European Journal of Economc and Poltcal Studes NYSE Specalsts Partcpaton n the Posted Quotes Bülent Köksal 1 Abstract: Usng 2001 NYSE system order data n the decmal prcng envronment, we analyze how the specalsts react to the changes n market varables whle makng partcpaton decsons to the posted quotes. We analyze the specalsts decson to undercut or add depth to the lmt order book. We dstngush bd and ask sde of the quotes. We fnd that the prmary factors that affect the partcpaton strategy of the specalsts to the current posted quotes are the changes n the best prces and depths on the lmt order book snce the prevous quotes. In addton, specalsts partcpate to the posted quotes more for low-volume stocks. Unlke some prevous studes, we fnd sgnfcant nventory effects provdng some evdence that the specalsts actvely manage ther nventores. Keywords: NYSE Specalsts, Dealer Tradng, Market Makers, Lmt Order Book, Posted Quotes. Introducton New York Stock Exchange (NYSE) specalsts are responsble for makng markets for the stocks assgned to them. Brefly, they should be wllng to trade when other traders are unwllng to trade, the bd-ask spread should not be too wde, they should ntervene to prevent large prce jumps, and create prce contnuty. Specalsts cannot trade for ther own accounts f there exst publc orders at the same prce or better. In addton, they should not trade wth lmt orders n order not to take the lqudty avalable to publc traders. 1 Central Bank of Turkey, bulent.koksal@tcmb.gov.tr 13
Bülent Köksal In ths paper, usng propretary 2001 NYSE system order data 2 n the decmal prcng envronment, we analyze the determnants of specalst partcpaton n the posted quotes over tme and across stocks by parttonng posted depth nto the specalst s contrbuton and the lmt order book s (LOB) contrbuton. To calculate the net specalst contrbuton, we frst estmate the LOBs at each pont n tme by usng the method descrbed n Kavajecz (1999). Then we compare the LOB to the posted quotes to determne the net specalst partcpaton n the posted quotes. The posted depth mght be comng from the LOB, specalst or both. Analyzng specalst partcpaton s very mportant because the nformaton about a partcular stock s dssemnated to the market by specalst quotes. In addton, they oversee a huge tradng actvty and there are potental conflct of nterests between the specalsts desre to make profts for themselves and ther oblgaton to be far to all publc traders. Our am s to contrbute to the debate about the role of specalsts n the tradng busness and address the queston of whether they partcpate to the posted quotes n a manner consstent wth ther affrmatve oblgatons. As dscussed above, we defne the specalst partcpaton to the posted quotes as ther contrbuton to the posted quotes n addton to the LOB. A specalst has three choces for both sdes of the posted quotes. He may not partcpate and let the posted quotes reflect the prces and depths on the LOB (0% contrbuton to the posted quotes from the specalst); he may add depth to the LOB at the best prces on the book (specalst percentage contrbuton s postve and less than 100%); and he may undercut the LOB whch mples that the posted quotes fully reflect the tradng nterest of the specalst (100% contrbuton from the specalst). As the prevous theoretcal and emprcal lterature show, the specalsts use all components of the posted prce schedule whle makng the market for ther stocks. 3 Accordngly, we examne the specalsts partcpaton strategy to the posted quotes by takng the smultanety of the bd and ask sde nto account. Ths work s related to a number of papers n the prevous lterature. Kavajecz (1999) examnes whether specalsts manage quoted depth to reduce adverse selecton rsk. Kavajecz and Odders-Whte (2001) nvestgates how specalsts update the prce schedules consstng of bd quotes, ask quotes, bd depths, and ask depths. The changes n the prce schedules posted by specalsts reflect the combned strateges of the lmt order traders, specalst, and floor brokers, but not the specalst s contrbuton alone. Therefore, changes n the prce schedules are not nformatve about the strategy of the specalst. Chung, Van Ness and Van Ness (1999) analyze the role of the lmt-order traders and specalsts by usng the TORQ dataset. They examne the effect of the quote type on spread, but they do not formally control for the exogenous varables that mght 2 The NYSE s SOD dataset gves detaled nformaton on the entry and processng of orders. Our sample has 148 securtes for the perod Aprl 2nd, 2001 June 29th, 2001. The orgnal sample was selected by systematc random samplng to ensure that all NYSE stocks are represented n the sample n terms of volume and prce. 3 See Lee, Mucklow and Ready (1993), Harrs (1994), and Kavajecz (1999). 14
European Journal of Economc and Poltcal Studes drve the specalst quotes and they do not dstngush between the bds and ask quotes. Fnally, Köksal (2010) also analyzes the net specalst partcpaton to the posted quotes by usng a multnomal logt framework. Our results provde evdence that the specalsts partcpate to the posted quotes n a manner, whch s consstent wth ther affrmatve oblgatons. Changes n the dfferences between best lmt prces and quote mdpont are statstcally and economcally sgnfcant. When the dfference between best lmt bd (ask) prce and quote mdpont ncreases, causng a decrease n lqudty from bd (ask) sde of the LOB, more than half of the specalsts step n to provde addtonal lqudty. Other prmary varables that affect the strategy of the specalsts are the changes n best LOB prces and the LOB depths at those prces. Specalsts partcpaton to the posted quotes decreases by transacton volume suggestng that specalst servces are needed more for lowvolume stocks. Fnally, we fnd sgnfcant nventory effects for some specalsts provdng evdence that the specalsts actvely manage ther nventores. The rest of the paper s organzed as follows. Secton 2 presents methodology and results for smultaneous equatons analyss. We dscuss cross sectonal analyss and ts results n Secton 3 and secton 4 concludes. Stock by Stock Analyss: Smultaneous Equatons Model Emprcal Methodology: Prevous theoretcal lterature show that the possblty of nformed tradng, specalst nventory, tme between trades, volatlty of the asset value, and LOB varables are mportant determnants of the market maker behavor. 4 In addton, there are some performance varables that the NYSE uses to evaluate specalst performance, such as average wdth of the quoted bd/ask spread and average depth of the quotes. Accordngly, we use the followng varables for the stock by stock tme seres analyss. We wll only dscuss the lqudty provder s bd sde varables, as ask sde varables are smlarly defned. LOB varables. Change n the Best Lmt Bd Prce s the current best lmt bd prce mnus prevous best lmt bd prce; Change n the Best Lmt Bd Sze s the current best lmt bd depth mnus prevous best lmt bd depth; LOB Asymmetry s the total sze of the sell lmt orders mnus total sze of the buy lmt orders n the LOB; 4 See for example, Kyle (1985), Stoll (1978), Ho and Stoll (1981, 1983), Easley and O'Hara (1992), Dupont (2000), Bondarenko and Sung (2003), Sepp (1997). 15
Bülent Köksal Change n the % Best Lmt Bd Gap s the defned as the change n the rato of the dfference between posted quote mdpont and best lmt bd prce to the posted quote mdpont snce the last quote revson (.e., Δ(Mdquote-Best Lmt Bd)/Mdquote)); (The relevant varable for sell sde s Change n the % Best Lmt Ask Gap and defned smlarly; Δ(Best Lmt Ask - Mdquote)/Mdquote)); Buy Order Placement s the sum of buy lmt orders placed snce the last quote revson; Buy Cancellaton Actvty s the sum of buy lmt orders cancelled snce the last quote revson; Other Varables. Buy volume snce the last quote revson s the total buy transacton sze snce the last quote revson; Change n the Specalst s Inventory snce the last quote revson. Ths varable s postve f the specalst has ncreased hs nventory,.e., he has bought more shares than he has sold, snce the last quote revson; Volatlty s the coeffcent of varaton of the transacton prces durng the last ten mnutes before the current quote; LOB dle tme s the tme n seconds between the last two LOB revsons; Prevous Percentage Spread s the rato of the spread to quote mdpont n the prevous quotes; Prevous Posted Bd Depth s the posted bd depth n the prevous quotes; Prevous Posted Ask Depth s the posted ask depth n the prevous quotes. As dscussed n the ntroducton, the specalst uses all varables n the posted quotes whle determnng hs strategy. In addton, prevous lterature fnds that there exst asymmetrc effects of the ndependent varables on bd and ask sde of the posted quotes. 5 Accordngly, we model the revson process of the specalst partcpaton to posted quotes as a system of two smultaneous equatons, where the dependent varables are the changes n percentage specalst partcpaton at the bd and ask snce the last quote revson. Percentage specalst partcpaton at the bd (ask) s smply the percentage of the depth that belongs to the specalst n the current total posted bd (ask) depth. These two choce varables reduce the dmensonalty of the specalst s decson problem. Table 1 dsplays the equatons and varables of the smultaneous equatons system. 5 See, for example Madhavan and Smdt (1991), and Panaydes (2004). 16
European Journal of Economc and Poltcal Studes Table 1. Equatons that are estmated by smultaneous equatons model Explanatory Varables Endogenous Varables Change n % Specalst Partcpaton at the Bd Change n % Specalst Partcpaton at the Ask Intercept X X Δ n % Spec. Partc. at the Bd Δ n % Spec. Partc. at the Ask Δ n the Best Lmt Bd Prce X X Δ n the Best Lmt Ask Prce X X Δ n the Best Lmt Bd Sze X X Δ n the Best Lmt Ask Sze X X Δ n the % Best Lmt Bd Gap X X Δ n the % Best Lmt Ask Gap X X Δ n the LOB Asymmetry X X Cumulatve Lmt Buy Order Placement X X Cumulatve Lmt Sell Order Placement X X Cumulatve Cancelled Lmt Buy Orders X X Cumulatve Cancelled Lmt Sell Orders X X Elapsed tme between last two LOB revsons X X Volatlty X X Buy volume snce the last quote revson X X Sell volume snce the last quote revson X X Δ n the Specalst's Inventory X X Prevous Percentage Spread X X Prevous Posted Bd Depth Prevous Posted Ask Depth X ndcates a rght hand sde varable ncluded n the relevant equaton. Identfcaton s an mportant problem n estmatng the smultaneous equatons models. Identfcaton s not a problem n our model, because the prevous posted bd (ask) depth varable appears only n the bd-sde (ask-sde) equaton and the model s exactly dentfed. We estmated our model for each stock n our sample by usng OLS and adjusted the standard errors usng the Newey-West autocorrelaton consstent covarance estmator. 6 X X X X 6 Other methods lke 2SLS or 3SLS produced smlar results. 17
Bülent Köksal Results The results from the smultaneous equatons analyss are presented n Table 2. We report the mean and medan of estmated coeffcents for all stocks. The last column reports the percentage of sgnfcant coeffcents at the 5% level. Table 2. Smultaneous Equatons Model Results for stock by stock estmaton For each explanatory varable n each equaton, the mean and medan of all coeffcent estmates across the stocks n the sample are provded. % column reports the percentage of sgnfcant coeffcents at the 5% level. All numbers for explanatory varables wth a * are multpled by 100000. Equatons Δ n % Spec. Partc. at the Bd Δ n % Spec. Partc. at the Ask Exogenous varables Mean Medan 5% Mean Medan 5% Intercept -0.05-0.03 56.94-0.03-0.02 59.72 Δ n % Spec. Partc. at the Bd 0.43 0.27 45.83 Δ n % Spec. Partc. at the Ask -0.48 0.09 35.42 Δ n the Best Lmt Bd Prce 3.18 0.24 39.58-11.43-4.04 59.72 Δ n the Best Lmt Ask Prce -3.27-0.05 36.81 9.97 2.37 52.78 Δ n the Best Lmt Bd Sze* 9.20 0.27 70.14-0.46-0.17 46.53 Δ n the Best Lmt Ask Sze* -12.00-0.10 31.94 1.20 0.32 72.92 Δ n the % Best Lmt Bd Gap 126.62 80.31 54.86-484.03-228.35 65.28 Δ n the % Best Lmt Ask Gap 49.11-23.86 39.58-33.24 21.72 54.17 Δ n the LOB Asymmetry* 9.70 0.57 55.56-1.00-0.67 73.61 Cumulatve Lmt Buy Order Placement* 8.80-0.07 33.33 0.63-0.13 45.83 Cumulatve Lmt Sell Order Placement* 12.00-0.14 41.67-0.54 0.02 40.28 Cumulatve Cancelled Lmt Buy Orders* -0.49 0.33 41.67-0.96 0.29 49.31 Cumulatve Cancelled Lmt Sell Orders* -17.00 0.30 36.11 2.40 0.03 40.28 Elapsed tme between last two LOB revsons* 39.70 3.40 32.64 11.30 1.50 34.03 Volatlty -0.12-0.01 22.22-0.01-0.01 22.92 Buy volume snce the last quote revson* -8.00-0.04 26.39 0.51 0.00 26.39 Sell volume snce the last quote revson* -10.00 0.00 25.69-0.85-0.06 29.86 Δ n the Specalst's Inventory* -13.00-0.08 21.53 0.96 0.05 31.94 Prevous Percentage Spread 12.44 4.48 41.67 5.37 3.24 55.56 Prevous Posted Bd Depth* 2.70 0.20 60.42 Prevous Posted Ask Depth* 0.78 0.05 62.50 18
European Journal of Economc and Poltcal Studes Endogenous varables have negatve (postve) coeffcents n the bd-sde (asksde) equaton. Postve coeffcents mply that when specalsts ncrease ther partcpaton on one sde of the market because of ther updated belefs about the stock value, they also ncrease ther partcpaton to the other sde of the market to support that relatvely weak sde of the market and to mantan prce contnuty. Negatve coeffcents ndcate that specalsts use both sdes of the market to mplement ther strateges. As an example, let s say that a specalst updates hs belefs about the value of the stock downwards or hs nventory s above hs target and he wants to decrease hs nventory. He ncreases hs partcpaton to the ask sde (to decrease hs holdngs of the stock), and he decreases hs partcpaton to the bd-sde (to avod buyng the stock). For example, when the specalst s partcpaton to the ask-sde ncrease by 10 percentage ponts, the effect of ths on the bd-sde s a decrease n hs partcpaton n bd sde by 4.8 percentage ponts. The reason why he uses both the bd and ask quotes to mplement hs strategy when he wants to decrease hs holdngs mght be that f he s caught wth a large postve nventory when the stock prces are declnng, he would suffer bg losses, so prce contnuty motve s not very strong n ths case, causng a decrease n partcpaton to the bd-sde. On the other hand, let us say that the specalst updates hs belefs about the value of the stock upwards. Then he ncreases hs partcpaton to the bd-sde (to ncrease hs holdngs of the stock and/or to mnmze hs losses to nformed traders), and asksde (to mantan prce contnuty). For example, when specalst s partcpaton to the bd-sde ncreases by 10 percentage ponts, the effect of ths ncrease on the ask-sde s an ncrease n hs partcpaton by 4.3 percentage ponts. There s a strong prce contnuty motve n ths case, because after he buys the stock, f he ends up sellng the stock (because of relatvely hgh partcpaton to the ask-sde) he only loses proft opportuntes from hgh future prces, rather than sufferng drect losses. One of the most mportant players n the posted quotes s the lmt order book (LOB). We use several LOB varables n our analyses. As the ntuton suggests, bdsde (ask-sde) LOB varables are more sgnfcant n explanng the partcpaton of the specalsts to the bd (ask) quotes, because bd-sde (ask-sde) LOB varables have a drect effect on the bd (ask) quotes. Consstent wth pror expectatons, specalsts use all avalable varables to mplement ther strateges. The specalst ncreases (decreases) hs percentage contrbuton to the bd-sde (ask-sde) when the best lmt bd prce ncreases consstent wth the proft motve. For example, a 1-cent ncrease n the best lmt bd causes the specalst to decrease hs partcpaton to the ask quotes by 11.43 percentage ponts ($0.01 x 11.43). Ths mples that he uses nformaton from the LOB,.e., he updates hs belefs about the stock value 19
Bülent Köksal upwards and decreases hs partcpaton to the ask-sde to decrease the probablty that he sells the stock. On the other hand, the specalst ncreases hs percentage contrbuton to the ask-sde when best lmt ask prce ncreases. Therefore, consstent wth hs affrmatve oblgatons, he supports the relatvely weak ask-sde n ths case. Besdes the changes n the best lmt prces, changes n the szes at those prces are sgnfcant too. When the best lmt bd (ask) sze ncreases, the specalst ncreases (decreases) hs contrbuton to the bd quotes. There mght be two explanatons. Frst, the specalst mght update hs belef about the stock value upwards (downwards) and ncrease (decrease) hs partcpaton to the posted bd quotes accordngly to ncrease (decrease) the probablty that he buys. Second, when the sze at the best prces ncreases, the specalst can hde behnd ths sze and safely ncrease hs partcpaton to the posted quotes, whch mproves hs performance by ncreasng the average depth that he quotes. The most drect way to see f the specalsts partcpate to the quotes n a manner consstent wth ther affrmatve oblgatons s by lookng at the best lmt bd and ask gap varables. A large gap between best lmt bd (ask) prce and posted quote mdpont ndcates a weak bd (ask) sde that needs support from the specalst. In addton, f the LOB provdes some nformaton about the future movements of the prce, a large gap between best lmt bd (ask) prce and posted quote mdpont mght provde bad (good) news for the stock snce the lmt order traders are less wllng to buy (sell) the stock. The medan of estmated coeffcents for the best lmt bd (ask) gap s postve for the bd (ask) sde, showng that when the lqudty from the LOB s not suffcent, specalsts step n to provde addtonal lqudty. The magntudes of the estmated coeffcents are also large ndcatng that these gap varables are two prmary varables that the specalst looks at whle determnng hs strategy. We use some actvty varables from the LOB. These varables are cumulatve buy and sell order placement and cancellaton snce the last quote revson. Lmt buy (sell) order placement varable s postve for the ask (bd) sde equaton ndcatng that, as cumulatve sze of lmt buy (sell) orders placed ncreases, the specalst ncreases hs contrbuton to the ask (bd) quotes because ask-sde(bd-sde) of the market s relatvely weak and needs support from the specalst. Smlarly, as cumulatve sze of cancelled lmt buy (sell) orders ncreases, the specalst adds more depth from hs own nventory to the posted bd (ask) as the medan of the estmated coeffcents ndcate. Harrs and Panchapagesan (2005) and Köksal (2010) fnd that an asymmetry n the LOB has sgnfcant explanatory power n predctng the future prce movements. LOB Asymmetry varable s defned as total sze of lmt sell orders mnus lmt buy orders and measures the overall asymmetry n the LOB. When LOB asymmetry ncreases 20
European Journal of Economc and Poltcal Studes because of a relatve ncrease (decrease) n lmt sell (buy) orders, the specalst decreases (ncreases) hs partcpaton to the posted ask (bd) quote, because ask-sde (bd-sde) of the market s relatvely strong (weak) now. Ths provdes evdence that specalsts do not try to undercut a heavy LOB on the sell-sde. The economc effect of the LOB asymmetry s small though, as the magntudes of the estmated coeffcents ndcate. As the volatlty of the securty prce ncreases, the specalsts decrease ther percentage contrbuton to both sdes of the quotes ndcatng nsuffcent prce stablzaton actvty. Ths provdes evdence n support of the theoretcal results of the Bondarenko and Sung (2003) who fnd that hgher volatlty ncreases the rsks assocated wth carryng nventory whch wll result n less specalst contrbuton to depth. The coeffcent of Elapsed tme between the last two LOB revsons varable s postve. Ths result provdes evdence n favor of the fndng of the Easley and O'Hara (1992) that f no actvty occurs n some tme nterval, the market maker rases hs probablty that no nformaton event has occurred. Accordngly, he ncreases hs partcpaton to the bd and ask quotes. The mean elapsed tme between the last two LOB revsons s around 10 and 140 seconds for hgh- and low-volume stocks, respectvely. A 1 mnute (60 seconds) no actvty tme causes the specalst to ncrease hs partcpaton to the bd (ask) quotes by 2.38 (0.678) percentage ponts. Mean coeffcent for the nventory varable s negatve (postve) for the bd (ask) sde. Therefore, when the nventory ncreases, specalsts decrease (ncrease) ther partcpaton to the bd (ask) sde to reduce the nventory. The mean of the estmated coeffcents for the percentage bd-ask spread n the last perod s postve for both bd- and ask-quotes. Bd-ask spread s an mportant measure of the specalst performance. Hence, specalsts decrease the current bdask spread when the prevous bd-ask spread s large. The estmated coeffcents are economcally sgnfcant too. When the percentage bd-ask spread n the prevous perod ncreases by 0.01, the specalst ncreases hs partcpaton to the bd (ask) quotes by 12.44 (5.37) percentage ponts. Therefore, when the spread s large n the prevous perod, the specalst tres to narrow the spread by ncreasng hs partcpaton n the posted quotes. As the posted bd (ask) depth n the prevous perod gets larger, the specalst adds more depth to the posted bd (ask) n the current perod. One possble explanaton s that the specalst completes hs buyng or sellng strategy n several steps. Another explanaton mght be that an ncrease n the prevous posted bd (ask) depth sgnals an ncrease (decrease) n the securty prce and the specalst ncreases hs partcpaton to the bd-sde (ask-sde) of the market. 21
Bülent Köksal Cross Sectonal Analyss Emprcal Methodology: Exstng lterature suggests that specalsts servces are more valuable for llqud stocks. 7 Based on ths and other theoretcal lterature dscussed n Secton 2.1, we use the followng varables to analyze how the specalsts partcpaton decsons to posted quotes vary across stocks. Log Mean Daly Volume s the logarthm of average daly volume over the sample perod; Log Market Captalzaton s the logarthm of market captalzaton as calculated by shares outstandng tmes stock prce. Relatve Tck s the tck sze (=$0.01) dvded by the mean prce over the sample perod; Volatlty s the average of the volatlty varable dscussed n Secton 2.1 over the sample perod; Average Percentage Spread s the average rato of the spread to quote mdpont durng the whole sample perod for each stock. To analyze how partcpaton of the specalsts to the posted quotes vares across stocks, we estmate the followng cross-sectonal regresson model: SpecPart LogMeanDalyVol LogMarketCap RelTck Volatlty 0 1 AvePercSpread 5 2 3 4 (1) where, for stock, partcpaton at the bd and ask, SpecPart s the sum of average percentage specalst LogMeanDa lyvol s the log of average daly volume, LogMarketC ap s the log of market captalzaton, Re ltck s the tck sze dvded by the average stock prce over the sample perod, the transacton prces from the tme seres analyss, percentage spread over the sample perod and s the error term. Volatlty s the average volatlty of AvePercSpr ead s the average Results Table 3 presents the coeffcent estmates of our cross sectonal regresson analyss. Coeffcent of logarthm of mean daly volume s sgnfcantly negatve ndcatng that specalst partcpaton to the posted quotes decreases as transacton 7 See for example, Grossman and Mller (1988), Glosten (1989), Huang and Lu (2004). 22
European Journal of Economc and Poltcal Studes volume ncreases. Ths mght ndcate ether that specalst servces are needed more for thnly traded, llqud stocks or partcpatng to posted quotes for low-volume stocks s more proftable. Table 3. OLS Results from Cross-sectonal Regresson of Specalst Partcpaton Ths table reports results from estmaton of equaton (1). Heteroskedastcty robust standard errors are reported n parentheses. ***, and ** denotes sgnfcance levels at the 1%, and 5% levels, respectvely. Dependent varable s the sum of average percentage specalst partcpaton at the bd and ask. Dependent Varable Sum of average percentage specalst partcpaton at the bd and ask Exogenous Varables Coeffcents Intercept 0.2305 (0.2349) Log Mean Daly Volume -0.0929 ** (0.0403) Log Market Captalzaton 0.0852 *** (0.0315) Relatve Tck -4.5804 (12.7732) Volatlty (Coeffcent of Varaton of the Transacton Prces) 0.0150 ** (0.0066) Average Percentage Spread 0.0271 *** (0.0075) Sample Sze 120 R 2 0.42 Log of market captalzaton has a sgnfcantly postve coeffcent mplyng that the specalst partcpaton s hgher for larger frms. One explanaton mght be that the possblty of nformed tradng s lower for larger frms because better publc 23
Bülent Köksal nformaton s avalable. Hence t s more proftable for the specalst to partcpate n the trades for these stocks to collect the bd ask spread. In addton, there s a postve relatonshp between the volatlty of the stock and the average percentage specalst partcpaton provdng evdence that specalsts ncrease ther partcpaton to smooth prces for volatle stocks. Fnally, we fnd a postve relatonshp between percentage spread and the specalst partcpaton. A wde spread leaves a lot of room for the specalst to undercut the LOB, therefore the specalst partcpaton ncreases as the spread becomes wder. Concluson We examne how the specalsts react to the changes n market varables whle makng partcpaton decsons to the posted quotes. Results ndcate that the prmary factors that affect the partcpaton strategy of the specalsts to the current posted quotes are the changes n the best prces and depths on the lmt order book snce the prevous quotes. In addton, specalsts partcpate to the posted quotes more for lowvolume stocks. We also fnd sgnfcant nventory effects provdng some evdence that the specalsts actvely manage ther nventores. Our results ndcate that on average specalsts partcpaton to the posted quotes s consstent wth ther affrmatve oblgatons. There exst some heterogenety across stocks, however, as reflected by dfferences n sgns and magntudes of the estmated coeffcents, ndcatng that some specalsts quotng strateges are costly for nvestors. Acknowledgement: Ths paper s based on my dssertaton at Indana Unversty. I gratefully acknowledge the helpful comments from Crag Holden, Robert Jennngs and Konstantn Tyurn. I thank Robert Jennngs for provdng the SOD data. Ths work was supported n part by Shared Unversty Research grants from IBM, Inc. to Indana Unversty. Any remanng errors are my own. References Bondarenko, O.; Jaeyoung, S. 2003."Partcpaton and Lmt Orders", Journal of Fnancal Markets 6, pp. 539-71. Chung, Kee H., Bonne F. Van Ness.; Robert A. Van Ness. 1999. Lmt Orders and the bd-ask Spread", Journal of Fnancal Economcs 53, pp. 255-87. Dupont, D..2000. "Market makng, prces, and quantty lmts", Revew of Fnancal Studes 13, pp. 1129-51. 24
European Journal of Economc and Poltcal Studes Easley, D.; Maureen O'Hara. 1992. "Tme and the process of securty prce adjustment", Journal of Fnance 47, pp. 576-605. Glosten, Lawrence R. 1989. "Insder tradng, lqudty, and the role of the monopolst specalst", Journal of Busness 62, pp. 211-35. Grossman, Sanford J., and Merton H. Mller. 1988. "Lqudty and market structure", Journal of Fnance 43, pp. 617-37. Harrs, Lawrence E. 1994. "Mnmum prce varatons, dscrete bd-ask spreads, and quotaton szes", Revew of Fnancal Studes 7, pp. 149-78. Harrs, Lawrence E., and Venkatesh Panchapagesan. 2005. "The nformaton content of the lmt order book: Evdence from nyse specalst tradng decsons", Journal of Fnancal Markets 8, pp. 25-67. Huang, Roger D., and Jerry W. Lu. 2004. "Do ndvdual nyse specalsts subsdze llqud stocks?", Unversty of Notre Dame, Workng Paper. Kavajecz, Kenneth A. 1999. "A specalst's quoted depth and the lmt order book", Journal of Fnance 54, pp. 747-71. Kavajecz, Kenneth A., and Elzabeth R. Odders-Whte. 2001. "An examnaton of changes n specalsts' posted prce schedules", Revew of Fnancal Studes 14, pp. 681-704. Köksal, B. 2010. "Partcpaton strategy of the nyse specalsts to the posted quotes", The North Amercan Journal of Economcs and Fnance 21, pp. 314-331. Kyle, Albert S.. 1985. "Contnuous auctons and nsder tradng", Econometrca 53, pp. 1315-35. Lee, Charles M. C.; Belnda M.; M. J. Ready. 1993. "Spreads, depths, and the mpact of earnngs nformaton: An ntraday analyss", Revew of Fnancal Studes 6, pp. 345-74. Madhavan, A.; Seymour S. 1991. "A Bayesan model of ntraday specalst prcng", Journal of Fnancal Economcs 30, pp. 99-134. Panaydes, M. 2004. "The specalst s partcpaton n quoted prces and the nyse s prce contnuty rule", forthcomng n the Journal of Fnancal Economcs. Sepp, Duane J. 1997. "Lqudty provson wth lmt orders and a strategc specalst", Revew of Fnancal Studes 10, pp. 103-50. Stoll, Hans R. 1978. "The supply of dealer servces n securtes markets", Journal of Fnance 33, pp. 1133-51. 25