Lets say I create a codec for broadcasts which contain totally random information at periodic intervals which will be ignored by my decoder. Now if I replace this random information with some more detail about the picture, but using less data than the random data that was injected in periodic intervals, theoretically I have produced better quality with lower bitrate!!
Taking a more practical outlook, I am sure out of the many tens of codec and decoders that are used out there today, I am sure some codecs could be very smart in compressing video data with better quality at lower bitrate.
So, its all in your codec and decoder!