AN APPLICATION OF FAMA FRENCH 3 FACTORS MODEL ON HOSE

1. INTRODUCTION1.1. Overview of ResearchThe Viet Nam stock market has been formed since July, 1998. Having some certain achievements, there is still riskiness and weakness. Currently, in Viet Nam, most of investment decisions of individual investor and business were based on recommendations of securities companies which using discounted cash flow method or relative method. However, with the current volatility of the market, these methods proved to be inefficient and could not predict the real market price of stock. So that investors could not make decisions in a more flexible way. Therefore, the study of the application of modern financial theory of investment in Viet Nam stock market in the current period is very important and urgent. Moreover, there were several researches in the world in applying of the theory of financial investment in the stock market, especially the empirical research on the stock market in emerging countries. These researches gave significant and extremely practical results. It reinforces more accuracy and empirical models. Seeing the need of the application of the asset pricing model to predict the stock market, I decided to investigate research topic: AN APLICATION OF FAMA FRENCH 3 FACTORS MODEL ON HOSE (HO CHI MINH CITY STOCK EXCHANGE).The Fama French 3 factors model has been one of the most widely used techniques in the global investing community for calculating the required return of risky asset but it seems very new in Viet Nam. This is the reason that motivates me empirically analyse the Fama French model in the context of Viet Nam stock market.This project aims to test whether Fama French model is a valid model for estimating risk and return of stocks listed on the HOSE.The study’s result might be useful recommendations for investor about how and what Fama French can be used for predicting risk and return and making investment decision in Viet Nam stock market.

JEAN MOULIN LYON 3 UNIVERSITY VIETNAM UNIVERSITY OF COMMERCE

MASTERS FINANCE AND CONTROL

THESIS

AN APPLICATION OF FAMA FRENCH 3 FACTORS MODEL ON HOSE

Supervised by: OUSSAMA LABILI

Hà Nội - 2013

I am sincerely thankful to my supervisor Oussama Labili from the Jean Moulin Lyon 3University, for his encouragement, guidance and feedback from the beginning to the finalstage

I would like to show my great gratitude to Mr Vu Manh Chien, Mr Vu Duc Khuong, MrNguyen Viet Dung for their detailed comments and valuable suggestions

In addition, I would like to show my gratitude to my classmates, Tuyet Dinh Thi Anh amongothers, for their useful discussions and assistance

I would like to express my sincere thanks to all of my lecturers for their teaching andguidance during my master course at the Jean Moulin Lyon 3 University and Viet NamCommercial University

Finally, I am also grateful to my family for their emotional, moral as well as financial support.Responsibility for any remaining errors lies with the author alone

1 INTRODUCTION 5

1.1 Overview of Research 5

1.2 Research Objectives 6

1.3 Research Structure 6

1.4 Previous Research in Viet Nam 6

2 LITERATURE REVIEW 8

2.1 The Fama French 3 factors model 8

2.2 Extended Fama-French 3 factors model- Carhart (1997) four factor model 10

2.3 Empirical evidence of Fama French 11

2.3.1 Empirical of Fama French model in developed markets 11

2.3.2.Empirical evidence of Fama French model in emerging markets 14

3 DATA 14

4 METHODOLOGY 16

5 EMPIRICAL RESULT 19

5.1 Test Result Summary 19

5.1 Testing for Errors 21

6 CONCLUSION 23

6.1 Research Summary 23

6.2 Contributions, limitations and recommendations 23

6.2.1 Contributions 23

6.2.2 Limitations 24

6.2.3 Recommendations 24

REFERENCES 26

APPLENDIX 1: 120 TICKER LISTED IN HOSE FROM JUL 2008 TO JUN 2013 30

APPLENDIX 2: RISK-FREE RATE (VIETNAM 2 YEAR BENCHMARK BOND – BID YIELD) AND VNINDEX 36

APPLENDIX 3: PORTFOLIO RETURN (ALL 120 TICKER, SH, SM, SL, BH, BM, BL) 37

APPLENDIX 4: TEST RESULT 40

HOSE : Ho Chi Minh City Stock ExchangeHASTC : Ha Noi Stock Trading CenterCAPM : Capital Asset Pricing Model

P/E : Price to Earning

BE/ME : Book equity/Market Equity

HML : High minus Low

SMB : Small minus Big

WML : Winner minus Loser

IPO : Initial public offering

SEO : Seasoned equity offering

HNX : Ha Noi Stock Exchange

1 INTRODUCTION

1.1 Overview of Research

The Viet Nam stock market has been formed since July, 1998 Having some certainachievements, there is still riskiness and weakness Currently, in Viet Nam, most ofinvestment decisions of individual investor and business were based on recommendations ofsecurities companies which using discounted cash flow method or relative method However,with the current volatility of the market, these methods proved to be inefficient and could notpredict the real market price of stock So that investors could not make decisions in a moreflexible way Therefore, the study of the application of modern financial theory of investment

in Viet Nam stock market in the current period is very important and urgent Moreover, therewere several researches in the world in applying of the theory of financial investment in thestock market, especially the empirical research on the stock market in emerging countries.These researches gave significant and extremely practical results It reinforces more accuracyand empirical models Seeing the need of the application of the asset pricing model to predictthe stock market, I decided to investigate research topic: AN APLICATION OF FAMAFRENCH 3 FACTORS MODEL ON HOSE (HO CHI MINH CITY STOCK EXCHANGE).The Fama French 3 factors model has been one of the most widely used techniques in theglobal investing community for calculating the required return of risky asset but it seems verynew in Viet Nam This is the reason that motivates me empirically analyse the Fama Frenchmodel in the context of Viet Nam stock market

This project aims to test whether Fama French model is a valid model for estimating risk andreturn of stocks listed on the HOSE

The study’s result might be useful recommendations for investor about how and what FamaFrench can be used for predicting risk and return and making investment decision in VietNam stock market

1.2 Research Objectives

With the motivation of specifying appropriate measurement for investors to evaluate risk andreturn, this study’s main objective is to evaluate the explanatory power of this model in VietNam stock market While most empirical tests of this models’ validity conducted with HOSEdata Therefore, this study will re-examine the forecasting capability of Fama French 3 factorsmodel for 120 companies listed on HOSE which is listed from July 2008 to June 2013 Thestudy will examine the effect of market premium, size premium and book value/market valuepremium to the stock and portfolio It is expected to estimating risk and return of stock andportfolio during the period of down-turn to find whether this model is suitable to evaluate cost

of equity in Viet Nam stock market or not?

In final section, summary and limitations of the study are provided

1.4 Previous Research in Viet Nam

In Vietnam, there are few previous studies on capital asset pricing model, as Vietnam stockmarket has been formed 15 years ago So I will summarize some typical researches on CAPMand Fama French in Viet Nam

In 2009, Doctor Quach Manh Hao, a lecturer at the National Economics University, has aresearch “Finding an asset pricing model in Viet Nam” posted in Financial Market imagine

He collected 50 stocks on HOSE and HASTC from 2007 to 2009 He found that there were 4most important factors that effect to stock return: Rm, P/E, liquidity and company size Heapplied Fama French model and found the result: R2: 55% and confidence level: 95%

In 2010, Vuong Minh Giang has a research ''An Empirical Examination of the Validity of theCAPM in the Vietnamese Stock Market” at Social Science Research Network The authorcollected data from 30 largest capitalization stocks in the period 2007-2009 The linear

relation between risk and rate of return in Vietnam stock market is tested for the falling period2007-2009 based on adjusted-price data As a result, the failure of the test shows that VN-Index, conventionally regarded as the "market portfolio", is not mean-variance efficient to theasset set being examined By optimization calculation, a general envelope containing theefficient frontier of the stock set in the Vietnamese case of short-sales restrictions is produced.Finally, some remarks are noted for stock pricing practice in the emerging market.

In 2011, a group of students in Ho Chi Minh Economics University conduct a research,namely “Application CAPM and Fama French Model in Viet Nam stock market” The authorscollected data from 88 stocks from 2007 to 2010 and reach a result as follow: R2 is 90.56%and confidence level is 95%

In my study, I would like to confirm that I cannot create a new model but basing on the result

of previous researches in Viet Nam and Fama French methodology, I want to conduct aintensive study with more observation in a longer period, through a bigger portfolio (120tickers) to find out more useful results

2 LITERATURE REVIEW

2.1 The Fama French 3 factors model

The capital asset pricing model is a set of predictions concerning equilibrium expected returns

on risky assets Harry Markowitz laid down the foundation of model portfolio management in

1952 The CAPM was developed 12 years later in articles by William Sharp, John Lintner andJan Mossi The time for this gestation indicates that the leap from Markowitz’s portfolioselection model to the CAPM is not trivial

CAPM tests the relation between market risk and asset return, assuming that investors arerisk averse and only mean and variance of their assets return are occupied Thus, the portfolio

is chosen if it can minimize loss, given expected return or maximize return at a certain level

of risk

The CAPM equation is as follows:

 Denote E(ri) and E(rM) as expected return of asset i and market portfolio, respectively

 Rf is the return of risk-free asset

 is the relationship between excess expected return of asset i and that of market portfolio as a fraction of total variance of market portfolio

In Roll’s critique, he also points out a number of drawbacks of CAPM such as its implication

of market efficiency or formation of market portfolio which challenge the practical of themodel Roll and Ross (1994) and Kandel and Stambaugh (1995) support for Roll’s view.They argue that CAPM provides an adequate evaluation of beta only in an efficient marketwhich is hardly measured in the real world Thus, it is uncertain about the predictingcapability of CAPM Besides, numerous studies suggest other factors which might outweighbeta in estimating companies’ return such as price to earnings (Basu, 1977), firm size (Banz,

1981 and Reinganum, 1981) Findings of other significant variables in explaining rate ofreturn imply the need to incorporate additional factors to modify CAPM (Lawrence et al,2007)

Fama and French (1992) confirm the insufficiency of CAPM by illustrating a weak relationbetween market beta and average US return of the period 1963-1990 by cross-sectional test.They point out other variables, namely E/P, leverage, firm size (market equity) and book tomarket equity (hereafter BE/ME) can explain the stock returns Combining these variables,two latter variables seem to absorb explanation power of the two former In short, firm size(market equity) and BE/ME could enhance CAPM’s efficiency in pricing stocks.

Fama French three factors model is presented as follow:

term; , , are parameters in the regression

The ex-post regression of Fama French three factors model is given by:

Where , are excess asset and market return, respectively

Fama and French’s later study (1993) expands the previous one in three dimensions First,they involve US government and corporate bonds in the set of tested assets instead of stockonly in Fama and French (1992) Second, one more variable, term-structure, which plays animportant role in bonds’ value, is added In their theory, bonds and stocks might be closelyrelated as markets are integrated Hence, they try to test whether a crucial variable of bondshas significant explanation power to stocks’ returns or not, and vice versa Third, themethodology used to examine assets’ returns is different If bonds is included in the cross-sectional regression, size and BE/ME variables have no relation to bonds Therefore, Black,Jensen and Scholes (1972)’s approach, a time series regression, is adopted while cross-sectional test in Fama French (1992) might be unsuitable in this case Portfolios are thereforeformed on size and book to market equity It is suggested that size factor be equal to themonthly average return of small companies minus that of big companies, denoted by SMB(small minus big) The other explanatory variable, book to market is the difference between

return of stocks with high book to market ratio and that of stock with low book to marketratio, given by HML (high minus low)

Fama and French (1995, 1996a) continue to prove the outperformance of Fama-French threefactors model compared to CAPM, using time series regression They point out the trend thatstrong firms with high earnings usually have low book to market equity and negative slope onHML In contrast, weak firms with low earnings tend to have high book to market equity andpositive slope on HML

2.2 Extended Fama-French 3 factors model- Carhart (1997) four factor model

Mark Carhart (1997) continues the momentum issue studied by Jegadesh and Titman (1993)

He constructs a risk factors relating to momentum effect (WML- Winner Minus Loser) andmakes a four factors model by adding momentum factor into the Fama French 3 factorsmodel The momentum effect is the effect that past winner or loser continues performing well

or poorly The WML factor is measured by the return of winner stocks portfolio minus thereturn of loser stocks portfolio As momentum strategy, the investors buy the stock withhigher returns as well as sell stocks with lower return over the previous 3 to 12 months couldlargely generate returns in the stock market

The theoretical model of 4 factors regression is as follow:

Where: WML is the return difference between past 1 year winner and loser portfolios, isparameters in the regression and the others are the same in Fama French model

From the result of Carhart’s study, we get the flowing suggestion We should not invest in thefunds that their rate of returns always are negative The funds that having the high return inthe year before will continuously generate the higher return in the year after, but it is notaccurate in a long term Management cost, total net asset, investment cost effect directly andinversely to fund’s return and take away the excess return of the fund having high return inthe year before in a long term

Toward the result of Carhart (1997), Daniel et al.(1997) and Wermers (1997) find evidencethat four factor model of Carhart (1997) does well in investigating the strategies that drive thepersistence in mutual fund performance Brave et al.(2000) reveals that the four factors model

of Carhart could explain the underperformance in return from a sample of IPO (initial publicoffering) and SEO (seasoned equity offering) companies.

In fact, results from some research in the global indicated that the four factors model havesignificant exposure in explaining the variations in average excess stock return R2 fromCarhart model is just slightly higher than Fama French model

In Viet Nam, a group of students in Ho Chi Minh Economics University (2011) conduct aresearch, namely “Application financial models in pricing investment portfolio in Viet Namstock market” They find that four factor model of Carhart have higher capability inexplaining the variations in average excess portfolio return R2 in CAPM, Fama French andCarhart is respectively 71.66%, 86.23%, 88.41%

2.3 Empirical evidence of Fama French

2.3.1 Empirical of Fama French model in developed markets

Evidence of Fama French three factors model can be seen mostly in US market such as Famaand French (1992, 1993, 1995, 1996a, 2006) In these studies, Fama and French showstatistically insignificant of the CAPM factor in estimating cost of equity during differentperiods of time in the US market Meanwhile, size and book to market factor are proved tohave a significant role in explaining asset return (ibid) In contrast, examining the validity ofthe two models in the US, Bartholy and Pearre (2005) find out that there is no significantdifference between their estimates Thus, they raise the issue that whether or not the other twovariables of Fama French should be adopted while they insignificantly contributed to cost ofequity evaluation

According to Drew and Veeraraghavan (2003) and Artmann et al (2012), the efficiency of anasset pricing model is strongly proved if its evidence is shown in sufficient new marketsrather than only in the biggest one Using Australia data for the period 1981-1991, Halliwell

et al (1999) show the significance of Fama French factors in explaining Australian averagereturns But the role of BE/ME is not as powerful as it has been proved in Fama French’sstudies Faff (2004) continues to research the validity of the Fama French model in Australiamarket, dealing with daily data of industry portfolio from 1 May 1996 to 30 April 1999 Hisfinding basically supports Fama French factors In the case that an estimated risk premium isconsidered, robustness of the model is decreased Expanding the examined time from 1982 to

2006, Brailsford et al (2012) prove the superior performance of Fama French model in

comparison with CAPM in evaluating Australia stocks Their empirical test results show thestatistical significance of book to market factor and the opposite in case of size effect inexplaining assets’ returns

However, studying Germany firms’ data for the period from 1969-2002, Schrimpf et al (2007)support the traditional asset pricing model, CAPM, rather than the modern one, Fama French

in the sense that it provides less error than its counterpart Also dealing with German marketbut for longer stage, 1960-2006, Artmann et al (2012) reject the explanatory power of bothCAPM and Fama French model There is limited statistical evidence on both size and book tomarket effect in German stock returns (ibid)

Chan and Chui (1996), when testing Fama French model for UK companies during the period1971-1990, found out the significance of book to market factor but insignificance of firm sizefactor in explaining stock returns Meanwhile, Zhang (2009) and Chabi-Yo and Fourseni(2009) demonstrate the strong capability in estimating UK stock returns of Fama-Frenchfactors Testing UK companies between 1975 and 2000, Hung et al (2003) combine FamaFrench factors with the CAPM, then examine whether the additional factors supportexplanatory power of CAPM beta or not They conclude that both CAMP beta and FamaFrench factors are significant in estimating the cross-section of UK stock returns Similarly,using UK data (1992-2001), Charitou and Constantinidi (2003) sum up with theoutperformance of Fama French model over the CAPM

Various applications of the model result in conflicting conclusions about the explanation ofFama French model Kothari et al (1995) argue that book to market has a marginal relation toexpected stock returns In their points of view, Fama and French (1992, 1993)’s results aredue to survivor bias in COMPUSTAT data This data source usually includes historical firms’data in which survivor companies used to have unexpected higher return than the dead ones.Thus, there might be a positive bias for the former (high book to market equity) incomparison to the latter On the contradictory, Chan et al (1995) and Fama and French(1996b) illustrate that survivor bias plays no role in explaining the relation between averagereturn and book to market equity In their arguments, omissions in COMPUSTAT data arefinancial institutions which have higher leverage than other types of firms The missing firms

do not cause survivor bias Additionally, Fama and French (1993) use value-weightedportfolios in their test Survivor bias does not matter in these portfolios as high weights aregiven to large companies

Black (1993) and MacKinlay (1995) state that the value premium of Fama-French’s factorsstems from data-snooping, whereas Lakonishok et al (1994) suggest, it is due to behavior ofinvestors As investors are used to buying stocks which are big and well-performed in thepast, thus, overprice these stocks Conversely, they tend to sell stocks which might be smallbut have high growth rate Thus, these stocks are underpriced La Porta (1996) shares thesame idea that investors might favor value stocks as they concentrate on these firms’ enduringdevelopment rather than those with rapid growth rate Fama and French (1992) suppose thatinvesting in growth stocks which are small and have high book to market ratios should rewardhigher returns as bearing higher risks.

Extending Fama and French (1992)’s research, Knez and Ready (1997) also apply Fama andMacbeth procedure in their study They compare a standard least square regression with aleast trimmed square regression which monthly omits 1% most extreme observations Theresults suggest that these trimmed observations are reasons for negative relation between firmsize and its average return Without them, coefficient of size in the regression for stocks’return is significantly positive They share the same idea with Fama French (1992)’sexplanation that investors who bearing risk investing in small firms should reward highreturns It is suggested that there should be systematic driven forces for risk premium whichalso help explaining firm’s development process Vassalow (2003) points out a significantfactor in explaining stock return that is news related to future GDP growth His studyillustrate that including the variable in CAPM model dramatically increases its explanatorypower Moreover, the existence of news related to future GDP growth might outweigh HMLand SMB role in pricing equities Thus, there might be a close relation between HML, SMBvariables and news about future GDP development

Initially stated by Banz (1981), size effect has been found in all common stocks listed onNYSE from 1926 to 1975 In his study, small firms are proved to bear higher risk than largeones These former stocks provide significantly higher average returns than the latter ones.According to Banz (1981), one possible explanation for the phenomenon is limitedinformation of small firms which leads to higher risk-adjusted returns for holding thesestocks Studying the same topic, Chan and Chen (1991) examine risk and return characteristic

of 19 industry groups on NYSE during the period 1956-1985 In their findings, productionefficiency, leverage and the ability to absorb external capital cause higher risk of small firmsthan large ones Zarowin (1990) also concludes a size effect in stocks’ returns when re-

examining the overreaction hypothesis stated by DeBondt and Thaler (1985) The hypothesis

is about the outperformance of loser to winner stocks Following Zarowin (1990), losers tend

to be small stocks and vice versa When winners are smaller than losers, they provide higherreturns than the latter

2.3.2.Empirical evidence of Fama French model in emerging markets

There is limited evidence on Fama French model’s validity in Asian countries As Drew andVeeraraghavan (2003) has observed, Chui and Wei (1998) are first examining the validity ofFama French in Asian market Their finding proves the significance of two more factors inFama French, firm size and book to market In their study, the relation between market betaand average return is weak, as opposed to that of Fama French factors Similarly, Drew andVeeraraghavan (2002, 2003) once comparing the validity of two models in emerging markets(Malaysia, Hongkong, Korea and Philippines) during the 1990s, point out the outperformance

of Fama French over CAPM model More recently, Taneja (2010) also confirms strongexplanatory power of Fama French in comparison with CAPM in Indian stock market, fromJune 2004 to June 2009 His result shows that variations in stock average return can becaptured better by multifactor Fama French model than single factor model CAPM UsingShanghai Stock Exchange data, Lin et al (2012) illustrates the superior explanation power ofFama French factors in estimating portfolios’ returns in China market However, it is shown

in their study that market factor seems to be more appropriate in evaluating individual stock’sreturn

3 DATA

Kothari, Shanken and Sloan (1995) report that beta from annual returns produce strongerpositive relation between beta and average return than beta from monthly return But Fama-French (1996) prove the contrast conclusion that annual and monthly returns have sameinferences about beta premium

This study uses historical monthly data of 120 Viet Nam companies based on the followingcriteria:

 Listed in HOSE at least 60 months (before July 31 2008) De-listed companies andthose which are listed after July 2008 are excluded to maintain the continuation ofportfolios

 Banks, insurance companies, securities firms are excluded because of their distinctivehigh leverage.

 Fiscal year ended on 31 December

All stock price used in this document are adjusted close price (A stock's closing price on anygiven day of trading that has been amended to include any distributions and corporate actionsthat occurred at any time prior to the next day's open)

VNIndex, July 2008 to June 2013 Data Source: cafef.vn

These figures, adjusted prices, shares outstanding, owner equity as well as other companies’financial information are obtained from http://cafef.vn and Thomson Reuters Eikon

After collecting all data basing above criteria, I regress data on monthly return of 120 testingstocks in 60 months, from July 2008 to June 2013

Vietnam 2-Year Benchmark Bond, July 2008 to June 2013 Data Source: Thomson Reuters

Risk-free rate of return is Vietnam benchmark bond Data on the risk-free interest isstatistically Bid Yield of 2-Year Viet Nam benchmark bond, because these bonds are tradedwidely and have the highest liquidity in the market

The sample is divided in a way that the proportions of small and big size, high and low B/E are equal Our definition of Small size (S), Big size (B), High BE/ME (H), Medium BE/ME (M), Low BE/ME (L) as following:

The dependent variable of the regression is the excess return of company (i) in month (t) Butbesides market factor, size (SMB – Small minus Big) and book-to-market (HML – Highminus Low) factors are two additional explanatory variables The exposures of firm (i) excessreturn to those risk factors are denoted by βiM, βiSMB and βiHML correspondingly

Additional Fama-French factors are formed basing on six value-weight portfolios constructed

by size and book to market ratio With regards to size, companies included in the sample will

be divided into two groups following its market capitalization rank, i.e big and small Themedian is used as a size breakpoint The book to market ratio or year t is computed bydividing book value at the beginning of testing period t with market value at the same time ofyear t

By ordering individual company’s book-to-market ratio, tested companies are categorized intothree groups: low, medium and high 30% bottom of the order contributes to the low groupwhereas 30% top belongs to high group The intersections of two-size and three book-to-market ratio groups create six portfolios, namely Small High, Small Medium, Small Low, BigHigh, Big Medium and Big Low For example SL is a portfolio of companies with small sizeand low book value/market equity

Determine and calculate SBM and HML Variable

SMB (Small minus Big) represents the risks related to the effect of the size factor SMB is theaverage monthly return of three small portfolios minus average return of three big portfolios,given by:

SMB =1/3(Small High + Small Medium + Small Low) – 1/3(Big High + Big Medium + BigLow)

HML (High minus Low) represents the risks related to the effects the book value/marketvalue factor on stock return HML is the average monthly return of two high book-to-marketratio portfolios minus average return of two low books to market ones.

HML = 1/2 (Small High + Big High) – 1/2 (Small Low + Big Low)

Rate of Return of portfolio:

Average monthly rate of return of the portfolio: based on the daily closing price of the stock,

I compute the return for each stock Because Viet Nam stock market has been formed sinceJuly 1998 so that the data to compute yearly return is not enough Daily return computing isnot chosen due to fluctuation amplitude effect This study uses historical monthly data of 120Viet Nam companies listed on HOSE over a 60-month-period, from July 2008 to June 2013 The return of each stock is computed by the flowing formula:

Where is the return of stock i at period t+1, is the market price of stock i at the end ofperiod t, and is the market price of stock i as the end of period t+1

Since then, the portfolio return is calculated by the average monthly return of all stock in theportfolio

Market Return and Free Rate Return:

Rf is rate of return earning by investing 2 years Viet Nam government bonds Rm is marketreturn that testing stock listed Therefore, in this research, I choose VNIndex to calculate Rm

for the stocks listed on HOSE

Market return is computed by the flowing formula:

R m

= VNindex t+1 - VNindex t VNindex t

Where is market return (VNIndex return), VNindex t is the value of VNIndex as the end of period t, is the value of VNIndex as the end of period t+1.

5 EMPIRICAL RESULT

5.1 Test Result Summary

We will test the time regression of three-factor Fama-French model presented in equation (1)

by running the Ordinary Least Squares tests using Gretl with 7 portfolio: allTicker-Rf (the portfolio of all 120 tickers), BH, BM, BL, SH, SM and SL The detail test results are

presented in Appendix 4: Test Result

We have the following summary table:

Dependent

Variable Independent Variables alpha beta_M beta_SMB beta_HML R2 Average R2

allTicker-Rf

HML, SMB, Rm_minus_Rf

First, we run regressions on the portfolio of 120 securities (allTicker-Rf) and we found that

= 89.43%, and the confident level of HML and Rm_minus_Rf is 99% However, the

confident level of SMB is only 90% The SMB having lower statistical significance indicatesthat it is less significant difference between small and big size performance, while the highBE/ME performance are significantly different from those of Low BE/ME As a result, theHML has higher predictive power than the SMB in forecasting excess return

Continue to conduct regression by small portfolios, which are divided based on the scalefactor and the ratio of BE/ME Preliminary assessment, in large group (BH, BM, BL), SMBregression coefficients showed no statistical significance, the remaining coefficients are verygood reliability

In small group (SH, SM, SL), only in the SL portfolio, HML regression coefficient is notstatistically significant

In 6 portfolio, SH has the highest R2 (90.52%), SL has the lowest R2 (70.95%)

Market risk premium can be explained from 70% to 90% change in the stock return with 99%confidence level Average of 6 portfolios is 82.7%, of the portfolio of 120 securitieswas 89.4%.

Alpha in all 7 model are close to zero and having statistically significant in 6 sub portfolios.The result reflecting the fact that 3 independent variable in the models have proved it power

to explain the excess return

Due to the small numbers of tickers in each 6 sub portfolios, the result of the 6 regressions may be bias to the specific performance of each sub portfolios However, the allticker portfolio (120 companies) provided us more reliable result in term of statisticalmeaning and the representativeness of Vietnam stock market We conclude that the SMB,HML, and market excess return are 3 significant variables, which are able to explain 89% ofspecific excess return in Vietnam stock market with confidence level of 90%, 99% and 99%respectively

sub-5.1 Testing for Errors