That would depend on the codec you choose to encode.
As an extremely simplified explanation:
When you render an image, Blender generates the information for Red Green and Blue and Alpha in 32bit float. Which is a lot of information. That image is not compressed in any way and all color information is available before processing to be encoded and saved.
Why not save then all of that information?
Not only it would take a lot of disk space, but you would need an extremely powerful device to play that amount of information in real time and sustain uninterrupted video streaming.
Compression schemes or codecs were invented to deal with the problem. Codecs allow for smaller files and facilitate playback at a steady framerate.
The efficiency and quality of the pictures is then determined by the codec used in the compression. Compression means that some information has to be thrown out.
In other words: when you choose a codec, you are choosing a way to discard information from an image.
One of the many ways is by using less information on the red and blue channels of the image than the green channel.
From the Wikipedia:
Chroma subsampling is the practice of encoding images by implementing less resolution for chroma information than for luma information, taking advantage of the human visual system's lower acuity for color differences than for luminance.
Different codecs use different ways to do chroma subsampling
For example, ProRes can have many flavors: 422, or 4444. ProRes422-variants use Chroma subsampling of 4:2:2 at 10 Bit Color depth. ProRes 4444 samples color in the 4:4:4 schema with a color depth of 12 Bit. Most DnxHD codecs use 422 but the ones with the higher bit depth will use 444. Different variants of Mpeg4 or H.265 will use 422 or 444.