Presented by Numerical Technologies Incorporated.| TITLE |
| ABSTRACT |
CreditBrowser® is a large-scale credit risk management system making use of the sophisticated financial technology. CreditBrowser® is currently the most serviceable credit risk management application with the best performance among various commercial packages.
PortfolioBrowserTM / CreditBrowser® brochure (8 pages, PDF 1,041KB, in English)
(Note) Yes: available/selectable. No: not available
Functional comparison across typical credit risk models CreditBrowser® (Numerical Technologies Incorporated.) CreditMetrics® (RMG) CreditRisk+ (CSFP) Theory Merton Merton Actuarial Computation Simulation Simulation Analytic Output VaR, Marginal VaR, CVaR, Marginal CVaR, RAROC, RAROA, etc. VaR, Marginal VaR, etc. VaR, etc. Evaluation by transaction MTM Yes Yes No DM Yes No Yes Yield spread based on credit rating and recovery rate Fixed value input Yes Yes No
(unenterable)Internal model Yes (for the loan of which spread is not accessible) No No
(unenterable)Correlation by obligor Country / industry sector Yes Yes No (on the implicit assumption by model and inoperable) Economic indicators Yes No No Financial / business ties Yes
(up to 3 parent companies definable for a single corporateNo No LGD (loss given in default) Fixed ratio input Yes No Yes Beta distribution Yes Yes No Truncated normal distribution Yes No No Indicator link Yes
(e.g., real estate price)No No Guarantor Max. 6 No No Multi-period Yes
(absorptive Marcov chain,up to 30 yrs in an increment of 1 yr)No No Rollover Yes
(support for short-term loans)No No Market risk Risk of individual stock incorporated in
(Merton and VaR models implemented).No No Scenario simulation Yes
(expected growth rate to be input per country, sector, and indicator)No No OLAP (On-line Analytical Processing) Yes
(obligor based and transaction based OLAP are supported)No No
Another strength of CreditBrowser® is its significant performance. CreditBrowser® achieves the top performance one-digit higher than any other existing commercial applications in the world at present. Even a 100,000-time precision simulation of a small-sized bank's portfolio can be completed within 24 hours. You don't have to rely on the simplified method in credit risk management any more.
CreditBrowser® is not just a credit risk modeling tool as shown in its basic system design concept:
CreditBrowser® = DataBrowserTM + Credit Risk Models
That is, CreditBrowser® comes with drilling down and hot spot analysis features as an applied model of DataBrowserTM. Benefited from our high performance computing technology, the built-in credit risk computing engine affords high capacity well enough to keep its serviceability even after allocating it to drilling down and computing. These outstanding features together with the advanced credit risk modeling distinguish CreditBrowser® from other risk management software products. As the study of credit risk model is still in the early stage of development, CreditBrowser®'s risk model can be selected among several options. Of course you may include internally developed model in CreditBrowser® along with your business line.
CreditBrowser®, the credit risk management system for major financial institutions, is a registered software with the the Information-technology Promotion Agency, Japan(IPA) (registration number 25295, registered on February 28, 1999), as well as a system that has been escrowed and copyrighted with the Software Information Center (SOFTiC). Its knowledge properties including copyrights is belong to Numerical Technologies Incorporated. CreditBrowser® is a client/server system, requires Microsoft Windows NT 4.0 for client. Supported system environment for server is, Windows 2000/2003 Server, HP Tru64 UNIX, Sun Solaris, and RedHat Linux 6.0.
| INDEX |
CreditBrowser® as an Advanced Business Application Extended from CreditMetrics
| DETAILS |
CreditBrowser® as an Advanced Business Application Extended from CreditMetrics
With regulators such as the Bank for International Settlements (BIS) and the Federal Reserve Board (FED), the sophisticated risk management is regarded as the second important issue after the market risk that has been restricted by the BIS since 1998. To make the risk management more sophisticated is the top priority especially for the Japanese financial institutions which are still suffering from piles of bad loans.
Goals to be required in today's risk management system are: (1) Provide more accurate method in credit risk measurement to overcome the weak point of first BIS regulation. (2) Achieve the credit exposure/profitability management on the basis of portfolio, and the efficient business operation. These goals are getting more concerned not only for the internal control purpose but for compliance and disclosure needs.
CreditBrowser® is a system developed theoretically based on J. P. Morgan's CreditMetrics type model, or the credit model developed by Robert Merton. However, CreditBrowser® offers additional features that is not available in CreditMetrics, for instance, (1) risk/return analysis (RAROC/RAROA), (2) extended system capability such as drilling down to support the actual business needs, and (3) compatibility with larger portfolios. It also includes a brand-new technology of the credit spread model (yield curve model per credit rating and per returning ratio, you don't have to rely on hard-to-observe corporate bond spread), and the like. CreditBrowser® is an extended and enhanced system beyond CreditMetrics both in theoretical and practical aspects.
Run a Larger Scale Monte Carlo Simulation
In credit risk quantification, there exists an engineering barrier of the large scale simulation technique in addition to a barrier of the most advanced financial mathematics represented by the credit derivatives and the latest portfolio theories. These barriers have prevented and prolonged many challenges to put the theory into a real system, even in the United States where the theoretical study would always precede. A credit risk quantification system for practical use is the Everest in financial modeling.
Backed up by the high performance computing (i.e., computing algorithm applied to quantum mechanics simulation, large scale parallel processing), CreditBrowser® can run a case-by-case simulation across the several hundreds of thousands of transactions for the first time in the world. This capacity easily covers the whole corporate loans at a major commercial bank. The most recent benchmarking test proved that CreditBrowser® calculates 150,000 each loan transactions in two hours on HP's 4-CPU AlphaServer 8400. Less than half of above process time was spent for computing the risk model, and the rest was used in structuring DataBrowserTM's multi-dimensional database. These figures may be enough to show the higher operation speed of CreditBrowser® .
Like CreditMetrics, CreditBrowser® supports multiple currencies. You may set the correlationship across countries or industries, and specific factors of each corporation. In the credit risk aggregation by account, case, or borrower, CreditBrowser® runs the 10000-time Monte Carlo simulation on a case-by-case basis by generating higher order correlation random numbers. CreditBrowser® is also responsible for quite large number of transaction cases, supported by the advanced pseudo random number generating techniques of the Mersenne Twister (longer period and higher order of equidistribution random numbers in 219937-1 period) and the singular value decomposition method (enabling triangular decomposition of large non-positive definite matrix).
CreditBrowser® as a Problem-identifying Tool
From a pragmatic perspective, a credit risk model itself, even if it is well established theoretically, may have no usage other than in pricing loan packages. Credit risk models are never effective before they are used with the database.
Thanks to DataBrowserTM's multi-dimensional database, CreditBrowser® enables you to flexibly drill down the data from the entire portfolio to each account and the lowest transaction case using the familiarized Microsoft Windows screens. The accounting items can be defined by department, branch, country, rating, currency, and industry depending on the credit management practices. At every layer you have drilled down over the several tens of thousand of account item combinations or several hundreds of thousand of individual loan, you can use CreditBrowser® 's features of probability density view, three-dimensional analysis of marginal risk, single risk, and returns, and analysis chart display by criteria. Other than the account item tree searching by the mouse, the drilling down feature enables you to jump from a hot spot specified on a two-dimensional chart. CreditBrowser® is a powerful tool that help you find the problem through drilling down the data on the entire corporation scale to the details.
Application of Credit Risk Management Technology
Revolution in credit management - from alternative decision to market- and probability-oriented attitude
- When it comes to credit management, the alternative judgement of "Do they go broke or not?" is the principle and credit is never given to a corporation who may have the potential to go bankrupt. This old credit culture no longer fits in financial institutions under the drastically changing environment of: (1) more and more intense profit competition, (2) demand for crediting on gray zone like venture businesses, and (3) commercialization of credit risk itself as the derivative or loan participation instrument. Their market sections have switched their index from basis point value (BPV) that links to the volume of the instrument, to the stochastic index like VaR. Today credit sections are also urged to shift to the stochastic credit culture like "Give credit to them comparing the expected value of the profit to that of the bad loan."
- Once the latest credit risk model is incorporated in the system, you will find even in a field involving high risk there are some sectors with lower risk due to the strong effect of credit variance. In the conventional sum-up method, the recognition of the total risk aggregation is not accurate because this effect is not counted. If you found a loan portfolio combination with relatively low risk, you can expect a high rate of return by the scale merit. This is not a brand-new technique but is seen in the repackage instruments of whole-loan mortgage and high-yield bond that have already been popularized in the U.S. in the late 1980s.
- Each market section at financial institutions shifted extremely to the maximum loss amount limit (VaR) from balance limit while being influenced by the successive large loss cases, growing derivative market, and second BIS regulation. If credit sections take their way to the stochastic determination from alternative determination, the setting method of credit line need to be reviewed; change to credit VaR (limit to the maximum expected loss) from the balance limit.
- The traditional credit management system takes long time to retrieve and view the credit information from the avalanche of forms or the large database by various criteria such as entire company, branch, industry, instrument, and currency. The second BIS regulation provides a clause stating the "involvement of senior management in risk management," which must be difficult to achieve without a tool that helps them grasp the complicated and huge amount of business lines throughout the company. Unlike a theoretical model, CreditBrowser® as the data sharing and data mining (in combination with DataBrowserTM) tool is absolutely the right staff in this area.
- Today, even a major and sound financial institution should be aware of observations by the market and authorities through its rating, equity price, and financing cost. Under this circumstance, the risk management section has extended its role beyond the cost center. It should be more than strategic and important for a financial institution to show the top management's concern with practicing the risk management with accurate quantification in line with the world's leading banks and regulators.
Coexistence of Arithmetic Capability and Usage
CreditBrowser® 's user interface is provided on a personal computer seamlessly connected to the network while the large server behind is transparent to the user. You can work on CreditBrowser® with familiarized Microsoft Excel or Microsoft Access.
Considering the actual operating environment at a financial institution, CreditBrowser® is designed to be independent from any specific database. The program of CreditBrowser® runs on a middle to high-end UNIX server as a safe and practical choice in today's computer technology level. We recommend to arrange the right hardware set capable of supporting the number of transaction cases and simultaneous log-on users. Please contact our SI vendors for the hardware requirements.
Current version of CreditBrowser® is a two-layer client/server system designed for the use at the credit management section in the head office. In the multi-client environment including branch offices, the three-layer system is preferable. Please contact us or our SI vendors for the optimum system for your case.
CreditBrowser® Version 1.5 Released in October, 1998
The primary feature of CreditBrowser® Version 1.5 is the built-in credit spread model component. This makes you free from the J. P. Morgan's bond spread unavailable other than in the U.S. markets. Support for the domestic credit (in Japan) is enhanced. It may be applicable to the computing of customer loan spread. Feel free to ask us the CreditBrowser® brochure.
Sample screens of CreditBrowser® Version 1.5
Specifications of CreditBrowser® Version 1.5
- Parallel Monte Carlo simulation in 10000-time full valuation is available by transaction case, borrower, and hierarchical account item (up to several hundreds of thousand combinations).
- Provided with longer period and higher order of equidistribution pseudo-random numbers Mersenne Twister (in 219937-1 period) that generates up to total amount of 109 of normal random number vectors.
- Combination of antithetic method (matching of all odd ordinal numbered moments) and second moment matching for improving convergence properties to cancel the deviation of each moment (expected value, variance, distortion, etc.)
- Inverse function method (but not the textbookish polar coordinate) is used for random uniform/random normal conversions to retain precision in computing eventual probability under 10-4.
- Singular value decomposition (SVD) (but not the Cholesky method with fatal problem) is used to handle non-positive definite matrix and underflow error.¥
- Precedence number computing by "introsort".
- Parallel convolution algorithm is used for vector arithmetic system.
- LRU cache algorithm is used for OLAP operation.
- Matrix arithmetic algorithm (unreleased) is used of which time complexity is adjacent to O(N) for the number of borrowers (N).
- Space complexity proportional to the number of combinations of (number of transaction cases + number of borrowers + OLAP).
- High speed computing - just in two hours for simultaneous operation of generating 4-dimensional OLAP hyper cube and modeling in book value and current value with 150,000 transaction cases and 50,000 borrowers (50,000-dimensional large scale Monte Carlo simulation).
- Three built-in credit risk models - book value, current value (J. P. Morgan type), and credit spread model with current value (without corporate bond spread).
- Drilling down (through OLAP) by various criteria from top level to each transaction case with user-settable hierarchy structure.
- Tagged text file for data entry - compatible with the generic database like Oracle, Access, and Sybase.
- External input available for return count and attribute information.
- 2-dimensional chart analysis available for risk/return and single risk/marginal risk.
- Hot spot analysis of the specified loan by clicking and jumping on 2-dimensional chart.
- Internal calculation of credit spread depending on the recovery rate and rating - for future use of customer spread calculation.
- Up to 4 types of rating table (S&P, Moody's, corporate-own rating) available.
- Support for multiple currencies (number unlimited).
- Scenario jump - displays the scenario using rating and z value by each corporate and country/business sector in a specific percentile from the probability density distribution.
- Data export in text file - enables to download the outputs into other system.
- Dynamic data exchange with Microsoft products - cut-and-paste operation
Internal logic of CreditBrowser® Version 1.5
CreditBrowser® Version 2.0 Released in February, 1999
Main improvements and additional features from the Version 1.5
- Color-coding of credit - Show each borrower in bullet by credit type on a chart.
- Internal generation logic of pseudo-cash flow - Support for amortizing loan and nonperforming loan.
- Principal rollover - Support for shorter term loan (ex. less than 180 days)
- Multi-period time horizon - Run absorption, multi-period type Monte Carlo simulation at one year interval up to 30 years.
- Equity price - Equity price risk integrated in by full Monte Carlo simulation on individual equity.
- Chain bankruptcies risk - Link three business sectors and define parentage at one level.
- WHAT-IF feature - Run ad-hoc simulation of new loan/recovery.
- Scenario jump - Display Monte Carlo scenario in a specific percentile.
CreditBrowser® Version 2.2 Released in August, 1999
Main improvements in Version 2.2
- Extended LGD formulas: In addition to (1) fixed ratio and (2) beta distribution (J.P. Morgan's), the new version offers (3) truncated normal distribution and (4) economic indicators-linked security evaluation. Particularly (4) allows you the direct entry of the amount of the pledged collateral to simulate the changes in cover ratio depending on the land price or stock price index.
- Scenario simulation: Enter the expected growth rate by each industry sector. You can include the economic trend in the simulation.
- Enhanced chain-reaction bankruptcy formula: Maximum of 3 affiliations can be defined for a single corporate by the improved definition method.
- Extended stock VaR calculation: Select (1) conventional market VaR approach (ignoring the credit risk) or (2) market & credit integrated Merton type VaR. Both will compute the risk of the stock price in full Monte Carlo on an individual stock basis.
- Two arithmetic modes: Select (1) single-precision for a simulation or test in one million unit and (2) double-precision for the production or a simulation in minimum monetary unit. The simulation count can be set from 100 to 100,000.
- High speed: Supports the precision simulation over 10,000 times and multi-period Monte Carlo. The multi-dimensional database DataBrowserTM built in CreditBrowser® implements the record level lock mechanism and striping (software RAID0) mechanism, improving the performance at more than 30 percent than the previous version according to the parallel processing configuration. The huge data of a whole bank portfolio can be stored in a split storage system.
- Less memory: Operable on the lower workstation or with Linux. Even a 100,000-time precision simulation of a small-sized bank's portfolio can be completed within 24 hours. You don't have to rely on the simplified method in credit risk management any more.
- Three-dimensional analysis chart (see below): As on the two-dimensional chart, click each dot to jump to the details of the case (drilling down). You may select which data item should appear on each axis.
3D Analysis Sample Screen
X-axis=Amount, Y-axis=marginal credit VaR, Z-axis=profit spread
Drilling-down available from each point on the chart.
CreditBrowser® Version 2.4 Released in October, 1999
Main improvements in Version 2.4
- CVaR (Conditional VaR, mean expected loss, expected shortfall below capital) and Marginal CVaR. For further information about CVaR, see: "Thinking Coherently," Risk 10 (1997) and "Risk Management Model Study Group Report," announced in September 3, 1999 by Financial Supervisory Agency (Japan).
- "Risk Contribution," an algorithm to allocate risk capital by each obligor. This feature is a special request from Sumitomo Bank, Co. Ltd.
CreditBrowser® Version 2.5 Released in March, 2000
Main improvements in Version 2.5
Copyright © Numerical Technologies Incorporated,
4-11-6 Jingumae, Shibuya-ku, Tokyo. All rights
reserved.
DataBrowser, CreditBrowser and/or other Numerical
Technologies products referenced herein are
either trademarks or registered trademarks
of Numerical Technologies. Other product
and company names mentioned herein may be
the trademarks of their respective owners.
