A big benefit of Systems Economics is that it provides a precise, rules-based approach to understanding and assessing economic and financial risk by associating each part of an economy with a specific point (coordinate) in the economy. In doing so, it also identifies the parts that all carry a shared risk.
Understanding risk
Every risk model, every portfolio construction process, and every sector-based market analysis begins with a decision nobody on the investment team actually made: how the underlying data is classified. GICS, SIC, and NAICS sort companies into industries by surface appearance and historical convention — what a company calls itself, or what business it was in when the code was assigned — not by what it actually does. Two companies can carry identical industry codes and perform entirely different economic functions inside the economy. Two companies with nothing obviously in common can perform nearly identical ones. Either way, the classification data looks ordered. Structurally, it isn’t.
That problem doesn’t fix itself with more data or better models layered on top of the same categories — it’s upstream of both. What’s changed since this argument was first made publicly in 2019 is the evidence for the alternative. Royalty Pharma, built from a $60 million start in 1996 on a functionally organized view of pharmaceutical royalty rights, has grown to roughly $26 billion in value, holding partial royalty rights on seven of the top thirty drugs sold in the United States. Cibus, applying the same functional-systems approach to plant biology through its Rapid Trait Development System, was named Fast Company’s second-most innovative agriculture company in 2024. Neither result depends on the industry an observer would file either company under. Both depend on what the underlying activity functionally does.
The alternative approach — call it a functional coordinate system — locates any business, job, product, or resource by what it does and where it sits in a recurring order of economic work, rather than by an inherited industry label. Every ongoing economic system, regardless of size, moves through the same four-phase cycle: it procures inputs, produces its output, exchanges that output, and manages the proceeds in a way that funds the next cycle. That cycle gives every entity in the economy — a single job or a multinational corporation — a position that can be computed from its own facts rather than assigned by an analyst’s judgment call. Function, in other words, becomes a third reference coordinate alongside geography and time: a way to locate an economic entity that stays stable regardless of which industry-code era or reclassification cycle happens to be in effect.
The clearest test of this approach so far sits inside index construction. A stratified-weighting methodology, built on this kind of functional classification, was tested against standard capitalization-weighted and equal-weighted benchmarks and published in two peer-reviewed studies — “An Index Methodology for Diversifying Business Risk” (The Journal of Index Investing, 2017) and a companion practitioner piece (Practical Applications, 2018). The finding: indexes constructed on functional groupings, rather than conventional sector classification, delivered meaningfully better risk-adjusted returns. The mechanism is straightforward once the classification problem is visible — a portfolio diversified across GICS sectors can still be concentrated in a single functional exposure if several of its sector-labeled holdings are, functionally, doing the same thing in the economy. Diversifying by function catches a category of risk that diversifying by sector code structurally cannot see.
None of this requires abandoning existing classification systems outright — GICS, SIC, and NAICS aren’t going anywhere, and they weren’t built for this particular job. But every risk model built entirely on top of them inherits a blind spot at the point where two functionally identical entities happen to sit in different sectors, or two functionally unrelated ones happen to share one. That’s worth a second look, independent of what any particular model concludes from the data it’s already using. The classification layer underneath the model is where the question belongs.
