The performance of imbalance-based trading strategy on tender offer announcement day

Transcription

1 The performance of mbalance-based tradng strategy on tender offer announcement day AUTHORS ARTICLE INFO JOURNAL FOUNDER Han-Chng Huang Yong-Chern Su Y-Chun Lu Han-Chng Huang, Yong-Chern Su and Y-Chun Lu (214). The performance of mbalance-based tradng strategy on tender offer announcement day. Investment Management and Fnancal Innovatons, 11(2) "Investment Management and Fnancal Innovatons" LLC Consultng Publshng Company Busness Perspectves NUMBER OF REFERENCES NUMBER OF FIGURES NUMBER OF TABLES The author(s) 218. Ths publcaton s an open access artcle. busnessperspectves.org

2 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 Han-Chng Huang (Tawan), Yong-Chern Su (Tawan), Y-Chun Lu (Tawan) The performance of mbalance-based tradng strategy on tender offer announcement day Abstract Ths study eamnes the process how the tender offer nformaton s ncorporated nto ntraday relaton between return and order mbalance. We frst eamne the relaton between lagged order mbalances and stock returns. The result shows that the mpacts of lagged one mbalance on returns are sgnfcantly negatve. It mples a lkelhood of mbalance-based strategy. We fnd that the relaton between order mbalance and volatlty s not strong enough, suggestng that market makers have power n mtgatng volatlty. We take a further step to eamne small frm effect durng prce formaton. The results show that nformaton asymmetry s severe n small frms. Based on the results, we develop an mbalance-based tradng strategy, whch yelds a statstcally sgnfcant postve return and outperforms buy and hold daly return on tender offer announcement day. A nested causalty approach, whch eamnes dynamc return-order mbalance relaton durng prce formaton process, eplans the mbalance based tradng strategy. Keywords: tender offer, order mbalance, nformaton asymmetry, volatlty. JEL Classfcaton: G14, G34. Introducton Over the past two decades, a consderable number of researches have been made on takeover. At frst, a majorty of lterature focuses on the stock abnormal return mmedately surroundng announcement dates (e.g. Agrawal et al., 1992; Kanel et al., 212). Recently, a small body of study has eplored long-run post acquston abnormal returns (e.g. Dutta and Jog, 29; Bessembnder and Zhang, 213). Nonetheless, to our knowledge, there s no study that eplores the behavor of the market mcrostructure on the announcement day. Accordng to Cao et al. (25) and Arnold et al. (26), the tradng pror to a tender offer announcement could be manly ntated by traders who hold prvate nformaton. Nevertheless, the majorty of nvestors are unnformed traders and they could only trade the stocks after hearng the news on the announcement day 1. Therefore, although the tradng pror to announcement s largely orgnated by nformed traders, the tradng on the announcement day could be manly ntated by unnformed traders. The tradng strategy we construct would be useful for unnformed ndvdual nvestors. Based on the form of offer, takeover could be dvded nto two parts: merger and tender offer. Accordng to Agrawal and Jaffe (2), mergers and tender offers should be nvestgated separately as they could have dfferent mplcatons for frm performance. Tender offers are dfferent from mergers manly n that acqurng frms of tender offers bd for target shares n the open market 2. Based on the takeover sample durng , Dong et al. (26) fnd that the percentage of tender offer (19.4%) s only one-ffth tme than that of merger (8.6%). Meanwhle, n the academc area, the studes 3 about the tender offers are less than those about mergers. Because there are nadequate researches on tender offers, n ths study, we fll the gap to eamne the convergence process as to how tender offer nformaton s ncorporated nto the bdder s stock prce on the announcement day. If tender offer nformaton cannot be ncorporated nto the prce mmedately 4, the unnformed traders are theoretcally able to develop a tradng strategy, whch yelds a postve return durng the announcement day. Motvated by Chorda and Subrahmanyam (24), we use ntraday transacton data for the tender offer on the announcement day to eamne the relatonshp between the order mbalances and ndvdual stock returns. We eamne the convergence process wth three dfferent tme ntervals (5-, 1-, 15- mn). In order to make sure that volatlty plays no role n the return-order mbalance relatonshp, we employ a tme-varyng GARCH (1, model to eamne the volatlty-order mbalance relatonshp. We epect that a large volatlty s followed by a large order mbalance. Moreover, we develop an mbalance-based tradng strategy, whch could earn a statstcally sgnfcant abnormal return. Fnally, a nested causalty between the order mbalance and return s nvestgated to eplore the ntraday dynamcs whch s essental n the convergence process. Han-Chng Huang, Yong-Chern Su, Y-Chun Lu, A large prevous lterature fnds that the average abnormal returns of takeover bdders tend to be negatve or close to zero. Therefore, ratonal unnformed nvestors should sell the bdder s stocks to make a proft. 2 Accordng to Rau and Vermaelen (1998), n the case of tender offers, bdder frms are often consdered as hostle and wth cash offer. Mergers occur through dscusson between the bddng frm and target frm, are often frendly, and are usually done through share offer (Loughran and Vjh, 1997; Martn and McConnell, See Mandelker (1974), Dodd and Ruback (1977), Bradley (198), Bradley et al. (1983), Lebler (1997), Ahn et al. (2, and Atanassov (213). 4 From the perspectve of market neffcency, Chorda et al. (25) shows that the market does not converge to effcency mmedately. Grossman (1975) and Grossman and Stgltz (198) fnd that the market prces cannot fully ncorporate all knowable nformaton. They argue that someone must be able to generate returns by eplotng the devaton of prces from fundamental values. 38

3 We have several margnal contrbutons. Frst of all, the tradng on the announcement day of the tender offer could be manly ntated by unnformed traders. If the nformaton cannot be ncorporated nto the prce mmedately, the unnformed traders could develop a tradng strategy, whch yelds a postve return. Secondly, on the announcement day of the tender offer, market maker behavor plays a very mportant role n mtgatng volatlty from dscretonary trades through nventory adjustments. Fnally, we nvestgate the nested causalty between order mbalances and returns as we eplore the ntraday dynamcs that s essental n the convergence process of the tender offer announcement. Our study s organzed as follows. Secton 1 descrbes data. Secton 2 ehbts the return-lagged order mbalances relaton. In secton 3, we dscuss the volatlty-order mbalance GARCH (1, relaton. Secton 4 presents the performance of order mbalance based tradng strategy. In secton 5, we ehbt the causalty relatonshp n eplanng return-order mbalance relaton and the fnal secton concludes. 1. Data We nclude tender offer acqurers from the Securtes Data Company (SDC) Merger and Acquston database. Our sample perod s from January 1, 2 through December 31, 27. Stocks are ncluded or ecluded n our samples accordng to the followng crtera. Frst, all stocks whose transacton data are not avalable n both SDC and TAQ are ecluded from our samples. Second, we delete assets from the followng categores: certfcates, Amercan Depostary Recepts, shares of benefcal nterest, unts, companes ncorporated outsde the U.S., Amercus Trust components, closed-end funds, preferred stocks and REITs, because of ther dfferent tradng characterstcs. Fnally, we have 15 samples. We use Lee and Ready (199 trade assgnment algorthm to derve 5-mnute, 1- mnute, and 15- Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 mnute order mbalances. Average return of our sample s -.249%, wth a medan of %. The standard devaton of return s.32, wth a mamum value s % and the mnmum s %. 2. Return-lagged order mbalances relaton We employ a mult-regresson model to eamne uncondtonal return-order mbalance relaton. 5 R t = 1 t- OI t-, ( where R t s the stock return at tme t of the sample stock. OI t are the lagged order mbalances at tme t of the sample stocks. We epect that whether lagged mbalances are postvely related to stock returns accordng to Chorda and Subrahmanyam (24). Sgnfcantly postve lagged order mbalances help us to develop an mbalance-based tradng strategy. We use another mult-regresson model to nvestgate the relaton between stock returns, contemporaneous and four lagged order mbalances. We epect a sgnfcantly postve mpact of contemporaneous mbalances on returns. Moreover, we conjecture how market makers dynamcally accommodate the mbalances pressure by eamnng whether there s a trend among three dfferent tme ntervals (5-, 1-, 15-mn). We run a multple-regresson model to eamne return-lagged order mbalances relaton. The results are presented n Table 1. At 5% sgnfcant level, we fnd that negatvely sgnfcant percentages of lagged one mbalance are 4.%, 4.7%, and 6.7% for 5-, 1- and 15-mn ntervals respectvely, whch are larger than those of postvely sgnfcant mbalances, namely 4.%, 2.7%, and.7% for 5-, 1-, and 15-mn. These results are nconsstent wth Chorda and Subrahmanyam (24). They argue that lagged order mbalances, especally the lagged one order mbalances, are sgnfcantly postve related to current stock returns due to the splt orders of lqudty traders. Table 1. Uncondtonal lagged return-order mbalance relaton Average coeffcent Percent postve 5-mn nterval Percent postve and sgnfcant Percent negatve and sgnfcant OIt-1-2.7E % 4.% 4.% OIt-2-1.9E % 2.7% 5.3% OIt-3 1.3E-8 46.% 4.% 6.% OIt-4-7.7E-8 44.% 2.7% 6.7% OIt-5-1.4E % 2.7% 5.3% 1-mn nterval OIt-1-5.4E % 2.7% 4.7% OIt-2-2.8E-7 44.% 2.% 6.% OIt E % 1.3% 4.% OIt-4-4.6E % 4.7% 2.7% OIt E-9 46.% 2.7% 2.% 39

4 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 Table 1 (cont.). Uncondtonal lagged return-order mbalance relaton Average coeffcent Percent postve 15-mn nterval Percent postve and sgnfcant Percent negatve and sgnfcant OIt-1-3.2E %.7% 6.7% OIt-2-7.6E % 1.3% 4.% OIt-3-2.5E % 1.3% 2.7% OIt-4-1.7E-8 48.% 2.7% 2.% OIt E % 4.% 2.7% Notes: Sgnfcant denotes sgnfcant at the 5% level. The possble eplanatons of our emprcal results are twofold. Frst of all, market makers have accommodated a hgh nventory level around the tender offer announcement day to mtgate mpacts from dscretonary nvestors. Another eplanaton s that, from prevous emprcal results, mpacts of the nformaton assocated wth the announcement of tender offers are not strong enough. That s why market makers do not face a great nventory pressure. We nclude contemporaneous and four lagged order mbalances n our regresson to eamne condtonal return-contemporaneous order mbalance relaton. The results are ehbted n Table 2. We fnd that the mpacts of contemporaneous order mbalances on returns are postvely sgnfcant for all tme ntervals at all sgnfcant levels. However, the mpacts of lagged one order mbalances are negatve for all tme ntervals at 5% sgnfcant levels. These results are consstent wth Chorda and Subrahmanyam (24). They use overreacton story to eplan the reason why negatve coeffcents of lagged one mbalance occur. Most of the nformaton about current stock returns s overreacted n contemporaneous order mbalance, therefore lagged one order mbalances, whch are autocorrelated wth contemporaneous mbalances, cause the current stock returns to reverse. Table 2. Condtonal contemporaneous return-order mbalance relaton Average coeffcent Percent postve 5-mn nterval Percent postve and sgnfcant Percent negatve and sgnfcant OIt E % 59.3%.7% OIt E-8 46.% 3.3% 9.3% OIt E % 4.7% 8.% OIt E % 4.7% 6.7% OIt E % 2.7% 7.3% 1-mn nterval OIt E-7 88.% 43.3%.7% OIt E % 3.3% 6.% OIt E % 2.% 5.3% OIt E-8 5.% 4.7% 4.% OIt E % 4.7% 3.3% 15-mn nterval OIt E-7 9.% 32.% 1.3% OIt E % 1.3% 6.7% OIt E % 1.3% 4.7% OIt E % 4.7% 4.% OIt E % 2.%.7% Notes: Sgnfcant denotes sgnfcance at the 5% level. There s one nterestng fndng n our emprcal results. Snce the percentage of postvely sgnfcant contemporaneous order mbalances s 59.3% and the percentage of negatvely sgnfcant coeffcents of lagged one order mbalance s only.7% n 5-mn nterval. It mples that dscretonary traders have a possblty to obtan prvate nformaton before the bdders announce to acqure ther targets through tender offer deals. If the nformaton they obtaned s true, they are gong to take a long poston, whch enhances a large postve mbalance and boost up stock prce. Market makers wth nventory and adverse selecton concerns react by rasng bd-ask quote together to accommodate large mbalances. Ths releases market makers nventory pressure. However, from our emprcal fndngs that nventory pressures caused by dscretonary traders are not as serous as they had epected. That s why they lower the quote prce to rebalance ther nventory levels, whch results n a 4

5 negatve coeffcent of lagged one order mbalance. Durng the convergence process, we observe the decreasng nfluence of contemporaneous order mbalances and the percentages of postvely sgnfcant coeffcents, whch have been decreasng from 59.3% n 5-mn to 32% n 15-mn. 3. Volatlty-order mbalance GARCH (1, relaton In order to make sure that volatlty plays no role n dynamc return-order mbalance relaton, we employ a tme varyng GARCH model to nvestgate volatlty-order mbalance relaton. R t N(, h ) t t t h A Bh C OI, 2 t t1 t1 t Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 (2) where R t s the return at tme t, and s defned as ln (P t /P t-1 ). OI t denotes the eplanatory varable of order mbalance. t s the resdual value of the stock return at tme t. h t s the condtonal varance at tme t. t-1 s the nformaton set n at tme t. s the coeffcent measurng the mpact of the order mbalance on volatlty of the return. We epected that nformaton clusters around announcement of tender offer. Informaton flows from dfferent vews of tender offer volatle stock returns. In order to eamne volatlty-order mbalance durng convergence process, we employ a tme varyng model. The results of dynamc volatlty-order mbalance relaton are ehbted n Table 3. Table 3. The dynamc volatlty-order mbalance GARCH (1, relaton Postve Percent postve and sgnfcant Percent negatve and sgnfcant 5-mn nterval 41.% 6.%.% 1-mn nterval 33.% 3.%.% 15-mn nterval 35.%.% 1.% Note: Sgnfcant denotes sgnfcance at the 5% level. We epected that there was a postve correlaton between volatlty and order mbalances, that s, a large volatlty s accompaned by a large order mbalance. Whle the results show that the relaton s not as sgnfcant as we had epected. At 1% Panel A: Returns compared wth zero H : 1. H 1 : sgnfcant level, only 8.%, 6.%, and 4.% of order mbalances have a sgnfcantly postve mpact on prce volatlty for 5-, 1-, 15-mn nterval respectvely. At 5% sgnfcant level, the sgnfcant number s even less. Moreover, there s no order mbalance has a sgnfcantly postve mpact on prce volatlty respectvely for all tme ntervals. As epected, we observe that the mpacts of order mbalances on return volatlty are weaker as the tme nterval gettng longer. We use market maker behavors to eplan the nterestng results. From our emprcal fndngs, we fnd that market makers wth an nherted oblgaton to reduce prce volatltes ndeed have abltes to mtgate large order mbalance effects from dscretonary traders on tender offer announcement date. Another possble eplanaton s that market makers have obtaned prvate nformaton before tender offer announcement. Therefore, they have enough nventores to mtgate large order effect. 4. Order mbalance based tradng strategy Table 4. Tradng proft under the bass of quote prce Accordng to our results n prevous sectons, we fnd that the contemporaneous order mbalances have sgnfcantly postve nfluence on stock returns, and the magntudes of mpacts decrease as the tme nterval ncreases. And the average daly open-to-close return of our 15 tender offer bdders on the announcement date s -.249%. In ths secton, we develop an order mbalance based tradng strategy for three dfferent tme ntervals. We trm off 9% of small order mbalances, matchng wth two defntons of prce, namely quote and tradng prces. We buy a share at ask prce when postve mbalance appears and sell t at bd prce when t turns negatve. We report the results n Panel A and the sgnfcance test n Panel B of Table 4. We generate an average return of -2.8%, -1.81%, and -2.1% wth a 5% sgnfcance for 5-, 1-, and 15- mn ntervals, respectvely. We conclude that the tradng strategy under the bass of quote prce underperforms daly return. We suspect that large bd-ask spreads play a role n the emprcal results. We then calculate on the bass of transacton prce. Sample Mean P-value 5-mn return of strategy mn return of strategy mn return of strategy

6 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 Panel B: Returns compared wth returns of buy-and-hold strategy 2. H : H 1 : Table 4 (cont.). Tradng proft under the bass of quote prce Mean Orgnal open-to-close return -.28 P-value 5-mn return of strategy mn return of strategy mn return of strategy Panel C: Dfferences n returns among the three ntervals 3. H : j H 1 : j 5-mn return P-value 5-mn return 1-mn return 1-mn return mn return We buy a share at tradng prce when a postve mbalance appears and sell t at correspondng tradng prce when t turns negatve. The results are reported n Panel A wth a sgnfcance test n Panel Panel A: Returns compared wth zero H : 1. H 1 : Table 5. Tradng proft under the bass of trade prce B of Table 5. We earn sgnfcant average postve returns of.49%,.17%, and.43% respectvely for 5-, 1-, and 15-mn ntervals. We conclude that they outperform daly returns. Sample Mean P-value 5-mn return of strategy mn return of strategy mn return of strategy Panel B. Returns compared wth returns of buy-and-hold strategy 2. H : H 1 : Mean Orgnal open-to-close return -.52 P-value 5-mn return of strategy mn return of strategy mn return of strategy Panel C: Dfferences n returns among the three ntervals 3. H : j H 1 : j 5-mn return P-value 5-mn return 1-mn return 1-mn return mn return In concluson, we fnd that an order mbalance base tradng strategy on tradng prce yeld statstcally sgnfcant postve returns and outperform the benchmark of daly returns. That s to say, when a company announces to acqure the other company by tender offer deal, we apply the mbalance based tradng strategy to earn abnormal returns. 5. Causalty relatonshp n eplanng returnorder mbalance relaton In order to eplan the story behnd mbalance-based tradng strategy, we employ a nested causalty to eplore the dynamc causal relaton between return and order mbalance. Accordng to Chen and Wu (1999), we defne four relatonshp between two random varables, 1 and 2, n terms of constrants on the condtonal varances of 1(T+ and 2(T+ based on varous avalable nformaton sets, where = ( 1, 2,..., T ), = 1, 2, are vectors of observatons up to tme perod T. Defnton 1: Independency, 1 2 : 1 and 2 are ndependent f 42

7 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 Var( ) Var(, ) 1( T 1 1( T 1 2 ~ ~ ~ Var(,, ) and 1( T 1 2 2( T ~ ~ ~ Var( ) Var(, ) 2( T 2 2( T 1 2 ~ ~ ~ Var(,, ) 2( T 1 2 1( T ~ ~ ~ (3) (4) Defnton 2: Contemporaneous relatonshp, 1 < > 2 : 1 and 2 are contemporaneously related f Var( ) Var( T, ) (5) 1( T 1( Var (, ) Var (,, ) (6) 1( T 1( 2( T T ~ and Var( T ) Var( T, ) (7) 2( 2( The bvarate VAR model: Var (, ) Var (,, ) (8) 2( T 2( 1( T T ~ Defnton 3: Undrectonal relatonshp, 1 = > 2 : There s a undrectonal relatonshp from 1 to 2 f Var( ) Var(, ) 1( T (9) 1( T and Var( ) (, ) 2( T Var 2( T ( Defnton 4: Feedback relatonshp, 1 < = > 2 : There s a feedback relatonshp between 1 and 2 f Var( ) Var( T, ) (1 1( T 1( and Var( T ) Var( T, ) (12) 2( 2( To eplore the dynamc relatonshp of a bvarate system, we form the fve statstcal hypotheses n the Table 6 where the necessary and suffcent condtons correspondng to each hypothess are gven n terms of constrants on the parameter values of the VAR model. Table 6. Hypotheses on the dynamc relatonshp of a bvarate system 11 ( L) 12 ( L) 1t 1t ( L) ( L), where 1t and 2t are mean adjusted varables. The frst and second t 2t moments of the error structure, t ( 1 t, 2t), are that E( ) t t, and E( t tk) for k and E( t tk) for k =, where Hypotheses H1: (L) = 21 (L) =, and 12 = 21 = H2: 1 < > 2 12 (L) = 21 (L) = H3 : 1 > 2 21 (L) = H3 * : 2 > 1 12 (L) = H4 : 1 < = > 2 12 (L) 21 (L) H5 : 1 >> 2 21 (L)=, and 12 = 21 = H6 : 2 >> 1 12 (L) = =, and 12 = 21 = H7 : 1 << = >> 2 12 (L) 21 (L), and 12 = 21 = The VAR test To determne a specfc causal relatonshp, we use a systematc multple hypotheses testng method. Unlke the tradtonal par-wse hypothess testng, ths testng method avods the potental bas nduced by restrctng the causal relatonshp to a sngle alternatve hypothess. To mplement ths method, we employ results of several par-wse hypothess tests. For nstance, n order to conclude that 1 => 2, we need to establsh that 1 < 2 and to reject that 1 > 2. To conclude that 1 <> 2, we need to establsh that 1 < 2 as well as 1 > 2 and also to reject 1 2. In other words, t s necessary to eamne all fve hypotheses n a systematc way before we draw a concluson of dynamc relatonshp. The followng presents an nference procedure that starts from a par of the most general alternatve hypotheses. Our nference procedure for eplorng dynamc relatonshp s based on the prncple that a hypothess should not be rejected unless there s suffcent evdence aganst t. In the causalty lterature, most tests ntend to dscrmnate between ndependency and an alternatve hypothess. The prmary purpose of the lterature cted above s to reject the ndependency hypothess. On the contrary, we ntend to dentfy the nature of the relatonshp between two fnancal seres. The procedure conssts of four testng sequences, whch mplement a total of s tests (denoted as (a) to (f)), 43

8 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 where each test eamnes a par of hypotheses. The four testng sequences and s tests are summarzed n a decson-tree flow chart n Fgure 1. To eplore dynamc return-order mbalance relaton durng prce formaton, we employ a nested causalty approach. In order to nvestgate a dynamc relatonshp between two varables, we mpose the constrants n the upper panel of Table 6 on the VAR model. In Table 7, we present the emprcal results of tests of hypotheses on the dynamc relatonshp n Fgure 1. Panel A presents results for the entre sample. In the entre sample, we show that a undrectonal relatonshp from returns to order mbalances s 9.4% of the sample frms for the entre sample, whle a undrectonal relatonshp from order mbalances to returns s 8.72%. The percentage of frms that fall nto the ndependent category s 3.2%. Moreover, 48.32% of frms ehbt a contemporaneous relatonshp between returns and order mbalances. Fnally, 3.36% of frms show a feedback relatonshp between returns and order mbalances. The percentage of frms carryng a undrectonal relatonshp from order mbalances to returns s almost the same as that from returns to order mbalances, suggestng that order mbalance s not a better ndcator for predctng future returns. It s not consstent wth many artcles, whch document that future daly returns could be predcted by daly order mbalances (Brown, Walsh and Yuen, 1997; Chorda and Subrahmanyam, 24). In addton, the percentage of frms ehbtng a contemporaneous relatonshp s about twelve tmes than that reflectng a feedback relatonshp, ndcatng that the nteracton between returns and order mbalances on the current perod s larger than that over the whole perod. Fg. 1. Test flow chart of a multple hypothess testng procedure Table 7. Dynamc nested causalty relatonshp between returns and order mbalances 1 2 1< > < = > 2 Panel A: All sze All trade sze 3.2% 48.32% 9.4% 8.72% 3.36% Panel B: Frm sze Small frm sze 3.% 5.% 12.% 6.% 2.% Medum frm sze 34.69% 42.86% 4.8% 12.24% 6.12% Large frm sze 26.% 52.% 12.% 8.% 2.% In order to provde the evdence showng the mpact on the relaton between returns and order mbalances, n Panel B, we dvde frms nto three groups accordng to the frm sze. Then we test the multple hypotheses of the relatonshp between returns and order mbalances. The results n Panel B ndcate that the undrectonal relatonshp from order mbalances to returns s 6.% n the small frm sze quartle, whle the correspondng number s 8.% n the large frm sze quartle durng the entre sample perod. The trend of sze-stratfed result s not obvous. 44

9 Concluson Snce we beleve that markets do not converge to effcency mmedately durng tender offers and nvestors are able to earn abnormal returns from eplotng devaton of prces from fundamental values. In our study, we eamne publc announcement of tender offer to eplore the ntraday relaton between tender offer return, volatlty and order mbalance. We fnd that the mpacts of lagged one mbalance on returns are negatve for three dfferent ntervals. Ths result s nconsstent wth Chorda and Subrahmanyam (24). The result can be attrbuted to market maker behavors because they have enough nventores to mtgate the effects from dscretonary nvestors n tender offers. Ths s also confrmed by a low average return from tender offers. However, we fnd a consstent result wth Chorda and Subrahmanyam (24) when we eamne condtonal contemporaneous return-order mbalance relaton. In order to make sure that volatlty plays no role n return-mbalance relaton, we employ a tme varyng GARCH (1, to eamne relaton between prce volatlty and order mbalance. We epect a postve References Investment Management and Fnancal Innovatons, Volume 11, Issue 2, 214 relaton between prce volatlty and order mbalances, but the results come out to be nsgnfcant. Moreover, we observe that the mpacts of order mbalances on return volatlty decrease wth the tme nterval. Our story s that market makers wth an nherted oblgaton to mtgate market volatlty play a good role durng tender offer market makng. Based on the emprcal results, we develop an mbalance based tradng strategy. We fnd that an mbalance based tradng strategy tradng on transacton prce yelds a statstcally sgnfcant postve return and outperform the benchmark of orgnal daly returns. We also employ a nested causalty approach to eamne dynamc return-order mbalance relaton durng prce-formaton process. Ths research could etend to other corporate announcement events such as seasoned equty offerng or spn off stocks. In addton, Barclay and Warner (1993) and Anand and Chakravarty (27) fnd that most of the cumulatve stock prce change s due to medum-sze trades. Therefore, f we focus on medum-sze trades, the performance of mbalance-based tradng strategy should be better than that on all-sze trades. 1. Agrawal, A. & Jeffrey, F.J. (2). The post-merger performance puzzle, SSRN Electronc Journal, 1, pp Agrawal, A., Jaffe, F.J. & Mandelker, G.N. (1992). The post-merger performance of acqurng frms: A reeamnaton of an anomaly, Journal of Fnance, 47, pp Ahn, H.J., Cao, C. & Choe, H. (2. Share repurchase tender offers and bd-ask spreads, Journal of Bankng and Fnance, 25, pp Anand, A. & Chakravarty, S. (27). Stealth tradng n optons markets, Journal of Fnancal and Quanttatve Analyss, 42, pp Atanassov, J. (213). Do hostle takeovers stfle nnovaton: Evdence from anttakeover legslaton and corporate patentng, nnovaton, Journal of Fnance, 68, pp Arnold, T., Erwn, G., Nal, L. & Bos, T. (2). Speculaton or nsder tradng: nformed tradng n optons markets precedng tender offers announcements, Workng Paper, Unversty of Alabama at Brmngham. 7. Barclay, M.J. & Warner, J.B. (1993). Stealth tradng and volatlty: Whch trades move prces, Journal of Fnancal Economcs, 34, pp Bessembnder, H. & Zhang, F. (213). Frm characterstcs and long-run stock returns after corporate events, Journal of Fnancal Economcs, 19, pp Bradley, M. (198). Inter-frm tender offers and the market for corporate control, Journal of Busness, 53, pp Bradley, M., Desa, A. & Han Km, E. (1983). The ratonale behnd nterfrm tender offers nformaton or synergy, Journal of Fnancal Economcs, 11, pp Brown, P., Walsh, D. & Yuen, A. (1997). The nteracton between order mbalance and stock prce, Pacfc-Basn Fnance Journal, 5, pp Cao, C., Chen, Z. & Grffn, J.M. (25). Informatonal content of opton volume pror to takeovers, Journal of Busness, 78, pp Chen, C. & Wu, C. (1999). The dynamcs of dvdends, earnngs and prces: Evdence and mplcatons for dvdend smoothng and sgnalng, Journal of Emprcal Fnance, 6, pp Chorda, T. & Subrahmanyam, A. (24). Order mbalance and ndvdual stock returns: Theory and evdence, Journal of Fnancal Economcs, 72, pp Chorda, T., Roll, R. & Subrahmanyam, A. (25). Evdence on the speed of convergence to market effcency, Journal of Fnancal Economcs, 76, pp Dodd, P. & Ruback, R. (1977). Tender offers and stockholder returns, Journal of Fnancal Economcs, 5, pp Dong, M., Hrshlefer, D., Rchardson, S. & Teoh, S.H. (26). Does nvestor msvaluaton drve the takeover market, The Journal of Fnance, 51, pp Dutta, S. & Jog, V. (29). The long-term performance of acqurng frms: A re-eamnaton of an anomaly, Journal of Bankng and Fnance, 33, pp

10 Investment Management and Fnancal Innovatons, Volume 11, Issue 2, Grossman, S.J. (1975). On the effcency of compettve stock markets where trades have dverse nformaton, Journal of Fnance, 31, pp Grossman, S.J. & Stgltz, J.E. (198). On the mpossblty of nformatonally effcent markets, Amercan Economc Revew, 7, pp Lee, C.M.C. & Ready, M.J. (199. Inferrng trade drecton from ntraday data, Journal of Fnance, 46, pp Lebler, R.J. (1997). Tender offers to nfluental shareholders, Journal of Bankng and Fnance, 21, pp Loughran, T. & Vjh, A.M. (1997). Do long-term shareholders beneft from corporate acqustons, Journal of Fnance, 52, pp Kanel, R., Lu, S., Saar, G. & Ttman, S. (212). Indvdual nvestor tradng and return patterns around earnngs announcements, Journal of Fnance, 67, pp Mandelker, G. (1974). Rsk and return: The case of mergng frms, Journal of Fnancal Economcs, 1, pp Martn, K.J. & McConnell, J.J. (199. Corporate performance, corporate takeovers, and management turnover, Journal of Fnance, 46, pp Rau, R. & Vermaelen, T. (1998). Glamour, value and the post-acquston performance of acqurng frms, Journal of Fnancal Economcs, 49, pp