When I play two videos e.g. side by side, the summed up volume sometimes ends up clipping.
What I normally do, is lowering the volume of both the audio tracks (which have the same loudness), say of 1dB, and that usually sounds good enough.
But, is there a formula to know exactly what is the volume reduction I need to apply on the tracks, to keep the same loudness of the individual audio tracks?
Video editing software usually measures amplitude not loudness on its metering, so 0dbFS (decibels to Full Scale) is the largest amplitude it can play back without clipping, averaged over a few samples (I think).
Different sounds have different perceived loudness, depending on where the energy is in the frequency spectrum, because our ears are more sensitive to some frequencies than others (babies crying at about 1kHz = very sensitive, old TVs whistling at 15kHz = less sensitive).
Generally I would give myself some headroom to avoid clipping when combining material, so make sure none of the speech is peaking above 5 or 6dB, and any commercial music files I usually immediately drop by 15dB or so, as most commercial music is mastered to peak very close near 0dB.
I usually assume somewhere between 3-6 dB drop if I've got two sources that are going to be combined, but it totally depends on the source material.
There is a way of measuring loudness (in LUFS) in Premiere, but just as you wouldn't drive a car just by watching the speedometer, you shouldn't balance audio just by watching the meters... The end product goes into someone's ears, so you need to judge it by listening.
Finally, if you are delivering your video for broadcast, not the web, most companies have very specific rules about audio levels, clipping, and loudness - check them before delivering.
