Multiband (read: multifrequency) compression arose from FM radio; in addition to leveling out the antenna-bound signal, it sounds better on cheap-o radios (or rather, the cheap-o speakers in the radios). Most of the music we record has it's energy in the low end (a computer graphic is good at portraying this), and your alarm clock and/or Ford Pinto will violently explode if it were....
I'm not answering your question. Multiband compression sounds "slick." You hear the changes in lows all the time, highs most of the time, and mids usually get lost. There's also a "presence" band (1,000Hz to 6,000, thereabouts) that is usually compressed, too. This happens with regular mastering, but different artists (and mixing houses) will sound different.
And radio is all about flow and "branding." You want everything to sound like it's been mastered at the same place. That's one of the reasons why radio (FM especially) will run an already mastered song through multiband compression. Plus, the DJ and ad spots now have the same presence.
There are some other reasons why this happened (in addition to the speaker thing, too.) You listen to radio in the car, and cars are all low-end rumble. The radio signal has to be both modulated and excited before hitting the airwaves, and compression helps with this (especially DC bias, nowadays). I would also think that compressing each band gets the noise-to-signal ratio much lower, which is helpful towards the perimeter of a signal's reach.
But mostly, it is the "slick" factor and the concept of branding (style of station).
This help?
p.s. Check out the Cool Edit Pro Forum. We're reverse engineering a "Prism II FM" multiband compressor/gate/limiter that we use at work (rock station) and creating a script (macro) that will do the same thing.
The T.C.Finalizer is used by stations on a budget. ORBAN is probably most popular overall, with the OptiMod platform. I don't remember who makes the processor we're playing with, but it is circa 1995 and sounds good to me!