Original cataloging over time
Source of cataloging asks who made each record's metadata. This page asks when and (for the subset of records a library made itself) those two questions together are the closest this data comes to measuring a library's own cataloging output.
The time comes from MARC 008/00-05, date entered on file: the day the record was
created in the holding library's system. On its own that is a record-arrival clock;
for copy cataloging it dates the copying, not the cataloging. But where 040 $a
credits the library holding the record, arrival and authorship are the same event.
Three things to hold onto before reading anything below:
- "Credited to this library" is not the same as "cataloged by this library." The
attribution comes from a hand-curated map of agency codes to
members, and a library cataloging in OCLC
Connexion often ends up with someone else's code in
$a. Every figure here is a floor: the map holds only codes somebody has confirmed, so a library's retired or unrecorded codes go uncounted, and adding one raises its line. Nothing here distinguishes "did little original cataloging" from "has codes we don't know yet". - A reload restamps every record it touches. The tallest years are almost always system migrations and retrospective-conversion projects, not bursts of cataloging. The date is only ever as old as the last system the records passed through.
- Volume of intake is not effort. A year of large vendor loads looks like a productive year until you divide through, which is what the second chart does.
How much, and when
One panel per library, each on its own vertical scale. Compare shapes across panels, not heights; the axis labels carry the magnitudes. Hover over a year for the records credited, the total that arrived, and the share.
As a share of what arrived
The same records divided through by everything that entered that year. This is the comparable view: it removes both collection size and the sheer volume of a load year, so a rise here means the library really was cataloging more of what it took in, not merely taking in more.
Because a share is bounded and already size-independent, these panels share one vertical scale, unlike the ones above: heights mean the same thing everywhere and can be read across. Each panel greys in every other library, so you get the cross-library comparison without the six-way line tangle.
What is left out
008, an
unparseable one, the placeholder 000000 (which is not the year 2000, and would
distort it badly if taken literally), or a year before 1966,
when MARC did not yet exist. Years holding too few records to publish are folded in
with them, so that a suppressed year cannot be recovered by subtraction.