Section B.
Special Analysis of Expenditures Data

Changes Since 1988-89 in Computer Science Instrumentation Expenditures
Instrumentation Expenditures per Faculty Member
Trends Since 1982-83 in Expenditures by the Smaller Units

Changes Since 1988-89 in Computer Science Instrumentation Expenditures

Total expenditures for research instrumentation in computer science varied considerably by the type of unit, reflecting in large part the changing role of central computer facilities in science and engineering research. Expenditures at central computer facilities dropped 61 percent since the last survey in 1988-89 (from $193 million to $76 million), while expenditures in computer science departments increased 50 percent to $51 million (table 1).

There is of course continuing demand for powerful large computers and for supercomputers, and their usage has been increasingly shared among researchers in the academic community due to the rapid increase in networking access to off-site research computers via electronic connections. Since the last survey in 1988-89, however, the phenomenal growth in the computational capabilities of personal computers and workstations, coupled with their continued decrease in price, has changed the purchase pattern for computers being used for research in all disciplines. The emphasis on many campuses is increasingly away from the purchase of large mainframe computers housed in central facilities, and toward the purchase of smaller, very powerful computers that can be located in the laboratories and offices of the researchers themselves.

Thus, while many of the largest computers and the supercomputers are still utilized heavily, and their aggregate cost is great, new purchases are increasingly the less expensive but very powerful computers to be housed in the researchers' own laboratories, many of which have networking capability for off-site research and collaboration. As a result, new research instrument purchases at central computer facilities were still very costly, but had dropped to 60 percent of computer science expenditures in 1993, down from 85 percent in 1988-89.

Other data from the survey corroborated this change. The number of computer facilities reported in the 1993 survey was lower than the number reported in the 1988-89 survey; followup calls were made to ascertain the reasons for the decline. On many campuses, there had been a change in the academic mission of many of the mainframe computers housed in central computer facilities: Many of the mainframes utilized as research instruments in the 1988-89 survey were reported in the 1993 survey as being used primarily for administrative purposes instead. As a result, any continued expenditures at these computer facilities were not considered to be research expenditures, and consequently were not reflected in this survey.

Although the overall importance of computer instrumentation in research has continued to increase, the reporting of expenditures for computer research instruments may have moved from being expenditures for the discipline of computer science to being expenditures in the originating disciplines.

This change is illustrated by the answers to the survey question that asked respondents to list the research instrument considered the top priority item of need in their units. Twenty-eight percent of all respondents listed computers as their top priority item of need. As would be expected, 99 percent of respondents from computer science listed a computer as the top priority item, but 35 percent of respondents from both the environmental sciences and engineering also listed a computer as their top priority item (table A-10).

Research instrumentation needs at the remaining computer science facilities used for research are of course quite large in the aggregate, reflecting the very high cost per-item for instruments in these units. The median cost for a top priority computer for the computer science facilities ($200,000) is far greater than the cost of the top-priority computer for any other discipline (table A-10).

Instrumentation Expenditures per Faculty Member

To make well-informed policy decisions about research instrumentation funding it is necessary to know the extent of total expenditures by the various fields. However, some other normalizing procedure must also be used to relate total expenditures data to the actual experiences in each unit. The responding institutions in this survey represent some of the very largest academic research institutions, and their departments and facilities are often larger than corresponding units in other institutions. Larger units tend to have more faculty and therefore more research activity, which increases instrumentation expenditures based on size considerations alone.

The survey questionnaire asked for number of faculty, which allowed NSF to tabulate median expenditures on a per-full-time faculty and per-full-time research-faculty basis, in addition to the analysis of spending on a per-unit basis.

The Question
Respondents to this survey were asked to give the "number (headcount) of FULL-TIME faculty members in your unit" and the "number (headcount) of FULL-TIME faculty members in your unit who are participating in separately budgeted research projects."

Expenditures per Faculty Member.—The median expenditure per faculty member in 1993 for the purchase of research instrumentation by the units covered by this survey was $9,844. The field of research made a large difference in the level of expenditures: Median outlay per faculty member ranged from a high of $18,730 in chemistry to a low of $3,063 in the agricultural sciences (figure 1).

The type of unit also made a difference: Median expenditures per-faculty in the larger units were considerably higher ($24,200) than the per-faculty expenditure at the smaller units ($8,700).

Expenditures per Research-Faculty Member.—Although one usually associates the faculty with research at these large research-intensive universities, all faculty do not necessarily conduct research, nor conduct it during every year. For this reason the median expenditures per research-faculty member among the units responding to this survey were higher than the median per-faculty expenditure, reaching just over $13,000 in 1993 (figure 2). As in other instrumentation expenditure data, the field of science made a large difference in the expenditures: Median expenditures ranged from a high of $24,313 per research-faculty in chemistry to $4,284 in agricultural sciences.

The size of the largest instrument in the unit also made a difference in the per faculty costs. The larger units expended $26,200 per research-faculty member, compared with an $11,800 outlay by the smaller units.

Trends Since 1982-83 in Expenditures by the Smaller Units

Data Considerations
Over the years since the first instrumentation survey in 1982, there have been additions made to the types and the cost range of the instruments that were considered in scope. Therefore, trend data back to 1982 are available only for one category of respondents-those in the smaller units (in which all instruments cost less than $1 million). As would be expected, in most fields the aggregate expenditures for research instruments are generally higher for the smaller units, because these units outnumber the larger units (in which at least one instrument costs $1 million or more) (table 1). The exception is in computer science facilities, where the bulk of the expenditure cost is in the larger units, which typically house large mainframe computers.

Expenditures By The Smaller Units For The Purchase Of New Instruments
Total expenditures.—Total expenditures for the smaller units increased from $398 million in 1982-83 to $935 million in 1993 (table 1), an average annual increase of approximately 9 percent. However, figure 3 demonstrates that the greatest rate of increase in current dollars occurred between the 1982-83 and 1985-86 surveys (an approximate 19 percent annual increase). Since the 1985-86 survey, expenditures for purchases of academic research instrumentation in current dollars have risen more slowly (approximately 4 percent per year).

The change in the rate of increase was more dramatic in constant dollars. The total average annual increase since 1982-83 was approximately 5 percent. However, the constant dollar annual rate of increase, which was approximately 14 percent between 1982-83 and 1985-86, fell to an annual increase of approximately 1 percent between 1985-86 and 1993 (figure 3).

Median expenditures per unit.—Normalized on a per-unit basis, the median expenditures of the smaller units also changed direction since 1985-86. Median expenditures increased at a rapid rate in current dollars between the 1982-83 and the 1985-86 surveys, but have been virtually level ever since (figure 4). In constant dollars, median expenditures per unit for the purchase of additional equipment have registered declines in every survey since 1985-86.

Expenditures by source of funds.—Federal funding as a proportion of total funding for the smaller units decreased from 50 percent in 1982-83 to 46 percent in 1993 (table A-5). State government contributions also declined (from 11 percent to 9 percent). Industrial contributions remained level, at 8 percent of total expenditures.

Institutions' own funding rose from 24 percent in 1982-83 to 27 percent. "Other" sources, which include foundation and individual donations and bonds, rose from 7 percent to 10 percent.

Expenditure Trends by Field.— Smaller units in engineering experienced the greatest absolute increase in research instrumentation expenditures in current dollars: The $260 million in 1993 was almost three times larger than the $93 million in 1982-83. Smaller units in the biological sciences also experienced a large absolute increase: The $279 million spent in 1993 was more than twice the $130 million spent in 1982-83 (table 1).

In terms of proportional change, the largest percentage increase in expenditures since 1982-83 was in the computer sciences. This is not surprising, given the relative newness of the research field in the early 1980s, and the phenomenal increase in research computer capacity in recent years. Starting from the low base of $16 million in 1982-83, computer science expenditures rose 262 percent to $58 million in 1993.

Agriculture was the only field of science that failed to double its expenditures since 1982-83, when expenditures were $27 million. Expenditures increased 52 percent to $41 million in 1993. That increase, however, was 6 percent in constant dollars, an increase of less than 1 percent a year.

Expenditures by the Smaller Units for Maintenance/Repair and Operation of Existing Instruments
Trend data for the smaller units are also available for maintenance/repair costs back to the first survey in 1982-83. Survey questions about operation expenses, however, were only introduced with the 1987-88 survey (table A-1).

Maintenance/repair.—Expenditures for maintenance/repair in the smaller units decreased in 1993 for the first time since 1982-83. The $162 million spent in 1993 was 6 percent less than the $173 million spent in 1988-89 (table A-1). Median expenditures per unit decreased from $38,000 in 1988-89 to $25,000 per unit in 1993 (table A-6).

Maintenance /repair includes expenditures for service contracts and field service, salaries of maintenance personnel, tools, supplies, etc. In previous surveys, service contracts and field service comprised slightly more than 40 percent of total maintenance/repair expenditures. (This question was not asked on the 1993 survey, because of respondent burden issues.)

Operation.—Data for operation expenses were collected beginning with the 1988-89 survey. Median expenditures for smaller units dropped between the two surveys from $60,000 to $40,000 per unit. In the 1988-89 survey, respondents were asked to provide a breakdown of costs between salaries to operate the equipment and all other costs. Salaries comprised most of the costs in the smaller units. More than half of the respondents, in fact, had no operation expenses other than salaries. The larger units, by contrast, had median costs of $217,000 for salaries and $80,000 for other expenses (table A-7).