📈 economics

Herding Prices: Informational Inertia and Price Dispersion

This paper analyzes a duopoly model where switching costs create "informational inertia" that causes consumers to stick to default sellers despite private signals, thereby generating price dispersion and altering the transmission of private information through purchase histories even as firms adjust prices under Poisson opportunities.

Original authors: Georgy Lukyanov, Ariza Azova

Published 2026-08-04

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Georgy Lukyanov, Ariza Azova

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Science of Guessing and the Power of "Staying Put"

Imagine you are walking into a room full of people trying to guess which of two mystery flavors of ice cream is actually the best. This is the world of observational learning, a branch of economics and game theory that studies how we learn from watching what others do. Usually, we assume that if everyone starts buying Flavor A, it's because they've all figured out it's the winner. But what if people aren't just guessing? What if they are stuck?

Two key ideas drive this story. First, there is private information: everyone has a little secret hint (like a taste test they did alone) that tells them what they think is best. Second, there are switching costs: the invisible "friction" or hassle of changing your mind. Maybe you're used to a specific brand, or your computer is already set up for it, so even if you think the other option is better, you might just stay put because the effort to switch feels too high. This paper asks a fascinating question: If everyone is stuck in their own little rut, does the crowd still learn the truth? Or does the market just spin its wheels, pretending to learn while actually saying nothing at all?

The Ice Cream Paradox: When Silence Speaks Volumes

In this paper, authors Georgy Lukyanov and Arina Azova set up a playful but precise experiment involving two ice cream shops, let's call them Shop A and Shop B. They are trying to figure out which flavor is the "real winner," but neither the shops nor the customers know for sure at the start. Customers arrive one by one, each getting a tiny, private clue (a "signal") about which flavor is better.

Here is the twist: every customer is secretly "assigned" to one shop before they even walk in. Maybe they have a loyalty card, or their phone is pre-loaded with Shop A's app. To buy from the other shop, they have to pay a "switching cost"—a mental or real fee for the trouble of changing.

The authors discovered something counterintuitive: The market can get stuck in a state of "Informational Inertia."

Imagine a crowd where half the people are glued to Shop A and half are glued to Shop B. Even if a customer's private clue screams, "Shop B is amazing!" they might still buy from Shop A because the hassle of switching is too high. If this happens to everyone, the shops keep selling ice cream, and the crowd keeps watching. But here's the kicker: nobody learns anything. The sales numbers just look like a random split of the crowd's original assignments, not a reflection of which ice cream is actually better. The market is "silent." It's like a classroom where everyone raises their hand to answer, but they are all just repeating what they were told to say yesterday, ignoring the new facts in front of them.

The Price Tag as a "Wake-Up Call"

So, how do you break this silence? The paper shows that prices are the key.

In a world where prices are stuck (fixed), the market can stay in this "inertial zone" forever. The shops keep selling, but the customers' private clues remain hidden. However, if one shop changes its price, it acts like a wake-up call.

Think of the price gap as a lever. If Shop A drops its price just enough, it overcomes the "switching cost" for the people glued to Shop B. Suddenly, those customers switch. When they do, their purchase reveals their private clue: "I switched because I really think Shop A is better!" The market wakes up, and learning resumes.

But there's a catch. If the price difference gets too huge, the market goes silent again, but for a different reason. If Shop A is so cheap that everyone buys from it, regardless of their private clues, then the next purchase tells us nothing new either. It's like a crowd all rushing the same exit; you can't tell who knew the way and who just followed the herd. The "sweet spot" for learning is a moderate price difference that is just enough to make people switch, but not so huge that everyone blindly follows the cheapest option.

The "Heterogeneity" Test: Is This Real or Just Math?

You might wonder: "Is this 'stuck' behavior just a math trick because the authors assumed everyone has the exact same switching cost?" To be sure, the authors added a layer of realism. They imagined that switching costs aren't identical for everyone; some people find it easy to switch, while others find it hard. They modeled this as a range of costs, with a "floor" (a minimum cost that no one can go below).

Their finding was robust: As long as there is a minimum cost to switch (even a tiny one), the "inertial silence" still exists. The market can still get stuck where no one learns. However, if that minimum cost drops to zero (meaning anyone can switch for free), the silence disappears, and the market starts learning again. This proves that the "stuck" behavior isn't just a glitch in the math; it's a real economic phenomenon caused by the friction of changing your mind.

The Game of "Staggered" Prices

Finally, the authors looked at what happens when shops don't just set one price and forget it. In the real world, prices change at different times (like when a contract expires or a manager approves a sale). This is called "staggered pricing."

They found that even in this messy, real-world scenario, a stable pattern emerges. Shops can settle into a rhythm where they occasionally change prices to break the silence, but they also know that changing prices too often or too drastically can mess up the information flow. The paper proves that on a realistic grid of possible prices (like the specific cents you see on a price tag), there is always a stable "equilibrium" where both shops play optimally, balancing the need to make money now with the need to gather information for later.

The Bottom Line

This paper doesn't just tell us that people are reluctant or constrained. It shows that friction creates silence. When it's too much trouble to switch, the crowd stops learning from each other, even if they are all trying to make the best choice. Prices aren't just about money; they are the signal that tells the market, "Okay, it's time to pay attention to the clues again."

The authors didn't just guess this; they built a rigorous mathematical model to prove it. They showed that this "inertial" state is a distinct possibility, separate from the usual idea of a "herd" where everyone blindly follows the leader. In fact, the market can be divided and active, yet completely uninformative. It's a reminder that sometimes, the most expensive thing in a market isn't the product itself, but the cost of changing your mind.

Technical Summary: Herding Prices: Informational Inertia and Price Dispersion

Problem Statement
This paper investigates a continuous-time duopoly where firms and consumers are initially uncertain about which of two products better satisfies a common need (a common-value setting). The core mechanism driving the analysis is the interaction between observational learning and switching costs. Consumers arrive sequentially, observe public prices and purchase history, receive a private signal regarding product quality, and are privately assigned to one firm (a "default"). Buying from the non-assigned firm incurs a switching cost.

The central problem is to characterize how switching costs alter the transmission of private information through observed market shares. In standard information cascade models, learning stalls when all agents herd on a single action regardless of their signals. Here, the authors identify a distinct form of stalled learning called informational inertia: the market remains divided between the two firms (reflecting initial assignments) even when private signals recommend the alternative, rendering observed purchases uninformative about the true state.

Methodology
The authors employ a continuous-time stochastic game framework with the following components:

State and Beliefs: The state $\theta \in \{A, B\}$ is fixed. Firms and consumers share a common prior $\mu$ . Firms receive no private signals; they only update beliefs based on public purchase history.
Consumer Dynamics: Consumers arrive via a Poisson process. Each has a private assignment $r \in \{A, B\}$ and a binary private signal $s \in \{+, -\}$ . They maximize expected utility minus price and switching cost $\kappa$ .
Pricing Regimes:
- Fixed Prices: Used as a benchmark to derive closed-form discounted demand kernels and analyze the learning process in isolation from strategic price changes.
- Staggered Pricing (Calvo): Firms receive independent Poisson opportunities to reset prices. This introduces dynamic strategic interaction where price changes alter the "experiment" faced by subsequent consumers.
Heterogeneity Perturbation: To address technical issues of consumer indifference and impossible purchase histories in the deterministic model, the authors introduce a privately heterogeneous switching cost $K_\epsilon = \kappa + \epsilon\eta$ , where $\kappa$ is a positive floor and $\eta$ is a continuous random variable. They analyze the limit as $\epsilon \to 0$ .

Key Results

1. Consumer Choice and Informational Regimes
The authors derive two assignment-specific posterior cutoffs, $a(d, \kappa)$ and $b(d, \kappa)$ , which partition the price gap space into distinct regions:

Dominance Regions: Large price advantages induce all consumers to buy from the cheaper firm, regardless of signals. These are uninformative.
Signal-Revealing Region: Moderate price gaps allow signals to drive choices, making purchases informative.
Informational Inertia Region: When the price gap lies within a specific band determined by the switching cost and signal precision, consumers with a positive signal assigned to the "wrong" firm (and vice versa) stay with their default.
- In this region, the probability of buying from Firm A is exactly the assignment share $\phi$ in both states.
- Consequently, Bayes' rule leaves the public belief $\mu$ unchanged despite ongoing sales.
- Crucially, this inertia is distinct from herding: the market is split, not unanimous, yet no information is aggregated.

2. Fixed-Price Benchmark and Equilibrium
Under fixed prices with zero switching costs ( $\kappa=0$ ) and a symmetric prior:

The authors derive a closed-form discounted-demand kernel.
If the feasible price range is bounded by the signal-revealing threshold ( $\bar{p} \leq h$ ), there exists a unique symmetric, atomless mixed-strategy price equilibrium.
The equilibrium price distribution is explicitly characterized. The support width and mean price depend on signal precision and the discount factor.
The analysis highlights that prices do not merely allocate current demand but determine the future path of beliefs by shifting the market between informative and inertial regions.

3. Heterogeneous Switching Costs and Robustness
The paper demonstrates that informational inertia is robust to private heterogeneity, provided there is a positive switching-cost floor ( $\kappa > 0$ ):

Introducing heterogeneity ( $\epsilon > 0$ ) removes consumer indifference and ensures every purchase has positive probability, eliminating "impossible" histories.
However, if $\kappa > 0$ , the exact inertia region (where $g_+ = g_- = \phi$ ) persists.
If the floor is zero ( $\kappa = 0$ ), exact inertia disappears, and every finite price gap becomes informative.
Thus, the mechanism relies on the existence of a non-negligible cost to switch, not on the absence of heterogeneity.

4. Staggered Pricing and Dynamic Equilibrium
In the dynamic setting with staggered price resets:

A price reset does not immediately change beliefs (as firms have no private info) but changes the informativeness of the next purchase.
The authors prove the existence of a stationary Markov equilibrium for the "regularized" game (finite price grid, heterogeneous costs).
The reachable state space is countable. The equilibrium allows for alternating episodes of learning and inertia.
As heterogeneity vanishes ( $\epsilon \to 0$ ), the sequence of equilibria converges to a public subgame-perfect equilibrium of the deterministic game, with boundary behaviors selected by the perturbation.

Significance and Claims
The paper claims to identify a specific mechanism—informational inertia—where switching costs prevent the aggregation of private information even in a divided market. This challenges the conventional view that stalled learning requires unanimous herding.

Distinction from Cascades: Unlike standard cascades where the market converges to a single action, inertia maintains market shares that mirror initial assignments rather than quality assessments.
Role of Prices: Prices are shown to be more than allocation devices; they act as switches that determine whether the market is in a learning-active or learning-silent regime.
Robustness: The inertia mechanism is shown to be an artifact of the floor of switching costs, not the determinism of the model. It survives the introduction of continuous heterogeneity, provided the cost to switch cannot be arbitrarily small.
Limitations: The authors explicitly state they do not claim "almost-sure learning." With binary signals and bounded heterogeneity, public information need not become arbitrarily rich. The contribution is identifying when learning stops and how repricing can restart it, rather than proving eventual convergence to the true state.

The paper concludes that in markets with switching costs and staggered pricing, the sequence of purchases transmits private information only intermittently, and the equilibrium price dispersion is driven by the strategic manipulation of these informational regimes.