Performance attribution is a key technique in identifying positive and negative contributors to a portfolio’s investment performance. So, it would be reasonable to assume that it is widely used to explain to management and investors where portfolio managers added and lost value, wouldn’t it?
Well, this isn’t necessarily the case. It depends on the type of portfolio. Performance attribution is widely used for equity portfolios, but not for bond portfolios. And many of those who are using fixed income attribution are dissatisfied to some extent with the results it produces.
This isn’t a recent development; this has been the situation for many years now. Why?
There are two commonly held perceptions about fixed income attribution 1) it’s hard to understand and 2) it’s hard to implement. These perceptions are often cited to explain why fixed income attribution is a less successful technique than equity attribution.
It’s only hard to understand because we make it so
With equity portfolios we are happy, in most cases, to view the performance return as the combined result of changes in the values of the portfolio’s assets, which are grouped into asset classes. And we are happy to ‘explain’ the portfolio’s performance in terms of whether we were overweight or underweight in each asset class relative to the benchmark, and in terms whether each asset class performed better in the portfolio or in the benchmark. All of which is intuitive and easy to understand.
We could take a similar asset class valuation approach with fixed income attribution, but it doesn’t work as well. Attribution methods are mathematical models of the real world. In the real world, portfolio managers use investment processes; attribution methods are mathematical representations of investment processes. These attribution models are approximations to the real world. There is usually a part of the portfolio’s return that cannot be explained by the model because it is an approximation. This unexplained amount is called the residual. The problem with taking an asset class valuation approach for fixed income attribution is that we sometimes end up with a residual that is so big it makes the analysis worthless.
This is because fixed income investment processes are more complex than equity processes, and so fixed income attribution models have to be more complex than equity attribution models to avoid large residuals. Fixed income attribution has to go further than identifying by how much asset values change over time; it tries to identify the reasons for the changes. This is far from being a simple exercise and so fixed income attribution is less intuitive than equity attribution, and so harder to understand.
But we don’t have to make it so hard to understand. We don’t have to use terms like twist, butterfly, carry, and roll-down. We don’t have to explain the portfolio’s performance in complex terms just because fixed income attribution uses complex mathematical models. Instead of talking about roll-down effect, let’s talk about how much of the change in the prices of the portfolio’s bonds resulted from the fact that as bonds get closer to the date on which they will be repaid, their market price gets closer to the amount that will be repaid. Over a three-month period, the bonds in a portfolio move three months closer to their maturity date, and so their market prices will move closer to the redemption amounts. And this will contribute to the portfolio’s performance over the three-month period. Now that’s easier to understand than talking about a roll-down effect.
We should not make performance analysis opaque for anyone who isn’t a performance analyst or portfolio manager. Our job is to make it transparent. When it comes to providing more transparency around fixed income attribution, we need to try harder.
It’s only hard to implement because we don’t go about it the right way
It is true that the implementation of fixed income attribution has a very mixed track record. Projects to implement equity attribution rarely fail, but this is not true of fixed income attribution projects. As a result, there is a widely held perception that the implementation of fixed income attribution is extremely difficult, so difficult that some performance teams are reluctant to attempt it.
The difficulty arises from two main factors.
First, there is a lack of standardized fixed income attribution models. Different vendor systems employ different models, and systems developed in-house are usually based on bespoke models. This means that differences arise between the attribution effects calculated by front-office systems used by portfolio managers, vs. those calculated by the middle-office systems used by performance teams. These differences can result in portfolio managers rejecting the fixed income attribution effects provided by performance teams as “incorrect,” and in the suspension of the implementation project until the differences can be explained.
Second, the complexity of fixed income attribution models makes them very sensitive to data integrity issues; different attribution models have different levels of complexity and therefore different levels of data sensitivity. The level of integrity in the data input to fixed income attribution systems has to be aligned with the level of complexity of the specific attribution model that is being implemented. Failure to ensure this before the project begins will result in large residuals, which in turn will result in project delays as efforts are diverted to bringing source data integrity up to the required level.
If is when these factors are not taken into consideration during implementation planning for fixed income attribution that the risk of project failure is highest.
It’s time for a fresh look at fixed income attribution
This post is the introduction to a series of posts that will take a fresh look at fixed income attribution. It’s about time we did this. Too many people in our industry feel intimidated by fixed income attribution, and by the challenge of implementing it. Yes it is hard to understand and yes it is difficult to implement; but we can make it a lot easier to understand and a lot easier to implement if we shift our perspective by taking a fresh look at it.
BI-SAM’s B-One is designed to centralize performance management processes across a single platform, including data management, performance measurement and reporting, attribution and risk analysis and GIPS composite management and compliance. Our attribution capabilities include support for multiple fixed income methodologies, including Single Duration analysis or Key Rate Duration analysis.