Collection Overlap Analysis

Both summarize shared holdings, but they normalize differently. The overlap (Szymkiewicz-Simpson) coefficient divides the intersection by the size of the smaller collection, so it measures how completely the smaller collection is contained in the larger; it can reach 1 even when the larger collection is far bigger. The Jaccard coefficient divides the intersection by the union, measuring shared titles relative to combined titles, so it stays low when collections differ greatly in size. For asymmetric comparisons, report whichever matches your question and state which you used.

Every overlap and uniqueness figure depends on correctly deciding whether two records describe the same title. Matching on stable identifiers such as ISBN, ISSN, or OCLC control number is far more reliable than matching on title and author strings, which break on punctuation, transliteration, edition statements, and format differences. Poor matching either misses true matches (inflating uniqueness) or merges distinct editions (inflating overlap). Because retention and weeding decisions hinge on these counts, investing in clean, identifier-based deduplication is the single most important quality step.

Not on its own. High overlap shows that other collections hold the same titles, but a responsible decision also needs use data and a group-wide count of retained copies. Shared-print programs typically commit to keeping a minimum number of copies of each title so that overlap does not silently erode the last accessible copies. Overlap analysis should therefore feed into a retention framework that weighs duplication against demand, access guarantees, and the risk of losing the only remaining copy, rather than driving weeding directly.