A Bertrand duopoly game with differentiated products reconsidered

Xiaoliang Li, Bo LiarXiv 2023 · DOI · arXiv

TL;DR

This paper provides the first rigorous analytical proof for the stability of a dynamic Bertrand duopoly game featuring differentiated products and boundedly rational firms. Using advanced symbolic computation, the authors reveal a striking paradox: while product substitutability usually stabilizes Cournot (quantity-based) markets, it actually destabilizes Bertrand (price-based) markets, potentially triggering chaotic price wars.

What Problem Does This Paper Address?

In economic theory, the "Bertrand model" describes a market where firms compete by setting prices. While classic models often assume linear demand for simplicity, real-world markets are non-linear and complex. Previous researchers (specifically Ahmed et al., 2015) proposed a more realistic model using a Constant Elasticity of Substitution (CES) utility function, but they could only explore it through numerical simulations because the underlying math was too "messy" for traditional analytical methods.

The authors of this paper aim to fill this gap. They seek to provide exact mathematical proofs for the existence, uniqueness, and local stability of market equilibria. Specifically, they want to understand how the degree of product differentiation—how "unique" a product feels compared to its competitor—affects the long-term stability of the market.

Methods

The researchers moved beyond standard calculus by employing symbolic computation tools. This allowed them to handle complex non-linear polynomial systems that are usually considered "unsolvable" in closed form. Key techniques include:

  • Triangular Decomposition: An algebraic method (an extension of Gaussian elimination) used to break down complex systems of polynomial equations into simpler, "triangular" sets that can be solved sequentially.
  • Partial Cylindrical Algebraic Decomposition (PCAD): A method used for quantifier elimination, allowing the authors to determine the exact regions in parameter space where the market remains stable.
  • Gradient Adjustment Mechanism: The model assumes firms are "boundedly rational," meaning they don't have perfect foresight. Instead, they adjust their prices incrementally based on local profit gradients (if raising the price slightly increased profit last time, they'll do it again).

Datasets

Not discussed in the paper. As a theoretical work in mathematical economics, the study relies on algebraic proofs and numerical simulations rather than empirical datasets.

Key Insights and Ideas

  • The Destabilization Paradox: The most significant finding is that increasing product substitutability (making products more similar) has a destabilizing effect. In Cournot models (where firms compete on quantity), similarity usually leads to stability. In this Bertrand model, similarity leads to intense price competition that can spiral into chaos.
  • Analytical Rigor: By using symbolic math, the authors proved the uniqueness of the non-vanishing equilibrium and established rigorous stability conditions for the first time, moving the field beyond "educated guesses" based on simulations.
  • New Bifurcation Discovery: The authors discovered the existence of a Neimark-Sacker bifurcation directly on the equilibrium. This is a specific type of mathematical transition where a stable point becomes an unstable one surrounded by a vibrating "loop," a detail missed by previous purely numerical studies.
  • Welfare vs. Price: In the case of identical marginal costs, the paper shows that while lower product differentiation leads to lower prices (good for consumers), it also leads to lower profits and lower overall social welfare.

Key Results

  • Equilibrium Uniqueness: The paper proves that for specific substitutability degrees ($\alpha = 1/2$ and $\alpha = 1/3$), there is exactly one unique, non-vanishing Nash equilibrium.
  • Stability Thresholds: The authors derived exact algebraic inequalities (e.g., Theorem 1 and Theorem 3) that define the "safe zone" for market parameters. If firms adjust prices too quickly or if marginal costs are too low, the system loses stability.
  • Transition to Chaos: Numerical simulations (Figure 5 and 6) confirm that as the substitutability degree ($\alpha$) increases, the system moves from a stable point to 2-cycle orbits, then to periodic orbits, and finally into full-blown chaotic attractors.

Limitations and Future Work

The primary limitation is that the analytical proofs are focused on two specific cases of substitutability ($\alpha = 1/2$ and $\alpha = 1/3$). This is because the general model involves exponential parameters that current symbolic computation tools struggle to solve generally.

Future work could involve:

  • Developing new algebraic algorithms to handle the general case for any value of $\alpha$.
  • Expanding the model from a duopoly (two firms) to a triopoly or an $n$-firm oligopoly.
  • Introducing heterogeneous costs or different types of consumer utility functions to see if the "destabilization paradox" holds true across other market structures.

Summary updated

Abstract

In this paper, we explore a dynamic Bertrand duopoly game with differentiated products, where firms are boundedly rational and consumers are assumed to possess an underlying CES utility function. We mainly focus on two distinct degrees of product substitutability. Several tools based on symbolic computations such as the triangular decomposition method and the PCAD method are employed in the analytical investigation of the model. The uniqueness of the non-vanishing equilibrium is proved and rigorous conditions for the local stability of this equilibrium are established for the first time. Most importantly, we find that increasing the substitutability degree or decreasing the product differentiation has an effect of destabilization for our Bertrand model, which is in contrast with the relative conclusions for the Cournot models. This finding could be conducive to the revelation of the essential difference between dynamic Cournot and Bertrand oligopolies with differentiated goods. In the special case of identical marginal costs, we derive that lower degrees of product differentiation mean lower prices, higher supplies, lower profits, and lower social welfare. Furthermore, complex dynamics such as periodic orbits and chaos are reported through our numerical simulations.
BibTeX
@misc{li2023bertrand,
  author = {Xiaoliang Li and Bo Li},
  title = {{A Bertrand duopoly game with differentiated products reconsidered}},
  year = {2023},
  eprint = {2301.01007},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2301.01007},
}
Figure