Chris Perkins has a column on the WotC website called The Dungeon Master Experience. It's by far the best column that the website currently has (if you're a DM, that is) and this week's pertains to something I've been thinking about lately: the balance between a "sandbox" style and a "railroaded" campaign. Most campaigns occupy a middle ground between the two, and I think that's best overall. From a player's perspective (I enjoy that I play almost as often as I DM) too much railroad makes me feel like my character's actions don't matter, and too much sandbox can make the whole venture seem aimless (not to mention the fact that it could easily catch the DM off guard, and I don't care how good of a DM you are, you're not going to be running as optimally if you're trying to improvise a random direction as opposed to something you've planned, or at least considered).
I have a feeling that most DMs think about this spectrum in relation to the current adventure, because after all the players are going to react to things happening in the present. It's tougher, however, to compromise throughout the span of an entire campaign arc (at least I think so). Generally it's either unplanned and the DM just goes with the flow (often leading to a string of semi-related adventures) or the DM has a good idea of where he wants the campaign to go. Planning for multiple (in this case, three) different arcs gives the DM focus because he's thought about the possible arcs and in general knows where they're headed (he just has no idea when and in what order the players will tackle them in), but it also gives the players a lot of choice in what they want to pursue because the DM has thrown several well-thought options out there. Obviously this will work best if the DM allows even the arcs that weren't pursued to "progress," perhaps making the situation "worse" and providing the PCs with more and more difficult choices.
Anways, here is a link to the article. It's a good read.