Effects of stock attributes, market structure, and tick size on the speed of spread and depth adjustment Kee H. Chung State University of New York (SUNY) at Buffalo Chairat Chuwonganant Indiana University-Purdue University at Fort Wayne 1
Motivation n n The bid-ask spread is an important measure of market quality because it represents the cost of trading in securities markets. Marketmakers adjust the bid-ask spread in response to new information embedded in order flow, trades, and return volatility, among other factors. 2
n n We know very little about the dynamics of the bid-ask spread. Prior studies offer little evidence as to the speed at which new information is impounded into the bid-ask spread. There is also limited evidence regarding how market structure and trading protocol, such as tick size, affect the speed at which new information is incorporated into the bid-ask spread. In this study, we provide such evidence. 3
Research Questions n n How quickly do specialist/dealer quotes incorporate new information? Do specialist quotes reflect changes in stock attributes more quickly than dealer quotes? How is the speed of quote adjustment related to stock attributes? Do stocks with greater information-based trading exhibit faster quote adjustments toward optimal spreads and depths? 4
n n Do stocks that are traded in less competitive markets (e. g. , fewer dealers) exhibit slower quote adjustments? Do liquidity providers move more quickly to optimal spreads and depths for stocks with more frequent trading and higher return volatility? 5
n n Do smaller tick sizes result in faster quote adjustments to new information? What is the relation between quote adjustment speeds and variable measurement intervals? 6
n n Answers to these questions would be of significant interest to market regulators because they could help design better market structures. Because marketmaker quotes (i. e. , bid-ask spreads) determine trading costs, the speed at which specialists/ dealers adjust quotes to new information is also of concern to traders. 7
How our study differs from previous studies? n n Hasbrouck (1988, 1991 a, 1991 b), Hasbrouck and Sofianos (1993), Madhavan and Smidt (1993), Dufour and Engle (2000) examine how marketmakers adjust quote midpoints to signed trades. Our study examines how quickly marketmakers adjust quote width (i. e. , the bid-ask spread) and depth (i. e. , number of shares at the bid and ask) to their optimal values in response to new information. 8
n n n Huang and Stoll (1996), Barclay (1997), Bessembinder (1999, 2003 c), and Chung, Van Ness, and Van Ness (2001) compare the execution costs of dealer and auction markets. Amihud and Mendelson (1987), Stoll and Whaley (1990), Masulis and Ng (1995) investigate the impact of market structure on return volatility. Affleck-Graves, Hedge, and Miller (1994) compare components of the bid-ask spread between auction and dealer markets. 9
n n n Heidle and Huang (2002) examine the impact of market structure on the probability of trading with an informed trader. Garfinkel and Nimalendran (2003) compare the impact of insider trading on effective spreads between NYSE and NASDAQ stocks. However, none of these studies examine how market structure affects quote adjustment speeds on the NYSE and NASDAQ. 10
n Damodaran (1993) estimates the speed of price adjustment for a sample of NYSE and NASDAQ securities using the partial adjustment model of Amihud and Mendelson (1987). n Thoebald and Yallup (2004) n We focus on spreads and depths 11
Some Conflicting Results n n Jones and Lipson (1999) find that quotes in NYSE- and AMEX-listed stocks adjust more quickly to the information contained in order flow than quotes in NASDAQlisted stocks. Masulis and Shivakumar (2002) show that price adjustments are faster by as much as one hour on NASDAQ using a sample of seasoned equity offering announcements by NYSE/AMEX and NASDAQ companies. 12
Our Main Findings n n n The speed of quote adjustment on the NYSE is faster than the speed of quote adjustment on NASDAQ. In both markets, quote adjustment speed is faster for stocks with a larger number of trades, higher share prices, greater return volatility, and smaller trade sizes. Stocks with greater information-based trading and in more competitive trading environments exhibit faster quote adjustments. 13
n n n The speed of quote adjustment after decimal pricing is significantly faster than the corresponding figure before decimal pricing in both markets. Quote adjustment speed increases with the length of variable measurement intervals. On the whole, our study provides evidence that stock attributes, market structure, and tick size exert a significant impact on the speed of quote adjustment. 14
Market Structure and Speed of Quote Adjustment n Masulis and Shivakumar (2002) hold that quote adjustment speed is likely to be slower on the NYSE for several reasons. • Limit orders on the NYSE cannot be updated instantaneously or conditioned on public information (such as the stock’s last transaction price) and this slow updating of limit orders can delay revisions in the specialist’s bid and ask quotes. • NYSE specialists may buy stocks when prices are falling because of their affirmative obligation to stabilize prices and this behavior can slow quote adjustment process. The specialists’ obligation to provide price continuity can reinforce this effect because it requires them to go tick by tick through the limit order book. 15
• Based on these observations and the fact that NASDAQ is essentially an electronic market in which dealers do not have affirmative obligations, Masulis and Shivakumar conjecture that quote adjustments on the NYSE are likely to be slower than those on NASDAQ. 16
n n Chung, Chuwonganant, and Mc. Cormick (2004) show that a large portion of order flow on NASDAQ is either internalized or preferenced. NASDAQ dealers do not have strong incentives to make quick quote adjustments in response to information shocks. 17
n n n Specialists on the NYSE can adjust quotes quickly after each trade because all order flow in a stock goes through one specialist. On NASDAQ however, dealers are less able to make quick quote adjustments to informed trading because one informed trader can trade simultaneously with several different dealers before the quotes are adjusted. Hence, NASDAQ dealers may be slower in detecting information-based trading than their counterparts on the NYSE. 18
n Specialists on the NYSE may be able to respond more quickly to changes in informed trading because they have face-to-face contact with floor brokers while such contact is not available to NASDAQ dealers because NASDAQ operates on an electronic screen-based system. 19
n Garfinkel and Nimalendran (2003) find that there is less anonymity on the NYSE specialist system compared to the NASDAQ dealer system. They find that when corporate insiders trade medium-sized quantities, NYSElisted stocks exhibit larger changes in proportional effective spreads than NASDAQ stocks. 20
Data n n n The NYSE’s Trade and Quote (TAQ) database. We use the trade and quote data for the three-month period from September 2002 to November 2002. Applied various filters to minimize data errors 21
22
Methodology n We partition each trading day into 13 successive 30 -minute intervals. We then estimate the following partial adjustment model for each stock: $SPREADi, t – $SPREADi, t-1 = a 1 i[$SPREAD*i, t – $SPREADi, t-1] + 1 i, t; (1) $SPREADi, t = the mean dollar spread of stock i during period t $SPREAD*i, t = the optimal dollar spread of stock i during t 23
Optimal Spread $SPREAD*i, t = 0 i + 1 ilog(NTRADEi, t) + 2 ilog(TSIZEi, t) + 3 ilog(PRICEi, t) + 4 i. RISKi, t; (2) NTRADEi, t = the number of transactions TSIZEi, t = the average trade size PRICEi, t = the average share price RISKi, t = the standard deviation of quote midpoint returns 24
Substituting Eq. (2) into Eq. (1) and after rearrangement, we obtain $SPREADi, t – $SPREADi, t-1 = 0 i a 1 i – a 1 i$SPREADi, t-1 + 1 i a 1 ilog(NTRADEi, t) + 2 i a 1 ilog(TSIZEi, t) + 3 i a 1 ilog(PRICEi, t) + 4 i a 1 i. RISKi, t + 1 i, t. (3) 25
We then estimate the model for each stock using time-series observations: $SPREADi, t – $SPREADi, t-1 = 0 i + 1 i$SPREADi, t-1 + 2 ilog(NTRADEi, t) + 3 ilog(TSIZEi, t) + 4 ilog(PRICEi, t) + 5 i. RISKi, t + 1 i, t. n (4) We measure the speed of quote adjustment by the estimate of – 1 i. 26
27
28
29
Matched Sample n We calculate MS for each NYSE stock against each of the 2, 888 NASDAQ stocks in our study sample: MS = [(Xk. N - Xk. Y)/{(Xk. N + Xk. Y)/2}]2, where Xk represents one of the four stock attributes and N and Y, refer to NASDAQ and NYSE, respectively. § Then, for each NYSE stock, we select the NASDAQ stock with the smallest MS. § This procedure results in 539 pairs of NASDAQ and NYSE stocks with similar attributes. 30
Speed of Quote Adjustment and Stock Attributes n Hypothesis 1: The speed of quote adjustment is positively related to both the number of trades and return volatility. Insofar as trades convey information on asset values, liquidity providers may update quotes more quickly for stocks that are more actively traded and have greater return volatility. 31
n Hypothesis 2: The speed of quote adjustment is positively related to share price. Chung and Chuwonganant (2002) show that the minimum price variation is more likely to be a binding constraint on absolute spreads for low-price stocks. Liquidity providers make slower adjustments toward optimal spreads for low-price stocks because the binding constraint prevents them from making such quote revisions. 32
n Hypothesis 3: The speed of quote adjustment is positively related to both adverse-selection risks (and costs) and the level of dealer competition. Liquidity providers are likely to make faster quote adjustments to new information for stocks with greater adverse-selection risks (and costs). This is because the dealer cost of quoting sub-optimal spreads is probably greater for such stocks. Similarly, we hold that liquidity providers make faster quote revisions to optimal spreads when competition is high 33
Measurement of adverse-selection costs and risks n n We use the spread component models developed by Glosten and Harris (1988), George, Kaul, and Nimalendran (1991), and Lin, Sanger, and Booth (1995) to measure adverse-selection cost. We use the algorithm in Easley, Hvidkjaer, and O’Hara (2002) to estimate adverseselection risk. 34
Glosten and Harris (GH) model George, Kaul, and Nimalendran (GKN) model Lin, Sanger, and Booth (LSB) model 35
Easley, Hvidkjaer, and O’Hara (EHO)’s model 36
The likelihood function: 37
Regression Model 38
Regression Model 39
40
41
Does tick size affect quote adjustment speed? n n The NYSE initiated a pilot decimalization program on August 28, 2000 with seven listed issues. The NYSE converted all 3, 525 listed issues to decimal pricing on January 29, 2001. The NASDAQ Stock Market began its decimal test phase with 14 securities on March 12, 2001. All remaining NASDAQ securities converted to decimal trading on April 9, 2001. 42
n n Although there is extensive literature on the effect of tick size on market quality, there is little evidence as to how tick size affects quote adjustment speed. In this study, we contribute to existing literature by investigating the impact of tick size on quote adjustment speed using data before and after decimal pricing. 43
§Hypothesis: The speed of quote adjustment during the post-decimal period is faster than the speed during the pre-decimal period. The penny tick size would be a binding constraint less frequently than the predecimal tick size, allowing liquidity providers to move toward optimal spreads more quickly. A smaller tick size results in greater price competition because it implies a smaller cost of both front running by sell-side intermediaries and stepping ahead of the existing queue by buyside traders. 44
n n For NYSE stocks, we consider the three-month period from May 28, 2000 to August 27, 2000 as the predecimal period and January 30, 2001 to April 29, 2001 as the post-decimal period. For NASDAQ stocks, we consider the three-month period from December 12, 2000 to March 11, 2001 as the pre -decimal period and April 10, 2001 to July 9, 2001 as the post-decimal period. 45
46
47
48
Regression Model 49
50
Speed of depth quote adjustment n n n Marketmakers post both the price and the quantity of shares that they are willing to trade. The analysis of price quotes alone is likely to show an incomplete picture of marketmaker behavior. We analyze how adjustment speed in depth quotes varies with stock attributes and tick size. 51
Estimating the Speed of Depth Adjustment 52
Regression Model 53
54
55
Regression Model 56
57
58
Summary n n Market structure exerts a significant impact on the speed of quote adjustment. Liquidity providers on the NYSE react more quickly to new information than liquidity providers on NASDAQ. Liquidity providers make faster quote adjustments new information for stocks with greater adverse-selection costs and quote competition. 59
n n n Stocks with a greater number of trades, greater return volatility, higher prices, and smaller trade sizes exhibit faster quote adjustments to new information. Liquidity providers on both the NYSE and NASDAQ react more promptly to new information after decimalization. Large tick sizes create friction in exchange markets, and thus delaying price discovery. Quote adjustment speed increases with variable measurement intervals. 60
Limitations and Future Studies n n We assume that optimal spreads and depths are determined by four stock attributes and that liquidity providers make quote adjustments accordingly. To the extent that optimal spreads and depths are also functions of other variables, our empirical models are subject to misspecification. 61
n n We assume that liquidity providers make quote adjustments every 30 minutes according to the target liquidity level (i. e. , the optimal spread and depth) projected by the value of four stock attributes during the same 30 -minute interval. It would be a fruitful area for future research to estimate the speed of quote adjustment using more frequent observations (e. g. , every five minutes) and to assess the robustness of the results. 62
Скачать презентацию Effects of stock attributes market structure and tick
3a253b5fee0f82a1b2be919d8692d392.ppt