Paper: arXiv 2307.14322
Authors: Zhiyu Cao, Zihan Chen, Prerna Mishra, Hamed Amini, Zachary Feinstein
Abstract
Financial contagion has been widely recognized as a fundamental risk to the financial system. Particularly potent is price-mediated contagion, wherein forced liquidations by firms depress asset prices and propagate financial stress, enabling crises to proliferate across a broad spectrum of seemingly unrelated entities. Price impacts are currently modeled via exogenous inverse demand functions. However, in real-world scenarios, only the initial shocks and the final equilibrium asset prices are typically observable, leaving actual asset liquidations largely obscured. This missing data presents significant limitations to calibrating the existing models. To address these challenges, we introduce a novel dual neural network structure that operates in two sequential stages: the first neural network maps initial shocks to predicted asset liquidations, and the second network utilizes these liquidations to derive resultant equilibrium prices. This data-driven approach can capture both linear and non-linear forms without pre-specifying an analytical structure; furthermore, it functions effectively even in the absence of observable liquidation data. Experiments with simulated datasets demonstrate that our model can accurately predict equilibrium asset prices based solely on initial shocks, while revealing a strong alignment between predicted and true liquidations. Our explainable framework contributes to the understanding and modeling of price-mediated contagion and provides valuable insights for financial authorities to construct effective stress tests and regulatory policies.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 4.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper introduces a novel dual neural network architecture with specific architectural choices and an explainable AI framework, indicating moderate to high math complexity. However, the empirical evidence relies solely on simulated datasets without mention of real-world backtesting, statistical metrics on live data, or code release, placing it lower on empirical rigor.
Research Flowchart
flowchart TD
A["Research Goal:<br>Model Price-Mediated Contagion<br>without Observable Liquidations"] --> B["Methodology:<br>Dual Neural Network Architecture"]
B --> C{"Input:<br>Initial Shocks"}
C --> D["1st Neural Network:<br>Map Shocks to Predicted Liquidations"]
D --> E["2nd Neural Network:<br>Map Liquidations to Equilibrium Prices"]
E --> F["Key Findings:<br>1. Accurate Price Prediction<br>2. Captures Linear/Non-Linear Effects<br>3. Explainable Framework"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a novel dual neural network framework that first predicts asset liquidations from initial shocks and then uses these to predict equilibrium asset prices.
Source sentence
we introduce a novel dual neural network structure that operates in two sequential stages: the first neural network maps initial shocks to predicted asset liquidations, and the second network utilizes these liquidations to derive resultant equilibrium prices.
Assumes 4
- Only initial shocks and final equilibrium asset prices are observable; actual liquidations are unobserved. data
Source sentence
Specifically, in practice, only the initial shocks and the final equilibrium asset prices are typically observable.
- The relationship between predicted liquidations and prices can be captured by a trainable neural network acting as a data‑driven inverse demand function. model
Source sentence
the second neural network directly produces a data-driven equivalent of the inverse demand function even in the absence of observable liquidation data.
- Predicted liquidations capture monotonic trends and can be linearly scaled to recover absolute values. model
Source sentence
To rectify this and enhance the model’s efficacy to predict absolute liquidation values, we introduce a linear scaling procedure.
- Asset price cross‑impacts exist and can be modeled within the dual‑network framework. model
Source sentence
our proposed approach can realistically capture price cross‑impacts during financial contagion.
Data
- Simulated dataset synthetic unknown
Source sentence
Experiments with simulated datasets demonstrate that our model can accurately predict equilibrium asset prices based solely on initial shocks
Costs
Not applicable
Finds 3
- Predicted liquidations are highly correlated with true liquidations (correlation >0.999 for linear, >0.9995 for exponential, >0.9992 for arctan). predictive-accuracy
Source sentence
Dataset Corr Coefficient Linear 0.9999 Exponential 0.9995 Arctangent 0.9992
- Linear regression on predicted prices recovers the true inverse‑demand coefficients (‑0.15 and ‑0.015) with non‑rejection of the null hypothesis. statistical
Source sentence
We fail to reject the null hypothesis that these coefficients are the true values (i.e., the coefficients are −0.15 and −0.015
- The model can capture price cross‑impacts, as evidenced by regression estimates matching the true coefficients for asset‑1 price. pricing-calibration
Source sentence
... Estimated coefficients of −0.149 and −0.015 for the scaled predicted liquidations of assets 1 and 2 respectively
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 16 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.