top of page

An Empirical Analysis of Strategic Behaviour Models

Zyanza Research Series

Comerton-Forde C., O'Brien M., Westerholm J., (2007), Australian Journal of Management

Original articles can be found here


Comerton-Forde, O’Brien and Westerholm (2007) demonstrate that a substantial proportion of intraday trading patterns can be explained by strategic trading behaviour of market participants [i]. An extensive empirical literature documents distinct daily trading patterns in equity markets [1]. There are significant seasonal intraday effects in intraday patterns whereby volume and volatility is concentrated at the open and close of the trading day with spreads widest at these times. One of the explanations put forward relates to strategic trading between informed and uninformed traders, in short an information asymmetry effects. It is this strategic interaction between various market participants that forces daily systematic patterns in stock markets to appear. Comerton-Forde et.al utilize these results and attempt to empirically explain the reasons behind these observations.


On the whole, strategic behaviour models suggest that it is the trading pattern and timing of trading of informed and liquidity traders that tend to dictate systematic intraday patterns. In particular, the authors look at three distinct predictions, namely (i) liquidity and informed traders tend to enter the market simultaneously, (ii) proportion of informed traders decreases through the day and (iii) the proportion of informed traders will be the highest at the open and close of the trading day. Furthermore, the authors test proposed suggestion that the proportion of informed trades is positively related to spreads and also that spreads are positively related to volume. Hence, the higher the trading volume, the higher the likelihood of informed traders to be present on the market and the higher the chance of the rest of the market operators to adjust bid-ask prices in protection against the informed players.


Datasets

Comerton-Forde et.al use detailed and unique Helsinki Stock Exchange (HEX) and the Finnish Securities Central Depository (FCSD) database to classify market participants into either informed or liquidity traders based on their stock picking capabilities. The database contains trade-by-trade information over 30 major HEX stocks from 12 April 1999 to 26 May 2000.


Our research extends the work done by Comerton-Forde et.al, as such it is in place to critically assess the methodology used and analyze the results derived by the authors. Fundamental feature of the approach is the trader’s classification into informed or liquidity based groups. The FCSD registry data separates all traders into twenty seven investor classes. For each investor class daily percentage change in shareholding is calculated as dividing the change in shareholding by the day’s opening shareholding for all thirty stocks in the data sample. Stock performance is estimated by calculating daily returns over the next three month using official market closing prices. Then, using Gompers and Metrick (2001) suggested approach, the authors control for other variables than can influence stock returns and ownership changes.


All in all, changes in shareholdings per investor class are matched with stock’s three month performance and nine additional variables [2]. Subsequently, investor classes that show a significant positive relationship between stock performance and change in ownership are classified as informed investors while all other classes are classified as liquidity traders. We tend to disagree with this approach. Given the structure of the database, each class as defined by FSCD contains a large number of individual (either nominee or individual) accounts. Classification between informed and liquid traders then is based on aggregated data. Out of twenty seven investor classes only two groups were found to have strong stock picking ability [3]. Within each group, there is still a scope for additional segregation. In short, we believe that in order to form informed as opposed to liquid trader groups, details per individual accounts are to be maintained.


Methodology

We proceed with detailed discussion of the methodology used by Comerton-Forde et.al. To examine intraday variation in bid-ask spread, trading activity, volatility and proportion of informed traders, the authors split the trading day into fourteen half-hourly trading intervals. For each interval they calculate standardized (i) time-weighted bid-ask spread, (ii) number of trades executed, (iii) standard deviation of the mid point prices and (iv) proportion of informed traders. It is the last variable that is of interest to us right now. The authors assume that liquidity demanders are traders that initiate the trade. In other words, they classify trade initiator to be investors that execute a buy above the midpoint or sell below the midpoint. Executions at the midpoint are discarded, without actually providing any reason. Then the proportion of informed traders is calculated as:



The null hypothesis is that the variable under investigation is uniform across trading day. Final regression framework estimated using general method of moments (GMM) with Newey-West correction has the following format:


Where, Wt is the variable in question, w is the fixed effect or a base variable set at 13:30-14:00 HEX lunch break, diis dummy variable representing time of the day equals one if observation t falls within the time frame and zero otherwise, βi estimated intraday coefficient for periods other than the base case.


Furthermore, to investigate the intraday effect on bid-ask spread the authors specify the following relationship:


Where, St is the spread during time period t, the rest of the variables are defined as above.



Summary

In summary, Comerton-Forde et.al provide a direct empirical test of the theoretical models based on strategic behaviour assuming two types of investors, informed and liquidity. All of the variables show significant intraday variations and distinct patterns. Volume of shares traded exhibit J-shaped pattern as opposed to the number of trades with reversed J-shape pattern. This behaviour is attributed to the nature of HEX and its place on the global market stage. North American exchanges open within HEX afternoon trading hours. Relative spreads and volatility measures exhibit reverse J-shape pattern indicating lower costs of trading during the afternoon hours. Additionally, informed as well as liquidity traders tend to trade simultaneously with a concentration of activity at the open and close of the trading session. The results suggest that the higher the trading activity the lower the bid-ask spreads. On the other hand, the higher the volatility and the proportion of informed traders the higher the bid-ask spread.



 

[1] To name a few, Foster and Viswanathan (1993b). Madhavan, Richardson and Roomans (1997), from theoretical models, Foster and Viswanathan (1993a,1994,1996), Admati and Pfleiderer (1988) [2] Details of the regression and definitions of all relevant variables can be found Comerton-Forde et.al pp.11-12. The authors use general method of moments GMM with autocorrelation and heteroscedasticity corrected using Newey-West Corrections [3] Across the period from 12 April 1999 to 26 May 2000, out of the total of 27 investor classes only 2 groups were classified as informed traders, namely (i) other financial institutions and (ii) residences in EU member states.

[i] Comerton-Forde C., O'Brien M.,Westerholm P.J. (2004)."An Empirical Analysis of Strategic Behavior Models", Working Paper Series, University of Sydney.

21 views0 comments

Comments


Commenting has been turned off.
bottom of page