Historical Developments in Probability Measures for Asset Pricing: From State Prices to Modern Pricing Kernels

Zhang Chen, Chen Kay · 2026

Read Historical Developments in Probability Measures for Asset Pricing: From State Prices to Modern Pricing Kernels by Zhang Chen, Chen Kay (2026) 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-2605-27658.txt.

Introduction

In their 2026 review, Zhang Chen and Chen Kay trace probability measures in asset pricing from state prices to modern pricing kernels. Pricing theory builds, transforms, or selects measures so prices read as discounted averages, covering risk-neutral valuation, martingale measures, stochastic discount factors, and benchmark pricing. Landmarks run from Bachelier and Arrow-Debreu through Black-Scholes-Merton to Delbaen-Schachermayer and Cochrane's summary.

Catalog record

CREATORZhang Chen, Chen Kay
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62343
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.27658

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Historical Developments in Probability Measures for Asset Pricing: From State Prices to Modern Pricing Kernels Zhang Chen, Chen Kay Abstract This review summarizes the historical development of probability measures in asset pricing, from early mathematical finance and state price theory to risk neutral valuation, martingale measures, forward measures, stochastic discount factors, incomplete market measure selection, benchmark pricing, robust and nonlinear pricing, and modern data driven probability transformations. The central theme is that asset pricing is not merely an exercise in estimating physical probabilities. Instead, pricing theory constructs, transforms, or selects probability measures so that market prices can be represented as expectations after discounting, numeraire normalization, marginal utility weighting, entropy penalization, calibration, or information conditioning. The paper gives special emphasis to the most influential publications that changed the field: Bachelier’s probabilistic model of speculation, Arrow and Debreu’s state contingent claims, Markowitz and the CAPM literature, Black Scholes and Merton’s option pricing theory, Harrison Kreps and Harrison Pliska’s martingale formalization, Delbaen and Schachermayer’s general fundamental theorem, Breeden and Litzenberger’s implied state price densities, Geman, El Karoui, and Rochet’s change of numeraire, Hansen and Jagannathan’s stochastic discount factor restrictions, Cochrane’s SDF synthesis, and the modern empirical and machine learning literature on learned pricing kernels. Text, attention, and sentiment based probability transformations are treated as recent information-adjusted forecasting extensions that sit alongside, rather than above, the established martingale, numeraire, SDF, and incomplete market frameworks. The paper also states common semimartingale market assumptions and collects key mathematical formulas for state prices, stochastic discount factors, Radon Nikodym densities, Girsanov changes of measure, Black Scholes Merton valuation, forward measures, implied risk neutral densities, incomplete market selection rules, coherent risk measures, benchmark pricing, learned SDFs, and information adjusted forecasting. It concludes that probability measures in asset pricing have evolved from static state price ratios into a broad language for valuation, hedging, risk premia, ambiguity, information, and empirical identification. Keywords: probability measure, asset pricing, risk neutral measure, equivalent martingale measure, state price density, stochastic discount factor, change of numeraire, incomplete markets, entropy, information adjusted measure. 1 Introduction The history of asset pricing can be read as a history of probability measures. In elementary forecasting, a probability measure describes the likelihood of future states. In asset pricing, however, the relevant measure is usually not only the physical measure observed from historical frequencies. Prices also encode discounting, risk preferences, marginal utility, constraints, liquidity, market incompleteness, and information. For this reason, modern asset pricing repeatedly replaces, tilts, or augments the physical probability measure, usually denoted by \mathbb{P} , with another measure or valuation object under which prices become conditional expectations. The simplest statement is familiar. If H_{T} is a payoff at maturity T , then in a complete arbitrage free market its time zero price can often be represented as V_{0}=\mathbb{E}^{\mathbb{Q}}\left[\exp\left(-\int_{0}^{T}r_{s}ds\right)H_{T}\right], where \mathbb{Q} is a risk neutral or equivalent martingale measure. Yet this compact equation hides more than a century of conceptual development. It requires a state space, traded claims, a discounting convention, a relation between prices and probability weights, and a theorem ensuring that no arbitrage is equivalent to the existence of a suitable probability measure. It also requires a selection principle when the market is incomplete, since there may be many such measures. The purpose of this review is to summarize the main historical developments in probability measures for asset pricing. The paper covers early mathematical foundations, the Arrow Debreu state price view, the Black Scholes Merton risk neutral transformation, martingale pricing theory, change of numeraire, term structure measures, incomplete market measures, stochastic discount factors, benchmark and real world pricing, robust and nonlinear approaches, and recent data driven reweighting methods. The final part of the review turns to data rich environments in which text, attention, and sentiment enter the information set used for forecasting or for estimating a pricing kernel. These recent approaches are presented as continuations of the broader measure change tradition: they modify the distribution used for prediction or conditioning, while no arbitrage valuation remains governed by martingale measures, numeraires, and stochastic discount factors. Several organizing distinctions are useful. The physical measure \mathbb{P} describes statistical dynamics. The risk neutral measure \mathbb{Q} removes risk premia from discounted asset dynamics when valuation can be done by replication. A forward measure uses a zero coupon bond or other numeraire so that forward prices become martingales. A pricing kernel or stochastic discount factor represents valuation under the physical measure by multiplying payoffs by a random marginal valuation term. In incomplete markets, equivalent martingale measures are no longer unique, so one may choose a measure through hedging error, entropy, variance, utility, calibration, robustness, or benchmark arguments. In data rich settings, text, attention, and sentiment can enter either as conditioning variables in a learned stochastic discount factor or as information adjusted probability tilts for forecasting. 2 Landmark Publications and Historical Emphasis This review emphasizes publications that changed how probability was used in asset pricing, not merely papers that introduced additional notation. A publication is treated as historically central when it did at least one of the following: introduced a new valuation object, established a measure existence or uniqueness theorem, connected pricing measures to economic equilibrium, made a probability measure empirically recoverable from market prices, or changed the empirical estimation of the stochastic discount factor. Table 1 gives a compact guide to the most important milestones. Period Representative landmark publications Measure-theoretic contribution 1900–1960s Bachelier; Arrow; Debreu; Samuelson [ 7 , 4 , 35 , 96 ] Stochastic price dynamics and state contingent claims; normalized state prices anticipate risk neutral probabilities. 1950s–1970s Markowitz; Sharpe; Lintner; Mossin; Ross [ 84 , 99 , 77 , 88 , 94 ] Physical expected returns are linked to covariance and factor risk; pricing kernels later unify these equilibrium restrictions. 1973–1979 Black and Scholes; Merton; Cox, Ross, and Rubinstein [ 14 , 86 , 32 ] Dynamic replication removes physical drift and makes risk neutral probabilities operational in option pricing. 1978–1995 Breeden and Litzenberger; Heston; Duffie, Pan, and Singleton [ 17 , 62 , 40 ] Option prices reveal risk neutral densities; stochastic volatility, jumps, and affine models enrich the pricing measure. 1979–1998 Harrison and Kreps; Harrison and Pliska; Delbaen and Schachermayer [ 59 , 60 , 36 , 37 ] Equivalent martingale measures become the rigorous expression of no arbitrage; completeness corresponds to uniqueness. 1980s–1990s Vasicek; Cox, Ingersoll, and Ross; Heath, Jarrow, and Morton; Geman, El Karoui, and Rochet [ 102 , 31 , 61 , 50 ] Term structure models and change of numeraire show that numeraires generate natural forward measures. 1990s–2000s Hansen and Jagannathan; Lucas; Cochrane; Fama and French [ 56 , 82 , 28 , 44 ] Stochastic discount factors become the common language for equilibrium, empirical factor models, and no arbitrage pricing. 1990s–2010s Follmer and Schweizer; Frittelli; Artzner et al.; Hansen and Sargent [ 46 , 48 , 5 , 57 ] Incompleteness, entropy, utility, coherent risk, and ambiguity transform measure choice into an optimization or robustness problem. 2010s–2026 Gu, Kelly, and Xiu; Chen, Pelger, and Zhu; Kelly et al.; Korsaye, Quaini, and Trojani; Wang, Cheng, and Wang; Avramov and He [ 54 , 27 , 75 , 76 , 104 , 6 ] High dimensional data estimate pricing kernels and learned measure changes; text, news,

Read the complete text