Probability Weighting Meets Heavy Tails: An Econometric Framework for Behavioral Asset Pricing
Akash Deep, Svetlozar T. Rachev, Frank J. Fabozzi · 2025
Read Probability Weighting Meets Heavy Tails: An Econometric Framework for Behavioral Asset Pricing by Akash Deep, Svetlozar T. Rachev, Frank J. Fabozzi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-16563.txt.
Introduction
Akash Deep, Svetlozar Rachev, and Frank Fabozzi build an econometric approach that joins behavioral probability weighting with heavy tailed asset returns. Investors distort probabilities in known behavioral ways, while returns show far heavier tails than Gaussian models allow. Their approach uses Student distributions with a weighting transform that preserves infinite divisibility. Dynamic pricing traits survive that combination by building.
Hundreds of thousands of records across dozens of assets over two decades test the model. Heavy tailed specifications beat Gaussian ones in most cases. Risk underestimation at extreme levels shrinks dramatically versus Gaussian practice. Joint estimation identifies tail and behavioral parameters with established asymptotic theory. Risk managers gain inference that respects both market tails and investor psychology.
Catalog record
| CREATOR | Akash Deep, Svetlozar T. Rachev, Frank J. Fabozzi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-16563
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Probability Weighting Meets Heavy Tails: An Econometric Framework for Behavioral Asset Pricing Akash Deep, Svetlozar T. Rachev, Frank J. Fabozzi Abstract We develop an econometric framework integrating heavy-tailed Student’s t t distributions with behavioral probability weighting while preserving infinite divisibility. Using 432,752 observations across 86 assets (2004–2024), we demonstrate Student’s t t specifications outperform Gaussian models in 88.4% of cases. Bounded probability-weighting transformations preserve mathematical properties required for dynamic pricing. Gaussian models underestimate 99% Value-at-Risk by 19.7% versus 3.2% for our specification. Joint estimation procedures identify tail and behavioral parameters with established asymptotic properties. Results provide robust inference for asset-pricing applications where heavy tails and behavioral distortions coexist. Keywords: Behavioral Finance, Heavy-Tailed Distributions, Student’s t Distribution, Probability Weighting, Econometric Estimation, Infinite Divisibility 1 Introduction Financial return series exhibit heavy tails that systematically reject Gaussian assumptions, as first documented by [ 19 ] and [ 11 ] and confirmed in subsequent econometric studies. While Student’s t and related specifications provide better empirical fit, their integration with behavioral asset-pricing models has been limited because behavioral distortions such as probability weighting often violate conditions required for arbitrage-free pricing and continuous-time econometric analysis. We contribute to this literature by developing an econometric framework that jointly models heavy-tailed returns and behavioral probability weighting while preserving infinite divisibility and compatibility with the Fundamental Theorem of Asset Pricing. We establish sufficient conditions under which bounded probability-weighting transformations of Student’s t processes retain infinite divisibility, and we propose estimation procedures that allow joint identification of tail and distortion parameters, along with results on their asymptotic properties. Empirical analysis on 432,752 daily observations across 86 assets (2004–2024) confirms the econometric relevance of our framework. Gaussian assumptions are universally rejected, and Student’s t distributions dominate Laplace and normal alternatives in 88.4% of cases. Moreover, misspecification has substantial consequences: Gaussian models underestimate 99% Value-at-Risk by 19.7% on average, while our specification reduces errors below 3.2%. The remainder of the paper is organized as follows. Section 2 reviews related econometric literature on heavy-tailed modeling and behavioral adjustments. Section 3 presents the theoretical framework, including conditions under which probability-weighting transformations preserve infinite divisibility. Section 4 describes the estimation and testing procedures and establishes their asymptotic properties. Section 5 reports the empirical results from model comparisons, risk backtests, and robustness checks. Section 6 discusses implications for econometric modeling and asset-pricing applications, and Section 7 concludes. 2 Related Econometric Literature The econometric evidence against Gaussian return assumptions is longstanding. [ 19 ] and [ 11 ] first documented leptokurtosis and excess tail probability in financial returns, findings that continue to be confirmed in high-frequency data [ 2 ] . Parametric alternatives emerged through variance-mixture models [ 23 , 5 ] and Student’s t specifications, while subsequent work formalized their infinite-divisibility properties [ 17 , 13 ] . These contributions provided the theoretical basis for integrating heavy-tailed distributions into continuous-time econometric models. Heavy tails have also been linked to stochastic volatility and jump processes. [ 21 ] jump-diffusion model and [ 15 ] stochastic volatility framework demonstrated how non-Gaussian features arise naturally in dynamic asset-pricing settings. Extensions based on Lévy processes [ 4 , 10 , 9 , 24 ] established rigorous econometric tools for modeling discontinuities and volatility clustering while maintaining arbitrage-free pricing. Recent econometric advances include time-change methods [ 7 ] , realized-volatility approaches [ 2 ] , and jump detection tests [ 1 ] , all of which reinforce the empirical necessity of heavy-tailed specifications. Another strand of literature emphasizes econometric testing and model comparison. [ 6 ] GARCH model, and later EGARCH [ 22 ] , demonstrated how conditional heteroskedasticity generates unconditional heavy tails. [ 14 ] conducted large-scale forecast comparisons, showing that models with heavy-tailed innovations systematically outperform Gaussian specifications. Extreme Value Theory methods [ 20 ] and flexible parametric distributions such as the Skewed Generalized T [ 3 ] have been widely employed for tail risk measurement and regulatory capital estimation. Behavioral finance introduces additional econometric challenges. Prospect theory [ 16 , 26 ] and cumulative probability weighting imply nonlinear distortions of return distributions. While these models capture empirically observed biases, their lack of infinite divisibility often renders them inconsistent with dynamic pricing frameworks. Recent work [ 25 ] has shown how bounded modifications of probability weighting can be embedded in rational dynamic asset pricing models while preserving arbitrage-free conditions. The literature just reviewed establishes three points critical to our contribution: (1) heavy tails are an econometric regularity of financial data; (2) mathematically consistent heavy-tail models exist but are rarely integrated with behavioral distortions; and (3) econometric procedures for jointly estimating tail and behavioral parameters remain underdeveloped. Our paper addresses this gap by introducing a Student’s t -based behavioral framework that preserves infinite divisibility, provides tractable likelihood-based estimation, and improves statistical inference for risk measurement and pricing applications. 3 Theoretical Framework A central requirement of econometric models in dynamic asset pricing is infinite divisibility , which ensures temporal consistency and compatibility with arbitrage-free pricing. If a return distribution is infinitely divisible, the joint distribution of returns across subperiods can be represented consistently for any partition of the time horizon. Definition 1 (Infinite Divisibility) . A random variable X is infinitely divisible if for every positive integer n , there exist independent and identically distributed random variables X_{1},X_{2},\ldots,X_{n} such that X\overset{d}{=}X_{1}+X_{2}+\cdots+X_{n} . While Gaussian and stable laws satisfy infinite divisibility, they fail to capture empirically observed return behavior. Student’s t distributions, by contrast, are both empirically superior and mathematically admissible. Theorem 1 (Student’s t Infinite Divisibility) . All Student’s t distributions with degrees of freedom parameter \nu>0 are infinitely divisible [ 13 ] . The Student’s t distribution with location \mu , scale \sigma , and degrees of freedom \nu has density: f(x;\nu,\mu,\sigma)=\frac{\Gamma\left(\frac{\nu+1}{2}\right)}{\Gamma\left(\frac{\nu}{2}\right)\sqrt{\nu\pi}\sigma}\left(1+\frac{(x-\mu)^{2}}{\nu\sigma^{2}}\right)^{-\frac{\nu+1}{2}} (1) We extend this framework by introducing a behavioral adjustment operator that preserves infinite divisibility. Let F_{t}(x) denote the cumulative distribution function of a Student’s t random variable and w:[0,1]\to[0,1] be a probability weighting function. We define: \mathcal{B}_{w}[x]=x\cdot\left(1+\theta\cdot\tanh\left(\beta\cdot\left(\frac{w(F_{t}(x))}{F_{t}(x)}-1\right)\right)\right) (2) where \theta\in[0,0.3] controls adjustment magnitude and \beta>0 determines sensitivity. Theorem 2 (Preservation of Infinite Divisibility) . Let X be an infinitely divisible Student’s t random variable and \mathcal{B}_{w} be the behavioral adjustment operator defined in equation (2). If the probability weighting function w is Lipschitz continuous and \theta\leq 0.3 , then Y=\mathcal{B}_{w}[X] remains infinitely divisible. This result ensures compatibility with the Fundamental Theorem of Asset Pricing and preserves tractability for likelihood-based estimation. 4 Estimation and Asymptotic Properties 4.1 Model Setup Let R_{t} denote asset returns at time t . We assume R_{t}\sim t_{\nu}(\mu,\sigma), a Student’s t distribution with location parameter \mu , scale parameter \sigma , and degrees of freedom \nu>2 , ensuring finite variance. To capture behavioral distortions, we introduce a probability weighting operator w(p;\alpha) , parameterized by \alpha\in(0,1) , following Prelec (1998). The behavioral-adjusted return distribution is defined by its cumulative distribution function (CDF): F_{bw}(r;\theta)=w(F_{t}(r;\nu,\mu,\sigma);\alpha), where \theta=(\mu,\sigma,\nu,\alpha) . Our objective is to estimate \theta consistently and establish asymptotic normality of the estimators. 4.2 Likelihood-Based Estimation For a sample \{R_{1},\ldots,R_{T}\} , the log-likelihood under the behavioral-adjusted Student’s t distribution is \ell_{T}(\theta)=\sum_{t=1}^{T}\log f_{bw}(R_{t};\theta), where f_{bw}(\cdot;\theta) is the density implied by the weighted CDF transformation. Direct evaluation requires numerical integration; however, under Lipschitz continuity of w(\cdot;\alpha) , the transformation preserves