... and they reasoned that being the loudest station on the dial would capture more listeners as they tuned past. (Back when radios had tuning knobs.)
But the loudness problem started before that, with programmers' meetings where they auditioned new tracks to add to the playlist. Each track might only get 10 or 20 seconds to make an impact, so the louder it was the better.
Just tossing another carrot in the stew here, but radio puts another level (no pun intended) on it. Radio stations must limit their peaks, so as not to overmodulate and encroach on neighboring stations—by law. So they need a hard limit (clip), and a limiter so that the clip is reached as a last resort. I'm pretty sure most rock stations compress as well, to keep levels hot and listers engages, especially those driving, but broadcast is not my expertise so I'll leave it at that.
A big aspect of the loudness war is making things sound relatively good on crappy equipment, like boom boxes. I've seen Andrew Scheps (Chili Peppers, Adele, Metallica) speak a few times, and he often addresses the loudness wars in an unapologetic way—once he said that he got satisfaction from walking by someone's iPhone with one of his songs playing, and it sounding good.
Also, I should make clear exactly what is meant by "dynamic range" in the sense of something truly squashed due to the loudness war. It's not simply a reduction in dynamics. It's most apparent when you view the waveform of a crushed tune. The overview waveform is often a solid block for the worst offenders, there is no dip at all in the big picture. It's really not exactly the kind of dynamic range choices many are talking about here. And don't get the idea we're only talking about heavy metal here. Here's Schep's mix of Adele's Don't You Remember, that I did a quick capture of:
And, as you can see in the Dynamic Range DB, Adele's album releases on vinyl have a much large DR figure than the same album released on CD. The target audience is considered.