Despite these successes, it became apparent upon reflection that the sociometric measures of ST value are not direct measures of the effect of cancellations on actual library use and the costs of this use. Even if a serial subscription is canceled, the backfile is retained, and the costs in lost use and replacement relate to the future and not to the past.

To gauge these future costs, it was decided to use the accounting technique of a standard cost system based on estimated costs derived from average past experience (Plank and Blensly 1989, 134). The essence of the accounting technique is to normalize the use of serials on the same annual basis as their subscription prices, so that the two measures can be logically employed together. This was accomplished in the following steps. First, UIUC Chemistry Library use was restricted to the post-1980 period to capture current trends, because science tends to concentrate on literature of the more recent period. Some of the serials dated back decades—even to the nineteenth century—and use over the earlier backfiles was estimated to be spotty, making the calculation of averages difficult. Moreover, pre-1980 use could be considered fully depreciated and essentially cost-free from an accounting standpoint. Post-1980 UIUC use accounted for 51,740 (73.8%) of the 70,072 total uses, and the correlation between pre-1980 and post-1980 use was 0.81, reflecting the stability of informetric distributions over time. Second, a use age was calculated by considering all serials predating 1980 to be 14 years old—1980 through 1993—and all post-1980 serials to have a use age equivalent to the number of years in their backfiles. Third, the post-1980 use of the serials was divided by their use age to yield an estimated annual use or an estimate of the use that would be lost during the first year due to cancellation. Finally, the 1993 subscription price of the serials was divided by this estimated annual use to calculate the cost-per-use of the titles.

The creation of these accounting measures enabled us to analyze the cost-per-use structure of the chemistry journals under investigation. For this purpose, we used two methods of summarizing the cost-per-use of a set of serials. One was the mean title cost-per-use. This is simply the average cost-per-use of the titles. However, this method assigns equal weight to those journals with low use but high cost-per-use and therefore overestimates the cost-per-use of a set of serials. The other method was mean document cost-per-use, and this was done by summing up the subscription prices of the serials in a set of serials and then dividing this sum by the total estimated annual use of these serials. This method assigns the proper weight to the serials with high use and low cost-per-use and provides a good statistic to compare against some benchmark.

Overall, the title cost-per-use of the entire universe of the 120 chemistry serials common to the LSU and UIUC data sets ranged from $1.79 to $4,079.27. Their mean title cost-per-use was $247.49, and their mean document cost- per-use was $41.30. Breaking up this universe into sets defined by publisher type made some extremely revealing comparisons possible. In making these comparisons, it was considered necessary to exclude one U.S. association title that proved to be an anomaly. This title was not a U.S. association journal in the usual sense but a translation of a Russian journal published by a U.S. association. With this exclusion, the statistics for title cost-per-use for the different categories of publishers were as follows: U.S. commercial titles ranged from $29.47 to $1,950.00, with a mean of $343.51; U.S. association titles ranged from $1.79 to $45.50, with a mean of $15.21; foreign commercial titles ranged from $3.29 to $4,079.27, with a mean of $254.69; and foreign association titles ranged from $16.50 to $147.78, with a mean of $67.68. The difference between these means was statistically significant at the 0.0001 level.

With respect to mean document cost-per-use, the figures were as follows: U.S. commercial—$108.14; U.S. association—$7.64; foreign commercial— $73.27; and foreign association—$32.05.

In considering these numbers, one should keep firmly in mind that cost-per-use figures are institutionally specific. Thus, in the fall of 1992, the UIUC chemistry program had 44 faculty members and 277 graduate students, whereas the LSU chemistry program had 35 faculty members and 100 graduate students (Goldberger, Maher, and Flattau 1995, 316 and 318). Due to this difference, LSU's cost-per-use figures would have been concomitantly higher.