One of the reasons clients have purchased FINCAD is to provide an independent valuation to compare against their other sources. This invariably raises the question: "Why don’t the numbers from FINCAD match the numbers from other valuation sources?" Having provided software and services to the finance industry since 1990, and used by more than 3,500 organizations in over 80 countries, FINCAD has incorporated the experience and knowledge gained over the years into all product offerings.
This is one of the most common questions asked of our Client Service Analysts. The challenge in answering this question is that those "other sources" can be quite varied -- counterparties, Bloomberg Finance LP*, Thomson-Reuters, internal systems, or a system offered by your favourite vendor.
Essentially, the user is comparing the results generated by FINCAD to the results generated from these other source(s). Servicing the financial industry for nearly 20 years has provided us with a great deal of experience in helping our clients resolve these discrepancies by providing world class support and completely transparent products. Comprehensive math documentation allows users to answer these questions on their own. However, if they encounter situations where the difference is significant, FINCAD is here to help.
This article investigates common reasons why such discrepancies occur, and contains tips on investigating and resolving valuation differences. In addition to this practical advice, it also considers why you should expect some difference to exist.
Potential Areas of Differences
It is usually very difficult to directly compare valuations done by different companies. There are three key areas that could cause differences:
- Models, methods and proprietary modifications including:
a. Yield Curves
b. Calibration - Terminology (different definitions)
- Market Data (different sources and quotation conventions)
Models, Methods and Proprietary Modifications
The first thing to check is if all parties use the same model. Quite often, one party will be using a different model. If different models are used, some difference is expected. However, sometimes it is not so simple. If the model is the same, the difference could be the result of the internal implementation of the model. This is something that is more difficult to resolve and sometimes impossible if a vendor is unwilling to provide detailed information on the inner workings of their model implementation. Of course this assumes that reviewing the underlying mathematical implementation in detail is something that you want to do. For most people, this is
not a priority and therefore this task would need to be given to someone else with the necessary skills.
FINCAD does not believe in black-box solutions – we provide complete transparency with math reference documents for each of our more than 1600 financial functions. Math References are detailed documents outlining the model, FINCAD’s implementation of it, an explanation of the inputs and outputs to the various FINCAD functions associated with that model, and a detailed example. Also included, is a list of the references used which can be reviewed for further information.
To evaluate the lastest version of FINCAD Analytics Suite, contact a FINCAD Representative
If different systems provide different results while utilizing the same model and are implemented in similar manner, two other key areas where differences could arise are yield curve building and calibration.
Yield Curve
When FINCAD is asked the question of why there is a difference, the yield curve is a likely culprit. More than likely, the difference is due to how the yield curve is built. Building a yield curve is as much an art as it is a science.
For example, there are different ways of bootstrapping (solving for unknown parameters) and interpolating between known points. Of course, you need to have market data and in most cases that data is a combination of cash or deposit rates for the short end, futures or forward rates for the middle and par swap rates or bonds for the long end of the curve. As yield curve building has many components, it also has many potential areas for differences including bootstrapping, interpolation and data. As an example of how many variables are involved in curve building, consider that FINCAD Analytics Suite provides five bootstrapping methods and three interpolation methods.
Given that the yield curve is the basis for most calculations, differences in models or market data are in addition to yield curve differences.
Further, since yield curves are an input to most subsequent calculations, any differences in curve construction can compound as you price a specific instrument. Models for exotic or structured instruments may also rely on other data, thereby leading to other areas of potential differences.
Calibration
Calibration involves finding values of the parameters such that the model is able to reproduce (as close as possible) market observed prices of calibration instruments. These calibrated model parameters are then used as inputs to price the instrument. If a valuation difference arises, the user will have to determine and understand which calibration method is being used, the model used to price the instrument and verify that the market data used in calibration and pricing is consistent. These areas can pose a significant challenge.
Embedded options are a good example where model differences can play a part in the pricing of instruments. The difference may be due to the need for calibration to be done and the subjective nature of calibrating against the necessary market data so that it can be used in pricing the instrument. Many exotic or structured products use models such as Hull-White term structure (either one or multiple factors), Heston and SABR models of stochastic volatility, and LIBOR Market Models (LMM) which rely on calibration. In addition, these models can be further modified providing another reason why "the numbers won’t match". For example, one valuation could be performed using the LMM model, while the valuation you are comparing against may have used the LMM model with stochastic volatility.
Terminology
Another area that can lead to differences is the terminology used. Different labels can mean the same thing while the same label can mean different things. In finance, there are many cases where an output term is understood by different people to mean something completely different. In some cases, the differences in these definitions are slight, but they still make a difference in the generated result, or in the interpretation of the generated result.
For example, the DV01 calculation in Bloomberg Finance LP is done by bumping the raw rates that are used in the curve building process whereas FINCAD calculates the DV01 by bumping the generated spot rates. In this case, the user of FINCAD software can be shown additional ways to obtain a more direct comparison.
Market Data
Another possible reason why the numbers from two different sources are not the same is market data. The market data used by each vendor to perform the calculation could lead to discrepancies because there are many different sources of data, and users may not know which source is being used (data and vendor) for a particular calculation. In addition, the assumptions used in the quote for the data may be different. For example, a rate may be quoted under different conventions (semi-annual vs annual or act/360 vs act/365). This might help explain why the numbers are different.
What Can I Do If There Is a Difference?
Once the user finds out that the numbers generated by a FINCAD solution don’t match with another system, the next question asked is: "What can I do about the difference?" The answer to this question is not easy. Ultimately trying to find why values do not match can be time consuming and tedious. You need to ensure that it makes economic sense to track down the difference. In addition, FINCAD cannot investigate what your other vendor is doing, but since our products are completely transparent we can help you understand exactly how your numbers were generated in FINCAD products.
When our analysts are asked to help a customer understand why the numbers generated using FINCAD don’t match the results from other systems, we investigate the situation by doing the following:
- Set a Tolerance Level
- Review Terminology
- Review Model Assumptions
- Compare Yield Curve Construction
- Check Market Data
- Verify Deal Details
- Set a Tolerance Level
Since identifying and resolving valuation differences can be time consuming and tedious, you need to ensure that it makes economic sense to track down the difference. Basically, it is up to you to determine how much of a difference you are willing to accept. As we have been discussing, it is likely that valuations generated by multiple vendors will not match exactly and, therefore, a tolerance must be set. A spreadsheet can be created where the user sets up an area to compare the values from multiple valuation sources. If the difference is outside the set tolerance level, the user is warned. It is not uncommon for tolerance levels to be in the 5% - 10% range but at the end of the day, it is a company decision. - Review Terminology
There are many cases where the definition of an input or result may be different. If a difference does arise, verify that the definition of all inputs and outputs is the same across vendors. - Review Model Assumptions
The user can verify the model, methods and data being used. If a difference is found in model, re-do the calculation with the same model. If the model is not available, document the reason for the difference. In some cases, this may be enough so long as you understand the reason for the difference and it is expected. This type of explanation may satisfy the people who rely on the results. - Compare Yield Curve Construction
Compare the yield curve construction methods used by each valuation source. As we have been discussing, there are many different ways of generating the yield curve. A good test to verify differences is to output the raw cash flows from the vendors you are comparing. This may not be suitable for all instruments; however, if it is, the user can look at the implied forward rates being calculated. If there are large differences in the calculated rates, it might point to a problem in the curve, and subsequently curve building. This is a good way to locate differences. - Check Market Data
A final area to check is to verify that the market data being used is the same. This does not only mean the data as quoted, but also the assumptions in quoting that data. Make a request to your vendor for the data and assumptions used in the quote. - Verify Deal Details
Verify that the deal details you have entered are correct. It is not uncommon to enter in a transposed date due to date settings in a spreadsheet.
Conclusion
As we have seen, there are many potential areas that could cause the difference in valuation. If you are comparing vendors and the values provided are different, none of them will be able to explain why they are different – but they must be able to explain where their numbers come from. Transparency into the numbers is critical. Vendors need to provide information on what data, assumptions and models were used in arriving at the value. At the end of the day it is up to the user to ask the necessary questions to try and resolve any differences they find.
As a trusted analytics provider, FINCAD has nearly 20 years of experience working with models used in derivatives valuations, and the comprehensive documentation provided in the product gives our users the transparency needed to understand how the models work and our implementation of them. If the documentation does not help resolve the issue, our clients can always call FINCAD for further assistance.
Disclaimer
Your use of the information in this article is at your own risk. The information in this article is provided on an "as is" basis and without any representation, obligation, or warranty from FINCAD of any kind, whether express or implied. We hope that such information will assist you, but it should not be used or relied upon as a substitute for your own independent research.
For more information or a customized demonstration of the software, contact a FINCAD Representative.
* Bloomberg is a trademark of Bloomberg Finance LP. FINCAD is not associated in any way with Bloomberg Finance LP.