Discrete Price Model (DPM)

Updated 3 April 2026

Discrete Price Model is a framework that incorporates price discreteness from tick-size granularity and integer-valued increments into mathematical market analysis.
It applies across financial microstructure, discrete-choice econometrics, and derivative pricing to capture the impact of finite pricing grids on market dynamics.
Empirical and algorithmic approaches, including Poisson-based modeling and branch-and-bound algorithms, offer precise estimation and optimization within DPM frameworks.

A Discrete Price Model (DPM) is any modeling framework that explicitly incorporates price discreteness—arising from tick-size granularity, finite price grids, or integer-valued increments—into the mathematical representation, estimation, or optimization of prices in markets or decision systems. DPMs are central in modern financial microstructure analysis, discrete-choice econometrics, option pricing under pathwise uncertainty, and competitive equilibrium theory under practical denomination constraints.

1. Foundations and Model Classes

Discrete Price Models arise in heterogeneous domains, each emphasizing distinct aspects of discreteness.

Financial Microstructure and Continuous-Time DPMs

In high-frequency finance, Shephard and Yang introduce a DPM for asset prices as a pure-jump, piecewise-constant semimartingale with integer-valued jumps and finite activity (Shephard et al., 2014). The construction rests on a marked Poisson random measure on (y, x, s)∈ℤ∖{0}×[0,1]×ℝ with a Lévy measure v and support in a squashed "trawl" set and its time-shifts. The process is

$P_t = V_0 + L(A_t) + L(B_t),$

where $A_t$ encodes fleeting moves and $B_t$ permanent ones; price reversals are governed by a trawl function $d(s)$ , controlling how quickly moves are canceled in continuous time.

Discrete-Choice Demand Models

In revenue optimization and pricing, the DPM captures product-specific and consumer-specific granularity. Marandi and Lurkin formulate expected demands using a discrete mixed logit model, where each price $p_j$ belongs to a finite set $P_j$ (Marandi et al., 2020). The utility for individual $i$ from product $j$ is modeled as

$U_{ij}(p) = \alpha_{i0} + \alpha_{i1} p_j + \beta_i^T x_j + \epsilon_{ij},$

with price sensitivities $\alpha_{i1} < 0$ , and multinomial probabilities mixing over discrete parameter segments.

Integer-Valued Observation-Driven Processes

For point-process or return modeling, the DPM may take the form of state-space models with discrete-valued increments. Holý and Tomanová use mixtures of double Poisson laws, while Barunik et al. develop a score-driven GARCH framework for integer price changes, using a zero-inflated Skellam law (Holý et al., 2021, Holý, 2022).

Non-probabilistic and Pathwise DPMs

Pascucci and colleagues introduce trajectory-based DPMs for discrete non-probabilistic markets, specifying only the admissible paths $A_t$ 0 and a family of self-financing portfolios $A_t$ 1 (Ferrando et al., 2014). Pricing intervals are determined by global sup-inf (min-max) functionals over all paths.

2. Key Model Structures: Price Discreteness and Dynamics

Semimartingale and Poisson Construction

Shephard and Yang’s DPM leverages a Poisson random measure producing integer-valued jumps, combining "permanent" (x≤b) and "fleeting" (x>b) components, the latter governed by a trawl function d(s) that determines the survival and reversal kinetics of price increments. The resulting $A_t$ 2 process exhibits:

Piecewise-constant paths,
Finite activity and finite variation,
Decomposition into martingale and finite variation terms,
Empirical moment fitting via jump-size distributions and variograms (Shephard et al., 2014).

Discrete Choice and Static Pricing Problem

Discrete pricing arises naturally in static multiproduct optimization under constrained price sets $A_t$ 3. The DPM computes the expected revenue

$A_t$ 4

subject to $A_t$ 5. Discreteness in feasible prices leads to a nonconvex, combinatorial optimization domain (Marandi et al., 2020).

Score-Driven and MA(1)-Filtered DPMs

Observation-driven approaches for high-frequency price changes encode integers via a zero-inflated Skellam distribution with score-driven volatility, smoothed by MA(1) terms to control for microstructure autocorrelation. Intraday patterns and duration effects are integrated using nonparametric spline correction factors (Holý, 2022).

3. Estimation and Computational Algorithms

Moment-Based and Maximum-Likelihood Estimation

Shephard and Yang’s DPM parameters can be fit to high-frequency data via closed-form estimators for the empirical jump-size pmf, power-variations, and variogram slopes, leveraging the exclusion of diffusion (Brownian) terms and the structure of the trawl (Shephard et al., 2014).

Holý–Tomanová’s mixture DPM employs MLE, recursively updating dynamic parameters (mean, dispersion, clustering index) with GAS/GARCH-type and multinomial logit recursions. Estimation involves nonlinear score-driven updates and derivative-free optimization (Holý et al., 2021).

Barunik et al. specify the full log-likelihood for sequential price changes, accounting for irregular timing and microstructure noise, and use numeric maximization for parameter estimation (Holý, 2022).

Exact Algorithms for Discrete Pricing

Marandi and Lurkin develop CoBiT, an exact global algorithm for static pricing under DPM:

Lower Bound via Trust-Region LO: At each iteration, a linearized trust-region subproblem provides a feasible, improving solution.
Upper Bound via Convex McCormick Relaxation: Biconvex reformulation with auxiliary variables and McCormick convex envelopes provides tight upper bounds.
Branch-and-Bound: Recursive interval partitioning on price variables closes the optimality gap.

Complexity is polynomial in $A_t$ 6 and $A_t$ 7 in practical cases, with empirical results confirming global optimality where off-the-shelf solvers fail (Marandi et al., 2020).

Pathwise Min-Max for Option Pricing

Discrete non-probabilistic DPMs define price intervals via global min-max functionals, leveraging dynamic programming and local up-down node conditions to ensure coherence and rule out arbitrage in a pathwise, model-free sense (Ferrando et al., 2014).

4. Applications and Empirical Findings

High-Frequency Asset Price Dynamics

Empirical fits of Shephard–Yang DPMs to futures data show:

“Reversal probability” (fraction of fleeting moves) from ≈0.35–0.81,
Jump intensity ∥v∥ ranging 0.014–0.28 ticks/sec,
Best-fit to variance signatures and jump-size histograms achieved with superposition-GIG trawl; exponential trawl underfits microstructural timescales,
Negative autocorrelation of returns on sub-second to minute scales,

Model extensions allow for seasonality, time-varying trawl, and multivariate joint dynamics (Shephard et al., 2014).

Statistical Properties of Discrete Efficient Price Estimators

Estimation of a fundamental price using rebate-adjusted, squared-volume imbalance at the best quotes yields estimators outperforming traditional mid-price or linear benchmarks in terms of explained variance, bias, and return autocorrelation. The DPM preserves key linear response properties on average, despite tick-size and rebate effects (Bonart et al., 2016).

Pricing with Discrete Consumer Choice

In multiproduct contexts, DPMs enable explicit optimization across discrete prices, incorporating rich consumer heterogeneity and finite price possibilities. CoBiT finds global optima in practical cases (J≤10), with stability in optimal prices and market shares once mixed-logit distributions are approximated with sufficient granularity (Marandi et al., 2020).

Price Clustering Phenomena

Holý–Tomanová’s DPM exposes a reversal in clustering-volatility relationships: at ultra-high frequency, higher volatility reduces clustering at tick multiples (contrasting with the positive daily effect), attributable to the dominance of algorithmic traders versus low-frequency rounders (Holý et al., 2021).

Discrete-Grid Nash Equilibria

In supply chain duopoly models, discrete price increments (Δ) directly affect equilibrium existence, uniqueness, and profitability. Finer Δ amplifies price competition and can eliminate pure-strategy equilibria (price war to marginal cost), while larger Δ enables stable positive-margin NE; segments of loyal customers further moderate competitive pressures (Wadhwa et al., 5 Jun 2025).

Min-Max Pricing of Derivatives

The pathwise DPM for option pricing provides intervals for superhedges and subhedges under transaction discreteness, with dynamic programming allowing practical computation. The framework accommodates limited arbitrage, coherence under local neutrality, and correspondence to risk-neutral prices when the trajectory set reflects martingale evolution (Ferrando et al., 2014).

5. Theoretical Implications and Limitations

DPMs highlight the structural and algorithmic consequences of discreteness:

The lack of a Brownian component and small-jump asymptotics in semimartingale DPMs preserves integer-valued increments, crucial for modeling high-frequency reversals.
Nonconvexities induced by finite pricing grids require tailored solution algorithms (e.g., McCormick relaxations for global optimization).
Trajectory-based DPMs show that price intervals can exist even with limited forms of arbitrage when sharp local neutrality conditions are met.

A plausible implication is that DPMs are essential wherever tick rules, regulatory denomination, or operational granularity dominate aggregate price formation or decision-making. However, they may require substantially more computational effort (branch-and-bound, combinatorial enumeration) compared to continuous relaxations.

6. Synthesis and Outlook

The Discrete Price Model framework, spanning stochastic processes, combinatorial optimization, and pathwise finance, provides a mathematically exact characterization of markets and systems governed by price discreteness. It reconciles efficient pricing with discrete mechanisms, clarifies the boundaries of Nash equilibria under denomination constraints, and enables robust empirical estimation under integer-valued, high-frequency data regimes.

The DPM paradigm is increasingly significant in environments where digital trading, microstructure features, and consumer choice coexist with real-world pricing limitations. Future advances may focus on integrating richer sources of heterogeneity, adaptive dynamics (e.g., time-varying trawl, randomization of grid size), and computationally scalable inference in ultra-granular contexts.