We investigate the discrepancy (or balanced coloring) problem for hypergraphs and matrices in arbitrary numbers of colors. We show that the hereditary discrepancy in two different numbers
a
,
b
∈
N
≥
2
a, b \in {\mathbb N} _{\ge 2}
of colors is the same apart from constant factors, i.e.,
\[
herdisc
(
⋅
,
b
)
=
Θ
(
herdisc
(
⋅
,
a
)
)
.
\operatorname {herdisc}(\cdot ,{b}) = \Theta ( \operatorname {herdisc}(\cdot ,{a})).
\]
This contrasts the ordinary discrepancy problem, where no correlation exists in many cases.