Why aren't the lengths of day and night exactly the same at the equinoxes?

I hate these questions

;>

Seriously, the actual 'equinox' date falls a few days before the actual astronomical equinox. I haven't really thought about this issue very carefully, but atmospheric refraction is probably one effect contributing to this. I can't really think off hand what the other contributing factors are. If I come across a good, clear explanation, I will post it here.