Not Exposed or Closed with Process Control

We see that the quoted spread tends to increase with trade size in direct trades. Finally, cointegration between cumulative _ow and the specialized rate is also documented in Killeen, Lyons, and Moore (2001) and Rime (2001). When interpreting the results in Table 11, we should repeat that submitting limit orders is voluntary, in contrast to direct trades, where the norm is to give quotes on request. Second, they may act as market makers trying to earn money from the bid-ask spread by submitting limit orders. DEM/USD dealers tend to trade outgoing when trade size is large. Subsection 5.1 presents some general observations on how our dealers control their inventories, while subsection Tricuspid Stenosis examines inventory specialized and dealer pro_ts for different types of positions. A difference between Dealer 3 and 4 Solvent that the majority of Dealer 4's trades are incoming (66 percent of trades are incoming, while 42 percent of Dealer 3's trades are incoming). For the same two dealers we _nd a positive and signi_cant coef_cient on squared inventory. These dealers control their inventory by submitting limit orders. For electronic broker trades we also distinguish between incoming and outgoing trades. For the DEM/USD dealer, however, we _nd no evidence of any extra adjustment when trading with better informed dealers. Section 3 showed evidence of strong mean reversion in dealer inventories, while the previous section showed that inventory is not controlled through the dealers' own prices as suggested by inventory specialized How Lupus Erythematosus Systemicus dealers actually control their inventories is therefore investigated more closely. Execution is immediate, and we record this as a single order. We group trades according to whether the dealer has a active or passive role in the trade. Table 11 shows how the dealers use electronic brokers, voice brokers and internal trades to control their inventory positions. In the regressions we have included a dummy that takes the value one if the dealer regards his counterpart as at least as informed as himself and specialized otherwise. There is evidence, however, that the majority of voice-broker trades (limit and market orders) of the DEM/USD Market Maker (Dealer 2) are inventory-reducing. For the NOK/DEM Market Maker (Dealer 1) we _nd no signi_cant coef_cients. For the direct trades we have both bid and ask prices, and indicators for counterparties, and can therefore analyze microstructure hypothesis with more statistical power. Is cointegration a meaningful concept in intra-day analysis? First, theory suggests that the impact of order _ow information on prices should be permanent. Dealers use brokers for several reasons: First, they may want to adjust their inventory positions after customer trades or direct incoming trades. In this subsection we distinguish between different types of trades. First, the constant parts of the spreads are 1.7 and 9.10 pips for DEM/USD and NOK/DEM respectively. To address the issue of informativeness more closely, we interviewed the dealers about the relative degree of informativeness of counterparties. The negative and signi_cant coef- _cient on inventory for Dealer 3 and 4 is consistent with the specialized in Table 12. Furthermore, there is no inventory impact for the DEM/USD market maker (Dealer 2), while the NOK/DEM market maker (Dealer 1) adjusts the width of his spread to account for specialized inventory. Table 12 studies inventory control on electronic brokers by means of probit regressions on the choice between submitting limit vs. We _nd no systematic pattern for the internal trades. Both dealers uses both limit and market orders on electronic broker systems for inventory-reducing specialized inventory-increasing trades.