Setting design is something I struggled with for years. Overall, I discovered that it could be a lot of fun creating a complex setting full of rich detail, many fleshed-out nations, complicated geography, and everything of that sort, long before you ever run a campaign there. However, in my own experience trying to actually create a campaign and run adventures in such a setting is about as fun and easy as pulling out your own teeth with a plastic spork... I think this happened to me because looking at the setting from such a top-down perspective really hurts my ability to limit myself to the necessary small details needed to make an individual campaign work. It results in adventures that are trying to introduce the grand setting and its complexities rather than be fun in their own right.
After going through that trauma, I have turned more towards Hussar's way of thinking. The important thing is to create a campaign first, and build a setting to fit the campaign. More than it is a world map or list of organizations and NPCs, a setting is a giant mass of adventures, plot hooks, and possibilities. The more you define everything from the start, the more you close off possibilities, so you should delay as long as possible before setting anything in stone. You don't need the setting to fully crystalize until the PCs are high level and traveling far and wide.
So, here is my setting creation process:
1) Create a theme or core idea for a campaign.
2) Brainstorm about possibilities of implementing the core theme, consequences of the theme, and anything that seems like it would be fun and would fit. Characters, adventures, places... anything works, from a guy running some tavern or a funny orc to grand cosmologies and pantheons. Write these all down in a big list. Let this happen over at least a few days or weeks, or even several months. The longer you gather material without solidifying anything, the better.
3) Pick out the things that need to happen first in order to get the characters involved with the theme from the beginning, and turn those into the starting point for the campaign.
4) Pick out my favorite bits from various ideas, and organize them into logical clumps that will eventually become scenes, organizations, locations, and adventures.
5) Map out a rough progression of how the campaign will progress from the starting point towards the scenes, adventures, and locations that you want to put in the setting. At this point, geography only comes in if it is absolutely mandated by an idea.
6) Start coming up with adventures specifically geared towards the first few levels of the game. You don't need anything detailed, since these may never actually happen, but you want there to be a cool possible adventure waiting around every corner.
7) Turn that collection of adventures into the basic geography of the land surrounding the game's starting point, so every adventure has a general location and every location has an adventure. Again, you don't need anything detailed, but a large volume of ideas to fall back on helps.
8) Find some players. Get them to make elaborate characters with detailed backstories and strong motivations. I like dedicating the entire first session of a campaign to character building and setting discussion.
9) Even if you have to throw everything you just did out, match the setting to the players' backstories and motivations. Trust me, it is better this way.
10) Begin the campaign.
11) Continue to brainstorm and write down ideas as you go, and integrate them into the setting and adventures whenever you think it is appropriate. Don't let anything you thought you liked before limit you, unless the PCs have already encountered it in game (and even then remember to distinguish what you thought they encountered and what they actually encountered).
12) As the party completes adventures, gets interested in plot hooks, and explores the world, slowly map things out in greater detail and let the world write itself.
I will note that I consider steps 2, 9, and 11 to be particularly important.
Of course, if you want something quicker and easier than that, you can always just start up a game of
Sid Meier's Civilization IV, play until you reach an interesting point, and then stop and record the details of the world geography, political borders, resource layout, state of affairs, relative technology, and the like, and then turn that data into a setting, choosing the starting place randomly. Then come up with something totally random for the first adventure and completely wing it from there. I don't think I could pull it off myself, but I imagine it would be a lot of fun.
