The phenomenon of concentration within the curriculum cores followed the same pattern as that between the curriculum cores. With respect to costs, the percentage of titles accounting for 75% of total costs ranged from a low of 19.2% in General Science to a high of 50.0% in both Biological and Agricultural Engineering and Electrical Engineering. For all 33 curriculum cores, the average percentage of titles accounting for 75% of costs was 34.4%. Concerning ST value, the percentage of titles responsible for 75% of faculty score ran from 9.6% in General Science to 50.0% in Industrial and Manufacturing Systems Engineering. The average percentage of titles accounting for 75% of faculty score in all 33 curriculum cores was 33.7%—very close to the average for 75% of total costs.
Given this concentration of costs and ST value, it was decided to investigate whether the library market for ST journals manifests in other subject areas two main features that had been found in chemistry by the analysis of the data from the 1993 SRP pilot project with the LSU Department of Chemistry, i.e.: (1) that ST value plays no role in the price of ST serials, and (2) that the market bifurcates, with costs tending to concentrate on the serials of the commercial, largely foreign publishers and ST value tending to concentrate on the journals of the U.S. associations. Other researchers had found these features with methods that used impact factor. For its proper utilization, impact factor should be employed with costs also controlled for size, and such methods have the disadvantage of masking the huge effects of the skewed distributions operative in the library market for ST serials. The most famous case concerned physics and the work done by Barschall (1988) and Barschall and Arrington (1988). Barschall, a University of Wisconsin–Madison physicist, divided cost measured in cents per 1,000 characters by impact factor and came to the following conclusion (Barschall l988, 57):
Barschall's findings were replicated by other researchers in other fields. Applying Barschall's method in chemistry, Christensen (1992) estimated that association journals were about 4 times more cost effective than commercial ones. Ribbe (1988; 1990; 1991) tried a slightly different approach in the geosciences, dividing cost per citable source item by impact factor, which he found to be highly correlated with the proportion of papers supported by grants from the National Science Foundation, U.S. Department of Energy, and the National Aeronautics and Space Administration. However, his results were similar, and he found that by his index association journals rated on the average 5.0 times more favorably than commercial ones in Geosciences, 3.3 times more favorably in Geology, and 4.4 times more favorably in Paleontology. Ribbe's findings were corroborated by Turner (1994) in a study of 274 journals of interest to estuarine and coastal scientists. Using both total citations and impact factor, Turner concluded (p. 724) that "on the average, professional societies often (but not always) publish relatively high impact articles at one-third to one-tenth the price of commercial for-profit publishers." Moline (1991) used the Spearman rank-order correlation to test the relationship of impact factor to cost in cents per 1,000 characters for mathematics journals. She made this test with her data defined into three different sets: commercial publishers; "other" publishers such as associations, university presses, and university mathematics departments; and commercial and "other" publishers together. No significant results at the 0.01 level were found for the first two sets, and with the commercial and "other" publishers combined Moline actually found a negative correlation of -0.38 significant at the 0.01 level. In a study of 5,399 journals in 12 scientific disciplines, Van Hooydonk (1995) found upon grouping the journals by country of publication that U.S. titles had a cost per article considerably lower than average but impact factors 1.5 times the average, constituting one of the few bargains in the ST serials market.
A number of researchers have utilized impact factor without correcting costs for size. For example, Baldwin and Baldwin (1989) visually compared impact factors to price for 1,048 journals in 15 subject categories, and came to the conclusion that (p. 128):
More questionable are the results of studies by Nisonger (1993) and Petersen (1992), because these researchers employed standard statistical techniques without correcting price for size. Nisonger's results fit the standard pattern, as he found no statistically significant correlation at the 0.05 level between institutional subscription prices and impact factors of genetics journals for the years 1980, 1985, and 1990. However, Petersen's findings for economics journals represent an anomaly, because of all the researchers whose work was reviewed, Petersen was the only one to find a positive and significant relationship of a journal's "impact" with its price.
To conduct his analysis, Petersen utilized a regression model in which 1990 price was made the dependent variable and the independent or causal variables encompassed the following factors: size (number of issues per year and pages per issue), circulation, presence or absence of advertising, type of publisher (commercial or nonprofit such as an association), country (U.S. or Canadian, British, European, or other), and "impact." For "impact" Petersen used a ranking constructed by Leibowitz and Palmer (1984) in the following complicated manner: a set of economics journals was chosen; this set was then ranked by total citations to these journals in 1980 to issues published between 1975 and 1979 to control for age; these total citations were then adjusted by excluding citations from noneconomics issues and reducing the weight of citations coming from the lesser-cited economics journals to emphasize the importance of the journal to the economics profession; and, finally, the adjusted citations were controlled for size by dividing them by the total number of characters published by the journals in the 1975–79 period. Not surprisingly, major discrepancies were found between this ranking and the most influential contemporaneous one established by peer ratings of academic economists.
Except for the positive relationship of "impact" to price, Petersen's findings followed the usual pattern: bigger journals measured by number of issues per year cost more; high circulation journals cost less; commercial journals cost more than nonprofit ones; and European journals cost more than U.S. or Canadian ones. As for his finding on "impact," this is highly dubious for the following reasons: he used 1990 prices unadjusted for size against 1980 citations adjusted for size by 1975–79 size measures; he violated Garfield's law of concentration by using a measure that deliberately excluded citations from other disciplines; and the ranking he used did not conform to contemporary peer ratings.
It should be pointed out that in none of the above studies did researchers take into account the effect of the higher impact factors of review journals.
To investigate the structure of the library market for ST journals, Evaluator runs were made for the purpose of measuring the trade-offs in costs versus ST value within the desired universe of serials—i.e., for every title named by the LSU faculty in the SRP survey—in all 33 curriculum cores. For this investigation, the Evaluator's second algorithm was once again chosen, and again the runs were made at the default settings of trying to reduce total costs by 75% while retaining 75% of total ST value. Of primary interest was the type of publisher involved in these trade-offs.
The lack of correlation between price and ST value became immediately apparent. The lowest trade-off in terms of loss of ST value for cost reduction was in Food Science, where a 24.3% reduction in costs could be achieved with the loss of 9.5% in total faculty score—a favorable ratio of 2.6 to 1. The highest trade-off in terms of loss of ST value for cost reduction was in General Science, where costs could be reduced by 73.1% for only a loss of 4.4% in total faculty score—a favorable ratio of 16.6 to 1. For all 33 curriculum cores the average cost reduction was 38.6% for an average loss in total faculty score of 7.7%—a favorable ratio of 5 to 1.