The radio resource allocation problem in wireless networks is a major research issue because the resource is scarce, and it must be shared among multiple network components. In this paper, we consider the resource allocation problem in downlink broadcast services in IEEE 802.16j WiMAX networks, where the objective is to serve as many mobile stations (MSs) as possible with the given resource budget. Solving the problem is not trivial because allocating different amounts of resources to the multihop relay base station (MR-BS) and relay stations (RSs) yields different network topologies and, thus, diverse routing paths from the MR-BS to the MSs. Existing approaches only provide inflexible heuristic resource allocation strategies or solutions based on a limited network scenario. More significantly, they focus on single-level relay networks (i.e., an RS can only relay data for MSs, not for other RSs). However, in practice, RSs can relay traffic from other RSs, and the resource allocation could be more efficient. This motivates us to formalize the problem as a multilevel broadcast receipt maximization (ML-BRM) problem. We show that the problem is NP-complete and propose a multilevel utility-based resource allocation (ML-URA) mechanism to solve the ML-BRM problem. In addition, we present a computation reduction mechanism that reduces the running time of the ML-URA substantially. The performance evaluation via simulations shows that the proposed ML-URA can serve more MSs than existing approaches, and the derived solution is very close to the optimum under different network scenarios and resource budgets. Finally, we show the approximation ratio between the ML-URA and the optimal solution.