Baoding, Hebei, China. *Corresponding author

Transcription

1 rd Inernaonal Conference on Economcs and Managemen (ICEM 2016) ISBN: Research on he Applcably of Fama-French Three-Facor Model of Elecrc Power Indusry n Chnese Sock Marke Yeld Predcon Ze-hong LI 1,a, Jun-je PAN 2,b,* 1,2 Deparmen of Economcs and Managemen, Norh Chna Elecrc Power Unversy, Baodng, Hebe, Chna a @qq.com, b @qq.com *Correspondng auhor Keywords: The Capal Asse Prcng Model (CAPM), Fama-French Three-Facor Model, Elecrc Power Indusry, Rae of Reurn, Regresson Analyss. Absrac. Fama-French Three-Facor model (FF3), as a represenave of mulfacor asse prcng models, s an mprovemen of CAPM. For a long me, Chna s elecrc power enerprses have wdely used he CAPM n he process of nvesmen decson-makng. Whle he nernal and exernal envronmen of Chna s elecrc power ndusry s undergong grea changes n recen years, he applcably of he CAPM need furher research. Ths arcle selecs he sock daa of Chnese elecrc power ndusry n he sock marke durng , hen uses CAPM and Fama-French hree facors model respecvely o conduc regresson analyss wh he sock yeld daa, and compare he applcably levels of he wo models, hus access he applcably of Fama-French hree facors model, he purpose s o fnd a suable sock marke yeld predcon model for he presen Chna s power ndusry. 1. Inroducon The capal asse prcng model (CAPM) s he mos wdely used asse prcng model, whch depcs he lnear relaonshp beween rsk and reurn, and have been wdely acceped by he researchers a home and abroad n erms of secures marke yeld predcon. A presen, he CAPM s sll wdely used n he equy capal yeld predcon as well as he consrucon projec nvesmen decson-makng n Chna s elecrc power ndusry, as he nernal and exernal envronmen of Chna s power ndusry has experenced grea changes n recen years, he applcably of he CAPM needs furher sudy. A he same me, f he CAPM no longer apply, we also need o fnd more evdence of he applcably of Fama - French hree facors model and oher mulfacor models. 2. The Inroducon of FF3 The bgges advanage of he CAPM s o buld a lnear relaonshp beween rsk premum and he expeced reurn, because of s smplcy and praccably, s wdely regarded by boh a home and abroad. Bu a he same me, he model also rases many quesons, o sum up, he quesons of he CAPM manly reflecs n he followng wo aspecs [1,3]: (1) The premse of CAPM s overmuch, complee capal marke hypohess s dffcul o acheve. CAPM model s based on several assumpons, and hese assumpons, ncludng perfec compeon marke and raonal agen hypohess, ec., are dffcul o acheve. (2) The CAPM only consders he nfluence of sysemac rsk of sock reurns, whch canno effecvely explan he cross-seconal dfferences of sock reurns. In vew of hs, n 1993, hrough he emprcal sudy of he U.S. sock marke, Fama and French pu forward FF3 whch ncluded he marke rsk facor, sze facor (SMB), and book-o-marke facor (HML) [4]. Laer n he sudy, by he nspecon of he sock marke of he Uned Saes, Ausrala, ec., Fama and French found ha here were sze premum and value premum n sock

2 markes around he world, whch proved he correcness of he model. In vew of FF3, a lo of research have been conduced, he concluson also generally suppor Fama and French s sudy. The model expresson of FF3: ( R R ) + s SMB+ h HML R = R f + β m f (1) In he formula, R s rsk premum rae of he sock, f Rm R f s he rsk premum n equy marke. β s bea coeffcen of he sock, whch s measuremen of he marke rsk of ha sock. s and h are he coeffcens of SMB and HML. 3. Sample Selecon and Research Desgn 3.1 Sample selecon R s he rsk-free rae of reurn, ( ) The purpose of hs paper s o sudy he applcably of he wo knds of asse prcng models of he yeld predcon of he elecrc power ndusry n Chna, n order o guaranee he accuracy of he emprcal research resuls, we should gve full consderaon o he he nfluence of ndusry envronmen changes o he applcably of he model, herefore, we wll conduc our research from 2007 o Reasons are as follows: (1) In 2002, Chna offcally sared markezaon reform of elecrc power sysem, nroduced compeon mechansm no he organzaon form and prce model of elecrc power enerprse. As of Aprl 2005, he consrucon of a new elecrcy prce sysem bascally compleed, and major achevemens had been made n he markezaon reform of elecrc power sysem. Ths reform weakened he monopoly of he power ndusry, ncreased he unsysemac rsk n he enerprse operaon. (2) In order o solve he problems of he crculaon of non-radable shares, Chna sared he reform of nroducng shareholder srucure n lsed companes n Aprl As of March 26, 2007, he reform had been compleed n all 53 elecrc power enerprses ha lsed before The shares rao of non-radable shareholders n lsed power enerprse fell by 4.90% on average aer consderaon compared wh ha before consderaon, on he bass of realzng he crculaon of non-radable shares, ndusry equy srucure has become more reasonable. In he process of daa collecon, he sample s colleced as follows: Durng , aer hese reforms, we seleced socks of 33 power enerprse, whch had been lsed n Shangha and Shenzhen A sock marke before December 31, 2006, as he research objec, among whch he ST and *ST companes and he lsed enerprse whch had defc ne worh had been elmnaed durng he sudy perods. The relevan sock ransacon daa came from CSMAR daabase and oher Inerne nformaon. 3.2 Research desgn (1) Model Consrucon and varable explanaon Ths arcle adops he mehod of comparave analyss. We wll buld he radonal capal asse prcng regresson model and he FF3 regresson model a he same me, and conduc regresson analyss respecvely usng he daa colleced, akng regresson resuls of CAPM for reference, we can evaluae he degree of effecve of FF3 on sock yeld predcon. Regresson model of radonal capal asse prcng: R R = α + β ( R R ) + ε (2) In he formula, perod ; Rm m R s he yeld of porfolo durng perod ; R s he rsk-free rae durng s marke rae of reurn durng perod ; β s he regresson coeffcen of marke rsk facor of porfolo ; α s he nercep and Regresson model of FF3: ε s resdual error.

3 R R = α + β ( R R ) + s SMB + h HML + ε (3) m In he formula, he meanngs of R, R, R m, α, β, ε are he same wh he above; s s relave o he regresson coeffcen of SMB of porfolo, whch s he sensvy of porfolo sze premum of all; h s he regresson coeffcen of HML of porfolo, whch s he sensvy of porfolo relave o value premum of all. (2) Block desgn and varable values As he value of each varable n he model, he calculaon mehod s as follows: R s he weghed average yeld of he porfolo n each perod, s n monh, he wegh o calculae he weghed average yeld s he proporon of curren marke of each sock accouned for he oal curren marke n he porfolo; R s monhly average marke reurns consderng cash dvdend renvesmen. R m s monhly rsk-free rae. To explore f here exs scale effec and value effec n he research objec, hs paper grouped he seleced shares of he elecrc power ndusry based on he sze and book-o-marke rao, he groupng mehod and calculang mehod of SMB and HML s he same wh he pracce of Fama and French, so as o provde more evdence for he evaluaon of model applcably. 4. Posve Analyss on he Applcably of Rae of Reurn Predcon Models of Elecrcy Indusry n Chnese Sock Marke 4.1 Applcably analyss on CAPM The regresson analyss resuls of CAPM are as follows (Table 1): Skcd SH SM SL BH BM BL Model Noe: 95% confdence nerval Table CAPM regresson resuls. Sandardzed coeffcen B Sandard error Sg. a RM-RF a RM-RF a RM-RF a RM-RF a RM-RF a RM-RF As we can see from Table 1, among he 6 porfolos, he regresson coeffcen of consan a n porfolo BH, BM and BL are sgnfcanly as much as 0, ndcang ha he ness of CAPM o he small sze companes s no well, CAPM can no explan he small frm effec, here sll exac oher rsk facors whch affec he cross-secon reurn of socks expec for, whch furher confrms he exsence of sze effec n Chnese elecrcy ndusry sock marke. 4.2 Applcably analyss on FF3 Below are he regresson analyss resuls of FF3 (Table 2, 3):

4 Table 2. Average rae of reurn comparson on porfolos. Hgh Mddle Low Small Bg To analyze daa n he Table above, longudnally, he rae of reurn (RI) of small companes s sgnfcanly hgher han ha of large companes, whch reflecs sgnfcan sze effec. However, compared wh he yeld dae durng , he value effec s no obvous n hs sage. Skcd SH SM SL BH BM BL Model Noe: 95% confdence nerval Table FF3 regresson resuls. Sandardzed coeffcen B Sandard error Sg. a RMRF SMB HML a RMRF SMB HML a RMRF SMB HML a RMRF SMB HML a RMRF SMB HML a RMRF SMB HML To analyze daa n he Table above, under he confdence level of 95%, he values of consan a n 6 porfolos are all as much as 0.As o he sgnfcance level, he regresson parameers of SMB n he 6 porfolos show good sgnfcance; however, here are wo porfolos (SM, BM) among he 6 porfolos whose regresson parameers show no sgnfcan, ndcang ha he value effec n hs sage s no as obvous as ha durng On he whole, FF3 can explan he cross-secon reurn of sock well n hs sage. 4.3 Goodness of comparson beween models Ths s he dagram of he ng resuls of wo models durng :

5 Fgure 1. Fng resuls of he wo models durng Noe: Predced values s he predcon of RI; RIRF refers o (RI-RF) 5. Concluson Overall, compared wh CAPM, FF3 s more applcable o he rae of reurn predcon of elecrcy ndusry n Chnese sock marke. I s mporan o noe ha he goodness of of FF3 has also declned durng compared wh ha n , he reasons may be ha as he explanaory ably of droped, he explanaory ably of SMB and HML s lmed. A he same me, s for sure ha compared wh he radonal sngle facor mode, he mul-facor models lke FF3 can beer explan he rsk facors he socks faced by. So how o mprove he exsng facors and fnd ou rsk facors of sronger explanaon ably o mprove he goodness of of he model wll be he drecon and focus n he followng research. Reference [1] Fama, Eugene F., Sock Reurns, Expeced Reurns, and Real Acvy, Journal of Fnance, Vol. 45 Issue 4, [2] Fama, Eugene F., Fama and French ncden capal markes II, Journal of Fnance 46, , [3] Fama, Eugene F., The Cross-Secon of Expeced Sock, Journal of Fnance 47, [4] Fama, Eugene F. And Kenneh R. French, Common rsk facors n he reurns on socks and bonds, Journal of Fnance Economcs 33, 3-56, [5] Gregory Connor, Sanjay Sehgal. Tes of he Fama and French Model n Inda. Fnancal markes group, [6] Belra, Andrea, d Tra, Massmo. The Cross-secon of Rsk Prema n he Ialan Sock Marke. Economc Noes, Vol. 31, pp , [7] Deng Chang-rong, Ma Yong-ka. Research and Tes of he Three Facor-Facor Model n Chnese Sock Marke. Sysems Engneerng-Theory Mehodology Applcaon, 2005 (03).