As consumers have limited capacity to process information, advertisers must compete for attention. This creates information congestion which produces social loss like unread advertisements. We apply population games and best response dynamics to analyze information congestion. Multiple equilibria impair traditional policies, and thus, non-traditional policies are examined to lead the system to a Pareto efficient equilibrium. We achieve this, for example, by changing the cost per message multiple times during the evolutionary process. In this process, policymakers gradually investigate externalities. However, these policies are costly, which confirms the inefficiency of advertising structures where advertisers send messages regardless of consumer interests.

There are two popular equilibrium concepts for finite extensive-form games, perfect equilibrium and sequential equilibrium. We find a relatively simple necessary and sufficient condition with which sequential equilibrium is perfect. We interpret the condition while referring to lexicographic domination proposed in Okada (1991). In particular, when each path includes at maximum two decision nodes in a game, regardless of the number of players in the game, any lexicographically undominated strategy combination is a perfect equilibrium, and any perfect equilibrium is a lexicographically undominated strategy combination. Inaddition, we indirectly discuss “perfect equilibrium” in games with uncountable actions via lexicographic domination, which is applicable to uncountable actions.

We analyze the necessary and sufficient condition with which sequential equilibrium (Kreps and Wilson 1982) is perfect (Selten 1975). This From-Sequential-equilibrium-To-Perfect-equilibrium (FSTP) condition consists of special types of strategy profiles (well-mixed strategy profiles) which are slightly weaker than completely mixed strategy profiles. Well-mixed strategy profiles are applicable to uncountable strategy sets. In addition, well-mixed strategy profiles enable us to check various rationalities based on completely mixed strategy profiles while requiring less data on payoffs of choices.