Weak Atomic Continuity and Subgame Perfection

summary

The gist

Subgame-perfect equilibrium need not exist in continuous games with perfect information when Nature moves Under standard regularity assumptions…

In short

The paper proves that weak atomic continuity of Nature's transition probabilities guarantees a pure-strategy subgame-perfect equilibrium exists for dynamic games with perfect information across finite and infinite horizons. This condition is necessary for this existence result uniformly over game primitives, meaning its violation allows for the construction of games without any subgame-perfect equilibrium.

Key concepts

Weak Atomic Continuity
This topological condition ensures that every atom in a limiting probability distribution is matched by atoms whose locations and probabilities converge to it. It is crucial because it implies weak continuity, which is stronger than ordinary weak convergence, and guarantees the existence of equilibrium strategies.
Subgame-Perfect Equilibrium (SPE)
An SPE is a set of strategies where every player's strategy constitutes a Nash equilibrium in every subgame. The paper proves that under specific conditions, such as weak atomic continuity, such an equilibrium can be found for dynamic games with perfect information.
Continuation Payoff Correspondence (Qt)
This represents the set of attainable payoffs achievable after making optimal choices at every realized state. The authors show that weak atomic continuity ensures this correspondence is non-empty, compact-valued, and upper hemicontinuous, allowing players to find optimal actions by integrating expected payoffs.
Necessity Proof
The necessity proof demonstrates that if the weak atomic continuity condition is violated, one can construct a dynamic game with no subgame-perfect equilibrium. This establishes the condition as a necessary requirement for guaranteeing SPE existence uniformly across all game primitives.

Terminology used across episodes

This episode discusses

The paper

Weak Atomic Continuity and Subgame Perfection · Read on arXiv

Enxian Chen, Wei He, Yeneng Sun, Hanping Xu

School of Economics, Nankai University · Department of Economics, The Chinese University of Hong Kong · Departments of Economics and Mathematics, National University of Singapore · Institute for Advanced Study in Mathematics, Harbin Institute of Technology

Subgame-perfect equilibrium need not exist in continuous games with perfect information when Nature moves. Under standard regularity assumptions, we prove that weak atomic continuity of Nature's transition probabilities yields a measurable pure-strategy subgame-perfect equilibrium for finite and infinite horizons. The condition, originating in the population genetics literature, is necessary for the existence of subgame-perfect equilibrium uniformly over the remaining primitives: whenever this condition is violated, one can construct a dynamic game with no subgame-perfect equilibrium, even in behavioral strategies. The proof preserves compactness and upper hemicontinuity of continuation payoff correspondences under integration without requiring convexity.

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: "Weak Atomic Continuity and Subgame Perfection".

Warren: Here's the gist of the paper. Subgame-perfect equilibrium need not exist in continuous games with perfect information when Nature moves Under standard regularity assumptions, we prove that weak atomic continuity of Nature’s transition probabilities yields a measurable pure-strategy subgame-perfect equilibrium for finite and infinite horizons The condition, originating in the population genetics literature, is necessary for the existence of subgame-perfect equilibrium uniformly over the remaining primitives Whenever this condition is violated, one can construct a dynamic game with no subgame-perfect equilibrium, even in behavioral strategies Core Finding and Necessity The paper establishes that weak atomic continuity of Nature’s transition probabilities guarantees the existence of a pure-strategy subgame-perfect equilibrium for dynamic games with perfect information across both finite and infinite horizons This condition is shown to be necessary for such an existence result uniform over the remaining game primitives The necessity is demonstrated by constructing a two-player, five-stage game with no subgame-perfect equilibrium, even in behavioral strategies, when weak atomic continuity is violated This construction preserves the parameter space, state space, and transition probability The Role of Weak Atomic Continuity Weak atomic continuity is defined as a condition requiring that every atom of the limiting distribution be matched by atoms whose locations and probability weights converge to those of the limiting atom This topology, introduced by Ethier and Kurtz (1994), is used to study convergence in population genetics while keeping track of the frequencies of individual genetic types The paper notes that weak atomic continuity implies weak continuity, which is stronger than ordinary weak convergence It also coincides with weak continuity for deterministic transition probabilities and for transition probabilities on a fixed finite state space The difference between the two conditions arises when probability concentrates at a limiting atom without a matching atom along the approximating histories Methodological Observation and Proof Structure The existence proof relies on the methodological observation that expected payoffs are obtained by selecting an equilibrium continuation after each realized state and integrating the resulting payoffs against the transition probability The authors prove that weak atomic continuity preserves nonemptiness, compactness, and upper hemicontinuity in this step These properties allow an optimal action to be supported by suitable equilibrium continuations without requiring the convexity of the payoff sets The integration result preserves compactness and upper hemicontinuity of the continuation-payoff correspondence without requiring convexity Results in Finite and Infinite Horizons The results are presented through two main theorems concerning the existence of subgame-perfect equilibrium Theorem 1 states that if Nature’s transition probabilities satisfy weak atomic continuity, a pure-strategy SPE exists, and at every date, Et has nonempty compact values and is upper hemicontinuous For the infinite horizon, the authors establish that the set of attainable continuation payoffs Qt is nonempty, compact-valued, and upper hemicontinuous The proof for the infinite horizon involves showing that for any truncation date N, QN t is non-empty and compact-valued, and that Qt = OtQt+1 Illustrative Example An illustrative example of the difficulty arises in Figure 1, where setting epsilon = 0 leads to a situation where the initial player’s limiting payoff is no longer attainable because distinct realizations of uncertainty become a single observed state This failure occurs when atomless distributions can shrink to a single state, and continuity of the payoff functions does not prevent this difficulty The example shows that under weak atomic continuity, the condition is satisfied by deterministic transition probabilities and for transition probabilities on a fixed finite state space The paper concludes by establishing a general existence theorem for dynamic games with perfect information under weak atomic continuity and proving its necessity for existence uniformly over the remaining game primitives The condition ensures that limits of attainable continuation payoffs remain attainable as Nature’s probability distributions vary with the history The authors show that any point of Qt(h) can be realized, thus Et = Qt Finally, they prove that no initial strategy is optimal under any profile that is an SPE in all proper subgames and the game has no behavioral SPE References Aliprantis, C.

Ruth: D., and K.

Warren: C.

Ruth: Border (2006).

Title and authors: Ruth: So we're looking at a paper called "Weak Atomic Continuity and Subgame Perfection" by Chen, He, and Sun. The title itself hints at what’s important here: the way uncertainty evolves over time.

Warren: Yeah, it sounds technical, but they are tackling subgame perfect equilibrium in continuous games with perfect information when Nature moves. That's a pretty dense setup to start with.

Iris: It starts by pointing out that you don't always have a pure-strategy subgame perfect equilibrium in these kinds of games when Nature is involved, even under standard regularity assumptions. The authors are trying to find the conditions under which one *does* exist.

Ruth: Exactly. They focus on this weak atomic continuity condition of Nature’s transition probabilities as the key ingredient that guarantees existence for both finite and infinite horizons. It comes from population genetics literature, which is interesting context for this kind of structural assumption.

Warren: So, instead of just looking at standard assumptions like compactness or continuity, they are isolating this specific type of convergence—the weak atomic continuity—as the necessary regularity condition to ensure that a measurable pure-strategy subgame perfect equilibrium can be found.

Iris: It seems they are building on older results from Harris and Hellwig, but adding this specific topological requirement to bridge the gap between finite and infinite horizon games under these continuous dynamics.

The paper's summary: Ruth: Now, looking at the core of what this paper does, it establishes that if Nature’s transition probabilities satisfy weak atomic continuity, then a measurable pure-strategy subgame perfect equilibrium exists for both finite and infinite horizons. That’s a pretty big claim to make in this area.

Warren: It means that instead of just assuming some general properties hold for the payoffs or the state space, this specific convergence condition on Nature’s transitions is enough to guarantee that an optimal strategy exists after every single move, even when you look ahead forever.

Iris: The paper also really stresses why this condition matters: it's necessary for existence uniformly over all the other primitives of these games. If you drop that continuity assumption, things break down and no subgame perfect equilibrium can be found, even in behavioral strategies.

Ruth: That’s a crucial point. They demonstrate this necessity by constructing a very specific two-player five-stage game where weak atomic continuity is violated and no subgame perfect equilibrium exists, even when players are allowed to use behavioral strategies.

Warren: That construction shows that you can keep the setup—the state space, the transition probabilities—the same while violating this condition and still have a game where no SPE exists. That really hammers home how essential this specific regularity is for the equilibrium concept to work in these dynamic settings.

Iris: And they also show that under this continuity assumption, they can preserve compactness and upper hemicontinuity of the continuation payoff correspondences even when you use integration, without needing those payoffs to be convex. That’s a strong methodological point for finding solutions.

The paper's improvements: Ruth: So what are the actual improvements or extensions they are suggesting beyond just proving existence under this condition? It seems the real meat is in how they handle the structure of those payoff correspondences themselves.

Warren: They show that this weak atomic continuity allows them to preserve compactness and upper hemicontinuity of those continuation payoff correspondences when they perform integration, which is a big deal because it means you don't need convexity for the optimal action to be supported by an equilibrium continuation.

Iris: That’s what I find interesting. Usually, finding an optimal action requires convexity in the payoff sets to guarantee you can find a solution easily. But here, they show that under weak atomic continuity, this requirement is lifted in this specific context of dynamic games with perfect information and Nature moving.

Ruth: They prove that any point of the attainable continuation payoff function at any history can actually be realized as an equilibrium payoff, which means the set of attainable payoffs is exactly what you need for finding the optimal strategy. That solves a major hurdle in continuous settings where limits don't always map nicely to attained values.

Warren: So, if we look at the infinite horizon case specifically, they establish that for any truncation date N, the set of attainable continuation payoffs up to that date is non-empty and compact-valued. Then they show that the payoff set at any time t is just related to the payoff set at time t plus one.

Iris: That structure—that Q t equals O t Q t+one —is what links all the stages together under this continuity assumption, allowing them to use those compactness results from the finite horizon argument to nail down the infinite horizon case <ref:2611.06620#pg2>.

Conclusion: Ruth: So, wrapping things up with "Weak Atomic Continuity and Subgame Perfection," we see that this specific condition on Nature’s transition probabilities is what guarantees a pure-strategy subgame perfect equilibrium for dynamic games with perfect information across both finite and infinite horizons.

Warren: The real implication is that this isn't just another existence proof under some mild assumptions; it’s a necessary structural requirement for the game to have an equilibrium at all when uncertainty evolves continuously. If you violate this, the game structure simply doesn't support a subgame perfect equilibrium.

Iris: And their construction showing no SPE even in behavioral strategies when continuity fails provides a really strong check for anyone trying to build models where they expect equilibria to exist under weaker regularity conditions. It sets a clear boundary for what kind of uncertainty structure is mathematically viable for these games.

Ruth: Exactly, and the paper’s final result confirms that this condition ensures limits of attainable continuation payoffs remain attainable as Nature’s probability distributions vary with the history, meaning any point in that payoff set can be reached.

Warren: So, to summarize, they proved existence under weak atomic continuity for finite and infinite horizons, showed it's necessary by construction, and demonstrated how this continuity handles the payoff correspondences without needing convexity. That’s the picture of what they achieved with "Weak Atomic Continuity and Subgame Perfection".

More episodes

← Home