The authors build a theoretical model that generates demand for collateral by Large Value Transfer System (LVTS) participants under the assumption that they minimize the cost of holding and managing collateral for LVTS purposes. The model predicts that the optimal amount of collateral held by each LVTS participant depends on the opportunity cost of collateral, the transactions costs of acquiring assets used as collateral and transferring them in and out of the LVTS, and the distribution of an LVTS participant's payment flows in the LVTS.

The authors conclude that the aggregate amount of collateral pledged to the LVTS is quite close to that predicted by the model, when benchmark values are used for opportunity and transactions costs that are based on anecdotal evidence, despite the fact that these costs are likely to vary among participants. If one LVTS participant that appears to face a lower opportunity cost of collateral is excluded from the analysis, the model predicts an aggregate level of collateral that is within 5 per cent of the amount actually held by LVTS participants, on average, between February 1999 and May 2003.

The authors also apply panel-data regressions to the level of collateral held in the LVTS. They find that the results are broadly supportive of the theoretical model. Sensitivity analysis of this model indicates that, when the opportunity cost of collateral increases, the amount of collateral that participants hold could be greatly reduced.