Direct load control has the potential of minimizing operational costs of electric utilities, which could thus reduce electricity prices for their customers. However, turning on and off specific appliances during peak demand periods has proven difficult to implement in practice due to system complexity, lack of fine-grained metering data, and scalability issues as the number of controllable loads increases. Therefore, in this study we want to investigate the extent to which it is possible to flatten power demands by operating load control at household granularity rather than appliance granularity. To this end we propose a hierarchical scheduling algorithm for communities of households, called HHS, which evenly distributes aggregate demands of background electrical loads: i) by leveraging on their cyclic on/off behaviors, and ii) without affecting the comfort levels of home occupants. Results obtained with realistic simulations show that our load management scheme can reduce peak demands by 30% in a community of 100 households.