I came across an interesting blogger, Jim from Jim's Embeddedtronics. There are lots of interesting posts in here, but I wanted to highlight one.  He has a machine to measure torque of a stepper motor at speed and he ran a series of tests on micro stepping.  

The end result is simple enough, micro stepping does not reduce the torque of  your stepper motor but rather it helps smooth out the motion of the motor.  I had run a somewhat similar test a while back and came to the same conclusion, but it is nice to see it verified with better equipment than a postal scale.

There is a lot of confusion about this online with people assuming that if you run 1/16 micro stepping you end up with 1/16 the holding torque.  The truth is that when the stepper is on a full step the holding torque is at full force.  And when the stepper is half way between poles (half step) the torque is reduced by 30% or so.  However as you push the rotor closer to a pole (full step) the holding force increases.  So micro stepping can't necessarily hold position accurately between the poles but it can't reduce the accuracy any lower than without micro stepping.  

Basically it is a win-win, when the motor is not pushed to its limit it increases accuracy and smooths out the motion.  But when the motor is pushed to its limit it works as good as if you had no micro stepping at all.  Basically, as usual, whatever you read online is probably wrong, to some extent.