Chapter 4 Experimental Random Arrival Markets with Competing Insiders
4.4 Results
4.4.4 The Effects of Order Flow on Price Changes
increase their rate of speculation as a whole by about 3 to 6 additional units in the next thirty second interval.
Over all experiments, insiders tended to over accumulate inventory during periods of low FCE prices and under sell inventory during periods of high FCE. This is reflected in the slope coefficients FCE<FIP Dummy and FCE>FIP Dummy not summing to zero, although both estimates are in the direction predicted by theory.
Table 4.5: Inventory Accumulation Rate of Uniformed Traders
Variable Coefficient
Lagged Rate of Inventory Accumulation per
Uninformed Trader 0.30***
Lagged Rate of Inventory Accumulation per
Informed Trader 0.58***
FCE<FIP Dummy 0.21**
FCE>FIP Dummy -0.11*
Constant -0.01
R2 0.57
* p<.1, ** p<.05, *** p<.01
(decrease), it is either because of an increased (decreased) rate of buy (sell) market orders eroding asks (bids) in the sell (buy) order book, or because of an increase (decrease) in the level of limit prices at which market orders transact. By submitting a higher bid or lower ask than the current market prices, an insider risks revealing the direction of his or her information prior to making a transaction. Placing market orders, on the other hand, also reveals a small amount of insiders’ information to the market, but does so after the insider has already transacted. Exactly why insiders have this preference is not fully explained by the Glosten and Milgrom model and is hence left as an assumption rather than a consequence of utility maximizing behavior.
Alton (Chapter 1) shows that the informational content of trades can be
estimated by looking at runs in trade initiation. When we apply the same methodology to our experimental data, we discover that, while the direction of trade initiation does appear to impact prices, the actual amount traded in each run, paradoxically, does not affect prices.
Result 7: Asymmetric information in Random Arrival Markets is not transmitted through signed order flow. The direction of order flow however, does impact prices.
Table 4.6: Effect of Signed Run Size on Traded Prices
Variable Coefficient
Positive Order Flow
Dummy 31.96***
Size of Trade Run 0.24
Constant -16.06***
R2 0.14
* p<.1, ** p<.05, *** p<.01
According to the results presented in Table 4.6, positive order flow (buy market orders) tend to increase traded prices by approximately 32 Francs per run and negative order flow tends to decrease prices by about 16 Francs per run. The fact that prices tend to increase more on a positive run than they decrease on a negative run, is due to a combination of parameter choices (tending to shift FCE prices upward over the course of an experiment) as well as mistakes made by insiders (such as in 080727 pd. 3, in which insiders accumulated too much inventory during the first half of the experiment and failed to unload all of it by the end of the experiment). The size of the signed order flow, on the other hand, which should be the only significant explanatory variable, is not significantly different from zero. While this indicated that prices generally move in the direction predicted by theory, it also indicated a general deficiency in the theory.
Figure 4.6: Scatter Plot of Price Changes Vs Size of Run Size
Figure 4.6 shows a scatter plot of the raw price change data in which it can be clearly seen that buy market orders tend to increase prices and market sell orders tend to decrease prices, but there is no linear relationship between the size of a run in trade initiation and its effect. A part of this result may be due to insiders’ choice of whether to transact via market or limit orders.
A common assumption regarding informed traders is that informed traders trade primarily through market orders. This is due either to the mathematical complexity involved in creating theories in which insiders submit both limit and market orders, or because market orders are believed to carry less information than limit orders and insiders never want to reveal their information to the market. Whatever the reason, we
-40 -20 0 20 40 60 80 100
-800 -600 -400 -200 0 200 400 600
Size of Signed Run (number of units traded)
Change in Price
state this as Hypothesis 6, and test it by comparing the proportion of limit orders sent by insiders to the proportion of limit orders sent by non-insiders.
Hypothesis 6: Informed subjects will always submit market orders.
Result 8: We reject hypothesis 6. Informed subjects submitted both market and limit orders in the same proportion as uninformed traders.
There were no significant differences between insiders and uninformed agents in terms of the proportion of limit and market orders that both types of agents submitted.
Both uniformed and informed trades submitted about 60% of their orders in the form of limit orders. This result can be seen in Table 4.7 below, which lists the number of limit and market orders submitted by insiders and uninformed traders.
Table 4.7: Market And Limit Order Submission
Number of Orders
Submitted by Insiders
Submitted by Uninformed
Limit Bids 2693 251 2444
Limit Asks 2196 245 1951
Market Buy 2063 132 1931
Market Sell 1497 160 1337
Excluded 080727 pd 3
Result 9 helps to explain why price movement does not seem to be directly related to signed order flow. Such a result may also explain why the measured effect of asymmetric information in the Australian Stock Market in Chapter One appears to be so small. If informed traders attempt to hide their identities in electronic limit order markets by placing both market and limit orders in the same proportion as the rest of the market, such behavior can attenuate the measurements of the effect of asymmetric information toward zero.