This question is motivated by an example in Section 4.7.3. Open-Loop and Closed-Loop Equilibria in Games with many players in Game Theory by Fudenberg and Tirole.

The definitions of closed-loop and open-loop in the book are given below.

Closed-loop and Open-loop strategies(Section 4.7.1 in FT): Our definition of a multi-stage game with observed actions corresponds to the closed-loop information structure, where players condition their play at time $t$ on the history of play until that date. The corresponding strategies are called closed-loop strategies, while open-loop strategies are functions of calendar time alone.

The following example is given to illustrate that an open-loop equilibrium can also be subgame-perfect equilibrium.

Section 4.7.3 in FT: Consider a game has a continuum of nonatomic individuals of each player type--a continuum of player 1s, a continuum of player 2s, and so on. (Let the set of individuals be copies of the unit interval endowed with Lebesgue measure for concreteness). Suppose further that each player $i$'s payoff is independent of the actions of any subset of opponents with measure $0$. Then if one individual player $j$ deviates, and all players $k\not=i,j$ ignore $j$'s deviation, it is clearly optimal for player $i$ to ignore the deviation as well. Thus the outcome of an open-loop equilibrium is subgame perfect.

My question is that, in the argument above, why do we need the part that it's optimal for player $i$ to ignore $j$'s deviation given others doing so?

To my understanding, to prove an open-loop equilibrium to be subgame perfect, we need to apply the one-shot deviation princinple. Therefore suppose player $j$ deviates, then because of the assumption and the fact that the measure of player $j$ is $0$, this will not affect the payoff of any player. Hence if a strategy profile constitutes an open-loop equilibrium, it's also subgame perfect.

However, the argument in the book seems to indicate that we not only need to check whether player $j$'s deviation is profitable, we also need to check whether there is a profitable deviation for player $i$ in the subgame in which $j$ already deviates. I am a little confused about how to apply the one-shot deviation principle.

To make my confusion more specific, consider an infinitely repeated version of the two-person prisoners' dilemma, and consider the following strategy profile: Both players will choose $C$, and if one player chooses $D$ at some $t$, both players will play $D$ for all $\tau>t$.

Suppose we want to check whether the above strategy profile constitutes a subgame perfect equilibrium, we apply the one-shot deviation principle by the following steps.

Step 1:We assume player $i$ deviates to $D$ at some time $t$, and no deviation has occurred until $t$, we check whether this deviation is profitable.

Step 2:Next we assume player $i$ deviates to $C$ in some subgame in which a deviation has already occurred and check whether such deviation is profitable.

Step 3:If the answer is no in both steps, we say that the above strategy profile constitutes a subgame-perfect equilibrium.

In this particular setting, my question is that, in step 1, do we also need to check whether there is a profitable deviation for player $-i$ after the deviation of $i$ from $C$ to $D$? How to distinguish between deviations in all subgames and one-shot deviation?

Thanks in advance for your help!


1 Answer 1


It's a bit more complicated. Treating open-loop strategies as closed-loop strategies- that happen to play a given action in each period irrespective of the history- will not give you a subgame-perfect equilibrium. So there is no given closed-loop equilibrium that you can test using the one-deviation principle for being a subgame perfect equilibrium.

Rather, the claim is that the outcome path of an open-loop equilibrium will also be the outcome path of a closed-loop subgame perfect equilibrium with a continuum of individually insignificant players. The textbook doesn't quite spell out the technical details, which can be found in [Fudenberg, Drew, and David K. Levine. "Open-loop and closed-loop equilibria in dynamic games with many players." Journal of Economic Theory 44.1 (1988): 1-18.].

The idea is that most subgames can actually not be reached by individual deviations, so how play would proceed there becomes strategically irrelevant. If players ignore deviations of measure-zero sets of agents- which they can do since payoffs will not depend on those agents- the strategic situation will not change and it will still be optimal to follow the given plan. Individual players cannot get the play into subgames in which a positive measure set of players play differently. But in order to get a subgame perfect equilibrium giving rise to the outcome, one would need to specify strategies appropriately for such hard-to-reach subgames too.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.