Reshaping the Graduate Education of Scientists and Engineers

APPENDIX B

STATISTICS ON GRADUATE EDUCATION OF SCIENTISTS AND ENGINEERS

Michael McGeary
Study Director, Committee on Science,
Engineering, and Public Policy

Contents


OVERVIEW

About 1,500 institutions of higher learning in the United States have programs leading to degrees in science and engineering. Of those, nearly 300 offer doctoral-degree programs in science and engineering. They also offer master's degrees, and more tha n 400 nondoctoral academic institutions offer master's-degree programs in science and engineering. In 1992, about 430,000 graduate students were in science and engineering programs; 87% of them were at the 300 doctorate-granting institutions.

In 1992, about 80,000 master's degrees and 25,000 doctoral degrees were earned in science and engineering fields. About one-fourth of the doctorates were awarded in each broad field of science and engineering: physical/mathematical sciences, life science s, social sciences, and engineering. The median time from the bachelor's degree to the PhD was 9.2 years. More than half of the master's degrees and 90% of the PhDs are awarded by the 150 universities that receive 90% of federal academic R&D funding.

About 5% of all science and engineering doctorate recipients in 1993 (14% of life-sciences PhDs) were supported by federal fellowships and traineeships. Another 61% (including 78% of physical scientists and 69% of engineers) received external support, pr imarily research assistantships and teaching assistantships. Many of the research assistantships were funded by federal grants. About one-quarter of the science and engineering doctoral recipients (including one-half the social scientists) were self-sup porting (including federally guaranteed loans).

More than one-third more doctorates in science and engineering were awarded in 1993 than in 1983. Seven-tenths of the net increase in doctorate awards went to foreign citizens with temporary visas, and most of the remaining increase was to US women. In 1993, nearly 30% of the doctorates were earned by women, up from about 25% in 1983. In 1992, 5.7% of PhDs were earned by members of underrepresented minorities in 1992, up from 4.1% in 1983; most of the increase was earned by Hispanics. Foreign citizens with temporary visas greatly increased their share of US doctorates, earning 18.5% in 1983 and 32% in 1993; almost all the net increase was accounted for by citizens of Asian countries. Nearly half of the engineering PhDs went to foreign citizens with t emporary visas.

THE GRADUATE STUDENTS

In 1992, the National Science Foundation (NSF) estimated that about 431,600 students were enrolled in graduate science and engineering degree programs (NSF, 1994a:Table 1). Most (87%) were enrolled in doctorate-granting institutions, a proportion that h as varied only slightly since the NSF survey began in 1975. Most (67%) were full-time students (this proportion was 72% in doctorate-granting institutions).

It is not possible to tell which of these graduate students were enrolled in master's degree programs and which in doctoral programs, although many PhD recipients have master's degrees (72% in 1993) (NRC, 1995:Appendix Table A-3).

Table B-1, a comparison of the distribution of science and engineering graduate students among fields by type of institution and enrollment status shows that life-sciences graduate students were s omewhat more likely than science and engineering graduate students overall to be at doctorate institutions and to be enrolled full-time. Social sciences and behavioral-science graduate students had the opposite pattern: they were somewhat more likely to be part-time and at master's institutions. Engineering graduate students were slightly more likely to be at doctorate institutions but more likely to be enrolled part-time.

Non-US Citizens

Nearly 110,000 (25.3%) science and engineering graduate students were not US citizens in 1992. About 93,000 of them were enrolled full-time. Their distribution among fields differed from that of US-citizen science and engineering graduate students. Table B-2 shows the distribution of full-time science and engineering graduate students by citizenship and broad field in 1992. Those who were not US citizens were more likely to be studying enginee ring or the physical sciences and less likely to be in life-science or social/behavioral-sciences programs.

As a result, those who were not US citizens constituted relatively high proportions in some fields--46% of all full-time graduate students in engineering and 39% of those in the physical/mathematic sciences--but low proportions in other fields--27% of al l full-time graduate students in the life sciences and 17% of those in the social/behavioral sciences or psychology.

Female Graduate Students

In 1992, more than 150,000 (35%) science and engineering graduate students were women (up from 25% in 1977). As Table B-3 shows, they were more likely to be enrolled in the life sciences or the s ocial/behavioral sciences and less likely to be in the physical sciences or engineering. In fact, half of all female science and engineering graduate students were in social sciences and psychology programs.

As a result, the majority (54%) of graduate students in the social/behavioral sciences were women, as were 44% of those in the life sciences. Only 15% of engineering graduate students and 27% of those in the natural (physical, environmental, mathematica l, and computer) sciences were female.

Members of Underrepresented Minorities

Fewer than 29,000 (9%) of science and engineering graduate students who were US citizens were members of underrepresented minorities--black, Hispanic, or American Indian. Compared with all US-citizen graduate students, they were much more likely to be s tudying social/behavioral sciences (53 versus 37%) and substantially less likely to be in the life sciences (13% versus 20%). Members of underrepresented minorities constituted 13% of US citizens in the social/behavioral sciences and about 7% of those in the other broad fields (see Table B-4).

Growth Trends in Full-Time Graduate Enrollment Since 1982

In 1992, there were nearly 291,000 full-time science and engineering graduate students, 30.6% more than in 1982. The growth by field is presented in the first column of Table B-5. Much of the ne t growth came from foreign citizens; as overall enrollment was increasing by almost 2% a year, foreign enrollment was growing by more than 5% a year (NSB, 1993:50). The second and third columns of Table B-5 compare the increases in full-time science and engineering graduate students who were foreign citizens with those who were US citizens in 1982-1992, by field.

Enrollment increases were also driven by the increased participation of women--3% a year, compared with 1% among men, during the 1980s. There were absolute decreases in the number of male graduate students in the life, environmental, and social science s and psychology (NSB, 1993:53).

Growth in First-Year and Beyond-First-Year Enrollments, 1982-1992

The NSF survey of graduate students and postdoctorates in science and engineering fields began to collect information on the number of first-year full-time enrollments in 1982. The data indicate that first-year enrollments increased at a lower rate than total full-time enrollments until about 1989, after which they increased more rapidly for several years. During 1982-1992, first-year enrollments increased by 17% and beyond-first-year enrollments by 37% (Table B-6). It is difficult to interpret those data. Are the recent large increases in first-year enrollments the result of reports in the middle to late 1980s of impending shortfalls in the number of PhDs or the tendency of more college grad uates to go to graduate school when economic conditions are poor? Also, how much of the higher rate of growth among beyond-first-year graduate students until recently was simply the manifestation of the steadily increasing degree requirements among scien ce and engineering PhDs, and how much was due to graduate students' deliberately delaying completion of their degrees as short-term responses to poor job-market prospects?

Sources and Mechanisms of Financial Support

In 1992, science and engineering graduate students were supported in a number of ways by a variety of sources. For each full-time student, the NSF survey asks for the "major" (i.e., largest) source of support (e.g., federal, institutional, and self) and the type (e.g., fellowship, and/or research assistantship). Table B-7 shows that the sources of support vary considerably from field to field. Although on the average 20% of full-time science an d engineering graduate students received their major support from a federal source, this was the largest source of support for 32% of graduate students in biology and nearly 36% of graduate students in the physical sciences. Only 7% of graduate students in the social or behavioral sciences and 10% of those in the mathematical sciences were supported primarily by federal funds. Nearly two-thirds of mathematical scientists and half of those in the physical sciences received their major support from their institutions (mostly in the form of research and teaching assistantships), but institutional funds were also an important source of graduate support in the other disciplines--between 32 and 45%. "Own funds" (including, however, federally guaranteed loans ) were the major source of support for large fractions of graduate students in some fields--46% of those in computer science and 40% of those in the social and behavioral sciences--but for relatively few in physics, astronomy, and chemistry (6%) or the bi ological sciences (13%). Only a few percent received foreign support (although those completing the survey might not always have known whether those funding their own way--thus classified as self-supporting--were receiving foreign support). Finally, abo ut 7% overall were receiving support from industry and domestic sources other than federal and institutional. About 11% of students in engineering and agricultural science were receiving such support.

The pattern of sources of support has not changed much over the last 10 years, as seen in Table B-8; federal, institutional, and other US sources of support were up by several percentage points ea ch, offsetting relative declines in foreign and own sources of support.

The sources of federal support for full-time science and engineering graduate students also varied by field (see Table B-9), although the pattern has not changed since 1982 (compar e NSF, 1992: Table C-18).

As for the mechanisms of support (including nonfederal), the latest detailed data by field are for 1991 (Table B-10).

Science and Engineering Graduate Students in Master's-Granting Institutions

In 1992, nearly 57,000 science and engineering graduate students were in institutions whose highest degree is the master's. Most of them (65%) were enrolled part-time. Table B-11 gives their d istribution and enrollment status by field and compares them with graduate students at doctorate-granting institutions. The table indicates that graduate students in master's degree institutions were much more likely to be in the social sciences and psyc hology and much less likely to be in the life sciences. In fact, 26% of all graduate students in the social sciences and psychology were in master's-degree institutions, compared with 11% of the remaining science and engineering graduate students.

THE INSTITUTIONS

In 1991, there were 3,611 institutions of higher education in the United States; they enrolled 14 million students and granted 1.9 million degrees, of which about one-fourth (470,000) were in science and engineering fields (NSB, 1993:38). The Carnegie F oundation for the Advancement of Teaching has classified those institutions into categories according to the size of their bachelor's-degree and graduate programs, amount of research funding, and--for liberal-arts colleges--selectivity of admissions. Table B-12 shows how many institutions were in each category in 1991.

Although comprehensive and liberal-arts institutions were the majority of institutions granting science and engineering bachelor's and master's degrees, the research and doctorate institutions accounted for large fractions of the degrees granted: 54% of all science and engineering bachelor's degrees and 53% of the master's degrees. Degree production is especially concentrated at the doctoral level (see Table B-13): nearly two-thirds of the sc ience and engineering PhDs awarded in 1991 came from the 71 Research I universities, four-fifths from the 105 Research I and II universities; and nine-tenths from the 153 Research I and II and Doctorate I universities. The same set of 153 universities al so receives 90% of all academic R&D funding (NSB, 1993:40; Appendix Table 2-5).

The 105 Research I and II universities produce nearly 80% of the science and engineering doctorates awarded each year (see Table B-14). This concentration of PhD production differs some by bro ad field, although the 71 Research I institutions accounted for at least the majority of PhDs granted in each.

There was substantial growth in the number of institutions with graduate degree programs since 1961 (see Table B-15). The number of institutions granting doctorates doubled between 1961 and 19 91; the number of master's-degree institutions more than doubled.

SCIENCE AND ENGINEERING MASTER'S DEGREES

Master's Degrees in Science and Engineering

The number of master's degrees awarded in science and engineering has grown every year since 1966 except in the 1979-1981 period (Table 1 in NSF, 1994b). In 1991, over 78,000 science and engineering master's degrees were awarded, up from 41,000 in 1966. Science and engineering master's degrees were 29% of all master's degrees awarded in 1966, a percentage that declined to less than 21 in 1976, increased to 25% in 1987, and fell again to 23% in 1991.

According to Table B-16, about one-third of science and engineering master's degrees are awarded in the social and behavioral sciences, three-tenths in engineering, one-fourth in the natural sc iences, and one-tenth in the life sciences (NSF, 1994b). (Compared with 1966, the social/behavioral sciences increased their share by almost 9 percentage points, gaining 5 points from the life sciences, 3 from engineering, and 1 from the natural sciences .)

Women

Women earned 13.3% of the science and engineering master's degrees in 1966, a percentage that increased steadily to nearly 36% in 1991 (see Table B-17). This varied by field.

The number of science and engineering master's degrees per 1,000 24-year-old women in the US population increased from 4 in 1966 to 15 in 1991 (NSF, 1994b:Table 57). Meanwhile, the number going to men went from more than 30 per 1000 in 1969-1970 to 26 i n 1991 (after reaching a high of 32 in 1970 and a low of 21 in the early 1980s). Overall, in 1991, 21 master's degrees in science and engineering were awarded for every 1000 24-year-olds in the US population, up from 15 in 1966.

Underrepresented Minorities

NSF has collected information on the race, ethnicity, and citizenship of master's degree recipients biennially since 1977. Table B-18 shows that the proportion of science and engineering master 's degrees awarded to members of underrepresented minorities (i.e., blacks, Hispanics, and American Indians) has increased slowly in the natural sciences and engineering, offset by declines in psychology and the social sciences, fields that traditionally have registered the largest shares of underrepresented minorities.

Non-US Citizens

Non-US citizens with temporary visas received almost 20% of the science and engineering master's degrees in 1991, almost double their share in 1977 (see Table B-19). Most were in physical or m athematical science and engineering programs, where they constituted about 30% of all master's-degree students. They were relatively unlikely to be in social-science or psychology programs.

SCIENCE AND ENGINEERING DOCTORAL DEGREES

Doctorates in Science and Engineering

The number of science and engineering PhDs awarded annually tripled between 1958 and 1968 to about 18,000. Between 1968 and 1974, the number leveled off or fell in various fields, although the aggregate number peaked in 1971-1973 at nearly 19,400 per ye ar, and fell to fewer than 18,000 per year by 1977, where it remained through 1981. Table B-20 shows the numbers of science and engineering PhDs awarded during 1983-1993 period.

In 1982, the number of science and engineering doctorates awarded annually rose above 18,000 for the first time since 1976. The growth rate was still low-about 2% per year--until 1988, when the increase was more than 5% over 1987. The number increased by around 3% per year from 1988 through 1993. Table B-21 shows the increases from 1988 to 1993 by field.

Women

The number of women awarded science and engineering PhDs increased from 4,624 in 1983 to 7,537 in 1993, or 63.0%. As a result, the proportion of PhD awards to women increased from 25.1% in 1983 to 29.9% in 1993. They varied from field to field (see Table B-22).

Women earning science and engineering PhDs were concentrated in particular fields. In 1993, nearly half (45%) were in the social and behavioral sciences, and 30% were in the life sciences (see Tab le B-23). The concentration in social and behavioral sciences was reduced over the previous 10 years, however, as more awards to women were made in the physical and mathematical sciences and engineering. Overall, the shift was 4.5 percentage points (to nearly 18%). The number of women receiving PhDs in the physical and mathematical sciences more than doubled from 1983 to 1993 (from 617 to 1,344). The number in engineering quadrupled (from 124 to 521), but fewer than 7% of women PhDs were in engine ering.

The percentage of science and engineering doctorates awarded to members of underrepresented minorities_American Indians, blacks, and Hispanics--has been very low and has increased slowly (see Table B-24). In fact, the percentage of doctorates received by black citizens declined during the last half of the 1980s.

Non-US Citizens

In 1993, more than 8,000 science and engineering PhDs went to foreign citizens with temporary visas--nearly one-third of all the doctorates awarded by US universities (see Table B-25). Only 18 .5% of PhDs awarded in 1983 went to foreign citizen with temporary visas. In 1993, they received just under half of the new doctorates in engineering, up from 42% 10 years earlier. They were awarded more than one-third (36%) of the PhDs in the physical and mathematical sciences and more than one-fourth (28%) of those in the life sciences.

The number of non-US citizens with temporary visas increased 138% over a decade--from 3,400 in 1983 to 8,087 in 1993 (Table B-25). That increase was greater in the physical and mathematical sc iences and in the life sciences. The rate of growth was lower than average in the social and behavioral sciences.

Table B-27 shows the distribution of foreign citizens with temporary visas, by field. In 1993, roughly one-third were in engineering, and another one-third were in the physical/mathematical sc iences. Around 15-20% were in each of the life sciences and the social sciences. Compared with 1983, temporary-visa holders moved away from the social sciences (by about 4.5 percentage points) and into the physical/mathematical sciences (by about 2 poin ts) and the life sciences (by about 2.5 points). The country of origin of temporary-visa holders shifted during the 1980s (see Table B-28). Citizens from eastern Asia increased their share of science and engineering PhDs awarded to temporary -visa holders from one-fourth to more than a half. Nearly all this increase was accounted for by students from the People's Republic of China, who increased their share from near zero in 1982 to 22% of all the science and engineering PhDs earned by tempo rary-visa holders in 1992.

The annual reports of the Survey of Earned Doctorates conducted by the National Research Council for NSF have documented a substantial increase since 1970 in the time it has taken to obtain a PhD, whether measured in years since the bachelor's degree or in years registered in graduate school. According to these reports, which calculate the median time-to-degree (TTD) of all those obtaining PhDs each year, the figure has increased by about 30% over the last 20 years.

According to the latest report of the Survey of Earned Doctorates, total TTD (years from bachelor's degree to doctorate), or TTTD, went up by 29.6% from 1967 to 1993, from 8.1 to 10.5 years (Table Table B-29). The trend in registered time-to-degree (RTTD) was similar--31.5% (from 5.4 to 7.1 years). However, 1967 was near the postwar total-TTD low of 8 years reached in 1970. TTTD increased for decades (from 7 years in 1920 to 9 in 1962), fell during the 1960s to the low of 8 years in 1970, and then resumed its upward trend after 1970 (Bowen and Rudenstine, 1992:Figure 6.3). Those figures are for all PhDs in all fields, including humanities, education, and the professional (which have had the highest TTDs historically). The patterns vary widely by field, even within the sciences and engineering. Engineering and physical s ciences have always had shorter than average completion times; social sciences, longer.

The increase in TTD has slowed considerably since about 1987, even though the recession of the early 1990s might have increased the incentive to stay in school a year or two longer.

In conducting the research for their recent book In Pursuit of the PhD, Bowen and Rudenstine (1992:113-119), noticed that TTD figures were lower for their sample of 10 schools. They consulted demographers who suggested a different method f or determining TTD that should be more accurate. The method used by the Office of Scientific and Engineering Personnel (OSEP) and others determines the median number of TTD years for all those receiving their doctorates in a particular year. The demogra phers pointed out that this permits a bias if the cohorts entering graduate school are increasing or decreasing in size over time. Each entering class of PhD candidates has some fast finishers and some slow finishers. In a period such as the late 1950s and 1960s, when the number entering PhD programs was growing every year, the proportion of fast finishers showing up for their degrees a few years later increased and made the decreases in TTD larger than they would have been if cohorts had been steady. Similarly, when cohort sizes decrease, as they did in 1974-1984, the proportion of fast finishers getting their degrees a few years later goes down, increasing the apparent TTD. More recently, enrollments have gone up again, and that accounts for at leas t part of the decrease in TTD medians in the past several years.

Bowen and Rudenstine corrected for that bias by calculating average TTD of entering cohort, rather than graduating cohort. They asked, how long on the average, did it take those entering a PhD program (or getting their bachelor's degrees in year X to ge t their doctorates? They found that use of the entering-cohort method gave an increase in TTD of about 10% over the preceding 15-20 years, not 30%. They admitted that any lengthening in the already-long TTD is a serious problem but said that its magnitu de and newness had been exaggerated.

A study of TTD by staff of OSEP reviewed the literature on the causes of increasing TTD (Tuckman, et al., 1990). They found that earlier studies had looked at sociological, demographic, economic, and institutional factors, although few had looked at the m all and undertaken a causal analysis. They developed a model of TTD with five vectors of variables: family background characteristics, individual abilities and interests, tuition and financial aid, institutional environment and policies, and economic and social forces. They tested the model in 11 fields using data from the Survey of Earned Doctorates and found a variety of factors that affected registered time-to-degree (RTTD) or total time-to-degree (TTTD), including the availability and form of stu dent support, labor-market conditions, sociodemographic characteristics of the doctorate recipients, and characteristics of the undergraduate and graduate institutions. Yet no factor or set of factors consistently explained the general upward trend in TT D. That might be because TTD is poorly measured (the study was based on the graduating, rather than the entering, cohort), or because the data are inadequate. They are aggregate data, and some measure the variables of interest only indirectly; other var iables, such as increasing complexity of subject matter or the incentive for faculty to keep students longer as cheap labor on research projects, are not measured at all.

As for negative consequences, the following have been mentioned (Tuckman, et al., 1990):

Bowen and Rudenstine (1992) also studied the effects of financial support in some detail in their 10-school sample. They found that it mattered. Students who received financial aid had much higher completion rates and shorter TTD than students who reli ed on their own resources. In the sciences, the form of the aid had an effect on completion and TTD; research assistantships had the best effect, fellowships a close second, and teaching assistantships the worst effect. They also found that the NSF fello wship program had been very successful in reducing median TTD (4.9 years versus 5.6 years for those who were not NSF fellows in an eight-university group). Interpreting such findings is problematic, however. Did the NSF fellows finish earlier because of the fellowship form of support itself or because they were selected through a rigorous process that selected more-motivated students?

In conclusion, both RTTD and TTTD have been increasing for a long time, with the exception of the 1960s. Presumably, the increases are caused in part by the increasing complexity of knowledge and techniques to be mastered in doctorate programs and in pa rt by less-desirable or less-excusable reasons (e.g., an increase in tuition costs and a decrease in federal aid, which force students to work more during graduate school, or a desire of faculty to keep students working on research projects). They are al so caused in part by the increasing participation of women and minority-group members, who generally have longer TTDs.

According to Bowen and Rudenstine, having outside aid does improve completion and TTD rates. The form of the aid--fellowships, research assistantships, or teaching assistantships--might have little independent effect.

Source of Support

The Survey of Earned Doctorates (SED) administered yearly by OSEP for NSF asks new PhDs to list their primary source of support during graduate school. The data for 1993 are displayed in Table B-3 0. It should be noted, however, that the nonresponse rate to this question was 34%, for unknown reasons (it was 23% in 1991, and 30% in 1992). It also should be noted that federally funded research assistantships are listed with other research assis tantships under "university" because students often do not know the source of support for their research assistantships. Federal loans are listed under "personal." "Other" includes national fellowships, employer funds, and support from foreign governmen ts, state governments, and other nonspecified sources. The "life sciences" include "health science" PhDs as well as the biological and agricultural scientists listed in the other tables in this appendix.

A brief analysis of the table shows that a relatively large percentage of the PhD recipients in social sciences are self-supporting--nearly half, compared with 10-15% of those in the physical sciences and engineering and a fifth of those in the life scie nces.

PhDs in the life sciences receive the most direct federal support, probably resulting from the large fellowship and traineeship programs of the National Institutes of Health.

Most PhDs in the physical sciences and engineering, and to a lesser extent the life sciences, receive their primary support from their universities. That includes federally funded research assistantships, as well as other research and teaching assistant ships.

POSTDOCTORATE EMPLOYMENT PLANS

According to the SED, among new science and engineering PhDs who had definite postgraduation plans, the percentage planning to work in academe (college or university) was 48% in the early 1960s (NRC, 1978:Table 30). That figure increased to 57.0% in 197 0 before falling steadily to 44.1% in 1980 (NSF, 1993b:Table 15) and 40.4% in 1993 (NSF, 1994f:Table 7). Meanwhile, the proportion of new science and engineering PhDs going to business and industry grew from about 22% in the 1960s to 26.5% in 1970 and 36 .2% in 1993.

Note that Table B-31 does not include those with definite plans for postdoctoral study in the United States, almost all at universities. These numbered 2,789 in 1970, 3,571 in 1980, 4,676 in 1 990, and 5,739 in 1993 (NSF, 1993b:Table 15, 1994f:Table 7).

It also should be noted that the percentage of science and engineering PhDs who had definite plans at the time of the SED survey fell from 76.6% in 1970 to 72.0% in 1980, 64.0% in 1990, and 60.1% in 1993 (NSF, 1993b:Table 15, 1994f:Table 7).

POSTDOCTORAL STUDY TRENDS

B-32: Postdoctoral Study Plans of Recipients of Science and Engineering Doctorates from US Universities, 1985-1992

B-33: Postdoctoral Study Plans of Recipients of Science and Engineering Doctorates from US Universities, by Field, 1992

B-34: Science and Engineering Postdoctoral Appointees in Doctorate-Granting Institutions, by Field, 1982-1992

B-35: Trends in Net Growth of Science and Engineering Postdoctoral Appointee Positions in Doctorate-Granting Institutions, by Field, 1982 and 1992

B-36: Appointments of Postdoctoral Scientists and Engineers Who Were Not US Citizens in Doctorate-Granting Institutions, by Field, 1982 and 1992

B-37: Federally Supported Science and Engineering Postdoctoral Appointees in Doctorate-Granting Institutions, by Field, 1982 and 1992

B-38 Sources of Support for Science and Engineering Postdoctoral Appointees in Doctorate-Granting Institutions, by Field, 1992

Previous Section | HTML Home Page | Next Section

NAS Home Page | NAP Home Page | Reading Room | Report Home Page