This happens when the Pan law in Preferences>Multitrack isn't set to Left/Right Cut (logarithmic). With the default -4.5dB cut setting (which is more use for a lot of things), you'll get a quieter mixdown, certainly.
But, I have to say that it really doesn't matter in the slightest if you're mixing in 32-bit (which you should be) - all you have to do to get your signal back is normalize the resultant file to whatever level you want. No quality loss or anything like that at all - you just end up with a far more precise final level anyway. This also applies if you've apparently overloaded the output and it goes beyond 0dB - the Floating Point mix can recover your mix correctly from either direction. Yeah, I know it sounds like magic, but it isn't really; it's all to do with scaling. Your sound device, because it has to work in the 'real' world, can't handle the Floating Point signal directly, so an overloaded mix will sound overloaded/clipped/distorted - but that's because it's been converted back to an integer signal, and they're the ones that can overload. But in Audition's virtual world, that's not a limitation at all...
Dear Steve, thank you very much for the helpful, fast reply. I appreciate your extended response as well:)) Oh yes... Adobe, you were so nice to put this option by default--so unlogical decision... alright.
And it will be always better for mix, if in the first picture you chose a 32 float instead of 16?
So, you mean Steve, I should always mix in 32 bit, even if the final result I want should be in 16? And it will be always better for mix, if in the first picture you chose a 32 float instead of 16?
Absolutely it will! Audition's mix engine runs natively in 32-bit Floating Point mode anyway - if you choose to save the result in any integer format at all (ie 16-bit), then you are throwing away the advantage of being able to rescale the result indefinitely.
So, even though the master takes up twice as much room, you've got two advantages. The first one I've already mentioned, and the second one is that you have an undithered master available. So, if you want to produce copies in other formats, like MP3 for instance, you don't have the dither applicable to one format applied in advance, and you're giving the encoder the best signal you can.
And the two can be combined as well - if you want, for instance, (as you really should) to make an MP3 that doesn't sound quite so bad, you can temporarily renormalize your master to -2dB to create it, and then normalize it back again to wherever you had it before. Or just leave it like that and save it, as you can always restore it next time you open it. And there's the rub - if you did this with a 16-bit integer file, you'd lose one bit of resolution permanently if you saved it without restoring it. No problemo with 32-bit FP...