《毕博上海银行咨询Final Deliverables technicalfinal2.ppt》由会员分享，可在线阅读，更多相关《毕博上海银行咨询Final Deliverables technicalfinal2.ppt（14页珍藏版）》请在三一办公上搜索。
1、Table of Contents,1Executive Summary2CRMS Project Review3CRMS Project Commercial Process Improvements4HAVICS System Overview 5HAVICS System Technical Components 5.1Rating Methodology Overview5.2Rating Decision Support Parameters5.3Pricing Methodology Overview5.4Pricing Decision Support Parameters5.5
2、Limits Methodology Overview5.6Limits Decision Support Parameters,Hanvit Pricing Model Development Documentation,PurposeThe purpose of the pricing model is to accomplish the following:Provide a way for Hanvit to record deal information and use this information to analyze potential loansSupply Hanvit
3、with a method to create loan scenarios(i.e.,run an analysis with specific data but do not necessarily save the data to the CRMS database).These scenarios will either be discarded or saved on a per user basis and not necessarily connected to a borrower in the CRMS databaseUse the module to track loan
4、s/deals over an extended period of timeAfford Hanvits policy makers information to set appropriate policies governing the system,Hanvit Pricing Model Development Documentation,MethodologyThe module calculates a loans value by computing the net present value(NPV)of its expected,credit-risk-adjusted c
5、ash flows.Alternatively,the Bank may use the model in determining pricing at which the loan provides NPV=0(This spread measures the amount of credit risk in a loan).,A few parameters are simplified to both decrease complexity and make the model function quickly:Payments are made quarterlyNo pre-paym
6、ents are assumedQuarterly time base-no flow rate used-no backward recursion necessary,Hanvit Pricing Model Development Documentation,Methodology(continued)The general formula for calculating NPV sums the cash flows adjusted for the probability of being in various rating states.,T=loan termCF=cash fl
7、owsDF=Discount FactorsPG=Grade Probabilityg=grade,Hanvit Pricing Model Development Documentation,InputsThere are a number of specific inputs required of the Pricing module.In some cases these are designed to handle various cases in a general format.Loan TypeTermRevolverCollateralCollateral Distribut
8、ionCollateral RatioRepayment MethodStraight Line(with optional grace period)BulletManual AmortizationBorrower Risk RatingBase RateSpreadFeesAnnual UndrawnUpfrontAnnual FacilityFacility Start and End Dates(and grace period length if required),Hanvit Pricing Model Development Documentation,OutputsThe
9、outputs of the pricing module are designed to facilitate loan scenario analyses:NPVPar Spread over Market Opportunity Rate(MOR)Term loan:NPV=0Revolver:drawn and undrawn spreads,Table of Contents,1Executive Summary2CRMS Project Review3CRMS Project Commercial Process Improvements4HAVICS System Overvie
10、w 5HAVICS System Technical Components 5.1Rating Methodology Overview5.2Rating Decision Support Parameters5.3Pricing Methodology Overview5.4Pricing Decision Support Parameters5.5Limits Methodology Overview5.6Limits Decision Support Parameters,Hanvit Pricing Model Development Documentation,Transition
11、Matrix SmoothingSince the pricing module is set to run on a quarterly basis,a quarterly transition matrix with the same rating format as Hanvits rating model is a required set of parameters.Due to initial data limitations,it is currently only feasible to directly observe a yearly transition matrix.A
12、 quarterly transition matrix is therefore derived from the yearly matrix.The initial estimation of individual elements within the yearly transition matrix were extremely noisy.A smoothing method was required before transforming the yearly transition matrix.That method consists of:1)Form block averag
13、e values in the middle of the matrix2)Zero out severe down migrations caused by 97-98 crisis3)Ensure unitary row sums by putting residual on main diagonal.Use this matrix as the input into the flow rate algorithm described below.,Transition Matrix DerivationAfter smoothing the yearly transition matr
14、ix,a flow rate algorithm is used to derive the fourth root representing a quarterly transition matrix consistent with default rates used in the rating model.The algorithm generates the following:1)Calculate matrix X so that A=(I+X),where I=Identity matrix.2)Calculate matrix ln(A)from ln(I+X)=X-.5X2+
15、.333X3-.25X4+3)Set any off-diagonal negative values to zero4)Adjust the diagonal values so that the rows once again sum to zero,creating the Y matrix,the powers of which are needed to take the exponent5)Create the fourth root of A=exp(.25*ln(A),Hanvit Pricing Model Development Documentation,Transiti
16、on Matrix Derivation(continued)The final quarterly transition matrix to be used in the initial model is shown here.,Hanvit Pricing Model Development Documentation,Transition Matrix(continued)Ongoing maintenance of the transition matrix will preferably require two separate undertakings.First,the year
17、ly transition matrix should be calculated every quarter,then smoothed and transformed to a quarterly transition matrix for potential use in the model.Second,an attempt to directly observe and tabulate a quarterly transition matrix should be made.It is likely that the directly observed quarterly tran
18、sitions will be too volatile initially for use in the model due to too few data points from the newly implemented Hanvit rating model.With increasing time,however,it is anticipated that this method will become the preferred source for the actual Pricing model transition matrix.,Hanvit Pricing Model
19、Development Documentation,Base Usage Rates AllotmentInitial usage rates were created from a starting point of a study using the previous Hanil Bank rating system and loans.That study showed usage rates varied from a low of about 3%to a high of around 33%.These ranges were applied to the new Hanvit r
20、ating system as the boundaries for BRR 1 to BRR 9.Usage rates were then interpolated among all the other ratings.The rate of 70%in default was derived from US banking experience.The preferred method of maintenance and refinement of base usage rates requires compiling actual bank experience showing a
21、verage usage rates by rating over time.,Adjustments for Expected UsageThe model requires an expected usage for a facility.Base usage rates by rating are adjusted linearly according to the difference between the expected usage and the base usage by rating.,Hanvit Pricing Model Development Documentati
22、on,NIEAverage net interest expenses(NIE)by rating are required as a component of cash flow calculations.Hanvit currently does not have a sufficient tabulation of lending expenses.In order to develop initial expense inputs,two sources were used.First,KPMG provided approximate US average expense estim
23、ates by rating.Additionally,Hanvit experience suggests much higher expense averages as evidenced by the 299 bps used internally at the Bank in constructing the Prime Rate.Pricing module assumptions were therefore created by starting with US experience and adding 125 bps on top of them across the boa
24、rd.Ongoing experience will suggest adjustments necessary to the quarterly NIE factors.Enhancements to these factors is strongly urged by compiling complete,actual expense estimates.,Hanvit Pricing Model Development Documentation,Credit PremiumsThe pricing module requires credit premiums by rating.Th
25、e initial premiums were developed by relating expected default rates in the transition matrices to risk premiums in the market.The process began with a reasonable sample of credit spreads on three year bonds bucketed by rating.(Coverage within this sample tended to be good for the higher ratings but
26、 thinner in the lower,more adverse ratings.)These spreads represent both expected and unexpected loss.Using the assumption of loss in the event of default equals 65%(see Section 5.2 Rating Decision Support Parameters),expected loss is extracted from the sample.Unexpected loss is derived from the ass
27、umption that unexpected loss increases both with the expected default rate and its volatility.To derive one year credit premiums,a least squares equation was fit to the the three year bond data,then one year default rates were simply plugged into the resulting equation.This method both smoothes and
28、extrapolates spreads that would be observed in the market.There is relative little market information to rely upon due to the state of development of the Korean bond market.As the market develops further,it is expected that one year credit premiums could be compiled from direct market observation.Ma
29、intaining credit premiums in the meantime requires the Bank to update the risk premiums observed in the bond market.,Hanvit Pricing Model Development Documentation,Forward Discount RatesWithin the Pricing model,forward discount rates were derived from the Market Opportunity Rate(MOR)currently in dev
30、elopment by Hanvit and expected to be piloted within the Bank around January 2000.The MOR,a compilation of government debt and bank borrowing rates,was designed to represent the rates at which the most credit worthy banks within Korea could borrow.For the purposes of the Pricing model,any liquidity
31、premiums added to the top of these rates should be removed before full relying on the MOR.To compile forward discount rates,1)term gaps in the spot rates of the MOR were extrapolated linearly,2)forward rates were backed out of the spot rates directly,then 3)a logarithmic fit was applied to smooth ou
32、t aberrations within the discount rates.This last step of smoothing follows usual convention at other large international banks.This method of calculating forward discount rates is a substitute for what is anticipated to come into being in Korea:a broad range of term structures within government debt and bank borrowing rates available within the market.,
