tanlsth, can you please try to fix your solution? Because mine is 3 pages long, after it has undergone a lot of error fixing (apparently it is very easy to make mistakes in this problem), and I still am not as sure of its validity as I usually am when I write down a detailed (unreadable) proof. You can find my proof at
http://www.cip.ifi.lmu.de/~grinberg/subsetcond.pdf ; here are its cornerstones:
Theorem 1. Let
be a set. Let
and
be two nonnegative integers such that
. Let
,
, ...,
be
subsets of
such that
for every
. Then, there exists a subset
of
such that
and
for every
.
Sketch of proof. Case 1 (when the union of the

covers all of

) is very easy (by a counting argument, every set

contains an element which is not contained in any other

, so you get your

), while Case 2 (when there is some

which does not lie in any of the

) is the actual argument, which mainly consists of replacing every

by
and applying induction over

(noticing that

itself collapses to a one-element set, and one-element sets can be ignored).
Ah, and of course, it is obvious that

is indeed the minimum (for

, a counterexample can be easily found).
PS. tanlsth, please tell me your real name in case you want it to be mentioned in the PDF.
darij