A method and an apparatus for load balancing a serving subcell providing a data service to one or more user equipments in a cellular communication system are provided. The method includes calculating a load metric by using a radio resource occupation rate of data traffic in one or more scheduling types except for a Best Effort (BE) scheduling type, determining whether the serving subcell is in an overload state by using the load metric, and triggering a load balancing algorithm when it is determined that the serving subcell is in the overload state. Accordingly, the load metric used for load balancing may be defined to have a value closer to an actual free load, and thus the load balancing may be efficiently performed.