This paper presents an encoder-based prediction model to study the impact of packet loss on H.264 scalable video coding (SVC). A Markov Chain (MC) with 2^N states is developed to describe the error propagation process inside a group of pictures (GOP). Using the packet loss rate reported by RTCP (RTP Control Protocol) and analyzing the inter-frame prediction rules, we evaluate packet-loss-induced quality degradation in SVC as the probability that a given scalable layer can be correctly received. Based on the proposed model, the performance of the SVC hierarchical B-frame structure, the zero delay encoding/decoding structure, and the advance video coding (AVC) IPPP structure (compatible base layer in SVC) are evaluated and compared under both random and burst packet loss events.