Stock Market Behaviour Around Profit Warning Announcements

Transcription

1 Sock Marke Behaviour Around Profi Warning Announcemens Henryk Gurgul

2 Conen 1. Moivaion 2. Review of exising evidence 3. Main conjecures 4. Daa and preliminary resuls 5. GARCH relaed mehodology 6. Empirical resuls 7. Conclusions

3 Moivaion There is very lile research focused on profi warnings, probably because signal conveyed in a profi warning is uncerain invesors mus deermine wha a profi warning implies abou cash flows There is subsanial evidence from he developed marke ha he warnings provide valuable informaion which can affec firm value. I seems o be very advisable, in his conex, o make sure wheher he same is rue in he case of one of he smaller sock marke like he Polish one The Warsaw Sock Exchange (WSE) is ofen regarded by foreign invesors as a marke wih high level of ransacions based on privae informaion and rumours. This common belief should be refued by providing empirical evidence o suppor he Efficien Marke Hypohesis on he WSE

4 Review of exising evidence Earnings announcemens The effec of earnings announcemens on sock prices has been documened by Brown (1978), Was (1978), and Rendleman and Jones (1982), among ohers Gurguland Majdosz(2004) found he relaed earnings announcemen effecs using he sample which comesfrom he WSE Timing of informaion issue Gennoeand Trueman(1996) demonsraedha managers seem o prefer o make earnings announcemens separaely from oher announcemens if he earnings have more favourable implicaions for he firm s fuure profiabiliy han less informed raders may believe Teohand Hwang (1991) proposed an explanaion of some oherwise anomalous empirical evidence concerning sock price reacions o announcemens

5 Review of exising evidence Earnings forecass and profi warnings Libby and Tan (1999) found ha analyss revise fuure earnings projecions o incorporae earnings warnings. I means ha he warnings provide valuable informaion which was no already processed by analyss Jackson and Madura (2003)repored ha prices begino adjus abou five days before a profi warning, and he marke response is no complee unil abou five days afer he warning he accumulaed response over he 11-day period endingfive days afer he announcemen is 21.7% he profi warning effec over he 2-dayannouncemen period is 32 imes he valuaion effec upon subsequen release of he acual earnings here is no evidence of a reversal afer his period

6 Main conjecures Firms ha issue profi warnings experience negaive valuaion effecs a he ime of he announcemen The negaive valuaion effecs associaed wih profi warnings are more pronounced when he informaion is provided closer o he acual earnings announcemen The amoun of informaion conveyed a he ime of a profi warning will be inversely relaed o he size of he firm Firms experience negaive valuaion effecs before hey announce a profi warning The negaive valuaion effecs coninue beyond he day on which profi warnings are announced

7 Daa and preliminarily resuls Daa source: he Parkiedaabase including over 400,000 iems ha concern firms lised on he WSE (colleced from Parkie Newroom, PAP, and ohers) A several keywords was used o filer from he daabase he news of ineres Time period under consideraion was limied o he en years( ) To guaranee ha our resuls no o be affeced by behaviour of prices of illiquid firms, we focus ouraenion on hose of firms which are large enough o be lised on he primary marke of he WSE Even definiion: he day on which firm (he board) announcesha he arge wih respec o profi (ne profi) will no be reached is reaed by us as an even dae. If he announcemen was issued afer he WSE had been closed, nex rading day was regarded as an even day

8 Daa and preliminarily resuls The iniial sampleconsised of over one hundred iems. Unforunaely, some of he evens had o be removed from he sample due o he roubles wih separaing anoher even price effec (confounding evens) The final sample includes 57 evens The sock price series The ime series of closing prices of he sampled firm are derived direcly from Parkie Afer conrolling for misakes (missing daa), sock price series were ransformed ino reurns by using he formula: R P = log P 1

9 Daa and preliminary resuls The firs glance (Marke Model): Afer defining he 7-day even window and he 100-day esimaion window, he Marke Model was used o generae he expeced reurns over he even window An abnormal reurn is deermined as a difference beween he acual reurn and he expeced reurn To check wheher he average abnormal reurn on a given day wihin he even window is saisically significan he es saisic is employed sa 1 N N i 1 = = AR ( ) SD AR i, where he sandard deviaion of mean abnormal reurns (in denominaor) is defined as 1 ( ) = ( 1) ( 1) ( ) N 1 N 1 1 SD AR N AR AR = 0 i= = 0 i= 1 0 i, 1 0 i, 2 1/ 2

10 Daa and preliminarily resuls The Corrado (non-parameric) es is also used o verify he conclusions drawn from he parameric es saisic Tes resuls for mean abnormal reurns wihin he even window Day Mean AR(%) -sa Median AR (%) Corrado sa * * * means he values saisically significan a he 5% level

11 GARCH relaed mehodology On he basis of he resuls obained by means of Marke Model we can see ha no even effec is observed apar from he negaive mean abnormal reurn on he announcemen day (even day). I seems o be, herefore, reasonable o use he 3-day insead of he 7-day even window Savickas (2000) pu forward he es saisic based on he expeced reurns and he condiional volailiy generaed from he GARCH model l ( ) R = µ + φ R + λ R + ε + θ ε, ε ~ 0, h h i, i j = 1 i, j i, j i m, i, j = 1 i, j i, j i, i, q 2 p i, = ω i + α 1 i, jε j i, j + β j 1 i, j h = = i, j k To deermine he proper order l, k, q, andp, we employ he ACF and PACF funcions as well as he Akaike Informaion Crierion

12 GARCH relaed mehodology The es saisic is defined as N ( 1) 1/ ( ) 2 i= 1, l = ASR N SR ASR d i where: SR i, AR SR i, i= 1 = and ASR = hˆ N i, N i, Analogically, he es saisic for he sub-period of he even window is given by N ( 1 ) / ( ) 2 i= 1 l = ASCAR N N SCAR ASCAR CAR m, s i, m, s m, s where: SCAR s = m i, i= 1 i, m, s i, m, s = and ASCAR s m, s = = m AR hˆ i, N SCAR N

13 Empirical resuls Tes resuls for mean abnormal reurns wihin he even window Day Mean AR (%) ld-sa p-value * Tes resuls for cumulaive abnormal reurns over he hree sub-period of he even window Sub-period (m,s) CAR(m,s) (%) lcar-sa p-value (-1, 1) (-1, 0) (0, 1)

14 Cross-secional regression analysis The se of variables ha probably have an explanaory power for heabnormal reurn a he announcemen day consiss of he hree variables, namely SIZE-naural logarihm of he firm s size TIMING dummy variable assigned a value of one for firms ha issued a profi warning lesshan wo monh in advance of he earnings announcemen, and zero oherwise MAJOR dummy variable assigned a value of one for firms ha announced a profi reducion greaer han 50%, and zero oherwise AR = α + α SIZE + α TIMING + α MAJOR + ε i,0 i,0 i,1 i,2 i,3 i The OLS-mehod is used o esimae he model parameers

15 Cross-secional regression analysis The esimaion resuls Variable name Esimae(%) Sd. error of esimae SIZE TIMING MAJOR R-square Noe ha he signs of he parameers collaborae wih hose of Jackson and Madura (2003), bu none of parameers saisically differs from zero

16 Conclusions I was shown ha he profi warnings lead o negaive valuaion effec The above-menioned resul is invarian wih respec o he used mehodology. Therefore, i seems o be robus The even effec is observed only on he day of announcemen and none oher days of he even window, hus -here is no marke response neiher he period prior o he announcemen nor he period afer i -he new informaion is incorporaed ino prices very quickly, wha suppors he EMH on he WSE There is no significan dependencebeween he even effec and he ime when he announcemen is released, firm size, nor he level of profi reducion The laer resul needs o be confirmed in he fuure