May 18, 2024
Fixed Income

Thoughts from the Digital Fixed-Income Evolution

The data science wave is here, and bond managers must adapt traditional investment processes to harness new technologies and amplify human talent.

As the fixed-income investing world continues its journey from a largely analog past to a digital future, investment managers must consider how to deploy the growing power of data science and AI across the business. Integrating new and emerging technologies—done effectively—has the potential to drive better and faster insights and decisions while making the most of human talent. We share a few observations from our own digital journey.

Designing Centers of Excellence Can Bolster Investing Capabilities

Data science can enhance a very wide range of investment activities, and we think it’s important for investors to choose the specific areas where they want to build their centers of data-science excellence. Those capabilities need to be developed over time, because the field continues to advance rapidly.

Designed and deployed effectively, centers of excellence can help strengthen investing capabilities, share new insights with clients and better learn from each other. In AB’s business groups, for example, we’re enhancing our distribution, operations, risk and compliance functions. From the strategic investing perspective, we can use data science tools to help create and stress-test asset allocations, incorporating diverse assumptions and client preferences.

Data science and innovation provide the capabilities to architect a next-generation investment platform—a crucial step in generating better investment performance and delivering on clients’ needs.

Data Sets Are Growing … and So Is the Need to Collect and Synthesize Them

In leveraging data science, it’s a good approach to collect as much information as possible, synthesizing it to measure its impact on decision-making. In recent years, the sources and amount of available data have exploded. To process all of it, we think it makes sense to evaluate data using the “four Vs”: volume, velocity, variety and veracity—the last one because it’s paramount that data be true.

A lot of the data sets we’re applying in our investment processes are unstructured. Web scraping, for example, enables investors to find data from news articles and the media and process it to inform decisions. One direct application is sourcing data directly from the web to assess the fiscal health of US states and cities, which offers insight into their creditworthiness. For instance, we’ve scoured websites of local schools and colleges for data on student enrollments and graduation rates that can indicate whether neighborhoods are thriving. That type of unstructured data, which isn’t sourced from a vendor, can yield unique insights to help us better understand risks and evaluate securities—that’s critical in both credit and equity investing.

Combining traditional data sets with unstructured data can lead to better insights. Analysts, for instance, can assess economies on a highly micro level, scrutinizing aspects such as individual shopping malls to map out trends, including consumer purchases or real estate transactions. These aggregate findings can be mapped to prices of real estate and securities as well as portfolio attributes. Compared with the labor-intensive era of Excel spreadsheets, data science is driving more and faster insights that are more accurate and increasingly powerful.

Both Fundamental and Quantitative Research Can Benefit

Data science has a lot of potential to improve both fundamental and quantitative research, which has a lot of appeal to us. Fundamental analysts can use data science tools as they seek better insights through a deeper understanding of the underlying story of a particular credit issue. New dimensions revealed by AI, if interpreted effectively by an analyst’s human expertise and experience, may lead to a better understanding of what’s really driving an issuer.

Data science can bring new insights to quantitative research, too. For instance, historically we needed long periods of data history to validate trends, but AI can achieve this effectively over shorter periods by using a more cross-sectional analysis. Incorporating the right data science techniques can fine-tune existing proprietary approaches using AI, making them more effective. That stands in contrast to asking direct, undifferentiated questions of a tool like ChatGPT in the hopes of yielding lasting insights.

Investors need to be nimble in applying data-science to analysis, because those being analyzed are becoming more aware that they’re under the microscope. Natural language processing (NLP) provides a good example. It’s now being used quite often to analyze corporate earnings statements, and we’re finding that company CEOs are choosing their words and phrasings carefully to project more optimism in investors’ analyses. That’s why we prefer to focus on analyst sentiment metrics, which we’ve found have provided a steadier and stronger signal over time.

AI-Driven Processes Can Lead to Better Leveraging of Human Talent

Data science is fundamentally changing investment processes, which not only makes them faster and more efficient but also enables better leveraging of human talent.

It can be deployed to digitize and integrate the steps in what are often time-consuming, linear investment approaches stretching from research through review, investment committee approval, order building and trading. As we saw it, creating a digital research hub was a vital first step. Because bond markets are highly fragmented, tools to identify pockets of liquidity are critical, too. And a virtual chatbot assistant can instantly integrate research and liquidity insights to recommended trades.

The time saving and efficiency enabled by data science is tangible. Think about deciding which new bond issues are eligible for client portfolios. There are many new issues, with offering memorandums as long as 100 pages. Processes that can read and apply predetermined criteria to documents let analysts focus on AI findings and key passages, making recommendations based on their expertise. That could save 50%–70% of analysts’ time, enabling them to access two or three times the number of new issues.

Automating the more routine parts of the investment process and arming portfolio teams with enhanced quantitative and fundamental research, in our view, fosters better decisions informed by human insight—and gives humans more bandwidth to develop them. That role will always be critical, so human experts still need to own the investment process and be accountable for it.

The focus on human insight applies to traditional fixed-income products, where portfolio managers select favored securities directly. It also applies to systematic products, where portfolio managers create a library of predictive factors that machine learning processes apply to select securities.

Talent Needs Evolve as Humans and AI Tackle New Challenges

One thing that data science won’t change is the need for fixed-income analysts and portfolio managers with strong investment acumen. The investment world is so complex, varied and hard to predict that these human skills and attributes will remain in demand. And investment strategies must still make intuitive, logical sense in terms of how they deploy capital in portfolios. Crucially, AI should remain a toolset to augment human decision-making, not replace it, and human experts must retain ownership for investment decision-making.

But the required skill sets are shifting as data science continues to evolve. For the new generation of fixed-income investors, the ability to code will be table stakes, in much the same way that strong Excel skills were table stakes for the previous generation. The ability to collaborate closely with in-house data scientists and technology experts will also be a critical attribute in helping organizations make the most out of data science and AI.

This effort will include developing new ways to weave AI into investment processes. As we see it, developing AI tools that can enhance scenario analysis will be a big opportunity. There’s more work needed, but we envision that machines can eventually learn to model economic variables such as inflation or even assets such as private debt, enabling human experts to make better-informed and faster investment and allocation decisions.

We also think customization will likely be another major focus in the future. Historically, the task of tailoring client portfolios has been a labor-intensive process. But for investment managers who can develop effective and efficient digitized processes, AI will provide stronger capabilities to deliver bespoke solutions to clients in a cost-effective way. These are just a few of the avenues we see ahead on the journey to harness the growing power of data science to pursue better outcomes.

Leave a Reply

Your email address will not be published. Required fields are marked *