FOUR RISK FACTORS RELATED TO THE COST OF EQUITY UNDER ASYMMETRIC INFORMATION: EVIDENCE FROM CHINESE CAPITAL MARKET

Transcription

1 Ming-Feng Hsu, In.J.Eco. Res., 215, FOUR RISK FACTORS RELATED TO THE COST OF EQUITY UNDER ASYMMETRIC INFORMATION: EVIDENCE FROM CHINESE CAPITAL MARKET Ming-Feng Hsu Assisan professor in he Deparmen of Banking and Finance, Naional Chiayi Universiy, No. 58, Sinmin Rd., Chiayi Ciy, Taiwan. Tel.: , FAX: , address: mfhsu@mail.ncyu.edu.w. Absrac We invesigae four risk facors on cos of equiy using companies lised in Shanghai and Shenzhen Sock Exchanges. Discreionary accruals and real earnings managemen are considered he fourh facor besides Fama- French s hree facors. The empirical findings show ha Fama-French s risk facors have posiive impacs ye he earnings risk facor has a negaive impac on he cos of equiy when he earnings qualiy is high, no maer how he level of informaion asymmery is. These findings indicae ha he higher he marke risk facor, he size risk facor and he book-o-marke risk facor, he higher he cos of equiy, while he higher he earnings qualiy, he lower he cos of equiy. Keywords: earnings qualiy; cos of equiy; informaion asymmery; Fama-French hree facors JEL classificaion: G32; G34; M; M41 1. Inroducion The facors affecing sock reurns include size, leverage, E/P raio, and book-o-marke raio. I is suggesed by Fama and French (1992) ha he size and book-o-marke raio have sronger impacs on sock reurns. Banz (1981) and Chan e al. (1991) find ha he size effec, marke equiy (ME), can explain he cross-secion porfolio reurns, and denoes ha he reurns from smaller firms are higher and he ones from larger firms are lower. Chan e al. (1991) furher sugges ha book-o-marke equiy raio has a srong role in explaining reurns on Japanese socks. Bae and Kim (1998) also sugges he book-o-marke equiy and firm size explain mos cross-secional sock reurns in Japan. Fama and French (1993) consruc porfolios wih differen size and book-o-marke risk facor o confirm heir explanaion for sock reurns. The hreefacor model wih marke risk, size risk and book-o-marke risk seems o deermine he crosssecion reurns on U.S. socks. Lieraure in he pas has long shaped he hinking on he relaionship beween securiy reurns and risk. The efficiency porfolios imply ha he cross-secional expeced reurns on securiies are a posiive linear funcion of he marke risk. Numerous sudies indicae ha firms wih high book-o-marke risk have high sock reurns (Chan e al., 1991; Fama and French, 1992; Capaul e al., 1993; Lakonishok e al., 1994). Haugen and Baker (1996) and Cohen e al. (22) sugges ha sock reurns have a posiive associaion wih size risk facor. Exensive sudies have shown he negaive relaionship beween accrual earnings and sock reurns (cos of equiy) (Sloan, 1996; Fairfield e al., 23; Richardson e al., 25; Chan e al., 26). Two reasons are suggesed by Fama and French (26). Firsly, if wo firms have he same sock prices and expeced growh, bu one has higher expeced earnings, i mus have higher cos of equiy. Secondly, if wo firms have he same sock prices and expeced earnings bu one needs more equiy invesmen o produce he earnings, i mus have lower cos of equiy. Therefore, earnings qualiy can influence he cos of equiy. Francis e al. (24) sugges ha he accruals qualiy is he mos imporan facor affecing he cos of equiy. Gray e al. (29) and Bhaacharya e al. (213) also find ha poor earnings qualiy will increase he cos of equiy. The previous sudies mainly focus on he influence of accruals qualiy on he cos of equiy. Recenly, here are sudies using real earnings measures o explore he relaed issues (Roychowdhury, 26; Cohen and Zarowin, 21; Chi e al., 21; Eldenburg e al., 211; Zang, 49

2 Ming-Feng Hsu, In.J.Eco. Res., 215, 212; Alhadab e al., 213; Hsu and Wang, 214; Hsu and Wen, 215). In paricular, Deng and Ong (214) sugges ha managers use real earnings managemen o arac uninformed rading o increase he liquidiy a lower cos of equiy during seasoned equiy offerings of real esae invesmen russ. In his paper, we apply A-shares in Shanghai and Shenzhen Sock Exchanges o examining he impacs of four facors on he cos of equiy. These facors are Fama-French s hree facors and earnings qualiy measured by accruals and real earnings managemen, wih excess porfolio reurns defined as he cos of equiy. Earnings managemen is considered a risk for asymmeric informaion in lieraure, so we sor sample firms by he absolue measure of earnings managemen ino wo groups of high and low quiniles o invesigae wheher hese facors produce differen impacs on he cos. The findings are as follows: (1) Three risk facors have significan and posiive impacs on he cos of equiy, whereas he earnings facor is significanly negaive for high earnings qualiy group. (2) No maer he level of informaion asymmery, hree facors sill have posiive associaion wih he cos of equiy excep he earnings facor for high earnings qualiy group. For he robusness es, we use monhly daa during he sample period o esimae he loading coefficiens of four facors hrough ime-series regressions. Based on earnings-facor loadings, observaions are sored ino he op and he las quiniles under he high-low informaion asymmery defined by bid-ask spread. The final resuls find hree facors sill posiively impac on he cos, bu he earnings risk facor is negaive under high earnings qualiy, no maer how asymmeric informaion is. The remainder of he paper is organized as follows. Secion 2 presens he relaed lieraure review. Secion 3 describes daa and research mehods. Our empirical resuls and discussions are shown in Secion 4, and final secion concludes. 2. Lieraure review Sock reurns have been idenified wih many paerns. Some sudies show ha sock reurns are relaed o size (marke equiy), book-o-marke equiy, earnings/price, sales growh and so on, which are no explained by he capial asse pricing model (CAPM). They can be capured by Fama-French s hree facors, which are he excess reurn of a marke porfolio, size facor defined as he difference beween small sock reurns and large sock reurns, book-o-marke equiy facor defined as he difference beween high book-o-marke and low book-o-marke sock reurns. Academics and praciioners srongly influenced by CAPM have long hough ha he expeced reurn of a sock is a posiive linear funcion of he expeced reurn on marke. Hence, excess reurn on marke will have a posiive associaion wih sock reurn in hree-facor model. Fama and French (1992) sugges ha if asses can be priced raionally, sock risks are mulidimensional. In paricular, boh size (marke equiy) and book-o-marke-equiy risk facors srongly impac on sock reurns. Generally speaking, size facor is posiively relaed o profiabiliy. Haugen and Baker (1996) and Cohen e al. (22) find ha sock reurns are posiively associaed wih profiabiliy afer conrolling book-o-marke raio. According o hese empirical findings, we can exrapolae ha he relaionship beween size facor and sock reurn is also posiive. Similarly, here are a lo of empirical sudies finding ha firms wih higher booko-marke equiy have higher sock reurns (Chan e al., 1991; Capaul e al., 1993; Lakonishok e al., 1994; Fama and French, 1992, 26). Sloan (1996) indicaes ha he persisence of earnings mainly depends on he relaive composiion of he cash flow and accruals, and shows ha accruals relaed o profiabiliy are 5

3 Ming-Feng Hsu, In.J.Eco. Res., 215, negaive, i.e. he higher accruals, he lower sock reurns. Xie (21) and Richardson e al. (25) exend he sudy of Sloan (1996) by connecing accruals o earnings persisence, and show ha discreionary accruals can predic sock reurns. Chan e al. (26) also find ha accruals may predic sock reurns if he marke considers discreionary accruals he reacion o he pas growh. Aboody e al. (25) also exend Fama-French s hree-facor model o examine he relaionship beween he fourh facor of earnings qualiy and he cos of equiy (excess sock reurns). The earnings qualiy has been considered o be informaion asymmery ha affecs he cos of equiy. Chan e al. (26) sugges ha low-qualiy earnings accompanied by high accruals are associaed wih low fuure reurns. Bhaacharya e al. (213) use samples of NYSE and NASDAQ firms o invesigae he associaion beween earnings qualiy and informaion asymmery, and find ha poor earnings qualiy is significanly relaed o higher informaion asymmery. In sum, he earnings qualiy as a proxy of informaion asymmeries is very likely o influence he sock reurns. The following hypoheses are formed. Hypohesis 1: Under high earning qualiy, he marke risk facor has a posiive impac on he cos of equiy, and vice versa. Hypohesis 2: Under high earnings qualiy, he size risk facor has a posiive impac on he cos of equiy, and vice versa. Hypohesis 3: Under high earnings qualiy, he book-o-marke risk facor has a posiive impac on he cos of equiy, and vice versa. Hypohesis 4: Under high earnings qualiy, he earnings risk facor has a negaive impac on he cos of equiy, and vice versa. 3. Research Mehod Daa of companies lised in Shanghai and Shenzhen Sock Exchanges are colleced during he period from 27 o 211. Excluding firms in financial indusries and wih insufficien daa, 2,212 sample firms are used o analyze he impac of he earnings qualiy besides Fama-French s hree facors on firm s equiy cos. Mos of he companies are from manufacuring indusry. Daily sock prices and monhly financial daa of he samples are from he daabase of Taiwan Economic Journal (TEJ). There are 84,994 valid firm-monh sample daa used in his sudy. The sudy uses individual sock abnormal reurns as he cos of equiy. We specify he following regression model: ARi, = β + β1mkt + β2smb + β3hml + β4mkt + β5smb + β 6 HML i,, (1) ARi, = β + β1mkt + β2smb + β3hml + β4em + β5mkt + β 6 SMB + β7hml + β8em, (2) where AR is he abnormal reurns of firm i in period ; KMT is he marke risk facor in period ; SMB is he size risk facor in period ; HML is he book-o-marke risk facor in period ; EM is he earnings managemen measures of firm i in period ; HEM is he low earnings qualiy of firm i in period. The measuremens of he variables are described below. KMT is calculaed by he marke rae of reurn minus he risk-free rae which is he oneyear deposi ineres rae of banks in China. SMB is calculaed by he rae difference beween large and small size groups, based on firm s marke value. HML is calculaed by he rae difference beween high-raio and low-raio porfolios; he samples are divided ino he op 3% 51

4 Ming-Feng Hsu, In.J.Eco. Res., 215, low book-marke raio, he medium 4% book-marke raio and he las 3% book-marke raio, based on firm s book-o-marke raio. EM indicaes ha absolue values of discreionary accruals and comprehensive real earnings indicaors are used as he earnings qualiy measures which are presened in Appendix A. HEM indicaes ha higher absolue value on earnings managemen represens poor earnings qualiy; he dummy HEM is one if i belongs o he group of op quinile and zero oherwise. 4. Empirical Resuls 4.1 Descripive saisics of empirical daa As shown in Table 1, he mean of abnormal reurns compued from hree-facor model (-.8) is larger han is median (-1.6), denoing significanly skewed o he righ. Mos of he companies have negaive abnormal reurns during he sample period. Inasmuch as he indices in 27 were he highes levels bu began o fall afer 27 on Shanghai and Shenzhen Exchange Socks, he esimaes from he model are consisen wih he volailiy in sock indices. I denoes here is informaion risk o uninformed invesors. Wih respec o hree-facor variables, he mean and median of he wo facors, besides he book-o-marke facor, are posiive and here is no significan skewness. There are risk premiums on boh markes, and reurns on small size firms are higher han hose on large size firms. The resuls are consisen wih he prior sudies. For earnings managemen, he means are slighly larger han he medians, so is disribuion is insignificanly skewed o he lef. Tab le 1 descripive saisics for he measures in he firs regression equaion Mean Median Maximum Minimum Sd. Dev. AR MKT SMB HML ATDA CREM Noe: AR= abnormal reurn; MKT= marke premium facor; SMB=size facor; HML=book-omarke facor; ATDA = abnormal oal discreionary accruals; CREM= comprehensive measures of real earnings managemen. 4.2 Impacs of four risk facors on cos of equiy under differen levels of earnings qualiy In his secion, we respecively use he op and he las quiniles based on earnings managemen as earnings qualiy measures o verify he relaionship beween hree- or four-facor variables and he cos of equiy. The firs wo columns in Table 2 examine he impacs of hree facors on excess porfolio reurns under differen earnings qualiy, alhough oher columns es he relaion beween earnings managemen, excep hree-risk facors, and cos of equiy. As he firs column for discreionary accruals shows, hree facors have posiive and significan associaion wih abnormal reurns, consisen wih he expecaion. The higher he marke risk facor, he size risk facor and he book-o-marke risk facor, he higher he abnormal reurns under high earnings qualiy. In spie of he higher marke and size risk facors will induce higher abnormal reurns when earnings qualiy is poor. For real earnings managemen shown in he second column, here is a similar resul. Earnings managemen measure is considered he fourh risk facor o invesigae is relaion 52

5 Ming-Feng Hsu, In.J.Eco. Res., 215, wih cos of equiy, as shown in he las wo columns. The hree-facor variables associaed wih abnormal reurns are sill posiive under high earnings qualiy, i.e. he higher he risk facors are, he higher he cos of equiy is when earnings qualiy is high. Furher, he size and ne-marke price risk facors do no give rise o higher cos of equiy under poor earnings qualiy. Under higher earnings qualiy, here is a significan and negaive relaionship beween he fourh facor of earnings managemen and abnormal reurns a he.1 significan level, regardless of accruals or real earnings managemen. Then, accrual earnings managemen and he cos of equiy are significanly and negaively relaed a 1% level, if he ineracion erm of earnings is considered. This resul is no expeced. The coefficien on poor qualiy agains high qualiy is posiive ( =.322) from he regression of real earnings managemen. Table 2 he regression resuls of he influence of four risk facors on he cos of equiy AR ATDA(1) CREM(2) ATDA (3) CREM(4) α *** *** *** *** MKT.928 ***.892 ***.933 ***.94 *** SMB.971 ***.755 ***.897 *** 1.7 *** HML.12 *** ***.63 *** EM *** *** MKT HEM.38 **.85 ***.31 ***.38 *** SMB HEM.291 ***.67 *.377 *** *** HML HEM *** -.91 ** EM HEM ***.613 *** AdjR Noe: AR= abnormal reurn; MKT= marke premium facor; SMB=size facor; HML=book-omarke facor; ATDA = abnormal oal discreionary accruals; CREM= comprehensive measures of real earnings managemen; HEM= he dummy variable of high earnings managemen. The - saisic values under he coefficiens are indicaed in ialics. *, **, *** denoe saisical significance a he 1 percen, 5 percen and 1 percen levels (wo-ailed), respecively. 53

6 Ming-Feng Hsu, In.J.Eco. Res., 215, 4.3 Impacs of four risk facors on equiy cos under differen levels of informaion asymmery The above findings are no compleely consisen wih our expecaion. In addiion o he direc impacs on cos of equiy as saed by Bhaacharya e al (212), earnings qualiy also has indirec impacs due o informaion asymmery. The monhly average sock bid-ask spread (BAS) is applied as a measure of informaion asymmery: askprice bidprice BAS = 1%, ( askprice + bidprice) / 2 i, where askprice is he daily selling price of he sock; bidprice is he daily bidding price of he sock. Higher absolue value of BAS represens higher level of informaion asymmery. The dummy variable HBAS is 1 if he group formed in he op quinile has high informaion asymmery and oherwise zero. Similarly for dummy LBAS in he boom quinile represens he low informaion asymmery group. We use wo groups of observaions o es he regression equaion (1) o (4) once again. Table 3 he regression resuls from he cos of equiy influenced by four risk facors under differen informaion asymmery AR α MKT SMB HML EM MKT HEM SMB HEM HML HEM EM HEM AdjR 2 HBAS ATDA (1) *** 1.8 *** ***.129 * *** CREM(2) ***.861 *** 1.2 ***.275 ***.218 *** ATDA (3) -5.4 ***.979 *** *** *** *** CREM(4) ***.862 *** ***.273 ** *** LBAS ATDA (1) *** *** 1.52 ***.518 ***.126 ***.397 *** CREM(2) *** ***.891 *** *** ** ATDA (3) *** *** 1.44 ***.559 *** ***.463 *** ** CREM(4) *** ***.754 *** ***.17 *** *** Noe: AR= abnormal reurn; MKT= marke premium facor; SMB=size facor; HML=book-omarke facor; ATDA = abnormal oal discreionary accruals; CREM= comprehensive measures of real earnings managemen; HEM= he dummy variable of high earnings managemen. The - saisic values under he coefficiens are indicaed in ialics. *, **, *** denoe saisical significance a he 1 percen, 5 percen and 1 percen levels (wo-ailed), respecively. From he resuls of high informaion asymmery group (HBAS) in Table 3, he hree-facor variables, which are marke risk, size risk and book-o-marke risk, are sill posiively relaed o abnormal reurns a 1% significan levels under high earnings qualiy, no maer which resuls of discreionary accruals or real earnings managemen are. Neverheless, under poor earnings qualiy, he negaive impacs of hree facors on he cos of equiy are no significan excep he 54

7 Ming-Feng Hsu, In.J.Eco. Res., 215, book-o-marke risk facor a he 1 % level. Meanwhile, he coefficien on he marke risk facor is significanly posiive. Considering he impac of four risk facors on he cos of equiy, he coefficiens on oher risk facors excep he earnings risk facor are sill posiive. In erms of he fourh facor of earnings managemen, he coefficiens on wo regressions are insignificanly posiive under high earnings qualiy, bu he coefficiens on poor earnings qualiy are negaive a he 1% significan level and heir values are ( ) and -.82 ( ), respecively. For he resuls of low informaion asymmery (LBAS), he coefficiens on he impac of hree facors on he cos of equiy are similar o hose in he HBAS group when earnings qualiy is higher. The impacs of four-facor variables on he cos of equiy are posiive and significan a 1% levels. The coefficiens on discreionary accruals and on book-o-marke risk facor in real earnings regression are negaive bu insignifican under low earnings managemen. Neverheless, under poor accruals qualiy, here are posiive impacs of marke risk and size risk on abnormal reurns, and he coefficien on earnings managemen risk is significan and posiive.151 ( ) a he 5% level. In he regression analysis of real earnings managemen, regardless of high or poor earnings qualiy, boh coefficiens on earnings managemen are 2.42 and.346 ( ), respecively. The coefficien on high earnings qualiy is larger han ha on poor earnings qualiy. According o he empirical resuls above, he impacs of hree facors on he cos of equiy are consisen wih he pas sudies under high earnings qualiy, no maer wheher here is influence of informaion asymmery. Three risk facors will generae higher abnormal reurns. Noneheless, he relaionship beween hree facors and he cos of equiy is uncerain when affeced by poor earnings qualiy, excep ha he marke risk facor has a cerain and posiive impac when he level of informaion asymmery is low. Generally, he cos of equiy is low if firms possess high earnings equiy, ye firms wih high real earnings qualiy canno ge he low cos of equiy under low informaion asymmery. Therefore, hese findings under high earnings qualiy lend suppor o hypohesis 1 o hypohesis 3, bu no o hypohesis 4. In order o confirm he relaionship beween earnings managemen and cos of equiy, we apply he mehod suggesed by Aboody e al. (25) o examining he impac of earnings qualiy on he cos of equiy hrough he level of informaion asymmery. To obain esimaed facor loadings for each firm-monh, he 36 monhly daa during sample periods are used o esimae four-facor coefficiens hrough ime-series regressions. According o he esimaed facor loadings of earnings qualiy in each monh, he highes quinile and he lowes quinile are defined as he poor qualiy and he high qualiy, respecively. The same regressions are used o examine he relaionship beween four facors and abnormal reurns for wo groups of informaion asymmery. The empirical resuls are shown in Table 4. Three facors are sill posiively bu he earnings risk facor is negaively associaed wih he cos of equiy under high earnings qualiy, no maer how he level of informaion asymmery is. Alhough he coefficiens on poor accruals qualiy are also negaive, only he coefficien on he low informaion asymmery reaches 5% significan level. In erms of poor real earnings qualiy, he coefficiens of earnings risk facors are.317 ( ) and.314 ( ), respecively, regardless of he level of informaion asymmery. I denoes ha he poorer he real earnings qualiy, he higher he cos of equiy. 55

8 Ming-Feng Hsu, In.J.Eco. Res., 215, Table 4 he regression resuls from he cos of equiy influenced by four risk facors hrough esimaed risk facors under differen informaion asymmery AR α MKT SMB HML EM MKT HEMΦ SMB HEMΦ HML HEMΦ EM HEMΦ AdjR 2 HBAS ATDA (1) ***.87 ***.948 ***.263 ***.76 *** *** CREM(2) ***.867 ***.961 ***.258 ***.78 *** ** *** ATDA (3) ***.88 ***.957 ***.288 *** ***.111 *** CREM(4) ***.882 ***.937 ***.228 *** ***.82 *** *** -.19 ***.516 *** LBAS ATDA (1) ***.93 ***.953 ***.226 ***.59 ***.183 *** CREM(2) ***.898 ***.977 ***.254 ***.63 *** ATDA (3) ***.97 ***.942 ***.246 *** **.83 ***.219 *** ** CREM(4) ***.922 ***.958 ***.217 *** ***.79 *** **.672 *** Noe: AR= abnormal reurn; MKT= marke premium facor; SMB=size facor; HML=book-omarke facor; ATDA = abnormal oal discreionary accruals; CREM= comprehensive measures of real earnings managemen; HEMΦ= he dummy variable of high earnings managemen based on esimaed earnings facor loadings. The -saisic values under he coefficiens are indicaed in ialics. *, **, *** denoe saisical significance a he 1 percen, 5 percen and 1 percen levels (wo-ailed), respecively. In sum, he impacs of four risk facors on he cos of equiy have he mos consisen conclusion under high earnings qualiy, ye he influence of earnings risk facors is no surdy and he impacs of hree risk facors are barely disinguishable from he level of informaion asymmery based on sock bid-ask spread when earnings qualiy is poorer. 5. Conclusions In he sudy, we use companies lised in Shanghai and Shenzhen Sock Exchanges from 27 o 211 o examine he relaionship beween four risk facors and he cos of equiy. These risk facors include earnings using abnormal oal discreionary accruals and comprehensive real earnings, and Fama-French s hree facors. Excess porfolio reurns from hree-facor model are used as he cos of equiy. Observaions during he sample period are spli ino wo groups of he op and he las quiniles based on he absolue value of earnings managemen measure. Under he high earnings qualiy, hree risk facors have a significan and posiive relaionship wih he cos of equiy, bu earnings risk facor is significanly negaive. Neverheless, under he poor earnings qualiy group he relaionship beween four facors and he cos is unclear. Bhaacharya e al (212) saed ha earnings qualiy has indirec impacs on he cos of equiy because of informaion asymmery. Using bid-ask spread as measure of informaion asymmery, samples are furher divided ino op and boom quiniles for high and low informaion asymmery. The same regressions are used o verify he relaionship again. The resuls find ha Fama-French s hree facors and he real earnings are posiively relaed o he 56

9 Ming-Feng Hsu, In.J.Eco. Res., 215, cos of equiy under he high earnings qualiy group, regardless of he levels of informaion asymmery. The findings for he fourh facor are no consisen wih our expecaion. Subsequenly, we use he monhly daa during sample periods o esimae four-facor coefficiens hrough ime-series regressions. Based on he loadings for earnings risk facor, he wo groups of he op and las quiniles of he loadings are used o es he relaionship beween four risk facors and he cos of equiy repeaedly. The empirical resuls find ha hree risk facors sill have a posiive impac on he cos of equiy bu earnings risk facors have negaive impacs under high earnings qualiy group, indicaing ha he higher he marke risk facor, he size risk facor and he ne-marke price risk facor, he higher he cos of equiy, ye he higher he earnings qualiy, he lower he cos of equiy. 6. Appendix A Measures of he earnings qualiy are specified as follows. Discreionary accruals are esimaed by modified Jones model as in Dechow e al. (1995) o ge abnormal oal discreionary accruals (ATDA ). TA 1 SALES ARi, PPE ATDA = ˆ + ˆ + ˆ + ˆ α α1 α2 α3, (a1) Ai, 1 ASSETS 1 ASSETS 1 ASSETS 1 where TA is firm i s oal accruals equal o ne income minus cash flow from operaion in period ; ASSETS -1 is firm i s oal asses in period -1; ΔSALES is firm i s change in sales beween period -1 and period ; ΔAR is firm i s change in ne accouns receivable beween period -1 and period ; PPE is firm i s oal fixed asses in period. To measure real earnings managemen, cash flows from operaion, producion coss and discreionary expenses are aken as measures of earnings manipulaion. The manipulaive mehod presened below could affec he level of hree measures. Firs, immoderae price discoun and overproducion will cause unusually high producion coss; herefore, lower cash flow from operaion. Second, reducing discreionary expendiure will lead o unusually low discreionary expenses, bu high cash flow from operaion. Consequenly, given fixed sales, manipulaive real earnings could cause unusually low cash flow from operaion and discreionary expenses while increasing abnormal producion coss. As a resul, he earnings managemen model suggesed by Roychowdhury (26) is proposed as follows: CFO 1 SALES SALES = ˆ α ˆ ˆ ˆ 1 2 3, (a2) A A A A 1 PCos A SALES SALES SALES i 1 = ˆ α ˆ ˆ ˆ ˆ , (a3) A A A A 1 1 DExpense 1 SALSE 1 = ˆ α ˆ ˆ 1 2, (a4) Ai, 1 Ai, 1 Ai, 1 where CFO is firm i s cash flow from operaion for period ; PCos is firm i s producion coss for period ; DExpense is firm i s discreionary expenses for period, including hose for adverising, R&D, selling and adminisraive expenses; SALES is firm i s sales for period ; ΔSALES is firm i s change in sales beween period and period -1; SALES -1 is firm i s sales for period -1 ; ΔSALES -1 is firm i s change in sales beween period -1 and period -2. Sandard cash flow from operaion, as esimaed in equaion (a2), is deduced from acual cash flow from operaion o ge abnormal value (ACFO). Similarly, acual producion coss minus sandard producion coss from equaion (a3) can ge abnormal producion coss (APC); acual discreionary expenses minus sandard discreionary expenses from equaion (a4) can ge

10 Ming-Feng Hsu, In.J.Eco. Res., 215, abnormal discreionary expenses (ADE). According o real earnings measures above, we can acquire a real earnings comprehensive indicaor (CBPM) defined as APC-ACFO-ADE. 7. References Alhadab, M., I. Clacher, and K. Keasey Real and accrual earnings managemen around iniial public offerings under regulaory environmens. Working paper, Universiy of Leeds. Aboody, D., J. Hughes, and J. Liu. 25. Earnings qualiy, insider rading, and cos of capial. Journal of Accouning Research 43(5): DOI: /j X x Banz, R. W The relaionship beween reurn and marke value of common socks. Journal of Financial Economics 9: Bhaacharya, N., H. Desa and K. Venkaaraman Does earnings qualiy affec informaion asymmery? Evidence from rading coss. Conemporary Accouning Research 3(2): doi: /j x Bhaacharya, N., F. Ecker, P. M. Olsson, and K. Schipper Direc and mediaed associaions among earnings qualiy, informaion asymmery, and he cos of equiy. The Accouning Review 87(2): Bae, K. H., and J. Kim The usefulness of earnings versus book value for predicing sock reurns and cross corporae ownership in Japan. Japan and he World Economy 1: Chan, K., L. K. C. Chan, N. Jegadeesh, and J. Lakonishok. 26. Earnings qualiy and sock reurns. Journal of Business 79(3): Cohen, R. B., P. A. Gompers, and T. Vuoleenaho. 22. Who underreacs o cash-flow news? Evidence from rading beween individuals and insiuions. Journal of Financial Economics 66: Chan, L. K.C., Y. Hamao, and J. Lakonishok Fundamenals and sock reurns in Japan. Journal of Finance 46(5): Ch W., L. L. Lisic, and M. Pevzner. 21. Is enhanced audi qualiy associaed wih greaer real earnings managemen? Working paper, Naional Chengchi Universiy. Capaul, C., I. Rowley, and W. F. Sharpe Inernaional value and growh sock reurns. Financial Analyss Journal 49(1): Cohen, D. A., and P. Zarowin. 21. Accrual-based and real earnings managemen aciviies around seasoned equiy offerings. Journal of Accouning and Economics 5: Deng, X., and S. E. Ong Real earnings managemen, liquidiy and REITs SEO dynamics. AREUEA-ASSA Annual Meeing, Philadelphia, PA, 3-5 January. Dechow, P. M., R. G. Sloan, and A. P. Sweeney Deecing earnings managemen. The Accouning Review 7(2): Eldenburg, L. G., K. A. Gunny, K. W. Hee, and N. Sodersrom Earnings managemen using real aciviies: Evidence from nonprofi hospials. The Accouning Review 86(5): DOI: 1.238/accr-195. Fama, E. F., and K. R. French The cross-secion of expeced sock reurns. Journal of Finance 47(2): Fama, E. F., and K. R. French Common risk facors in he reurns on socks and bonds. Journal of Financial Economics 33: Fama, E. F., and K. R. French. 26. Profiabiliy, invesmen and average reurns. Journal of Financial Economics 82: doi:1.116/j.jfineco

11 Ming-Feng Hsu, In.J.Eco. Res., 215, Francis, J., R. LaFond, P. Olsson, and K. Schipper. 24. Cos of equiy and earnings aribues. The Accouning Review 79(4): Fairfield, P. M., J. S. Whisenan, and T. L. Yohn. 23. Accrued earnings and growh: Implicaions for fuure profiabiliy and marke mispricing. The Accouning Review 78(1): Gray, P., P. Koh, and Y. H. Tong. 29. Accruals qualiy, informaion risk and cos of capial: Evidence from Ausralia. Journal of Business Finance & Accouning 36(1)&(2): doi: /j x Haugen, R. A., and N. L. Baker Commonaliy in he deerminans of expeced sock reurns. Journal of Financial Economics 41: Hsu, M., and K. Wang The dual roles of insiuional and insider s ownership in accrual and real earnings managemen. 214 Inernaional Conference on Business and Informaion, Osaka, Japan. Hsu, M., and S. Wen The roles of insiuional invesors and insiders in earnings managemen around IPO firms in Taiwan. Inernaional Journal of Economics and Financial Issues: forhcoming. Lakonishok, J., A. Shleifer, and R. W. Vishny Conrarian invesmen, exrapolaion, and risk. Journal of Finance 49(5): Roychowdhury, S. 26. Earnings managemen hrough real aciviies manipulaion. Journal of Accouning and Economics 42: Richardson, S. A., R. G. Sloan, M. T. Soliman, and İ. Tuna. 25. Accrual reliabiliy, earnings persisence and sock prices. Journal of Accouning and Economics 39: Sloan, R. G Do sock prices fully reflec informaion in accruals and cash flows abou fuure earnings? The Accouning Review 71(3): Xie, H. 21. The mispricing of abnormal accruals. The Accouning Review 76(3): Zang, A. Y Evidence on he rade-off beween real aciviies manipulaion and accrualbased earnings managemen. The Accouning Review 87(2): DOI: 1.238/accr