Paper: arXiv 2506.04656

Authors: Qian Hui, Sidney I. Resnick, Tiandong Wang

Abstract

Accurately identifying the extremal dependence structure in multivariate heavy-tailed data is a fundamental yet challenging task, particularly in financial applications. Following a recently proposed bootstrap-based testing procedure, we apply the methodology to absolute log returns of U.S. S&P 500 and Chinese A-share stocks over a time period well before the U.S. election in 2024. The procedure reveals more isolated clustering of dependent assets in the U.S. economy compared with China which exhibits different characteristics and a more interconnected pattern of extremal dependence. Cross-market analysis identifies strong extremal linkages in sectors such as materials, consumer staples and consumer discretionary, highlighting the effectiveness of the testing procedure for large-scale empirical applications.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10 (capped from 7.0: no out-of-sample test; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)

Why this score: The paper employs advanced extreme value theory and asymptotic statistics, evidenced by formulas for angular measures and bootstrap-based hypothesis tests, indicating high mathematical density. It also demonstrates strong empirical rigor by applying the methodology to real-world financial datasets from the U.S. and China, conducting cross-market sectoral analysis, and using a structured bootstrap procedure with tuning parameters for large-scale classification.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Identify Extremal Dependence<br>in Financial Markets"] --> B["Methodology: Bootstrap Testing<br>on Multivariate Heavy-Tailed Data"]
  B --> C["Data Inputs:<br>US S&P 500 & Chinese A-Share<br>Absolute Log Returns"]
  C --> D{"Computational Process:<br>Bootstrap Inference Testing"}
  D --> E["US Markets Outcome:<br>Isolated Clustering of Assets"]
  D --> F["China Markets Outcome:<br>Interconnected Pattern of Dependence"]
  E & F --> G["Cross-Market Analysis:<br>Strong Linkages in Materials,<br>Consumer Staples, & Discretionary"]
  G --> H["Final Finding:<br>Effective Procedure for<br>Large-Scale Risk Analysis"]

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 systematic, largely automated bootstrap-based classification procedure for extremal dependence structures in financial time series.

Source sentence
Motivated by the framework introduced in Wang and Resnick (2025a), this paper focuses on a systematic and largely automated classification procedure for classifying asymptotic dependence structures that allows categorization of a large number pairs of financial time series into four canonical dependence types

Assumes 5

  • The data consist of an i.i.d. sample satisfying multivariate regular variation. statistical
    Source sentence
    Assuming an iid sample Z1, . . . , Zn with common distribution satisfying (1).
  • Absolute log returns are computed from every‑other‑day prices. data
    Source sentence
    we compute the absolute log returns for each of these 66 stocks using their every-other-day prices, resulting in a dataset of n = 822 observations per stock.
  • Threshold k is chosen by a minimum‑distance method with a formulaic adjustment. model
    Source sentence
    Initially, we chose the bootstrap sample thresholds required by our classification procedures by applying the minimum distance method
  • Absolute log returns are used instead of raw log returns to avoid diminishing data length. data
    Source sentence
    The reasons we chose to use absolute log returns as opposed to log returns include simplicity, reluctance to diminish the data length
Show 1 more
  • Serial dependence is mitigated by using every‑other‑day sampling. market
    Source sentence
    Therefore, to mitigate serial dependence in stock returns, we compute the absolute log returns for each of these 66 stocks using their every-other-day prices

Data

  • Absolute log returns of selected U.S. S&P 500 stocks market-prices unknown
    66 U.S. S&P 500 constituents
    Source sentence
    apply the bootstrap classification method to these absolute returns to assess asymptotic dependence between asset pairs
  • Absolute log returns of selected Chinese A‑share stocks market-prices unknown
    66 Chinese A‑share stocks
    Source sentence
    apply the bootstrap classification method to these absolute returns to assess asymptotic dependence between asset pairs

Recognised sources: Yahoo Finance

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • The U.S. market exhibits stronger clustering of extremally dependent assets, while Chinese stocks show a more interconnected extremal dependence pattern. statistical
    Source sentence
    the procedure reveals more isolated clustering of dependent assets in the U.S. economy compared with China which exhibits different characteristics and a more interconnected pattern of extremal dependence
  • Consumer‑related sectors (materials, consumer discretionary, consumer staples) display the strongest cross‑market extremal dependence. statistical
    Source sentence
    Between China and the U.S., sectors such as materials, consumer discretionary, and consumer staples show extremal dependence
  • Within China, the utilities sector shows stronger extremal dependence on consumer discretionary and staples sectors, suggesting a shift toward consumption‑driven economy. statistical
    Source sentence
    the Chinese utilities sector exhibits somewhat stronger extremal dependence on the consumer discretionary and staples sectors
  • Financials and industrials sectors exhibit weak extremal dependence between the two countries. statistical
    Source sentence
    financials and industrials sectors exhibit weak extremal dependence between the two countries

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 34 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.

Cited References (118)

  1. On estimation and order selection for multivariate extremes via clustering (2025) — link
  2. Empirical properties of asset returns: S tylized facts and statistical issues (2001)
  3. Stock market prices do not follow random walks: E vidence from a simple specification test (1988)
  4. Bootstrap Methods for Testing Asymptotic Dependence in Multivariate Heavy-Tailed Data (2025)
  5. The World is Flat: A Brief History of the Twenty-First Century (2005)
  6. Modeling multivariate extreme value distributions via M arkov trees (2024)
  7. Handbook of Mathematical Functions (1970)
  8. Maxima of P oisson-like variables and related triangular arrays (1997)
and 110 more
  1. Growth of preferential attachment random graphs via continuous-time branching processes (2008)
  2. Efficient network generation under general preferential attachment (2015) — link
  3. Kernel PCA for multivariate extremes (2022)
  4. Estimating uncertainty in respondent-driven sampling using a tree bootstrap method (2016)
  5. Regularly varying multivariate time series (2009)
  6. Universal techniques to analyze preferential attachment trees: Global and local analysis (2007) — link
  7. Twitter event networks and the Superstar model (2015) — link
  8. Consistency of the PLF it estimator for power-law data (2020)
  9. Directed preferential attachment models: Limiting degree distributions and their tails (2020)
  10. Extreme values of independent stochastic processes (1977)
  11. Processes of rth largest (2018)
  12. Functional Laws for Trimmed Levy Processes (2016) — link
  13. Decompositions of Dependence for High-Dimensional Extremes (2016) — link
  14. Transformed-Linear Models for Time Series Extremes (2021) — link
  15. The igraph software package for complex network research (2006)
  16. Living on the multi-dimensional edge: Seeking hidden risks using regular variation (2011) — link
  17. Living on the multi-dimensional edge: Seeking hidden risks using regular variation (2013)
  18. Generation and detection of multivariate regular variation and hidden regular variation (2014) — link
  19. Hidden regular variation under full and strong asymptotic dependence (2016) — link
  20. Tail estimates motivated by extreme value theory (1984)
  21. Clustering multivariate time series using energy distance (2023) — link
  22. On a minimum distance procedure for threshold selection in tail analysis (2018) — link
  23. On a minimum distance procedure for threshold selection in tail analysis (2020) — link
  24. Conditional extreme value models: fallacies and pitfalls (2017) — link
  25. Empirical tail copulas for functional data (2021)
  26. Modelling dependence with copulas (2001)
  27. Estimation of the cure rate for distributions in the Gumbel maximum domain of attraction under insufficient follow-up (2021) — link
  28. Non-parametric cure rate estimation under insufficient follow-up using extremes
  29. Cure models in survival analysis
  30. Towards estimating extremal serial dependence via the bootstrapped extremogram (2012)
  31. Extremal dependence of random scale constructions (2018) — link
  32. Foundations of Modern Analysis (2010)
  33. Risk measures and multivariate extensions of Breiman’s theorem (2012)
  34. The impact of competition on prices with numerous firms (2016) — link
  35. When does the mean excess plot look linear? (2011) — link
  36. Fitting Heavy Tailed Distributions: The poweRlaw Package (2015)
  37. Patterns of Link Reciprocity in Directed Networks (2004) — link
  38. Visualizing dependence in high-dimensional data: An application to S&P 500 constituent data (2017) — link
  39. Ratios of Ordered Points of Point Processes with Regularly Varying Intensity Measures (2017) — link
  40. Ratios of ordered points of point processes with regularly varying intensity measures (2019) — link
  41. Trimmed L'evy Processes and their Extremal Components (2018) — link
  42. Trimmed L 'evy processes and their extremal components (2019) — link
  43. Reciprocity in social networks with capacity constraints (2014) — link
  44. Reciprocity in Social Networks with Capacity Constraints (2015) — link
  45. An assessment of preferential attachment as a mechanism for human sexual network formation (2003) — link
  46. Social networks (communication arising): Sexual contacts and epidemic thresholds (2003)
  47. Degree distributions of growing networks (2001) — link
  48. Organization of growing random networks (2001)
  49. Heavy-Tailed Time Series (2020)
  50. Konect: the K oblenz network collection (2013)
  51. KONECT – The Koblenz Network Collection (2013) — link
  52. Handbook of Network Analysis; The KONECT Project (2018) — link
  53. Handbook of Network Analysis; The KONECT Project (2021)
  54. Facebook friendships network dataset – KONECT (2017) — link
  55. Asymptotic independence and support detection techniques for heavy-tailed multivariate data (2020) — link
  56. SNAP: A General-Purpose Network Analysis and Graph-Mining Library (2016)
  57. SNAP Datasets : Stanford Large Network Dataset Collection (2014)
  58. Modeling End-User Behavior in Data Networks (2011)
  59. Finite sample and asymptotic distributions of a statistic for sufficient follow-up in cure models — link
  60. Hidden Regular Variation: Detection and Estimation (2010)
  61. Hidden regular variation and detection of hidden risks (2011)
  62. Modeling total expenditure on warranty claims (2011) — link
  63. Email networks and the spread of computer viruses (2002) — link
  64. Multivariate tail estimation with application to analysis of CoVar (2013)
  65. Stable Distributions - Models for Heavy Tailed Data (2015)
  66. The second-order version of Karamata’s theorem with applications (2013) — link
  67. Empirical Processes: Theory and Applications (1990)
  68. The Art of Finding Hidden Risks; Hidden Regular Variation in the 21st Century (2024)
  69. Maxima of bivariate triangular arrays: Between asymptotic complete dependence and asymptotic independence lies the conditional extreme value model (2013)
  70. Hidden regular variation of moving average processes with heavy-tailed innovations (2014)
  71. Nonstandard regular variation of in-degree and out-degree in the preferential attachment model (2016) — link
  72. On the use of the method of collective marks in queueing theory (1965)
  73. Queueing systems, volume 2: Computer applications (1976)
  74. Tauberian theory for multivariate regularly varying distributions with application to preferential attachment networks (2015)
  75. Asymptotic normality of degree counts in a preferential attachment model (2016) — link
  76. Asymptotic normality of degree counts in a preferential attachment Model (2016)
  77. Clustering of M arkov chain exceedances (2012)
  78. Clustering of M arkov chain exceedances (2013)
  79. Markov Kernels and the Conditional Extreme Value Model (2012) — link
  80. Transition kernels and the conditional extreme value model (2014) — link
  81. Asymptotics of Markov kernels and the tail chain (2013)
  82. Transition Kernels and the Conditional Extreme Value Model (2014) — link
  83. R: A Language and Environment for Statistical Computing (2019) — link
  84. quantmod: Quantitative Financial Modelling Framework (2019) — link
  85. Nonstandard regular variation of in-degree and out-degree in the preferential attachment model (2014) — link
  86. Surveillance, Epidemiology, and End Results (SEER) Program Research Data, 1975-2016 (2019) — link
  87. CAS: Context-aware Background Application Scheduling in Interactive Mobile Systems (2017)
  88. An Analytical Framework to Characterize the Efficiency and Delay in a Mobile Data Offloading System (2014) — link
  89. Predicting User Roles in Social Networks using Transfer Learning with Feature Transformation (2016)
  90. Three efficient algorithms for implementing the preferential attachment mechanism in Yule-Simon stochastic process (2010)
  91. Random Graphs and Complex Networks (2014)
  92. On the Evolution of User Interaction in Facebook (2009)
  93. Are extreme value estimation methods useful for network data? (2019) — link
  94. Are extreme value estimation methods useful for network data? (2019) — link
  95. Threshold Selection for Multivariate Heavy-Tailed Data (2017) — link
  96. Multivariate Regular Variation of Discrete Mass Functions with Applications to Preferential Attachment Networks (2016) — link
  97. Multivariate Regular Variation of Discrete Mass Functions with Applications to Preferential Attachment Networks (2016)
  98. Asymptotic normality of in- and out-Degree counts in a preferential attachment model (2015) — link
  99. Asymptotic normality of in- and out-degree counts in a preferential attachment model (2017) — link
  100. Consistency of H ill estimators in a linear preferential attachment model (2017)
  101. Common Growth Patterns for Regional Social Networks: a Point Process Approach (2019) — link
  102. Common Growth Patterns for Regional Social Networks: A Point Process Approach (2021) — link
  103. Asymptotic dependence of in- and out-degrees in a preferential attachment model with reciprocity (2022)
  104. Asymptotic dependence of in- and out-degrees in a preferential attachment model with reciprocity (2022) — link
  105. Preferential Attachment with Reciprocity: Properties and Estimation (2023) — link
  106. Random Networks with Heterogeneous Reciprocity (2022) — link
  107. Random Networks with Heterogeneous Reciprocity (2023) — link
  108. Distinguishing forms of asymptotic dependence in heavy tailed data (2025)
  109. Supplementary materials of `` D istinguishing forms of asymptotic dependence in heavy tailed data" (2025)
  110. Extreme value analysis without the largest values: what can be done? (2017)

0 of 118 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.