In the daily work of producing market values and risk measures, the rates and prices used must be the same across departments, or the figures disagree. Control functions independent of trading, and the risk management function, manage market data by setting in a policy where the values come from and how they are checked.

Why inconsistent market data causes problems

Interest rates, exchange rates, prices, volatility and the other values observed in the market and used to calculate market values and risk measures are called market data. The same position gives a different market value and different risk measures if it is run on different market data. A financial institution therefore decides where to get market data, how to check it and how to keep records of it, and manages it on that basis.

Consider a hypothetical example in which a position changes in market value by 1 million yen when interest rates move 1bp (0.01%). If the 10-year rate used to value it is loaded 10bp higher than the actual rate, the market value is off by about 10 million yen. The direction of the error depends on the direction of the position, so it can make market value and profit and loss look larger or smaller than they really are.

The examination manual of the US Federal Reserve Board (the "examination manual" below; the sections cited date from February 1998 to April 2003) lists, as a common weakness in valuation procedures, heavy reliance on traders' inputs without independent verification of the valuation inputs1. Differences in valuation assumptions or methods between the front office (sales and trading) and the back office (operations and settlement) are another source of error. Such differences can lead to errors in financial reporting and in the assessment of market risk2.

The procedure that uses end-of-day market data to fix market values and profit and loss is covered in Daily Valuation and P&L Sign-Off.

Sources of market data

The sources of market data depend on how much a product trades. Products traded on an exchange and actively traded OTC products can use values from independent sources. For OTC products that trade less, fewer sources are available. The sources listed in the examination manual are as follows.

Source Main use Points to note
Exchange prices Listed products A basis for checking traders' values. Liquidity considerations make the valuation of quotes more complex3
Prices from distribution services Actively traded OTC products Like exchange prices, a basis for checking traders' values4
Broker and dealer quotes OTC products (few quotes are available for products that trade less) A provider's value is not necessarily the same as the internal value. Understand the provider's position and check that the value is reasonable5
The institution's own inputs Values that cannot be loaded automatically An independent control function enters them, although traders may take part in deciding the values6

For OTC products that trade less and for customized products, independent third-party prices are often unavailable, so prices are produced with valuation models and similar methods7. Among the inputs to models, observed option volatility is one of the points prone to errors and manipulation8.

UK rules (the Prudential Regulation Authority (PRA) rulebook) require a financial institution that calculates an additional valuation adjustment (AVA, an adjustment that brings a market value to a prudent value) from market data to consider the same range of market data as in independent price verification. The sources to consider are as follows9.

  • Exchange prices in liquid markets
  • Trades in identical or very similar instruments
  • Executable quotes
  • Consensus service data
  • Broker quotes as a reference
  • The valuation of collateral by the counterparty

The handling of purchased data also has to be decided. The Financial Services Agency (FSA) "Comprehensive Guidelines for Supervision of Major Banks, etc." list points of focus for the price valuation of marketable credit products. When the bank obtains price valuations from brokers or external vendors, the point of focus is whether the bank asks for as much information as possible on their valuation methods and verifies that the prices are reasonable10. The FSA's "Principles for Model Risk Management" (aimed at systemically important financial institutions) treat third-party data and parameters used in models as subject to management in the same way as vendor models. Its examples include asking for as much information as possible and preparing a plan for when the data cannot be used11.

From loading to use

Market data is loaded from its sources, aligned to a common time, checked against independent values, recorded, and then passed to the systems that use it, in that order. The order and the division of work between departments differ by institution.

SourcesData managementUsing systemsExchange, broker,distributionservice valuesLoad the values(including inputs)Align the time,fix end-of-dayvaluesCheck againstindependentsourcesKeep records ofdifferences andresultsSystemsuse themCalculate values,risk, reports
Figure 1Flow of market data from loading, through checking, to use
Figure description

Of the three lanes (sources, market data management, and using systems), the figure shows that the checking work is concentrated in market data management.

In the examination manual (February 1998), the basic approach to loading is to update values automatically when an automated feed is available. Values that cannot be loaded that way are entered by a control function independent of trading6. The items to confirm for controls over loading include how incomplete loads are detected, whether users can override market data, and how the accuracy of manually entered rates and prices is maintained14.

The examination manual expects the valuation time to be aligned. Products in the same portfolio are to be valued at the same time even if they trade in different places. A hedging product and the product it hedges are also to be valued at the same time even if their exchanges close at different times. Quotes that are fresh at the time of valuation are used15. When prices diverge because the cash and futures markets close at different times, a theoretical price may be used16. Which time's value is used as the closing price is an item to put in the valuation policy17.

How to check values

Loaded values are checked against values from sources independent of the traders' inputs. The checks are done at two frequencies, daily and periodic.

In the daily check, the control function resolves and records the differences between the traders' inputs and independent market values. For each difference, it keeps the reason and the effect on profit and loss in a record of differences, and reports large differences to senior management18. The periodic check is one in which a function independent of trading re-verifies the price inputs. It is called independent price verification.

What to check depends on the type of value. The following table collects examples from the examination manual and other sources.

What to check Example methods
Gaps, overrides and stale values in loading The items to confirm for gaps and overrides in loading, and the use of quotes that are fresh at the time of valuation1415
Dispersion of quotes When there are several broker quotes for an illiquid OTC trade, state in the policy how the average is taken19
Values entered directly by brokers The control function checks their accuracy and reasonableness19
Consistency with past values Check portfolio prices for consistency with past values20
How the risk management function obtains its data Whether it obtains data directly without going through the front office. If it uses front office data, whether it verifies that the data is reasonable21

Keeping records of the results

The results of the checks are kept as records that can be traced later. Writing to the data is controlled by limiting who can do it. The examples in the examination manual are as follows.

  • The reason for each difference between the traders' inputs and independent market values, and its effect on profit and loss (the record of differences)18
  • A document that records the monthly valuation results in enough detail to be traced later22
  • Limits on write and edit permissions, periodic review of permissions, and audit reports on user access23

Internal audit checks whether the front office is involved in identifying and measuring risk. One example is to look at the method of registering data, the access rights to the risk measurement system, a review of the history, and similar items24.

Separate databases and reconciliation

The statements below concern data in general, including customer and trade files, and are not limited to market data. The more independent databases there are, the more room there is for errors in the data25. An institution that keeps its customer and trade files in one database can have stronger data controls than one with several data sources. Still, it is rare for a single system to handle everything from input through processing to control for all OTC and listed products, and institutions may combine purchased systems with applications they built themselves26.

When data from several sources is combined, checks and reconciliation are put in place so that the data is not corrupted. If independent databases exist, reconciliation arrangements are needed at each point where several files come together27. When the back office uses a database and software independent of the front office, the differences are reconciled regularly28. Examiners check the types and frequency of the reconciliations performed and the tolerance for differences25.

The overall structure, including how market data is distributed to each system, is covered in Front-to-Back System Architecture.

The guidelines and rules that bear on market data management are as follows.

  • The FSA's Comprehensive Guidelines for Supervision of Major Banks, etc. are not rules on market data itself, but they include points of focus on the price valuation of credit products and on measuring VaR. They also include points of focus on risk data aggregation, which ask G-SIBs and D-SIBs (global and domestic systemically important banks) to meet the accuracy and completeness needed. To reduce errors, automating most of the aggregation is also a point of focus29.
  • The FSA's "Principles for Model Risk Management" (November 12, 2021) cover the management of external data. They apply to G-SIBs and D-SIBs designated by the FSA, among others30.
  • The PRA's rules list the sources of market data for additional valuation adjustments and list the closing-price time among the items to include in the valuation policy917.

Notes

  1. Board of Governors of the Federal Reserve System, "Trading and Capital-Markets Activities Manual," Section 2060.1 (February 1998), pages 5–6, "Revaluation." Common weaknesses include heavy reliance on traders' inputs and no independent verification of the inputs. ↩

  2. Examination manual, Section 2060.1 (February 1998), page 6, "Revaluation." Differences in valuation assumptions or methods between the trading department and the back office lead to erroneous financial reporting and market risk assessment. ↩

  3. Examination manual, Section 2060.1 (February 1998), page 6. Exchange prices are an objective check on traders' values, although liquidity considerations make the valuation of quotes more complex. ↩

  4. Examination manual, Section 2060.1 (February 1998), page 6, "Revaluation." Prices from electronic distribution services for actively traded OTC products give a similar check on traders' prices. ↩

  5. Examination manual, Section 2100.1 (April 2003), page 3. A broker or dealer quote is not necessarily the same as the value used internally, and the provider is to be identified as a counterparty or an independent price service paid for valuations. ↩

  6. Examination manual, Section 2100.1 (February 1998), page 2, "Daily Mark-to-Market Process." A separate control function enters values that automated feeds cannot supply, although traders may take part in determining the values. ↩ ↩2

  7. Examination manual, Section 2100.1 (April 2003), page 4, "Illiquid Instruments and Transactions." Illiquid and customized trades usually have no independent third-party price, and a valuation model is likely to be used. ↩

  8. Examination manual, Section 2060.1 (February 1998), page 6, "Revaluation." Valuation of illiquid products is subjective and leaves more room for errors and fraud, and observed option volatility is one of the weak points. ↩

  9. PRA Rulebook, Trading Book (CRR) Part, Chapter 4, Article 3 "Sources of Market Data" (version of November 2, 2024), paragraphs 1 and 2. These are UK rules and do not apply directly to Japanese financial institutions. ↩ ↩2

  10. Financial Services Agency, "主要行等向けの総合的な監督指針" (Comprehensive Guidelines for Supervision of Major Banks, etc.), III-2-3-3-2(3)①ハ (page 126) (in Japanese). The point of focus concerns the price valuation of marketable credit products when valuations are obtained from brokers or external vendors. ↩

  11. Financial Services Agency, "モデル・リスク管理に関する原則" (Principles for Model Risk Management) (November 12, 2021), Principle 7, 7.1 and 7.2 (page 13) (in Japanese). English translation published by the FSA: FSA. Vendor models and similar items include third-party data and parameters. The examples of management include asking vendors for detailed information and drawing up a contingency plan. ↩

  12. Financial Services Agency, "主要行等向けの総合的な監督指針" (Comprehensive Guidelines for Supervision of Major Banks, etc.), III-2-3-3-2(2)④ and ⑤ (page 125) (in Japanese). The bank chooses the observation period and input data appropriately and verifies the results, and takes into account that VaR may be understated when past experience is insufficient or the data is unreliable. ↩

  13. Examination manual, Section 2110.1 "Capital Adequacy" (April 2000), page 21. The types of stress tests listed are (1) tests that use market data from the time of past events, (2) tests that use "market data" the institution creates for assumed events, and (3) institution-specific analysis. ↩

  14. Examination manual, Section 2040.4 (February 1998), page 1, item 10(a). For feeds from external services such as Reuters, Telerate and Bloomberg and for manual entry, the items to confirm are how incomplete feeds are detected, whether users can override market data, and how data accuracy is maintained. ↩ ↩2

  15. Examination manual, Section 2100.1 (April 2003), page 3, "Liquid Instruments and Transactions." Products in the same portfolio, and a hedging product and the product it hedges, are valued at the same time. Quotes are fresh at the time of valuation. ↩ ↩2

  16. Examination manual, Section 2100.1 (April 2003), page 4, "Illiquid Instruments and Transactions." When prices diverge because the cash and futures markets close at different times, a theoretical price may be used, and it may be appropriate to use a pricing model. ↩

  17. PRA Rulebook, Trading Book (CRR) Part, Article 105 "Requirements for Prudent Valuation" (version of June 1, 2024), paragraph 2(a). The items to include in the valuation policy list the closing-price time. ↩ ↩2

  18. Examination manual, Section 2060.1 (February 1998), page 6, "Revaluation." Differences between traders' inputs and independent market values are resolved and recorded, with the reason and the effect on profit and loss, and large differences are reported to senior managers. ↩ ↩2

  19. Examination manual, Section 2100.1 (April 2003), page 4, "Illiquid Instruments and Transactions." With several quotes, a simple average or an average excluding extreme values is stated in internal policy, and a middle or back-office department checks data that brokers enter directly. ↩ ↩2

  20. Examination manual, Section 2100.1 (April 2003), page 4. In the monthly revaluation, the reasonableness of portfolio prices can be checked by consistency with past values, among other methods. ↩

  21. Bank of Japan Financial System and Bank Examination Department, Center for Advanced Financial Technology, "市場リスク管理態勢の整備" (Building a market risk management framework) (September 2019), page 37, "フロント部門に対する牽制機能の発揮" (Acting as a check on the front office) (in Japanese). A point to note is whether the risk management function obtains data directly and verifies the reasonableness of front office data. ↩

  22. Examination manual, Section 2100.1 (April 2003), page 3. The results of the monthly valuation are documented in enough detail to form a complete audit trail. ↩

  23. Examination manual, Section 2040.4 (February 1998), page 1, item 6. The items to confirm are whether write and edit commands are limited to a few people, whether access rights are reviewed periodically, and whether audit reports on user access exist. ↩

  24. Bank of Japan Financial System and Bank Examination Department, Center for Advanced Financial Technology, "市場リスク管理態勢の整備" (Building a market risk management framework) (September 2019), page 42, points for internal audit (in Japanese). The examples for checking whether the front office is involved in risk measurement are the method of registering data, access rights and a review of the history. ↩

  25. Examination manual, Section 2040.3 (February 1998), page 1, item 6. Where the front office, back office and credit control use independent databases, examiners check the types of reconciliation, its frequency and the tolerance for differences, and more independent databases leave more room for errors. ↩ ↩2

  26. Examination manual, Section 2040.1 (February 1998), page 2. An institution that relies on one database for customer and trade files can have stronger control of data integrity. A single system handling all OTC and listed products is rare. ↩

  27. Examination manual, Section 2040.1 (February 1998), page 2. When data from several sources is combined, checks and reconciliation are performed. If independent databases exist, reconciliation arrangements exist at each point where several data files come together. ↩

  28. Examination manual, Section 2040.1 (April 2003), page 4. Where the back office uses a database or software independent of the front office, differences must be reconciled regularly. ↩

  29. Financial Services Agency, "主要行等向けの総合的な監督指針" (Comprehensive Guidelines for Supervision of Major Banks, etc.), III-2-3-6-1 and III-2-3-6-2 (pages 142–144), item (2)① (in Japanese). The points of focus are whether the bank produces risk data that meets the accuracy and completeness needed for reporting in normal times and in a crisis, and whether most of the data is aggregated automatically. ↩

  30. Financial Services Agency, "モデル・リスク管理に関する原則" (Principles for Model Risk Management) (November 12, 2021), II(1) "対象金融機関" (Covered financial institutions) (page 3) (in Japanese). The principles apply to Japan's G-SIBs, Japan's D-SIBs, and Japanese subsidiaries of FSB-designated G-SIBs that hold the FSA's model approval. ↩