Two reasons. First, by putting the photons from four pixels into one pixel, you essentially increase your signal to noise ratio. Second, it was thought that since all of your detail comes from the Luminance layer, it's okay if the RGB chrominance layer is lower-res, even a bit blurry. That assumes you are taking separate luminance at full 1x1 binning and combining later. If you shoot just RGB and extract L, you'd want full res.
I suppose a third reason might be file size and memory concerns, but that's hardly an issue these days with huge, cheap HDDs.
That's the common explanation, but I shoot everything full res.