Paper: arXiv 2404.11722
Authors: Yifan He, Abootaleb Shirvani, Barret Shao, Svetlozar Rachev, Frank Fabozzi
Abstract
This research extends the conventional concepts of the bid–ask spread (BAS) and mid-price to include the total market order book bid–ask spread (TMOBBAS) and the global mid-price (GMP). Using high-frequency trading data, we investigate these new constructs, finding that they have heavy tails and significant deviations from normality in the distributions of their log returns, which are confirmed by three different methods. We shift from a static to a dynamic analysis, employing the ARMA(1,1)-GARCH(1,1) model to capture the temporal dependencies in the return time-series, with the normal inverse Gaussian distribution used to capture the heavy tails of the returns. We apply an option pricing model to address the risks associated with the low liquidity indicated by the TMOBBAS and GMP. Additionally, we employ the Rachev ratio to evaluate the risk–return performance at various depths of the limit order book and examine tail risk interdependencies across spread levels. This study provides insights into the dynamics of financial markets, offering tools for trading strategies and systemic risk management.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical models like ARMA-GARCH with Normal Inverse Gaussian distributions and applies option pricing theory, indicating high complexity. It uses high-frequency limit order book data from LOBSTER and analyzes three specific stocks, demonstrating strong empirical rigor.
Research Flowchart
flowchart TD A["Research Goal<br>Define & Extend BAS/Mid-Price to TMOBBAS/GMP"] --> B["Data<br>High-Frequency Trading Data"] B --> C["Distribution Analysis<br>Test Normality & Heavy Tails<br>(3 Methods)"] C --> D["Dynamic Modeling<br>ARMA(1,1)-GARCH(1,1) + NIG Distribution"] D --> E["Risk Assessment<br>Option Pricing & Rachev Ratio"] E --> F["Outcomes<br>Tail Risk Interdependencies<br>Trading & Risk Management Tools"]
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 introduces two new constructs, TMOBBAS and GMP, extending the traditional bid‑ask spread and mid‑price, and applies dynamic modeling and option pricing to them.
Source sentence
In this study, we expanded the concepts of BAS and mid-price to two new constructs: TMOBBAS, a development of the BAS, and GMP, a development of the mid-price.
Assumes 5
- Return series follow an ARMA(1,1)-GARCH(1,1) process with a normal inverse Gaussian (NIG) distribution. model
Source sentence
We adopted an ARMA(1,1)-GARCH(1,1) model with an NIG distribution to model the dynamics of the return series.
- Historical log‑returns are treated as independent and identically distributed (IID). statistical
Source sentence
Within this framework, we treated the dataset as a series of historical log returns, assuming they were IID.
- The double subordination option pricing model assumes the subordinators T(t) and U(t) are inverse Gaussian Lévy processes. model
Source sentence
Consider the case where the subordinators T(t) and U(t) are inverse Gaussian (IG)
- Depth d of the order book is defined as the number of price levels considered beyond the best bid/ask. data
Source sentence
In this study, we refer to this quantity as the “depth” and denote it by the symbol d.
Show 1 more
- The high‑frequency LOB data from LOBSTER are representative of market dynamics for the selected stocks on the chosen day. market
Source sentence
... to select HFT data samples for Amazon (AMZN), Apple (AAPL) and Google (GOOG) from 9:30:00 AM to 4:00:00 PM on June 21, 2012.
Data
- LOBSTER high‑frequency order‑book data order-book public
Source sentence
... to select HFT data samples for Amazon (AMZN), Apple (AAPL) and Google (GOOG) from 9:30:00 AM to 4:00:00 PM on June 21, 2012.
Costs
Not applicable
Finds 4
- Log‑return distributions of both TMOBBAS and GMP exhibit heavy tails and significant deviations from normality. statistical
Source sentence
CIs for each depth. This revealed that the log‑return distributions for both TMOBBAS and GMP have heavy tails, indicative of significant deviations from normality.
- Estimated tail index ξ is significantly greater than zero for all depths, supporting a heavy‑tailed power‑law decay. statistical
Source sentence
From Figure 9, we observe that the estimated values of ξ are significantly greater than zero, and their corresponding 95% CIs for each depth extend strictly to the right of zero.
- Deeper order‑book levels are associated with lower implied volatility in options priced on TMOBBAS. other
Source sentence
As a result, options priced in deeper markets tend to reflect lower implied volatility levels, as market participants perceive less risk and uncertainty in price fluctuations.
- The Rachev ratio of TMOBBAS and GMP varies with depth, providing insight into risk‑return performance across order‑book levels. performance
Source sentence
Figure 18 illustrates the dependence of the Rachev ratio of TMOBBAS and GMP for AMZN, AAPL, and GOOG on the depth.
Checked by others
Supported by
- The $κ$-generalised Distribution for Stock Returns
Compare the findings
This paper: Log‑return distributions of both TMOBBAS and GMP exhibit heavy tails and significant deviations from normality.
That paper: The heavy‑tailed stable distribution fits better than the normal distribution in over 95% of cases.
- The Pitfalls of Continuous Heavy-Tailed Distributions in High-Frequency Data Analysis
Compare the findings
This paper: Log‑return distributions of both TMOBBAS and GMP exhibit heavy tails and significant deviations from normality.
That paper: The normal distribution at 1‑second frequency fails to capture out‑of‑sample outliers on 56 % of days, resulting in numerically zero likelihood.
- Information-Theoretic Approach to Financial Market Modelling
Compare the findings
This paper: Log‑return distributions of both TMOBBAS and GMP exhibit heavy tails and significant deviations from normality.
That paper: The MMM predicts MSCI log‑return density as a Student‑t distribution with four degrees of freedom, which is reported as highly significant.
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 (52)
- Limit Order Books (2016)
- Th 'e orie de la sp 'e culation (1900)
- Residual life time at great age (1974)
- Normal inverse G aussian distributions and stochastic volatility modelling (1997)
- Mastering R for Q uantitative F inance (2015)
- Information-theoretic optimality of observation-driven time series models for continuous responses (2014)
- Generalized autoregressive conditional heteroskedasticity (1986)
- Option valuation using the fast F ourier transform (1998)
and 44 more
- Pricing European options with a log S tudent’s $t$-distribution: A G osset formula (2010)
- A subordinated stochastic process model with fixed variance for speculative prices (1973)
- Price dynamics in a M arkovian limit order market (2013)
- Financial Modelling with Jump Processes (2004)
- Comparison of tail index estimators (1998)
- Dynamic Asset Pricing Theory (2010)
- Modelling Extremal Events: For Insurance and Finance (1997)
- Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation (1982)
- The behavior of stock-market prices (1965)
- Stochastic F inance: An I ntroduction in D iscrete T ime (2004)
- Rank-1/2: a simple way to improve the OLS estimation of tail exponents (2011)
- The Origins of Value: The Financial Innovations that Created Modern Capital Markets (2005)
- Assessing confidence intervals for the tail index by Edgeworth expansions for the Hill estimator (2007)
- A forecast comparison of volatility models: Does anything beat a GARCH (1, 1)? (2005)
- Trading costs and returns for US equities: Estimating effective costs from daily data (2009)
- A simple general approach to inference about the tail of a distribution (1975)
- More light on the kurtosis and related statistics (1972)
- Adaptive robust procedures: A partial review and some suggestions for future applications and theory (1974)
- LOBSTER: Limit Order Book Rreconstruction System (2011)
- Risk premia: Exact solutions vs. log-linear approximations (2013)
- Stochastic behavioral asset-pricing models and the stylized facts (2009)
- On the distribution of stock price differences (1967)
- Portfolio selection (1952)
- Quantitative Risk Management: Concepts, Techniques and Tools— R evised Edition (2015)
- The estimation of the parameters of the stable laws (1975)
- Statistical inference using extreme order statistics (1975)
- The generalized hyperbolic model: Estimation, financial derivatives, and risk measures (1999)
- Stable Paretian Models in Finance (2000)
- Advanced Stochastic Models, Risk Assessment, and Portfolio Optimization: The Ideal Risk, Uncertainty, and Performance Measures (2008)
- An asset return model capturing stylized facts (2011)
- Proof That Properly Anticipated Prices Fluctuate Randomly (1965)
- L'evy Processes and Infinitely Divisible Distributions (1999)
- L' e vy Processes in Finance: Pricing Financial Derivatives (2003)
- Capital asset prices: A theory of market equilibrium under conditions of risk (1964)
- Mutual fund performance (1966)
- An agent-based approach to financial stylized facts (2007)
- Bitcoin Volatility and Intrinsic Time Using Double Subordinated L evy Processes (2023) — link
- Multiple subordinated modeling of asset returns: Implications for option pricing (2021)
- Equity premium puzzle or faulty economic modelling? (2021)
- An Inquiry into the Nature and Causes of the Wealth of Nations (2002)
- Performance measurement in a downside risk framework (1994)
- Analysis of Financial Time Series, 3rd Edition (2010)
- A note on the mean correcting martingale measure for geometric L 'e vy processes (2011)
- Empirical characteristic function estimation and its applications (2003)
0 of 52 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.