Sean has it right. Every project is a bit different, maybe every hole, and the calculation varies. No doubt that all factors being equal, as David M say, you should route it to walk it. But, as Sean and Brent say, sometimes that just isn't possible, given either parcel shape or subsurface rock, wetlands, etc.
As to Ian's point, I recall that exact discussion on the second hole at Cowboys. We considered a low tee, but then took it up the hill for that snapshot view. It was going to be 100% carts, it had to be one of the most dramatic drops in the DFW market (outside Dallas National) and would be unique in that way in taking the most advantage of the landscape. So, was it wrong to consider it in terms of regional context, possible revenue enhancer for the Owner? (You know, Jerry really needs that money!)
I also recall another project where we had 400 acres to choose from, and my associate laid holes over the entire property. A quick tally of the extra irrigation main and cart path showed the extra expense of a loose routing, and we found similar holes closer in to save money. As a lower end public course, walking was a larger component among the cost conscious customers. So, staying as tight as possible has many natural benefits.
Lastly, the course I played most as a kid had a few walk arounds to get from one tee to another, along with consecutive par 5 holes and an 18th hole par 3 (Old Orchard, Mt. Prospect, IL, since changed, reduced, etc.) The designer part of me always thought some of that was cool, because it seemed to break the rules about going right to the next tee. In one case, the crossover also brought you closer to the rest room/snack bar.
Over time, seeing so many golfers get lost (in the CCFAD era, so many golfers played each new course once a year and weren't as familiar as a club course might be) sort of knocked the idea of crossovers out of my mind as a good idea. I could see why the "standard" of an short trip, easy to find next tee made loads of sense.