This web site was copied prior to January 20, 2005. It is now a Federal record managed by the National Archives and Records Administration. External links, forms, and search boxes may not function within this collection. Learn more.   [hide]

Survey Methodology:
Survey of Graduate Students and Postdoctorates in Science and Engineering

1. Overview top

This section provides a brief discussion of the methodology used in the conduct of the Survey of Graduate Students and Postdoctorates in Science and Engineering. For a more in-depth discussion of the methodology used in conducting this survey, please see the full report.

a. Purpose

The Survey of Graduate Students and Postdoctorates in Science and Engineering (also known as the Graduate Student Survey, or GSS) survey provides data on the number and characteristics of graduate science and engineering students enrolled in U.S. institutions. The results of the survey are used to assess trends in financial support patterns and shifts in graduate enrollment and postdoctoral appointments. The GSS collects data from all institutions offering graduate programs in any science, engineering, or health field. Data items are collected at the academic department level and include counts of full-time graduate students by source and mechanism of support, with data on women; part-time graduate students by sex; and citizenship and racial/ethnic background of all graduate students, including first-time students. In addition, count data on postdoctorates are requested by source of support, sex, and citizenship, with separate data on those holding first-professional doctorates in the health fields; and summary information on other doctorate nonfaculty research personnel.

b. Respondents

The GSS is an establishment-based survey of approximately 600 academic institutions. The survey is administered to institutional coordinators at educational institutions with programs leading to graduate degrees in S&E fields. A separate survey form is completed for each department granting S&E postbaccalaureate degrees.

c. Key variables

Count data are available by the following:

2. Survey Design top

a. Target population and sample frame

The data collected in the fall 2001 GSS provide national estimates of graduate enrollment and postdoctoral employment at the beginning of academic year 2001–02 in all academic institutions in the United States that offer doctorate or master's degree programs in any science or engineering field. Included are data for all branch campuses, affiliated research centers, and separately organized components such as medical or dental schools, schools of nursing, public health, etc. The survey universe consisted of 721 reporting units at 602 graduate institutions. Included were 238 master's-granting institutions and 483 reporting units associated with 363 doctorate-granting institutions.

b. Sample design

The survey is a census of all eligible institutions. The final 2001 survey universe consisted of 721 responding units at 602 institutions. The complete survey universe has been surveyed since fall 1988. From 1984 through 1987, master's-granting institutions were surveyed on a sample basis.

c. Data collection techniques

This was the fourth year in which respondents had the option of reporting data using the Graduate Student Survey Web-Based Data Reporting System. Sixty-six percent (467 reporting units) chose to report data using the Web system. Both institutional coordinators and designated department-level respondents had the option of submitting data through the Web system. This system provided immediate feedback on the data to the coordinators and departmental respondents, which helped ensure that their data were accurate and complete. Using the Web system allowed respondents to upload their own data files, speeding up the reporting process. The remaining 34 percent of the reporting units choose to report on paper survey forms.

In either version, paper or Web-based, institutional coordinators were asked to review a list of departments and to indicate any changes in their departmental 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 questionnaires would therefore not be submitted. The revised departmental listing forms were used to track departmental responses.

A separate survey form was completed for each eligible department either centrally by the institutional coordinator or at the departmental level by a designated respondent. Paper forms were returned to the data processing contractor for data entry, editing, and tabulation. Arithmetic errors, inconsistencies across items, and sharp year-to-year fluctuations were referred to the institutional coordinators for correction or clarification.

Information is collected for the fall of the year in which data collection occurs. Initial mail-out is in November of the year for which information is requested; collection typically ends in July of the following year.

d. Estimation techniques

Missing data for partially nonrespondent departments were imputed using the department's previous year's data, where available, or data from peer institutions in cases where data had not been reported the previous year. Data for completely nonrespondent departments were imputed using data from the previous year, where available.

3. Survey Quality Measures top

a. Sampling variability

Since the 2001 survey was distributed to all institutions on the sample frame, there was no sampling error.

b. Coverage

Coverage at the institutional level is believed to be at or close to 100 percent, given that it is relatively easy to identify schools that grant master's or doctoral degrees. At the departmental level, coverage should also be quite good, since institutional coordinators would normally be in a position to identify relevant departments. Minor coverage problems could arise through incorrect classification of departments with respect to whether they fall within an S&E field.

c. Nonresponse

(1) Unit nonresponse. Of the 721 responding units included in the fall 2001 survey, 710, or 98.5 percent, were able to provide at least partial data. At the departmental level, 11,852 departments responded, or 99.0 percent of the 11,967 departments surveyed. A total of 115 departments, or 1.0 percent of the departmental total, required complete imputation.

(2) Item nonresponse. Of the responding departments, 1,731, or 14.5 percent, had one or more data cells imputed. Imputation for missing data elements was based on the prior year's data, where available. Otherwise, they were imputed from data for peer institutions.

d. Measurement

Possible sources of measurement error in this survey are incomplete information and failure to follow survey definitions.

Institutions may have incomplete information on their students. For example, to the extent that certain types of support—e.g., self-support—are not channeled through the institution, it is likely that they are underreported. In addition, institutions are not always able to identify foreign sources of support.

Respondents may not follow instructions regarding definitions of categories. For example, institutions can vary in defining mechanisms of support (e.g., fellowships versus traineeships), and may report students according to the institution's definition rather than the survey's. Similarly, some misreporting of postdoctorates may occur within medical schools with clinicians being reported as postdoctorates. For the purposes of this survey, postdoctorates devote their primary effort to their own research training through research activities or study. Clinical fellows are generally excluded, unless the primary purpose of the appointment is research training under the supervision of a senior mentor. Additionally, institutions may have difficulty in breaking down financial information by discipline—e.g., a student may be enrolled in one discipline while receiving support from another.

4. Trend Data top

The National Science Foundation has collected data on graduate S&E enrollment and postdoctoral appointees since 1966. From fall 1966 through fall 1971, data from a limited number of doctorate-granting institutions were collected through the NSF Graduate Traineeship Program, which requested data only on those S&E fields supported by NSF. Beginning with the fall 1972 survey, this data collection effort was assigned to the Universities and Nonprofit Institutions Studies Group and was gradually expanded during the period 1972–75 to include additional S&E fields as well as all institutions known to have programs leading to the master's or doctorate degree. Due to this expansion, data for 1974 and earlier years are not strictly comparable with 1975 and later data. No attempt has been made to inflate the data for 1966–74 to reflect universe totals.

During the fall 1988 survey cycle, the criteria for including departments in the survey universe were tightened, and all departments surveyed were reviewed. Those departments not primarily oriented toward granting research degrees were no longer considered to meet the definition of "science and engineering." As a result of this review, it was determined that a number of departments, primarily in the field of social sciences, not elsewhere classified, were engaged in training primarily teachers, practitioners, administrators, or managers rather than researchers; these departments were consequently deleted from the database. This process was continued during the fall 1989–97 survey cycles and expanded to ensure trend consistency for the entire 1975–97 period. As a result, total enrollments and social science enrollments for all years were reduced.

The definition of "medical schools" was revised during the fall 1992 survey cycle to include only those institutional components that are members of the Association of American Medical Colleges. Tables generated after the fall 1992 survey differ from their counterparts in earlier years in that they exclude schools of nursing, public health, dentistry, veterinary medicine, and other health-related disciplines; they should not be compared with tables from earlier years.

Although NSF has attempted to maintain consistent trend data, some modifications in the survey questionnaire have been made to respond to changing issues over the past 25 years. As a result, some data items are not available for all institutions in all years. Major changes in the data collected are as follows:

For these reasons, and because institutions may revise their data for earlier years, only the latest trend data should be used in historical analyses. Analysts wishing to do trend analyses not covered in this report are encouraged to contact the project officer listed below for additional information.

5. Availability of Data top

a. Publications

The data from this survey are published annually in Graduate Students and Postdoctorates in Science and Engineering, available on NSF's Division of Science Resources Statistics (SRS) website. Information from this survey is also included in Science and Engineering Indicators and Women, Minorities, and Persons with Disabilities in Science and Engineering, both produced by SRS. Selected data items from this survey for individual doctorate-granting institutions are included in SRS's Academic Institutional Profiles series.

b. Electronic access

Data from this survey are available on the SRS Web site and from WebCASPAR. Public use data files are available for the years 1973 through 2001. For a detailed description of the file format, see the Guide to the Graduate Student Survey Data Files at

c. Contact for more information

Additional information about this survey can be obtained by contacting:

Emilda Rivers
Mathematical Statistician
Division of Science Resources Studies
National Science Foundation
4201 Wilson Boulevard, Suite 965
Arlington, VA 22230

Phone: (703) 292-7773


[1] For purposes of this survey, native residents of a U.S. possession and applicants for U.S. citizenship holding a green card are counted as U.S. citizens. See section 4, Trend Data, for information on changes in the citizenship by the survey over time.

Last Modified: Mar 02, 2004 Comments to