5D Library › Books › 5DL-B-040

宏观流动性体制转换、极值尾部相依与杠铃资产配置

藤 Copula、隐马尔可夫模型与主权缓冲动力学

Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation: Vine Copulas, Hidden Markov Models and Sovereign Buffer Dynamics

  • 范雅福 Fan Yafu
  • Version 1.0.0 · 2026-10-30
  • CC-BY-NC-ND-4.0
  • 5DL-B-040
DOI10.5281/zenodo.23828049

摘要

危机来临时,资产会一起下跌吗?储备与安全垫要留多少才够?本书是教学与研究用的数学模型,不是投资建议。它依次讨论:用汉密尔顿的马尔可夫切换模型与隐马尔可夫模型识别波动率体制(前向—后向、鲍姆—韦尔奇与维特比算法均用标准库实现;期望持续期 1/(1−p) 只在无记忆的假设下成立,十年日数据即使知道真实状态,危机体制持续期的 90% 区间也宽达 23 到 48 天;实时滤波概率比事后平滑概率晚两天到一周,收益本身厚尾时信息准则会多选出一个并不存在的体制);极值理论(广义极值与广义帕累托分布、皮坎兹—巴尔克马—德哈恩定理、希尔估计量与平均超额图),条件极值方法的回测(8,000 个预测日中正态模型突破 99% VaR 165 到 183 次,条件 EVT 为 87 次,期望 80 次),以及库皮克检验在 250 天样本中把握度只有约四分之一;Copula 与尾部相依(高斯 Copula 的尾部相依系数为零而非「相关为零」,在 1% 分位条件共同暴跌的概率仍有 0.125,只是趋于零),福布斯—里戈本的「没有传染,只有相互依存」与朗金—索尔尼克的熊市极值相关两方证据,以及藤 Copula:与高斯 Copula 两两秩相关相同时,四项资产同时跌破 1% 分位的概率约高十倍。作者要求证明「90% 短债 + 10% 期权」的杠铃配置随机占优;书中证明它不可能被股票占优,且在期权作废的概率高于股票跌破保底的概率时(本书模型即如此)也不可能一阶占优于股票,并在跳跃扩散模型的四种期权定价下显示它对股票、对 90/10 线性组合都不构成一阶或二阶占优,期权偏贵时反而被 90/10 线性组合二阶占优;CPPI 在乘数 6 时有 6.4% 的年份跌破保底线。主权储备一章核对了格林斯潘—吉多蒂规则与 IMF 的 ARA 指标(固定汇率与浮动汇率权重不同,同一虚构经济体分别为 108% 与 163%),并用两体制流动性模型说明缓冲的边际保护递减。

第七章与《礼记·王制》「三年耕,必有一年之食;九年耕,必有三年之食。以三十年之通,虽有凶旱水溢,民无菜色」、《墨子·七患》所引《周书》与《汉书·食货志》的常平仓并读,用程序显示:若粮食每年损耗超过 1/27(约 3.7%),「九年之蓄」永远达不到。第八章给出体制、尾部、相依与缓冲四步核查表,以及一项「条件 EVT 与条件正态 99% VaR」的预注册回测方案(唯一主要结局为两模型突破率之差,最小关注效应 0.25 个百分点,20 个序列、约 9.2 年锁定数据,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。

关键词:隐马尔可夫模型;马尔可夫切换;鲍姆—韦尔奇算法;维特比算法;体制持续期;极值理论;广义帕累托分布;希尔估计量;期望损失;条件极值方法;库皮克检验;克里斯托弗森检验;Copula;尾部相依;高斯 Copula;金融传染;藤 Copula;随机占优;杠铃配置;外汇储备

Abstract

When a crisis comes, do assets fall together? How large must reserves and safety cushions be? This book presents mathematical models for teaching and research and is not investment advice. It covers the identification of volatility regimes with Hamilton's Markov-switching model and hidden Markov models (forward–backward, Baum–Welch and Viterbi algorithms, all in the standard library; the expected duration 1/(1−p) holds only under the memoryless assumption, and even with the true states known, ten years of daily data leave a 90 % interval of 23 to 48 days for the duration of the stress regime; real-time filtered probabilities lag the smoothed ones by two days to a week, and with fat-tailed returns an information criterion adds a regime that does not exist); extreme value theory (generalized extreme value and generalized Pareto distributions, the Pickands–Balkema–de Haan theorem, the Hill estimator and the mean-excess plot), a backtest of the conditional EVT method (in 8,000 forecast days normal models break the 99 % VaR 165 to 183 times, conditional EVT 87 times, against 80 expected) and the power of Kupiec's test, only about one in four in a 250-day sample; copulas and tail dependence (the Gaussian copula's tail-dependence coefficient is zero, which is not "zero correlation": the conditional probability of a joint crash at the 1 % quantile is still 0.125, it merely tends to zero), the evidence on both sides, Forbes and Rigobon's "no contagion, only interdependence" and Longin and Solnik's extreme correlation in bear markets, and vine copulas: with the same pairwise rank correlations as a Gaussian copula, the probability that four assets all fall below their 1 % quantiles is about ten times higher. The author asked for a proof that a barbell of "90 % short-term bills + 10 % options" stochastically dominates; the book proves that it cannot be dominated by a stock and, when its options expire worthless more often than the stock falls below the floor (as in the book's model), cannot first-order dominate the stock either, and shows in a jump-diffusion model under four option prices that it neither first- nor second-order dominates the stock or a 90/10 linear mix, and that when options are rich it is itself second-order dominated by the 90/10 mix; CPPI with multiplier 6 ends below its floor in 6.4 % of years. The chapter on sovereign reserves checks the Greenspan–Guidotti rule and the IMF's ARA metric (fixed- and floating-rate weights differ, giving 108 % and 163 % for the same fictional economy) and uses a two-regime liquidity model to show the diminishing marginal protection of a buffer.

Chapter 7 reads these beside the "Royal Regulations" of the Book of Rites ("in three years of tilling there must be one year's food; in nine years, three years' food; reckoning over thirty years, even with disasters of drought and flood the people do not look starved"), the Book of Zhou as quoted in the Mozi's "Seven Calamities", and the ever-normal granary of the Book of Han, with a program showing that if grain loses more than 1/27 (about 3.7 %) a year, a "nine-year store" can never be reached. Chapter 8 gives a four-step checklist (regime, tail, dependence, buffer) and a pre-registered backtest of conditional EVT against conditional normal 99 % VaR (single primary outcome the difference in violation rates, smallest effect of interest 0.25 percentage points, 20 series and about 9.2 years of locked data, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated.

Keywords: hidden Markov model; Markov switching; Baum–Welch algorithm; Viterbi algorithm; regime duration; extreme value theory; generalized Pareto distribution; Hill estimator; expected shortfall; conditional EVT; Kupiec test; Christoffersen test; copula; tail dependence; Gaussian copula; financial contagion; vine copula; stochastic dominance; barbell strategy; international reserves

目录 · Contents

  • 序:缓冲、尾部与体制English 中文Preface: Buffers, Tails and Regimes
  • 第一章 隐马尔可夫模型与波动率体制English 中文Chapter 1 Hidden Markov Models and Volatility Regimes
  • 第二章 极值理论:从尾部指数到期望损失English 中文Chapter 2 Extreme Value Theory: From Tail Index to Expected Shortfall
  • 第三章 Copula 与尾部相依English 中文Chapter 3 Copulas and Tail Dependence
  • 第四章 危机中的相关性与藤 CopulaEnglish 中文Chapter 4 Correlation in Crises and Vine Copulas
  • 第五章 杠铃配置与随机占优English 中文Chapter 5 Barbell Allocation and Stochastic Dominance
  • 第六章 主权缓冲与流动性体制English 中文Chapter 6 Sovereign Buffers and Liquidity Regimes
  • 第七章 原典并读:「九年之蓄」与常平English 中文Chapter 7 Reading the Classics: the "Nine-Year Store" and the Ever-Normal Granary
  • 第八章 实践与预注册检验English 中文Chapter 8 Practice and a Pre-Registered Test

How to cite · 引用

Fan, Y. (2026). Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation: Vine Copulas, Hidden Markov Models and Sovereign Buffer Dynamics [宏观流动性体制转换、极值尾部相依与杠铃资产配置] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828049

BibTeX
@book{fan2026tail,
  title = {{Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation}: Vine Copulas, Hidden Markov Models and Sovereign Buffer Dynamics},
  author = {Fan, Yafu},
  year = {2026},
  publisher = {5D Library},
  edition = {Version 1.0.0},
  doi = {10.5281/zenodo.23828049},
  url = {https://doi.org/10.5281/zenodo.23828049},
  note = {Chinese title: 宏观流动性体制转换、极值尾部相依与杠铃资产配置}
}
DOIhttps://doi.org/10.5281/zenodo.23828049
DOI(全部版本 all versions)https://doi.org/10.5281/zenodo.23828048
ORCIDhttps://orcid.org/0009-0004-3519-1203
5D Library ID5DL-B-040
版本 Version1.0.0 · 中英双语版 Bilingual edition
出版日期 Date2026-10-30
网址 Onlinehttps://5dlibrary.com/books/tail-risk-copulas-macro-liquidity-regimes/
SHA-256 · tail-risk-copulas-macro-liquidity-regimes_bilingual_v1.0.0.pdffb3fdf1498b2938e6375d19a93f5e55fb142cafb1ceb5c5115ea68dfda05d6de
SHA-256 · tail-risk-copulas-macro-liquidity-regimes_en_v1.0.0.pdff721cb35c10deb7add7666e7b91ce1fa06e6e5405b5e1a5912aa65b6a8ef43e6
SHA-256 · tail-risk-copulas-macro-liquidity-regimes_zh_v1.0.0.pdf8667b297207c956cc67990b48f136e467d7031c766c29e9d473a974087543cf6
Licence · 许可Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0). https://creativecommons.org/licenses/by-nc-nd/4.0/

Disclosure. The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and therefore has a commercial interest in the subject of this book. The book promotes no course, card deck or service; the Chapter 8 trial is designed to be run and analysed by researchers independent of the author's business and does not recruit from the author's community, students or customers; all the code and the study protocol are published openly for anyone to check. The author develops and maintains quantitative research scripts and provides related open-source research and independent consulting services; while writing this book he may have held routine observation positions in US equity index funds (such as SPY) or related derivatives. All models in this book are offered solely for mathematical and historical-statistical discussion; they constitute no investment advice or trading signal, and the book recommends no specific trade or asset.