I was interested to read the following in a report just released by the Research Information Network in the UK about the completeness of catalogue coverage of research collections :
The study shows significant progress: librarians estimate that 50% of material in their research collections is now covered by online catalogues, compared with 31% five years ago. But much more remains to be done before all the significant material held in UK libraries that may be of value to researchers can be readily traced through online catalogues. Librarians are keen to pursue this work, and we recommend it should remain a high priority for them. [Uncovering Hidden Resources:Progress in extending the coverage of online catalogues - PDF]
This seemed high to me on a first read - before I realised that the survey was restricted to research collections. How do they define research collection?:
Respondents were asked how many special collections for scholarly research their library holds in total. It should be noted that no attempt was made to define the term special collections for scholarly research, so the responses received were based on the interpretation of the responding librarians. In some cases, respondents considered their library’s entire stock to be a special collection, and hence the number of collections held by the responding libraries was found to vary considerably, ranging from 1 to 5000. It should be noted that many of the figures supplied were approximate and that no indication of the size of collections may be inferred from these data. The most common numbers (mode) of collections held were one or three, and the median number of collections was found to be six. PDF]
The survey was based on responses from 96 libraries. Over half of these were academic, and the balance was divided between public (26%) and specialist (21%). This represented a response rate of 26% for academic libraries, and 12% for public.
The results are not broken down by type of library, or by size of library. This might have been helpful.