The financial crisis and household savings in South Africa : an econometric analysis

THE FINANCIAL CRISIS AND HOUSEHOLD SAVINGS IN SOUTH

AFRICA: AN ECONOMETRIC ANALYSIS

ltumeleng Pleasure Mongale

Thesis submitted for the requirements for the degree Philosophiae Doctor in Economics at the (Mafikeng Campus) North-West University

Supervisor: Co-supervisor: Mafikeng South Africa November 2012 Prof

J. Mukuddem-Petersen

DECLARATION

I declare that "The financial crisis and household savings in South Africa: An econometric analysis" is my own work, that it has not been submitted for any degree or examination in any other university, and that all the sources I have used or quoted have been indicated and acknowledged by complete references.

Full names ... Date ... .

Signed ... . Signature ... Date ... . Supervisor Signature ... Date ... . Co-Supervisor ii

ACKNOWLEDGEMENTS

I personally wish to acknowledge a substantial number of people who contributed to the

completion of this work. Firstly, a debt of gratitude is due to my supervisor, Professor Janine Mukuddem-Petersen and co-supervisor, Professor Mark A. Petersen for their

splendid guidance, suggestions, motivation and mentorship. I am particularly grateful for their informative feedbacks and constructive inputs during the writing of this thesis.

I acknowledge the emotional support provided by my ·family; my wife, Millicent, and children, who motivated me to study hard and also for the amount of time and space that they granted me to focus on my studies.

I am also grateful to all the members of the Financial Modeling and Optimization Research Group (FMORG), who provided valuable assistance in the form of comments and criticisms during presentations.

Furthermore, I am grateful to the NWU for providing me with funding during the duration of my studies.

I could not have achieved the completion of this dissertation without God's love, care

and grace.

PREFACE

One of the contributions made by the North-West University at Mafikeng (NWU-MC) to the activities of the financial economic community in South Africa has been the establishment of an active research group (FMORG) that has an interest in institutional

finance, modelling and economic crises.

Under the guidance of Pratts. Mark A. Petersen and Janine Mukuddem-Petersen, this

group has recently made valuable contributions to the existing knowledge about the modelling and optimization of financial institutions.

The work in this dissertation originated from our interest in Modeling, the 2008-2009 Financial crisis as well as household savings in South Africa. From the onset it became apparent that little work has been done on this topic although it has been identified as an area of potential growth.

Some of the outcomes of this project were collected in five research outputs which include one accepted chapter in a book entitled the "Economics of Debt", published by NOVA in New York, (see Appendix G) and one conference proceeding paper and 3 research articles that are submitted for possible publication.

ABSTRACT

The "global" financial crisis (GFC) emerged during 2008 and it was mainly triggered by the sub-prime mortgage crisis (SMC) in the United States of America. The main aims of this thesis is to conduct an econometric analysis of the financial crisis and household savings in South Africa and also to provide a rationale that will facilitate a policy attention on Domestic Resource Mobilisation (ORM) through household savings. The study uses quarterly time series data for the period 199401 to 2011 Q2 obtained on-line from the South African Reserve Bank (SARB). The research is based on the Keynesian saving function, which is a complement of the consumption function. The model will be estimated by using a cointegrating vector autoregressive (CVAR) framework, which allows for endogeneity of the regressors. To check robustness on the cointegration results, the study employs the second empirical technique based on Generalized Impulse Response Function (GIRF) analysis and Variance Decomposition. The regression equation of household savings is expressed as a function of household

disposable income, household debt to disposable income, real GDP, interest rate,

inflation rate and foreign savings.

The variables are tested for the presence of a unit root by the application of the

Augmented Dickey-Fuller (ADF), Phillips-Perron (PP) Kwiatkowski, Phillips, Schmidt

and Shin (KPSS) tests. The findings of the study are that all variables have unit roots. The cointegration model emphasises the presence of a long run equilibrium relationship

between dependent and independent variables. The CVAR reveals the short run of the

dynamic household savings model. Taking this into consideration, the study concludes

that household debt has a huge influence on the level of household savings.

The econometric analysis also revealed that household savings in South Africa actually improved during the period associated with the GFC. It could be postulated that South African households responded to their deteriorating financial situations by reducing their average spending and increasing their savings. Variance decomposition analysis revealed that 'own shocks' constitute the predominant source of variations in household

saving therefore household savings can be explained by the disturbances in

macroeconomic variables in the study.

The study recommends the promotion of household savings and economic growth in order to reduce the dependence of South Africa on foreign savings. ORM is therefore enhanced by a higher level of household savings, which can facilitate higher levels of investment and economic growth.

Keywords: Household savings, Cointegrated Vector Autoregression, Generalized Impulse Response Function, Diagnostic tests, South Africa

JEL Classification C32; E21; G01

GLOSSARY OF TERMS

Annuity is a term used to describe a type of contract by which financial institutions, such as insurance companies, agree to provide regular income for life. The word annuity comes from annual payments.

Debt income ratio is the ratio between a person's income and how much they pay monthly in instalments and revolving debt.

Gross domestic savings refers to the sum of public savings and private savings. Public saving is the difference between total revenue and current expenditure of the consolidated public sector.

Household savings is the difference between a household's disposable income and its consumption. The household saving rate is calculated by dividing household saving by

household disposable income.

National savings rate is an estimate of income that remains after the subtraction of consumption-related spending. It includes household, corporate, public services and government savings.

Shares refer to part ownership of a company that may be held by individuals or other companies.

"Saving" versus "savings." The saving means an increase in one's assets, an increase in net worth, whereas savings refers to one part of one's assets, usually deposits in savings accounts, or to all of one's assets. Saving is an activity occurring over time, a flow variable, whereas savings is something that exists at any one time, a stock variable. It is often contrasted with investment.

Stochastic process is where what happens is not exactly predictable, as it is affected by apparently random factors. It is often not known whether the apparent randomness is really so, or whether it is the result of the action of forces that are determinate, but so numerous that it is impracticable to model their effects.

ADF AIC BRICS CVAR

ORM

GOS

Augmented Dickey Fuller Akaike Information Creteria

Brazil, Russia, India, China, South Africa

Cointegrating Vector Autoregressive Domestic Resource Mobilisation Error Correction Model

Gross Domestic Savings.

Global Financial Crisis

Generalised Impulse Response Function Kwiatkowski, Phillips, Schmidt and Shin

Life-cycle hypothesis

Organization for Economic Corporation and Development South African Reserve Bank

South African Saving Institute Sub-prime mortgage crisis Vector Autoregression Variance Decomposition Vector Error Correction Model

TABLE OF CONTENTS

DECLARATION ... ii

ACKNOWLEDGEMENTS ... iii

PREFACE ... iv ABSTRACT ... v

GLOSSARY OF TERMS ... vii

LIST OF ACRONYMS ... viii TABLE OF CONTENTS ... ix LIST OF FIGURES ... xiii LIST OF TABLES ... xiv CHAPTER 1 ... 1

INTRODUCTION ... 1

1.1 Introduction I background I rationale ... 1

1.2 Problem statement. ... 5

1.3 Aims I objectives ... 6

1.4 Research questions and hypothesis ... 7

1.5 Significance of the study ... 7

1.6 Limitations I delimitations ... 8

1 . 7 Structure of study ... 8

CHAPTER 2 ... 9

LITERATURE REVIEW ... 9

2.1 Introduction ... 9

2.2 Theories of savings ... 11 2.2.1 The Keynesian theory ... 11

2.2.2 Life-cycle hypothesis ... 13

2.2.3 Permanent Income Hypothesis (PIH) ... 13 2.2.4 Endogenous growth model ... 14

2.3 Empirical literature ... 15

2.3.1 Determinants of household savings ... 15

2.3.1.1 Disposable income ... 16 2.3.1.2 Household debt ... 18 2.3.1.3 Interest rate ... 19 2.3.1.4 Inflation rate ... 21 2.3.1.5 Real GDP ... 21 2.3.1.6 Foreign savings ... 23

2.4 Impact of the GFC on developing economics ... 25

2.5 Summary ... 25

CHAPTER 3 ... 26

METHODOLOGY ... 26

3.1 Introduction ... 26

3.2 Econometric model ... 26

3.3 Description of variables I data ... 29

3.3.1 Definition and justification of variables ... 29

3.3.1.1 Household savings ... 29

3.3.1.2 Household disposable income ... 29

3.3.1.3 Household debt to disposable income ... 29

3.3.1.4 Real GDP ... 30

3.3.1.5 Interest rate or prime overdraft ... 30 3.3.1.6 Inflation rate ... 31 3.3.1.7 Foreign savings ... 31 3.4 Estimating the model ... 34

3.4.1 Unit root test ... 37

3.4.1.1 Augmented Dickey-Fuller (ADF) test.. ... 38

3.4.1.2 Phillips-Perron (PP) test ... 40

3.4.1.3 Kwiatkowski, Phillips, Schmidt and Shin (KPSS) test.. ... 41

3.4.2 Johansen cointegration test ... 42

3.4.3 Vector error correction model (VECM) ... 46

3.5 Diagnostic and stability tests ... 48

3.6 Dynamic analysis ... 55

3.6.1 Generalised Impulse Response Function (GIRF) ... 55

3.6.1 Variance decomposition ... 58

3. 7 Summary ... 59

CHAPTER 4 ... 61

EMPIRICAL RESULTS AND DISCUSSION ... 61

4.1 Introduction ... 61

4.2 Descriptive statistics ... 61

4.2 Analysis of results ... 63

4.2.1 Visual inspection ... 63

4.2.1.1 Household savings ... 63

4.2.1.2 Household disposable income ... 64

4.2.1.3 Household debt ... 66

4.2.2 Stationarity tests ... 70

4.2.3 The Johansen cointegration test.. ... 76

4.2.4 Vector Error Correction Model (VECM) ... 79

4.2.5 Diagnostic and stability tests ... 83

4.2.6 Generalised Impulse Response Function (GIRF) ... 88

4.2.7 Variance decomposition ... 91

4.3 Summary ... 93

CHAPTER 5 ... 95

CONCLUSIONS AND FUTURE DIRECTIONS ... 95

5.1 Introduction ... 95

5.2 Summary of findings ... 95

5.3 Conclusions ... 97

5.4 Recommendations and policy implications ... 100

REFERENCES ... 102

APPENDICES ... 112

Appendix A: Unit root tests ... 112

Appendix A1: ADF test ... 112

Appendix A2: Phillip-Perron (PP) Test.. ... 113

Appendix A3 : KPSS ... 114

Appendix B: Cointegration tests ... 116

Appendix C: Vector error correction test.. ... 120

Appendix D: VEC stability condition check ... 123

Appendix E: Diagnostic tests ... 124

Appendix E1: Breu sch-Godfrey serial correlation LM test.. ... 124

Appendix E2: Heteroskedasticity test: ARCH ... 125

Appendix E3: Heteroskedasticity test: White (no cross) ... 126

Appendix E4: Ramsey RESET test ... 127

Appendix F: Data ... 128

Appendix G: Research output ... 130

LIST OF FIGURES

Figure 2.1: Circular flows of income and output ... 12

Figure 2.2: Keynesian function ... 17

Figure 2.3: Household debt, saving and net wealth as a percentage of personal disposable income ... 19

Figure 4.1: Savings by households in South Africa ... 63

Figure 4.2: Log household disposable income ... 65

Figure 4.3: Differenced log household disposable income ... 66

Figure 4.5: Correlogram of household debt at first difference ... 68

Figure 4.6: Log inflation rate ... 69

Figure 4.7: Log foreign savings ... 69

Figure 4.8: Inverse roots of AR characteristic polynomial ... 82

Figure 4.9: Histogram of residuals ... 84

Figure 4.10: CUSUM test ... 87

Figure 4.11: CUSUM of squares ... 88

Figure 4.12: Generalised impulse fesponse functions ... 89

Figure 4.13: Actual, fitted and residual values ... 94

LIST OF TABLES

Table 4.1: Summary of the descriptive statistics of the variables ... 62

Table 4.2: ADF test for unit roots ... 71

Table 4.3: Phillips-Perron test for unit roots ... 73

Table 4.4: KPSS ... 74

Table 4.5: VAR lag order selection criteria ... 76

Table 4.6: Johansen test for cointegration ... 77

Table 4.7: The results of VECM estimation ... 79

Table 4.8: Breusch-Godfrey serial correlation LM test ... 85

Table 4.9: Heteroskedasticity test: ARCH ... 85

Table 4.10: Heteroskedasticity test: White (no cross) ... 85

Table 4.11: Ramsey RESET test ... 86

Table 4.12: Variance decomposition of LHSAV ... 92

CHAPTER 1

INTRODUCTION

1.1. Introduction I background I rationale 1.2. Problem statement

1.3. Aims and objectives

1.4. Research questions and hypothesis 1.5. Significance of the study

1.6. Limitations I delimitations 1.7. Structure of dissertation

1.1 Introduction I background I rationale

The "global" financial crisis (GFC) emerged during 2008 and it was mainly triggered

by the sub-prime mortgage crisis (SMC) in the United States of America. In this regard, Petersen, Senosi and Mukuddem-Petersen (2012) indicated that the SMC shook the foundations of the financial industry by causing the failure of many iconic Wall Street investment banks and prominent depository institutions. This crisis stymied credit extension to households and businesses, thereby creating credit crunches and, ultimately, a global recession. In addition, according to Masilela

(2009), the GFC is to a large extent a debt crisis. Particularly, the crisis can be traced to the collapse of the complex, yet opaque structured securitisation market in

the United States, with spill-over effects predominantly in the United Kingdom and other developed countries, mainly in Europe. This collapse was rooted in the area of mortgage backed securities. Cheap credit, borrowers with poor paying ability, a housing price bubble and weak regulation all contributed to the crisis. The aforementioned are some of the factors encapsulated by the IDIOM hypothesis 1 postulated by Petersen et al. (2010).

1

The hypothesis claims that the sub-prime mortgage crisis was largely caused by the intricacy and design of sub-prime agents, mortgage origination and securitisation that led to information (asymmetry, contagion, inefficiency and loss) problems, valuation opaqueness and ineffective risk mitigation).

The same sentiment is shared by McKinsey (2010), who concluded that after 2000

debt grew significantly in most mature economies. The borrowing increased

substantially in the case of households, especially through mortgages. The household debt level increased significantly in relation to the household disposable

income. This was ignored because the ratio of debt to assets appeared stable

before the crisis due to the rising house prices. The interesting aspect is that the

causes of this crisis are similar to those of the SMC. The common feature is that the

crises were preceded by long periods of rapid credit growth, low risk premiums, abundant availability of liquidity, soaring asset prices and the development of

bubbles in the real estate sector (European Commission, 2009).

The ensuing decrease in the world prices, combined with a drop in the world

demand, led to a fall in the production for most sectors. Consequently, the fall in

production brought about unemployment and a drop in household income (Chitiga,

Mabugu and Maisonnave, 2010). Regarding the latter, Naude (2009) reiterates that

there is now a widespread agreement that global economic reform should at least

include addressing global imbalances in savings. As Subbarao (2009) puts it:"what

is clearly beyond debate though is that this Great Recession of 2008/09 is going to

be deeper and the recovery longer than earlier thought".

The initial expectation was that African economies would be isolated from the contagion of the crisis. However, McCarthy (2009) argued that this was an unrealistic presumption and that the crisis will have serious fallout in Africa. In

particular, this sentiment was proved by Chitiga et al., (2010), who concluded that

the effects of the world economic crisis on South Africa are harsh even in the

moderate scenario. Within this context, the GFC has hit some of the key drivers of

growth, namely trade, investment, mining and manufacturing sectors and

approximately 959 000 jobs were lost in South Africa. While poor borrower quality

has been less severe in South Africa than in advanced economies, it is still a

significant contributor to the stress currently faced by consumers. Many South Africans lost their homes and vehicles in 2008. In fact, South Africa entered a

recession in 2009. One of the lessons from the crisis is that true long-term

sustainability at the macro- and household level cannot be attained through

investments based on accumulated savings. As expected, the US, the source of the global crisis, has the lowest saving rate of any major country, roughly 10% of GDP

(Masilela, 2009).

Compared to some of its 02 counterparts, South Africa's 15% saving ratio is low. The saving ratio of India is in the region of 30%. China has had by far the highest overall saving rate in the world since 2000. Its saving rate has increased even further since, to nearly 50% of GDP. Gross capital formation (investment) is also high in China, but because saving exceeds investment, it is running a net surplus, which translates into a current account surplus (Horioka & Wan, 2007). These rapidly growing economies have clearly benefited from a robust savings environment. As alluded to earlier, South Africa had a national savings rate of about 15%. This was before the GFC and it includes household, corporate, public services and government savings in South Africa. In essence, according to Mcivor (2009), the aforementioned national savings rate is low by international standards.

Savings as a percentage of GDP is 15%, versus 34% in other developing countries and 34% worldwide.

The South African Reserve Bank (SARB) (2010) indicates that the national savings rate deteriorated from a high of 17.5 % in the final quarter of 2002 to a low of 12.9%

in the third quarter of 2007 as consumption expenditure increased. During the period of the crisis, in the first half of 2008, the savings ratio of general government declined from 2.9% to a rate of dissaving3 of 1.2% in the first half of 2010. The contra-cyclical expansionary fiscal policy adopted by general government increased

government's recurrent expenditure, while the recessionary conditions adversely

affected income. The household savings ratio improved from 1.0% in the first half of 2008 to 1.7% in the second half of 2009, mainly reflecting a decline in household expenditure over this period. Strict lending criteria and uncertainties brought about by the financial crisis contributed to a more cautious approach to spending (SARB,

2010).

2

It is an acronym referring to a grouping of emerging economies, which include Brazil, Russia, India,

China and South Africa. 3

Decreasing net assets by spending beyond one's income. This may be done either by spending money taken from a bank balance or the proceeds of selling assets, or by incurring debts.

On the other hand, the period after the GFC is different. In the main, the savings ratio recovered from 14.2% in the first half of 2008 to 16.5% in the first half of 2010. The improvement of the national saving rate lowered South Africa's shortfall of savings to finance capital formation. The improvement in the first half of 2010 was due to the increased consumption expenditure associated with the easier monetary and fiscal policies, and an improvement in households' wealth as asset prices continued to rise. The improvement of the national savings rate during 2008 and 2009 and in the first half of 2010 was primarily due to an increase in the savings ratio of corporate business enterprises. The savings ratio of corporate business enterprises increased from 10.4% in the first half of 2008 to 16.1 % in the first half of 2010. The curtailment of dividend payments and cost containment, together with growth in domestic demand, contributed to the improvement of the savings rate (SARB, 2010).

In South Africa, the threat of an increasing level of inflation continues. This is likely to put pressure on the SARB to increase the interest rates. All these, together with increases in food and fuel prices contributed to increased risks to financial stability as corporate and household budgets became more strained. According to SARB (2008), the high level of household indebtedness continues to be a source of vulnerability to the financial system in the event of an adverse shock to the repayment capacity of households. As a percentage of financial assets, household debt decreased from 28.6% in the first quarter of 2008 to 27.6% in the second quarter. Debt servicing costs continued to increase in the first half of 2008, resulting in income gearing4 increasing from 11.3% in the first quarter of 2008 to 11.6% in the second quarter. Over the same period, capital gearing5 decreased slightly from 19.5% to 19%.

To date, various approaches have been used by researchers such as Balchin (2009) and Allen and Giovannetti (2010) to examine the impact or the effects of the GFC in Africa and Sub-Saharan Africa. However, only a handful of studies, such as Karolyi (2002) and Sheng and Tu (2000), used a quantitative approach to study the impact of the financial crisis on the economy. Despite an increased interest in the

4

Income gearing is the interest costs as a ratio of disposable income.

5

effects of the GFC in South Africa, there is little empirical research that has been conducted on this specific topic. According to my knowledge, there is no study that has focused on the impact of the financial crisis on a macroeconomic variable such

as savings in South Africa. This study appears to be the first of its kind to consider

the impact of GFC on household savings and also the first one to investigate this matter using the CVAR model to estimate the relationship between household

savings and its determinants. Furthermore, it is definitely the first to analyse this

problem by employing the GIRF and variance decomposition analyses.

Ultimately, this novel econometric study attempts to contribute to the literary body of

knowledge by exploring the impact of the GFC on household savings (from 1994 to

2011) in the South African economy as well as examining the imbalances in

savings and consumption. Due to the envisaged geographic impact of this crisis, it

was initially coined as the "global" financial crisis; however, it is now more

commonly being referred to as the financial crisis. For the purpose of this study both of these terms will be used interchangeably in the thesis.

1.2 Problem statement

One of the policy responses to the crisis raised at the United Nations Economic Commission for Africa (UNECA) (2009), is the improvement of domestic resource mobilisation (ORM). The use of domestic resources for development purposes is

becoming more and more important as access to foreign resources becomes

increasingly difficult. The main problem with ORM in South Africa has been that not

enough savings are being generated to facilitate the required investment. With its

large structural savings/investment gap, South Africa depends on foreign savings to

support investment and growth (Bureau of Economic and Business Affairs U.S.

Department of State, 2002). The types of savings available do not easily make

financial intermediation possible (Aryeetey, 2004). According to Allen and

Giovannetti (2010), improved levels of household savings will contribute to

domestic resources, which will act as a mitigating factor against financial crises.

The right response to the crisis would be to mobilise domestic resources, although

this will require that functional institutions should offset the potential trade-offs

1 . 3 Aims I objectives

Economic growth is among the most important factors affecting the lives of people in a country. Therefore, given the close relationship between savings and growth, the analysis of savings behaviour becomes naturally important for policy analysis (Agrawal, Sahoo & Dash, 2008).

1.3.1 The aims of are as follows:

1.3.2.1 to identify the parameters related to the 2008/2009 financial crisis and to analyse how South African household savings levels are affected before, during and after the crisis;

1.3.2.2 to examine the main determinants that affect the level of household savings in South Africa;

1.3.2.3 to estimate an econometric model for household savings in South Africa for the period 1994 to 2010 in order to analyse the relationship that exists between household savings and other macroeconomic determinants, as well as other related variables;

1.3.2.4 to consider the impact of various shocks to household savings in South Africa and to overcome the shortcomings of other papers in this regard; and

1.3.2.5 to provide a rationale that will facilitate a policy attention on ORM through household savings.

1.3.2 The objectives of the study include:

1.3.2.1 The critical screening of literature related to the 2008-2009 financial crisis in developing countries and the identification of those parameters related to household savings. A graphical analysis of possible trends in the South African level of household savings in relation to the periods 'before', 'during' and 'after' the aforementioned crisis.

1.3.2.2 Identify the determinants that affect the household savings in South Africa (the results of related studies and economic theory will be used to

guide this selection).

1.3.2.3 The application of multivariate CVAR methodology will be employed in

identifying the determinants of household savings.

1.3.2.4 To use the (GIRF) to evaluate the impact of the 2008-2009 financial

crisis on the level of South African household savings. Furthermore, to use the variance decomposition analysis to evaluate which variables mostly contribute to an explanation of the shocks in household savings.

1.3.2.5 To formulate a ORM rationale using household savings.

1.4 Research questions and hypothesis

1.4.2 Research questions

1.4.2.1 How is the household savings trend affected before, during and after the

GFC?

1.4.2.2 What are the main determinants of South Africa's household savings

rates?

1.4.2.3 What are the macroeconomic implications of household savings in South

Africa?

1.4.2.4 Which variables mostly contribute to an explanation of the shocks in the

household savings model?

1.4.2.5 Can ORM be enhanced by increased household savings?

1.4.3 Hypothesis

The GFC will have a negative effect on the level of household savings in South

Africa. The impact will result in hampered access to foreign resources.

1.5 Significance of the study

This study is important because the frequency and the severe consequences of

economic crises are intolerable. The search for something to remedy this situation

2006). According to Aron and Muellbaue (2000), it is important to understand personal saving behaviour in order to formulate policies to raise the domestic saving rate in line with the needs of economic growth.

This research will therefore be helpful to policy makers to consider potential responses to mitigate the impact of the crisis by advocating for policies on the importance of pre-emptive interventions to encourage household savings rather than household lending in order to buy assets such as houses. Understanding the nature of household savings is important and critical in designing policies to promote savings and investment in a developing economy such as South Africa.

The results of the study will be of importance to investors in understanding how South African households are dealing with the GFC. Furthermore, it will be a useful tool to the South African Government agencies and other private regulation companies. Specifically, they will be able to use this information as a guide on how the GFC is affecting domestic resources for development. The study also serves as a wakeup call for the development of domestic resources mobilisation appeal in order to serve as a mitigating tool for future financial crises.

1.6 Limitations I delimitations

There are some limitations experienced in relation to the availability of some data. In this regard, this study was forced to exclude an important variable, namely unemployment, due to the unavailability of quarterly data for the period under consideration. There were no delimitations.

1. 7 Structure of study

The study will be structured as follows: Chapter 1 is the introductory chapter. Chapter 2 focuses on the literature review, where previous studies that focused on the GFC and other related topics are discussed. In Chapter 3, methodology that is undertaken is presented. The methodology consists of the presentation of the econometric models, variables and data description. Chapter 4 presents the empirical results. It is mainly the presentation of the economic and statistical outputs. Chapter 5 presents the findings and the conclusion of the investigation.

CHAPTER 2

LITERATURE REVIEW

2.1 Introduction

2.2 Theories of savings 2.3 Empirical literature

2.3.1 Determinants of household savings

2.4 Impact of the GFC on developing economies 2.5 Summary

2.1 Introduction

This chapter reviews the theoretical and empirical literature concerning the behaviour of household savings in different countries. The purpose of this section is to explore the literature that seeks to address the effects of the GFC on the South African economy. The aim is to review several studies as guide to the choice of the appropriate variables, model and method used in this study. This is done by highlighting the arguments and findings by different authors and what their implications to this study will be. Several theories, such as the Keynesian model, Life Cycle Hypothesis (LCH) and the Endogenous growth model will be discussed in order to build the theoretical perspectives of the study. The research will be based on various strands of literature about the crisis and household savings.

Chitiga et al. (2010) studied the effects of GFC on the economy and households. They used the dynamic Poverty and Economic Policy (PEP 1-t) standard model to evaluate the impacts of the world economic crisis on South Africa. In contrast, the present study will use the CVAR model, GIRF and variance decomposition to study the effects of the GFC on household savings in South Africa. As far as the situation of household savings before, during and after the GFC is concerned, see Masi le la

(2009), SARB (2008) and SARB (2010). For this will be helpful in the sense that it will help to direct the analysis for the effect of GFC on household savings.

Kaufmann et al. (2010) used Johansen (1998) and Juselius' (2006) cointegration

vector autoregressive (CVAR) model to test the hypothesis that increasing

expenditures on energy by households cause the consumer to fall behind on

mortgage payments and thereby increase mortgage delinquency rates. Similarly,

Kulshreshtha and Parikh (2000) used the CVAR model to estimate the relationship between the long run structural relationships of coal consuming sectors in India. In the present study, the same approach is used to test the hypothesis that GFC will have a negative impact on the level of household savings in South Africa. This will be determined by investigating the long and short run relationships among the chosen variables.

The empirical framework also follows a similar approach, which is used by Wang, Yang and Bessler (2003) to examine long run relationships and short run dynamic causal linkages among the five largest emerging African stock markets and the US market, with particular attention to the 1997-1998 global emerging market crises.

They applied GIRF analysis, which was developed by Koop, Pesaran and Potter

(1996), which is invariant to the ordering of the variables in the VAR model. The

same approach was also used by Pesaran and Shin (1998) in an analysis of

cointegrated VAR models. GIRF based on estimated VAR provides further insight

into how innovations in a particular market within the system affect other markets through dynamic interactions among markets. This study will employ GIRF to provide insights into the speed and the direction of impact by GFC on household

savings, with particular attention to the 2008/2009 GFC. The combination of CVAR

and the GIRF framework followed in this study was also used by Hurley (2010) to examine the dynamic relationships between East Asian holdings and the US securities and the US short term and long term interest rate. In addition, this study will employ variance decomposition, which will help to indicate the amount of information each variable contributes to the other variables in the Vector Error Correction Model (VECM) (Kar & Manda I, 2010).

The following writers are also considered in the literature review because they highlight the impact of the GFC on South Africa and other developing countries.

Authors like Aryeetey (2004) and Culpeper and Bhushan (2008) argue that the GFC and the subsequent recession are generally expected to make access to foreign resources more difficult. Furthermore, it is postulated that this financial crisis is also affecting the mobilisation of domestic resources. In the main, such resource mobilisation can come from both the public and the private sector. The public sector

does this through taxation and other forms of public revenue generation. The private sector mobilises resources through household and business savings,

working through financial intermediaries to convert these into productive assets.

This study shares the same sentiment with these authors that Domestic Resource Mobilisation (ORM 6) is important, because it can help to facilitate higher levels of investment and economic growth. Truman (2009) maintains that a country will be better off if it has preserved room to manoeuvre and to respond to external shocks through the use of domestic policy instruments, primarily fiscal and monetary policies.

2.2 Theories of savings

Saving fundamentally is about choosing between current and future consumption.

Savings theories traditionally predict that the current consumption is related not to the current income, but to a longer term estimate of income. Many financial

institutions, particularly in developed countries, incorporate these theories into their product designs (Ashraf, Gons, Karlan & Yin, 2003). The economic theories in relation to savings are further discussed under the following categories. Household savings literature is in most cases based on the following major theories.

2.2.1 The Keynesian theory

Just like the Keynesian economics, which was developed against the background of the world depression of the 1930s, this study is inspired by the GFC of 2008/2009. Keynes' theory provided the basis for economic policies to combat unemployment by stimulating aggregate demand (Froyen, 2005). The importance of savings in economic growth is captured in the simple Keynesian model: Conditions

6

Refers to the generation of. savings from domestic resources and their allocation to socially productive investments.

for equilibrium output. The model indicates that with the national product Y

measuring national income, we can write:

Y:=C+S+T

2.1

According to Froyen (2005), equation 2.1 is an accounting definition or identity

stating that national income, all of which is assumed to be paid to households in

return for factor services, is either consumed (C), paid out in taxes (T) or saved (S).

National product can also be defined by equating it to consumption plus realised

investment Ur) plus government spending, which can be written as:

Y := C +Ir+ G

2.2

The condition of equilibrium is therefore obtained by combining equations 2.1 and

2.2 as follows:

C+S+T:=Y=C+I+G

2.3

or

S+T=l+G 2.4

The Keynesian model can also be explained by the circular flow of income and output and it is illustrated in Figure 2.1 as follows:

Figure 2.1: Circular flows of income and output

ational Income (Y)

Con umption (C)

a ing

_ _ ___;;;....;...;...-.i Financial Markets Inve tment l)

_ _ _ _

T

_

ax.

_

e

_..;..

T

....;..

)

_ - i

G

S

Go

emment

pending

(G)

· overnment ect r

Figure 2.1 illustrates that saving is a very important component of the circular flows. The saving flows into financial markets, which means that part of the income that is

saved is held in the form of some financial assets, such as bank deposits, currency,

etc.

2.2.2 Life-cycle hypothesis

The Life-cycle hypothesis (LCH) postulates that individuals' consumption and saving decisions during a given year are the result of a planning process that considers their lifetime economic circumstances. Modigliani (1966) predicts that individuals hold their consumption constant over their lifetime. They save during their working years and withdraw their savings during retirement. This implies that the amount an individual saves each year depends not only on the income they received that year, but also on income that they expect in the future and the income they received in the past. Modigliani's model also showed that aggregate saving depends primarily upon the rate of growth of the economy and demographic

variables, such as age groups, birth rates, dependency ratio and the life

expectation. Now, for the purpose of this study, mainly the economic or financial variables such as interest rate, inflation rate, etc will be considered.

2.2.3 Permanent Income Hypothesis (PIH)

In terms of the permanent income hypothesis, Friedman (1954) argues that consumption is proportional to a consumer's estimate of permanent income. More recent theories use 'hyperbolic discounting' and 'mental accounting' to model this view. Hyperbolic discounting models suggest that individuals would improve their

welfare by seeking mechanisms that commit them to save in the future. Mental

accounting theories suggest that individuals often behave as if money is not

fungible7 and hence savings levels can be affected by a mere framing of decisions.

7

Friedman differentiates between permanent and transitory components of income as determinants of savings. This hypothesis is basically defined in terms of long-time income expectation over a planning period and steady rate of consumption over a lifetime given the present level of wealth. Muradoglu and Taskin (1996) then argue that this transitory income is the difference between actual and permanent income, since individuals are assumed not to consume out of this income category, marginal propensity (tendency) to save on transitory will be unity.

2.2.4 Endogenous growth model

Endogenous growth theory is considered in this study because, according to Culpeper and Bhushan (2008), endogenous theories of growth are extremely pertinent for ORM. There is a clear association between the endogeneity of such growth models and the endogeneity implied by ORM. Endogenous theories of growth reinforce the notion implied by the Harrod-Demar and Solow's neoclassical growth models8 that increased savings and investment enhance growth (even,

according to the latter, only if in the short term). The theory is considered to be relevant to the concept of ORM as it argues that economic growth is generated from within the system as a direct result of internal processes. The theory also notes that the enhancement of a nation's human capital will lead to economic growth by means of the development of new forms of technology and efficient and effective means of production.

At the same time, ORM can reduce the impact of movements of capital especially of short term 'hot money' during the periods of financial crises. According to Stiglitz (2006), when crises occur, they unleash a wave of capital outflows and increased uncertainty in international transactions. Mitigating these reversals or, at least, dealing with them in an orderly way would greatly speed up the resolution of crises and resumption of economic growth. Within this context, the study considers it an important policy to deal with the economic crises mainly at the domestic level. Specifically, it is important through the results of this study to encourage the

8

According to lnvestopedia.com, it is an economic theory that outlines how a steady economic growth rate will be accomplished with the proper amounts of the three driving forces, namely labour,

adoption of sound policies by persuading investors that the adoption of some restraints on capital flows is not necessarily a sign of unfriendliness to investment. 2.3 Empirical literature

The choice of the explanatory variables used in the study is guided by theory and the literature.

2.3.1 Determinants of household savings

In order to have an insight into the determinants of household savings, function-related saving studies are reviewed in relation to the proposed variables of the model. Several studies present different views about the determinants of savings,

for example, Lewis (2002) indicates that as government saves more, private savers tend to save less (limited Ricardian equivalence9), so that the recent reduction in

government dissaving has been associated with lower private savings. Household savings are responsive to changes in corporate savings behaviour, so that as firms save more, households tend to save less and the extensive financial liberalisation that has occurred in South Africa appears to be negatively related to the private and household savings rate. On the other hand, Kazima (2004) maintains that the aggregate savings of a country have three components, namely private, corporate and the public sector savings. Each of these components has its own specific determinants. These determinants further add to the complexity of analysis. In different countries, the three components assume different weights, implying that aggregative techniques for analysing saving patterns of societies at different stages of economic development may not provide sufficient insight into the saving behaviour across countries.

Various government actions can have a bearing on savings. Among these, the effect of fiscal policy has especially been the centre of debate. Theoretical views on

9

Ricardian equivalence asserts that, to the extent that individuals are rational and far-sighted, they will realise that a permanent rise in government spending today or an increase in government

dissaving must be paid for either now or later. They will therefore increase saving by an equal

this relationship span a broad range of the neo-classical version of the lifecycle model and assert that a decline in government saving will tend to raise consumption

and discourage saving by shifting the tax burden from present to future generations,

and predict that a decline in government saving will cause a decline in national

saving (Ozcan et al., 2003). Some of the proposed variables of the model are discussed as follows based on the empirical literature.

Based on the arguments raised by the empirical literature, the Keynesian saving

function will be used as the basis for the formulation of the model of this study. The following variables will also be included as other determinants of household savings in the South African economy:

2.3.1.1 Disposable income

According to Miller and VanHoose (2001), in the Keynesian theory of household

saving, a key determinant of households' annual saving flow is their disposable

income. The basic idea is that as disposable income rises, households can

increase their saving. This notion is captured by the following saving function:

S=-a+(l-b)Yv

2.5

This function captures the saving-income relation, the flipside of the

consumption-income relation that forms one of the key building blocks for Keynesian economics.

The two key parameters of the saving function are the intercept term (-a), which

indicates autonomous saving, and the slope, which is the marginal propensity to

save (1 - b) and indicates induced saving.

Figure 2.2: Keynesian saving function

s

s

lope

=

=

-a

+

y

-a

I

Disposable

Income

Source: Fro yen (2005) page 103

The saving function shows the level of saving (S) at each level of disposable income (Yv). The slope of the saving function is the marginal propensity to save (MPS)10 (1 - b), the increase in saving per unit increases in disposable income. The intercept for the saving function (-a) is the (negative) level of saving at a level of disposable income. The Keynesian saving function, which defines savings as a linear function of income, will be used as the basis for the formulation of the model of this study. More variables will be included as other determinants of household savings in the South African economy.

The above theory is also supported by Park and Kwanho (2009), who indicate that the empirical literature using cross-country data finds two economic variables, namely the level of per capita income and its growth rate to be particularly significant determinants of saving rates. For example, a 1 % point increase in per capita income growth raises the national saving rate by 1 % in industrialised countries and 0.5% in developing countries. Richer countries tend to save more than poorer countries and faster-growing countries tend to save more than slower-growing countries. The expected signs are therefore positive for both variables. Intuitively, this makes sense since the average household in poor countries will tend

10

The additional saving caused by an increase in disposable income; the change in saving divided by the

to save less of its income than the average household in rich countries, due to lower consumption levels and a higher marginal utility of consumption. Furthermore, economic growth boosts the income of the working-age population, which tends to save, relative to the population of the retirees, which tends to dissave, pushing up aggregate saving in the process.

Similarly, the study investigating the determinants of household saving in China, by

Horioka and Wan (2007), followed the Life-Cycle Hypothesis (LCH), which

postulates that the household saving rate will be the function of the growth rate of per capita income. Therefore, the household disposable income will be used as one of the independent variables. This is in line with the Keynesian saving function.

2.3.1.2 Household debt

Prinsloo (2002) maintains that there is an inverse relationship between an increase in the utilisation of consumer credit (household debt in terms of this study) and

household savings over time. This relationship can be demonstrated by the

following graphical illustration. It is clear from the illustrations in Figure 2.3 that the deterioration in household savings ratios over the years coincided with the greater use of credit by the households. Even though net wealth is not one of the variables, it shows that it has a major influence on decision-making between household debt and savings. This is somehow in line with both the lifecycle and income hypotheses discussed earlier, which indicated that individuals maximise consumption over time subject to the constraints of expected income and wealth.

Figure 2.3: Household debt, saving and net wealth as a percentage of personal disposable income Percent 70 300 60 50 250 40 30 200 20 10 0 150 1975 1980 1985 1990 1995 2001 - Househok:i debt

- Ne eatth (rig -hand scale) - Household saving

Source: Prinsloo (2002), page 73

Similar results were also achieved by Chakrabarti et al. (2011 ), who analysed changes in household debt and savings during the 2007 recession in New York. They realised that in response to their deteriorated financial situations, households reduced their average spending and increased their savings. The higher saving rate appears to reflect a considerable decline in household debt, as households paid down mortgage debt. This suggests that rebuilding net wealth is an important driver of households in making decisions between increasing debt and lowering saving or vice versa.

2.3.1.3 Interest rate

According to Gutierrez and Solimano (2007), the real interest rate is the equilibrating variable between the supply of loans (savings) and the demand of

loans for productive purposes (investment). An investment boom is created when banks or the monetary policy keeps the interest rate below the 'natural rate'11.

Gutierrez and Solimano (2007) go further by indicating that the association between interest rates and savings is also ambiguous theoretically (income and substitution effects may work in opposite directions). The income effect produced by higher interest rates may be positive or negative depending on whether the saver is a net wealth holder or a net debtor. The positive income effect of an increase in interest rates for a net wealth-holder may run in an opposite direction than the substitution effect that induces a cut in current consumption (substituting for future consumption).

It is clear from this discussion that empirical evidence of the effects of interest rates on savings has proven to be inconclusive. In certain cases, some authors have explored the sensitivity of savings to the rate of interest as a function of income levels. At lower income levels, people cannot smooth consumption over time and at higher income levels it is possible to save and dissave. In terms of this evidence, the inter-temporal elasticity of substitution between present and current consumption varies with the level of wealth.

Lastly, Mikesell and Zinser (1973) point out that interest rates are more significant in determining the channels into which savings will flow in the developed and developing countries than in altering saving propensities. According to these authors, this is mainly caused by the fact that the relationship between interest rates and aggregate saving involves a number of complex theoretical and econometric problems, such as separating income and substitution effects of interest changes, quantifying the role of expectations and planning horizons in saving decisions, and solving a difficult econometric identification problem.

Finally, according to Horioka and Wan (2007), the real interest rate is included to test for the impact of financial variables, and its coefficient is expected to be positive if the substitution effect is more than the income effect.

11

It is a concept developed by the Swedish economist Knut Wicksell, which refers to the interest rate that equilibrates the demand for loans (investment) with the supply of funds (savings).

2.3.1.4 Inflation rate

As far as inflation is concerned, Muraduglu and Taskin (1996) raise two contradictory arguments. They posit that inflationary expectations may encourage expenditures on durable goods at the expense of savings and, on the contrary, it decreases the real value of financial wealth fixed in nominal terms. This will in a way force households who try to restore their wealth-income position to increase their savings. Generally, empirical literature provides ample evidence that supports both views. That being the case, this variable is still very important for this study, because the evidence comes from studies that are based on Asian countries and the US, which are also slightly outdated, and none from African countries.

The inflation rate is included as a proxy for the price uncertainty and macroeconomic stability. The aforementioned reasons for the inclusion of the said independent variables will also be true for this study. Furthermore, a similar choice of independent variables, that is, interest rate and inflation rate, was selected by

Modigliani and Cao (2004) in the Chinese saving puzzle and the lifecycle

hypothesis.

2.3.1.5 Real GDP

Real GDP is chosen as one of the regressors in this study because many authors have been interested in the relationship between savings and economic growth. Mason (1988) indicates that savings contribute to economic growth by freeing up resources that can be employed to raise the productive capacity of the economy by increasing the amount of capital equipment. At same time, freeing up resources does not guarantee that they will be usefully applied. This is based on the assumption that not all investment decisions are good. Some will contribute nothing

to increase productive capacity; however, in general, additional investment

contributes to higher output per worker. If investment opportunities are poor relative to those available in another country, residents may choose to invest their savings in foreign, rather than domestic ventures.

The same sentiment is echoed by Kazmi (2004), who indicates that savings provide the most important economic link between the past, the present and the future of a country. He emphasises that the stock of savings of a nation sets a limit on the level of its gross investment and its growth rate. As a result, a low rate of savings, if maintained over a longer period of time, entraps an economy in a circle of low investment, low growth, low productivity and low real per capita income. A higher rate of savings is therefore a prerequisite for a country to get out of a low-saving

-low-growth equilibrium. Based on this, this study advocates the endogenous

theories of growth for South Africa by reinforcing the notion that increased household savings and investment enhance growth.

High savings and high investment rates have long been a defining structural characteristic of East Asian economies. Saving and economic growth served as key ingredients in the East Asian miracle that transformed eight market-based regional economies, namely China, Indonesia, Japan, the Republic of Korea, Malaysia,

Singapore, Taipei and Thailand between 1960 and 1990. While other ingredients such as sound macroeconomic policies, limited price distortions, human capital accumulation and so on contributed to the economic miracles, high saving and investment rates lay at the heart of the miracle. Furthermore, there were tangible,

measurable differences between the region and other parts of the developing world (Park & Kwanho, 2009).

Contrary to the above, a study done by Shan and Sun (1998) finds that Australia's low levels of saving are not responsible for the excessive external liabilities in Australia. This finding adds some empirical support to the argument in the literature that domestic investment is independent of domestic saving and hence increasing domestic saving alone is not a necessary and sufficient condition for increased

domestic investment and future economic growth.

Even though there is evidence of a strong and positive association between saving ratios and real per capita growth, the problem of economic development is not as simple as raising savings rates. There is no clear consensus on how to interpret such a relationship. The problem is the direction of causality. On the one hand,

On the other hand, there are supporters of growth to saving models, namely Modigliani (1970), Harrod-Damar, Carroll and Weil (1994).

Gutierrez and Solimano (2007) cast the controversy of causality of savings and economics in terms of two leading theoretical perspectives: the 'Marx-Schumpeter-Keynes view' versus the 'Mill-Marshall-Solow view'. The first view maintains that investment and savings are the two variables that drive output growth. Within this context, savings adjust passively to meet the level of investment required to hold macroeconomic equilibrium and deliver a certain growth rate of output. According to this view, growth leads to savings. In contrast, in terms of the Mill-Marshall-Solow approach, that channel of causality is reversed as it assumes that all savings are automatically invested and translated into output growth under wage price flexibility and full employment. As a result, in terms of the Mill-Marshall-Solow approach, savings lead to economic growth. The two schools deliver alternative lines of causality between savings, investment, innovation and growth.

2.3.1.6 Foreign savings

Another important variable to be considered in this study is foreign savings. It is deemed important because the wish is that it will give the effects of capital inflows on the South African economy. Taking the debate further, Gutierrez and Solimano (2007) maintain that empirically, national savings and growth are positively associated, especially in the case of developing countries. In terms of causality, the research on the determinants of savings has generally considered growth as a determinant of national savings, suggesting that the causality runs from growth to national savings. Subsequently, a typical regression is one in which national savings is the dependent variable of the regression and GDP growth is a right-hand side explanatory variable. Evidence of the association of GDP growth and foreign savings is mixed. There are episodes of high growth with relatively low levels of foreign saving rates, which are found in some East Asian economies. On the other hand, there are also episodes of low growth and high foreign savings, which are found in low income countries in Africa and Latin America that receive sizeable levels of foreign aid. According to Muraduglu and Taskin (1996), the access to foreign borrowing is expected to supplement domestic savings and fill the gap

between domestic investment and national savings. Based on this, the capital inflows are therefore expected to reduce household savings.

As far as this study is concerned, the regression is envisaged to have household savings as the dependent variable, GDP and foreign savings on the right-hand side

as part of other explanatory variables. This might help us to determine the evidence

on their association in the case of the South African economy, since the episodes of low growth and high foreign savings have been found in low income countries both in Africa and Latin America that receive sizeable levels of foreign aid.

Finally, another study that is of great interest to this study, is an analysis of the

savings behaviour in five South Asian countries, namely India, Pakistan,

Bangladesh, Sri Lanka and Nepal (Agrawal, Sahoo and Dash, 2009). Even though some of the variables used are not relvent, this analysis provides important policy

insights for increasing domestic savings rates in these countries. Their econometric

analysis shows that the main factors positively affecting total savings rates in these countries are income per capita or its growth rate and access to banking facilities. They also find that the dependency ratio and availability of foreign savings have a statistically significant negative effect on savings. Therefore, the recent increase in savings rates in South Asia is largely explained by the increasing per capita income or growth, declining dependency rates (fewer children per couple) and improved availability of banking facilities.

The changes in the real interest rate have only a small effect on the savings rate and even the direction of change is unclear. Changes in real interest rate were found to be positive but negligible in Bangladesh and Nepal and negative in India, Pakistan and Sri Lanka. These results suggest that trying to influence the savings rates by manipulating interest rates is not likely to be a practical policy option in these countries, as interest rate changes have only a minor impact on savings rates. However, a greater use of foreign savings can reduce savings and do need to be controlled, especially if it is due to increases in imports of consumption goods (Agrawal, et al, 2009).

2.4 Impact of the GFC on developing economics

According to Lin (2008), developing countries that are able to gain access to capital will pay higher interest rates, because of the flight to safety and greater risk aversion of lenders. As noted above, the global slowdown will reduce demand for commodities and manufactured goods, cutting into export earnings. And as labour markets slacken, foreign workers are likely to suffer disproportionate impacts on their earnings, which will reduce remittances. About half of all developing countries have been running current account deficits of 5% of GDP or more, and in some cases the deficits are around 10%. These economies will be highly vulnerable to swings in these various sources in external financing.

2.5 Summary

A selective theoretical and empirical literature review has been carried out to highlight some of the theories concerning household savings and the justification of the choice of the variables that will be used to estimate the CVAR model (see Chapter 3). Even though the review is not exhaustive, it nonetheless summarises some of the major findings in the field. The next chapter is the methodology, which outline both the formulation of the model and methodology of the study.

CHAPTER 3

METHODOLOGY

3.1 Introduction to the methodology

3.2 Econometric model

3.3 Description of variables I data

3.4 Estimating the model

3.5 Diagnostic and stability tests

3.6 Dynamic analysis

3.7 Summary

3.1 Introduction

This chapter first presents the econometric model, definitions and justification of the variables that affect savings. The study then, introduces the estimation technique (CVAR) that will be employed and all the tests that will be performed to examine the properties of the variables. In particular, the study uses the time series data to estimate the household savings function of the South African economy. In order to achieve the stable dynamic household savings model, several tests will be introduced which will help to verify the nature of the model.

3.2 Econometric model

The econometric model of this study is based on the Keynesian model. Just like the Keynesian economics, which was developed against the background of the world depression of the 1930s, this study is motivated by the GFC of 2008/2009. This study investigates the effects of the GFC on household savings in the South African

economy.

Keynes developed a mathematical relation between saving and income by the household sector. According to Miller and Van Hoose (2001 ), in the Keynesian