S ISTEMA N ACIONAL DE G ESTIÓN DE R IESGOS
4.2. Reestructuración: el Sistema Nacional de Gestión de Riesgos en Ecuador a partir de 2007.
4.2.1. Transformaciones del sistema político ecuatoriano.
by another dealer for the entire order. Hence, it may take some time before the entire order is fulfilled. The dealer can also choose to withdraw his limit order. This difference may explain the significant coefficient on absolute trade size. There is evidence that Dealer 1 submits limit orders on electronic broker systems (incoming trades) to adjust his inventory position after an incoming direct trade. Dealer 1 also tends to trade outgoing if the previous trade was incoming. Both Dealer 2 and Dealer 4 tend to submit limit orders when their absolute inventory position is large. However, in table 2.5 we find no evidence that these dealers shade quotes to control their inventory position. Moreover, our findings are consistent with those of Manaster and Mann (1996). Similar to futures scalpers, the inventories of Dealer 2 and Dealer 4 show strong mean reversion, while there is no sign of quote shading in incoming trades. As noted by Manaster and Mann (1996) for futures scalpers, our FX dealers are not merely passive order fillers, but are profit-seeking individuals with heterogeneous levels of information and/or trading skills.
Table 2.9 outlines the different options available for dealers through broker trades. Trades that increase the absolute size of the inventory are accumulating, while trades that decrease the absolute size of the inventory are decumulating. The dealers use outgoing trades both to reduce and to increase their absolute inventory position. This clearly illustrates that their target inventory position varies. The target inventory position can vary for two reasons: i in anticipation of a
customer order or ii for speculative reasons. While both motives are relevant for Dealer 1 and
Dealer 3, only the last motive is relevant for Dealer 2 and Dealer 4. For Dealer 2 and Dealer 4 there is a tendency for outgoing trades through D2000-2 to increase the absolute inventory position, while incoming trades through EBS decrease the absolute inventory position. To some extent, a similar trading pattern applies to Dealer 1. For Dealer 3, there is no similar pattern. Most of his electronic broker trades are executed using D2000-2. During the week of our study, only a very small number of NOK/DEM trades were executed through EBS.
Table 2.9: Accumulating and decumulating broker trades Incoming Outgoing
Electronic Electronic
broker trades broker trades Voice- D2000-2 EBS D2000-2 EBS broker Sum Dealer 1: Decumulating trades 39 154 113 98 36 440 Accumulating trades 70 96 163 79 21 429 Dealer 2: Decumulating trades 7 131 44 52 NA 234 Accumulating trades 28 20 147 12 NA 207 Dealer 3: Decumulating trades 24 0 20 2 9 55 Accumulating trades 35 0 20 2 7 64 Dealer 4: Decumulating trades 66 122 35 15 NA 238 Accumulating trades 44 46 73 20 NA 183
Trades that increase the absolute size of the inventory are accumulating, while trades that decrease the absolute size of the inventory are decumulating. Voice-broker trades are not signed.
2.6.2 Information effect
The size of quantity demanded and bid and ask spread in direct trades
Microstructural models suggest that, in absence of an inventory effect, a positive relationship be- tween spread and absolute trade size would be evidence of an information effect. We use the D2000-1 conversations which ended with a trade to examine this. Roughly 70% of all D2000-1 conversations ended with a trade. This indicates that the quotes are highly competitive. Transac- tions are always executed at the bid or the ask price. The advantage of using the bid and ask quotes from the conversations is that all noise between transactions are excluded.
Table 2.10 presents descriptive statistics of the observed bid and ask spread from D2000-1 conversations which ended with a trade. The average spread quoted by Dealer 1 is 1.985 pips (or 0.0001985 DEM). The DEM/USD spread varies between 1 and 4 pips. The average spread quoted by Dealer 3 is 15.515 pips (or 0.0015515 NOK). The NOK/DEM spread varies between 5 and 35 pips. Measured in percent the average DEM/USD spread is 0.011. The average NOK/DEM spread is more than three times higher and equals 0.037%. Note that both the DEM/USD spread and the NOK/DEM spread are positively correlated with absolute quantity traded.
Table 2.10: Descriptive statistics: Bid and ask spread in direct trades Dealer 1 Dealer 3 DEM/USD NOK/DEM Average 1.985 15.515 Median 2 12.5 Stdev 0.507 8.601 Min 1 5 Max 4 35 Correlation between
spread and absolute size 0.455 0.760
Observations 62 67
The numbers are in pips. Average is the average spread observed from D2000-1 conversations which ended with a trade. Similarly, we calculate the median, standard deviation, minimum and maximum values. Correlation between spread and absolute size is the simple correlation between the spread and absolute trade size.
To test whether the quoted spread increases with quantity, we run the following regression for Dealer 1 and Dealer 3:
Spreadt β0 β1 abs Qjt εit
(2.15) where abs Qjt is the absolute quantity. The results in table 2.11 show that the bid and ask spread
increases with quantity demanded by the initiator of the trade. First, consider the results for Dealer 1. Typically, the spread for a USD one million request (DEM/USD) is 1.90 pips 1 747 0 138 .
He tends to increase the spread to 3.13 pips 1 747 10 0 138 when the quantity requested
increases to USD ten million. The dealer increases the spread with 1.4 pips per USD 10 million 10 0 138. Note that this increase in quoted spread is much smaller than estimated by Lyons
(1995) for the effective spread (see table 2.5). Lyons finds that the effective spread increases with 2.8 pips per USD ten million traded. Hence, Lyons dealer increases his spread twice as much as
2.6. MICROSTRUCTURE EFFECTS UNDER THE NEW MARKET STRUCTURE 51