Graduate Students and Postdoctorates in Science and Engineering: Fall 2004.

Appendix A: Technical Notes

Survey Universe Top.

The data collected in the 2004 Survey of Graduate Students and Postdoctorates in Science and Engineering represent national estimates of graduate enrollment and postdoctoral employment as of fall 2004 in all U.S. academic institutions that granted doctorates or master's degrees in any science, engineering, or selected health-related field.[1] The survey collects data for all branch campuses, affiliated research centers, and separately organized components, such as graduate or medical schools. The survey universe consisted of 710 reporting units at 589 graduate institutions: 221 reporting units at 221 master's-granting institutions and 489 reporting units at 368 doctorate-granting institutions.[2]

Data on graduate science and engineering (S&E) enrollment and postdoctoral appointees have been collected since 1966. From fall 1966 through fall 1971, the NSF Graduate Traineeship Program collected these data from a limited number of doctorate-granting institutions and requested data on only those S&E fields supported by NSF. The NSF Universities and Nonprofit Institutions Studies Group began collecting these data with the fall 1972 survey. Between 1972 and 1975, eligibility definitions were changed to include health-related fields and to increase the number of S&E fields surveyed. In addition, the survey was broadened to include all institutions known to have programs leading to a doctorate or master's degree in science and engineering. Because of these changes, data for 1974 and earlier years are not comparable with 1975 and later data.

Table A-1 shows the number of institutions, reporting units, and departments at each level included in the data, as well as the total enrollment reported for each year between 1966 and 2004. The data for 1966–74 are not inflated to reflect universe totals. From 1984 to 1987, the survey included master's-granting institutions on a sample basis. For those years, the survey design incorporated a stratified random sample with a certainty stratum that included all doctorate-granting institutions; all master's-granting, historically black colleges and universities; and all land-grant institutions. NSF divided the remaining master's-granting institutions into two sample strata based on enrollment size. In 1988 NSF reestimated data for sampled institutions for the years 1984–87 on the basis of 1983 and 1988 data. During the 1989 survey cycle, NSF revised the S&E field definitions, resulting in the deletion of some departments; NSF adjusted data for 1975–88 to conform to the revised definitions. The fall 1988 survey universe included for the first time since 1983 all eligible institutions, not just a sample. Since 1988 the survey universe has included any institutions establishing S&E and selected health-related doctoral or master's programs and has excluded any that have closed all their S&E graduate programs (see "Survey Methodology," below).

Tables A-2 and A-3 present data on departmental coverage by S&E and health-related fields for doctorate- and master's-granting institutions for the years 1997–2004.

Survey Instrument Top.

The 2003 and 2004 survey instruments are essentially the same. The survey instrument consists of two forms: (1) NSF Form 811 is a computer-generated list of departments or programs specific to each institution surveyed. The list is based on the departments known to have existed in the previous survey cycle, and each school coordinator is asked to update it. (2) NSF Form 812 is a series of questions completed by the school coordinator or departmental respondent that obtains counts of students and data on key characteristics of interest, such as race/ethnicity, sex, citizenship, and source of funding. In addition to the questionnaires, which include instructions, the survey package contains the following:

Since the 1998 survey, the school coordinators and departmental respondents have had the option of providing data using the Web-based data collection system. In 2004, 620 of the 710 reporting units elected to do so.

Survey Methodology Top.

In addition to the verification of the information cited above, the acknowledgment postcard also requested that the school coordinators indicate how the data were collected, whether the data were maintained centrally or collected from individual departments, and whether they were derived from a computerized database or were hand tabulated. The majority of reporting units indicate a combination of sources for their data. Over the years, the use of computerized systems has shown a small but gradual increase.

School coordinators were requested to review the departmental listing provided in the survey packet on NSF Form 811 and indicate any changes in their school structure, such as departments newly formed, phased out, split, or merged, and to check off any departments that had neither graduate students nor postdoctorates and for which survey forms would therefore not be submitted. The school coordinators returned revised Form 811s to the survey contractor for use as a checklist in tracking departmental responses.

Schools completed a survey questionnaire for each department either centrally or at the department level and returned the questionnaire to the survey contractor for data entry, editing, and tabulation. The survey contractor referred arithmetic errors, inconsistencies between items, and sharp year-to-year fluctuations to the school coordinators for correction or clarification.

Response Rates Top.

From 1975 through 2003, NSF calculated response rates for reporting units (schools) and departments but did not calculate an institutional response rate. For schools, the response rate was calculated as the total number of responding schools divided by the total number of eligible schools. A school was considered responding if one or more of its eligible departments responded. The departmental response rate was calculated as the number of departments not requiring full imputation divided by the total number of eligible departments. A department was considered responding if it reported at least one data item.

For 2004, new response-rate calculations were evaluated and tested for schools and departments/programs, and a response-rate calculation was developed for institutions. These response rate calculations were changed to adhere to the American Association for Public Opinion Research's standards for computing response rates (see calculation Response Rate 3, AAPOR 2004: 29).[3] The 2004 response rates are not directly comparable to response rates reported for prior years.

In 2004, 578 of the 589 (98 percent) eligible academic institutions were complete or partial respondents. Of that number, 560 of the 589 (95 percent) were complete respondents, reporting data on 90 percent or more of their departments, and 18 (3 percent) were partial respondents, reporting more than 50 percent but less than 90 percent of their departments. Eleven institutions (2 percent) were nonrespondents, reporting less than 50 percent of their departments.

For 2004, 702 of the 710 (99 percent) schools in the survey were complete or partial respondents. Of that number a total of 674 (95 percent) of the 710 schools in the survey were complete respondents, reporting 90 percent of more of their departments. Of the remaining 36 schools, 24 (3 percent) were partial respondents, reporting more than 50 percent but less than 90 percent of their departments, and 12 schools (2 percent) were nonrespondents, reporting less than 50 percent of their departments.

For 2004, at the department level, 11,998 (98 percent) of the 12,240 eligible departments were either complete or partial respondents. A total of 10,524 (86 percent) were complete respondents, responding to all applicable data cells in items 5, 6, 7, and 8 and not requiring imputation. A total of 1,474 (12 percent) were partial respondents, requiring imputation for one or more data cells in Items 5, 6, 7 (excluding cells pertaining to first-time, full-time students) or 8. A total of 242 departments (2 percent) did not respond to the survey.

Table A-4 shows departmental response rates for 1975–2004. Response rates for 1975–2003 are calculated using the previous methodology. The table shows two sets of response rates for 2004. The first set uses the methodology employed from 1975–2003. The second set follows the new methodology, which adheres to AAPOR standards.

Imputation techniques were used to fill in missing data for departments. Tables A-5 and A-6 show the number of departments in doctorate- and master's-granting institutions that required total or partial imputation, graduate enrollment and postdoctorates imputed, and imputation rates for 2002, 2003, and 2004. Tables A-7 through A-9 provide imputation rates by specific data items for 2004.

Changes in Data Items Top.

Over time, changes have been made to the content of the survey to keep it relevant to the needs of the data users. Such changes prevent precise maintenance of trend data. Therefore, some data items are not available for all institutions in all years. Major changes in the data collected and the year the changes became effective include the following:

Data Revisions Top.

During the fall 1988 survey cycle, a review of the survey universe and of the S&E definition resulted in the exclusion of departments that were not primarily oriented toward granting research degrees. A number of departments—mostly those in the field of "Social sciences, not elsewhere classified" were found to be primarily engaged in training teachers, practitioners, administrators, or managers rather than researchers; thus, they were no longer eligible for the survey. During the 1989–2004 survey cycles, this process has continued, and all ineligible departments identified were removed to ensure trend consistency for the entire 1975–2004 period. These changes resulted in a reduction in total enrollments and social science enrollments for all years. Table A-10 shows the net effect on enrollment data of these adjustments over the years.

Since the fall 1992 survey cycle, an institution's previous year's data for the highest S&E degree granted were changed to reflect the institution's highest S&E degree in the current survey cycle. This change resulted in a smaller decrease in enrollment at doctorate-granting institutions than at master's-granting institutions, given that over the years a number of master's-granting institutions had become doctorate-granting institutions.

Since the 1992 survey cycle, the definition of medical schools has included only those institutional components with membership in the Association of American Medical Colleges. Data collected before 1992 are not comparable with data collected after the fall 1992 survey.

List of Technical Tables top

Table Table Name Excel Spreadsheet (.xls) Portable Document Format  (.pdf)
A-1 The NSF data collection series: 1966–2004 .xls .pdf
A-2 Science, engineering, and health departments in doctorate-granting institutions, by detailed field: 1997–2004 .xls .pdf
A-3 Science, engineering, and health departments in master's-granting institutions, by detailed field: 1997–2004 .xls .pdf
A-4 Departmental response rates: 1975–2004 .xls .pdf
A-5 Imputation for nonresponse in doctorate-granting institutions, by field and graduate enrollment or postdoctoral status: 2002–04 .xls .pdf
A-6 Imputation for nonresponse in master's-granting institutions, by field and graduate enrollment or postdoctoral status: 2002–04 .xls .pdf
A-7 Imputation rates of full-time graduate students for all departments at all graduate institutions, by source, mechanism of support, and female students: 2004 .xls .pdf
A-8 Imputation rates of graduate students at all graduate institutions, by race/ethnicity, enrollment status, and sex: 2004 .xls .pdf
A-9 Imputation rates of science and engineering postdoctorates, by source of support, and nonfaculty research staff with doctorates at all graduate institutions: 2004 .xls .pdf
A-10 Graduate enrollment data as originally published and as modified through the fall 2004 graduate student survey cycle: 1975-2004 .xls .pdf


[1] See for additional survey information and for available data related to graduate students and postdoctorates in science and engineering.

[2] For purposes of this report, the term "reporting unit" is essentially equivalent to a school (such as graduate schools, medical or dental schools, nursing schools, schools of public health), an affiliated research center, a branch campus, or any other organizational component within an academic institution that grants an S&E or health-related degree.

[3] American Association for Public Opinion Research (AAPOR). 2004. Standard Definitions, Final Dispositions of Case Codes and Outcome Rates for Surveys. Lenexa, Kansas: AAPOR.

Previous Section. Top of page. Next Section. Table of Contents. Help. SRS Homepage.