Shock Propagation and Macroeconomic Fluctuations

This paper demonstrates that macroeconomic fluctuations and tail risk arise from the interference of overlapping "productivity waves" propagating through a production network, where the dominant transient eigenvalue of the network's dynamics, rather than the static degree distribution, determines the magnitude of aggregate volatility and the significance of granular shocks.

The Gist

Here is an explanation of the paper "Shock Propagation and Macroeconomic Fluctuations" using simple language, analogies, and metaphors.

The Big Idea: The "Traffic Jam" of the Economy

Imagine the economy as a massive, complex highway system where every car (firm) is connected to others. If one car speeds up or slows down, it affects the cars behind it, which affects the cars behind them, and so on. This is called a production network.

For decades, economists have asked: "If a few big trucks (large firms) have a bad day, does the whole highway gridlock?"

The traditional answer (the "Static" view) was: Yes, absolutely. If a giant truck breaks down, it stops the flow for everyone connected to it. If the network has a few massive hubs (like a giant Amazon or Walmart), a shock to them causes huge economic crashes. This is based on the idea that the economy instantly adjusts to the new reality.

This paper says: "Not so fast."

The authors argue that the real world isn't instant. It takes time for a shock to travel through the network. While the first shock is still rippling through the system, a new shock often arrives. These waves crash into each other, cancel each other out, and create a mess that is actually less volatile than the static models predict.

---

The Core Metaphor: The "Productivity Wave"

Imagine you are standing in a crowded stadium doing "The Wave."

1. The Static View (The Old Way):
   Imagine the stadium is empty, and one person stands up. Then, everyone else instantly stands up in a perfect chain reaction. Then, that person sits down, and everyone instantly sits down.

   • Result: The movement is huge and dramatic. If the person in the "hub" seat stands up, the whole stadium jumps.
   • Economist's Fear: If the big hubs get hit, the whole economy jumps violently.

2. The Dynamic View (The New Paper):
   Now, imagine the stadium is full, and the "Wave" takes 10 seconds to travel from one side to the other.

   • The Twist: Before the first wave even reaches the other side, a second person on the opposite side starts a new wave.
   • The Interference: The first wave (moving right) crashes into the second wave (moving left). They interfere. Sometimes they add up; sometimes they cancel each other out.
   • The Result: The crowd never reaches the "perfect" standing or sitting position. They are just jiggling in the middle. The total movement is smaller and less extreme than if they had waited for the first wave to finish before starting the next one.

The Key Concepts Explained Simply

1. Overlapping Adjustments (The "Time-Averaging")

In the real world, companies don't wait for the economy to settle down before the next news story hits.

• Analogy: Imagine you are trying to paint a wall. You start painting the left side. Before you finish, someone hands you a bucket of paint for the right side. You are now juggling two jobs at once. The final result is a mix of both, not a perfect left side followed by a perfect right side.
• The Paper's Finding: Because shocks arrive before the economy finishes adjusting to the last one, the "noise" of the economy actually smooths out. The waves cancel each other.

2. The "Dominant Eigenvalue" (The Speed of the Wave)

This is a fancy math term for "How fast does the economy forget a shock?"

• Fast Mixing (Small Eigenvalue): The economy is like a fast-flowing river. A rock thrown in creates a splash, but the water clears quickly. New rocks don't interfere much.
• Slow Mixing (Large Eigenvalue): The economy is like a thick, sticky swamp. A rock thrown in creates a ripple that lasts for a long time. If you throw a second rock, it hits the lingering ripple of the first one.
• The Surprise: The paper finds that if the economy is "sticky" (slow to adjust), the big hubs matter less. Why? Because the "clutter" of old, unfinished waves drowns out the new shock. The system is so busy dealing with yesterday's problems that today's big shock gets lost in the noise.

3. Fat Tails vs. The Speed of Convergence

Economists love "Fat Tails." This means a few firms are massively bigger than everyone else.

• Old Belief: Fat tails = Big Risk. If the biggest firm fails, the economy crashes.
• New Belief: Fat tails only matter if the economy is fast enough to process the shock.
  • If the economy is slow (converges slowly), the "Fat Tail" firms lose their power. Their shock gets diluted by the overlapping waves of other firms.
  • If the economy is fast, the Fat Tail firms still cause massive crashes.

The "Back-of-the-Envelope" Conclusion

The authors did some math to guess how much of the real-world economic volatility is actually caused by these "Big Firm" shocks.

• Static Model Prediction: Big firms cause about 30% to 100% of economic swings.
• Dynamic Model Reality: Because of the "wave interference" (overlapping adjustments), big firms might only cause 10% to 15% of the swings.

In plain English: The economy is much more resilient to big shocks than we thought, but only because it's messy and slow to adjust. The chaos of the real world actually acts as a shock absorber.

Why This Matters for You

1. It changes how we predict recessions: We might be overestimating the danger of a single giant company failing because we ignore the fact that the economy is constantly juggling other problems.
2. It highlights the importance of "Speed": The paper suggests that how fast the economy adjusts is just as important as who is big. If the economy gets slower (more sticky), the big players become less dangerous.
3. It's a warning for data: We can't just look at a map of who sells to whom (the network structure). We also need to know how fast the money and goods actually move through that map. Without knowing the speed, we can't predict the risk.

Summary in One Sentence

The economy is less fragile than we thought because it's too busy dealing with yesterday's problems to fully react to today's big shocks, causing the waves of trouble to cancel each other out.

Technical Summary

Here is a detailed technical summary of the paper "Shock Propagation and Macroeconomic Fluctuations" by Antoine Mandel and Vipin P. Veetil (March 2026).

1. Problem Statement

The paper addresses a central question in macroeconomics: What determines the contribution of granular (firm-level) shocks to aggregate volatility?

Existing literature, notably following Acemoglu et al. (2012), posits that aggregate volatility is driven primarily by the cross-sectional heterogeneity of the production network (specifically, fat tails in the degree distribution). These models typically assume a static benchmark: the economy instantaneously jumps from one general equilibrium to another after every shock realization.

The authors argue this assumption ignores a critical dynamic mechanism: overlapping adjustment. In reality, new productivity shocks arrive before the economy has fully adjusted to previous shocks. This creates a "time-averaging" effect where productivity waves from different vintages interfere with one another. The paper asks: How does this intertemporal interference dampen aggregate volatility and tail risk compared to the static equilibrium benchmark?

2. Methodology and Model Structure

A. Economic Environment

• Network Structure: The economy consists of $n$ firms linked as buyers and sellers of intermediate inputs, modeled via a Cobb-Douglas production function with Hicks-neutral productivity shocks ($\epsilon_t$).
• Adjustment Dynamics: Unlike static models, the authors embed the shocks into the decentralized adjustment dynamics of Gualdi and Mandel (2016). Firms update prices and input demands based on local conditions rather than solving for a global equilibrium instantly.
• Shock Process: Productivity shocks arrive as an i.i.d. stream at discrete time steps.
• State Evolution: The log-output vector follows a linear Markov process:
  $$\log q_{t+1} = A \log q_t + \epsilon_t + b$$
  where $A$ is the input-share matrix (with spectral radius $\rho(A) < 1$).

B. The "Wave" Representation

The core innovation is representing aggregate output as a superposition of partially propagated waves.

• Static Benchmark ($y^*_t$): Assumes infinite processing depth before the next shock. Output is $y^*_t = \gamma^\top (I-A)^{-1} \epsilon_t$.
• Dynamic Series ($y_t$): Shocks arrive before full adjustment. Output is a convolution of past shocks:
  $$y_t = \gamma^\top \sum_{j=0}^{\infty} A^j \epsilon_{t-j}$$
  This creates a "productivity wave" that travels through the network layers.

C. Spectral Decomposition

To analyze the dynamics, the authors decompose the network operator $A$ into its eigenmodes.

• Perron Component ($\lambda_1$): Represents the common growth trend (removed via demeaning).
• Dominant Transient Mode ($\lambda_2$): The second largest eigenvalue (in modulus) governs the rate of convergence to equilibrium.
  • If $|\lambda_2|$ is close to 1, the network mixes slowly (shocks persist).
  • If $|\lambda_2|$ is small, the network mixes quickly.

The aggregate output is approximated by the projection onto this dominant transient mode, reducing the system to a scalar AR(1) process driven by the projected shock $\eta_t$.

3. Key Contributions

1. Intertemporal Interference Mechanism: The paper formalizes how overlapping shocks create "interference." When a new shock arrives, it interacts with the unfinished propagation of previous shocks. These waves can partially cancel each other out, dampening the realized aggregate fluctuation.
2. Re-evaluation of Granular Origins: The authors challenge the view that degree heterogeneity (fat tails) is the sole determinant of macro volatility. They show that the macroeconomic significance of fat tails is mediated by the rate of convergence (spectral gap) of the network.
3. Static vs. Dynamic Volatility Bounds: They prove that static aggregate volatility strictly bounds dynamic aggregate volatility (and tail risk) from above. The dynamic economy exhibits lower volatility because the "time-averaging" of shocks reduces the effective amplitude of fluctuations.
4. Finite-Horizon Distributional Analysis: Using quadratic forms of Gaussian vectors, they derive the exact finite-sample distributions of volatility, showing that the static distribution stochastically dominates the dynamic one under certain conditions.

4. Key Results

A. Volatility Attenuation

The ratio of dynamic variance ($\phi$) to static variance ($\phi^*$) is given by:
$$R = \frac{\phi}{\phi^*} = \frac{(1-\lambda_2)^2}{1+\lambda_2}$$

• Implication: As the convergence rate slows ($\lambda_2 \to 1$), the attenuation factor $R \to 0$. Even if the static benchmark predicts massive volatility due to fat tails, a slow-converging network can dampen realized volatility to near zero.
• Tail Risk: The dampening effect is even more pronounced for tail risk. Extreme events require a rare, sustained alignment of waves across vintages; overlapping adjustments make such alignment exponentially less likely.

B. The Role of Degree Heterogeneity

In the static model, fatter tails (lower power-law exponent $\alpha$) increase volatility by increasing cross-sectional exposure.
In the dynamic model, the effect is ambiguous because $\alpha$ influences two opposing channels:

1. Exposure Channel: Fatter tails increase the loading $b(\alpha)$, raising volatility.
2. Mixing Channel: Fatter tails often create bottlenecks (hubs), slowing down convergence (increasing $\lambda_2$). This increases the overlap/interference, which dampens volatility.

• Result: A more heterogeneous network does not necessarily imply higher aggregate volatility in a dynamic setting; the net effect depends on the trade-off between exposure and mixing speed.

C. Empirical Calibration

Using back-of-the-envelope calculations with US production network data:

• If the network converges relatively fast ($\lambda_2 \approx 0.2$), granular shocks might explain ~1/6 of observed volatility.
• If convergence is slower ($\lambda_2 \approx 0.5$), the implied contribution drops to nearly zero.
• This suggests that standard static estimates of granular origins may significantly overstate the true macroeconomic impact of firm-level shocks.

5. Significance and Implications

• Theoretical Shift: The paper moves the debate from "who is large?" (cross-sectional concentration) to "how fast does the network mix?" (spectral properties). It highlights that time is a critical dimension in aggregation, not just a sequencing device.
• Policy and Measurement: It implies that measuring aggregate volatility requires knowledge of the spectral gap of the production network, not just the degree distribution. Current data limitations (lack of firm-level spectral estimates) make precise quantification difficult.
• Granular Shocks: The findings suggest that the "granular origins" of business cycles may be weaker than previously thought. The "noise" of firm-level shocks is heavily filtered by the time it takes for the economy to adjust, preventing the full amplitude of cross-sectional heterogeneity from manifesting in aggregate data.
• Methodological: The use of Neumann series and spectral decomposition to model overlapping adjustments provides a rigorous framework for analyzing non-equilibrium dynamics in network economies.

Conclusion

Mandel and Veetil demonstrate that the macroeconomic impact of micro shocks is not a simple function of network topology but a dynamic interplay between cross-sectional exposure and temporal mixing. The "time-averaging" of shocks through overlapping productivity waves acts as a powerful dampener, meaning that the macroeconomic significance of fat-tailed degree distributions is contingent upon the speed of equilibrium convergence.