Performance Attribution: A Look Back at a Timeless Article

Summary: Patrick W. Fowler

Original article by David Spaulding, DPS, CIPM

The 2009 article “Performance Attribution: An Introduction” by David Spaulding continues to provide a comprehensive overview of performance attribution, a key tool in investment performance measurement. The article helps performance and attribution professionals understand and communicate the effectiveness of their strategies.

Introduction to Performance Attribution

Performance attribution is an essential analytical tool for investment managers. Its purpose is to break down the sources of portfolio returns, allowing portfolio managers to determine what decisions contributed to outperformance or underperformance relative to a benchmark. This analysis becomes crucial in client communications and portfolio reviews, offering transparency on the outcomes of investment strategies.

Attribution itself isn’t a new concept; it is something professionals in many fields perform. For example, police officers and fire officials conduct causal investigations after accidents or fires, and sports teams review game strategies post-performance. Similarly, in the investment world, portfolio managers implement strategies and later assess their effectiveness, which is the core of attribution analysis.

Types of Attribution: Relative vs. Absolute

David distinguishes between relative attribution and absolute attribution, also referred to as “contribution.”

• Relative attribution involves reconciling to the excess return, defined as the portfolio’s return minus the benchmark return. This analysis seeks to understand the sources of relative outperformance or underperformance.
• Absolute attribution, on the other hand, focuses on the total return, examining what contributed to the overall performance of the portfolio, rather than its performance relative to a benchmark.

In practice, relative attribution is more commonly used by portfolio managers, especially when their performance is benchmarked against a market index. Absolute attribution may be more useful for highlighting specific investments or strategies that performed exceptionally well or poorly, without the comparative context of a benchmark.

Analyzing Portfolio Effects: Allocation and Selection Decisions

Investment strategies can typically be categorized as top-down or bottom-up approaches:

• Top-down managers begin by analyzing macroeconomic data and political events to predict how various sectors or industries might perform. They make allocation decisions based on their expectations of which sectors will outperform or underperform, by adjusting the portfolio’s sector weightings compared to the benchmark sector weightings accordingly.
• Bottom-up managers, in contrast, focus primarily on selection decisions of individual securities based on their fundamental qualities. Although these managers do not necessarily make explicit allocation decisions, differences in portfolio weightings relative to a benchmark can still impact returns. Therefore, even bottom-up managers should consider the effects of their allocation choices.

Performance attribution models help analyze two primary effects: allocation and selection. Allocation effects measure the impact of over- or under-weighting certain sectors relative to the benchmark. Selection effects assess the effectiveness of picking securities within those sectors.

Attribution Models: Brinson-Hood-Beebower (BHB) and Brinson-Fachler (BF)

Two of the most widely used models for equity attribution are the Brinson-Hood-Beebower (BHB) model and the Brinson-Fachler (BF) model. Both models offer frameworks for attributing excess returns to allocation and selection decisions, though they differ in key aspects.

• BHB Model: Developed in 1986, this model calculates attribution using three effects: allocation, selection, and interaction. The allocation effect is derived by multiplying the difference in portfolio and benchmark weights by the benchmark return for each sector. The selection effect, meanwhile, focuses on the portfolio manager’s skill in picking better-performing securities relative to the benchmark. The interaction effect accounts for situations where a manager’s allocation decision interacts with their selection decision.
• BF Model: The BF model, differs in how it calculates the allocation effect. The selection and interaction effects are identical in both models. Rather than using the sector’s benchmark return in the calculation, the BF model uses the sector’s benchmark return relative to the overall portfolio’s benchmark return. This approach may lead to significant differences in the allocation effect, especially when sectors perform poorly but still outperform the benchmark. The BF model rewards overweighting sectors that outperform the portfolio’s benchmark, regardless of whether their absolute return is positive or negative. The interaction effect accounts for situations where a manager’s allocation decision interacts with their selection decision.

Example Application of BHB and BF Models

Dave provides mathematical examples in the original paper demonstrating how these models operate. For instance, if a portfolio manager over-weights the technology sector and that sector outperforms, the BHB model would show a positive allocation effect. If the sector underperforms, the allocation effect would be negative.

The BF model, however, would consider the performance of the overall benchmark. If the technology sector performed poorly but still outperformed the overall benchmark, the BF model produces a positive allocation effect, rewarding the manager for their relative foresight. This divergence in results highlights the importance of understanding the model used, as different models can yield significantly different interpretations of the same investment strategy.

Challenges and Considerations

• Interaction Effect: One controversial aspect of attribution is the interaction effect, which results from the combination of allocation and selection decisions. Some argue interaction effect should be merged with selection or allocation, arguing that it complicates the analysis without adding meaningful insights. Others prefer it to be reported separately.
• Holdings-Based vs. Transaction-Based Attribution: Holdings-based models use the portfolio’s starting weights for the period, ignoring any transactions (such as buys, sells, or dividends) that occurred during the period. Transaction-based models account for these activities, providing more accurate and reconcilable results. Transaction-based attribution is preferred by most firms for its precision, especially in portfolios with high turnover.

Conclusion

The article emphasizes that performance attribution is not just a reporting tool but a critical part of the investment management process. It helps portfolio managers understand which decisions contributed to their performance and allows them to refine their strategies. Whether using the BHB or BF model or opting for holdings-based versus transaction-based attribution, it’s crucial to choose the model that best aligns with the investment philosophy and strategy of the portfolio manager. Additionally, maintaining clean data and ensuring reconciliation with official records are vital for producing accurate attribution results.

Performance attribution provides a powerful framework for ongoing portfolio analysis, enabling investment professionals to better understand and communicate the value of their decisions to clients and stakeholders.

Have an opinion? Please share it with Patrick Fowler.

GIPS is a registered trademark owned by CFA Institute.

Industry Dates and Conferences

• November 7-8 – Autumn EMEA Meeting of the Performance Measurement Forum – Barcelona, Spain
• November 20 – Fall Meeting of the Asset Owner Roundtable (AORT) – Charleston, SC
• November 21-22 – Fall North American Meeting of the Performance Measurement Forum – Charleston, SC

For information on the 2024 events,
please contact Patrick Fowler at 732-873-5700.

2025 Events Announcement

• April 23 – Spring Asset Owner Roundtable (AORT) – North America
• April 24-25 – Spring Meeting of the Performance Measurement Forum – North America
• May 20 – Women in Performance Measurement in-person meeting – The Heldrich Hotel, New Brunswick, NJ, U.S.A.
• May 21-22- PMAR North America – The Heldrich Hotel, New Brunswick, NJ, U.S.A.
• June 19-20 – Spring EMEA Meeting of the Performance Measurement Forum
• September 17-18 – PMAR Europe – London
• November 6-7 – Fall EMEA Meeting of the Performance Measurement Forum
• December 3 – Fall Asset Owner Roundtable (AORT) – North America
• December 4-5 – Fall Meeting of the Performance Measurement Forum – North America

Institute / Training

Access TSG’s Online Training
Content With One Pass

Our classes cover a wide range of performance measurement concepts, including the Fundamentals (Rates of Return, Attribution, Benchmarking, Risk, and the GIPS standards), and deeper dives into Attribution to include Equity Attribution, Fixed Income Attribution, Multi-Level Attribution, and Multi-Period Attribution. Students will also have access to the newly released Python for the Performance Measurement Professional class. Whether you want to get new members of your performance team trained, or you’re looking to fill in gaps of experienced staff, these classes fit every experience level. This is also a great way to give non-performance professionals a solid overview of performance methodologies and jargon.

The multi-pass gives students unrestricted access to TSG’s entire suite of on-demand training classes and conference recordings available on our online training Institute. This includes more than 80 lessons and over 50 hours of content that’s directly beneficial to investment professionals. To learn more, contact Andrew Tona at (ATona@TSGPerformance.com).

Additional Comment:

With respect to the defined terms and the glossary, we believe this guidance should be consistent with other GIPS guidance, where defined terms that can be found in the glossary are denoted with small capital letters. The use of an initial capital letter for defined terms is confusing.

We hope you found these comments helpful.

That’s a Good Question

I wanted to double check something that I think is implicit in these two bullets:

• Presenting more than one net return stream is acceptable.
• The model fee used to calculate net returns should be the highest fee, specifically stating it should not be the average.

Taken between the two, “the highest fee” seems like it would have to be the highest fee appropriate to the investor, and not the highest fee that anyone in the strategy was paying, right? i.e – if you have a composite with a combination of institutional/subadvisory accounts and direct/private wealth accounts, and have different fee schedules for the different classes, then you could still use a lower institutional schedule with institutional investors, a higher private wealth fee with private wealth investors, and potentially (though I don’t know why you would choose to do this) use the same GIPS Report for both by just showing a Net – Institutional and Net – Private Wealth column on the report, right? And that it just has to be the highest fee for each class used as a model, rather than the highest fee actually assessed to a composite account, even if that’s a private wealth account in a composite that’s also pitched to lower-fee institutional investors?

That’s always been my understanding under GIPS 2020, but I wanted to double check!

Response:

To satisfy SEC requirements, the net returns presented should reflect the deduction of a model fee that is equal to the highest fee charged to the intended audience to whom the advertisement is disseminated. The SEC’s Marketing Rule states if the fee to be charged to the intended audience is anticipated to be higher than the actual fees incurred by accounts represented in composite performance, the adviser must use a model fee that reflects the anticipated fee to be charged in order to not violate the rule’s general prohibitions. So, you are correct in your assessment that the highest fee can be based on the targeted audience, and reporting should be adjusted accordingly.

However, when firms use a model fee to calculate net returns, the GIPS Standards, permit a net return that is equal to or lower than what net returns would have been had actual fees incurred been used to calculate net returns; therefore, depending on what the accounts were charged throughout the track record, the GIPS Standards could be in conflict with the SEC requirements. In certain cases – e.g., when initial performance reflects a discounted fee – the Standards seem to have a lower bar, IMO, because they allow a net return that does not reflect the current fee schedule. IMO, when the SEC Marketing Rule was issued, the GIPS governing body addressed the conflict by advising firms to present more than one net return – one that met the SEC requirements and one that met the GIPS requirements – but I think that was in dispute for a while as some thought presenting more than one net return would be confusing. So, during last week’s conference, when the SEC panelist said that it was okay to present more than one net return, I considered it a key point.

I believe I’ve answered your questions but let me know if I can be of further assistance.

Please submit your questions to Patrick Fowler.

Potpourri

In The News

From David Spaulding, DPS, CIPM

In the August issue, we reviewed a book on Pete Rose, and had a poll regarding whether he should be admitted to the Baseball Hall of Fame.

Pete died on Monday, September 30, at the age of 83. This article, from the Wall Street Journal, suggests it’s time for him to be admitted. https://www.wsj.com/opinion/pete-rose-is-dead-now-reinstate-him-baseball-player-sports-betting-62ea02ff?st=paoEeQ&reflink=desktopwebshare_permalink

When I learned of his death, I thought it would not be appropriate, as it might be deemed an insult to wait for his death. However, perhaps the writer is correct. It’s at least something to think about, yes? Only time will tell.

Article Submissions

The Journal of Performance Measurement^® Is Currently Accepting Article Submissions

The Journal of Performance Measurement is currently accepting article submissions on topics including performance measurement, risk, ESG, AI, and attribution. We are particularly interested in articles that cover practical performance issues and solutions that performance professionals face every day. All articles are subject to a double-blind review process before being approved for publication. White papers will also be considered. For more information and to receive our manuscript guidelines, please contact Douglas Spaulding at DougSpaulding@TSGperformance.com.

Submission deadlines

Winter Issue: January 13, 2025

Spring Issue: March 10, 2025