Ending Mandatory Retirement for Tenured Faculty: The Consequences for Higher Education (1991)

The figures in Chapter 2 showing the age distribution of faculty members are based on data from three surveys: the National Survey of Postsecondary Faculty (NSOPF), conducted by the National Center for Education Statistics of the U.S. Department of Education; the Survey of Doctorate Recipients (SDR), conducted by the National Research Council for the National Science Foundation and other federal sponsors; and the faculty survey conducted by the Higher Education Research Institute (HERI) of the University of California at Los Angeles. The NSOPF and the SDR also provide data on the age distribution of faculty members by type of institution and by academic field. The HERI data provide information from a larger number of faculty members, although faculty were included in that survey using nonscientific sampling techniques. The HERI faculty age distribution, which looks similar to the NSOPF and SDR age profiles, was included because so few large faculty data bases are available. In this appendix we describe the three data bases used; Table A-2 shows the basic characteristics of each.

The SDR is a longitudinal survey of doctorate holders in the sciences, social sciences, engineering, and humanities. It is designed to collect information on the demographics, employment, and supply of those doctorate holders in the United States. As the most recent SDR methodological report (National Research Council, 1989a:1) describes the survey:

The SDR sample population is selected from all research doctorates granted in the United States. The total sample size for the 1989 survey was 91,327; 48,408 usable responses were received. Our age distribution tables and figures are based on the replies of 19,117 doctorate holders reporting employment as faculty members.

Since the SDR is a survey of doctorate recipients, it is not representative of all faculty. In particular, we did not use the SDR to estimate the age distribution of faculty at 2-year colleges because approximately 75 percent of 2-year college faculty do not have doctorates. The SDR is more representative of the faculty at 4-year colleges and universities, about 70 percent of whom have doctorates.

Prior to 1987 individuals were selected for the sample who had earned their doctorates within the past 42 years. For each new survey the oldest two groups were dropped and replaced by a sample of people who had received doctorates in the previous 2 years, thus maintaining the 42-year coverage span. In 1987 and 1989, in response to concern about the retirement rates of doctorate holders, the oldest groups were retained in the sample when new doctorate recipients were added. Thus, the 1989 sample contained individuals who received a doctorate between January 1, 1942, and June 30, 1988 (National Research Council, 1989a; 1990b).

The exclusion from the sample of individuals earning doctorates prior

to 1942, reflecting the survey's original intent "to represent all working-age doctorates living in the United States" (National Research Council, 1990b:2), creates a possible source of bias in measuring the number of doctorate holders or faculty over age 70. According to SDR statistics, the median age at which doctorate holders earned their degrees in 1942 was 28, so the majority of faculty earning doctorates in 1942 would have been in their 70s at the time of the 1989 survey. Therefore, we believe any error in the age distribution of employed faculty resulting from the absence of individuals earning doctorates prior to 1942 is likely to be small.

The SDR has been conducted as a mail survey since it began. Response rates have declined from 75 percent for the first survey in 1973 to 55 percent in 1989. As the response rate decreases, the probability increases that the data received do not accurately represent the population. (For a detailed discussion of the need to improve the SDR response rate, see National Research Council, 1989b.) The low response rate for recent surveys gave the committee additional reason for caution in using SDR results. We note that a 1989 pilot study of the effectiveness of computer-assisted telephone interviewing as a way to obtain interviews from nonrespondents to the mail survey suggests that improvement in the overall response rate is possible (National Research Council, 1990b), and we commend efforts to increase the response rate in order to make the SDR more useful for future researchers.

When faculty respondents to the SDR are divided into subcategories by age and institution type or field, the unweighted numbers of faculty over age 60 in some categories drop below 50. The unweighted numbers of faculty aged 65–69 and 70 or older in some fields are single digits or zero. The committee therefore limited its analysis of faculty age distributions to broad fields of study and broad categories of institutional types. We view the data on the proportion of older faculty by field with some caution. The committee also checked the data on faculty age distribution by institution type and field against results from the NSOPF. In Chapter 2 we present results only from the SDR, since the NSOPF has a smaller sample size and therefore requires similar caution.

The SDR is not very useful for looking at retirement issues. The number of faculty responding that they were employed in 1987 and retired in 1989 was too low for us to calculate retirement rates. The survey asks only for current employment status, not for when that status changed. Therefore, the data show only that a respondent's retirement occurred sometime between the last survey on which he or she reported employment and the first on which he or she reported retirement. For individuals responding to consecutive surveys, this would give a 2-year range of possible retirement ages. However, respondents checking "retired" who did not respond to the previous survey may have retired during the previous 4 or more years.

Although the SDR is a longitudinal survey, many respondents do not respond in successive survey years: For example, only about 65 percent of the 1989 survey respondents also responded in 1987.

Furthermore, the SDR survey form is not designed to collect information on retirement from a specific job, such as a tenured faculty position. It asks respondents to indicate whether they are employed full time, employed part time, on a postdoctoral appointment, unemployed and seeking full-time or part-time employment, not employed and not seeking employment, or retired and not employed. Faculty who officially retire and continue to work part time as part of a partial retirement program or faculty who engage in research or consulting work after retiring from a tenured position might therefore not indicate their retirement on the survey form.

The NSOPF is a survey of instructional faculty in higher education; the National Center for Education Statistics (NCES) conducted this study for the first time during the 1987–1988 academic year. The NSOPF had three components: a survey of institutional policies and practices, which was sent to institutions; a survey of faculty at those institutions; and a survey of department chairs at those institutions. Faculty provided information on their backgrounds, responsibilities, compensation, and attitudes. Institutional and department-level respondents provided information on faculty composition, turnover, recruitment, retention, and tenure policies (National Center for Education Statistics, 1990b).

The stratified random sample of 480 institutions used in the survey was selected (National Center for Education Statistics, 1990b:94): ". . . [from all] accredited nonproprietary U.S. postsecondary institutions that grant a two-year (A.A.) or higher degree and whose accreditation at the higher education level is recognized by the U.S. Department of Education." The sample included religious colleges, medical schools that are independent of a 4-year college or university, other specialized postsecondary institutions, and 2- and 4-year colleges and universities. The sample was drawn from the 1987 Integrated Postsecondary Education Data System (IPEDS), which contained 3,159 institutions meeting the sample criteria (National Center for Education Statistics, 1990b).

Of the 480 institutions in the sample, 449 (94 percent) provided lists of their part-time and full-time instructional faculty members. A stratified random sample of 12,569 faculty was selected from these lists. On the basis of the responses received, NCES estimated that 11,071 of the respondents met eligibility criteria as regular instructional faculty; 8,382 eligible faculty responded, for a faculty response rate of 76 percent (National Center for Education Statistics, 1990b). This response rate does not take into

account the 6 percent of sampled institutions that declined to participate and whose faculty were therefore not included.

The NSOPF included questions asking faculty whether they anticipated retiring within the next 3 years, but because the survey population included current faculty only, the NSOPF produced no actual data on faculty retirements. Consequently, the committee chose to limit its use of NSOPF data, as it did the SDR data, to showing the age distribution of current faculty.

The NSOPF has two advantages over the SDR as a source of data on faculty members. First, it is a survey of faculty, rather than of doctorate holders, and it is therefore more representative of the population of faculty members. Second, its sample design did not contain any selection criteria likely to exclude the oldest faculty, such as the exclusion from the SDR of individuals who earned doctorates prior to 1942.

The NSOPF also has two disadvantages for the purposes of collecting information on faculty demographics. First, since the 1997–1988 survey was the first conducted, the NSOPF could not provide information on changes in faculty age distributions over time. The NCES has announced plans to repeat the survey in the 1991–1992 academic year, and the committee notes that in the future this survey may be a useful source of information on changes in faculty demographics, activities, and attitudes and in institutional policies.

Second, despite its higher response rate, the NSOPF provides data on fewer faculty than the SDR. As in the SDR, the number of faculty in the highest age categories is small, which limits analysis of the data by age and any other category such as type of institution or field of study. Further research on the demographics, responsibilities, and attitudes of older faculty would be possible if the 1991–1992 NSOPF oversampled faculty over age 60 in order to obtain a larger number of responses from older faculty members.

The HERI was conducted in 1989 with funding from the National Science Foundation, the Exxon foundation, and individual colleges and universities. It began as a survey of faculty at 150 colleges and universities, but the investigators then invited all the 2-year and 4-year colleges and universities in the country to participate in exchange for a fee to cover reporting data back to individual institutions. Thus, some institutions in the HERI survey were self-selected paying participants rather than part of a random sample chosen by standard statistical methods.

In all, 432 institutions participated. HERI asked institutions to provide a complete list of their faculty members; if the institution did not do so, HERI obtained lists of faculty from outside vendors. The lists obtained from both institutions and vendors reflect different institutional definitions of faculty status. Some included librarians, part-time faculty, and administrators in addition to regular full-time instructional faculty members. HERI distributed survey questionnaires to 93,479 faculty members listed and received 51,574 usable responses (55 percent).

The use of nonscientific sampling techniques limits the usefulness of HERI data, although HERI survey procedures contain no obvious sources of bias by age. The committee notes only that the age distribution of faculty responding to HERI is remarkably similar to the faculty age distributions calculated from SDR and NSOPF data. Since few large faculty data bases are available, we included the HERI age distribution in Chapter 2 as an additional check on the overall age distribution of U.S. faculty.

Modification of faculty surveys to provide additional data on faculty retirement behavior would assist policy makers at state and federal levels as well as those at colleges and universities not only in considering faculty retirement policy but also in predicting and preparing for possible changes in faculty supply and demand. The committee therefore recommends that the sponsors of faculty surveys oversample older faculty and, when relevant, retired faculty to ensure an adequate data base for estimating the number of faculty over age 70 and studying faculty retirement patterns. We also encourage survey sponsors to develop questions that measure when faculty retire and address such retirement issues as retirement benefits and factors affecting the decision to retire.

The committee notes that state retirement systems and private pension plan providers may also be in a position to collect data of use to government policy makers, colleges and universities, and researchers considering retirement issues. TIAA-CREF has already done a number of studies on retirement policy issues, including surveys of older and retired faculty members. However, the TIAA-CREF data base of participants and many state retirement system data bases do not contain any means of separating data on faculty from data on other participants in their pension plans. We recommend that pension plan providers seek ways to assist colleges and universities, policy makers, and researchers by coding data in a way that permits studies of faculty retirement behavior.