Existence of Bayesian Equilibria in Incomplete Information Games without Common Priors

summary

The gist

Existence of Bayesian equilibria in incomplete information games without common priors focuses on introducing and proving the existence of Bayesian equilibria in finite-player incomplete information…

In short

This research introduces a condition called absolute continuity of beliefs as a sufficient requirement for finding Bayesian equilibria in incomplete information games where players have inconsistent beliefs and no common prior. The study shows that this condition is enough to guarantee an equilibrium exists, even in games with discontinuous payoffs, by relating the original game to a simpler surrogate game.

Key concepts

Absolute Continuity of Beliefs
This is a specific condition requiring the existence of a product probability measure that dominates each player's belief about the payoff state and others' types. It is distinct from having a common prior and serves as the key sufficient condition for equilibrium existence in this type of game.
Surrogate Game
The framework reformulates the original game into a simpler 'surrogate Bayesian game.' The common prior is replaced by this dominating product probability measure, which automatically satisfies absolute continuity of information, making equilibrium analysis easier.
Bayesian Equilibrium (BE) = Nash Equilibrium (NE)
The core result establishes that a Bayesian equilibrium exists if and only if a Nash equilibrium exists for some dominating product probability measure. This means the set of Bayesian equilibria is equivalent to the set of Nash equilibria in the surrogate game, up to null sets.

Terminology used across episodes

This episode discusses

The paper

Existence of Bayesian Equilibria in Incomplete Information Games without Common Priors · Read on arXiv

Denis Kojevnikov, Kyungchul Song

Tilburg University · University of British Columbia

DOI: 10.1016/j.jet.2026.106268

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: We're looking at the paper: "Existence of Bayesian Equilibria in Incomplete Information Games without Common Priors".

Warren: Here's the gist of the paper. Existence of Bayesian equilibria in incomplete information games without common priors focuses on introducing and proving the existence of Bayesian equilibria in finite-player incomplete information games where players hold mutually inconsistent beliefs without a common prior.

Ruth: This research is significant because it extends classical results by establishing absolute continuity of beliefs as a sufficient condition for equilibrium existence in a broad class of games, including those with discontinuous payoffs.

Warren: Pruning redundant reasoning tokens The paper introduces the notion of absolute continuity of beliefs, which is shown to be sufficient for the existence of Bayesian equilibria in a wide class of games without a common prior.

Ruth: Absolute continuity of beliefs requires the existence of a product probability measure that dominates each player’s belief about the payoff state and the other players’ types.

Title and authors: Ruth: Okay, so what the paper actually summarizes is that the core idea is to characterize Bayesian equilibria as the intersection of Nash equilibria across a whole family of surrogate games <ref:2504.16240#pg3>. This family of games is indexed by these product probability measures that dominate the players’ beliefs <ref:2504.16240#pg3>.

Warren: It basically says that if we can find just one of those dominating measures, there will be a Nash equilibrium in the surrogate game, and that existence implies the existence of a Bayesian equilibrium in the original game <ref:2504.16240#pg3>. That’s a powerful connection between two different types of games.

Iris: The key mechanism they use to make this work is reformulating the original game into a surrogate Bayesian game where the common prior becomes that dominating measure itself <ref:2504.16240#pg3>. This automatically satisfies absolute continuity of information, which is a big step because it bypasses the need for a common prior in the first place.

Ruth: So they are essentially using this surrogate game setup to create a scenario where the standard assumptions of equilibrium existence, like absolute continuity of information, are met by construction <ref:2504.16240#pg3>. It’s about finding a way around the common prior hurdle.

Warren: And they highlight that this general framework can handle games with payoffs that are discontinuous in actions, which is something the older literature struggled with because of continuity requirements <ref:two thousand five hundred four point one six two four zero#pg3. That's a major extension of what we usually see established in these existence proofs.

Iris: I see how they apply this to incomplete information Cournot games, showing that the results hold even in those settings where you might expect things to be rougher <ref:two thousand five hundred four point one six two four zero#pg3. It gives concrete evidence that this machinery isn't just theoretical fluff.

Ruth: So the summary boils down to: we define a set of games, find a Nash equilibrium in one of those games, and that existence guarantees an equilibrium in the original game <ref:2504.16240#pg3>. It’s a constructive way to prove existence without needing a common prior assumption <ref:2504.16240#pg1>.

Warren: It shifts the focus from assuming priors to finding the right measure, which is an interesting shift in strategy.

The paper's summary: Ruth: Now let’s talk about what they suggest as improvements or extensions to this framework, because this isn't just a statement of existence; it’s setting up a new way to analyze games <ref:two thousand five hundred four point one six two four zero#pg3.

Warren: They point to Theorem three point one as the main characterization, which states that Bayesian equilibria exist if and only if a Nash equilibrium of the surrogate game exists for some dominating product probability measure <ref:two thousand five hundred four point one six two four zero#pg3. It’s this equivalence that they argue is the most important structural result in this paper.

Iris: The real strength, in my view, is how they characterize the set of Bayesian equilibria as the intersection of those Nash equilibrium sets across all possible dominating measures <ref:two thousand five hundred four point one six two four zero#pg3. It tells us precisely what the set looks like—it’s a specific geometric intersection defined by these belief structures.

Ruth: And they show that this characterization is useful because it means that for any given measure, you can find a version of the Nash equilibrium in the surrogate game that actually corresponds to a Bayesian equilibrium in the original game <ref:two thousand five hundred four point one six two four zero#pg3. That’s a practical step toward finding an actual strategy.

Warren: They also tackle games with discontinuous payoffs, which we touched on before, by using Assumption three point one which allows for upper semicontinuity in the player's own action instead of requiring joint continuity <ref:two thousand five hundred four point one six two four zero#pg3. That’s a specific technical tool they developed to handle rough payoff functions.

Iris: And there’s this interesting point about when the condition itself breaks down, which is where the limitations come in <ref:two thousand five hundred four point one six two four zero#pg3. They show that absolute continuity of beliefs isn't always true if we have uncountable type sets, which means this method stops being universally applicable in those extreme cases.

Ruth: So they’re offering a very powerful tool for games without common priors, but they have to be careful where the underlying set structures become too complex for their measure-theoretic tools to handle gracefully <ref:two thousand five hundred four point one six two four zero#pg3.

The paper's improvements: Ruth: So wrapping up, the main point of this paper is that we can establish existence for Bayesian equilibria in games without a common prior by using absolute continuity of beliefs as a sufficient condition <ref:2504.16240#pg3}. This result is solidified in Corollary 3.1, which states that if those assumptions hold, the equilibrium exists > <ref:2504.16240#pg3>

Warren: And they extend this nicely to specific games like the incomplete information Cournot Game GC with a Bayesian equilibrium existing if the belief map profile satisfies absolute continuity of beliefs <ref:2504.16240#pg3>. That shows it’s not just abstract theory, but something applicable to specific models we already study >

Iris: From my side, I think the characterization is the set being equal to the Nash equilibria in a surrogate game up to null sets <ref:2504.16240#pg3>. That’s a very strong statement about what's actually there, even if it's defined using these abstract measures >

Ruth: So this framework offers a general path to prove existence in incomplete information games without common priors by showing that BE equals NE for each measure <ref:2504.16240#pg3}. It’s a key result for establishing the existence of Bayesian equilibria in these complex settings >

Warren: It’s a very solid result because it bypasses the need for a common prior assumption by providing an alternative, measure-theoretic condition for when an equilibrium can exist <ref:2504.16240#pg1>. The paper "Existence of Bayesian Equilibria in Incomplete Information Games without Common Priors" is a solid contribution to the literature on this topic.

Iris: It's definitely a piece that connects different areas of research, showing how belief continuity can be the link between different types of assumptions <ref:2504.16240#pg1>. That connection is what makes this work interesting >

Conclusion: Ruth: So we’re wrapping up on "Existence of Bayesian Equilibria in Incomplete Information Games without Common Priors." Basically, the paper shows that you don't need a common prior belief to prove an equilibrium exists if you use absolute continuity of beliefs <ref:2504.16240#pg3>.

Warren: Right. They establish that if you can find just one of those dominating product probability measures, a Nash equilibrium in the surrogate game will exist, which automatically guarantees a Bayesian equilibrium in the original game >

Iris: And they make that work by reformulating the whole thing into a surrogate Bayesian game where that measure becomes the common prior itself <ref:2504.16240#pg3>. That’s a clever move to bypass the need for an actual shared starting belief >

Ruth: It lets them set up a scenario where standard assumptions, like absolute continuity of information, are met by construction, rather than just assuming they hold true upfront >

Warren: And they show that this framework works even when the payoff functions are discontinuous in actions, which is a big deal because those kinds of rough payoffs usually cause problems for traditional equilibrium proofs >

Iris: But there's a caveat here, and it's important. They point out that absolute continuity of beliefs isn't always true if you have uncountable type sets, so this method has limits where the underlying type structure gets too messy >

Ruth: So the main result is that they characterize the set of Bayesian equilibria as being equal to the Nash equilibria in those surrogate games for any dominating measure <ref:2504.16240#pg3>. It’s a very constructive way to prove existence without needing that common prior assumption >

Warren: That means we shift the focus from assuming priors to actually finding the right measure, which is a new way of thinking about strategy in these games >

Iris: I think the connection they make between belief continuity and equilibrium existence is what makes this work interesting for econometricians like us, showing how different types of continuity can link up >

Ruth: It's a solid contribution to understanding when we can guarantee an equilibrium in incomplete information settings without relying on the common prior assumption >

Warren: That’s a solid result because it provides an alternative condition that allows for existence proofs where continuity assumptions usually fail >

Iris: It’s definitely a piece that connects different areas of research, showing how belief continuity can be the link between different types of assumptions >

Ruth: For those interested in how beliefs interact with strategy without needing a common starting point, this paper is the reference <ref:2504.16240#pg3>.

Warren: Next up, we’re looking at some work on discrete choice models and how social interactions affect decisions. Stick around for that.

More episodes

← Home