This paper deals with the construction of cooperative equilibria for stochastic dynamic games, where the players cannot observe the actions of their opponents. For a particular class of dynamic games with payoffs defined as the limit of average gains one obtains a relatively strong result showing the existence of perfect equilibria which are also Pareto-optimal. Such a general result is much harder to get when the payoffs are obtained as the expected discounted sum of transition rewards. However it is shown with the help of an example that one can elaborate cooperative equilibria based on the use of trigger strategies, when the trigger mechanism includes a long enough memory.