Economies don’t just “grow” — they learn. They change what they’re good at, one step at a time.

A new paper in Nature Communications by James McNerney and colleagues asks a deceptively simple question:

If economies tend to move into related industries in the short run, what does that imply for their long-term development path?

And the answer quietly rewires how we think about economic complexity, development, and the famous shift from farms to factories to high-tech.

---

What this research is about (in human terms)

Think of an economy like a person building a career.

• If you’re a software engineer, you’re more likely to move into data science than into professional ballet.
• Similarly, if a country already exports textiles, it’s much easier to move into clothing, then technical fabrics, than to suddenly jump into rocket engines.

Economists call this pattern the Principle of Relatedness (PoR):

Economies diversify preferentially into activities that are related to what they already do.

So far, research in economic complexity has split into two big camps:

1. Short-term diversification camp
   • Uses networks of related products or industries.
   • Shows that countries enter new products that sit close to their existing ones in this network.
2. Complexity metrics camp
   • Builds indices like the Economic Complexity Index (ECI) and Fitness.
   • Claims these scores measure how many hidden “capabilities” (skills, institutions, technologies) an economy has.

Until now, these two camps talked to each other mostly in story form, not in equations.
This paper builds a simple dynamical model that connects them.

---

The core idea: a dynamic map of structural change

The authors treat an economy as a vector of abilities across many activities — basically:

“How good is country X at product or activity p right now?”

Then they do two key things:

1. Build a product relatedness network
   • Two products are “close” if many countries tend to export them together.
   • For example, cloth garments and knitwear sit close together; crude oil and precision optics do not.
2. Let abilities diffuse on this network over time
   • If you’re strong in a product, your strength tends to “spill over” into nearby products.
   • Mathematically, they model this with a graph Laplacian — the same type of operator that describes diffusion, heat flow, or consensus in networks.

For the science-oriented crowd:
They decompose this dynamic system into eigenmodes of the relatedness network. Each eigenmode represents a pattern of change with its own characteristic timescale. The slowest modes dominate long-run behavior.

For everyone else:
Imagine you drop a bead on a complex landscape and let it roll. Short-term bumps and wiggles matter, but over a long period, the bead mostly moves along the broadest valleys and ridges. Those broad directions are what these eigenmodes capture.

---

Two coordinates that explain decades of development

Out of this model, two slow, fundamental coordinates pop out:

1. A – Average ability / diversity-like coordinate
   • A is basically a weighted average ability across all products.
   • In practice, it lines up strongly with how diverse a country’s export basket is.
   • Higher A → country exports more types of products competitively.
2. b – Composition coordinate (ECI*)
   • b captures what kind of products dominate the basket, not just how many.
   • One side of b corresponds mostly to agricultural and resource-heavy goods.
   • The other side corresponds mostly to manufactured and more advanced products.
   • This b coordinate turns out to be mathematically and empirically very close to the familiar Economic Complexity Index (ECI).

So instead of a mysterious “complexity score,” the paper gives us a 2D map:

• Horizontal axis (A): How many different things you can produce competitively.
• Vertical axis (b): How far you’ve shifted from basic/raw to manufactured/advanced goods.

Then the authors take 56+ years of UN Comtrade export data (1962–2018) for ~249 countries, compute these two coordinates each year, and watch how economies travel across this map.

---

What they find: economies follow a recognizable path

When you plot all country-year points on this A–b plane, a clear picture emerges:

1. Poor, undiversified economies
   • Low A, low b
   • Few export products, mostly agricultural or raw resources.
   • Think early-stage industrialization.
2. Climbing economies (Region 1 in the paper)
   • Over decades, they tend to move up and to the right:
     • A increases: They diversify into more products.
     • b increases: Their export mix shifts from agriculture → manufacturing.
   • This neatly matches classic Kuznets-style structural change: farms → factories → eventually services (though services are harder to see in trade data).
3. High-income, diversified economies (Region 2)
   • High A, moderately high b
   • These are your OECD-type economies: Japan, Germany, etc.
   • They stay relatively diversified, and their export composition stabilizes.
4. Rich but undiversified economies (Region 3)
   • Low A, relatively high income, often resource-rich or tax havens.
   • Example: oil-heavy economies. They make a lot of money but in few sectors.
   • Over time, their b tends to drift downward; it’s hard to stay specialized only in the “fancy” parts of the map forever.

Crucially:

• Income tracks b more than A.
  • Shifting into high-b products (manufactures and advanced goods) matters more for income than just adding more product categories.
• Diversity tracks A more than b.
  • You can become much more diverse without necessarily becoming much richer, depending on what you diversify into.

So the model’s long-run behavior reconstructs major stylized facts of development from a single short-term rule:

“Move into related stuff.”

---

The surprise: complexity metrics might not measure “complexity”

Now for the twist that matters a lot for policy and for the economic complexity community.

When the authors compare their two coordinates to standard complexity metrics, they find:

• Group 1 (diversity-like metrics)
  • Country Fitness, Production Ability, entropic measures, and even diversity itself all align with A.
• Group 2 (composition metrics)
  • ECI, and related indices align strongly with b — the agriculture→manufacturing direction of long-run change.

In other words:

• These indices cluster into two types:
  • One type mostly tells you how many things a country does.
  • The other type mostly tells you what kind of things they do.

That’s a big deal, because complexity metrics are often sold as:

“Sophisticated tools that infer hidden, latent capabilities from trade data.”

This paper suggests something more grounded and less mystical:

Many complexity metrics mostly summarize long-run structural change —
not the underlying hidden capabilities themselves.

They still capture something useful, but:

• ECI ≈ progress along the long-run development axis implied by the PoR.
• Fitness ≈ strength and breadth of the activity basket.

So ECI and Fitness are not rivals trying to estimate the same magical “true complexity.”
Instead, they measure different aspects of structural change, and each is incomplete without the other.

---

Why this is cool (for both nerds and normal humans)

1. It unifies two fragmented literatures

Researchers working on:

• Diversification and relatedness, and
• Complexity indices

have been in somewhat separate bubbles.

This model shows they’re really describing short-term vs long-term consequences of the same underlying dynamics.

2. It turns a jungle of metrics into a simple 2D map

Instead of juggling multiple complexity metrics and arguing over which is “best,” you can:

• Plot countries in the A–b plane, and
• See both diversity and composition of activities at once.

Policy teams can then ask:

• Are we only diversifying (A up) without moving into higher-value activities (b flat)?
• Are we over-specializing in a high-b niche that might be unstable?

3. It demystifies weird rankings

Why does oil often show up as “low complexity” while pottery or artwork appear “high complexity” in some indices?

Seen through this lens:

• It’s not saying oil extraction is technically simple.
• Instead, the metrics are capturing that countries tend to diversify away from oil as they move along their development path, while high-income countries tend to support specialized artisan products.

Once you treat these metrics as structural-change summaries, not intrinsic difficulty scores, these anomalies make sense.

4. It opens the door to better dynamic models of development

The authors deliberately use a simple model based on local, related moves. But in the real world:

• Countries sometimes make big leaps into unrelated industries via FDI, state-backed bets, or technology transfer.
• Policies, geopolitics, and institutional shocks can bend or break the typical PoR dynamics.

Because this framework is explicitly dynamical, you can now:

• Add extra terms to model long-distance jumps,
• Simulate different policy strategies (e.g., forcing jumps vs following the network), and
• See how that changes the long-run A–b trajectories.

---

What’s next?

Here are some directions this work hints at:

• Beyond country exports
  • Apply similar ideas to networks of industries, occupations, patents, or skills, not just trade.
  • That could help cities, regions, or even universities map their own “structural change” paths.
• Better policy dashboards
  • Instead of obsessing over one index like ECI, policy teams could track:
    • movement in A (are we diversifying?), and
    • movement in b (are we shifting into more income-associated activities?).
• Richer models of transformation
  • Incorporate leaps to unrelated sectors, subsidies, infrastructure shocks, and climate constraints.
  • See which strategies move you to a better part of the A–b plane faster and more safely.

Bottom line:
This paper doesn’t just tune an index. It offers a simple, dynamic story of how economies learn, and it shows that many popular “complexity” tools might really be maps of development’s long-run footprints rather than direct X-rays of hidden capabilities.

---

Check out the cool NewsWade YouTube video about this article!

Article derived from: McNerney, J., Li, Y., Gomez-Lievano, A. et al. Bridging the short-term and long-term dynamics of economic structural change. Nat Commun 16, 10225 (2025). https://doi.org/10.1038/s41467-025-65043-0