In a previous life, I used to lead and manage performance measurement teams. Swelteringly long sun-soaked afternoons in Singapore left me well acquainted with the nuisances of yield curve attribution, as I battled across from fund managers anxious to reflect their month-end numbers as favourably as possible (normally just before bonus season). In those days, performance measurement—and in particular, fixed income attribution—were heavily data intensive, but what always struck me as odd was that for such highly numeric processes, there was so much subjectivity in the returns. A case of art more than science, with as much inference as interpretation. Due to this there was always an underlying desire from the industry for “increased accuracy,” “transparency,” and “real time attribution.” At that time people confidently predicted these were all imminent.
Fast forward a few years and, whilst catching-up with some performance industry veterans, I was struck by how many challenges remained the same. The same issues around timeliness, data accuracy, return divergences, manual processes, and single point in time snapshots are all still proving hugely obstructive in moving the industry forward. At the same time, we also discussed changes that seemed truly transformative, almost futuristic. It feels like performance measurement is finally at an inflection point. Ask anybody with an opinion and there are dozens of trends about to emerge at any one time, but of them all these are the ones I feel are most likely to transform our industry:
- Shift from ABOR to IBOR returns reporting This one literally has performance aficionados choking on their oat milk lattes, but increasingly, fund managers are asking us why they can’t adopt IBOR data for calculating their returns. Aside from the purity of fund accounting data, is there really any reason a manager using a sophisticated order management system with robust data validation processes could not use their IBOR to generate client reporting returns? As long as the models are consistent, documented and the returns externally verified, it would be logical for this data set to be used when calculating returns as it would lend support to some of the other trends highlighted below.
- Multi-factor attribution Traditional attribution models focused on three factors: asset allocation, stock selection, and the market timing of investment decisions. However increasingly sophisticated end clients are pushing for even more explanation of their returns, leading to new models emerging that include factors such as risk, style, and (of course) ESG. Legacy OMS and performance systems are not designed to capture these factors, or apply them, causing a new breed of excel-based attribution being born. Storing these factors for longer term computation (1yr, 3yr, 5yr, SI) have their own challenges for today’s performance measurement teams.
- Dynamic attribution Dynamic attribution, also known as time-weighted attribution, is an approach to performance attribution that measures the impact of investment decisions over shorter intervals of time. Traditional attribution models calculate the performance of an investment portfolio over a fixed period, such as a month or a quarter. This approach assumes that all investment decisions made within that period have an equal impact on performance. However, in reality, the timing of investment decisions can greatly influence returns. Dynamic attribution addresses this limitation by evaluating the contribution of individual investment decisions over shorter time intervals, often on a daily or intraday basis. It captures the effect of timing decisions, such as buying or selling securities, by considering the actual periods during which those decisions were implemented.
- Visualisation Historically, attribution reports were in table formats with portfolio vs. benchmark and a breakdown of three attributes, but increasingly, managers are looking at new ways to present this information. This is partly driven by the inclusion of different metrics, and partly due to the changing demographics of the audience. Firms are utilising new (and existing) tools such as PowerBI, Tableau, Qliksense, outside of their performance reporting systems to engage with their clients.
- Shift toward alternatives Increasing investment in real assets, private assets, and private credit by mainstream asset managers is creating challenges in reporting to end clients, who are used to traditional performance attribution results for vanilla asset classes. For example, a pension fund trustee receiving a detailed breakdown of their equity returns vs. skeletal information on their recent toll road investment, may be inclined to feel there are some pages missing from their report. In reality, a lack of comprehensive, reliable, and timely data for products which may be priced by only synthetics or proxy sources, against non-standardised benchmarks make feeding into “standard” performance models difficult. Combined with the inherent heterogeneity of real assets (for example, wind farms in different parts of the world may have completely unique characteristics), these factors are presenting real problems for performance measurement analysts in providing meaningful attribution for their clients.
- Digital assets Ironically, despite the more “immediate” nature of digital assets, they face many of the same challenges as those of real assets. Lack of consistent data, complex instruments, and the diversity of assets (how many ways can you model a non-fungible token or NFT?) result in the same headaches for performance analysts trying to measure them. However, two other aspects are influencing this asset class more than others: Volatility —digital assets can be notoriously volatile, creating large fluctuations in prices or returns, which can disproportionately affect the overall return of a portfolio; and regulation—the digital asset regulatory landscape is still evolving and differs across jurisdictions, which engenders inconsistency in assessing returns within this asset class.
With the focus on more frequent reporting cycles, the adoption of AI and ML, the mass customisation of reporting, and investors’ growing desire for more transparency and disclosure from investment managers, I wonder if predictive performance attribution may also be lurking on the not-too-distant horizon? In a world that has gone “Moneyball” crazy, where even football (or as my American colleagues refer to it, soccer) is embracing predictive stats such as xG (the expected number of goals scored by a certain player or team), how long is it before clients want to know what the predicted return of a particular environmentally friendly investment is? And then what the corresponding societal impact of it was?
These trends may sound futuristic but some, if not all, are coming. In my opinion, overnight transformation is unlikely. The majority of fund managers still operate on a “live” OMS, whereas performance measurement systems use daily data for month-end reporting. These disparate systems will likely have different pricing hierarchies, classifications, prices, and even cut off times. Day-weighted performance calculations help to provide more alignment to the IBOR screen returns, but the month-end cycle still requires the fund manager to cast their mind back across multiple decision points to recall when they made an investment choice, making it hard to align dynamic reporting aspects to current models.
Additionally, GIPS compliance and historical return calculations don’t easily lend themselves to systemic changes to alternative performance calculation methodologies. Reporting cycles are geared up to meet month-end client deliverables. Re-engineering this end-to-end process will require more than just replacing a calculation engine. It will require different data sets, run times, factor considerations, models, data storage, and access solutions. As my colleague Roger Emms recently put it in his excellent blog on digital assets, we envisage a period of hybrid adoption, where traditional systems and processes are augmented with new ancillary systems and data sources, prior to main-stream transition.
Here at Citisoft we’re discussing these trends with our clients and helping them understand how to begin that transition. Whilst the idea of moving forwards may seem daunting, the reality is that performance measurement is fluid, and will move and adapt over time. At the core of any attribution model is a desire to analyse and determine the returns of the portfolio over a period of time vs a benchmark or universe. The next generation of performance measurement is likely to change again, and more quickly than the last, so the critical aspect in any planning now is to ensure flexibility in the future underlying data sets in order to meet those changing client demands.
Comments