In the above publication, we derive, under two different stochastic volatility models, the expression of the variance of the error associate to the use of the squared return as proxy of daily volatility.
On the very last line of p. 256 we have an expression for the expectation of the the square of zt
This is correct if zt had not been standardized. Given that zt is standardized as we describe at the end of Section II, this expectation should be unity (this mistake has been found by Prof. David Giles). On p. 257 it then follows that the expection of et is zero and we have unbiasedness. This, of course, then affects and simplifies the calculation for the variance that follows. In particular, we have that, if SV-t model (M2) holds, the correct formula for the variance of et , is