The objective of this chapter is to define the term informed trading and to determine the adequate measure for identifying informed trading on the trader level for this study. It starts with a brief introduction to the information paradigm, followed by a classification of traders along different motives for trading and the corresponding level of information. Common measures of informed trading will be briefly presented.111 The choice of method follows (i) the need to identify the level of information for each transaction and (ii) data restrictions, i.e. there is no possibility to rebuild the order book.


Adverse Selection Limit Order Market Maker Order Book Market Order 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 111.
    Measures for information asymmetry can be distinguished into two groups: (i) microstructure measures and (ii) corporate finance measures (e.g. analysts’ forecasts, market-to-book or earnings-price ratio). This study solely focuses on microstructure measures. The latter have been additionally implemented by Clarke/ Shastri (2000) and Van Ness/Van Ness/Warr (2001a).Google Scholar
  2. 112.
    Fama (1970), p. 383 distinguishes three forms of information efficiency depending on the level of information incorporated in prices (weak, semi-strong, strong market efficiency). According to Ross/Westerfield/Jaffee (1996), p. 343ff. securities markets can be expected to be at least semi-strong efficient, meaning prices will reflect all publicly available information.Google Scholar
  3. 113.
    The initial contribution was made by Bagheot (1971). Copeland/Galai (1983) and Glosten/Milgrom (1985) show that asymmetric information alone is sufficient to explain the existence of bid-ask spreads. Kyle (1985) analyzes informed traders and their strategic trading behavior. Easley/O’Hara (1987) present a dynamic model for dealer behavior under the assumption of asymmetrically informed traders.Google Scholar
  4. 114.
    See Bagheot (1971), p. 13. The trisection did not prevail as consensus in the literature. In most of the theoretical and empirical literature, informed traders and uninformed (liquidity) traders are distinguished. Noise traders are included in the category of uninformed traders.Google Scholar
  5. 115.
    For an analysis of noise traders, see Black (1986).Google Scholar
  6. 117.
    See Harris (2003), p. 222ff.Google Scholar
  7. 118.
    See Hasbrouck (1990), p. 232. For Germany §§ 12–14 WpHG define what securities are eligible for insid trading and who insiders are, and finally prohibit insider trading. For a detailed description of these paragraphs, see Behr (2000), pp. 24-33.Google Scholar
  8. 119.
    See Madhavan (2000), p. 217. Chung/Charoenwong (1998), p. 15ff. study how trading of corporate insiders affects liquidity measured through the bid-ask spread. Results reveal that spreads increase with trading activity by insiders. Specialists increase their spreads depending on the size of incoming orders. This is in line with results for informed traders.Google Scholar
  9. 120.
    See Harris (2003), pp. 176–194.Google Scholar
  10. 121.
    See Harris (2003), pp. 222–243.Google Scholar
  11. 122.
    Bagheot (1971), p. 12 states: “Every time one investor benefits from a trade, after all, another loses.”Google Scholar
  12. 123.
    In contrast to informational frictions, real frictions hinder all traders alike. See Stoll (2000), p. 1483.Google Scholar
  13. 124.
    See Hasbrouck/ Schwartz (1988), p. 11 or Schwartz/Whitcomb (1988), p. 48.Google Scholar
  14. 125.
    See Harris (1990), p. 6f. Harris (1998), p. 3 finds evidence for uninformed traders acting as pre-committed traders when spreads are wide and the time frame until trading ends is still distant.Google Scholar
  15. 126.
    Admati/ Pfleiderer (1991) provide a detailed theoretical analysis of sunshine trading, i.e. its effects on liquidity, volatility, and trader profits in a market with information asymmetry. Sunshine trading changes the nature of information asymmetry in the market, as sunshine traders, who can convince the other market participants that they are uninformed, receive better prices. However, uninformed traders that are not able to signal their status credibly are worse off as their costs increase.Google Scholar
  16. 127.
    The trader types and motives follow Harris (2003), p. 199.Google Scholar
  17. 128.
    See Oesterhelweg (1998), p. 10. Dennis/Weston (2001), p. 1 explain that institutions are more likely to be informed than private investors as they generate an informational predominance through economies of scale in information processing and acquisition.Google Scholar
  18. 129.
    See Demsetz (1968), p. 35ff. He demonstrates that the bid-ask spread is the result of two equilibrium prices for buys and sells respectively.Google Scholar
  19. 130.
    See Copeland/ Galai (1983), p. 1457f. Hasbrouck (1988), p. 229 states that the inventory and the information paradigm are not competing, but both effects can be seen in practice. He models the simultaneous existence of both effects.Google Scholar
  20. 131.
    While Demsetz (1968) and Tinic (1972) argued that the spread exists due to immediacy services, Stoll (1978) and Amihud/Mendelson (1980) formally modeled this component.Google Scholar
  21. 132.
    Based upon results provided by Bagheot (1971), Copeland/Galai (1983), Glosten/Milgrom (1985), and Easley/O’Hara (1987) provided theoretical models, while Glosten/Harris (1988) presented the first empirical analysis. Copeland/Galai (1983), p. 1459f. define the optimum spread of the market maker as a trade-off between reducing the costs of adverse selection when widening the spread and losing profitable trading opportunities with uninformed traders accordingly. Glosten/Milgrom (1985), p. 72 demonstrate that a spread exists even if the market makers’ fixed and variable costs are zero and competition ensures that his profits are also zero. Easley/O’Hara (1987), p. 88f. include trade size as a signal for adverse selection in their model.Google Scholar
  22. 133.
    See Stoll (1978), p. 1153, Glosten/Milgrom (1985), p. 72, and Hasbrouck (1988), p. 230.Google Scholar
  23. 134.
    See Stoll (1989), p. 118f. Stoll (1989), p. 129 finds that the inventory holding cost is only 10% of the spread, although he analyzes a market with a specialist (NYSE). De Winne/Majois (2004) compare different spread decomposition models for Euronext and discuss problems arising from implementing models that explicitly compute an inventory component. They find that these models do not provide consistent results for the order-driven market structure of Euronext.Google Scholar
  24. 135.
    See Roll (1984), Stoll (1989), and George/Kaul/Nimalendran (1991).Google Scholar
  25. 136.
    See Glosten/ Harris (1988), Lin/Sanger/Booth (1995), Huang/Stoll (1997), Madhavan/Richardson/Roomans (1997).Google Scholar
  26. 137.
    Roll (1984), p. 1127 provides a model based on serial covariance of transaction prices. He presents a procedure that allows inferring the effective spread directly from a series of transaction prices. The model by Stoll (1989), pp. 116–123 introduces the concept of the realized spread, defined as the difference between the price at which a market maker buys at one point in time and the price at which he sells at a later point in time. Based on the covariance of the price change and the quote change, the realized spread is estimated as a fraction of the quoted spread of the market maker. George/Kaul/Nimalendran (1991), p. 628 and p. 635f. demonstrate that Roll (1984) and Stoll (1989) provide results that reveal a downward bias as they exclude any time variation in returns. Accordingly, they provide a spread decomposition method that includes time variation.Google Scholar
  27. 138.
    See Glosten/ Harris (1988), p. 128. Lin/Sanger/Booth (1995) also provide a two-way decomposition. Huang/Stoll (1997) develop a three-way and a two-way decomposition model.Google Scholar
  28. 139.
    See Hasbrouck (1988, 1991a, 1991b, 1993, and 1995)Google Scholar
  29. 140.
    Van Ness/ Van Ness/ Warr (2001b), p. 3 demonstrate that the adverse selection component of the spread depends on the market structure of an exchange. Chung/Van Ness/Van Ness (2004), 269f. argue that-in line with their results on the proportions of limit orders that do not reflect a participation of the specialist (1999)-spread decomposition results for the bid-ask spread that include limit orders should be interpreted with care.Google Scholar
  30. 141.
    Brockman/ Chung (1999), p. 235ff. find for a sample of 345 instruments traded at the Stock Exchange of Hong Kong (SEHK) that approx. 32% of the spread are due to adverse selection. They implement the model developed by Lin/Sanger/Booth (1995), which decomposes the spread into two components, the adverse selection component and order processing costs.Google Scholar
  31. 142.
    Iversen (1994) analyzes average spreads of DAX instruments in IBIS and on the floor for data in 1991. Treske (1996) investigates spread components in DAX instruments in IBIS. The adverse selection component is on average 22%, while small DAX companies reveal a component as high as two thirds of the spread. Wolff (2003) analyzes the spread components in Xetra comparing results for two different trading segments.Google Scholar
  32. 143.
    DanÍelsson/ Payne (2001), p. 1 explain the popularity of spread decomposition models due to the fact that the existing literature on the inventory and asymmetric information paradigm clearly defines the components of the spread and that most databases did not provide any information concerning the depth in the order book.Google Scholar
  33. 144.
    The results vary between 8% and 40%: George/ Kaul/ Nimalendran (1991), 8–13%, Lin/Sanger/Booth (1995), approx. 35%, Huang/Stoll (1997), approx. 10%, Madhavan/Richardson/Roomans (1997), up to 40% and Glosten/Harris (1988), 25–40%.Google Scholar
  34. 145.
    Clarke/ Shastri (2000) analyze the models provided by Madhavan/Richardson/Roomans (1997) and Huang/Stoll (1997), concluding that the former yields implausible estimates 14% of the time and the latter about 60% of the time. Van Ness/Van Ness/Warr (2001a) support the analysis of Clarke/Shastri (2000), as they find implausible estimates for Madhavan/Richardson/Roomans (1997) 18% of the time and 50% for Huang/Stoll (1997). However, the Glosten/Harris (1988) model shows 0% implausible estimates and the Lin/Sanger/Booth (1995) model reveals implausible estimates only 0.5% of the time.Google Scholar
  35. 146.
    The initial model as well as extensions have been implemented for different topics and markets by Easley/ Kiefer/ O’Hara (1996, 1997a and 1997b), Easley/O’Hara/Paperman (1998), Brockman/Chung (2000), Grammig/Schiereck/Theissen (2000 and 2001), Dennis/Weston (2001), Easley/Engle/O’Hara/Wu (2001), Easley/Hvidkjaer/O’Hara (2002 and 2004), Hanousek/Podpiera (2002), Heidle/Huang (2002), Chung/Li (2003), Jain/Jiang/McInish/Taechapiroontong (2003), Chung/Li/McInish (2005), Venter/De Jong (2004), Lei/Wu (2005), Brown/Hillegeist (2006), and Goldstein/Van Ness/Van Ness (2006).Google Scholar
  36. 147.
    See Easley/ Kiefer/ O’Hara/ Paperman (1996), pp. 1408–1415.Google Scholar
  37. 148.
    See, among others, Brockman/ Chung (2000), Grammig/Schiereck/Theissen (2000) and (2001), Hachmeister/Schiereck (2006), and Ma/Hsieh/Chen (2001).Google Scholar
  38. 149.
    Recent critics voiced by Boehmer/ Grammig/ Theissen (2007) based on trade misclassification due to implementing trade classification algorithms do not play a role in automated order-driven markets: Trades can only take place at the posted bid and ask prices, as per definition price improvement is not possible. The trade initiator of any trade can be identified through the available order entry and execution timestamps for all orders involved in the trade.Google Scholar
  39. 150.
    See Sandas (2001), p. 721ff.Google Scholar
  40. 151.
    See Huang/ Stoll (1996b), p. 12f. Perold (1988), p. 5f. calculates the so-called implementation shortfall, where the relevant midpoint to be implemented is the midpoint at the time of the order entry decision and not the order entry itself. However, exchange databases do not include the time of the decision but the time of order entry. In a fully automated electronic trading system that allows order entry with no technical delays, the order entry timestamp can reflect the time of their decision. In that case, the results of the implementation shortfall and the effective spread would be similar.Google Scholar
  41. 152.
    See Huang/ Stoll (1996b), p. 14 or Bessembinder (1999), p. 395.Google Scholar
  42. 153.
    Bessembinder (1999) and (2003b) implement a thirty-minute time frame, while Bessembinder (2003a) chooses a ten-minute difference. Huang/Stoll (1996b), pp. 22–27 implement five-and thirty-minute differences and find that thirty-minute price reversals are only slightly larger than price reversals calculated based upon the five-minute difference.Google Scholar
  43. 154.
    See The rule became effective on 30 January 2001. To ensure comparability of results, especially to the US equity markets, this study follows the SEC rule.Google Scholar
  44. 156.
    See Bessembinder (1999), p. 393.Google Scholar
  45. 157.
    Huang/ Stoll (1996a, 1996b), Bessembinder (1999, 2003a, 2003b), Weston (2000), SEC (2001), Theissen (2002), Handa/Schwartz/Tiwari (2004), Boehmer (2005), Frey/Grammig (2006).Google Scholar
  46. 158.
    Bessembinder (2003b) discusses the issues for assessing trade execution costs for the US equity exchanges. The described issues, i.e. the estimation of trade direction or the general delay for quoted spreads, are not relevant for the Xetra trading systems. See Chapters 6.2.2 and 6.2.3.Google Scholar
  47. 159.
    Spread decomposition models were implemented by Wolff (2003), the structural models by Grammig/Schiereck/Theissen (2000 and 2001), Hachmeister/Schiereck (2006), and Frey/Grammig (2006), and the ad hoc method also by Frey/Grammig (2006).Google Scholar
  48. 160.
    See Frey/ Grammig (2006), p. 1028. Chung/Li (2003), pp. 266–270 demonstrate that the adverse selection component of the spread estimated based on Glosten/Harris (1988) and Lin/Sanger/Booth (1995) and the results for PINF reveal a significant positive relation. Barclay/Hendershott/McCormick (2003) implement the Hasbrouck (1991a, 1991b) VAR method and the ad hoc method, with both results pointing in the same direction. Chung/Li/McInish (2005), p. 1667 analyze the PINF and Hasbrouck (1991a, 1991b) VAR model and find that results are comparable.Google Scholar

Copyright information

© Deutscher Universitäts-Verlag | GWV Fachverlage GmbH, Wiesbaden 2007

Personalised recommendations