The relationship between financial development and household welfare case study in five asian countries

To detect the nexus of financial development and household welfare, the Pedroni cointegration test is run to find out the long-run relationship between financial development and househol

VIETNAM-NETHERLANDS PROGRAMME FOR MASTER IN

PHAN THI KHANH VAN

This paper was submitted in partial fulfillment of the requirements for

Master’s degree in Development Economics

Ho Chi Minh City, July 2013

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM-NETHERLANDS PROGRAMME FOR MASTER IN

Dr DUONG NHU HUNG

Thispaper was submitted in partial fulfillment of the requirements for

Master’s degree in Development Economics

ACKNOWLEDGEMENT iii

ABSTRACT iv

ABBREVIATION v

LIST OF FIGURES vi

LIST OF TABLES vi

CHAPTER I: INTRODUCTION 1

1 Problem statement 1

2 Research objectives 3

3 Research questions 4

4 Justification of the study 4

5 Scope of the study 4

6 Structure of the study 4

CHAPTER II: LITERATURE REVIEW 5

1 Definitions of key concepts 5

1.1 Financial development 5

1.2 Household welfare and Poverty 5

2 Theoretical literature 6

2.1 Direct relationship 7

2.2 Indirect relationship 8

3 Empirical studies 9

CHAPTER III: ECONOMETRICS REVIEW 12

1 Stochastic Process, Stationarity and Random Walks 12

2 Unit Root Test 13

3 Cointegration 13

4 Granger Causality Test 14

5 Panel Unit Root Test 15

6 Panel Cointegration 16

7 Instrumental Variables Regression (IV) 18

8 Generalized method of moments (GMM) 19

CHAPTER IV: DATA AND RESEARCH METHODOLOGY 22

1 Data 22

2 Research methodology 23

CHAPTER V: ANALYSIS RESULTS 26

1 Data descriptions 26

2 Empirical results 31

CHAPTER VI: CONCLUSIONS AND POLICY IMPLICATIONS 36

1 Conclusions 36

2 Policy implications 37

3 Limitations and directions for further studies 38

3.1 Limitations 38

3.2 Directions for further studies 39

REFERENCES 40 APPENDICES a

Appendix 1: Description of FD and PR variables (1998-2011) a Appendix 2: Panel Unit Root Test of variables b Appendix 3: Pedroni Cointegration Test j Appendix 4: GMM m

me theoretical advice and introduced me to Dr Le Van Chon for intensive support

Moreover, I would like to express my gratitude to Dr Le Van Chon and Dr Phung Thanh Binh who supported my study and my mind when I lost inspiration during research period Not only the teaching staff but also the class MDE 17 classmates are those I really appreciate Thanks to their care, their understanding and their sharing, I know exactly what I should do

The more importantly, I would like to express my appreciation to my direct supervisor

Dr Duong Nhu Hung He is a very kind teacher who always cares me and encourages me during the research period He is kind to me with his scientific guidance, soft but invaluable advice till the final stage of the study

Last but not least, during the time doing this study, I encountered both mental and fiscal problems At that time, my family, especially my mother, my grandmother and my aunts who always advise me to try my best and give me spiritual assistance I also send my sincere thanks

to my husband who is always with me I am proud of his patience and his sympathy He has given me a chance to concentrate on my studying instead of housework

ABSTRACT

The study presents the empirical result of the relationship between financial development and household welfare, which has been the hotly debated issue recently To detect the nexus of financial development and household welfare, the Pedroni cointegration test is run

to find out the long-run relationship between financial development and household welfare In empirical study, it is affirmed that there exists the long-run relationship between financial development and household welfare However, the impact of financial development on household welfare cannot be shown through Pedroni cointegration test Thus, 2SLS GMM is deployed to identify the impact of financial development on household welfare

Keywords: financial development, household welfare, cointegration, two stage least

squares,

2SLS: Two Stage Least Squares

ADF: Augmented Dickey Fuller

ADRL: Autoregressive Distributed Lag Model

AIC: Akaike information criterion

AR: Auto-regressive

DCBS: Domestic credit provided by banking sector as a percentage of GDP

DCP/GDP: Domestic credit to the private sector as a ratio of gross domestic product DCPS: Domestic credit to the private sector as a percentage of GDP

DF: Dickey Fuller

DMBA: Domestic money bank assets

EG: Economic growth

FD: Financial development

GDP: Gross Domestic Product

GMM: Generalized method of moments

HLSS: Household Living Standards Survey

IMF: International Monetary Fund

IV: Instrumental Variable

M2/GDP: money and quasi money as percentage of GDP

M3: the broadest definition of money

OECD: Organization for Economic Co-operation and Development

OLS: Ordinary Least Square

PP: Phillips-Perron

PR: Poverty reduction

SBC: Schwarz’s Bayesian criterion

SME: Small, medium-sized enterprise

VAR: Vector Auto-regressive

VECM: Vector Error Correction Model

WB: World Bank

WDI: World Development Indicator

WEF: World Economic Forum

LIST OF FIGURES

Figure 1: Money and quasi money (M2) as percentage of GDP 2

Figure 2: Household per capita consumption (constant 2000 US$) 3

Figure 3: Financial Sector Development and Poverty Reduction 6

Figure 4: Line graph of proxies of FD and household welfare from 1960 to 2011in five Asian countries 28

Figure 5: Relationship between FD and per capita consumption in five Asian countries (1960-2011) 30

LIST OF TABLES Table 1: Empirical studies about the causal nexus of FD and PR 10

Table 2: Proxy variables 23

Table 3: Description of FD and household welfare variables (1960-2011) 27

Table 4: t-statistics panel unit root tests 32

Table 5: t-statistics panel unit root test: Variables at the first difference 32

Table 6: Pedroni cointegration tests: Variables from 1960 to 2011 33

Table 7: Pedroni cointegration tests: Variables from 1998 to 2011 33

Table 8: Two stage least squares estimator between FD and PR 34

CHAPTER I: INTRODUCTION

1 Problem statement

It is undeniable that the relationship between financial development (FD) and economic growth (EG) has been one of the most attractive areas of research in the field of economic development over recent decades Some related studies have been employed, yet this relation remains controversial issues In fact, there have been some conflicts on the relationship between finance and growth in earlier literature In fact, Robinson (1952) and Lucas (1988) dismiss the role of finance in understanding EG while McKinnon (1973) and Miller (1988) insist on this relation between FD and EG In recent researches, the harmony of vital roles of finance in enhancing growth has been reached For example, Kirkpatrik (2000) states that good financial system that mobilizes savings and allocates resources to more productivity contributes to growth by supporting capital accumulation, promoting investment efficiency, and improving technology

Furthermore, many people believe that EG reduces absolute poverty because the more growth the economy reaches, the more jobs would be generated for the poor or the fewer differentials in wage between the skilled and unskilled labor at a later stage of development (Galor and Tsiddon, 1996) benefits the poor Then, a consensus emerged recently is that EG overall leads to poverty reduction (PR) through the improvement of household welfare

However, these close relationships between FD and EG or between EG and household welfare do not mean that FD contributes to PR (Beck et al, 2007) through the improvement of household’s welfare The explanation follows that the goal of EG in most developing countries is linked with both PR and income distribution In other words, if FD stimulates EG

by increasing income of the rich, which results in worsening income equality, FD will not benefit the poor This debate appeals many researchers to conduct studies on relationship between FD and household welfare

In addition, this paper aims to examine the relationship between FD and household welfare in five Asian countries including Indonesia, Malaysia, Philippines, Thailand, and Vietnam The reason why the research focuses on a set of these five Asian countries is that there is little research on FD and PR in Asia Due to the limit of short time series data, the research will identify the relationship between FD and PR in panel data, especially panel five Asian countries with the assumption that these countries are nearly at the same foundation of development

In the context of these countries including Vietnam, it can be seen that finance sector has a rapid development in both quantity and quality, especially in the global economy According to Odhiambo (2008), the ratio M2/GDP indicates the real size of financial sector

of a developing country Figure 1 provides some statistics of financial depth of five Asian countries in the period 1990-2010 Most of them (except Indonesia which has a slightly decrease in M2/GDP and Philippines which is nearly stable) have an increase in finance sector In fact, the slightly increasing financial sector in such three countries – Malaysia and Thailand – has been seen while it is noted that Vietnam is the country, which has a dramatic improvement in financial sector, particularly the ratio M2/GDP has been moving from around

21 percent in 1992 to nearly 110 percent in 2011

Figure 1: Money and quasi money (M2) as percentage of GDP

(Source: World Development Indicator – World Bank) Moreover, household consumption per capita in five Asian countries has also gained

as the illustration of figure 2 In detail, only in Malaysia, it can be seen the dramatic increase

in household welfare which is expressed by the per capita consumption from around 1,800 US$ in 1997 to 2,800 US$ in 2011 even though there is a steep reduction to over 1,500 US$

in 1998 It reaches nearly to the double in per capita consumption Similarly, household per

capita consumption in four countries is increasing The slope of this line is not so steep as Malaysia’s

In sum, it can be seen that finance sector improves, household per capita consumption increases Hence, the household welfare can be said to be improved In other words, FD and household welfare have a positive relationship

Figure 2: Household per capita consumption (constant 2000 US$)

(Source: World Development Indicator – World Bank) Thus, it is suggested to raise the question of whether a well-functioning financial system will help to enhance the welfare of household or not For answering this question, as well as providing policy implications, the paper particularly focus on investigating the causal relationship between FD and household welfare in the five Asian countries

2 Research objectives

In the paper, the specific research objectives are to:

i To examine whether there is any relationship between FD and household welfare

in these five Asian countries

ii How does FD affect household welfare?

iii To draw some policy implications in order to help authorities have a general view

and then propose some necessary/appropriate interventions

3 Research questions

This paper attempts to answer following research questions:

- Main question: What is the relationship between FD and household welfare in these

five Asian countries? How does FD influence household welfare?

4 Justification of the study

This paper attempts to identify the relationship between FD and household welfare, which has been one of the hottest issues in research field recently In academics, most of researches focus on the nexus of finance-growth, while this paper tries to make an effort to go further to the relationship between FD and household welfare

In addition, this paper attempts to update this relationship in these five Asian countries, which should be taken a significant consideration because researches on this field mainly have been done in African regions, some Europe countries especially in Turkey and some Asian countries such as India, China Hence, this research will establish a new foundation for further study and for local authorities to propose some necessary and appropriate policies to improve the standard living of household

5 Scope of the study

The study will examine the causal nexus of development finance and household welfare in five Asian countries including Malaysia, Indonesia, Philippines, Thailand and Vietnam with the data time series spanning from 1960 to 2011

6 Structure of the study

The rest of the paper will be organized in four more sections Section 2 presents the thereotical review of the relationship between FD and household welfare; some empirical studies are also mentioned in this section Section 3 presents a brief discussion about econometrics review Next, section 4 describes data and research methodology Section 5 dicusses the findings and discussions Finally, Section 6 concludes and suggests some practical policy implications; limitation and direction for futher studies are considered at the end

CHAPTER II: LITERATURE REVIEW

In this chapter, some theories and studies of financial development and household welfare are reviewed In addition, this chapter also covers some empirical study on the relationship between financial development and household welfare In general, this chapter comprises three main parts: definitions of key concepts; theoretical review of the relationship between financial development and household welfare; and some empirical studies

1 Definitions of key concepts

1.1 Financial development

There are several definitions of FD in many researches FD is a concept related to activities of the stock market (Chinn and Ito, 2007), which financial contracts are enforceable (Mendoza, Quadrini and Rios-Rull, 2007) and the process of innovations and improvements

of financial institutions or organizations in the financial market (Hartmann et al., 2007)

In 2011, Noureen Adna addresses at one international conference that all the factors such as policies, factors and the institutions that make a contribution to the efficiency of financial intermediaries and the efficiency of financial market are related to FD Its definition

is quite consistent with that of the report of World Economic Forum (WEF) in the same year

Similarly, in the new research of Imran and Khalil (2012), “financial development can

be defined as a process of improving the quantity, quality and efficiency of financial intermediary services”

1.2 Household welfare and Poverty

As mentioned in Merriam-Webster dictionary, welfare is a concept which refers to the state of doing well especially in respect to good fortune, happiness, well-being, or prosperity Consequently, household welfare simply refers to household well-being or household prosperity The failure for achieving a minimal capabilities or doing primarily important functions means poverty The concept of poverty has been raised for a long time Actually, there are numbers of definitions of poverty Poverty is referred, on basics, to the fact that needs of individuals or households might be satisfied in a range of limited resources

According to World Development Report of WB in 2000, poverty is defined as the state of that human suffers the physiological deprivation and social deprivation as well in their life Moreover, based on citation of United Nation’s Economic and Social Council in

Weisfeld-Adams et al (2008), poverty is fundamentally referred to an offense of human dignity It means that there is limited access to take park in social activities and limited resources to meet some certain basic needs such as clothes, food, water, education, credit Then, poverty reduction is ultimately aimed at encouraging the pro-poor growth in main sectors such as infrastructure, agriculture

2 Theoretical literature

Several studies have tried to examine the impact of FD on household welfare There are two main ways that are relevant (see Figure 3) First, the direct approach is to examine the link between FD and PR without other intermediary concepts, then sticks to household welfare Second, the indirect approach when investigating the connection between FD and PR also considers several concepts such as savings, growth It is believed that FD helps resource allocation efficiently, improves corporate governance effectively, mobilizes more savings and facilitates the exchange of goods and services Then, it leads to the improvements of the poor because FD creates more opportunities for them to be employed, and consequently their consumption is becoming smoothing, which enhances their well-being The detailed flow of process of financial sector development and PR or welfare is demonstrated as follows:

Figure 3: Financial Sector Development and Poverty Reduction

FD could strengthen the productiveness of the poor households’ assets, therefore, enhance their productivity Credit accesses and other financial service give low income households a chance to switch from low-risk, low-return assets for preventive purposes (like jewelry), to higher risk and higher return assets, (for instance education, or an agricultural instrument), with generally long-term income improving effects (Dehejia & Gatti, 2002) Savings facilities’ supplying can allow the poor to build up of reserves securely over time to finance comparatively large, incoming investments or expenditures For instance, the credit’s accessibility can reinforce the poor’s productive assets by allowing them to invest in productivity like new and higher technological implements, or to invest in their education and health Moreover, the poor can use their savings to facilitate smooth consumption for unexpected changes in their life (Holden and Prokopenko, 2001; Odhiambo, 2009) Furthermore, they can occasionally attain their savings’ return In all, those features can be principally essential for the poor to improve their condition (DFID, 2004)

FD increases the possibility for accomplishing sustainable livelihoods Credit’s accesses can decrease the susceptibility of the low income households in the case of no savings or insurance when shocks come As discussed above, savings facilities can allow the poor to accumulate their funds for unexpected thing like diseases or unemployment Hence, the shocks might be avoided, and the probability of being poor, as a result, might be minimized significantly (Zhuang et al., 2009)

Second, the FD allows the poor households to build up reserves or to borrow money

to establish micro-enterprises (DFID, 2004) Credit access can be a determinant in the construction or development of small and medium businesses Therefore, it generates employment and raises incomes (DFID, 2004) In developing countries, the small, medium and micro-sized enterprises are the most important instrument for PR or welfare improvement It is due to the fact that creating job is the principal channel to improve prosperity whilst SMEs are obviously employment-intensive (Zhuang et al., 2009) However, Zhuang et al (2009) also maintained that the accessibility of credit for SMEs is lower

compared with large enterprises Similarly, cost of credit for SMEs is higher In fact, there are good explanations for those observations From a lender’s viewpoint, it is desirable to supply credit to large enterprises In addition, SMEs do not have ability to offer collateral to make loans In all, enhancing SME credit has a significant role of PR

2.2 Indirect relationship

Several researches established the indirect relationship between the financial FD and the household welfare This indirect relationship is considered by examining the role of FD to the EG and investigating the contribution that growth leads to improvement of household welfare

First, the contribution of FD to growth is examined by many recent studies

Theoretically, EG is affected by some certain financial variables in the way of increasing

savings of financial assets It results in the accumulation of the capital formation Indeed, several empirical researches such as Odhiambo (2008), Liang and Teng (2006), and Kar et al (2011) have supported that concern

Second, the linkage between growth and household welfare (or it can be said to be poverty reduction) is also focused with attention in recent years Dollar and Kraay (2001) used the data related to the lowest income quintiles, claimed that growth benefits the poor more than other income quintiles and therefore reduce income inequalities as well as PR Klasen (2008) used both income and non-income indicator to support that growth could reduce income inequalities Donaldson (2008) also claimed that growth is good for the poor

However, Holden and Prokopenko (2001) indicated EG does not have any relationship with poverty alleviation or household welfare in some situation They argued that in high growth countries, the beneficiaries may not be the poor In those cases, the issues about the income inequality increases This means that the rich are richer while almost the poor become poorer Likewise, Basu and Mallick (2008), when examining the rural Indian case, they found that grow reduced poverty does not appear in that location

Indeed, the associations between concepts in indirect methods are currently debated (Kar et al., 2010) Therefore in this research, the causal relationship between FD and PR are examined by applying the direct method

Quartey (2005) used VECM and descriptive statistics to investigate the relationship between FD and PR in Ghana by using annual data from 1970 to 2001 He tested the causal direction between financial sector development and domestic resource mobilization; financial sector development and poverty reduction; and domestic resource mobilization and poverty reduction So, the answers for this main research question are that: even though there are no linkages between FD and mobilized savings in Ghana, he found that financial sector development seems to cause PR Besides that, his findings are also included that: First, the impact of FD on per capita consumption statistically is insignificant although the sign is positive; second, there seems to exist on a long-run cointegration relationship between FD and per capita consumption

Similarly, Odhiambo (2009) studied the causal nexus of FD and PR in Zambia from

1969 - 2006, but he used different method which is ARDL model In this research, he used three proxies of FD, which are broad money supply (M2/GDP), domestic credit to the private sector as a ratio of gross domestic product (DCP/GDP) and domestic money bank assets (DMBA) and use per capita consumption as a proxy of PR When M2/GDP is used as a proxy

of FD, he found that PR might cause per capita consumption However, when the DMBA and the DCP are used as proxy of FD, FD tend to cause per capita consumption respectively

Moreover, in the research of Odhiambo & Ho (2011), ARDL method is used to find out the relationship between FD and PR in China from 1978 to 2008 When using DCP/GDP ratio as a proxy for FD, they found the distinct causal direction in short run that FD causes per capita consumption Whilst using M2/GDP ratio for proxy of FD, there still have bidirectional causality from FD to per capita consumption in short run; but inversely per capita consumption induces FD in long run

Kar et al (2010) used the annual data of IMF and OCED online database spanning from 1970-2005, and using VECM model toexamine the causal nexus of FD and economic growth Three proxies of FD were used to investigate respectively were M2/GDP ratio,

DCP/GDP ratio and private credit/GDP ratio Some findings that the influence of FD on the per capita consumption could be found in the case of short run and weak; and they concluded that the relation of FD and per capita consumption is too blurry in short run once comparing

to the causality in long run

Finally, Inoue and Hamori (2010) investigated how financial deepening affected poverty reduction in India using state-level panel data in India However, they used a different method that is a dynamic generalized method of moments (GMM) estimation Ultimately, they found the evidence supporting for the relationship between FD and PR; EG and PR Moreover, they concluded that the higher inflation and the more international openness affect negatively on the poor

Table 1: Empirical studies about the causal nexus of FD and PR

No Authors Methodology Data Findings

1 Quartey (2005) Descriptive

statistics and Granger causality

Annual data

of Ghana from 1970-

2001

-FD does not cause savings mobilization

- FD induces per capita consumption

- Saving mobilization causes per capita consumption

2 Odhiambo

(2009)

of Zambia from 1969-

3 Kar et al

(2010)

of Turkey from 1970-

4 Inoue and

Hamori (2010)

Dynamic generalized method of moments (GMM)

State-level panel data in India (28 states in India)

-FD and EG induce PR

- The higher inflation and the more international openness affect negatively on the poor

2008

-When using DCP/GDP ratio as a proxy for FD, they found a distinct causal direction in short run that FD causes PR

- Whilst using M2/GDP ratio for proxy of FD, there still have bidirectional causality from FD to PR

in short run; but inversely PR induces

FD in long run

In sum, this chapter has captured a general picture of the nexus between FD and household welfare as well as PR in theoretical and empirical aspects as well In practice, there is a variety of methodologies employed to detect this relationship Most of them, which used VECM method to find the causal relationship between FD and per capita consumption, found the same finding that FD induces per capita consumption

CHAPTER III: ECONOMETRICS REVIEW

In this chapter, the econometric issues related to my study will be presented In particular, this chapter will describe basic concepts on stochastic process, stationarity, randon walk in time series and some advanced methodologies and concepts such as panel cointegration, generalised method of moments

1 Stochastic Process, Stationarity and Random Walks

In time series econometrics, it is equally important that the analysts should clearly understand the term “stochastic process” “Stochastic process is a collection of random variables ordered in time” (Gujarati, 2003) All basic assumptions in time series models are related to the stochastic process In the context of time series regression, the idea that historical relationships can be generalized to the future is formalized by the concept of stationary

According to Gujarati (2003), a key concept underlying stochastic process that attracts many analysts’ attention is named stationary stochastic process In general, when there exist a constant mean value and a constant variance over time, a stochastic process can be called stationary Otherwise, it is called a nonstationary time series Stationarity is very important in the context of time series models because if the series is nonstationary, all the typical results then are invalid Regressions with nonstationary time series may have no meaning and are therefore called spurious (Asteriou, 2007)

The simplest model of a variable with a stochastic trend is the random walk There are two kinds of random walks: (1) random walk without drift, (2) random walk with drift, which are defined as below:

𝑌𝑡= 𝑌𝑡−1+ 𝑢𝑡 (1)

𝑌𝑡 = 𝛼 + 𝑌𝑡−1+ 𝑢𝑡 (2) Where 𝑌𝑡 : is the random variable at the year t

𝑢𝑡 : is a white noise error term

𝛼 : is the drift parameter

2 Unit Root Test

Most macroeconomic time series are trended and therefore in most cases are nonstationary Consequently, a test for nonstationarity is a need Dickey and Fuller (1981) devised a procedure to formally test (named DF test) However, the error term is unlikely to

be white noise, so Dickey and Fuller extended their test procedure suggesting an augmented version of the test (named ADF test), which includes extra lagged terms of the dependent variable in order to eliminate autocorrelation

The three possible forms of the ADF test are given by the following equations:

Thanks to the unit root test for stationarity of time serties data, the result becomes more reliable and less spurious

3 Cointegration

According to Asteriou (2007), most of macroeconomic variables are stationary at the first difference When two variables are nonstationary, then stochastic trends can be represented However, in the case that two variable are related, it is expected that these two variables move together and when two stochastic trends are combined, it should be possible

to find a combination of them which eliminates the nonstationarity At that time, it is said that two variables are cointegrated

The definition of two variables integrated of order one is that two variables are cointegrated if there exists a parameter 𝛼 that is followed by the equation 𝑢𝑡= 𝑦𝑡− 𝛼𝑥𝑡stationary process

In general, according to the brief summary of Binh (2010), there are three cases as belows:

(i) if two nonstationary variables are integrated of the same order, but not

co-integrated, we should apply Vector Autoregressive model (VAR model) for the differenced series These models just provide short-run relationships between them

(ii) If two nonstationary variables are integrated of the same order, and co-integrated,

which suggests that there must be Granger causality in at least one direction To determine the direction of causation, it may be determined by using the error correction mechanism (ECM) model The ECMenables us to distinguish between short-run and long-run Granger causality

(iii) If two nonstationary variables are integrated of the different orders, or

non-cointegrated or non-cointegrated of an arbitrary order, it is often suggested to employ the Toda and Yamamoto version of Granger causality or Bounds Test for Cointegration within ARDL

4 Granger Causality Test

The Granger causality test of two stationary variables is expressed as followings

Where 𝑢𝑦𝑡 and 𝑢𝑥𝑡 are uncollerated white-noise error terms

The optimal lag length is popularly determined by using the Akaike’s information

by OLS and then the F Wald test is applied to test the importance of the coefficients on the lagged terms in the unrestricted models as described in the following null hypothesis

5 Panel Unit Root Test

As mentioned above, the augmented Dickey-Fuller (ADF) is very well-known to test the unit root However, it has the drawback of low power in rejecting the null hypothesis of non stationarity test due to the short-spanned data In recent years, some researchers such as Levin, Lin and Chu (LLC) (2002), Im et al (IPS) (2003), Maddala and Wu (1999) and Hardi (2000) developed panel-based unit root test to overcome the problem of traditional ADF because they have higher power than the tradition one In fact, they take advantage of the additional information by pooled cross-section time series

According to Al-Iriani (2006), among different kinds of unit root test, LLC and IPS are proposed to be the most popular because both are based on ADF principle Their model takes the following form:

𝐻1: 𝛽𝑖 = 𝛽 < 0, ∀𝑖

In contrast, the IPS is said to allow for the heterogeneity in these dynamics Hence, the null hypothesis to be test is

𝐻0: 𝛽𝑖 = 0, ∀𝑖

Against the alternative hypothesis

𝐻1: 𝛽𝑖 < 0 for at least one 𝑖

6 Panel Cointegration

There are many types of tests for panel cointegration such as Kao (1999), Pedroni (1997, 1999, 2000) and Larsson et al (2001) According to Asterious (2007), Kao’s test imposes homogenous cointegrating vector and AR coefficients However, this test does not allow for multiple exogenous variables in the cointegrating vector and it does not address the problem of identifying the cointegration vectors and the cases where more than one cointegrating vector exists Then, Pedroni’s test was arised to fix these drawbacks This approach differs from Kao’s in assuming trends for the cross-sections and considering as the null hypothesis that of no cointegration He proposed seven statistics: four of them (panel v-statistic, p-statistic, t-statistic non-parametric, t-statistic parametric) are based on pooling along the “within” dimension; the rest (group p-statistic parametric, t-statistic non-parametric, t-statistic parametric) are based on pooling the “between” dimension

Here is the general equation for cointegration regression

𝑦𝑖,𝑡 = 𝛼𝑖+ 𝛽1𝑖𝑥1𝑖,𝑡+ 𝛽2𝑖𝑥2𝑖,𝑡+ ⋯ + 𝛽𝑀𝑖𝑥𝑀𝑖,𝑡+ 𝑒𝑖,𝑡

𝑡 = 1, … , 𝑇; 𝑖 = 1, … , 𝑁; 𝑚 = 1, … , 𝑀 Where

T: the number of observations over time

N: the number of individual members in the panel (cross-section)

M: the number of independent variables

The first-difference of the original series are taken to compute the panel-p and panel-t and then the residuals of the following regression is estimated

∆𝑦𝑖,𝑡 = 𝑏1𝑖∆𝑥1𝑖,𝑡+ 𝑏2𝑖∆𝑥2𝑖,𝑡+ ⋯ + 𝑏𝑀𝑖∆𝑥𝑀𝑖,𝑡+ 𝜋𝑖,𝑡The long-run variance of 𝜋̂ (symbolized as 𝐿𝑖,𝑡 ̂ ) is formulated as follows 11𝑖2

𝑠̂𝑁,𝑇∗2 ≡ 1

𝑁∑ 𝑠̂𝑖

∗2 𝑁

𝑖=1

Then, the calculation of the seven statistics is applied as follows:

1 The panel v statistic

1

1

2 1 , 2 11 1

2 / 3 2 ,

i T

i T

t

t i N

i T

1

1

2 1 , 2 11 1

i T

t

t i N

2 / 1

1

2 1 , 2 11 1

i T

t

t i N

* 1 , 2 11 1

2 / 1

1

2 1 , 2 11 1

t t N

i T

1

2 1 , 1

2 / 1 1

t t i N i T

2 / 1

1

2 1 , 2 1

2 / 1 1

,

2

/

7 The group t statistic (parametric)

t

t i N

i T

* 1 ,

2 / 1

1

2 1 , 2 1

2 / 1

7 Instrumental Variables Regression (IV)

Instrumental Variables (IV) estimation is used when the model has endogenous X’s It

is also used to address some threats: (1) omitted variable bias from a variable that is correlated with X but is unobserved, so cannot be included in the regression, (2) simultaneous causality bias, (3) errors-in-variable bias

Let consider the following model:

𝑦 =∝ +𝛽1𝑥1+ 𝛽2𝑥2+ ⋯ + 𝛽𝑘𝑥𝑘+ 𝑢 Where 𝐸(𝑢) = 0 and 𝑐𝑜𝑣(𝑥𝑗, 𝑢) = 0 for 𝑗 = 1,2, … , 𝐾 − 1

If the residual is correlated with 𝑥𝑘, there exists the potentially endogenous problem, which causes OLS inconsistent

This reason leads to the IV estimator solution The instrument 𝑧1 is chosen to satisfy two conditions: (1) the instrument must be exogenous or it can be said that 𝑐𝑜𝑣(𝑧𝑗, 𝑢) = 0, (2) the instrument must be informative or relevant This means that the instrument must be correlated with the endogenous regressor 𝑥𝑘 and conditional on all exogenous variables in the model

The considered IV regression above is simple with one endogenous explanatory variable and one instrument Similarly, if the regression with two endogenous explanatory

under-identified when there are fewer instruments than endogenous regressors and vice versa

In practice, it is a good idea to have more instruments than what is really needed Wooldridge (2005) assure that the two stage least squares (2SLS) obtained by using all instruments simultaneously in the first stage regression is the most efficient IV estimator

8 Generalized method of moments (GMM)

According to Gujarati (2004), there are three methods of parameter estimation: (1) least squares, (2) maximum likelihood, and (3) method of moments and its extension, the generalized method of moments (GMM)

Assume the following model:

𝑌𝑖𝑡 =∝ 𝑋𝑖𝑡+ 𝛿𝑌𝑖,𝑡−1+ 𝜇𝑖+ 𝑢𝑖𝑡

in which N countries are observed over T period, 𝜇𝑖 is country-specific effects and the disturbances 𝑢𝑖𝑡 are assumed to be independently distributed across countries with a zero mean

Since 𝑌𝑖𝑡 is a function of 𝜇𝑖, this implies that 𝑌𝑖,𝑡−1 is also a function of 𝜇𝑖 Therefore,

𝑌𝑖,𝑡−1 is correlated with the error term through the presence of 𝜇𝑖 As a result OLS is biased and inconsistent To fix this problem, Anderson and Hsiao (1981, 1982) suggested first differencing the model to get rid of the residual 𝜇𝑖, and then using an instrument variable method (IV method) This proposed method leads to consistent but not necessarily efficient estimates because it does not make use of all available moment conditions and not take into account the differenced structure on the residual disturbances It is also a reason that Arellano and Bond (1991) proposed a more efficient estimation procedure His idea is to take the first differences to get rid of the individual effects and use all the past information of dependent variables as instruments GMM is introduced to overcome this problem The reason lies in the possible correlation between the lagged dependent variable and the unobserved country specific effect

However, Blundell and Bond (1998) found that it has some drawbacks that this has poor finite sample properties in terms of bias and precision Arellano and Bover (1995) and Blundell and Bond (1998) proposed a system based approach to overcome these limitations The combination of the standard set of equations in first difference with suitably lagged levels as instruments and with an additional set of equations in the levels with lagged first differences as instruments forms the system GMM The Arellano and Bover test is used to

detect the autocorrelation, in which there is a null hypothesis that no autocorrelation exists It

is also applied to the differenced residuals The more important thing is that the test of AR(2)

in first difference since autocorrelation in levels will be detected

As discussed above in previous sections, conventional IV estimators such as two stage least squares (2SLS) is considered special cases of GMM estimator In fact, the method is nearly the same Due to the fact that the finding of a compatible and reasonable instrument variables encounter with problems, it is suggested that the lagged variables be used as instrument variables

The basic model:

𝑦 = 𝛽𝑋 + 𝑢, 𝑢~(0, Ω), X ( N x k ) Let define Z ( N x l ) where l ≥ k the Generalized Method of Moments IV (IV-GMM) estimator The l instruments give rise to a set of l moments:

𝑔𝑖(𝛽) = 𝑍𝑖′𝑢𝑖 = 𝑍𝑖′(𝑦𝑖 − 𝑥𝑖𝛽), 𝑖 = 1, 𝑁 The averaging over N of g is estimated by

𝛽̂ = (𝑋2𝑆𝐿𝑆 ̂𝑋)′ −1𝑋̂𝑦 = (𝑋′ ′𝑃𝑍𝑋)−1𝑋′𝑃𝑍𝑦

If and only if an equation is over-identified, with more excluded instruments than included endogenous variables, we may test whether the excluded instruments are appropriately independent of the error process That test should always be performed when it

is possible to do so, as it allows us to evaluate the validity of the instruments

A test of over-identifying restrictions does a regression the residuals from an IV or 2SLS regression on all instruments in Z Under the null hypothesis that all instruments are uncorrelated with u the test has a large sample chi square distribution where r is the number

Besides that, according to Hansen (1982), there is Hansen test which is tested the validity of the instruments or it can be said that Hansen test is a test whether the instruments are uncorrelated with the error term 𝑢𝑖𝑡 Hansen test has the null hypothesis that over-identifying restrictions are valid If the p-value of the test is less than our critical significance level, we reject the null hypothesis that the over-identifying assumptions are valid Hansen test displays outside as J-statistic It is referred to the value of the GMM objective function evaluated using an efficient GMM estimator This statistic acts as an omnibus test statistic for model mis-specification A large value of J-statistic indicates a mis-specified model The p-value of this test is calculated by the chi square function of the J statistic and the degree of freedom In general, the degree of freedom is formulated equally to the number of moment conditions minus the number of coefficients It can be expressed by the formula

𝑝_𝑣𝑎𝑙𝑢𝑒 = 𝑐ℎ𝑖𝑠𝑞(𝐽_𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐, 𝑑𝑒𝑔𝑟𝑒𝑒 𝑜𝑓 𝑓𝑟𝑒𝑒𝑑𝑜𝑚)

In sum, all related econometric concepts and methods are presented in this chapter This chapter gives us more clear vision about econometrics

CHAPTER IV: DATA AND RESEARCH METHODOLOGY

The literature review of the nexus between FD and household welfare and econometrics review have been described in the previous chapter This chapter continues to present an overview of data and research methodology used for this study

1 Data

There have been various measurements to address the level of FD In several recent studies, the ways to measure the level of FD are varied According to Hassan, Sanchez, & Yu (2011), they used some popular proxies for FD such as domestic credit provided by banking sector as a percentage of GDP (DCBS), domestic credit to the private sector as a percentage

of GDP (DCPS), the broadest definition of money (M3) as a proportion of GDP or the ratio

of gross domestic savings to GDP For this study, DCBS, DCPS and the broad money supply ratio (M2/GDP) are proxies for FD

Hassan et al (2011) suggested choosing M3/GDP because it is more likely to reflect the ability of the financial systems to provide transaction services than to the ability to channel funds from savers to borrowers (Khan & Senhadji, 2003, p.ii93) However, due to the lack of data, in this study, I select the ratio M2/GDP even although it is not a good demonstration of the level of FD The support of my selection is the study of measuring banking sector development of WB, which assume the ratio M2/GDP one of traditional indicators of financial sector development This ratio captures the degree of monetization in the system

Additionally, two more proxies DCBS and DCPS are used to show the size of the financial market Based on the study of Levine (1997), banks are assumed to provide five main functions in the market; hence, the higher DCBS means the higher degree of dependence on the banking sector for financing Similarly, a high DCPS indicates a higher level of domestic investment and a higher development of FD as well

In this research paper, per capita consumption is used as a proxy for welfare In the previous research, the authors have used GDP per capita as a proxy for welfare; however, it does not account for the observed connection between poverty and growth Meanwhile, as suggested by WB, it had better to use consumption as a measurement of PR than income First, consumption is said to be a better indicator because actual consumption has a close link with ones’ well-being to meet their basic needs while income is only one among other

poverty more exactly In reality, the income flows of the poor may fluctuate during a year whereas their consumption pattern seems to be certain and steady for a year Therefore, it is realized that per capita consumption is the suitable proxy for welfare

Consequently, table 2 demonstrates clearly the details of all proxies of FD and welfare

of the five Asia countries (including Indonesia, Philippines, Malaysia, Thailand and Vietnam) which will be collectedannually from World Development Indicator (World Bank) in the period of time from 1960 to 2011

Table 2: Proxy variables

Variable Proxy Label Time Data

sector2 as a percentage of GDP DCBS Annual

World Bank Domestic credit to the private sector3 as

World Bank

Household

welfare

Household final consumption expenditure per capita (constant 2,000 US$)

Bank

2 Research methodology

In respective to the objective of this study, the question whether FD is causally related

to household welfare might be investigated The general form to show the relationship between FD and household welfare is:

1 “Money and quasi money comprise the sum of currency outside banks, demand deposits other than those of the central government, and the time, savings, and foreign currency deposits of resident sectors other than the central government This definition of money supply is frequently called M2.” (World Bank)

2 “Domestic credit provided by the banking sector includes all credit to various sectors on a gross basis, with the exception of credit to the central government, which is net The banking sector includes monetary authorities and deposit money banks, as well as other banking institutions where data are available (including institutions that do not accept transferable deposits but do incur such liabilities as time and savings deposits).”(World Bank)

3 “Domestic credit to private sector refers to financial resources provided to the private sector, such as through loans, purchases of nonequity securities, and trade credits and other accounts receivable, that establish a claim for repayment For some countries these claims include credit to public enterprises.” (World Bank)

POV = f (FD+)

Or

𝑃𝑂𝑉𝑖𝑡 = 𝛼 + 𝛽𝐹𝐷𝑖𝑡 + 𝜀𝑖𝑡Where

POV is the symbol of household welfare, in particular it calculated by per capita consumption

FD is the symbol of financial development, in which there are three measurements: DCBS (domestic credit provided by banking sector), DCPS (domestic credit to private sector), M2GDP (money and quasi money as a percentage of GDP)

In particular, the three equations are formed below:

𝑃𝑂𝑉𝑖𝑡 = 𝛼 + 𝛽𝐷𝐶𝐵𝑆𝑖𝑡+ 𝜀𝑖𝑡𝑃𝑂𝑉𝑖𝑡 = 𝛼 + 𝛽𝐷𝐶𝑃𝑆𝑖𝑡+ 𝜀𝑖𝑡𝑃𝑂𝑉𝑖𝑡 = 𝛼 + 𝛽𝑀2𝐺𝐷𝑃𝑖𝑡 + 𝜀𝑖𝑡

When the panel cointegration is detected through the Pedroni’s cointegration test, this means that there exists a long-run relationship As previewed in previous section, Pedroni proposed seven statistics of residual-based panel cointegration test composed of two main groups: within-dimension-based (panel-p and panel-t) and between-dimension-based (group-

p and group-t) He also assumed that the rho-statistics is the most powerful one to test the cointegration in the large sample size (Pedroni, 1999) whilethe panel-t statistics and the group-t statistics are more reliable in a small sample (Pedroni, 2004)

Before doing the Pedroni’s, the panel unit root test is run In this study, we use LLC and IPS panel unit root test because both are based on ADF principle The difference between LLC and IPS lies on that LLC assumes homogeneity in the dynamics of the autoregressive coefficients for all panel members while IPS is said to allow for the heterogeneity in these dynamics

In reality, recent studies have applied a variety of techniques to estimate systems of linear equation simultaneously The well-known technique commonly used is the class of instrument variables (IV) methods, of which 2SLS is the most important special case In this study, 2SLS is employed to detect the relationship between FD and household welfare

The instrumental variables chosen are the lagged variables of POV It is sufficient because it meets two requirements that it is highly correlated with DCBS, DCPS, or M2GDP and uncorrelated with itself of the t time

When using 2SLS, we should pay attention to the J-statistics because if the J-statistic value is too high, it also displays the mis-specification of the model or if the J-statistic is identically zero it will be positive for an over-identified equation The J-statistics is the test of over-identifying restrictions The null hypothesis of this test is that over-identifying restrictions are valid The p-value of J-statistics is calculated mentioned in previous section

If the specification is correct, we expect is the failure of rejection of the null hypothesis

This chapter has presented the data and research methodology As for the data, the study collects the macroecnomic data from WDI - WB As for the research methodology, the study employs the GMM two-stage least square model