Thanks to those who have already sent answers to my question.
However, I think I've found another resolution to the
discrepancy. Sze, in Physics of Semiconductor Devices,
2nd edition, page 17, gives a formula for the effective
density of states in the conduction band (equation 12).
The formula is proportional to (m_de)^(3/2)*M_c, where m_de
is the "density of states effective mass", which he defines as
the geometric mean of the effective masses in the three directions.
The factor M_c is the "number of equivalent minima in the
conduction band". I can't find a place to look up the value
of M_c. However, from the numbers in Appendix H, I find that
you need to use M_c = 5.9 in order to obtain the given number
for the effective density of states. But what if, instead,
we were to absorb M_c into the effective mass? Then, instead
of .33, we would need an effective mass of 1.08, very close
to Kittel's value of 1.06. Therefore I propose that Sze and
Kittel are using two different definitions of the effective
mass. Does this sound plausible to you experts?
(By the way, I note that on page 18, Sze defines an effective
mass for holes that sums over the effects of "light" and "heavy"
holes. This seems similar to absorbing M_c into the effective
mass. Apparently it's conventional to do this for holes but
not for electrons...)