Open:FactSet Forum

Quantamental approach – Fundamental Investing with new tools for the manager By David Marra

equities
fundamentals
machine-learning
quantitative

(Kevin Close) #1

Traditionally, there have been two investment camps, fundamental and quantitative. In the past, the majority of leading asset managers have primarily had a tent in one camp or the other. With the proliferation of powerful, easy-to-access quantamental tools revolutionizing the investment process, the concept of fundamental vs. quantitative has become anachronistic. A new research paradigm is needed.

Fundamental managers have traditionally relied on fundamental data (e.g., earnings, cash flows, etc.) and analysis to arrive at a qualitative investment story or belief; while quantitative managers have relied on a wide variety of data, often a mix of financial and non-financial, to arrive at a quantitative, statistically-based signal of expected return behavior.

Today, a fundamental manager may use quantitative overlays based on a machine learning analysis of fundamental data like earnings or cash flows. This is an example of what we mean by quantamental, the reshuffling of tools in the fundamental investment research toolbox to discover new insights from a traditional, fundamental data source.

For fundamental managers, quantamental investing can mean ‘having your cake and eating it too’ because it allows managers to increase their edge from a familiar and time tested information source.

Quantitative investors, quantitative hedge funds in particular, have recently embraced ‘alternative data’, at least at the experimental level. While there is no single definition of alternative data, the term generally refers to information that has not been available until recently or information that has not previously been used by investment practitioners, or both. Social media data, data gathered by satellites, weather forecast data, are just three examples of vast and expanding landscape of alternative data.

The hope that quantitative investors have about these forms of data is that they will provide either a new investment insight or a more timely investment signal than fundamental information which is available at the quarterly or monthly frequency.

While there is little doubt that a larger information menu and more timely information should benefit managers with the ability to process it, alternative data sources need to be evaluated like any other data source. All the usual caveats necessarily apply. How is the data gathered? What is the quality and scope of the data gathering process? How error prone is the process and the data itself? Do we have enough data points, enough time-series history, to infer anything with confidence about the value of the signal? Can we say with confidence when the signal works and when it does not work? And so on.

Not surprisingly alternative data, being the new kid on the block, with limited history, and often gathered by young commercial entities, have multiple hurdles to overcome with regards to these various caveats.

Thus, a sound first step fundamental managers can readily consider is to apply new analytic tools like artificial intelligence (AI) and machine learning (ML) to their proven data sources i.e., fundamental data. Such a measured approach should have wide appeal for the fundamental community for obvious reasons. Little disruption to existing investment processes, low incremental investment requirement, large opportunity to learn and to extract incremental insights and edge.

Consider the quantamental research results below from Arialytics, an investment research firm that uses artificial intelligence to extract new insights from fundamental and alternative data sources. This particular research uses a large fundamental dataset to forecast future GICS sub-industry returns.

We clearly see such a relationship exits and there is a large performance spread between the low and high portfolios. The Low portfolio (low signal) has an annualized return of 6.4% and an alpha of negative 1.52% compared with the High portfolio which has a return of 13.4% an alpha of positive 5.50% for the 5 year holding period. A similar spread is apparent over the two year holding period as well.

image

A major implication is that this alpha is achieved from a signal that exploits fundamental data. It illustrates the edge that can be derived from fundamental data using an advanced implementation of the latest quantitative toolbox.

Alternative data holds great promise, in particular for gaining more timely insights into performance compared to fundamental information. The following are some questions to ask when considering any new dataset for your investment process. Is the data coverage relevant to my investable universe? Is there enough historical data to run out-of-sample testing across multiple market regimes? What is the collection process and does it produce reliable data? And, very importantly, what is the incremental value of the the data to our investment strategies over and above what we already have and use?

Change in investment processes rarely occur in a moment but rather evolve over time to take advantage of advances in research methods, data, and technology. Quantamental research offers managers the potential to have their cake and eat it too, to exploit the edge offered by AI and ML technologies while adhering to the data and investment processes that have served fundamental investors so well.

About the Author:

David Marra is the Managing Director and Head of Research at Arialytics, an AI investment research firm in Rye, New York serving the professional investor community.

Correspondence: david.aria@arialytics.com

www.arialytics.com