Here is a quick guide to help folks when starting a new campaign as a DM:

1) Set max number of players:  Is this a closed campaign or an open PUG?

2) Set starting level:  Level 1 start is tough b/c everything can one-shot a player.

Level 3 is a good starting point for low level PC's

Level 5 means that PC's have built up some amount of reputation

3) Set Character Gen. Rules

     (a) Gestalt?

     (b) Roll stats or points:  30 is normal, 40 allows for Multi-ability dependent characters

     (c) Allowable races:  Racial ECL should not exceed starting level +1.  For example, if you have a level 2 start, don't allow a race with 3 racial HD and a Level Adjustment of +1.  With one character level, this is a level 5 character.

     (d) Will you allow characters to buy down ECL?   For example, an Aasimar has a Level Adjustment of +1.  This means they cannot be level 20....ever.  Unearthed Arcana has a vareit rule where at level 6, you give up 6,000 XP (don't advance to level 7) to "buy off" that +1.  At that point, you can continue to advance as normal.  A level 20 character can buy off a total of +3 (Zal may want to explain this one a bit more).

     (e) Starting gear/Wealth:  DMG has a table for average wealth by level.  For more cash, you let characters roll their starting gold into that, or let them have max starting gold.  Also, what limits will you put on gear?  Will you allow characters to invest everything, and let them have a Bag of Holding to carry everything?  Will you allow them to start with the ring of not eating and sleeping?

     (f) Feats/Homebrew:  Will you change/disallow any feats?  This is usually done on a case by case basis.  Magazine feats are usually excluded.

I hope this helps lots.  See you all in the world.

On this note, there are a couple of things I want to add to Paladin's post:

When I make a campaign, I have a story or goal in mind. I always start there. From the rough idea, I try to flesh it out a little bit: who is it effecting, how would it ideally go down, and how does a party of adventurers fit in. I also begin to flesh out the world itself: what else is going on in the world, are they going to conflict, or just add more stuff to do,

Paladin went over the character aspect of a new campaign. I just wanted to point out that the character build options are generally one of the last things I think about when doing a campaign (generally thought of on the fly the day of :D)

The thread is titled "when starting a campaign", but my comments are more in line with "building" a campaign so, if anyone wants to try that, here goes:

1) Start with a general theme or setting
Mark Rosewater once said that "restrictions breed creativity", and that certainly applies to world building.  If I try to "make a D&D campaign" then I just stare at the whitespace until I narrow it down.  But if I want to make a campaign that is "thematically based on dragons" then my brain starts firing away.  That one, vague restriction is easy to work with because there's so much design space, and any attempt to narrow it down can be developed into an encounter.  (There are a million ways that something can be "based on dragons"--so that's a million possible encounters which, given that simple restriction, are surprisingly easy to think up.)

2) Build a map of the starting country
Start with just the country.  It will give you a sense of scale and more restrictions that breed more creativity.  However, I'll often make a map, look at it, be inspired to run a certain adventer (chapter) that can't be done on the current map, and then change the map to fit the adventure.  Eventually, you'll have a good map with lots of good adventures.  Don't make your map too big; you're just aimlessly increasing your workload. "A map is only important if players go to the locations shown on it." ~4th edition DMG.

3) Plan major events, not a pre-programmed plotline
A lot of DMs will first come up with a scripted linear story and will then script actions for players (or the party) within the story.  I think this is how fantasy books are written.  However, game players will not follow your scripted actions by some magical coincidence, so you'll have to railroad them pretty hard in order to advance your scripted story.  I think a better approach is to plan a series of events that happen, and players can respond to those events however they want.  This allows the DM to create "parallel" content (which is  usable even if the players don't do what I expect), whereas scripting different stories for different party decisions is basically impossible.  This worked very well on the first day of the Legacy campaign, where  the life of the crown prince--and the lives of the players--were threatened by the prince's own foolishness.  The players had to deal with the threat, while deciding how much energy they wanted to put towards saving the prince.  However, it would have been bad game design for me to put this encounter to the players with the prince's death or survival already scripted.  (Do that often enough and players will feel like they're just watching you play D&D with yourself.)  So I ran the battle with the water elemental without knowing if the prince was going to survive or not (a pretty cool experience for the DM).  If the prince did survive then he would help the party later.  If he's sucked into the vortex and drowns, that's obviously the end of him.  Parallel content for the win.

There's actually a section in the DMG about building a campaign world. To add to Opal's ideas about starting with a country, it mentions that there are two main ways to design maps and areas: Starting micro and starting macro. Some people find it easier to start with countries and kingdoms, then break it down as necessary and as things develop from there, others find it easier to start with a farm or shire, plot a couple major landmarks, but then fill in the local areas as the party explores. The latter involves the kind of "Fog of war" that most games include as an exploration mechanic when maps are not readily available.

I'm sure i have more to say on this, but I'm too tired to think of it right now. Will check back later

-Z