To the best of my knowledge:
For the standard PrPro output formats, the Audio will be down-sampled to 16-bit. Now, when you Import, PrPro will Conform those files to 48KHz 32-bit floating point (in CFA format) for editing.
Other than from an audio editor, such as Adobe Audition, I think that the 48KHz 16-bit will be as good as it gets.
Thank you both for posting a response. I appreciate the insight Bill - it is good to know that the 24 bit depth is conserved while adding effects. That is good enough reason for me to keep getting the higher quality audio for input. I'm not too concerned about outputing a 16 bit file - that is plenty good for me.
When working with *24bits in a *32 bits floating-point internal environment you have *8 extra bits (these are *wordlengths and have nothing to do with the sample frequency as stated above!) and gives you virtually a endless headroom ( even above 0 db on individual channels).
But at the end (after the master and out of your soundcard that is ***16 or 24 bits) the wordlength gets changed (shortened from 32 to 24 or *** 16 bits) and the sample frequency sampled down eventually as well, the master output has to stay a *little under 0 db *depending of the type of sound.
To answer your question: "It makes perfect sense to work with larger wordlenghts, because no matter you go down to 16 bits at the end, you have a much better sound quality and bigger headroom during mixing and at the end even when it's downsized to 16 bits."
***16 bits in Premiere Pro