[Doc]

yeah, I agree with the point as quoted by Scottie

Also, as you say - you just can't play scared. 14th of 21 players left, I'm guessing that you probably had in the region of 10-20BB left in your stack. When you're in that area you're virtually in all-in or fold territory. Obviously you far prefer being the person moving rather than calling, but there are many big factors here that should make you call:
1) The player is very aggressive
2) It was folded to him in the SB, so he is facing a random hand heads up - with fold equity almost any hand is profitable here on a move all-in for him against most players
3) You have an excellent hand
4) He is out of position and therefore moving now gives him the greatest chance of winning the pot
5) You say you often tighten up at this stage - is it possible that he has picked up on this - or observed you reluctant to defend your blinds?
6) If you're as short stacked as I suspect, you may not get a better spot AND chips to do something with it

For whatever reason, Mickey didn't want me linking to the articles on short-handed play I wrote - but I'd suggest you seek them out and consider the figures in them.

You could also do worse than look at some poker tools:
Poker Calculator
S&G Calculator

Using the former with what I'd say are his likely hands (Any Pair, Any Ace, any suited King, any picture cards) you're a 55% favourite. In fact you are probably further ahead than that - most players wouldn't move all-in with Aces or Kings there - looking instead to extract more money from you. Equally there may be other, less-premium, hands that they'd also play.

If the BB is X% of your stack (before the hand), then you're basically wagering (1-X) to win twice your stack. That is odds of 2/(1-X).
With 55% chance of winning that gives you an expectation of:
1.1/(1-X).
That equates to an edge of (X+0.1)/(1-X)
If the big blind is 5% of your stack, then you have an edge of 15.8%
If it is 10% of your stack, then you have an edge of 21%
If it is 15% then you have an edge of 29% etc

Those are huge edges to be passing up, especially when you factor in the payout differential between the higher places and the lower ones.