There is no good answer to that. Many animations are built with G-Buffers so they can be layered in the video editing program. If there is a depth of field thing happening, they asset can be even lower in resolution because it is intentionally blurred in the editor. 

But, as a rule, rendering at half size, and then sizing up will work ok. Push it beyond that and you will notice artifacts in near still scenes. But, if there is motion involved, viewers won't notice it. Motion trumps detail - every time. In fact, super detail is only for stills, never motion.