The available channel bandwidth may vary due to many factors including network traffic congestion. Therefore, a video server should be able to decide how to reduce the transmission rate so as to maximize the signal quality at the receiver. The video encoder can scale the spatial resolution, temporal resolution or SNR to decrease the rate. In this paper, we study which option or which particular combination of these three scalability options needs to be employed to maximize the perceived signal quality for a given video sequence. The existing research in this area primarily concentrates on down-conversion of MPEG video by frame-dropping or by changing the quantization scale factor as opposed to considering the full spectrum of scalability options.