A Herding-Based Model of Technological Transfer and Economic Convergence: Evidence from Central and Eastern Europe

This paper proposes a micro-founded herding-based model of technological diffusion to explain economic convergence, demonstrating how peer effects and individual incentives drive Total Factor Productivity growth in Central and Eastern European economies as they adopt technologies from the global frontier.

The Gist

Imagine the global economy as a massive, high-speed train race. Some countries are on the front of the train, zooming ahead with the latest technology (like the US or Germany). Other countries are on cars further back, trying to catch up.

For a long time, economists had a hard time explaining exactly how the slower cars speed up. The old theory was like saying, "They just get lucky and invent new things." But this paper argues that catching up isn't about inventing the wheel; it's about copying the people in front of you.

Here is the paper's story, broken down into simple concepts and analogies:

1. The Core Idea: The "Herding" Effect

The authors propose a new way to look at how countries learn new technology. They call it a "Herding Model."

Think of a flock of birds or a school of fish. When one bird spots food, it dives. The bird next to it sees that and dives too. Soon, the whole flock is diving, not because they all saw the food, but because they saw their neighbors diving.

• In the economy: A country doesn't need to invent a new smartphone or a new factory machine from scratch. It just needs to see that a richer neighbor is using it and making money.
• The Mechanism: The paper suggests that countries "adopt" technology based on two things:
  1. Personal Gain: "Hey, this new machine makes me more money!"
  2. Peer Pressure: "Everyone else is using it; I better get on board or I'll be left behind."

2. The Mathematical Magic (Simplified)

The authors built a math model to describe this "copycat" behavior.

• The Starting Line: Every country starts with a certain level of technology (let's call it $A_0$).
• The Finish Line: There is a "frontier" country (like Germany or the US) that is constantly moving forward, getting faster and smarter.
• The Race: The model calculates how fast a country can close the gap. It turns out, the speed isn't constant. At first, it's slow. Then, as more people adopt the tech, it speeds up (like a snowball rolling downhill). Finally, as they get very close to the leader, it slows down again because it's harder to catch up when you're almost there.

They created a formula that looks like a smooth curve, showing exactly how a country's productivity grows over time as it adopts these "borrowed" technologies.

3. The Real-World Test: Central and Eastern Europe

To see if their theory works, the authors looked at Central and Eastern European (CEE) countries (like Poland, Lithuania, Romania, etc.) and compared them to Germany (and later, the US).

• The Data: They used real numbers about how much money these countries make per hour worked.
• The Result: The model fit the data surprisingly well! It showed that these countries are indeed "herding" toward the German/US technology level.
• The Prediction: Using their formula, they predicted what these economies will look like in 2030 and 2050.
  • Example: Romania started with very low tech but has a high "adoption speed," so the model predicts it will grow very fast.
  • Example: Slovenia started with higher tech but grows a bit slower because it's already closer to the top.

4. Why This Matters (The "So What?")

Most economic models are like black boxes: you put numbers in, and numbers come out, but you don't know why.

• This paper's advantage: It opens the box. It says, "The reason these countries are growing is because of this specific social behavior (herding) and the speed at which they copy each other."
• The Catch: The model is simple. It assumes everyone in the country is the same and ignores things like bad laws, corruption, or wars. It's a "clean" model, not a "messy" real-world one.

The Bottom Line

This paper tells us that economic growth isn't just about magic inventions. It's about social learning. Countries get rich by watching their neighbors, seeing what works, and quickly copying it.

The authors built a tool to measure exactly how fast different countries are doing this "copying," allowing us to predict who will catch up to the rich countries first and who might get left behind. It's a bit like predicting which students in a classroom will finish the test first based on how quickly they start copying the smart kid's answers!

Technical Summary

1. Problem Statement

Standard neoclassical growth models (e.g., Solow) treat technological progress as an exogenous variable, failing to explicitly explain the mechanisms by which technology diffuses between heterogeneous economies. While empirical evidence suggests that convergence in developing economies is driven by the adoption of existing technologies from advanced frontiers rather than frontier innovation alone, standard models lack a micro-founded description of this transfer process. The paper addresses the need for a model that:

• Bridges the gap between macroeconomic growth theory and micro-level agent interactions.
• Explains Total Factor Productivity (TFP) growth as a process of technology adoption (diffusion) rather than exogenous innovation.
• Provides a tractable analytical solution for economic convergence toward a moving technological frontier.

2. Methodology

The authors propose a hybrid framework combining neoclassical growth theory with agent-based modeling (ABM) using a herding mechanism.

A. Theoretical Framework

1. Production Function: The model starts with a standard Cobb-Douglas production function: $Y = AL^{1-\alpha}K^\alpha$, where $A$ represents technological productivity.
2. Herding Mechanism: Instead of assuming innovation drives growth, the model assumes growth is driven by the adoption of existing technologies.
   • Agents (innovators/firms) transition from "non-adopters" to "adopters."
   • This transition is governed by a herding interaction (based on Kirman's model), where the probability of adoption depends on:
     • Individual incentives: A baseline rate of adoption ($\sigma$).
     • Peer effects (Social influence): The proportion of existing adopters ($x$) influencing others, scaled by a parameter $h$.
3. Dynamics:
   • The authors derive a continuous differential equation for the fraction of adopters $x(t)$ using a Fokker-Planck approximation of the Master equation.
   • They introduce a variable change $s(t) = 1 - x(t)$ (representing non-adopters) to solve the system.
   • The growth rate of TFP is defined as $g_A(t) = \gamma s(t)$, where $\gamma$ is the contribution of remaining non-adopters.

B. Analytical Solution

The model yields an explicit analytical solution for TFP, $A(t)$, which describes a nonlinear convergence path.

• Static Frontier: Initially, the model assumes a fixed maximum productivity $A_m$.
• Moving Frontier: To align with reality (where advanced economies also grow), the authors assume the frontier $A_m$ grows exponentially ($A_m(t) = A_{0m}e^{\gamma_m t}$).
• Final Model Equation: The resulting differential equation for $A(t)$ is a logistic growth model with a moving carrying capacity:
  $$\frac{dA}{dt} = \gamma A(t) \left( 1 - \frac{A(t)}{A_{0m}e^{\gamma_m t}} \right)$$
  This equation has an explicit solution (Eq. 19 in the paper) characterized by parameters:
  • $A_0$: Initial productivity.
  • $\gamma$: The speed of diffusion/convergence.
  • $A_{0m}, \gamma_m$: Initial and growth rate of the frontier economy.

C. Empirical Strategy

• Data: OECD Productivity Database (TFP measured in USD per hour worked, PPP converted).
• Sample: 11 Central and Eastern European (CEE) countries (Bulgaria, Croatia, Czechia, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, Slovenia).
• Reference Economies: Two scenarios were tested:
  1. Germany as the technological frontier.
  2. United States as the technological frontier.
• Estimation: The authors used the Levenberg–Marquardt algorithm (a blend of gradient descent and Gauss–Newton methods) to perform non-linear least squares fitting of Eq. (19) to historical data.
• Projections: The fitted parameters were used to project TFP values for 2030 and 2050.

3. Key Contributions

1. Micro-Founded Diffusion: The paper successfully derives a macroscopic TFP dynamic equation from a microscopic herding interaction model, providing a theoretical basis for "catch-up" growth that is often treated as a "black box" in standard models.
2. Analytical Tractability: Unlike many agent-based models that rely solely on simulation, this approach yields a closed-form analytical solution, allowing for clear interpretation of parameters and easy empirical implementation.
3. Moving Frontier Integration: The model explicitly accounts for the fact that the technological frontier is not static but grows over time, offering a more realistic representation of long-term convergence.
4. Parsimonious Parameterization: The model reduces complex convergence dynamics to a small set of interpretable parameters, specifically the diffusion rate $\gamma$, facilitating cross-country comparisons.

4. Results

The model was fitted to CEE data with high precision. Key findings include:

• Convergence Patterns: The model successfully captures the observed convergence of CEE economies toward the technological frontiers (Germany and the US).
• Parameter Rankings:
  • Romania exhibited the highest convergence speed ($\gamma \approx 0.149$ when benchmarked against Germany), followed by Lithuania and Estonia.
  • Croatia showed the slowest convergence speed among the sample ($\gamma \approx 0.086$).
  • The ranking of countries by $\gamma$ generally correlates with their projected TFP levels in 2050.
• Sensitivity to Benchmark: Changing the reference country from Germany to the US altered the absolute values of the parameters and slightly shifted the ranking of countries, highlighting the sensitivity of the model to the choice of the frontier economy.
• Projections:
  • Under the Germany benchmark, projected 2050 TFP values range from ~126 (Croatia) to ~171 (Romania).
  • Under the US benchmark, projected values are higher, ranging from ~126 (Croatia) to ~192 (Romania).
• Limitations Identified: The simplicity of the model (single diffusion channel, homogeneous agents) means it abstracts from institutional factors, policy differences, and structural heterogeneity.

5. Significance and Conclusion

• Theoretical Integration: The paper provides a significant bridge between neoclassical growth theory and complex systems/agent-based modeling. It demonstrates how social interaction mechanisms (herding) can generate aggregate economic regularities (convergence).
• Policy Relevance: By isolating the diffusion speed ($\gamma$) as a key driver, the model suggests that policies enhancing the "social influence" or "peer effects" of technology adoption could accelerate economic catch-up, independent of pure capital accumulation.
• Future Directions: The authors suggest extending the model to include financial components (e.g., private debt, which they note is crucial for CEE growth), heterogeneous agents, and network structures to better reflect real-world innovation systems.

In summary, the paper offers a robust, analytically solvable framework for understanding economic convergence, validating that a herding-based diffusion mechanism can effectively model the technological catch-up of Central and Eastern European economies.