---
title: Where models fail: causality and self-reference in financial economics
url: https://www.ml-quant.com/papers/arxiv/2311.16570/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2311.16570
source_url: https://arxiv.org/abs/2311.16570
featured: 2024-01-30
citations: 4
topic: Econometrics & Forecasting
---


# Where models fail: causality and self-reference in financial economics

A research paper suggests that the use of unidirectional causation in capital market studies may be flawed, and a better understanding of empirical finance could be achieved by recognizing the limitations of current quantitative finance tools.

- Source: https://arxiv.org/abs/2311.16570
- Identifier: arXiv:2311.16570
- Released: 2023-11-28
- First featured: Quant Letter No. 35 (2024-01-30): https://www.ml-quant.com/issues/2024-01-30/
- Citations (Semantic Scholar): 4
- Published in: Journal of Economic Methodology
- Topic: Econometrics & Forecasting

## Related

- [TKAN: Temporal Kolmogorov-Arnold Networks](https://www.ml-quant.com/papers/ssrn/4825654/): The article presents Temporal Kolomogorov-Arnold Networks (TKANs), a new neural network design that merges the benefits of Recurrent Neural Networks and Long Short-Term Memory for improved multistep time series forecasting.
- [Robust agents learn causal world models](https://www.ml-quant.com/papers/arxiv/2402.10877/): The research suggests that intelligent agents must learn an approximate causal model to generalize to new domains, impacting fields like transfer learning and causal inference.
- [Forecasting Bid–Ask Spreads in Foreign Exchange: Analysis and Machine Learning Prediction](https://www.ml-quant.com/papers/ssrn/4701477/): Machine learning algorithms can effectively predict transaction costs in the foreign exchange market, which can greatly vary based on factors like time of day or week.
- [Deep Generative Modeling for Financial Time Series with Application in VaR: A Comparative Review](https://www.ml-quant.com/papers/arxiv/2401.10370/): The paper introduces new methods for conditional time series generation in financial risk modeling using deep generative methods, and presents a framework to assess the quality of the generated series.
- [The econometrics of happiness: Are We Underestimating the Returns to Education and Income?](https://www.ml-quant.com/papers/arxiv/1807.11835/): Value Rounding Behavior: The study addresses the issue of response scale simplification in surveys, particularly by less educated respondents, and introduces a model to estimate latent subjective wellbeing.
- [A Bayesian theory of market impact](https://www.ml-quant.com/papers/arxiv/2303.08867/): The research explains how large orders, split into smaller ones (meta-orders), affect prices in financial markets, suggesting that the square-root impact law originates from the over-estimation of order flows from meta-orders.