The performance in terms of signal-to-interference ratio (SIR), teletraffic, and spectral efficiency of a combined macrocellular and microcellular network is investigated when either both types of cells share the same channel set, or when the channel set is partitioned between the macrocells and the microcells. The analysis is for time-division multiple access (TDMA) with frequency hopping, power control, and discontinuous transmission, and the radio channel is composed of an inverse fourth-power path loss law with log-normal fading. We commence by introducing a single microcell into a hexagonal cluster of macrocells before considering clustered microcells. Both omnidirectional and sectorized cells are examined. We find that high reuse factors are required when channel sharing is employed. When channel partitioning is used, no co-channel interference occurs between the microcells and the macrocells allowing them to be planned independently. The reuse factors in the microcells and macrocells therefore do not need to be increased beyond conventional values. The outcome is that by opting for channel partitioning, the improvement in spectral efficiency compared to channel sharing is two to three times greater.