The probability of informed trading based on VAR model

Transcription

1 Universiy of Wollongong Research Online Faculy of Commerce - Papers (Archive) Faculy of Business 29 The probabiliy of informed rading based on VAR model Min Xu Beihang Universiy, xumin_828@sina.com Shancun Liu Beihang Universiy, liushancun@buaa.edu.cn Publicaion Deails Xu, M. & Liu, S. (29). The probabiliy of informed rading based on VAR model. Inernaional Conference on Business Inelligence and Financial Engineering, 29. BIFE '9 (pp ). Piscaaway, USA: IEEE. Research Online is he open access insiuional reposiory for he Universiy of Wollongong. For furher informaion conac he UOW Library: research-pubs@uow.edu.au

2 The probabiliy of informed rading based on VAR model Absrac The paper researches he represenaive variable of he probabiliy of informed rading, selecing CCER highfrequency rading daa of Shanghai sock exchange from o , adoping VAR model. Differen from previous sudies, he paper firsly accouns for he dynamic relaionship beween rade and price. Then, he conen of informaion in rading volume, duraion and rading direcion are considered in our model. Finally, i ges he probabiliy of informed rading and analyzes his variable. The resuls show: he probabiliy of informed rading is abou.7273; he more asymmeric informaion is, he larger spread is; he probabiliy of informed rading is he well-known U-shape; i is he bigges before he announcemen. Keywords probabiliy of informed rading, informed rader, spread, inraday paern, announcemen Disciplines Business Social and Behavioral Sciences Publicaion Deails Xu, M. & Liu, S. (29). The probabiliy of informed rading based on VAR model. Inernaional Conference on Business Inelligence and Financial Engineering, 29. BIFE '9 (pp ). Piscaaway, USA: IEEE. This conference paper is available a Research Online: hp://ro.uow.edu.au/commpapers/353

3 29 Inernaional Conference on Business Inelligence and Financial Engineering The Probabiliy of Informed Trading based on VAR Model Min Xu School of Economics and Managemen Beihang Universiy, Beijing 83, China School of Accouning & Finance, Wollongong Universiy, NSW 25, Ausralia Shancun Liu School of Economics and Managemen Beihang Universiy, Beijing 83, China Absrac-The paper researches he represenaive variable of he probabiliy of informed rading, selecing CCER high-frequency rading daa of Shanghai Sock Exchange from o , adoping VAR model. Differen from previous sudies, he paper firsly accouns for he dynamic relaionship beween rade and price. Then, he conen of informaion in rading volume, duraion and rading direcion are considered in our model. Finally, i ges he probabiliy of informed rading and analyzes his variable. The resuls show: he probabiliy of informed rading is abou.7273; he more asymmeric informaion is, he larger spread is; he probabiliy of informed rading is he well-known U-shape; i is he bigges before he announcemen. Keywords-probabiliy of informed rading; informed rader; spread; inraday paern; announcemen I. INTRODUCTION Based on asymmeric informaion heory, researches yield wo imporan empirical predicions: () asymmeric informaion and bid-ask spread have posiive relaionship, and (2) asymmeric informaion has posiive impac on he price effec of rade-relaed informaion. For he researches on he firs predicaions, he popular approach is o find ou he represenaive variable for asymmeric informaion, such as spread which has limis, because of he dispersal. Oher researches on he second aspec focus on VAR model from Hasbrouck (99) and is relaed models. Hasbrouck (99) found ou he change of price depends on rading size and he direcion. Glosen and Milgrom (985), Diamond and Verrecchia (987) explored he imporance of ime. In Glosen and Milgrom (985) research, hey focused on he effec of direcion of rading on asymmeric informaion. In summary, volume, duraion and direcion all have effecs on price and conain essenial informaion abou sock value. The informaion inflow ino he marke hrough informed raders. How many poenial informed raders are in he marke and how much real informaion hey have? These quesions refer o he probabiliy of informed rading. Easley, Kiefer, O Hara and Paperman (996) (EKOP model) proposed PIN as a direc measuremen for probabiliy of informed rading, however heir researches only ook rading volume ino consideraion. We invesigae he price permanen effec on sock reurn, adoping Hasbrouck (99) VAR model, considering he lagged endogeneic variables, puing duraion and direcion in he model. Based on hese, he reurn was divided ino wo pars and derives he porion aroused by informed rading. Then, he paper acquires he represenaive variable of probabiliy of informed rading and analyzes i. Secion inroduces he model and mehodology. The selecion of daa is in Secion. The esimaion of model is in Secion /9 $ IEEE DOI.9/BIFE

4 Secion is he conclusion. II. MODEL AND METHODOLOGY A. Hasbrouck Model A ime, he rading x happens and he marke maker quoes bid and ask price q b and a b a b a q laer. r = ( q + q )/2 ( q + q )/2 indicaes he revision of quoed price. Hasbrouck assumes ha he public informaion arrives a he marke afer x and before he revision of quoed price. Hasbrouck proposed he following VAR model: ν, = i i + i i + ν, i= i= = i i + i i + ν 2, i= i= r ar bx x cr d x indicaes he public informaion and ν indicaes he unexpeced rading volume 2, which includes he rading from privae informaion and noise owing o liquidiy. x indexes rading direcion. B. Reurn-volume Model R and volume, and Z = i i + λ + ( i i= i= diln( T i) ex i iln( T i)) V i v, V = i i + λ + ( i i= i= giln( T i) kix iln( T i)) V i ν 2, R ar DV c V br D V f V index he sandard reurn and T indicaes he duraion (+ second). D is a dummy variable. C. Reurn Decomposiion The raio of 5 exp( ar ) / exp( R) i= i i measures he uninformed componen of reurn, while 5 exp( ar i i) / exp( R) = inf is he i= informed componen of reurn. D. The Probabiliy of Informed Trading and Spread The microsrucure heories figure ou spread has posiive relaion wih privae informaion. The regression is as following: s = + inf+ 2V + 3s + e 3 Chordia e al. (22) have shown ha order imbalance is also an imporan indicaor of he dealer s invenory pressure. The regressive equaion is as following: s = + inf + V + s + q + e The effecive spread is expressed by s and q indexes he order imbalance. E. The Inraday Paern of Probabiliy of Informed Trading We can explain he U paern of spread and volume using he asymmeric informaion commendably. The high spread in he opening and he closing of he marke show high quaniy of informed raders. In he heory of Dufour and Engle (2), he duraion is smaller and he rading is more acive when here are more informed raders. Meanwhile, Hasbrouck (99) found he more informed rading is, he more volumes are. No maer from volume and rading frequency, he volumes increase in he opening and closing of he rading day. We suppose he U paern of probabiliy of informed rading. F. The Announcemen Impac on he Probabiliy of Informed Trading The announcemen of company publics he informaion and has impac on asymmeric informaion. For every announcemen, we selec six days as he announcemen days and compare heir inf. The wo days before announcemen, he announcemen day and he day afer announcemen day, he second and hird days afer announcemen day are defined as before announcemen ( α ), during announcemen ( α ) b and afer announcemen ( α ) separaely. a III. DATA AND VARIABLES This sudy adops CCER high-frequency rading daabase of Beijing Sinofin Informaion Service, and he high-frequency rading daa d 75

5 beween Jul., 23 and Dec. 3, 23 of SSE 5 Index socks as sample. We selec sock 662, because of is mid-capializaion. Oher socks resuls are similar and we can provide hem if needed. Basing on Engle (2), he sandard reurn and volume are: R = l n( r / T + ) and V = ln( v / T + ). In he equaion R, b a ( q + q )/2 r =. v is he volume. b a ( q + q )/2 The rading hours of Shanghai Sock Exchange are from every Monday o Friday, 9:3am-:3am and 3:pm-5:pm. D equals when he rading happens beween 9:3 and :, oherwise, D equals. We use Lee and Ready 99 mehod o judge he direcion of rading, buying or selling. The paper runcaes he lagged iems a 5, basing on Akaike Informaion Crierion and Schwarz Crierion. In he regression of spread, s = [(price midpoin)/ midpoin]*2.the midpoin is he average of bes bid and bes ask price. q m ν i i= =, m is he numbers of rading before a he same direcion and he same price. q is he cumulaed bid-ask pressure ( q =, in he opening of he day or if here is a change of rading direcion and rading price). IV. RESULTS AND DISCUSSIONS A. The Represenaive Variable of he Probabiliy of Informed Trading We adop weighed leas squares mehod o esimae he model in (2), because he model exiss heeroskedasiciy esing by whie heeroskedasiciy es. a, a, 2 a, 3 a and 4 a are ,.225,.9345, and separaely. The probabiliy of informed rading is B. The Regressive Resuls beween Probabiliy of Informed Trading and Spread We regress he effecive spread, probabiliy of informed rading, one lagged spread and volume, and he resul is in able (Regression wih spread ). Afer ha, we include he order imbalance ino he regression and he resul is in Regression wih spread 2. TABLE.THE PROBABILITY OF INFORMED TRADING AND SPREAD Regression wih spread Regression wih spread 2 Coef. Mean Coef. Mean (.4756).322 (-2.66) (-.95).938 (.3636) (.244).463 ( ) -.36 (-2.736) (-7.872) 3.32E-6 4 ( ) The coefficien of volume is negaive and ha of informed rading is posiive, which is he same as expecaion. I shows he increase of volume reduces spread, while he increase of informaive rading enlarges spread because he ohers are afraid of rading wih informed raders. The lagged spread has negaive impac on spread and order imbalance has insignifican effec on spread. C. The Inraday Paern From figure, he probabiliy of informed rading has U paern. Figure. The inraday paern of he probabiliy of informed rading In he firs hour of opening, he exen of asymmeric informaion is high, abou 5.8%. The insiuion and informed raders rade quickly and effecively o uilize he cumulaed privae informaion. The informaion becomes public informaion as heir rading. In he middle of he rading day, he probabiliy of informed 752

6 rading goes down o %. A he las hour of rading, he probabiliy of informed rading increases again, even %. D. The Announcemen The quarerly announcemen happened on 3 h Ocober and he effecs are shown in able. The probabiliy of informed rading is highes before he announcemen. The probabiliy reduces o.966% as he privae informaion urns o public. TABLE. THE PROBABILITY OF INFORMED TRADING BEFORE AND AFTER ANNOUNCEMENT Before On Afer Ohers Mean Median Sd α =.8459, b α =.75887, d α = a From he analysis, he publicaion of announcemen has he meanings as a mehod of informaion ransparence. The publicaion of announcemen can reduce he informed raders in he marke, reduce rading coss, order he marke and make he raders rading in a fair marke. V. CONCLUSION The paper invesigaes volume impac on price, adoping CCER high-frequency rading daa from Jul., 23 o Dec.3, 23, selecing Shanghai Sock Exchange 662 as sample, using VAR model in Hasbrouck (99). Afer ha, we decompose reurn and achieve he probabiliy of informed rading and analyze he variable. The empirical resuls show he probabiliy of informed rading is.7273 in he sample. The asymmeric informaion has posiive effec on spread and volume has opposie effec. The probabiliy has U shape in he rading day. Before he announcemen, he asymmeric informaion is highes which shows he ransparence of marke afer he announcemen. Foundaion for he Auhor of Naional Excellen Docoral Disseraion of PR China (No.2466) and Naional Naural Science Foundaion of China (No.7676). REFERENCE [] Hasbrouck J., Measuring he Informaion Conen of Sock Trades, Journal of Finance, vol. 46(), pp ,99. [2] Glosen Lawrence R, Milgrom Paul R, Bid Ask, and Transacion Prices in a Specialis Marke wih Heerogeneously Informed Agens, Journal of Financial Economics, vol. 4, pp. 7-,985. [3] Diamond D.W and Verrecchia, R.E. Consrains on Shor-Selling and Asse Price Adjusmen o Privae Informaion, Journal of Financial Economics, vol. 8, pp ,987. [4] Easley, D.; Kiefer, N.M. and O Hara, M., Paperman, J.B. Liquidiy, Informaion and Infrequenly Traded Socks, Journal of Finance, vol. 5, pp , 996. [5] Chordia, T.; Roll, R.and Subrahmanyam, A. Order Imbalance, Liquidiy, and Marke Reurns, Journal of Financial Economics, vol. 65, pp. 3, 22. [6] Dufour, A. and Engle, R.F. Time and he Price Impac of a Trade, Journal of Finance, vol. 55, pp , 2. [7] Engle, R.F. The Economics of Ulra-High-Frequency Daa, Economerica, vol. 68, pp. 23, 2. [8] Lee G and Ready M J. Inferring Trading Direcion from Inraday Daa, Journal of Finance, vol. 46, pp , 99. ACKNOWLEDGMENT This research is suppored by A 753