What is FIA?
FIA (’Fixed Income Attribution’) is a computer program that performs attribution on portfolios of equities, bonds, bills and other more complex fixed income securities. The program can either be run remotely on our server, or installed at your site.
Why FIA?
FIA was written in response to the many difficulties involved in using current-generation attribution systems. Typically, these are expensive, time-consuming to set up, difficult to use and resource-intensive.
FIA is none of these things. It is a radically new approach to the provision of fixed income performance analytics that requires (i) minimal data (basically, two files and some yield curve data); (ii) hours instead of months to set up and run.
If you have worked with other attribution systems, you may be surprised at how easy it is to get results with FIA.
We believe this system changes the nature of fixed income analytics in a major way. By making a range of capabilities that were previously only available to the largest players available to any investment manager, FIA brings greater transparency and better investment process management to the industry as a whole.
What are FIA’s strengths?
FIA leverages your existing performance system
FIA is intended to extend the capabilities of an addition to an existing performance calculation system, not to replace it. The program does not require that you modify your existing back office systems and processes in order to provide an attribution capability.
Instead, FIA uses pre-calculated security level returns as a starting point for attribution. By design, the attribution returns for any given security are then guaranteed to recombine to the given official return.
As a result, there is no need to perform run time-consuming reconciliation between the output of your attribution system and the back office. FIA’s design ensures that the overall returns from an attribution analysis are always identical to the signed-off returns report.
The program works equally well with returns from entirely different sources, allowing you to mix externally supplied security-level benchmark data with locally generated portfolio data from any of your in-house performance measurement systems.
FIA is simple to run
FIA can be called as a stand-alone program, integrated into your code, or called remotely over the Web.
FIA requires minimal data
FIA has extremely modest data requirements. Typically, security-level weights and returns, and some basic information about the securities held, are all you need to run attribution. Sovereign-curve yield curve data is provided for most markets.
FIA has the widest possible range of attribution models
Although FIA is simple to run, no compromises have been made on depth of analytical coverage or security types handled. FIA’s attribution and risk handling capabilities equal or surpass those of any current competing product, while instrument coverage is virtually unlimited. The program offers
- Equity attribution: Brinson-Fachler and Brinson-Hood-Beebower attribution (asset allocation and stock selection), including multi-level asset allocation on any user-defined classification scheme. Interaction return can be rolled into stock selection return if required.
- FX attribution: returns from currency movements , using both naive and Karnosky-Singer attribution models
- Fixed income attribution: returns are decomposed at the security level into
- carry returns (aggregated, running yield and pull-to-par, risk-free and credit carry)
- risk-free curve returns (aggregated, duration/non-parallel shifts, shift/twist/curvature, key rate durations, principal component analysis, CCB decomposition)
- credit returns (credit curve analysis, sector curve and security-specific returns)
- miscellaneous returns (inflation return, cash return, convexity return, rolldown return, price return, paydown return, others)
- Hybrid attribution: a combination of equity attribution for initial asset allocation decisions, with the remainder of returns decomposed into fixed income effects
- Duration allocation attribution: curve returns decomposed into global market direction returns, a sector allocation return, and security-specific returns. This analysis can be combined with other analyses – for instance carry can be analysed using an allocation term, risk-free return using key rate durations, and credit return using duration allocation analysis. DTS (Duration Times Spread) and van Breukelen analyses are also available.
- The system is highly configurable and any combination of the above effects can be chosen. The most commonly used fixed income attribution models (duration, Campisi, Tim Lord, KRD) models are available via a group of templates, allowing such analyses to be set up very quickly.
If for any reason a security type cannot be modelled, you can use the unattributed security type to include the security in attribution reports.
The program’s subportfolio modelling capability allows very straightforward handling of carve-outs and other hierarchical portfolio structures, including synthetic securities.
FIA requires no installation
FIA can be supplied
- as a locally hosted application on an in-house server;
- as a library called through an API, if you wish to integrate the system into an application;
- as a SaaS (Software as a Service) application, via a RESTful API
In the third case, there is no requirement for any other software to be installed on your machine, or for the program to be connected to any of your in-house systems. FIA can be run from an inexpensive netbook if required.
FIA can be run on equities or fixed income securities
With the Brinson allocation setting active, FIA will run Brinson attribution on portfolios of equities, decomposing active sector returns into asset allocation and stock selection returns. Brinson allocation can also be applied to portfolios of fixed income securities, in which case any fixed income returns are treated as stock selection returns. You can therefore use FIA for all your portfolio attribution requirements, including equities and fixed income.
FIA is user-extendable
FIA's OpenPricing interface allows the user to define and use new pricing types. This means that
- Security pricing can be updated without the need for a new release of the core program.
- New security types can quickly be set up and added to the system.
- The user can add new security types while keeping all IP in house. For instance, if you have an internal/proprietary/3rd party pricing library, it can be used by FIA for attribution.
In addition, FIA's OpenRisk interface allows the user to define and use new sources of risk that fall outside those generated by yield curves, such as interest, inflation, and optionality (all of which are supplied as standard).
The availability of OpenPricing and OpenRisk ensures that FIA is future-proof.
What does FIA not do?
FIA does not provide exact security pricing
For the majority of securities, FIA will make slight simplifications in pricing or use other methods to approximate a security’s return.
This approach differs from your back office, which must take exact account of the complications imposed by non-standard coupons or coupon payment dates that fall on a weekend and prices its security holdings accordingly. However, the ability to price at this level of detail requires substantial and costly resources in manpower, data and computing hardware.
The prices generated by FIA may then differ slightly from the official values, and so be unsuitable for cases where 100% accuracy is needed, such as for deal ticket pricing. However, they are easily accurate enough to calculate a security’s return, and hence an attribution report, to a very high degree of accuracy. Typically, the returns generated by FIA are within a fraction of a basis point of the official returns, even when these simplifications are taken into account. The difference in magnitude between these discrepancies and the aggregated returns from each source of risk for each security over the entire calculation period is usually so small that there is no appreciable effect on the overall result.
FIA uses the back office’s returns as a starting point. It then decomposes these quantities into returns generated by each source of risk in the portfolio, which is a function that the back office is not equipped to perform. Any differences between the sum of the returns from each source of risk and the total return are always displayed as a calculation residual, so if any significant discrepancies appear they can be easily identified and examined.
FIA does not calculate trading return
Since FIA uses pre-existing security level returns, it cannot calculate value added by buying or selling securities at levels different to the end-of-day revaluation. Returns generated by trading activities must be calculated using your existing performance system. Note that the system can calculate pricing return, which is return generated by using different end of day revaluations for the same security in portfolio and benchmark.
How can FIA be deployed?
FIA has two core features that make it deployable across a very wide range of environments:
- Written entirely in cross-platform C++
- No database
This means that we can offer the widest possible range of user options:
Local deployment
FIA can be deployed as a third-party application within your local IT infrastructure. The program is most commonly run under Linux and Windows environments, but ports to other operating systems should be straightforward.
The core engine is typically run from a wrapper in one of a number of languages - we offer Java, Python, C# and C++, and others can be set up.
Cloud deployment
FIA’s functionality is also available via a RESTful API, for clients who do not wish to run or deploy FIA on a local machine. We supply example code showing how to call the API from Python.
Each client is supplied with a user-name and a password to access our server. After uploading the raw files for one or more portfolios to our file store, configuration data is supplied either as via a parameter query or as a JSON object.
At the conclusion of the calculation, a set of reports are written to the client's file store, from where they may be downloaded.
Application integration
Depending on the client's requirements, we can make source code or object code available for integration into third party applications.
Post-processing reports
All options may be followed with a range of reporting options, including
- raw CSV and JSON files, suitable for importing into a data warehouse;
- XLS files branded with client logos, suitable for immediate distribution to clients
- interactive reports, allowing the user to use slice-and-dice, drill-down, and crosstabulate absolute and relative performance returns