The Network Origins of Large Economic Downturns

Daron Acemoğlu, Asuman Ozdaglar, Alireza Tahbaz-Salehi

TL;DR

This paper demonstrates that large economic crashes don't necessarily require massive "black swan" events or heavy-tailed shocks. Instead, the structure of an economy’s input-output network can amplify small, routine micro-shocks into major downturns, effectively turning "thin-tailed" inputs into "heavy-tailed" aggregate outcomes.

What Problem Does This Paper Address?

Economists have long struggled with the "small shocks, large cycles" puzzle: how do relatively minor, localized disruptions (like a strike in one industry or a localized crop failure) occasionally snowball into national catastrophes like the Great Depression?

Traditional macroeconomic models often rely on two explanations:

  1. Aggregate Shocks: Something hits everyone at once (e.g., a global pandemic or a massive change in interest rates).
  2. Heavy-Tailed Shocks: Individual firms are prone to extreme, catastrophic failures that are large enough to move the needle on their own.

Acemoglu, Ozdaglar, and Tahbaz-Salehi propose a third way. They argue that the network structure of the economy—specifically how sectors rely on each other for inputs—is the primary culprit. If the network is "unbalanced" (meaning a few sectors act as critical hubs for everyone else), even standard, normally distributed shocks can propagate and aggregate into deep recessions.

Methods

The authors use a theoretical mathematical model to analyze a static, multi-sector economy.

  • Production Model: They employ a Cobb-Douglas production framework where each sector uses labor and intermediate goods from other sectors.
  • Network Representation: The economy is modeled as a directed, weighted graph where nodes are sectors and edges represent the flow of goods.
  • Influence Vectors: They utilize the concept of "Bonacich centrality" (an "influence vector") to measure how much a shock to one sector affects the entire system.
  • Large Deviation Theory: To study "tail risks" (the probability of extreme events), they analyze the limiting behavior of aggregate output as the number of sectors ($n$) grows to infinity. This allows them to see how quickly the probability of a crash vanishes in different network configurations.

Datasets

Not discussed in the paper. This is a purely theoretical paper focused on mathematical proofs and stylized examples. It does not use empirical datasets or training data.

Key Insights and Ideas

  • Balanced vs. Unbalanced Economies: In a "balanced" economy—where no single sector is a dominant supplier—the Law of Large Numbers holds. Shocks average out, and the chance of a large downturn becomes infinitesimal as the economy grows. However, in "unbalanced" economies (like a "star" network where one sector supplies many others), this averaging fails.
  • Volatility is Not Tail Risk: A crucial insight is that aggregate volatility (standard deviation of GDP) is not a sufficient metric for predicting crashes. Two economies might have the same day-to-day volatility, but if one has a more "hub-heavy" network, it will experience deep recessions much more frequently.
  • The Shape of Shocks Matters: The network interacts differently with different types of shocks.
  • If shocks are normally distributed (thin-tailed), the overall risk is determined by the "Euclidean norm" of the network's influence.
  • If shocks have exponential tails, the risk is dominated by the single most central sector in the economy.
  • Network as a "Tail-Heavy" Generator: The most striking finding is that a specific network structure can make an economy subject to "thin-tailed" shocks (like a normal bell curve) behave exactly like an economy subject to "heavy-tailed" shocks (like a Pareto distribution). The network itself creates the "black swan" effect.

Key Results

  • Proposition 2 (The Irrelevance Result): In balanced economies, the probability of a large downturn decays exponentially fast. Essentially, in a perfectly diversified network, large crashes are impossible without a massive aggregate shock.
  • Proposition 3 & 4 (The Interaction): The authors provide formulas showing that in unbalanced networks, the probability of a crash decays much more slowly (sub-exponentially).
  • Example 3: They demonstrate a case where an economy with a non-trivial network and thin-tailed shocks exhibits the same frequency of extreme downturns as a collection of non-interacting sectors hit by heavy-tailed shocks. This proves that network propagation is a substitute for shock extremity.

Limitations and Future Work

  • Static Model: The model is a snapshot in time. It doesn't account for how firms might adapt over time, such as finding new suppliers if a "hub" sector fails (dynamic rewiring of the network).
  • Cobb-Douglas Assumptions: The results rely on specific production functions that assume constant returns to scale and specific substitution elasticities. Real-world supply chains might have "bottlenecks" that are even more fragile than this model suggests.
  • Lack of Empirical Testing: While the theory is robust, the paper does not map these findings to real-world input-output tables (like those from the Bureau of Economic Analysis) to rank the "crash-readiness" of specific national economies. This is a clear path for future empirical research.

Summary updated

Abstract

This paper shows that large economic downturns may result from the propagation of microeconomic shocks over the input-output linkages across different firms or sectors within the economy. Building on the framework of More specifically, we show that an economy with non-trivial intersectoral input-output linkages that is subject to thin-tailed productivity shocks may exhibit deep recessions as frequently as economies that are subject to heavy-tailed shocks. Moreover, we show that in the presence of input-output linkages, aggregate volatility is not necessarily a sufficient statistic for the likelihood of large downturns. Rather, depending on the shape of the distribution of the idiosyncratic shocks, different features of the economy's input-output network may be of first-order importance. Finally, our results establish that the effects of the economy's input-output structure and the nature of the idiosyncratic firm-level shocks on aggregate output are not separable, in the sense that the likelihood of large economic downturns is determined by the interplay between the two.
BibTeX
@misc{acemoglu2013network,
  author = {Daron Acemoğlu and Asuman Ozdaglar and Alireza Tahbaz-Salehi},
  title = {{The Network Origins of Large Economic Downturns}},
  year = {2013},
  doi = {10.2139/ssrn.2294153},
  howpublished = {SSRN},
}

Loading related content...

Figure