This paper proposes a cooperative game theoretic resource allocation scheme for downlink OFDMA wireless networks. We define a user¿¿s payoff as a function of his data throughput. Then the source allocation can be modeled as a cooperative game to maximize the aggregate payoff of the users. To achieve the Nash bargaining solution (NBS) of the game, a suboptimal subcarrier allocation is first performed by assuming an equal power allocation. An optimal power allocation algorithm then maximizes the sum payoff while ensuring QoS demand (the minimum data throughputs) of the users. Through comparing with the other two resource allocation algorithms, i.e., the maximizing system capacity and the max-min fairness algorithm, simulation results show that the cooperative game acquired a good tradeoff between the fairness and the overall system capacity. The user¿¿s QoS demand is ensured, and the system resource is utilized efficiently.