I would think, too, it has a lot to do with system requirements for the O/S. It takes a lot of cpu resources, just for a computer to sit idle these days.
This. And only this, really. The DAWs have to run on newer OSes, which, unless the maintainers are bordering on insane, usually means discontinuing support for older OSes. The newer OSes have higher overhead than older ones, which means you need a faster CPU to avoid the OS's overhead getting in the way at inopportune times.
And audio is one of the hardest workloads you can come up with in terms of how it affects the OS's scheduler. It requires that the disk I/O be scheduled far enough in advance that the data is ready when you need it, and requires that the I/O to the audio hardware occurs within a fairly narrow window of time—after the audio hardware has read the previous data at that spot in its ring buffer, and before it wraps around to that spot on the next pass through the buffer—over and over without ever writing data too quickly or too slowly. And it is more demanding than most other difficult tasks. Tiny glitches in video playback are usually tolerated, but tiny errors in audio will drive you bonkers. Not to mention that video capture is trivial compared with audio because there is rarely any sort of latency concern, so the ring buffers can be much larger, with sloppier timing.
The worst offenders are drivers that hold interrupts off for long periods of time (some hardware SD card controllers, non-USB floppy drives, etc.), consume large bursts of CPU (USB mass storage in general), produce large numbers of interrupts (graphics chips), or some combination of the above (many Wi-Fi drivers, IIRC)—particularly when those devices share an interrupt with your audio interface. In the presence of such abuse, a faster CPU can often make the difference between smooth recording/playback and glitches by simply being so fast that the poor performance of those drivers doesn't matter.
So yeah, operating systems and, in particular, crappy drivers.
Oh, and more powerful plug-ins available in newer DAWs, of course, but those only matter if you use them. When you kick in stuff like convolution reverb, that sucks down serious CPU power. But I'm assuming the question was about base requirements, which probably shouldn't take into account the higher requirements of people who use insane numbers of high-power plug-ins.
