Paper: arXiv 2309.00025

Authors: Aleksy Leeuwenkamp, Wentao Hu

Abstract

We propose a set of dependence measures that are non-linear, local, invariant to a wide range of transformations on the marginals, can show tail and risk asymmetries, are always well-defined, are easy to estimate and can be used on any dataset. We propose a nonparametric estimator and prove its consistency and asymptotic normality. Thereby we significantly improve on existing (extreme) dependence measures used in asset pricing and statistics. To show practical utility, we use these measures on high-frequency stock return data around market distress events such as the 2010 Flash Crash and during the GFC. Contrary to ubiquitously used correlations we find that our measures clearly show tail asymmetry, non-linearity, lack of diversification and endogenous buildup of risks present during these distress events. Additionally, our measures anticipate large (joint) losses during the Flash Crash while also anticipating the bounce back and flagging the subsequent market fragility. Our findings have implications for risk management, portfolio construction and hedging at any frequency.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces advanced mathematical constructs involving copulas and theoretical proofs of consistency and asymptotic normality, indicating high mathematical complexity, and applies these measures to real high-frequency data with practical implications for risk management, showing strong empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>New Dependence Measures"] --> B["Methodology:<br>Non-linear, Local, Invariant Construction"]
  B --> C["Estimator:<br>Nonparametric, Consistent, Asymptotic Normal"]
  C --> D["Data Input:<br>High-Frequency Stock Returns"]
  D --> E["Computational Process:<br>Measures around Flash Crash & GFC"]
  E --> F["Outcome:<br>Quantified Tail Asymmetry & Risk Buildup"]
  F --> G["Outcome:<br>Anticipated Joint Losses & Fragility"]

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 new set of non‑linear, local dependence measures with a nonparametric estimator that is well‑defined, easy to estimate and applicable to any dataset.

Source sentence
We propose a set of dependence measures that are non-linear, local, invariant to a wide range of transformations on the marginals, can show tail and risk asymmetries, are always well-defined, are easy to estimate and can be used on any dataset.

Assumes 5

  • Stationarity within each expanding window model
    Source sentence
    While it does require assuming stationarity within each window we aim to clearly show the time-dependent dependence structure through the changes between each window.
  • Continuous random variables statistical
    Source sentence
    Hence, the estimators are always well-defined if one assumes continuous random variables.
  • No imposed linear dependence, symmetry of copula or margins, radial symmetry, tail (in)dependence, or existence of moments statistical
    Source sentence
    Because the estimators are nonparametric assumptions like linear dependence, symmetry of the copula, symmetry of the margins , radial symmetry of the copula, tail (in)dependence and the existence of moments are not imposed.
  • 5‑minute aggregation mitigates microstructure effects data
    Source sentence
    To reduce the impact of microstructure effects on estimation the 5 minute aggregation is chosen in accordance with the high-frequency econometrics literature
Show 1 more
  • Expanding window chosen for simplicity and computational efficiency model
    Source sentence
    Lastly, for the sake of simplicity and computational efficiency the expanding window method was chosen.

Data

  • Pelger (2020) TAQ dataset · Pelger (2020) / author website market-prices public
    2004-01-02 to 2016-12-30 · 332 S&P 500 stocks (balanced panel)
    Source sentence
    For the empirical results we use the data from Pelger (2020)12. The dataset contains TAQ (Trade-and-Quote) data of S&P 500 stock returns from January 2nd 2004 to December 30th 2016 aggregated at the 5 minute level.

Recognised sources: NYSE TAQ

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The proposed measures anticipate large joint losses during the Flash Crash and also flag the subsequent market fragility. predictive-accuracy
    Source sentence
    Our measures anticipate large (joint) losses during the Flash Crash while also anticipating the bounce back and flagging the subsequent market fragility.
  • Early jumps in the dependence measure occur minutes before the official start of the Flash Crash, providing a timely warning signal. predictive-accuracy
    Source sentence
    the first jump of the measure occurs already at 14:20, the second jump at 14:40 and the third jump at 14:45. The timing of the first jump is 10 minutes before Menkveld and Yueshen (2019) define the start of the Flash Crash
  • Compared with correlations, the new measures reveal tail asymmetry, non‑linearity, lack of diversification and endogenous risk buildup during distress events. statistical
    Source sentence
    Contrary to ubiquitously used correlations we find that our measures clearly show tail asymmetry, non-linearity, lack of diversification and endogenous buildup of risks present during these distress events.
  • During the early phase of the Lehman bankruptcy (Sept 15, 2008) the upper tail dependence exceeds the lower tail, indicating strong positive comovement across stocks. statistical
    Source sentence
    early on September 15th the upper ι and δ exceed the lower ones. As an immediate market reaction to the bankruptcy of Lehman Brothers it shows initially markets reacted positively with large positive and strongly comoving returns across S&P500 stocks.
  • Non‑parametric estimates differ markedly from Gaussian copula based estimates, especially in the tails where the Gaussian model often fails to react. statistical
    Source sentence
    In Figure 3 it is again evident from the difference between the Gaussian based δ and the nonparametric estimates that in the tails returns have a dependence structure which differs strongly from that of a Gaussian copula.

Checked by others

Supported by

  • Robust Bernoulli Mixture Models for Credit Portfolio Risk (2024)
    On the question: Do Gaussian copula models underestimate tail risk compared to non‑parametric or alternative copulas?
    Compare the findings

    This paper: Non‑parametric estimates differ markedly from Gaussian copula based estimates, especially in the tails where the Gaussian model often fails to react.

    That paper: Simulation study shows that Gaussian copula underestimates tail risk compared to Clayton copula, and a mixed siBMM class yields intermediate AVaR values.

    A shows Gaussian copula fails in tails versus non‑parametric estimates; B’s simulation confirms Gaussian underestimates tail risk versus Clayton, with mixed model in‑between.

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.