty line for five regions in Thailand (Bangkok, Central exclude Bangkok, North, North-east, and South) are shown in Table 6.1. It can be seen from Table 6.1 that people in the north and the North-east of Thailand have the lowest poverty line over the past 10 years, whereas people in Bangkok (the capital city) have the highest poverty line.

Table 6.1 Poverty Line (THB per Person per Month)

Region 2007 2008 2009 2010 2011 2012 2013 2014 2015

Bangkok (BKK) 2565 2694 2676 2756 2901 2994 3047 3133 3132 Central 2220 2390 2382 2490 2610 2696 2775 2832 2827 (excl. BKK) North 1782 1936 1938 2040 2160 2226 2314 2387 2377

North-east 1717 1882 1883 2005 2130 2188 2273 2355 2355

South 2042 2219 2239 2344 2492 2577 2651 2735 2724

131

6.3.2 Overview descriptive statistics

This section will provide some statistics on the development of household debt during 2009 to 2015. For a clear picture of the change in debt status, it will focus on indebted households only (see Chapter 3 for a discussion on the debt status of all households in the sample). Firstly, this section will look at indebted households’ financial status in Thailand over the period of 2009 to 2015. Secondly, it will focus on household groups with the most precarious financial status by income group, region, and social class.

Development of indebted households’ financial status in Thailand during 2009 to 2015

In this part, the focus is on the changes in indebted households’ financial status in Thailand during 2009 to 2015. Four financial ratios: (1) monthly expenditure to income ratio (Equation 6.1); (2) debt to annual income ratio (Equation 6.2); (3) monthly debt service payments to income ratio (monthly debt service ratio: Equation 6.3); (4) monthly income after debt service payments to household’s poverty line (debt performance: Equation 6.4) are calculated as household’s financial soundness indicators.

퐸 EIR = (6.1) 퐼 where 퐸퐼푅 = Expenditure-to-income ratio

퐸 = Monthly expenditure

퐼 = Monthly income

퐷 DYR = (6.2) 푌 where 퐷푌푅 = Debt-to-annual-income ratio

퐷 = Stock of debt

푌 = Annual income

132

퐷푆푃 DSR = (6.3) 퐼 where 퐷푆푅 = Debt service ratio

퐷푆푃 = Monthly debt service payments

퐼 = Monthly income

퐼 − 퐷푆푃 DP = (6.4) 퐻푃퐿 where 퐷푃 = Debt performance

퐼 = Monthly income

퐷푆푃 = Monthly debt service payments

퐻푃퐿 = Household’s poverty line

Table 6.2 and Figure 6.1 represent indebted households’ average financial status from 2009 to 2015. They show that, over this period, indebted households had a high share of consumption to total income, which reflects a low ability to save. In addition, this group of households had a higher level of indebtedness reflected from rising debt- to-annual-income ratio.

First of all, regarding the balance between monthly expenditure and monthly income among indebted households, from 2009 to 2015, Table 6.2 and Figure 6.1 show that indebted households’ average expenditure-to-income ratio remained in a range of 0.94 to 0.98; these ratios suggest that savings on average are between 2-6% of income. The high share of consumption in income leaves little room for savings, and thus limits the capacity of households to take on debt. Furthermore, the sample of households with debt has a savings rate lower than that for the nation as a whole. The National Income Accounts, as shown by Figure 6.2, reveals that the average saving-to-disposable-income ratio for the period of 2001 to 2015 was 8.37%, and this average value had dropped from 14.32% for the period of 1990 to 2000.

Secondly, from Table 6.2, indebted households’ average debt-to-income ratio increased from 0.77 times in 2011 to 0.82 and 0.83 times in 2013 and 2015, respectively as a result of the greater growth in debt than income. Although there was an increase in household debt-to-income ratio, average debt service ratio stayed stable at around 0.3 times. Moreover, average debt performance was improving from 2009 to 2015 because 133 household income grew much faster than the household poverty line (Table 6.2 and Figure 6.1).

Table 6.2 Overview Statistics of Indebted Households’ Financial Status

Financial ratio (times)

2009 2011 2013 2015

Monthly expenditure to income ratio 0.97 0.98 0.94 0.96

Debt to annual income ratio 0.79 0.77 0.82 0.83

Monthly debt service ratio 0.28 0.28 0.27 0.29

Debt performance 2.57 2.72 2.99 3.08 Note: Using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figure 6.1 Indebted Households’ Income, Expenditure, and Debt for the Period of 2009 to 2015

Index Monthly expenditure to income (RHS) Debt to annual income (RHS) Monthly income Monthly expenditure (2009 = 100) Debt Monthly debt service payment Times Household's poverty line 150 0.98 1.0 0.97 0.96 140 0.94

130 0.9 120

0.83 110 0.82

0.79 0.8 100 0.77

80 0.7 2009 2011 2013 2015 Note: Using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

134

Figure 6.2 Household Saving-to-Disposable-Income Ratio from National Income Accounts

% 20

15 Average (1990 – 2000): 14.32%

Average (2001 – 2015): 8.37%

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Source: Office of the National Economic and Social Development Board. Author’s calculations.

The next part will identify households who have the poorest financial status by income group, region, and social class, respectively.

Households with the most unsound financial status

To identify households with the unsound financial status, this study focuses on the level of the debt service ratio (which reflected how heavy debt service burden is (equation 6.3)), debt performance (which reflected the distance between income after debt service payment and poverty line (equation 6.4)), and the expenditure-to-income ratio (which reflected the balance between monthly income and monthly expenditure (equation 6.1)). This research defines households with the most unsound financial status as households with the greatest debt service ratio, the lowest debt performance, and the highest expenditure-to-income ratio. Furthermore, for grouping households, three main categories, namely income group (income deciles from the lowest to the highest), region, and social class, will be considered respectively.

First of all, with regards to household financial status by income group for the period 2009 to 2015, the bottom 30% of households by income (group 1 – 3) have had the weakest debt status in terms of debt service ratio and debt performance. Table 6.3 shows that these households have heavy debt service burden, which is more than 30% of their income. According to D’Alessio and Iezzi (2013), households with a debt 135 service ratio of more than 30% and households with income after debt service payment less than poverty line are more likely to experience debt difficulties of repaying their debt.

In addition, households in the bottom 30% by income have poorer debt performance than households with higher income over the period (Table 6.3). The poorer debt performance means that these households have a shorter distance between their income after debt service payments and their subsistence levels of income (poverty line). Among household group 1 and 2, their average debt performance is less than one, which means their income after debt service payment cannot cover the minimum cost of living. Moreover, these low-income households have a monthly expenditure-to-income ratio greater than one, so they have no chance to save from their monthly income (Table 6.3).

Focusing on the debt status of rich households, although the top 10% richest households have a similar level of debt-to-income ratio as compared to the bottom 10% of the poorest households, they have a lower level of debt service ratio and have better debt performance (Table 6.3). According to the statistics of the composition of debt by income group in Chapter 4, more than 80% of the loans of the poorest households’ (i.e. the bottom decile by income distribution) are in the form of personal consumption loans and agricultural loans (both of which are unsecured loans), while around one-third of debt of households in the top decile by income distribution is in the form of mortgages on property (secured loans).

136

Table 6.3 Household Financial Status by Income Group

Monthly expenditure to income ratio (times)

2009 2011 2013 2015

Group 1 (Poorest) 1.83 2.10 1.98 1.93

Group 2 1.28 1.29 1.22 1.23

Group 3 1.10 1.13 1.09 1.08

Group 4 1.01 1.03 0.98 1.01

Group 5 0.98 0.98 0.94 0.95

Group 6 0.92 0.91 0.90 0.92

Group 7 0.88 0.87 0.85 0.90

Group 8 0.85 0.84 0.82 0.86

Group 9 0.78 0.77 0.79 0.81

Group 10 (Richest) 0.66 0.64 0.65 0.69

Debt to annual income ratio (times)

2009 2011 2013 2015

Group 1 (Poorest) 1.37 1.25 1.28 1.25

Group 2 0.79 0.80 0.75 0.83

Group 3 0.64 0.63 0.66 0.71

Group 4 0.59 0.58 0.58 0.60

Group 5 0.59 0.60 0.59 0.64

Group 6 0.60 0.62 0.65 0.67

Group 7 0.69 0.65 0.74 0.86

Group 8 0.76 0.81 0.85 0.87

Group 9 0.97 0.89 1.02 0.93

Group 10 (Richest) 1.03 1.00 1.06 1.03

Monthly debt service ratio (times)

2009 2011 2013 2015

Group 1 (Poorest) 0.62 0.59 0.60 0.59

Group 2 0.40 0.38 0.40 0.39

Group 3 0.29 0.31 0.31 0.31

Group 4 0.27 0.28 0.27 0.28

Group 5 0.25 0.26 0.24 0.27

Group 6 0.23 0.25 0.25 0.26

Group 7 0.23 0.23 0.23 0.26

137

Group 8 0.24 0.24 0.22 0.25

Group 9 0.25 0.23 0.23 0.24

Group 10 (Richest) 0.22 0.24 0.22 0.24

Debt performance (times)

2009 2011 2013 2015

Group 1 (Poorest) 0.41 0.52 0.48 0.64

Group 2 0.71 0.87 0.86 1.02

Group 3 1.02 1.06 1.20 1.31

Group 4 1.25 1.27 1.43 1.64

Group 5 1.46 1.53 1.69 1.85

Group 6 1.84 1.81 2.07 2.15

Group 7 2.17 2.23 2.46 2.63

Group 8 2.71 2.72 3.02 3.15

Group 9 3.64 3.75 3.95 4.12

Group 10 (Richest) 7.94 8.55 8.54 8.18 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figures 6.3-6.5 display the trend of financial indices from 2009 to 2015 for the bottom 30% of households by income (group 1-3) in terms of income, debt outstanding, and debt service payment. The indices are calculated from the level of each financial variable and are set as 100 for the year of 2009. Figures 6.3-6.5 show that household group 1-3 have a sharp increase in their debt level compared to an increase in their income, especially among groups 2 and 3. Focusing on household groups 2 and 3, their debt outstanding indices increased from 100 in 2009 to around 150 and 160 in 2015, which are equivalent to a growth rate of 8.3% per annum and a growth rate of 10% per annum respectively (Figure 6.4). These growth rates are higher than their average income growth at 7.5% per annum (Figure 6.3 shows that income indices increased from 100 in 2009 to around 145 in 2015 for both household groups).

138

Figure 6.3 Household Income by Income Group

Index (2009 = 100) Group 1 (Poorest) Group 2 Group 3 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figure 6.4 Household Debt by Income Group

Index (2009 = 100) Group 1 (Poorest) Group 2 Group 3 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

139

Figure 6.5 Household Debt Service Payment by Income Group

Index (2009 = 100) Group 1 (Poorest) Group 2 Group 3 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Second, focusing on the household financial status by region, the distribution of household debt also is distinct by region. In particular, during 2009 to 2015, households in the North-east have had the weakest debt status in terms of debt service ratio and debt performance, whereas households in Bangkok have had the strongest debt status (Table 6.4). Table 6.4 shows that, over the period, the average debt service ratio for households in the North-east is greater than 30%, while the average value for households in Bangkok is around 20%. Moreover, for people in the North-east, we can see from Table 6.4 that the high debt service ratio together with a high expenditure-to-income ratio (close to one) reflects their limited capacities to save.

140

Table 6.4 Household Financial Status by Region

Monthly expenditure to income ratio (times)

2009 2011 2013 2015

Bangkok (BKK) 1.01 0.89 0.93 0.85

Central (excl. BKK) 0.96 1.01 0.92 0.97

North 0.92 0.94 0.89 0.93

North-east 1.00 1.02 0.99 0.99

South 0.99 0.96 0.94 1.03

Debt to annual income ratio (times)

2009 2011 2013 2015

Bangkok (BKK) 0.93 0.92 1.00 0.78

Central (excl. BKK) 0.77 0.79 0.89 0.86

North 0.84 0.81 0.77 0.90

North-east 0.76 0.76 0.78 0.75

South 0.70 0.62 0.77 0.95

Monthly debt service ratio (times)

2009 2011 2013 2015

Bangkok (BKK) 0.21 0.21 0.21 0.19

Central (excl. BKK) 0.25 0.26 0.23 0.27

North 0.32 0.32 0.27 0.30

North-east 0.31 0.31 0.33 0.32

South 0.22 0.21 0.22 0.25

Debt performance (times)

2009 2011 2013 2015

Bangkok (BKK) 4.14 4.85 5.29 5.20

Central (excl. BKK) 2.80 2.60 3.41 3.22

North 2.38 2.59 2.77 2.64

North-east 2.04 2.20 2.23 2.49

South 3.29 3.76 3.81 3.52 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

141

Consider, next, the trend of financial variables. Figures 6.6-6.8 show that although people in the North-east had an increase in their income over the period, they had a greater increase in their debt level and especially debt service payments. Their debt service expenses index increased from 100 in 2009 to around 155 in 2015, which is equivalent to a growth rate of 9.2% per annum (Figure 6.8). This growth rate is high compared to their average income growth at 6.7% per annum (Figure 6 shows that the income index increased from 100 in 2009 to around 140 in 2015). Additionally, there is a low proportion of households with the perception of having secure income for households in the North-east compared with households in other regions (Table 6.5). From this evidence, it can be seen that households in the North-east have the persistently poorest debt status.

Figure 6.6 Household Income by Region

Index Bangkok (BKK) Central (excl. BKK) North Northeast South (2009 = 100) 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

142

Figure 6.7 Household Debt by Region

Index Bangkok (BKK) Central (excl. BKK) North Northeast South (2009 = 100) 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figure 6.8 Household Debt Service Payment by Region

Index Bangkok (BKK) Central (excl. BKK) North Northeast South (2009 = 100) 160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

In addition, focusing on the period 2013 to 2015, households in the South of Thailand have a poorer financial status reflected by an increase in debt-to-income ratio, and debt service ratio (Table 6.4). In that period, people in the South faced a decline in

143 their income, but they had an increase in debt and debt service payment (Figure 6.6 – 6.8).

This is partly from a drop in the price of rubber, which is the main source of income for households in the South of Thailand (Bank of Thailand, 2015; Bank of Thailand 2016c). Figure 6.9 shows that the rubber price declined significantly from 2011 to 2015. Moreover, focus on the balance between income and expenditure, the drop in rubber prices caused people in the South to lose their ability to save reflected by the average expenditure-to-income ratio that was close to one (Table 6.4).

Figure 6.9 Rubber Price

Baht per Kg 200

150 Rubber price

100

Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17

Sep-13 Sep-14 Sep-09 Sep-10 Sep-11 Sep-12 Sep-15 Sep-16 Sep-17

May-09 May-10 May-11 May-12 May-13 May-14 May-15 May-16 May-17

Note: Rubber (Asia), RSS3 grade, Singapore Commodity Exchange Ltd (SICOM) Source: Singapore Commodity Exchange Ltd (SICOM)

Table 6.5 Proportion of Households with the Perception of Having Secure Income by Region

Bangkok Central North North-east South (BKK) (Excl. BKK)

All households 64.2 62.9 51.5 45.2 62.2

Indebted households 68.2% 68.5% 55.0% 49.1% 70.0% Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Third, regarding household financial status by social class, the level of household indebtedness also differs by sector. During the period of 2009 to 2015, Table 144

6.6 shows that agricultural households and retirees have had the weakest debt status in terms of debt service ratio and debt performance, whereas professionals have the strongest debt status.

For agricultural households, they normally have fluctuating income, which depends on the weather. They are usually the main targets of government help, as mentioned in Chapter 5. In the past, the government provided debt relief to selected agricultural households. There was a debt-relief program for Thai farmers, which was implemented in 2012 by the government. Farmers with outstanding loans of less than THB 500 000 were given two options of debt relief: (1) debt repayment suspension for three years; (2) three-percentage-point reduction in the interest rate for three years (Ministry of Finance, 2012). This debt-relief program explains why agricultural households’ debt status in terms of debt service burden and debt performance improved in 2013 and then got worse, significantly, in 2015 (Table 6.6).

Focusing on retired households, from 2009 to 2015, the higher indebtedness among this group of households came along with a higher debt service ratio, which can lead to a concern about the readiness of becoming an ageing society in the future. Retired households tended to have increasing indebtedness reflected by a higher debt- to-income ratio (Table 6.6). Although retirees had a smooth increasing trend of income during 2011 to 2015 (their income is partly from pension funds), their debt and debt service payment jumped up significantly (and Figures 6.10- 6.12). Their debt outstanding index increased from 100 in 2009 to around 180 in 2015, which is equivalent to a growth rate of 13.3% per annum (Figure 6.11). This growth rate is high compared to their average income growth at 8.3% per annum (Figure 6.10 shows that income index increased from 100 in 2009 to around 150 in 2015). However, it is not possible to ascertain why Thai retirees had increasing indebtedness over the period.

In addition, both agricultural households and retirees have had the highest expenditure-to-income ratio, which reflects the poorest ability to save (Table 6.6). Furthermore, it seems that agricultural households and retirees have a low proportion of households with the perception of having secure income compared with households in other social classes (Table 6.7). From these statistics, it can be seen that agricultural households and retirees have the persistently poorest debt status with low ability to accumulate their wealth.

145

Table 6.6 Household Financial Status by Social Class

Monthly expenditure to income ratio (times)

2009 2011 2013 2015

Agricultural 1.05 1.03 0.95 1.02

Non-agricultural 0.94 0.97 0.94 0.97

Professional 0.81 0.83 0.82 0.82

Worker 0.98 0.95 0.93 0.93

Retired 1.00 1.11 1.08 1.08

Debt to annual income ratio (times)

2009 2011 2013 2015

Agricultural 0.80 0.72 0.73 0.84

Non-agricultural 0.91 0.84 0.91 1.01

Professional 1.13 1.22 1.35 1.11

Worker 0.63 0.62 0.66 0.65

Retired 0.63 0.75 0.79 0.78

Monthly debt service ratio (times)

2009 2011 2013 2015

Agricultural 0.37 0.35 0.32 0.37

Non-agricultural 0.28 0.28 0.28 0.30

Professional 0.26 0.27 0.27 0.25

Worker 0.21 0.20 0.21 0.21

Retired 0.28 0.32 0.32 0.33

Debt performance (times)

2009 2011 2013 2015

Agricultural 1.66 2.02 2.13 1.95

Non-agricultural 3.07 3.40 3.70 3.29

Professional 5.62 5.75 6.15 6.18

Worker 2.11 2.26 2.53 2.73

Retired 2.14 1.86 2.30 2.66 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

146

Figure 6.10 Household Income by Social Class

Index Agricultural Non-agricultural Professional Worker Retired (2009 = 100)

180

160

140

120

100

80 2009 2011 2013 2015

Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figure 6.11 Household Debt by Social Class

Index Agricultural Non-agricultural Professional Worker Retired (2009 = 100)

180

160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

147

Figure 6.12 Household Debt Service Payment by Social Class

Index Agricultural Non-agricultural Professional Worker Retired (2009 = 100)

180

160

140

120

100

80 2009 2011 2013 2015 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Table 6.7 Proportion of Households with the Perception of Having Secure Income by Region

Non- Agricultural Professional Worker Retired agricultural

All households 52.6 57.9 80.3 56.4 40.8

Indebted households 52.9 59.7 79.5 59.6 45.4 Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

In summary, regarding different household groups by income group, region, and social class, this study can highlight household groups with the most unsound financial status in terms of debt service ratio, debt performance, and expenditure-to-income ratio. In the case of Thailand, low-income households, households in the North-east, agricultural households, and retirees have poorer debt status and lower ability to save than other household groups. These groups of households have a greater increase in debt outstanding and debt service burden than the increase in their income. Moreover, some of them (households in the North-east, agricultural households) do not have secure income.

148

In the next section, the determinants of household debt performance will be analysed, using quantitative methods. The methodology and the list of variables for the analysis are explained first. Then the empirical results presented and discussed.

6.4 Methodology

In this chapter, for household indebtedness measurement, this study applies the proxy of household debt performance used by Keese (2009). It is the ratio of income after debt service payments to household’s minimum subsistence income level, and this research also uses the poverty line as the minimum subsistence level of income (Equation 6.4). Keese (2009) claimed that this ratio reflected the distance between income after debt service payments and the subsistence level of income. In addition, this ratio can also be used as a household’s gauge of debt sustainability.

To explain dynamic changes in household debt performance, first, household cohorts will be created using the household head’s characteristics (birth year and gender) and region. Tables 6.8–6.9 show the details of the household head’s characteristics and region, respectively.

Table 6.8 Household Head's Characteristic Groups

0. After 1989 (<= 19 in 2009)

1. 1980 – 1989 (20 – 29 in 2009)

2. 1970 – 1979 (30 – 39 in 2009) Birth year 3. 1960 – 1969 (40 – 49 in 2009) Household head's characteristic 4. 1950 – 1959 (50 – 59 in 2009)

5. Before 1950 (>= 60 in 2009)

1. Male Gender 2. Female

Table 6.9 Region Groups

1. Bangkok (BKK)

2. Central (Exclude BKK)

Region 3. North

4. North-east

5. South

149

Second, to finalise the number of cohorts, this study decided to drop households with too young household heads (household heads who were born after 1989) because of an insufficient number of observations. Then, the final number of cohorts could be clarified as follows:

- Five birth-year groups (without group 0) by two genders by five regions imply a total of 50 cohorts.

- Balanced panel with four years per cohort (2009, 2011, 2013, 2015) implies 200 observations in total.

- Dropping observations with the number of HH less than 100 for reliable statistics (38 observations)

- The final number of cohorts for balanced panel data is 37 cohorts with 148 observations.

Thirdly, to find relationships between debt performance and a set of explanatory variables, pseudo-panel data will be set using above cohorts. The data from the pseudo- panel is then used in a multiple linear regression with fixed effects (cohort effects) to investigate the contribution of each of the explanatory variables to the level of household debt.

푌̅푐푡 = 푋̅′푐푡훽 + 훼0 + 훼푐 + 푢푐푡 (Equation 6.5) where 푐 = Household cohorts by household head’s birth year, household head’s gender, and region

푡 = Time period (2009, 2011, 2013, and 2015)

푌̅푐푡 = Average debt performance of cohort c in period t

푋̅′푐푡 = K-dimensional vector of explanatory variables (average values) of cohort c in period t

훽 = Coefficient vector of interest

훼0 = Constant term

훼푐 = Fixed effect of cohort c

푢푐푡 = Error term of cohort c in period t

150

Table 6.10 shows the description of the endogenous variable (Y-variable) and explanatory variables (X-variables).

Table 6.10 List of Variables

Variable Description Measurement

Monthly income after debt service payment to Y-variable Debt performance poverty line ratio C-Y balance Monthly expenditure to income ratio Mean Number of household members with no income, excluding household members who help family Dependency businesses, to total number of household member ratio 1. Commercial banks (benchmark) 2. Bank for Agricultural and Agricultural Cooperatives: BAAC 3. Government Housing Bank and Government Saving Bank: GHB and GSB X-variables Main source of 4. Non-banks credit Proportion 5. Cooperatives (% of HH) 6. Village fund

7. Informal

Education level Obtaining at least a bachelor’s degree

Homeownership Having homeownership (both home and land)

Lastly, the relationship between each explanatory variable and debt performance will be discussed along with its significant level. In addition, another multiple linear regression with random effects will be estimated to examine the robustness of the parameters estimated.

6.5 Empirical results

After the pseudo-panel regressions (one regression with fixed effects, and another one with random effects) regarding the dynamic-debt-performance analysis were completed, Table 6.11 was composed showing the results from these panel regressions. As the Chi-squared value from the Hausman test is not great enough to reject the null hypothesis (the differences in the coefficients are not systematic), this study will discuss empirical findings based on the model with random effects. Firstly, the parameter estimates from the panel regression model with random effects suggest that a higher 151 level of expenditure-to-income ratio causes poorer debt performance (shorter distance between income after debt service payments and the subsistence level of income). Therefore, this research can claim that the indebted hand-to-mouth consumers tend to struggle with both poor debt performance and low ability to save. Secondly, from Table 6.11, there is a negative relationship between the dependency ratio and debt performance. This result may imply that more household members with no income can lead to more financial burdens (e.g. educational expenses, health expenses, and cost of living), and then higher indebtedness and debt service burden.

Thirdly, regarding the role of credit access, the results in Table 6.11 suggest that a higher proportion of households, who have to rely on an informal source of credit, leads to poorer average debt performance. It can be explained by different borrowing interest rates from different sources of credit. Figure 6.13 shows that, during 2009 to 2015, the effective interest rate (calculated from borrower’s annual interest payment and debt outstanding) of the informal source of credit was the highest one at around 35% per annum. Whereas the effective rates of other sources of credit, excluding non-banks, were lower than 10% per annum. Regarding the role of Bank for Agricultural and Agricultural Cooperatives (BAAC) this is a specialised financial institution owned by the government. The main target group of BAAC is agricultural households who have insecure and fluctuating income. Their income from crops depends on the market prices and weather. The ‘village fund’ was launched in 2001 by the Thai government and is managed by the local committee of each village. The government aimed to help households in rural areas who cannot get loans from formal sources (Boonperm et al., 2013). From Figure 6.14, borrower’s debt performance of BAAC and village fund was the lowest one at around 2 times (i.e. the income after debt service payments can cover 2 times of the minimum subsistence income), whereas borrower’s debt performance of commercial banks and cooperative was the highest, at around five to six times (i.e. the income after debt service payments can cover five to six times of the minimum subsistence income).

152

Table 6.11 Debt Performance Models (with Fixed Effects/ Random Effects) With fixed With random Variable effects effects Expenditure to income ratio (mean) -1.682*** -2.1***

Dependency ratio (mean) -3.57*** -4.118***

Main source of credit: BAAC (proportion: % of HH) -4.534*** -7.865***

Main source of credit: GHB and GSB (proportion: % of HH) -0.545 -1.625

Main source of credit: non-banks (proportion: % of HH) -1.526 -4.675***

Main source of credit: cooperatives (proportion: % of HH) -1.013 -3.063***

Main source of credit: village fund (proportion: % of HH) -3.569** -6.735*** Main source of credit: informal source (proportion: % of -3.532** -5.585*** HH) Homeownership ratio (proportion: % of HH) 3.195*** 3.095***

Obtaining at least a bachelor’s degree (proportion: % of HH) 4.855*** 2.918***

Constant 5.507*** 8.993***

Number of cohorts 37 37

Number of observations 148 148

Within R2 0.582 0.558

Between R2 0.819 0.921

Overall R2 0.763 0.840

Hausman Test: Chi2 13.1

Prob. > Chi2 0.218 Notes: *** Significant at the 1% level.

** Significant at the 5% level.

* Significant at the 10% level.

153

Figure 6.13 Average Effective Borrowing Interest Rate by Source of Credit % per annum 50 2009 2011 2013 2015 Average (2009 - 2015)

42.0 40

34.2 32.731.9 30.1 30

12.7 11.611.6 10.7 11.5 10 7.4 7.4 7.5 7.1 7.1 6.3 6.2 6.8 6.4 5.2 5.9 5.7 5.2 5.5 5.4 5.4 5.6 5.6 5.3 5.8 4.6 3.9 4.2 4.6 4.3

0 Commercial banks BAAC GHB and GSB Non-banks Cooperatives Village fund Informal

Notes: Effective borrowing rate = (Annual interest payment/Debt outstanding)x100. Using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Figure 6.14 Average Borrowers’ Debt Performance by Source of Credit Times 8 2009 2011 2013 2015 Average (2009 - 2015)

6 5.7 5.8 5.5 5.3 5.3 5.0 5.0 4.8 4.7 4.7 4.5 4.5 4.5 4.4 4.4 4 3.5 3.2 3.3 3.2 3.0

2.4 2.5 2.3 2.1 2.0 2.0 2.0 2.1 1.8 1.9 1.9 1.9 2 1.7 1.8 1.7

0 Commercial banks BAAC GHB and GSB Non-banks Cooperatives Village fund Informal

Notes: Using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

154

Fourthly, focusing on the importance of homeownership, Table 14 shows a positive relationship between homeownership (including land ownership) and debt performance. To explain this relationship, Table 6.12 shows that households with homeownership in both house and land have a lower proportion of households who used to be rejected by a formal source of credit than households with no homeownership. This means households with homeownership have greater access to formal sources of credit together with lower interest rates.

Table 6.12 Homeownership (House and Land) and Credit Access Households with Households with no

homeownership homeownership Proportion of households who used to be rejected by a formal 6.26 13.37 source of credit (% of households who have applied for loans) Effective borrowing interest rate (%) 29.10 33.79 Note: Using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

Finally, moving on to the impact of education on debt performance, this thesis finds a positive relationship between undergraduate education (or higher) and debt performance. Table 6.13 shows that households (household heads) with at least bachelor’s degrees have better financial status than households with no degrees in terms of ability to save (lower expenditure-to-income ratio) and debt performance (longer distance between income after debt service payments and subsistence level of income). Moreover, this study finds that households with a higher level of education have a higher financial literacy score (Table 6.14). From these findings, it can be stated that higher education (at least bachelor’s degrees) is correlated with a more sound level of both financial management and financial literacy.

155

Table 6.13 Financial Status of Households with Degrees and with No Degrees

Monthly expenditure to income ratio (times)

2009 2011 2013 2015 Households with degrees 0.79 0.82 0.84 0.82 (At least bachelor’s degrees) Households with no degrees 0.99 1.00 0.96 0.98

Debt performance (times)

2009 2011 2013 2015

Households with degrees 6.87 7.23 7.58 7.35

Households with no degrees 2.18 2.27 2.46 2.53 Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2009 - 2015, National Statistical Office. Author’s calculations.

Table 6.14 Average Financial Literacy Scores and Education levels

Education level

Post- Pre/ primary Secondary Bachelor’s secondary/ school school degree Diploma Average Financial literacy score 12.5 13.9 14.9 15.6 (out of 22) Note: Indebted households only, using population weights. Source: Household Socio-economic Surveys of 2013Q1, National Statistical Office. Author’s calculations.

6.6 Conclusion

To conclude, in this study the determinants of the changes in household debt performance from 2009 to 2015 have been presented. The research constructed household cohorts (groups of households) using household head’s birth year, household head’s gender, and region. Additionally, according to the literature and data availability, it has focused on five factors, which are (i) balance between monthly expenditure and monthly income to find the effect of consumption behaviour, (ii) dependency ratio, (iii) source of credit to examine the significance of credit access; (iv) homeownership, and (v) education level.

According to the results from the pseudo-panel analysis using panel regressions with random effects, firstly, it has found that, over time, households with hand-to-mouth

156 consumption behaviour (the expenditure-to-income ratio close to one) have weaker debt performance compared to those who have a lower expenditure-to-income ratio. People with a high expenditure-to-income ratio are people with poor financial management skills or people who must spend all of their income, leaving little for future consumption. This finding aligns with the study by Betti, Dourmashkin, Rossi, and Yin (2007), which said that households with poor financial management and myopic consumption behaviour had a higher chance of being over-indebted. Moreover, it affirms the analysis undertaken in Chapter 5, which suggests that people with higher income-to-expenditure ratio tend to have a higher chance of being over-indebted.

Secondly, focusing on the effect of dependency, this analysis showed that more household members who do not have their own income (more children or retirees) could cause lower debt performance over time. This finding confirms the result from the previous studies (Betti et al., 2007; Xiao & Yao, 2014). Again, this result affirms the static analysis in Chapter 4, which states that households with more financially dependent members tend to have poorer debt performance. Currently, the government and the Stock Exchange of Thailand have encourage Thai new couples to have a good financial plan before having kids (Stock Exchange of Thailand, 2019a). The plan includes the target number of kids, related future expenses (e.g. food, health, cloth, and education), and saving-and-investment plan. Additionally, this is the first time that the dynamic relationship between dependency and debt status has been quantitatively proved for Thailand.

Thirdly, with regards to the importance of credit access, this study has found that a higher proportion of people who rely on an informal source of credit can cause lower average debt performance because of the higher interest rates in informal credit markets. This result is also consistent with the research by Betti, Dourmashkin, Rossi, and Yin (2007). In addition, in the case of Thailand, the government has tried to create more channels of credit-access through specialised financial institutions and village funds. The Thai government aims to help households, who cannot obtain credit from banks (i.e. people in rural areas and people with low credit profiles such as farmers), to get loans to increase output (e.g. agricultural loans). However, the government might do more to encourage good financial management among these groups of households along with more credit-access channels. This issue will be discussed in Chapter 7.

157

Fourthly, focusing on the impact of homeownership, according to the results from the regressions, there is a positive dynamic correlation between homeownership (including land ownership) and debt performance. To explain this relationship, over time, homeownership can prevent households from having debt difficulties owing to their having real collateral. This is to say, homeowners can use their properties as debt collateral to secure loans from formal sources at lower interest rates than people without homes. This finding is aligned with the previous study by Keese (2009). However, regarding the static statistics from Figure 5.2 in Chapter 5, there is a higher homeownership rate among households who are objectively over-indebted but not subjectively over-indebted than households who are subjectively over-indebted but not objectively over-indebted. Therefore, households should get the benefits from their homeownership (i.e. more credit access at lower costs), and, with good financial discipline, this allows them to avoid the need for excessive borrowing.

Lastly, with regards to the significance of education, households with (at least) undergraduate degrees tend to have healthy financial status in terms of debt performance and ability to save. One possible reason for the correlation between the level of education and debt performance may be that these households have better financial literacy and financial management that contributes to superior management of household debt compared to those who do not have similar qualifications. In earlier chapters, I found that this was a plausible explanation of the data. If we refer to Table 5.4 in Chapter 5, we can see that those with a bachelor’s or a master’s degree have a more accurate perception of their financial circumstances, so it seems reasonable to conclude that this result holds over time.

The next chapter will summarise the empirical findings from all the previous chapters (Chapter 4 – 6), and draw implications for policymaking.

158

Chapter 7: Conclusions and Policy Implications

7.1 Conclusions

This thesis has shed light on the household debt situation in Thailand plus the implications of the above on household welfare. Thailand is an interesting case study on household debt analysis for at least three reasons. First, over the past 10 years, Thailand experienced a rapid increase in the level of household debt. Household debt grew at an annual average rate of 15% from 2009 to 2013. Thailand, during this period, experienced the highest growth rate of household debt amongst its peers in Asia. Second, there are few studies of households’ financial behaviour in the case of developing countries. As explained in the introductory chapter, there are three research gaps in this field of study: (i) household indebtedness assessments in terms of both subjective and objective measures, (ii) analysis using both static and dynamic approaches, and (iii) the significance of households’ debt behaviour. These research gaps lead to the research questions to figure out the characteristics of households that tend to be over-indebted and to find differences between subjective over-indebtedness and objective over-indebtedness. The third reason motivating this study’s interest in Thailand is that micro data from official household surveys (Thailand’s Household Socio-Economic Surveys) are available for doing some studies to close the research gaps. Using this kind of data, this study can compare and contrast between different groups of households with both static analysis and dynamic analysis. Therefore, such analyses can make essential contributions in the understanding of households’ debt behaviour in developing countries in terms of households with inconsistent debt assessments (subjective and objective assessments), and practical policy implications, including lending practices.

This chapter begins with a summary of the household debt situation in Thailand from 2009 to 2015. Then it will conclude with the empirical findings of the analysis.

7.1.1 Thailand’s household debt situation from 2009 to 2015

For the overview of the household debt situation in Thailand, firstly, this study summarised the development in household debt during the period of 2009 to 2015 using household-level data (household surveys of 2009-2015). For overview statistics, from 2009 to 2015, the proportion of indebted households had been declining over the period

159 resulting in lower average growth in debt for all households relative to the average growth in their income. However, focusing on indebted households only, their levels of indebtedness as reflected from the average debt-to-income ratio increased due to the rate of growth of debt being greater than the rate of growth of their income. Moreover, with regards to the ability to save, on average, indebted Thai households have low ability to save due to a high level of the expenditure-to-income ratio that is close to 1 (i.e. the ratio which is closer to 1 reflects lower ability to save). The saving rate of indebted households is only around 2-6% (estimated from the gap between monthly income and expenditure). However, the average debt service ratio (i.e. monthly debt service payment to income ratio) from 2009-2015 remained stable, and debt performance—the ratio of household’s monthly income after debt service payments to household’s poverty line - was improving slightly because of the higher growth in household income relative to the growth in the household poverty line. From these statistics, during 2009-2015, Thai indebted households had higher indebtedness along with low saving ability, while their debt service burden relative to their income was unchanged.

In this thesis, household financial status is characterised in terms of debt service ratio, debt performance, and expenditure-to-income. Additionally, different groups of households have been sorted by income, region, and social class to consider the impact of these factors on household characteristics. Chapter 6 presented evidence of the fact that low-income households, households in the North-east of Thailand, agricultural households, and retirees have the most precarious debt status. These household groups have heavier debt service burden (higher debt service ratio) and a shorter gap between income after debt service payment and the poverty line (lower debt performance) than the others. Over the period 2009-2015, they have experienced a greater increase in their outstanding debt and debt service payments than the increase in their income. In addition, these groups of households have a low ability to save as reflected by a high expenditure-to-income ratio.

In addition, the statistics from Household Socio-Economic Surveys of the first quarter of 2013 showed that people in the North-east, agricultural households, and retirees had a lower proportion of households with the perception of having secure income than the others. This meant that, besides their unsound financial status, these household groups had fluctuating income and tended to face difficulties in estimating their cash inflows. 160

Next, there will be a summary of the empirical results of the analysis chapters in this thesis.

7.1.2 Empirical results

Turning to the empirical findings from the analysis chapters in this thesis, there are three main analytical parts in this thesis: (i) the drivers of household indebtedness (with the findings presented in Chapter 4), (ii) differences between subjective over- indebtedness and objective over-indebtedness (Chapter 5), and (iii) the dynamic analysis of household indebtedness (Chapter 6). The data for the foregoing analyses have been drawn from the official household surveys that include information on the household’s balance sheet and other characteristics (i.e. region, social class, and household head’s age and gender) as explained in Chapter 4. Chapters 4 and 5 utilised information from the household surveys of the first quarter of 2013, which was combined with the data from the Bank of Thailand’s supplement survey that includes information on financial literacy and financial status (e.g. concerns about next debt repayment and the perception of having secure income) to identify the drivers for indebtedness. Chapter 6 uses household surveys over the period 2009-2015 to construct pseudo-panel data to test the robustness of the findings presented in Chapters 4 and 5. There are at least two advantages of using such data: 1) panel analysis can show dynamic relationships between an endogenous variable and explanatory variables, and 2) we can control unobserved household heterogeneity by using panel data. The analyses in this thesis are mainly quantitative, drawing on static analysis using ordinary least squares (OLS) coupled with logit estimations in Chapter 4 and 5, and dynamic analysis using panel regressions in Chapter 6.

In this thesis, with regards to household indebtedness, a household’s debt performance has been defined as the ratio of household’s monthly income after debt service payments to household’s minimum subsistence level of income (Keese, 2009). Moreover, with regards to household over-indebtedness, I define subjectively over- indebted households as households expressing concerns about their next debt repayment (Anderloni & Vandone, 2008), whereas I define objectively over-indebted households as households with monthly income after debt service payments less than household’s minimum subsistence level of income (Keese, 2009, D’Alessio & Iezzi, 2013). In this

161 thesis, the poverty line has been employed as such level of income (D’Alessio & Iezzi, 2013).

With regards to theories related to the analysis of household indebtedness, there are three main points of view, and each has been examined within the thesis: (i) neoclassical theories (the life cycle hypothesis and the permanent income hypotheses), (ii) credit rationing, and (iii) behavioural finance. This study has found that all the theories play significant roles in accounting for household indebtedness in Thailand.

Role of the life cycle hypothesis and the permanent income hypotheses

This section focuses on the results related to the life cycle hypothesis and the permanent income hypotheses. The empirical results of Chapter 4 show that households tend to have better debt performance when they are at working ages (i.e. those between 21-60 years of age) and expect higher income along their working paths. These results support the Life Cycle and Permanent Income Hypothesis and are aligned with the findings of the literature, which state that young households with low income are more likely to have poor debt performance (Vante, 2006; Betti et al., 2007). Moreover, the results show that, in the case of Thailand, households who perceive themselves to have secure income are more likely to have superior debt performance (that is a higher ratio [and one strictly greater than unity] of monthly income after debt service payments to the poverty line). Additionally, the results from the models in the analysis of the drivers of household indebtedness (Chapter 4) suggest that households with regular income- and-expenditure accounts are less likely to be both subjectively and objectively over- indebted. In the case of developing countries, this is the first time that the roles of having secure income, and doing income-and-expenditure accounts are statistically proved. The latter finding can be linked to the importance of financial management and will be discussed later in the policy implication part.

Again, the unique contribution of the analysis in this thesis is corroborating the subjective and objective assessments of over-indebtedness. The literature has analysed these measures separately. Thus, the analysis of differences between subjective and objective over-indebtedness, which is presented in Chapter 5, is new in this regard. It sheds fresh light on the drivers of household indebtedness. In reality, there are people who have inconsistent debt assessments (i.e. some people are objectively over-indebted but not subjectively over-indebted, or some people are subjectively over-indebted but

162 not objectively over-indebted). Regarding the different effect sizes of having secure income and doing income-and-expenditure accounts on subjective and objective over- indebtedness, the empirical results of Chapter 5 show that doing income-and- expenditure accounts coupled with the perception of having secure income (regular income) have similar sizes of effects on both the chance of becoming subjectively over- indebted and objectively over-indebted. Therefore, households who undertake income- and-expenditure accounts and those who have a secure source of income seem to be rational in the neoclassical way (the Permanent Income Hypothesis and the Life-Cycle Hypothesis) because they can realise their cash flows over the lifetime and seem to have consistent debt assessments (due to the similar sizes of effects of these factors on both subjective and objective over-indebtedness).

Focusing on the effect of dependency on debt status, this thesis defines the dependency ratio as the ratio of household members with no income to total members. The results of Chapter 4 show that the greater the number of members who do not have their own income (more children or retirees) in these households, the worse is their debt performance. To link this finding to the role of household head’s age, it can be claimed that households with young household head and low income together with higher dependency ratio tend to have poorer debt performance than the others. This finding confirms the results from the previous studies (Betti et al., 2007; Xiao & Yao, 2014), which stated that single-adult households and young head of household with many children are likely to have poor debt status.

Moreover, from the estimations in the analysis of differences between subjective and objective over-indebtedness (Chapter 5), although the effect of the dependency ratio on objective over-indebtedness is strongly significant, the effect on subjective over- indebtedness is not. This implies that households with more children and retirees may think that they can manage their cash flow, but they may not realise that they might have greater financial burdens with more dependent members (e.g. educational expenses, health expenses and cost of living).

In addition, the results of Chapter 5 also showed that the expectation of deterioration in the household’s economic status (e.g. employment status, occupation, and income level) could make households more concerned about their next debt repayment, whereas it has less effect on households’ objective debt status. This finding

163 supports the previous study of the role of household’s economic expectation (Keese, 2010).

Role of credit rationing

Focusing on the role of credit rationing, the results of Chapter 4 and 6 showed that households, which could obtain loans from the formal sector, have a higher probability of being over-indebted. This result is aligned with the previous study (Betti et al., 2007) that states that the credit constraint has a positive relationship with over- indebtedness. This is the first time this result has been statistically confirmed for Thailand. A higher proportion of people who rely on an informal source of credit are significantly related to lower average debt performance because of the higher interest rates in informal credit markets. In Chapter 5, it was also found that credit constraints have larger effects on the chance of being subjectively over-indebted than on the chance of being objectively over-indebted. In other words, households, that obtained loans from informal sources, tended to be more concerned about their debt repayments than households who obtained the same amount of loans from the formal sources.

In addition, according to the literature, some researchers use different types of loans to explain different levels of household indebtedness (Keese, 2009; Xiao & Yao, 2014). This thesis divided loans into five types: (i) agricultural loans, (ii) personal consumption loans, (iii) business loans, (iv) education loans, and (v) mortgage loans.

Regarding the effect of different types of loans, agricultural business loans have the greatest effect on debt performance and objective over-indebtedness. This is also the first time that the significance of this type of loan is statistically proved in the case of Thailand, and it reflects the inherent poverty of those in the agricultural sector combined with the fact that loans of that type are unsecured by collateral. Lenders’ credit standard, combined with credit rationing, then imply higher interest rates for such loans. This limits households’ ability to deal with income shocks. This finding also explains why the Thai government intervenes in the agricultural sector by way of subsidies, which have the effect of stabilizing incomes and ameliorating the impacts of the capital market in that sector. This matter will be discussed further in the policy implication part.

With regards to the impacts of personal consumption loans and business loans, personal consumption loans lead to a higher chance of becoming both subjectively over- indebted and objectively over-indebted at similar sizes. Whereas, business loans are 164 more strongly correlated with worse objective debt performance than they are with subjective concerns about being able to repay debts. This result suggests that entrepreneurs are more confident than the objective data suggest ought to be the case. For example, some of them believe that they can manage cash flows as even they have irregular inflows.

Focusing on to the effect of education loans, the results of Chapter 5 show that households take loans for the education of their young members with the judgement that this kind of loans is burdensome as indicated by the comparison between subjective and objective measures of over-indebtedness. Education loans, which are another form of non-collateralised loans, have an opposite effect on subjective and objective over- indebtedness. This type of loans has the greatest positive effect on the chance of being subjectively over-indebted. However, such loans have a negative impact on the likelihood of being objectively over-indebted. This pair of facts implies that there is likely to be some degree of welfare loss to the Thai economy as a result of the imperfection in the education loans market. The reason for this is that borrowers who anticipate themselves to be subjectively worse off as a result of taking out such loans, overestimate the likelihood of actually (i.e. objectively) being over-indebted. This causes subjectively over-indebted households to take out fewer education loans than they would if they knew the real probability of being over-indebted.

As a middle-income developing country, education loans (or student loans) in Thailand may not only make households earn more income in the future, but also have positive externalities on the whole economy through the creation of a higher level of human capital. Concerning household under-investment in education and the externality benefits of education on the economy, this evidence suggests that a wedge is driven between subjectively motivated educational choices of households and the level of education loans that is optimal for Thailand. This is a significant matter for future research.

Regarding the effect of mortgage loans, this kind of loan is collateralised and predictably has a smaller effect on indebtedness than the other kinds of loans. However, based on the findings from Chapter 5, mortgage loans have larger effects on the likelihood of being subjectively over-indebted than on the likelihood of being objectively over-indebted. Furthermore, focusing on the impact of homeownership, according to the results from the pseudo-panel analysis in Chapter 6, there is a positive 165 dynamic relationship between homeownership (including land ownership) and debt performance. To explain this correlation, over time, homeownership can prevent households from having debt difficulties (i.e. over-indebtedness) because homeownership significantly allows them to have more access to formal sources of credit (due to debt collateral) with lower borrowing rates. This finding is aligned with the previous study by Keese (2009) who claims that homeowners can earn the benefit from asset accumulation. Next, the role of behavioural finance will be considered.

Role of behavioural finance

Apart from the findings related to the neoclassical theories (i.e. the Life Cycle and Permanent Income Hypothesis) and credit rationing, with regards to the point of behavioural finance, the results of Chapter 4 showed that financial literacy and hyperbolic discounting behaviour have impacts on debt performance and objective over- indebtedness. Regarding the importance of financial literacy, the results affirmed the previous findings of the literature (Lusardi & Tufano, 2009; Disney & Gathergood, 2012) which claimed that households with poor financial literacy had higher costs of borrowing, and also had a higher chance of being over-indebted. The result of hyperbolic discounting behaviour is also aligned with the literature (Gathergood, 2012), which points out the positive relationship between self-control problems (hyperbolic discounters and impulsive spenders) and over-indebtedness. According to the findings from this thesis, this is the first time in Thailand that the significance of financial literacy and financial discipline on household’s debt status is proved using quantitative techniques.

As the finding in Chapter 5 that financial literacy (interest rate calculations) and hyperbolic discounting behaviour have impacts on objective over-indebtedness only, in the case of developing countries like Thailand, people may be objectively over-indebted since they do not realise the real cost of borrowing. Some households get loans from informal sources with high interest rates. However, they cannot do interest calculations. They only know the amount of each repayment, but they do not know how large the interest burden is. As a result, this group of people (i.e. people with low financial literacy and poor financial management who borrow from informal sources of credit) tend to be objectively over-indebted but not subjectively indebted. These results confirmed this study’s initial expectations, and demonstrated the importance of the

166 comparison between subjective and objective over-indebtedness. Moreover, we can make another link between the role of education and the significance of financial literacy: households with (at least) undergraduate degrees tend to have healthy financial status in terms of debt-performance and ability-to-save since they are deemed to have better financial literacy and financial management than those who do not have the degrees.

Regarding households’ financial management more generally, the results from the analysis of the drivers of household indebtedness (Chapter 4) suggest that having insufficient savings to deal with income shocks (i.e. saving that can fund monthly expenditure less than three months), combined with having problems with other financial commitments (such as rent and utility bills) over the past 12 months are significantly correlated with both debt performance and over-indebtedness. The significance of this finding is that it provides a useful indicator for banks and authorities to anticipate which kinds of households are likely to default on their debts - namely, those who are falling behind in meeting their basic household expenditures. This implication will be discussed next.

Although insufficient savings and having problems with other financial commitments have significant impacts on the chance of being over-indebtedness, the results of Chapter 5 showed different effect sizes on subjective and objective over- indebtedness. Insufficient savings and having problems with rent or utility bills have larger effects on the likelihood of being subjectively over-indebted than on the likelihood of being objectively over-indebted.

Furthermore, focusing on the relationship between hand-to-mouth consumption behaviour (i.e. the expenditure-to-income ratio close to one) and household indebtedness, the results of Chapter 5 and 6 suggested that a higher level of expenditure-to-income ratio leads to a higher chance to be objectively over-indebted than to be subjectively over-indebted and causes weaker debt performance over time, respectively. People with a high expenditure-to-income ratio can be people with poor financial management or people who enjoy spending at the present. This finding aligns with the study by Betti, Dourmashkin, Rossi, and Yin (2007), which said that households with poor financial management and myopic consumption behaviour have a higher chance of being over-indebted.

167

In addition, the results of Chapter 6 showed a gender dimension to household indebtedness. Regarding the relationship between gender and over-indebtedness, women tend to be more risk-averse than men even though there is no difference in their objective debt status. This finding aligns with the previous research by Jianakoplos and Bernasek (1998), Lenton and Mosley (2008), and Keese (2010) who found that women were more financially conservative than men.

The next section will discuss policy implications related to the empirical findings from this study.

7.2 Policy implications

According to the study on household indebtedness in Thailand, the empirical findings lead to six policy implications. The first implication follows from the finding that having secure income can improve debt performance and can make households less likely to be over-indebted. The implication is that a better income safety net can lead to better household economic wellbeing. Authorities can promote policies that make households feel that their income is secure. Linking this point to the finding that agricultural business loans have the greatest effect on debt performance, for agricultural households, which face fluctuating agricultural product prices and uncertainty in weather, the Thai government might introduce a long-term crop insurance scheme instead of the short-term crop subsidy scheme that it currently employs. This finding thus contributes to the current debate on this issue in Thailand (for an analysis of the ‘pros and cons’ of a government rice subsidy scheme, see Permani & Vanzetti, 2016). Permani and Vanzetti (2016) claimed that the previous rice pledging scheme (two-year scheme) by the government in 2011 motivated Thai farmers to increase their crop amount with poor quality due to short harvest time; although, the scheme helped low- income farmers to have higher revenue in the short run. Therefore, the long-term crop insurance scheme with the condition of consistent crop quality will let agricultural households have a kind of long-term secure income. The scheme will also encourage them to maintain the quality of their products instead of over-production with low quality in the short term, which is a result of the impulse from the short-term subsidy scheme.

The second implication follows from the finding that people who rely on an informal source of credit tend to be over-indebted due to the higher borrowing rates. In Thailand,

168 credit constraints operate in the formal credit market of the economy and drive some households to borrow in the ‘grey’ or informal market at high interest rates. In recognition of the negative impacts of the informal credit sector, the Thai government has tried to create more credit-access channels through specialised financial institutions and via specially created village fund. These channels can help households in rural areas and households with low credit profiles to get loans (e.g. agricultural loans). However, the opening of these channels, while necessary may be insufficient to encourage their uptake - the reason being the relative financial illiteracy of people making use of informal credit markets. Therefore, promoting sound financial management and good financial literacy, along with providing more credit channels, would lead to more credit access for households with more financial responsibility. The importance of encouraging people to have sound financial management and good financial literacy will be discussed later.

Thirdly, based on the finding that households who are falling behind in meeting their basic household expenditures (such as rent and utility bills) tend to have debt defaults, authorities can consider the utilisation of a leading indicator that can signal impending default risk. According to Bank of Thailand (2017c), with regards to the current credit lending practice regulated by the Bank of Thailand, loan applicants’ income is the main factor which is examined by financial institutions for granting credit to borrowers (loan applicants). For example, in the case of personal loans, if loan applicants’ average monthly income is less than THB 30 000, the maximum personal credit amount is 1.5 times of the income. While a personal credit amount can be up to five times of loan applicants’ monthly income if the income is THB 30 000 or more, the findings from this thesis about the significance of financial commitments difficulties on households’ debt status, the record of having problems with rent and utility bills (e.g. late payment) over the past twelve months can be used as a leading indicator of debt default and can be added to credit scoring applications. People, who miss their rent or utility bills, have a higher chance of debt default. Additionally, a question set on financial literacy and financial management also can be included in the applications to calculate borrows’ financial literacy score and to evaluate unsound financial behaviours (e.g. hyperbolic discounting behaviour and hand to mouth-to-mouth consumption behaviour). This additional information will be useful for financial institutions, such as commercial banks and specialised financial institutions (i.e. state-owned financial

169 institutions), and will allow them to provide the right amount of credit to the right borrower.

The fourth implication is the importance of promoting good financial discipline and financial literacy. There will be a robust contribution to social welfare in terms of household financial wellbeing if the authorities raise the issue of the capabilities of households in terms of good financial management and financial literacy over time. For example, the government can target household groups with unsound financial status (e.g. low-income households with insecure income, households in the North-east of Thailand, agricultural households, and households who nearly get retired) as targeted households. Then the government may have a set of campaigns that encourages the targeted people to do income-and-expenditure accounts regularly, and the government can enlighten people to know how to calculate the real cost of borrowing. Once households organise their income-and-expenditure accounts regularly, they will be able to manage their cash flows better and plan how much they want to save (saving goals). This, then, permits them to have sufficient savings to deal with income shocks. Regularly doing income-and-expenditure accounts is one way to allow households to have proper financial management and lessen the chance of being both subjectively and objectively over-indebted. Additionally, when households can perceive the exact cost of borrowing, they try to compare borrowing rates from different sources of credit and can get loans at the lowest costs. Over recent years, the Bank of Thailand has encouraged Thai people to do their own income-and-expenditure accounts and has raised the importance of interest calculations and saving discipline (Bank of Thailand, 2016d). The Bank of Thailand’s current target groups are low-income households, retirees, and illiterate households in urban areas who have access to the internet (through the Bank of Thailand’s website and its Facebook page). Therefore, it will be more efficient with higher impact if the government can cooperate and expand the campaign to other rural areas in Thailand.

In addition, currently, the Bank of Thailand has announced the second phase of the Debt Clinic Programme in May 2019 (Bank of Thailand, 2019d). The programme is the cooperation between the Bank of Thailand and Sukhumvit Asset Management (also known as SAM), and it aims to help indebted households who have problems dealing with credit card and personal loans. People who join this programme will have a chance to restructure their loans via SAM. Moreover, according to the Bank of Thailand’s 170 announcement (2019d), the second phase of the Debt Clinic has covered the loans from banks and non-banks (i.e. financial companies and credit fonciers), while the first phase covered only the loans from banks. The research undertaken in this thesis confirms that the rollout of the Debt Clinic Programme to a wider class of borrowers is warranted.

Furthermore, focusing more on the role of financial literacy on household financial sustainability, the data suggest that the greatest impact will be obtained by investing in primary and secondary schools. Once individuals have proper financial literacy, they will be able to reckon their current debt status and have the information needed to manage cash flows for the future repayments. Otherwise, if households are not concerned about their debt serviceability even though they have a weak financial status, they have a higher chance of facing debt unsustainability. This situation means these households may have to sacrifice their minimum standard of living when they are trying to repay their debt (Betti et al., 2007). Therefore, sharpening households’ financial literacy and use of financial innovations can be thought of as intermediate policies aimed ultimately at improving households’ financial status and diversifying financial service channels, respectively.

Moreover, there is the link between the role of financial literacy and the role of different types of loans. As some households need credit access to secure homeownership and higher education, the authorities can clarify the costs and benefits of different types of loans. The comparison between productive loans (e.g. business loans, mortgage loans, education loans) and non-productive loans (e.g. personal consumption loans) can be considered. To illustrate this point, on the one hand, households can use productive loans for generating their current and future income, and for accumulating their real and human capital. On the other hand, households cannot earn any income with non-productive loans, but they can use this type of loans for smoothing their consumption. The authorities, together with financial institutions, could publish advertisements about the different details of each type of loans (e.g. purposes, terms, and interest rates) on their websites and through social media channels (e.g. Facebook and Twitter). Then households would have the opportunity to understand the significance of different types of loans. For example, if households wanted to earn more income in the future and to accumulate their wealth, they could plan to apply for education loans or mortgage loans. Generally, productive loans are usually chosen for medium-to-long term effects, which will lead to household wellbeing in the future. The 171 key to this achievement is that households have to take full financial responsibility after getting the loans, e.g. they have to be able to pay back their loans.

The fifth implication is from the finding that the effect of the dependency ratio on objective over-indebtedness is highly significant, but the effect on subjective over- indebtedness is not. This implies that some households do not perceive the burden from there being more members with no income (many children or retirees). Therefore, it affirms the importance of the current campaign by the government (Stock Exchange of Thailand, 2019a), which encourages individuals to be able to manage their cash flows wisely by knowing the current financial status and predicting cash flows in the future. For example, recently, the government has tried to educate couples on financial readiness before having children. Couples should know the financial burden of having children (for example, food and education expenses). Additionally, the authorities may also encourage households to prepare themselves before getting retired in terms of financial cushions (e.g. savings) and a health safety net (e.g. health insurance). Once couples and retirees have financial readiness, they will have economic wellbeing and will be able to smooth their consumption under unexpected circumstances.

Last but not least, according to the finding that education loans have a negative relationship with the probability of being objective over-indebtedness, the sixth implication is the importance of education. It suggests that there is social value in providing income-contingent loans for study. Such loans can raise household income, and education can also have positive impact on society providing a higher quality of human capital. In the case of Thailand, the government can play an important role in promoting a higher education level of Thai students. In addition to mandatory free education for 15 years (UNICEF, 2019), the government can grant a kind of income- contingent education loans, which are linked to borrower’s future income (Chapman et al., 2009; Chapman & Lounkaew, 2010). Chapman and his colleagues are interested in student loan schemes in both developed and developing countries. In the case of Thailand, they analyse the pros and cons of Thailand’s current Student Loan Fund granted by the Ministry of Finance. This student loan fund is annually provided to Thai universities to help Thai students from low-income families (i.e. annual family income less than THB 150 000). The fund covers tuition fees and education-related expenses for upper secondary, vocational, and undergraduate education. To repay the loans, two years after graduation, the loan recipients are asked to pay the loan back within 15 years 172 with a progressive repayment rate (from 1.5% in year 1 to 13% in year 15) and a flat interest rate of 1%. Chapman, Lounkaew, Polsiri, Sarachitti, and Sitthipongpanich (2009) claimed that the student loan fund in Thailand was good in terms of borrowers’ repayment hardship because students with this loan would bear a small proportion of the loan repayment to total income, which is less than 5% (estimated from their average income at the age of 24-38 years). However, Chapman, Lounkaew, Polsiri, Sarachitti, and Sitthipongpanich (2009) and Chapman and Lounkaew (2010) were concerned about the high level of the government’s implicit subsidy because of the low interest rate (1%) and the two-year repayment grace period. The implicit subsidy is estimated from the present value of loan loss to the present value of total loan amount and collection cost. They suggested that the Thai government could reduce the implicit subsidy from the current student-loan scheme by imposing income-contingent-education-loan scheme. The income-contingent education loan is another type of student loan, which is linked to borrower’s income, and the loan repayments are deducted automatically from income.

An example of a well-known income-contingent education loan is HECS-HELP, which is a student loan programme in Australia. HECS stands for Higher Education Contribution Scheme, and HELP stands for Higher Education Loan Programme. According to Study Assist (2019), the HECS-HELP scheme provides student loans to Australian citizens who enrol in the Commonwealth supported places at tertiary institutions. The Australian Government will pay the cost of course fees directly to the educational institutions. To pay back the HECS-HELP loan, the repayments of this loan will be taken automatically from income once the income level meets the compulsory repayment threshold (Study Assist, 2019). The current threshold is AUD 51 957, and the repayment rates are progressive (i.e. higher income, higher repayment rate). In the case of Thailand, for instance, the government can choose potential high-school students (i.e. students with an outstanding study profile and social responsibility) and grant them such education loans until they obtain a bachelor’s degree. The loan repayments will be deducted from their monthly income after they graduate and start to work with threshold income (the current minimum salary for new employees with a bachelor’s degree is THB 15 000 per month). This type of loan, therefore, can mitigate the downside risk of the underinvestment in education, and the issue on the proper amount of this loan and the affordable interest rate is an important matter for further studies.

173

7.3 Main contributions of the study

There are three main contributions to the broad literature made by this thesis on household debt and further knowledge in the case of developing countries. Firstly, this thesis has raised the importance of the differences between subjective over-indebtedness and objective over-indebtedness. According to the analysis in this thesis, it shows that some households have inconsistent assessments of their debt status.

To illustrate this point, on the one hand, although some households have a sound financial status, they are overly conservative and worry about the financial burden from seeking loans (even the loans are productive loans such as education loans and business loans). This group of people tends to be subjectively over-indebted but not objectively over-indebted. Therefore, these households may under-invest in their education and businesses.

On the other hand, there are some people who enjoy spending and obtaining loans regardless of their financial status. This type of person tends to be objectively over-indebted but not subjectively over-indebted. Therefore, focusing on subjective or objective over-indebtedness only may not be enough. Moreover, in the case of Thailand, this is the first time that micro data (household surveys) has been used to compare and contrast the household characteristics which are the drivers of household subjective over-indebtedness and objective over-indebtedness.

Secondly, this study applies both static analysis and dynamic quantitative analysis for the statistical robustness of the results. According to the results from this study, it can be confirmed that there is a significance of education, credit access, and the dependency level on debt performance, which leads to the potential policies for improving household economic wellbeing, as discussed in the policy implication section of this thesis. Moreover, the dynamic analysis using pseudo-panel data in this thesis leads to a contribution to the panel analysis of household indebtedness in Thailand where there is no sufficient genuine-panel household survey. The relationships over time from such dynamic analysis can fulfil the relationships at one point of time from the static analysis.

Thirdly, with regards to the determinants of household indebtedness, in the case of developing countries, financial institutions can apply the results of this study for credit-scoring applications. Information about having a late payment on rent or utility

174 bills over the past 12 months, together with the evaluation on financial literacy and financial management, can be added to borrowers’ credit profiles, which lets financial institutions (lenders) know their customers better. More useful details in borrowers’ credit profile can improve the issue of the asymmetric information in the markets.

7.4 Areas for further research

For the further study in this area, first of all, in the case of developing countries, similar dynamic analysis of household indebtedness using genuine-panel household surveys can be done to confirm the empirical findings for this group of countries. With the genuine-panel household surveys, the same set of households can be tracked over time. Dynamic analysis using genuine-panel data can show the robust dynamic relationship between household indebtedness and its determinants, and the analysis will ensure that unobserved cross-section heterogeneity is controlled (Wooldridge, 2010). However, the lack of suitable panel household surveys in Thailand leads to an attempt to constructing pseudo-panel for dynamic analysis (panel analysis). Doing dynamic analysis with pseudo-panel data is challenging because of the way of constructing proper household cohorts (i.e. grouping households with the same characteristics). As mentioned in Chapter 6, some researchers (Verbeek & Nijman, 1992; Dargay & Vythoulkas, 1999; Verbeek, 2008; Tsai et al., 2014) discussed the issue of the reliability of this type of data set. They suggested that conditions such as cohort sizes, time variation in cohort means, and variation between cohorts needed to be held for tracking household cohorts as the same individuals over time.

Second, in the future, panel analysis can be done using more years of the surveys. This thesis employed four years of household surveys (SES of 2009, 2011, 2013, and 2015) because of data availability. The analysis using longer-time-series data will permit confirmation of long-run relationships.

Third, further analysis of macro-welfare benefits to suggested policies, especially in the case of income-contingent education loans, would also be useful. For example, cost-benefit analysis of this type of loan could be done using macro-data.

175

References

Aloui, R., Aïssa, M. S. B., & Nguyen, D. K. (2011). Global financial crisis, extreme interdependences, and contagion effects: The role of economic structure? Journal of Banking & Finance, 35(1), 130-141.

Anderloni, L., & Vandone, D. (2008). Households over-indebtedness in the economic literature (Working Paper No. 2008-46). Milan: University of Milan.

Anderloni, L., & Vandone, D. (2010). Risk of overindebtedness and behavioral factors (Working Paper No. 2010-25). Milan: University of Milan.

Attanasio, O. P. (1999). Consumption. Handbook of macroeconomics, 1, 741-812.

Bank of Canada (2017). Financial system review (June). Retrieved from https:// www.bankofcanada.ca/wp-content/uploads/2017/06/fsr-june2017.pdf.

Bank of England (2017). Financial stability report (June). Retrieved from https:// www.bankofengland.co.uk/-/media/boe/files/financial-stability-report/2017/ june- 2017.pdf?la=en&hash=EB9E61B5ABA0E05889E903AF041B855D79652644.

Bank of Thailand (2008). Thailand's economic conditions in 2008. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReport_2008.pdf.

Bank of Thailand (2009). Thailand's economic conditions in 2009. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReport_2009.pdf.

Bank of Thailand (2010). Thailand's economic conditions in 2010. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReport_2010.pdf.

Bank of Thailand (2011). Thailand's economic conditions in 2011. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReport_2011.pdf.

Bank of Thailand (2012). Thailand's economic conditions in 2012. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReport_2012.pdf. 176

Bank of Thailand (2013a). Financial access surveys of Thai households. Retrieved from https://www.bot.or.th/English/FinancialInstitutions/Highlights/Pages/2013%20Fi nancial%20Access%20Survey%20Final.pdf.

Bank of Thailand (2013b). Financial stability report 2013. Retrieved from https://www.bot.or.th/English/FinancialInstitutions/Publications/FSR_Doc/FSR2 013e.pdf

Bank of Thailand (2013c). Thailand's economic conditions in 2013. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/AnnualReportEng2013.pdf.

Bank of Thailand (2014a). Financial stability report 2014. Retrieved from https://www.bot.or.th/English/FinancialInstitutions/Publications/FSR_Doc/FSR2 014e.pdf

Bank of Thailand (2014b). Thailand's economic conditions in 2014. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/Annual_eng_2014.pdf.

Bank of Thailand (2015). Thailand's economic conditions in 2015. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/annual2015_V2.pdf.

Bank of Thailand (2016a). Financial access surveys of Thai households. Retrieved from https://www.bot.or.th/English/FinancialInstitutions/Highlights/FSMP2/2016%20 Financial%20Access%20Survey%20Final.pdf.

Bank of Thailand (2016b). Monetary policy report (March 2016). Retrieved from https://www.bot.or.th/English/MonetaryPolicy/MonetPolicyComittee/MPR/Mon etary%20Policy%20Report/MPR_E_March2016.pdf

Bank of Thailand (2016c). Thailand's economic conditions in 2016. Retrieved from https://www.bot.or.th/English/MonetaryPolicy/EconomicConditions/AnnualRep ort/AnnualReport/annual_2016_V2.pdf.

Bank of Thailand (2016d). Thailand’s financial literacy survey report 2016. Retrieved from https://www.1213.or.th/th/aboutfcc/knownfcc/Documents/ThaiFLsurvey59. pdf. (Thai version only) 177

Bank of Thailand (2017a). Financial stability report 2017. Retrieved from https:// www.bot.or.th/English/FinancialInstitutions/Publications/FSR_Doc/FSR2017e. pdf.

Bank of Thailand (2017b). Monetary policy report (March 2017). Retrieved from https://www.bot.or.th/English/MonetaryPolicy/MonetPolicyComittee/MPR/Mon etary%20Policy%20Report/MPR_March2017.pdf

Bank of Thailand (2017c). Regulations procedures and conditions for undertaking business of personal loan under supervision for financial institutions. Retrieved from https://www.bot.or.th/Thai/FIPCS/Documents/FPG/2560/EngPDF/ 25600205.pdf.

Bank of Thailand (2018a). EC_MB_039: Loans to household. Retrieved from http://www2.bot.or.th/statistics/BOTWEBSTAT.aspx?reportID=775&language= ENG.

Bank of Thailand (2018b). Financial stability report 2018. Retrieved from https://www.bot.or.th/English/FinancialInstitutions/Publications/FSR_Doc/FSR2 018E.pdf.

Bank of Thailand (2018c). Monetary policy report (March 2018). Retrieved from https://www.bot.or.th/English/MonetaryPolicy/MonetPolicyComittee/MPR/Mon etary%20Policy%20Report/MPR_EN_Mar2018.pdf.

Bank of Thailand (2019a). FM_FT_002: Foreign exchange turnover of commercial banks. Retrieved from https://www.bot.or.th/App/BTWS_STAT/statistics/ BOTWEBSTAT.aspx?reportID=798&language=ENG.

Bank of Thailand (2019b). FM_RT_008: Historical data for interbank transaction volume, rates, and amount outstanding. Retrieved from https://www.bot.or.th/App/BTWS_STAT/statistics/ReportPage.aspx?reportID=6 26&language=eng.

Bank of Thailand (2019c). Performance of the Thai banking system in 2018.

Bank of Thailand (2019d). The second phase of the Debt Clinic Programme with SAM. Retrieved from https://www.bot.or.th/Thai/PressandSpeeches/Press/News2562/ n2862t.pdf. (Thai version only)

178

Barajas, A., Dell’Ariccia, G., & Levchenko, A. (2007). Credit booms: The good, the bad, and the ugly. unpublished: International Monetary Fund.

Barba, A., & Pivetti, M. (2009). Rising household debt: Its causes and macroeconomic implications—A long-period analysis. Cambridge Journal of Economics, 33(1): 113-137.

Barnes, S., & Young, G. (2003). The rise in US household debt: Assessing its causes and sustainability (Working Paper No. 206). London: Bank of England.

Beck, T., Levine, R., & Loayza, N. (2000). Finance and the Sources of Growth. Journal of Financial Economics, 58(1-2), 261-300.

Beck, T., Büyükkarabacak, B., Rioja, F. K., & Valev, N. T. (2012). Who gets the credit? And does it matter? Household vs. firm lending across countries. The BE Journal of Macroeconomics, 12(1).

Bernanke, B., & Gertler, M. (1990). Financial fragility and economic performance. The Quarterly Journal of Economics, 105(1), 87-114.

Bernard, J.-T., Bolduc, D., & Yameogo, N.-D. (2011). A pseudo-panel data model of household electricity demand. Resource and Energy Economics, 33(1), 315-325.

Betti, G., Dourmashkin, N., Rossi M., & Yin Y. P. (2007). Consumer over-indebtedness in the EU: Measurement and characteristics. Journal of Economic Studies, 34(2), 136-156.

Bingxi, S., & Lijuan, Y. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper No. 46:51-57). Basel: Bank for International Settlements.

Blundell, R., Browning, M., & Meghir, C. (1989). A microeconometric model of intertemporal substitution and consumer demand. Retrieved from https://warwick.ac.uk/fac/soc/economics/research/workingpapers/1989- 1994/twerp324.pdf.

Boonperm, J., Haughton, J., & Khandker, S. R. (2009). Does the Village Fund Matter in Thailand? (Policy Working Paper No. 5011). Retrieved from https://openknowledge.worldbank.org/bitstream/handle/10986/4202/WPS5011. pdf.

179

Browning, M., Deaton, A., & Irish, M. (1985). A profitable approach to labor supply and commodity demands over the life-cycle. Econometrica: Journal of the Econometric Society, 503-543.

Campbell, J. Y., & Mankiw, N. G. (1989). Consumption, income, and interest rates: Reinterpreting the time series evidence. NBER macroeconomics annual, 4, 185- 216.

Cecchetti, S. G., Mohanty, M. S., & Zampolli, F. (2011). The real effects of debt (Working Paper No. 352). Basel: Bank for International Settlements.

Chantarat, S., Lamsam, A., Samphantharak, K., & Tangsawasdirat, B. (2017). Thailand's Household Debt through the Lens of Credit Bureau Data: Debt and Delinquency. Retrieved from https://ideas.repec.org/p/pui/dpaper/61.html.

Chapman, B., Lounkaew, K., Polsiri, P., Sarachitti, R., & Sitthipongpanich, T. (2009). Thailand's Student Loan Fund: An analysis of interest rate subsidies and repayment hardships: Centre for Economic Policy Research.

Chapman, B., & Lounkaew, K. (2010). Income contingent student loans for Thailand: Alternatives compared. Economics of Education Review 29: 695 - 709.

Chotewattanakul, P., Sharpe, K., & Chand, S. (2019). The drivers of household indebtedness: evidence from Thailand. Southeast Asian Journal of Economics, 7(1): 1-40.

Claessens, S., Dell’Ariccia, G., Igan, D., & Laeven, L. (2010). Cross-country experiences and policy implications from the global financial crisis. Economic Policy, 25(62): 267-293.

Cohen, P. (2016). A Bigger Economic Pie, but a Smaller Slice for Half of the U.S. New York Times.

180

Credit Suisse (2016). Global Wealth Data book 2016. Retrieved from http://publications.credit- suisse.com/tasks/render/file/index.cfm?fileid=AD6F2B43-B17B-345E- E20A1A254A3E24A5.

D'Alessio, G., & Iezzi, S. (2013). Household over-Indebtedness: Definition and measurement with Italian data (Working Paper No. 149). Rome:Bank of Italy

Dargay, J. M., & Vythoulkas, P. C. (1999). Estimation of a dynamic car ownership model: a pseudo-panel approach. Journal of Transport Economics and Policy, 287-301.

Davies, M. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper No. 46: 19-30). Basel: Bank for International Settlements.

Deaton, A. (1985). Panel data from time series of cross-sections. Journal of Econometrics, 30(1), 109-126.

Deaton, A. (1989). Saving and liquidity constraints. National Bureau of Economic Research. No. w3196.

Debelle, G. (2004). Macroeconomic implications of rising household debt (Working Paper No. 153). Basel: Bank for International Settlements.

Del-Río, A., & Young, G. (2005). The impact of unsecured debt on financial distress among British households. Bank of England Working Paper 262.

DeVaney, S. A. (1994). The usefulness of financial ratios as predictors of household insolvency: Two perspectives. Financial Counseling and Planning, 5, 5-24.

Diamond, D. W., & Rajan, R. G. (2001). Liquidity risk, liquidity creation, and financial fragility: A theory of banking. Journal of political Economy, 109(2), 287-327.

Disney, R., & Gathergood, J. (2012). Financial literacy and consumer credit portfolios (Working Paper No. 12-06). Nottingham: University of Nottingham.

Dynan, K. E., & Kohn, D. L. (2007). The rise in US household indebtedness: Causes and consequences. Federal Reserve Board Finance and Economics Discussion Series, 2007-37, August.

181

Endut, N., & Hua, T. G. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper 46: 107-116). Basel: Bank for International Settlements.

European Commission. (2010). Over-Indebtedness: New evidence from the EU-SILC special module (Research Note No. 4/2010). Brussels: European Commission.

Flemming, J. S. (1973). The consumption function when capital markets are imperfect: The permanent income hypothesis reconsidered. Oxford Economic Papers, 25(2), 160-172.

Friedman, M. (1957). The permanent income hypothesis. In A Theory of the Consumption Function, 20-37. Princeton University Press.

Gardes, F., Duncan, G. J., Gaubert, P., Gurgand, M., & Starzec, C. (2005). Panel and pseudo-panel estimation of cross-sectional and time series elasticities of food consumption: The case of US and Polish data. Journal of Business & Economic Statistics, 23(2), 242-253.

Gathergood, J. (2012). Self-control, financial literacy and consumer over-indebtedness. Journal of Economic Psychology, 33(3):590-602.

Geanakoplos, J. (2010). The leverage cycle. NBER macroeconomics annual, 24(1), 1- 66.

Georgarakos, D., Lojschova, A., & Ward-Warmedinger, M. E. (2010). Mortgage indebtedness and household financial distress. ECB Working Paper 1156.

Gumy, J. (2007, 5-7 July). Explaining overindebtedness in the European Union. Paper presented at BHPS 2007 Conference Colchester, UK.

Haas, O. J. (2006). Overindebtedness in Germany (Working Paper No. 44). Geneva: International Labour Office.

Hall, R. E. (1978). Stochastic implications of the life cycle-permanent income hypothesis: theory and evidence. Journal of political Economy, 86(6), 971-987.

Huang, B. (2007). The use of pseudo panel data for forecasting car ownership (Munich Personal RePEc Archive Paper No. 7086). Retrieved from https://mpra.ub.uni- muenchen.de/7086/.

182

International Monetary Fund (2002). Thailand: Selected issues and statistical appendix. IMF Country Report No. 02/195.

International Monetary Fund (2012). Dealing with household debt. In World economic outlook: growth resuming, dangers remain (pp. 89 – 124). Retrieved from http://www.imf.org/external/pubs/ft/weo/2012/01/.

International Monetary Fund (2017). Global financial stability report: Is growth at risk? Retrieved from https://www.imf.org/en/Publications/GFSR/Issues/2017/09/27/global-financial- stability-report-october-2017.

Jianakoplos, N. A., & Bernasek, A. (1998). Are women more risk averse? Economic inquiry 36(4): 620-630.

Jordà, Ò., Schularick, M., & Taylor, A. M. (2016). The great mortgaging: housing finance, crises and business cycles. Economic Policy, 31(85), 107-152.

Kang, T., and Ma, G. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper No. 46: 95-106). Basel: Bank for International Settlements.

Keese, M. (2009). Triggers and determinants of severe household indebtedness in Germany. Ruhr Economic Paper, 150.

Keese, M. (2010). Who feels constrained by high debt burdens? Subjective vs. objective measures of household indebtedness. Ruhr Economic Paper, 169.

Kenan Institute Asia. (2015). Study Report: Literacy improvement for better finance in Thailand. Bangkok: Kenan Institute Asia.

Keynes, J.M. (1936). The General Theory of Employment, Interest, and Money London: Macmillan. pp161-162.

Laibson, D. (1997). Golden eggs and hyperbolic discounting. The Quarterly Journal of Economics, 112(2), 443-477.

Lam, N. T. (2017). Even It Up: How to tackle inequality in Vietnam. Oxford: Oxfam. http://oxf.am/ZLuU

Lenton, P., & Mosley, P. (2008). Debt and health. Working Paper. Department of Economics, University of Sheffield ISSN 1749-8368.

183

Liv, D. (2013). Study on the drivers of over-indebtedness of microfinance borrowers in Cambodia: An in-depth investigation of saturated areas. Cambodia Institute of Development.

Loayza, N. V., & Ranciere, R. (2006). Financial development, financial fragility, and growth. Journal of Money, Credit and Banking, 1051-1076.

Longstaff, F. A. (2010). The subprime credit crisis and contagion in financial markets. Journal of Financial Economics, 97(3), 436-450.

Lucas, R. E., & Sargent, T. J. (1981). Rational expectations and econometric practice (Vol. 2). University of Minnesota Press.

Lusardi, A., & Tufano, P. (2009). Debt literacy, financial experiences, and overindebtedness (Center of Financial Studies Working Paper No. 2009/08). Retrieved from http://nbn-resolving.de/urn:nbn:de:hebis:30-63796.

Martin, R. (2011). The local geographies of the financial crisis: From the housing bubble to economic recession and beyond. Journal of Economic Geography, 11(4), 587-618.

Mian, A., & Sufi, A. (2011). House prices, home equity-based borrowing, and the US household leverage crisis. American Economic Review, 101(5), 2132-2156.

Ministry of Finance (2012). Press Release No. 046/2012. (Thai version only)

Modigliani, F. (1966). The life cycle hypothesis of saving, the demand for wealth and the supply of capital. Social Research, 33(2), 160-217.

Moffitt, R. (1993). Identification and estimation of dynamic models with a time series of repeated cross-sections. Journal of Econometrics, 59(1-2), 99-123.

Muthitacharoen, A., Nuntramas, P., & Chotewattanakul, P. (2015). Rising household debt: implications for economic stability. Thammasat Economic Journal, 33(3), 66-101.

Muthitacharoen, A. (2016). Gauging Households’ Debt Tolerance: Evidence from Thailand. Applied Economics Journal, 23(1), 59-74.

OECD INFE. (2011). Measuring financial literacy: Core questionnaire in measuring financial literacy: Questionnaire and guidance notes for conducting and internationally comparable survey of financial literacy. Paris: OECD.

184

Oxfam 2017. An economy for the 99%. Oxfam Briefing Paper. Retrieved from https://www.oxfam.org/en/research/economy-99.

Permani, R., & Vanzetti, D. (2016). Rice mountain: Assessment of the Thai rice pledging program. Agricultural Economics, 47(3), 273-284.

Prescott, E. C., & Townsend, R. M. (1984). Pareto optima and competitive equilibria with adverse selection and moral hazard. Econometrica: journal of the econometric society, 21-45.

Reserve Bank of Australia (2017). Financial stability review (April). Retrieved from https:// www.rba.gov.au/publications/fsr/2017/apr/pdf/financial-stability-review-2017- 04.pdf.

Rinaldi, L., & Sanchis-Arellano, A. (2006). Household debt sustainability: what explains household non-performing loans? An empirical analysis.

South Africa Reserve Bank (2017). Financial stability review (First edition). Retrieved from https://www.resbank.co.za/Lists/News%20and%20Publications/Attachments/77 86/First%20edition%20FSR%20April%202017.pdf.

Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71(3), 393-410.

Stock Exchange of Thailand (2019a). Financial planning before having kids. Retrieved from https://www.set.or.th/set/financialplanning/lifeevent.do?name=lifeevent_ detail_child-1&innerMenuId=10. (Thai version only)

Stock Exchange of Thailand (2019b). Market capitalization. Retrieved from https://www.set.or.th/en/market/market_statistics.html.

Study Assist (2019). Understanding your HECS-HELP debt. Retrieved from https://www.studyassist.gov.au/hecs-help.

Subhanij, T. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper 46: 136-161). Basel: Bank for International Settlements.

185

Tan, W. L. (2009). Household debt: Implications for monetary policy and financial stability (Working Paper No. 46:117-123). Basel: Bank for International Settlements.

Terrones, M., & Mendoza, E. (2004). Are credit booms in emerging markets a concern. World Economic Outlook, 147, 166.

Thai Bond Market Association (2019). Daily market summary. Retrieved from http://www.ibond.thaibma.or.th/EN/Reports/Daily/Summary.aspx.

Tsai, C.-H., Mulley, C., & Clifton, G. (2014). A review of pseudo panel data approach in estimating short-run and long-run public transport demand elasticities. Transport Reviews, 34(1), 102-121.

United Nations International Children’s Emergency Fund (2019). Public Expenditure Tracking Survey (PETS) on the 15-Year Free Education Program: Kingdom of Thailand. Retrieved from https://www.unicef.org/thailand/media/281/file/ Public%20Expenditure%20Tracking%20Survey%20(PETS)%20on%20the%201 5-Year%20Free%20Education%20Program:%20Kingdom%20of%20Thailand. pdf.

Vatne, B. H. (2006). How large are the financial margins of Norwegian households? An analysis of micro data for the period 1987-2004. Norges Bank. Economic Bulletin, 77(4), 173.

Verbeek, M., & Nijman, T. (1992). Can cohort data be treated as genuine panel data? In Panel data analysis (pp. 9-23). Physica-Verlag HD.

Verbeek, M. (2008). Pseudo-panels and repeated cross-sections. In The econometrics of panel data (pp. 369-383). Springer, Berlin, Heidelberg.

Warr, P. G. (2007). Long-term economic performance in Thailand. ASEAN Economic

Bulletin, 24(1), 138-163.

Weis, C., & Axhausen, K. W. (2009). Induced travel demand: Evidence from a pseudo panel data based structural equations model. Research in Transportation Economics, 25(1), 8-18.

Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data. MIT press.

186

World Bank (2018). Poverty and shared prosperity 2018. Retrieved from https://www.worldbank.org/en/publication/poverty-and-shared-prosperity.

World Economic Forum (2019). The global risks report 2019. Retrieved from https://www.weforum.org/reports/the-global-risks-report-2019.

Xiao, J. J., & Yao, R. (2014). Consumer debt delinquency by family lifecycle categories. International Journal of Bank Marketing, 32(1), 43-59.

187

Appendix A

Table A1 Model Specifications for Subjective Over-indebtedness Model (Logit Model) Variable Model1 Model2 Model3 Model4 Perception of having secure income (dummy) -0.260*** -0.209*** -0.351*** -0.354*** Expenditure to income ratio (level) 0.470*** 0.423*** 0.251*** 0.215** Expenditure to income ratio (squared term) -0.030*** -0.025*** -0.025*** -0.023*** Insufficient savings (dummy) 0.636*** 0.714*** Dependency ratio (level) -0.033 0.068 0.324** 0.222* Age (level) 0.108*** 0.145*** 0.045** 0.046** Age (squared term) -0.001*** -0.001*** -0.0004** -0.0004** Main source of fund (BAAC) 3.196*** 3.233*** Main source of fund (GHB and GSB) 2.308*** 2.259*** Main source of fund (non-banks) 2.832*** 2.738*** Main source of fund (cooperatives) 2.430*** 2.427*** Main source of fund (village fund) 2.915*** 2.795*** Main source of fund (informal source) 3.898*** 3.784*** Credit constraint (in case of working purposes) 0.553*** 0.454*** Credit constraint (in case of emergency) 0.465*** 0.401*** Mortgage loans to annual income ratio 0.302*** 0.282*** Agricultural business loans to annual income ratio 0.330*** 0.316*** Non-agricultural business loans to annual income ratio 0.400*** 0.402*** Education loans to annual income ratio 0.711*** 0.809*** Personal consumption loans to annual income ratio 0.216*** 0.208*** Other purpose loans to annual income ratio 0.442 0.397 Financial literacy score (total) -0.048*** -0.030*** Financial literacy score (knowledge) 0.018 Financial literacy score (behaviour) -0.282*** Financial literacy score (attitude) -0.05 Financial literacy (interest rate calculations) -0.046 Hyperbolic discounting behaviour 0.044 Having problems with other financial commitments 1.296*** 1.205*** 1.194*** (dummy) Doing income-and-expenditure accounts -0.354** -0.312** -0.494*** Dummy variable for social class (using professional as a benchmark) Agricultural business 0.830*** 0.857*** 0.622*** 0.630*** Non-agricultural business 0.716*** 0.705*** 0.991*** 0.925*** Worker 0.620*** 0.558*** 0.905*** 0.861*** Retired 0.222 0.106 0.471** 0.500** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 0.567*** 0.596*** 0.03 0.064 North 0.741*** 0.773*** 0.166 0.093 North-east 1.058*** 1.092*** 0.368 0.259 South 0.604*** 0.635*** 0.399 0.271 Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 1.014*** 0.323 Married (widowed/ separated/ divorced) 0.870*** 0.494** Constant -6.298*** -6.881*** -4.708*** -7.014*** N 9430 9430 9430 9430 Wald Chi2 705.740 786.790 1315.443 1341.828 Pseudo R2 (McFadden’s) 0.102 0.107 0.283 0.294 Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. ***, **, * Significant at the 1%, 5%, and 10% level, respectively.

188

Table A2 Model Specifications for Objective Over-indebtedness Model (Logit Model) Variable Model1 Model2 Model3 Model4 Perception of having secure income (dummy) -0.297*** -0.245*** -0.293*** -0.341*** Expenditure to income ratio (level) 2.250*** 2.168*** 2.215*** 2.259*** Expenditure to income ratio (squared term) -0.004*** -0.004*** -0.004*** -0.004*** Insufficient savings (dummy) 0.406*** 0.494*** Dependency ratio (level) 1.433*** 1.466*** 1.773*** 1.749*** Age (level) 0.087*** 0.147*** 0.090*** 0.053*** Age (squared term) -0.001*** -0.001*** -0.001*** -0.0005*** Main source of fund (BAAC) 0.991*** 0.875*** Main source of fund (GHB and GSB) 0.242 0.085 Main source of fund (non-banks) 0.432*** 0.236* Main source of fund (cooperatives) 0.458* 0.318 Main source of fund (village fund) 0.996*** 0.822*** Main source of fund (informal source) 0.742*** 0.611*** Credit constraint (in case of working purposes) 0.309*** 0.242*** Credit constraint (in case of emergency) 0.167* 0.115 Mortgage loans to annual income ratio 0.161** 0.119* Agricultural business loans to annual income ratio 0.754*** 0.729*** Non-agricultural business loans to annual income ratio 0.554*** 0.544*** Education loans to annual income ratio -0.846*** -0.755*** Personal consumption loans to annual income ratio 0.244*** 0.210*** Other purpose loans to annual income ratio 0.422* 0.378 Financial literacy score (total) -0.131*** -0.111*** Financial literacy score (knowledge) -0.185*** Financial literacy score (behaviour) -0.137*** Financial literacy score (attitude) -0.039 Financial literacy (interest rate calculations) -0.304*** Hyperbolic discounting behaviour 0.056* Having problems with other financial commitments 0.769*** 0.696*** 0.749*** (dummy) Doing income-and-expenditure accounts -0.187 -0.108 -0.485** Dummy variable for social class (using professional as a benchmark) Agricultural business 1.736*** 1.765*** 1.479*** 1.504*** Non-agricultural business 0.471** 0.466** 0.529** 0.477** Worker 0.636*** 0.604*** 0.691*** 0.664*** Retired 0.483** 0.255 0.315 0.488** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 1.880*** 1.927*** 1.818*** 1.866*** North 2.374*** 2.415*** 2.164*** 2.205*** North-east 2.570*** 2.637*** 2.309*** 2.309*** South 1.939*** 2.024*** 2.033*** 1.963*** Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 1.859*** 1.402*** Married (widowed/ separated/ divorced) 1.231*** 0.909*** Constant -9.865*** -10.195*** -8.916*** -10.743*** N 9430 9430 9430 9430 Wald Chi2 941.506 1030.569 1076.452 1038.479 Pseudo R2 (McFadden’s) 0.278 0.266 0.303 0.313 Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. ***, **, * Significant at the 1%, 5%, and 10% level, respectively.

189

Appendix B

Table B1 Model Specifications for Over-indebtedness Model (Logit Model) Variable Subjective Objective Perception of having secure income (dummy) -0.321*** -0.325*** Expenditure to income ratio (level) 0.192** 2.259*** Expenditure to income ratio (squared term) -0.021*** -0.004*** Insufficient savings (dummy) 0.656*** 0.467*** Dependency ratio (level) 0.15 1.736*** Age (level) 0.038* 0.051*** Age (squared term) -0.0004** -0.0005*** Main source of fund (BAAC) 3.378*** 0.886*** Main source of fund (GHB and GSB) 2.405*** 0.129 Main source of fund (non-banks) 2.851*** 0.252* Main source of fund (cooperatives) 2.560*** 0.338 Main source of fund (village fund) 2.893*** 0.843*** Main source of fund (informal source) 3.836*** 0.578*** Credit constraint (in case of working purposes) 0.414*** 0.226** Mortgage loans to annual income ratio 0.291*** 0.124* Agricultural business loans to annual income ratio 0.298*** 0.723*** Non-agricultural business loans to annual income ratio 0.414*** 0.544*** Education loans to annual income ratio 0.846*** -0.744*** Personal consumption loans to annual income ratio 0.188*** 0.198*** Other purpose loans to annual income ratio 0.474 0.405* Financial literacy (interest rate) -0.055 -0.308*** Hyperbolic discounting behaviour 0.045 0.056* Having problems with other financial commitments (dummy: Y = 1, N = 0) 1.070*** 0.684*** Doing income-and-expenditure accounts -0.462*** -0.464** Sex (dummy: female = 1, male = 0) 0.146* -0.137 Expectation of household’s economic status (dummy: worse = 1, otherwise = 0) 0.962*** 0.403*** Expectation of government help for credit default (dummy: Y = 1, N = 0) -0.126 -0.031 Dummy variable for social class (using professional as a benchmark) Agricultural business 0.633*** 1.491*** Non-agricultural business 0.903*** 0.460* Worker 0.852*** 0.640*** Retired 0.501** 0.499** Dummy variable for region (using Bangkok as a benchmark) Central (exclude Bangkok) 0.079 1.929*** North 0.168 2.285*** North-east 0.359 2.394*** South 0.193 1.984*** Dummy variable for household head’s marital status (using single as a benchmark) Married (couple) 0.354* 1.377*** Married (widowed/ separated/ divorced) 0.426** 0.952*** Constant -7.023*** -10.716*** N 9430 9430 Wald Chi2 1353.17 1062.13 Pseudo R2 (McFadden’s) 0.3081 0.3154 Notes: Households with respondent, who is not the household head or household head’s spouse, are dropped. ***, **, * Significant at the 1%, 5%, and 10% level, respectively.

190