In this paper joint optimization of layers in the layered video coding is investigated. Through theoretical analysis and simulations, it is shown that, due to higher interactions between the layers in a SNR scalable codec, this type of layering technique benefits most from joint optimization of the layers. A method for joint optimization is then proposed, and its compression efficiency is contrasted against the separate optimization and an optimized single layer coder. It is shown that, in joint optimization of SNR scalable coders when the quantization step size of the enhancement layer is larger than half the step size of the base layer, an additional improvement is gained by not sending the enhancement zero valued quantized coefficients, provided they are quantized at the base-layer. This will result in a non-standard bitstream syntax and as an alternative for standard syntax, one may skip the inter coded enhancement macroblocks. Through extensive tests it is shown that while separate optimization of SNR coders is inferior to single layer coder by more than 2 dB, with joint optimization this gap is reduced to 0.3{\textendash}0.5 dB. We have shown that through joint optimization quality of the base layer video is also improved over the separate optimization. It is also shown that spatial scalability like SNR scalability does benefit from joint optimization, though not being able to exploit the relation between the quantizer step sizes. The amount of improvement depends on the interpolation artifacts of upsampled base-layer and the residual quantization distortion of this layer. Hence, the degree of improvement depends on image contents as well as the bit rate budget. Simulation results show that joint optimization of spatial scalable coders is about 0.51 dB inferior to the single layer optimized coder, where its separate optimization counterpart like SNR scalability is more than 2 dB worse.