When is the value of public information positive in a game?

The value of public information is studied by considering the equilibrium selections that maximize the weighted sum of players' payoffs. We show that the value of information can be deduced from the deterministic games where the uncertain parameters have given values. If the maximal weighted sum of equilibrium payoffs in deterministic games is convex then the value of information in any Bayesian game derived from the deterministic games is positive with respect to the selection. We also show the converse result that positive value of information implies convexity. Hence, the convexity of maximal weighted sum of payoffs in deterministic games fully characterizes the value of information with respect to considered selections. We also discuss the implications of our results when positive value of information means that for any equilibrium in a game with less information there is a Pareto dominant equilibrium in any game with more information.