#p versus e

[ebookevaluation-blog]

Which subjects more popular in print, which in electronic?

Christianson and Aucoin took the top 10% of most-used titles in each format, then collected subject headings for these, then tallied which subjects came up most often in each format. Would make more sense to count every time a book was used, rather than counting each book once. They also give a caveat that a subject might not be high in the usage stats because they don’t have many books on that subject. But it was a paired list of titles, so a subject would be equally non-represented in both formats.

Slater 2009, not surprisingly, finds history and language/linguistics favor print, but also education and biology. Sciences, technology, math, and engineering preferred online, as well as media and communications. This is by comparing PEU between netLibrary and print collections.

Slater 2010 says we should not assume differences in print and electronic use are due to how users are using a book (i.e., print for extensive reading, e-books for quick look-up). There’s a bit of research showing print usage isn’t always extensive either.

Slater 2010 says if the differences between subjects are due to different users and their ability to adopt new technologies, we would expect the differences between disciplines to decline over time. But if the differences are due to different information needs in different disciplines, the disparity would remain.

[ebookevaluation-blog]

problems with circulation as a measurement

in-house use of print does not get counted, whereas a similar activity for an e-book would (i.e. flipping through to decide if you want to read it). Post-checkout you might use a book a lot and it only counts as one use, whereas for e-books each time you open it gets counted.

Different checkout periods for different users really mess up the stats as a true measure of use.

Remember to check if a book has been out for years. Circulation may be low because the book is actually not available.

Slater 2010 cites a study from 1986 with interviews of how people use print books. Their study found that 60% of interviewees said they read 10% or less of a book they check out, and 77% said they read “small parts” of books. He also mentions than Christianson and Aucoin where among the first to question the assumption that a print checkout implies extensive reading. It might just mean skimming or reading bits.

[ebookevaluation-blog]

How to compare usage p v. e

Kimball et all points out we can’t compare checkouts to section views, of course, since these measure different things. They use a study by Michael Levine-Clark that shows that most users say they read only a chapter at a time, or less than a full chapter. Likewise, most print checkouts also do not imply cover-to-cover reading. But e-book use would count if someone flipped through a book and decided not to use it; while for print this would only be in-house use and hence not measured. Kimball et all concludes he might as well compare checkouts to e-book accesses (section views?) because both are equally incomplete measures. They say it is “the traditional comparison” without citing who has used this measure p.20.

Keep in mind also, that a platform that limits simultaneous users is going to have lower e-book usage than one that allows unlimited simultaneous users. Counting simply the number of books used gets around this problem.

Sullivan and Leach start with the premise that you can’t compare amount of use. They will compare whether titles were used or not and divide books into “PnotE, EnotP, Both, Neither.” Also additional variables to consider: subject classification, year, publisher, edited vol versus monograph.

They found that 10% of books in their dataset were used in print but not online, while 26% were used online but not in print. The question then is, what is different about the books that were used in one format and not the other? How to measure this?

Knowlton says it is “nearly impossible to reliably compare print circulation to e-book usage.” p.21 This is true, and a relief.

Knowlton did a chi-square test comparing difference between books used and not use. (I don’t actually understand what numbers he compared, but essentially the difference between the amount of use within the “used” group is not statistically significant and the difference between “used” and “non-used” is.) at a significance of p<.001.

Another nice contribution by Knowlton is Percentage of Expected Use. This gets as the extensiveness of the use of each subject, as it takes into account the number of uses as well as the number of available titles, and it is comparable between print and electronic. He mentions Aguilar and Ochola as people who explain this methodology. (I have both of these articles elsewhere.) It would you History makes up X% of our collection and accounts for X% of use. But how on earth do you get the subject classifications for your entire collection?! You can’t. Maybe triangulate and look only at ebrary Academic Complete and then also look at a history-specific collection in some other way, like does it get more usage per book than the history titles in ebrary?

Slater 2009 also compares the PEU for print to PEU for e-books, by subject. Rather than looking at the differences between them, he looks at correlation between the two and actually finds a positive correlation (p=.025, r=.475). Meaning subjects with heavy usage in print also have heavy usage in e-books. It’s a single independent variable regression, I think.

Slater also does look at where the biggest differences are between relative use (PEU) of e-books and print. He, not surprisingly, finds history and language/linguistics favor print, but also education and biology. Sciences, technology, math, and engineering preferred online, as well as media and communications.

You want to be able to compare print usage for a specific year. This isn’t possible for us in Millennium, because of how our circulation statistics are stored, but I’ve love to have JO’s data!

Littman and Connaway also just look at whether titles were used at all. According to Cox, they may have been the first to suggest this strategy.

could you look to see if there is a negative correlation of print book usage (falling) to e-book usage (rising)? It seems like you’d need to look at the entire collection of both in order to test this. And also, the question of whether they are growing at different rates. Amy Fry posits that if print use is declining it may be because you’re buying fewer new books. People are more likely to use new books than older ones.

It seems like Amy Fry had records of in-house use.

[ebookevaluation-blog]

How to find comparable groups of print and electronic

Kimball seems to use the package’s own subject classification (e.g., “sorted by ebrary categories p.21) for e-books and LC call numbers for print books, but they do not say what call numbers they used.

I think it’s common to focus on a particular collection for which you do have duplication and try to extrapolate from that. Harvard (Sullivan & Leach) had a one-time fund that allowed them to purchase Project MUSE 2014 collection in print and online. This was 29,700 books. Or perhaps the one-time finding was for online only, and they happened to already own most of the titles in print? The overlap is not complete, but substantial (22,701 titles).

Duke e-books have been used in several studies because of their purchase model. Goodwin explains: an annual cost for all of their e-books released in a given year, with an add-on cost to get the print. So libraries are likely to have the exact same titles in both formats. (Be aware, though, that users might have different format preferences for university press books than other kinds.)

If your library’s purchasing has shifted towards electronic, you cannot just compare all you print books to all your e-books, as you may be lacking in newer print books. (You probably will still have older e-books, though).

Sullivan & Leach do a nice check where they look at how usage of the print titles that are also in Project MUSE compares to usage of a control group (presumably the collection as a whole). I think this shows whether print usage is going down in general versus going down specifically because the books are available online.

Knowlton says most studies look at collections where the exact same titles are held in both formats. This is clearly easier to do, but it will encompass only a small part of the library’s collection.

A lot of e-book packages contain backlist titles that are not as likely to get used. Better to limit your study to titles with a publication date equal to the year you bought the books (Knowlton). Obviously exclude print books that don’t circulate, such as reference or special collections.

Littman and Silipigni Connaway used netLibrary books that they already had in print.

Slater 2009 used “several hundred” books they had in print and online

Slater cites a study by Stephens, Melgoza and Wan, which took the Safari collection (a subject-specific collection) and compared it to print books with the same range of publication dates and with call numbers in QA75-76.9 (computer science). I like this idea.

Amy Fry looks at all the books acquired within the same time period. How does she account for e-books that were already old at the time acquired? She specifically says her method differed from Knowlton’s in that he looked at publication year rather than date of acquisition. She looks at books acquired in a given time period, which seems like it would exacerbate the problem of e-book collections of backlists.