I can't speak for every field, but at least for my own field of
information systems, where conferences count for zero, at least
among the most research-intensive universities:
Counting conference publications or not is in no way a judgment
either way of the quality of the papers. In information systems, it
is well known that some high-quality conferences (such as ICIS,
HICSS and AoM-OCIS) regularly field higher quality papers than many
journals. However, such publication often counted as zero in
promotion and tenure considerations.
What is going on is that in our field (and I suspect also in similar
fields) *conferences are not considered terminal publication
outlets--only journals are*. That is, when you present a paper in a
conference, even when it is published in the proceedings, you are
expected to later publish a significantly revised and significantly
extended version of that paper in a journal article (and I would
guess that in 90% of the time, this is what happens, at least for
high-quality papers). A high-quality conference paper is expected to
yield a high-quality journal article. Thus, *to avoid
double-counting*, conference publications are ignored in promotion
and tenure considerations.
From what I understand, in fields like computer science where
conferences are terminal publication outlets (that is, conference
papers are often not republished in journals), then it naturally
makes sense that the conference papers should be considered the
measure of a researcher's productive quality.
As for disciplines that do not count conference papers, I
cannot comment because my discipline (Computer Science) looks at
top tier conference publications in a similar way to journal
publications. However, I'd argue that anyone who does not value
a publication purely because the venue is called a "conference"
regardless of the impact/restrictiveness is making a mistake.
I've seen people include the acceptance rates on their CV to
avoid this situation.