Research On The Liquidity Saving Mechanism Of Interbank Payment Syste

Interbank payment system,as one of the most important financial infrastructures to the country,is offered by central bank or professional institution for banks operating interbank transfers.The security,efficiency and service level of the system directly affects domestic settlement,financial market and cross border payment ability of the country’s currency.The status of RMB,USD,EURO,JPY and other main international currency in international financial market highly correlates with the support of High Value Payment System(HVPS),CIPS,Fedwire,CHIPS,TARGET 2,BOJ-NET and other cross border payment system.The security,efficiency and service level of cross border payment system has always been one of the financial infrastructure problems which is concerned by every country’s central bank,Bank for International Settlements,and commercial banks in decades.Almost every country’s payment system has been going through the progress from Deferred Netting Settlement(DNS)to Realtime Gross Settlement(RTGS)and finally to Hybrid Settlement,which allows development from efficiency pursuing to security pursuing and at last the compatibility of efficiency and security.The design of Liquidity-Saving-Mechanism(LSM)is the fundamental,basic factor,and the spirit of the popular design in hybrid settlement throughout the crossborder payment system design as well.High level liquidity management ability can lower the settlement cost of bank,rise the efficiency and thus become the important reason why a country’s currency can attract other governments,financial institutions,cross border companies and even individuals to use the currency as the settlement tool.With the growing of the China National Advanced Payment System,the number of the sub-system increases while the settlement funding of commercial bank continues to separate.The problem in liquidity management has not been improved but getting more complicated.Research on this problem can be vitally important in improving RMB payment system.The essence of liquidity management problem in cross border payment system is the daily high frequency,side-to-side,massive cross border payments between commercial banks,which forms a real-time varied multilateral digraph.The digraph take bank as point,direction of payment as edge and amount of payment as magnitude.In principle,for every payment instruction,the bank should leave the same amount of settlement funding in advance.In practice,on the other hand,when a bank submitting payment instructions,it will also receive payments from other banks simultaneously.If the settlement funding is not enough,outcome payment instructions can be made to wait in queue until other banks income payments fill the gap and thus release the instructions in queue afterwards,which eventually realize the dynamic balance of settlement funding.Since there is time cost on settlement funding,banks tend to leave as less settlement funding as possible under the premise of satisfying a certain level of payment efficiency which can be measured by queuing time and thus turn the funding into interest-earning asset that can earn a higher income and eventually improve the ability of profit making.The problem appears when all banks tend to use other banks money and leave less settlement funding,which leads to a lower the whole system operation efficiency.There are ways in solving this problem.The first method is to force all the banks to reserve a certain level of liquidity by rule while the second way is to improve service so as to save liquidity for banks as far as it can be by improving the design of the system and the level of management without increasing cost and maintaining the same payment efficiency at the same time.Liquidity saving mechanism is a realization of the second method.This paper studies three main types of liquidity saving mechanism: multilateral offset of RTGS stream queuing payments,offset of high frequency DNS stream payments and improvements on liquidity in initial stage.The research into multilateral offset of RTGS stream queuing payments mainly focuses on the situation that HVPS has not adopted the hybrid settlement mode.There is only RTGS stream but no high frequency DNS stream and for RTGS stream,there is only real time bilateral offset but no multilateral offset or time-set bilateral offset which has a wider coverage.Based on sorting the current available researches and heuristic algorithms in use,the paper suggests a multilateral offset heuristic algorithm which can be applied to HVPS in real use.By developing simulation program and introducing simulation data into simulation,it can be proved that even if multilateral offset function is introduced into the system at a time in the peak of queuing payments,the amount of payment instructions can be reduced by 4.2 billion RMB based on the current mode of HVPS.If set a rather strict level to the net debit limit for an individual payment,the amount of payment instructions can be reduced by 1.7 billion RMB according to simulation.These two numbers take 13% and 5.4% of the peak amount of queuing payment instructions.Considering the massive amount of queuing payment instructions,the peak of which is 31.3 billion RMB for simulation,to introduce multilateral offset function will be quite meaningful.The research on offset of high frequency DNS steam payments mainly contributes to the following aspects:Firstly,it offers a multi-period BCP model which can be used to describe the settlement decision problem and theoretically prove its advantage over single stage BCP model(the BCP model mentioned in the section I of the paper can be considered as single stage BCP model).Traditional research papers usually take BCP problem as a static problem without time dimension.The counterexample in section III proves the importance of time variation in decision,that settle all the available payment instructions is not necessarily better than leave some of instructions for later use.Given that liquidity takes 7% of the payment amount,which is the working interval of the new algorithm,the optimal decision of the multi-period BCP model can lower the payment instructions in queue by 330-370 million RMB which takes 18% of the optimal gap.The optimization proves that the new model improves the accuracy of the problem’s description.Secondly,to solve the problem that the future payment instructions cannot be known by the mean time,with the help of model predictive control method using control theory,a heuristic algorithm is proposed to figure out the optimal decision under multi-period BCP model.The continuously input of the new status given by the new decision,together with the information that what really happens replaces the prediction,can be offered by the model predictive control method to solve the uncertainty of the prediction on future payment instructions.The overall effect of decision can be improved by dynamically updating the decision of next stage.By developing simulation and decision optimization program,simulation data is used in simulation experiments.It has been proved that the decision given by solving multi-period BCP model using model predictive control method is obviously better than the traditional single period method.Under the same assumption that the liquidity takes 7% of the payment amount,it can lower the payment instructions in queue by 10.7 billion RMB per second.Thirdly,since the algorithm is useful,improvements are made by setting the level of liquidity and other related parameters using sensitivity analysis,which allows the algorithm to be more effective.The contribution of the research to the improvement of liquidity distribution in the initial stage is that it proposes a method to optimize the distribution of liquidity based on the multi-period BCP model,and proves it is obviously better than the traditional one by developing simulation and decision optimization program,and simulating on real data as well.Including CHIPS and CIPS,almost all the payment systems that cannot directly use clearing account opened at central bank are facing the calculation problem of liquidity distribution at the initial stage.The normal distribution method is basically made according to the amount of each bank’s payment instructions in practice.It can be straight-forward and relatively fair.Taking advantage of the characteristic that multi-period BCP model takes all-day payment instruction decision as an optimal problem,in this section,the decision goal of minimizing the queuing payments is proposed.Introducing liquidity distribution variation into the model,it has been proved by the simulation based on “new model +real data” that the liquidity distribution given by the new model can lower the amount of queuing payments by 10.8billion RMB queuing payments per second.Facing such a big optimization gap,even using relatively simple method of taking average value for the beforehand estimation of liquidity distribution in the initial stage of the next day,it can fill up about 42.8% of the optimization gap.Specifically,the reduction of over 5billion RMB queuing payments per second is a big number for any payment system and a big improvement as well.CIPS,which is the RMB cross-border payment system in our country,is facing the problem of liquidity distribution in the initial stage everyday by now.High frequency DNS stream is just adopted in 2018.The current operation follows the traditional method of liquidity distribution.It is believed that by introducing the new method,the liquidity of the system,as well as the liquidity saving mechanism of high frequency DNS steam,can be significantly improved.At last,taking a comparison of all the three methods of liquidity saving mechanism mentioned above,two policy views focusing on improving the liquidity saving mechanism payment system’s construction are proposed:Firstly,improving design of the system is no better than improving the management rules.The tendency that building up system is dominating the market and developing the second phase is improving appears in today’s payment management and service aspect.Sometimes,good management is more effective than a good system.A good liquidity distribution algorithm in initial stage takes 0 costs in modification of the system and yet it brings good to the system far more than a complicated system update.Specifically speaking,optimization of liquidity distribution model can potentially promote the efficiency of the whole payment system by 10.8 billion per second,The stable level can be achieved by now is approaching 4.6 billion per second.It performs much more better than the first two methods.Secondly,improves the application of optimal theory in payment system construction and operation management aspects.Optimal theory has a significant influence on improving the design of payment system and operation management as well.It is necessary to have more youngsters with related professional knowledge in the construction of RMB payment system.The ability of system design and operational management can be improved then.And eventually raises service and management level of RMB payment infrastructure,which promotes the internationalization of RMB.