Sorry, guys, I don't have a magic equation and, actually, I don't work much with lights either. I am familiar with the issues, though. The problem with high-THD devices like ballasts and switched DC power supplies is that you can throw your 3-phase neutral theory out the window.

Traditionally, the current on a balanced 3-phase circuit will cancel completely, leaving virtually no current on the neutral. So, we plan only for the unbalanced load, sometimes even putting in smaller neutrals. Switched power supplies, however, work differently. They basically lop off the peak of each phase in turn, and turn our nice pretty sine waves into square waves. So instead of ending up with sin-wave shaped current on the neutral, we end up with a giant square wave that's all peak and no trough. (If you looked at the neutral current on an unfiltered power supply, you'd see it as a 180Hz square wave on the neutral.) Ampacities are rated by RMS current, and rule-of-tumb power RMS assumes a sine wave. For square waves, however, RMS = peak, and is about 1.4x higher. There's some other voodoo goes into it, but the max theoretical current on a 3-phase neutral is somewhere around 167% IIRC.

We call the artifacts from this harmonics, since the net effect is a combination of waves at different frequencies. You can see it pretty clearly here as you add in prime frequency plus 3rd harmonic, 5th, 7th, etc, up to 25th, and all those harmonics added together turn a sine wave into a square wave. (This is mathematical, btw, and includes 2nd, 4th, etc, which aren't present on 3-phase systems, but the net is the same.)

http://en.wikipedia.org/wiki/Square_wave
[Linked Image from upload.wikimedia.org]

If these were computer racks with filtered switched DC power supplies, I'd convert watts to amps using the manufacturer's (or measured) power factor and call it a day. The filters on modern server power supplies *usually* compensate for the harmonics to the point where it's still a reasouable facscimile of a sine wave going out the back and neutral current is less than line current (which makes doubling the neutral unnecessary). If the ballasts are unfiltered, though, the easiest way to deal with it is to double the neutral. Make sure to account for the neutral when calculating voltage drop, too.