We all know how cameras try to meter the scene as if it is 18% gray. That's why we should dial in +1 or even more when shooting on a snowy day and -2 for a night scene.
What about a cloudy or overcast day?
When you google "exposure compensation cloudy day" for example, the vast majority of the posts that come up say to dial in +1/3 to +1 EV. I think this school of thought stems from the underexposure possibility when bright cloudy sky is included in the scene, and should be stated as such. Otherwise, it is misleading advice. (The logic is similar to a snow scene.)
But assuming there is nothing in the scene (like white cloud) that can fool the meter, wouldn't you want to UNDEREXPOSE the scene? When you hold an 18% gray card in front of the camera on a cloudy day and meter, the camera will actually overexpose by trying to match the luminance of a bright sunny day. The result will be a bright, properly exposed picture, but not one representative of a cloudy day. That is why I think downward exposure compensation is necessary to catch the feel of a cloudy day. But I'm puzzled that this practice is not preached and wonder why.
At the end of the day you are the photographer and your photo is an expression of your vision. If you prefer the final image under or overexposed, then that is your choice. As long as you are not a photojournalist, then you are an artist -- do what makes you happy. Journalists should play by stricter rules and should not turn day into night or vice versa, everyone else go forth and have fun.
Guidelines are just that guidelines, they help people who are new, and can and should be ignored by people as they learn more and want to experiment.
All of this having been said, you may find that getting a good (as defined by your histogram) exposure in camera, and then darkening it in post, will provide richer shadows, as more information is available in the captured frame.
The only reason you'd want to dial -x is if you'd otherwise overexpose to the point of saturation or your shutter speed gets too long (motion blur). Otherwise, you want to maximize the signal to noise ratio. You can then produce the correct dark feeling in post, where the negative EC also attenuate the noise floor equally.
This doesnt apply to the style of shooting "out of camera" end results. In that case you are right.
...wouldn't you want to UNDEREXPOSE the scene?
You don't want to underexpose the "scene" ...you want to underexpose what your meter is telling you is the correct exposure for the scene. The problem is that TTL metering is inherently flawed.
Walk up to a white wall that will fill your viewfinder. Take an image at what the camera tells you is the correct exposure. Look at the image. When you find the answer as to why that image is gray, not white, you'll understand the benefit of a hand held incident meter.