From the HITEC website:
Q. What is servo deadband?<div id="main_text">A. Deadband reflects the time it takes for a servo to respond when stick movement is given. Standard servos have a deadband of around 8us and high performance servos have a deadband of 1~3us.

This would appear to be a definition of servo latency with values presented for deadband. Other sites seem to have some confusion between deadband and resolution.
Deadband is usually added to keep the servo from dithering as many JR digitals do and has nothing to do with the amount of time it takes for the servo to respond to input change. I still do not understand why this data has to be ferreted out from obscure posting on the manufacturers website. Since the results are so much better for JR as demonstrated in the link provided by Ed, why not sound off more? Certainly there is a lot of lowcost pressure on the top servo companies. It would appear that the problems with the budget servo being compared to the 821 had a lot more to do with backlash than deadband.

I would offer the following definitions.
Deadband: The servo input signal value change from steady-state required to initiate a change in output value
Accuracy: How closely the output signal matches the input signal once the deadband has been exceeded
Latency: The amount of time between a change in the input value and an initial change in the output value. Usually looked at from an overall system point of view, not just in a servo.

John</div>