Credit growth, macroeconomic factors and ftock performance the case of hose 2002 2010

The paper analyzes the dynamic interactions among credit growth, some fundamental macro-economic factors exchange rate, inflation, industrial production, interest rate, gold price and th

CREDIT GROWTH, MACROECONOMIC

FACTORS AND STOCK PERFORMANCE:

THE CASE OF HOSE 2002-2010

A thesis submitted in partial fulfillment of the requirements for the degree

Dr PHAM HOANG VAN

Dr NGUYEN TRONG HOAI

HO CHI MINH CITY, MARCH 2011

TABLE OF CONTENT

2

2 1

INTRODUCTION -4

LITE TJJ EW : J 0 Y N EpT j 2.2 THEORETICAL LITERATURE -11

2.3 EMPIRICAL LITERATURE -17

3 5 5.1 5.2 5.3 6 7 TJ-((()()L() ; . - - 2

RESEARCH METHODS -26

DATA DESCRIPTION -35

DATA ANALYSIS -37

RESULT ANALYSIS: -41

DESCRIPTIVE ANALYSIS -41

STATIONARY PROPERTY OF TIME-SERIES DATA -44

BIVARITATE RELATIONSHIPS: EG APPROACH & GRANGER CAUSALITY -45

MULTIVARIATE RELATIONSHIP: JJ PROCEDURE, VECM, VARIANCE DECOMPOSITION & IMPULSE RESPONSE FUNCTIONS -48

2

The paper analyzes the dynamic interactions among credit growth, some fundamental

macro-economic factors (exchange rate, inflation, industrial production, interest rate, gold price) and the

performance of Vietnam Ho Chi Minh Stock Exchange using time series econometrics ofcointegration and causality tests In the analysis, we explore further with VAR-based variance

decomposition and impulse response functions to capture the direct and indirect effects of

' innovations in one variable and other ones in the same model The interesting results come out with negative reaction of stock price to credit growth in first 11 months of study period before

any reverse trend occurs And then it will still remain the sign in longer term However, itseems no significant evidence to prove the positive short-run impact of credit rate on stockprice increase So Interest subsidy policy after global crisis (2008) is not the major reason torescue equity market In addition, the variation of key variables including interest rate,inflation and exchange rate has significant impacts on stock volatility in long run Oneimportant policy implication is that authorities should be cautious in implementing monetarypolicies exposed to inflation risk as it has a consistent adverse influence on stock change inboth short and long term

Keywords: Credit growth, Macro-economics, Stock performance

and Impulse Response Function

namely REE and SACOM traded two days per week From 1 March 2002, market traded daily

with two order-matching sessions Till 31 December 2007, 138 stocks were listed and traded fivedays per week through a fully-computerized trading system, Automatic Order-Matching and Put-

Through Trading system In general, the total capitalization in HOSE accounts for over 40%

GDP During operating time, Vn-Index had a sharp fluctuation peak to 1 137 in February of 2007and then turned down sharply, which shocked almost investors and policy authorities According

to some former studies, the root cause is originated from market participant’s over-expectation

on booming price

Despite HOSE’s certain achievements over year, it is still fragile due to its own highrisk, big price volatility and poor trading system As one of Asian emerging markets,HOSE has experienced ups and downs because of the significant influences from externaland internal factors In reality, many controversial problems signal the market inefficiency andinstability in

- terms of information asymmetry, a weak legal framework, the lack of transparency in financial

reporting, too much Government intervention in trading transactions and herding investor

behaviors Indeed, these weaknesses are the challenges of Vietnam economy towards financial

liberalization process Thus, HOSE in particular or Vietnam stock exchange in general hopefullywill develop into a strongly efficient capital-raising channel in the near future

In the development period, the required tasks for policy makers from now on is to domore qualified researches on Vietnam stock market and prevailing problems for timelyadjustment Observing the economic changes since 2002, the rapid domestic credit has grown

by nearly 10 times, from about 2 hundred thousand up to more than 2 million billion dong.Simultaneously, stock market boomed aggressively in 2007 when VN Index created history(Figure 1.1) And the question whether these two factors have any links has raised the interest infurther estimation Then its empirical result below can explain the relationship between domestic

• credit growth and stock volatility, especially in the period right after the peak in 2007 until the

broadening monetary policy in 2009

Figure 1.1: VN-Index & Credit aggregate in 2002M1-2010M3

5

1,200

1,000

600 400

400000

Sources.‘ IMF (2010j

6

Let’s discuss more about Vietnam economic background and why credit growthaccelerated the period prior to 2008 A relevant update from World Bank (2007) gave some

explanation related to the so-called “Impossible trinity” of simultaneously maintaining nearly

fixed foreign exchange, independent monetary policy and an open capital account Under this

implementation, the incident following increasing capital inflows was foreign exchange

depreciation or domestic currency appreciation that kept appealing more investors The economy

was put in a challenge of excess liquidity Due to its negative impact on Vietnam

competitiveness of export and growth slow-down, the Government intervened by purchasing

foreign exchange and selling securities Then it moved foreign exchange market muchflexibly However, the foreign reserve accumulation and VND appreciation forced SBV tochoose monetary and credit expansion in term of sterilization This also hid potentialexposure to inflation and then unpredictable capital outflows, most seriously a crisis (2008)when the stock value returned its real value So it raises the interest in finding the real impact ofcredit growth on stock performance scientifically over the period which will be presented inthe following parts

1.2

The scientific challenge:

The research will discover whether credit shocks have significantly affected HOSEperformance Particularly, the credit growth under interest support program in 2009 aimed ateconomic stimulation after global crisis in 2008 And thesis also generalizes how lagged lengthbetween credit growth and stock price reaction would be Further estimation will uncover which

of key macroeconomic and trend variables has remarkably influenced on stock price in bothshort and long run since 2002 Based on the empirical result, stock investors, academiceconomists and authorities can refer to the findings for their own decision-making However,

7

some existing limitation of data and quantitative tools to interpret it in economic meaning are necessary for further research then.

1.3

Goal and objectives of the research:

The overall goal of the project is to provide scientific results as trustworthy referencesfor stock investors And then it gives some appropriate policy recommendations related tocredit applicable for the development of Ho Chi Minh Stock Exchange and Vietnameconomy as a whole

To meet the goal, the first specific objective of the research is to identify thecointegration and causality of domestic credit growth to HOSE’s performance Next is toanalyze

, the lag length between the prominent change in some monetary policies and its impact on

stock price market over 2002-2010 Especially and specifically, how Decision TTg on Interest Rate Support for Organizations to expand their Production and Businessaffected stock price will be discussed in the paper By application of VAR and monetarytransmission mechanism (MTM), the research functions forecasting the stock volatility

131-2009-QD-The second objective is to explain more about interactions among credit growth, otherkey macroeconomic indicators and HOSE index The testing will identify how significant and

which relationships, negative or positive, each independent macro-variables impact on the

dependent stock price

The last is to give some recommendation for both stock exchange managers and

government policy makers

Research Ouestions:

In order to achieve the above objectives, let’s try to answer the following questions:

1 Is there any relationship between credit growth rate and stock price index in

long-term as well as short-long-term?

2 Among potential substitute investment channels via foreign exchange, gold, moneyand overseas stock markets, does domestic interest rate have the immediate effect

on the stock price variation?

3 Associated with credit, are all selective macroeconomic factors necessary for

forecasting HOSE price change in the long-run? If yes, what is the sign ofindividual relationship between stock price and the others?

1.5

Structure of the thesis:

The thesis will follow introduction section with four other chapters Chapter 2 reviewsthe applicable theories of stock price determination as well as empirical studies about the

relationship between security index and macro-economic indicators Chapter 3 describesresearch methodology including data collection, variables of interest, econometric modeltogether with empirical procedures Chapter 4 analyses the research results according tomethods recommended previously It answers the thesis hypotheses whether the causality ofCredit growth to stock price change exists, which market the main substitute for stockinvestment

• channel is and whether other macro-variables have significant impacts on stock variations And

chapter 5 closes the study with a conclusion, policy implication and opens with its limitationfor further studies

9

CHAPTER 2: LITERATURE REVIEW FOR STOCK’S RELATIONS WITH

bank financing channel, it will in turn boost aggregate demand and output, eventually push upstock price In line with the balance sheet channel, Bernanke and Gertler (1995) concerned theexternal finance premium, which they defined as the bridge between the externally-raised cost

of funds and the opportunity cost of internal funds Yet, this seems not significant in the case

of Vietnam because most credit has recently been granted to big state-owned enterprisesregardless of the consideration of their financial position Briefly, domestic credit growth viabanking loans is the channel the State Bank uses to inject liquidity to the whole market

2.1.2 Macro-economics

Macroeconomics is the study of the performance, structure and behavior of the entire

economy as a whole It is different from microeconomics which focuses more on individuals

and

how they make economic decisions Actually, macro-economy is so complicated because thereare many factors influence on it We usually analyze macro-economy by primarily looking at

national output (GDP), unemployment and inflation Besides, there are consumption, interestrate, foreign exchange, international trade and international finance which are modeled toexplain economic relationships The study herein will employ some of the mentionedindicators to see how they affect stock market

2.1.3 Stock performance

Stock performance is a measure of the returns on shares over a period of time There are

several measures of stock performance and each includes its own characteristics and benefitsduring an analysis of returns In order to understand stock performance of HOSE in Vietnam,the thesis will employ stock index (VN-Index) as its proxy

2.2

Theoretical literature:

2.2.1 Arbitrage Pricing TheorY (APT)

This research is drawn from Arbitrage Pricing Theory (APT) Developed by Stephen

Ross (1976), the theory asserts that an asset’s expected return is a linear function ofunanticipated change in a small number of macroeconomic factors The sensitivity of theasset’s return to the individual elements is determined empirically by way of statisticaltechnique The APT does not identify the specific factors that would affect asset returns Yetfour factors are frequently evaluated in most application of the APT: (1) inflation, (2) industrialproduction, (3)

risk premium and (4) the term structure of interest rate Different from Capital Asset PricingModel’s risk measure by asset “beta” against market index, the four forces are primary

11

'

influences on stock market Based on the traditional discounted cash flow valuation, the

unanticipated variations of inflation rate and industrial production are related to the real value of

future cash flow The other two variables seem intuitively to be more linked to the adjusted discount rate Risk premium measures investor attitudes toward risk perceptionabout general level of economic uncertainty while the term structure of interest rate influences

risk-on discount rate for multiply year cash flows However, the study did not show the evidence todeny the rest of unsystematic variables affecting asset returns

2.2.2 Discounted Cash Flow (DCF)

These four above factors find based on a common financial theory, namely, DiscountedCash Flow valuation (DCF) which was identified by Chen Ross & Roll (1986) It is one of the

foundations to estimate the true value of common stock investment Since the study aims to

identify key determinants of stock’s true value rather fundamental analysis, Dividend Discount

Model (DDM) is suitable for review in this section The model assumes that the time money

value of common stock is the present value of all future dividends In the following equation[2 1], interest rate and inflation affect discount rate and future cash flows while industrialproduction reflects the health of business performance via dividend payment to shareholders

[2 1]

Dt: Dividend accrued from period I

Ke: Cost of Equity

Assume that the dividend growth rate holds constant (g%), the time value of stock is written inform of the following function.

[2.2]

Where D : Dividend in period 1

: Cost of Equity

g : Constant growth rate

Based on the above function, there are 2 fundamental determinants, cost of equity and expectedgrowth rate, to examine the intrinsic value of a stock They are independent to each otherpursuant to Keilly and Brown (1997) Growth rate depends on retention rate (b) and return on

equity investment (ROE) via the function: g = b * ROE while stands for risk factors Total

risk is a combination between systematic risk (market risk) and unsystematic risk specific financial risk) Systematic risk is the variability of return on stocks or portfolioassociated with changes in return on the market as a whole whereas unsystematic one is alsothe return variation but not explained by general market movements It is avoidablethrough diversification of investment stock portfolio In regard to market risk, an aversioninvestor can look at macroeconomic and political conditions for their investment decisions.Damodaran (2002) demonstrated determinants of as a function of interest rate, inflation,exchange rate, news about economy and political stability On the other hand, disregardingreinvestment rate b determined by individual firm’s dividend policy, ROE in the equation ofgrowth rate should be concerned It is affected by the financing decisions of each firm via

(firm-13

ROE=ROC+D/E.[ROC-r(1-t)] [2.3]

Where ROC: Return on capital investment

D/E: Capital structure (Debt over Equity)

r: Interest expense on debt

t: Tax rate on ordinary income

Therefore, expected growth rate of a stock will eventually base on both monetary and fiscalpolicy changes since r is related to monetary policy and t refers to fiscal one From all above-mentioned equations, we can conclude that the change in stock price would be partially

determined by macro-economic variables related to interest rate, inflation, industrialproduction, exchange rate, money supply, credit aggregate, taxation and government budget.The research will employ some similar fundamental indicators based on this theory

2.2.3 Inflation and Stock Market

Beside APT and DCF, two alternative studies about why inflation affects stock price

were discussed in Feldstein (1980) and Modigliani-Cohn (1979)

Feldstein’s “Inflation and the Stock Market” indicated the inverse relation between higher rate of inflation and sustainable reduction in the ratio of share prices to pretax earnings

' But what was the cause to the failure of price increase during a decade (1980s) of steady inflation in US? This adverse effect resulted from the features of State tax laws, particularly

historic cost depreciation and the taxation on corporate-source income Further, the study

analyzed the difference between the effect of high constant inflation rate and its increaseexpected for future If the steady-state of inflation rate is higher, stock prices would go up atfaster rate And an increase in expected future of inflation rate would lead to a concurrent fall in

the ratio of share price to current earnings (PE) Then the price rises at higher rate of inflationbut PE ratio will be permanently lower as inflation pushes up the effective tax rate oncorporate revenue Despite Feldstein’s affirmation on the inverse influence of inflation onstock price via

US tax regime, he did not deny the anticipated factors — the slowdown in productivity growth,

booming cost of energy and increasing international competition — decreasing pretaxprofitability Sharing the same view of the negative relation between inflation and stock price,

Modigliani-Cohn approached it in different ways The reason why the ratio of market value to

profits declined in the late 1960s is two errors in evaluating common stocks The first is thatinvestors capitalize equity earning at a rate which is equal the nominal interest rate not

economically correct real rate The second is that investors fail to allow for the gain to

shareholders accruing from depreciation in the real value of nominal corporate liability Inother words, the inflation-induced errors are exposed to a permanently depressing effect onreported earnings even to the point of turning real profit into growing losses In short, the rate

of inflation is an indispensable proxy to measure the uncertainty of an economy in general orstock market in particular

2.2.4 Monetary Transmission Mechanism (MTM):

One more applicable theory within the analysis is Monetary Transmission Mechanism

(MTM) It describes how a monetary policy change affects some important macroeconomicvariables The specific channels of MTM operate through the effects of monetary policy oninterest rates, exchange rates, equity and real estate prices, bank lending and corporate balance

15

sheet According to Mishkin (2006), the increase in money supply may lead the rise of pricelevel and potential growing real output in the short-run, which can occur through four typicalchannels like interest rate, credit, exchange rate and asset price channel In case of Vietnam, the

credit channel explains much of the way in which credit as money supply affects real output Itis

more important than interest rate channel resulted from a study of Le Viet Hung, an expert from

Foreign Exchange Department (2008)

2.2.5 Efficient Market Hvpothesi EMH):

Lastly, one of the most influential theories of asset pricing is Efficient Market

Hypothesis (EMU) developed by Fama (1970) It asserts that financial markets are

' informatively efficient when market prices incorporate and reflect all relevant available

information There are three forms of market efficiency If all available information is related topast prices, it is classified into “weak form” And if it refers to all public information, the market

is defined as “semi-strong form” efficiency Otherwise it would be “strong-form” when shareprices adjust to all public and private information In the last form, investors cannot consistently

earn excess returns over a long period of time The theory has a lot of disputes andcontroversy, especially from technical analysts as they base their expectation on past prices,earnings track records and other indicators to predict the market trends In reality, it is veryhard to become a strong-form efficient market due to asymmetric problems

According to the above-mentioned theoretical review, there remain two signs of the

• dynamic interactions running from the key macro-economic factors to stock price The

negative sign refers to the causality of interest rate, inflation and exchange rate to stock index.And the positive exhibits its correlation with money supply and industrial production Differentfrom the

previous analytical evidence, the thesis is developed to find out some specific relationships

among key macro elements and the proxy variable of stock performance in the case of Ho ChiMinh Stock Exchange First of all is whether a consecutive credit growth, main independent

variable, over studying period has any impact on VN-Index immediately or within 12 months

The research interest is raised from whether the same upward trend of data series (Figure 1) overyears presents a cointegration in long term and a causal relation in short term Second, asobserved Vietnam investment market from 2002 up to now, almost individual investors have

changed the strategies toward money market when there are negative risk signals from stock

market Banking deposit is one of safe substitutes That is the reason why the research wants to

answer if money market is the most active investment alternative for equity in short-run Onemore discussion is the author’s ambition to understand whether 4 selective variables ofinterest, credit growth, foreign exchange, interest rate and inflation, have forecasting abilities

to stock volatility in long-run

2.3

Empirical literature:

Ibrahim and Yusoff(200 1) from Malaysia employed mainly VAR, Johansen-JuseliusCo- integration, Monetary Transmission Mechanism and AS-AD model to analyze therelationship between Kuala Lumpur Composite Index (KLCI) with selective macro-indicatorincluding M2 monetary aggregate as money supply, real industrial production to capturereal economic

• activities, CPI and exchange rate (Ringgit/USD) expressed in natural logarithms All the data

were collected in the period of Jan 1977-Aug 1998 As a result, they concluded that moremoney supply leads to more positive effects on stock price in short-run and vice versa inlong-run

Conversely, changes in stock price also drive the increase in demand for real money, interestrate and subsequently the value of domestic currency Furthermore, domestic currencydepreciation presented by exchange rate appreciation is both contractionary andinflationary Therefore, Malaysian authorities should pay attention to stabilizing theirexchange rate and monetary policies due to its adverse repercussion on financial market.

CooperMaysami, Howe and Hamzah (2004) tested the interactions between chosen factors

including interest rate, industrial production, price level, money supply, exchange rate and

Singapore Exchange Sector Indices of Finance, Property and Hotel by applying Johansen’sVECM model (1990) in multivariate context, a full information maximum likelihood estimationmodel Based on intuitive financial theory (Chen et al 1986 & Fama 1981), the model allows for

' testing a whole system of equation in one step and avoids carrying over errors from the first into

the second step And the conclusion drawn from its empirical study is that Singapore stockmarket and SES All-S Equities Property Index formed significant relationships with all

macroeconomic variables identified while the other two did with only selected ones Specially,the financial sector is significant affected by inflation rate, exchange rate and both long and

short-term interest rate whereas its relation to money supply is weaker in comparison with

Singapore’s stock market as a whole From this point, the authors came up with a statement that

“stock picking” could lead to superior earning capability

Another research is from the case of Norwegian stock market Anderson and Lauvsnes(2007) did a research on co-integration between stock price index and domestic credit instead ofmoney supply used by other previous studies for forecasting purpose Various methods and

comparisons were employed in their study First, they compared forecasts from VAR models

with and without imposing co-integration restrictions with simple uni-variate models in different

18

forecasting horizons Then it stepped up with comparing Mean Square Errors MSE and Mean

Absolute Errors MAE from 4 models: VAR log-differences, VAR - VECM, Random walk,

ARIMA(1,1,I ) in Box-JENKINs analysis According to many problems happening since1980s,

they particularly found financial turbulence with booms in stock markets, a strong overallgrowth in credit and asset prices (property & stock), concerns about sustainability of domesticprivate credit via excessive borrowing causing to increases in cost (e.g interest rate) andreduction of income as the result In next step, these scholars tried answering relevanthypothesis: How to improve stock index and credit growth forecasts? What are the fragileendogenous relationships in economy (Minskyl987)? Does previously detected co-integrationbetween these variables contribute to forecast accuracy? Additionally, they addressedcomparison between the out-of- sample forecast and previous in-sample results Theyscientifically figured out the results that Stock index is better predicted when co-integration isimposed Credit variable is better predicted with multivariate models than the uni-variate ones

In the case of Thailand, many scholars from different approaches identified macro-economic

elements affecting their stock market performance In 2007, Tantatape and Komain used

co-integration model to test the long-run relationship and ECM to determine short-run deviation

from long-run equilibrium The study collected monthly data of Money supply, FX, IPI andOil price, Inflation rate logarithm (1992-2003) In addition, Mahmood and Dinniah (2007)expanded their study of 6 Asia Pacific countries consisting of Malaysia, Korea, Thailand,Hong Kong,

Japan and Australia to make a comparison at regional level With the similar methodology(ECM) and observation periods but less independent variables (exchange rate, inflation rate and

• industrial output), the results came up with the co-integration between stock price and

exchange rate in Hong Kong market and industrial output in Thailand during

1993-2002 And

19

macroeconomic indicators in other countries seemed to produce a negligible impact on stock

return From different approaches, Nguyen Dinh Tho (2010) employed Arbitrage pricing theory

"(APT) to explore the behaviors of Thailand stock price against the changes of macroeconomic

factors Moreover, the research discovered its relationship with exchange rate, IP growth rate,inflation rate, current account balance, international-domestic interest rate difference and thechanges in domestic interest rate He collected monthly dataset before Asia Financial Crisis(from Jan 1987 to Dec 1998) to compute all with basic test of correlation coefficient to avoidmulti-colinearity, auto-correlation, ADF unit root test and finally OLS The reliable result was

thanks to various tests on flexible combination of endogenous and exogenous variables Thestudy ended with “Industrial production and exchange rate systematically affect stock returns

while the returns on value weighted SET index, used as a proxy of market portfolio, fails toshow its significance”

Regarding a limited number of empirical studies in Vietnam, Loc, Lanjouw and Lensink

(2008) conducted auto-correlation together with variance-ratio tests to discover whetherVietnam stock market holds weak-form efficiency It also answers the question if there is apossible bias of the results caused by thin-trading that characterizes the market Theemployed data was collected on weekly basis (2000M7D28-2004M12D31) The findingsscientifically reported that Stock Trading Center (STC) is not weak-form efficient in thin-trading period, even in case, its corrections of thin-trading are made In the same year, Long(2008) used daily data of closing stock price and return rate of index (%) in 2000-2007 period

to examine stock return volatility against regime changes By testing in GARCH model, hispredominant outcome proved that Financial Liberalization has a negative influence on stockreturn variation However, it is not easy to separate the influence of Financial Liberalization

on stock performance from that of the

20

growing number of IPOs (Initial Public Offering) because they coincided during research period.Then Khuyen (2009) investigated the efficiency of Vietnam stock market via the lagged impact

of twelve macro-economic monthly variables in 2000M12-2009M6 These factors representfor Real Production, Foreign Trade and Money Market indices The time series wereprocessed in order by Stationary test, Cointegration and Granger Causality tests in both longand short run The results form multivariate analysis reinforced that Vietnam stockperformance is not informationally efficient Thus, it is possible for any trader to earnabnormal returns by understanding good or bad news of macro-economics Besides, thearticles confirmed that the financial crisis worsened the inefficiency of stock market viamonetary variables As a result, the market performance is not well functioning in scareresource allocation It leads to less attractiveness to encourage almost foreign and domesticinvestors

Learning and developing from such above literatures, the research exploits further theperformance of Vietnam stock market via HOSE price index Moreover, the other differentiation

is the selection of main independent variable, using credit growth instead of money supply in aprevious studies of Malaysia and Vietnam in order to explain their relationships if any Besides,

the study keeps employ the similar fundamental economic factors including exchange rate,

inflation rate, interest rate, domestic deposit, industrial production to Asia Pacific researches.Another is the input of compelling variables related to HOSE stock’s co-integrations with USstock index (Dow Jones) — the proxy for a developed market, China stock index (SSE) — the

proxy for an emerging market and gold price All the up-to-date data (2002-2010) will be

' collected for the quantitative estimation of stock performance via the below conceptual

• framework (Figure 2.1) Among many econometric approaches applied in the reviewed studies,

the paper will only choose the 3 most suitable methods, Cointegration, Causality tests and

21

Impulse response functions, to test the target interactions between stock price and other

economic factors

Let’s discuss three main hypotheses that the paper will econometrically explain infollowing parts The first is to satisfy the author’s question whether the positive interactionbetween credit growth and stock price has existed as the scientific results of Norway andVietnam’s former studies In the recent context of Vietnam, Government has applied monetarypolicy to stimulate the whole economy after Global crisis Particularly, this is GovernmentStimulus Package of 1 billion US dollars starting from 2009 So wonder if the enforcement

during that time took effect somehow toward stock performance and the entire economy If yes,

it should be employed afterwards The major query is translated into the below hypothesis

l Credit growth has a significantly positive effect on the HOSE index within 12months The hypothesis will be tested using the following equation:

VN-Index = I(Credit growth rate t, t-1, t-2, I-n) in logarithmThe second developed from the controversy whether money market is a maininvestment alternative when stock market is not attractive in short-term As supervised overtime, almost local investors at HOSE often shift their investment to some alternatives such asbank deposit to earn interest or avoid VND depreciation, gold or foreign exchange trade toobtain margin and speculation in real estate with a big capital The thesis will test on thesechannels except real estate because of data limit and the big volume of capital And below isdescribed in term of hypothesis

22

2 In Vietnam, money market is the most active investment substitution for stock

exchange compared to other studies markets in short-run

Finally, the studies will go further to explore the individual relationship of some selectivemacro indicators with stock index It can be informative and useful for policy makers, stockinvestors together with economic scholars then Whether the selected macro factors havepredictable abilities for stock volatility is what the next sections will answer

3 The credit growth rate is a more significant determinant of the stock performance

than the other macroeconomic factors of interest

VN-Index = f(Domestic credit, Exchange rate (VND/USD), Inflation rate, Industrial Production) in logarithm

In summary, these three hypotheses above will be further analyzed by using either economic

theories or quantitative techniques throughout the thesis And the following framework initially presents the overview of all inclusive variables, resources as well as econometric tools

MONTHLY DATA:-IMF

-GSO-Reuters

STOCK INDEX

ESTIMATED STOCK MARKET

PERFORMANCE:

REFERENCE FOR POLICY

MAKERS & INVESTORS

INFLATION RATE

INTEREST RATE

INDUSTRIALPRODUCTION

GOLD PRICE

VAR

MODEL:

UNIT ROOT TEST (ADF) —

STATIONARY CO-INTEGRATION (EG &

CHAPTER 3: RESEARCH METHODOLOGY

To investigate the relationship between stock performance and credit growth for the case

of Vietnam, this thesis establishes three hypotheses (i) there remains a positive short-run andnegative long-run relationship between credit growth rate and VN-index; (ii) Money marketrate has a Granger cause to the variation of stock price at significant level compared with

other factors; (iii) In the long-run multivariate relation, credit factor accounts for high

percentage effect on stock index This chapter is divided into two main sections, researchmethodology and data collection The first section will introduce four econometrictechniques to test for the validity of three proposed hypotheses First, the unit root test is used

to test for stationary of all variables Second, with the bivariate and multivariate cointegrationtests, the hypothesis (i) about the long run relationship between credit growth rate and VN-indexwill be answered Third, the bivariate Granger causality tests are applied to check the directionalrelationship between variables, especially the effect of interest rate on stock change -hypothesis (ii) The bivariate Granger causality test is one of ways to exclude less significantvariable(s) out of the suggested model Fourth, the short run adjustment of stock price is explored

by employing both vector error correction models and Impulse response function to answer forthe hypothesis (iii) Next, the second part will discuss how the thesis deals with proxyvariables, data collection and data analysis

It starts with a summary of time series process through some technical tools employed

in the thesis like Vector Auto-regression, Unit root test for data stationary, Cointegration testfor long-term relationship, Granger test for causality, and Error correction model andVariance

• Decomposition - Impulse Response Function

25

3.1 Econometric techniques:

3.1.1 Analvsis of time series:

According to Asteriou (2007), there are various aspects to examine time series but the

two most common analysis types to fully exploit data structure are time series forecasting

(uni-variate analysis) and dynamic modeling tbi(uni-variate and multi-(uni-variate analysis) Different from

other econometrics the first type does not concern much with building structural models,

understanding economic issues or testing hypothesis but developing forecasting frameworks Byusing diversified criteria (AIC, SBC, RMSE, Correlogram and Fitted-actual value comparison),

it can explore data dynamic inter-relationship over time In contrary, the second type involvesmore than one variable and concerns both economic comprehension and hypothesis testing

However, it can be slow to adjust to external shocks Thus, the process must capture adjustment

techniques Recently, the modeling has become popular thanks to two Nobel works of CliveW.J Granger (Cointegration) and Robert F Engle (ARCH) Nowadays, it has also significantlycontributed to economic policy formulation

One more basic term in almost time series assumption is “stochastic process” which

means a collection of random variables ordered in time, usually denoted as Y, (Gujarati, 2003)

And Stock & Watson’s assumption (2007) stated if the future is like the past, the historical

relationships can be used to forecast the future change Otherwise, it may be unreliable.Therefore, this can be formalized by the concept of stationary in time series regression It is

• usually applied before further causality or cointegration estimations The next part will provide

some popular testing methods of this data stationary

26

3.1.2 Testing for Unit Roots by Correlogram, ADF & PP:

To test the stationary of time series data, we usually apply two typical ways; one is

graphical analysis (Autocorrelation or Correlogram) and the other formal tests (Dickey-Fullerand Phillips-Perron test for unit root)

Supported by Eview software, the data can be analyzed in term of a graph of

autocorrelation for various time lags for a variable (Hanke, 2005) It helps the researchers answer

4 questions whether the historical data is random for forecasting the future, non-stationary,

stationary or seasonal Based on selecting appropriate p (Autocorrelation coefficient) and q(Partial Correlation) in the ARIMA models, we can determine data characteristics for further

estimations

1 If p between Y, and Y,-k for any lag k are close to zero, the series is random

2 If p is significantly different from zero for first several lags and then drops toward

zero when the number of lags increases, it is non-stationary

3 If p is also far from zero and suddenly dies out as lag order rises, the data is

stationary

4 If p occurs at the multiples of seasonal lag, the pattern has seasonal elements

This graph-based correlogram is getting very useful for time series forecasting and otherbusiness implication However, in almost academic studies, it is essential to conduct formal

statistics like t-statistics, Ljung-Box statistics and especially ADF or PP Unit root tests

27

ADF (Augmented Dickey Fuller) is an extensive test procedure from its simple type,

DF, as the error term is unlikely to be white noise The test includes extra lagged terms ofdependent variable to eliminate autocorrelation The lag length on these extra terms can bedetermined by AIC or SIC Below is the general model which should be started in ADF test

Where

o : Intercept

/T : Trend variable

6Yi - i Lagged term of the dependent variable

After answering a set of questions in regard to the model appropriateness, if it is not satisfied, therestricted equations will be tested then

The other approach of PP was developed by generalizing the above ADF test procedure

It allows fairly mild assumptions related to the error distribution

28

The PP creates an adjustment to t-statistics of coefficient from AR (1) regression to

account for serial correlation in ed whereas the ADF corrects this by adding laggeddifferenced terms in model

3.1.3 Cointegration and Error Correction Models (ECM):

Before conducting a short-term causality between two time series, it is necessary to check

their integration and cointegration initially Cointegration can prove whether the series arecorrelated to itself and the other in long-run or not The concept of cointegration was first

introduced by Granger (1981), elaborated by Engle — Granger (1987) and most recently byJohansen (1995) Trended time series, macroeconomic data, can potentially create mainproblems in empirical econometrics due to its spurious regressions Thus, to solve it, we need to

take difference the series successively at I(1) or I(2) until stationary is shown for regressionanalysis But if the estimated data are cointegrated at levels I(0), Error Correction Model (ECM)

should be immediately used in evaluating their dynamic interactions The ECM resultsconveniently explain both the short-run dynamics and long-run equilibrium adjustment of thestudied variables

In case two variables, say X and Y, are related, we would expect their co-movementand then their stochastic trends, cumulate error processes, would be similar to each other Inother word, the trends can cancel to each other In econometric language, we call thecharacteristic a cointegration between X and Y (Asteriou, 2007)

If two variables have a long-term relationship, there will be a common trend linking them

' together And then we can expect a linear combination of Y, and X, which can be directly drawn

from the following regression:

29

, And the equation of residuals:

If in is stationary at level I(0), we can conclude that the two concerned series are cointegrated

From the previous econometric studies, we can generalize about 3 typical procedures of

nonstationary series below:

1 For 2 series integrated of the same order but not cointegrated, VAR can be used for their differenced forms to test short-run relationships Or we can call it Standard version of Granger causality.

2 For 2 series integrated of the same order and cointegrated, we can immediatelyconclude that there is at least one uni-directional Granger causality Then ECM

enables us to estimate the direction of causation and distinguish between long-run and

short-run Granger causality Optionally, it is known as Cointegration and Error

Correction version of Granger causality (or ECM).

3 For 2 series integrated of the different orders (or non-cointegrated), it is suggested

employ Soda and Yamamoto version of Granger causality.

As the expected research result is that the differenced X and Y are cointegrated by its OLS’ error term stationary at I(0) in the case 2, we will provide the review of ECM

specification as follows

’‹ — i + s [3.7]

30

Where

a : the short-run effect ( impact multiplier)

:i : the adjustment effect (error-correction coefficient) measures how much of the

equilibrium takes place each period or how much of its error is corrected Asteriou andHall (2007) explained it in different speed of the adjustment to equilibrium below

n = l means that 100% of the adjustment occurs within the period or it is full andinstantaneous

ii — 0.5 means that 50% of the adjustment occurs within the period

n = 0 means that there is no adjustment within the period

In order to test either long-term or short-term causal relationship, ECM test procedure

was proposed by Engle and Granger (1987) with 4 basic steps:

-Step 1: Test the time series for their order of integration

-Step 2: Estimate the long-term possible cointegrating relationship

-Step 3: Check for the integration order of the residuals

-Step 4: Examine the ECM model

3.1.4 VAR and Granger causalitv test:

According to Sim (1980), Vector Autoregressive Model (VAR) was developed from theidea that all variables are treated as endogenous once no distinction between endogenous and

exogenous ones Suppose there are 2 time series in which Y, is influenced by not only its pastvalues but the past and present ones of Xx Meanwhile, Xt, in turn, has the same impacts Thefollowing bivariate VAR model is given by:

[3.9]

Assume that both Yr and Xt are stationary and error terms up and uxt are uncorrelated

white-noise Since the longest lag length is unity, the above equations constitute a first order VAR

model And they are not reduced-form as the dependent variable has a contemporaneous impact

on the independent and vice versa

According to Asteriou (2007), VAR possesses 3 advantages The first is its simplicitybecause of no worry about the distinction between endogenous and exogenous variables The

second is its separate estimation of each equation with OLS method The last good point is its

forecasting function and foundation for causality tests

However, the model faces some criticism due to its disadvantages It starts with the

weakness of non economics-based theories as restrictions on any parameters are not employed.Statistical inference, yet, is used instead in order to drop insignificant coefficients and makemodels consistent with the underlying theory Next, VAR is argued due to the loss of degrees offreedom Thus, if the sample size is not enough large, the big number of parameters withl2 lags

32

for each will consume the degrees It may cause problems in study The other dispute focused on the interpretation of its obtained coefficients due to the lack of theoretical background.

As above-mentioned, VAR helps test the direction of causality Granger (1969)composed a simple causality test which has been widely applied in economic analysis Heassumed that Yr Granger-causes Xi provided that Xt «an be predicted with more accuracy bythe past values of Y„ all other terms hold unchanged We can formulate the two equationsbelow

where error terms are uncorrelated white-noise

Suppose Yr and Xt are not stationary, the Standard Granger Causality is performed with below procedure

Step 1: Testing for the unit root of time series (ADF or PP tests)

Step 2: Testing for cointegration between them (EG or Johansen approach)

Step 3: Taking the first differences of both (i.e., AYt and AX,)

Step 4: Determining the optimal lag length of the US-differenced variables

Firstly, supported by econometric software (Eview), we can automatically determine the optimal lag length of AY, and AXt in) based on AIC or SIC criteria

After that, run OLS, obtain the RSS of these regressions (restricted ones) and label it as

RSSp and RSSp

Secondly, manually identify the lag lengths of both AY, and AXt <) from unrestrictedequations based on AIC or SIC criteria

Similarly, run OLS, obtain the RSS of these regressions (unrestricted ones) and label it as

Step 5: Setting the null hypotheses

For (1) and (3): H0: y, =0 or X, does not cause Yr

For (2) and (4): H0: ,=0 or Yr does not cause X,

Step 6: Calculating F-statistic for the normal Wald Test on coefficient restrictions by

34

1.5 VAR-based Variance Decomposition and Impulse Response Function:

The approach helps to identify how much each economic variable affects on itself and the

other independent variables in term of percentage of the factor’s forecast error variance inlong run period And Impulse Response Function can point out the relationship signs amongall employing variables according to Lastrapes and Koray (1990)

3.2

Data description:

Almost monthly secondary data (2002M1-2010M3) will be collected from reliable resources of

IMF, GSO and Thomson Reuters Specifically, the study will use Industrial Production as a

proxy for GDP from GSO In general, there are 99 monthly observations for the empiricalresearch And Vietnam stock performance in this study will be specified in form of a function ofthe selective variables below:

log VNI = I(log CR, log FX, log RAT, log INF, log IP, log DEP, log DOW, log SSE, log

GOLD)[3 16]

35

Nominal exchange rate VND/USD

Month-end money market interest rate

Inflation rate as a proxy of consumer price level (CPI)

Industrial Production calculated on monthly basic from annual Gross Domestic Production

Domestic deposit aggregate

DOW: Dow Jones Index representing Stock price index of US — a leading mature stock

market in the West and alternative investment channel of foreign investors.The

Dow Jones Industrial Average is a price-weighted average of 30 blue-chipstocks that are generally the leaders in their industry It has been a widelyfollowed indicator of the stock market since October l , 1928 (Bloomberg)

SSE: Shanghai stock price index as proxy for China security market — a representative

in the Eastern emerging market and alternative investment channel of foreigninvestors Prepared and published by SSE, SSE indices are the authoritativestatistical indices widely followed and quoted at home and abroad in measuring

the performance of China's securities market SSE Index Series consists of 17

indices, including 5 constituent indices, 2 composite indices, 7 class and sector

indices and 3 other indices In order to promote the long-term infrastructure

construction and standardization process of the securities market, in June 2002,SSE restructured the original SSE 30 Index and renamed it SSE Component

Index, or SSE 180 Index

GOLD: Month-end domestic gold price

All these variables are summarized in the table 3 l below

Table 3.1: Variables summary

Dependent Variable

Ind endent Variables

3.3

Data aaaJvsis:

Time-series data of both dependent variable (VN Index) and independent macro factors

will be employed in order to identify their stationary via Augmented Dicky-Fuller and PhillipsPerron Unit Root Test; their lag length via the Schwarz Information Criterion (SIC) or Akaike

Information Criterion (AIC) and the causality Granger & ECM tests

37

The main methodology throughout the study is the application of VAR model Then the

econometric results through quantitative analysis present the implication of every interaction

among macro factors, credit growth and stock price fluctuation

Depending on data availability, the study uses quantitative analysis to test hypothesizes

First, this research uses Unit Root (ADF and PP) in Eview to test whether 2 time series of

stock performance and credit aggregate are stationary and co-integrated in first difference or not

As there are 2 variables tested each time, it is better off to apply EG approach instead of

• Johansen approach It can be broken down into 4 sub-steps below

Stepl: Run OLS of the equation

Step2: Generate residuals from these OLS equation, namely E to examine the property ofregression residuals

Step3: Use ADF test to regress the changes of E (namely d(E)) on lagged E (namely 1)) and the lagged term d(e(- l))

e(-Step4: Conclude that stock index and credit growth rate are co-integrated if DF test

statistic is less than the critical value at 5% or 10% It means we reject null hypothesis of noco- integration

In case it can satisfy the VAR prerequisite condition, we will go ahead to examine their

causality

Second, Error Correction version of Granger causality will be used to test therelationship between HOSE stock performance and Credit growth rate (2002-2010) As theroutine VAR assumes both lags of independent and dependent variable are the same Thiskind of selection would lead to model misspecification Therefore, optimal lag length should bemanually determined based on min value of AIC in restricted AR models

Then estimate the above equations by OLS and obtain dsso and ??SSRC

Next, we construct their unrestricted equations and similarly get fissci and RSStic

At this step, we will set the null hypothesis and calculate the F statistic for the normalWald Test to define that CR causes VNI Alternatively, in the thesis, we choose the automaticprocedure in Eview to identify bivariate Granger cause of the other 8 employed variables below

to stock index

39