The Best and Brightest for Science

Is There a Policy Problem Here?

WILLIAM ZUMETA AND JOYCE S. RAVELING

University of Washington

Key words: college entrance examinations, engineering education, enrollment trends, graduate students, graduate study, public policy, science careers, science education, scientific personnel

1. INTRODUCTION

Scientific research is exacting work, requiring not only long years of training but also high intellectual ability. Beyond this, one widely subscribed theory holds that within the scientific enterprise it is less the many “worker bees” than the few genuinely creative minds who make the most critical contributions (Kuhn 1962). In any case, few would argue but that the health of the enterprise requires a steady inflow of top-flight talent.

Since around 1993, graduate enrollments in most fields within the natural sciences and engineering1 have been declining, according to National Science Foundation data. It can be argued that this is simply a response to labor market signals and could be offset by recruiting more non-U.S. citizens. But the enrollment fall-off would more disturbing if it were found that the numbers of the very best U.S. students embarking on S/E graduate studies were falling faster than the total decline. Moreover, most careful observers would probably say that it is risky to cede away too many of the best minds raised here when many other countries’ universities and research institutions show signs of becoming increasingly attractive to their own natives. In addition, other developed countries appear to be competing more actively now than in the past in the international market for top scientific talent.

In this chapter, we present preliminary evidence showing some signs of a fall-off in interest in advanced studies[2] in science and engineering by top young U.S. talent, make some assumptions for purposes of the present discussion about what additional research may show, and then focus on exploring conceptually the broad contours of the policy options available to policymakers and the constraints operating upon these. To preview the conclusions, as with many problems there do not appear to be a set of feasible, low cost, high efficacy options to choose from in this complex policy arena.

First, it is not at all clear that simply facilitating the “natural” operation of the labor market by improving the information available to prospective students, although desirable and probably now technically feasible, will help much with the problem of declining attraction of the “best and brightest.” Indeed, it seems likely that it is market signals that are dissuading many in this key group from pursuing advanced S/E studies. Policies designed to work directly on the supply side of the problem, such as enhancing fellowship and other graduate student support, would likely have an impact but would also have serious unintended consequences. By themselves, such policies ignore the demand side roots of the problem and indeed are likely to lead to further supply/demand imbalances[3] in the longer run. While we find that certain modest supply side steps are desirable, the major conclusion is that a carefully targeted effort probably should be made on the demand side of the issue. We give some attention then to sketching the design parameters of a feasible policy response that provides some promise of enhancing the attractiveness of advanced studies in S/E to the nation’s most talented young college graduates.[4]

2. BACKGROUND: A BRIEF HISTORY OF THE UPS AND DOWNS OF THE S/E PHD LABOR MARKET

The essentials of this history are well known.[5] The Cold War, in particular the Sputnik era, the scientific and technical possibilities of the atomic age, and the need to teach the baby boom generation of college students produced a burst of demand for scientists and engineers in the late 1950s and the 1960s, fed in large measure by public R & D and higher education spending. The scientific, technological and career prospects this burst created, along with the graduate fellowships and assistantships it supported, stimulated a large gain in graduate enrollments in the S/E fields. PhDs awarded in the sciences and engineering approximately trebled from 1960 to 1970 (Fechter and Gaddy 1998).

A sharp cutback occurred in the growth rate of R & D spending in the late 1960s and 1970s. Some fields experienced real declines for several years. Also, in the early 1970s, college and university enrollments stopped growing rapidly as the cohorts associated with the big postwar jump in births passed on through the system. The federal government reduced its support of graduate fellowships quite abruptly (Breneman 1975). In response, PhD awards began to fall off considerably in engineering, mathematics, and the physical science disciplines and leveled off in the biological sciences (Fechter and Gaddy 1998; National Research Council, annual publication). Fewer graduate students created problems for the academic research economy which needs these relatively low cost apprentice scientists to produce its research products cost-effectively. Hence, partly as a replacement mechanism, the number of non-citizen graduate students in the S/E fields climbed sharply (although there had long been gradual gains in foreign graduate students and PhD recipients at U.S. universities). Such an increase was possible because U.S. science and universities were very attractive to foreign students.

Another noteworthy response to the abrupt downturn in demand developed in the S/E PhD labor market. For decades, a small share of young scientists, in large measure the most promising ones, opted for additional post-PhD training as postdoctoral appointees (sometimes called fellows or associates or simply “postdocs”) in academic or occasionally other research laboratories. [6] Postdoctoral appointments are temporary research appointments undertaken by young scientists, usually immediately after obtaining the PhD, ostensibly to deepen and/or broaden their experience in research. The usual pattern is to move to a different university from the one where the young scientist obtains the doctorate in order to maximize new learning and cross-fertilization of ideas. In the past, typically, the young scientist remained in this temporary student/research associate status for a year or two before moving on to a faculty or other career research position (NRC, 1969). As late as the 1960s, the proportion of new PhDs taking such immediate postdoctoral research/study appointments was fairly modest, 20-30% in chemistry, physics and most of the biosciences, though already around 50% in the biomedical disciplines. Historically, those young scientists who chose the postdoctoral route were the “cream of the crop,” disproportionately those from the best graduate departments, national fellowship winners and the like (NRC, 1969; Zumeta, 1985).

By the late 1970s a distinctly different pattern was developing in the postdoctoral ranks. First, the numbers of postdoctoral appointees and their share of the new PhD cohort climbed abruptly in a pattern coinciding closely with measures of the slackness of the PhD labor market (Zumeta 1985: especially pp. 22-26). Second, the typical time spent in the ostensibly temporary postdoctoral status began to grow (NRC 1981), suggesting that young scientists were backing up in a relatively stagnant postdoctoral pool with insufficient outlets. Third, when he examined the self-reported motivations of more recent postdocs (1970s) compared to those of earlier eras, Zumeta found that those of more recent vintage saw themselves not only as becoming or extending their postdoctoral stay in order to await improved market conditions but at the same time were investing further in their human capital, which they felt the market eventually would reward.[7] But Zumeta also found that the big increase in the number of postdoctoral appointees included not just those who, from all indications,[8] were the most promising of the PhD crop as in the past, but also for the first time a substantial number of young scientists who seemed to be less promising and to report that they became postdocs because they had few other options (Zumeta 1985; chapter 3).

In 1985, Zumeta expressed concern that the disappointing labor market returns to postdoctoral training might be dissuading the most able students from pursuing this training although it showed signs of improving their subsequent research productivity (see note 7). Although market conditions improved (from the standpoint of new PhD scientists) somewhat in the mid- and late 1980s – albeit in part because new PhD output had fallen off considerably in many of the S/E fields – this improvement was short-lived. Even as the economy boomed in the late 1980s, several analytical reports appeared forecasting a strong academic market for additional PhD scientists and engineers by the mid-1990s as colleges and universities saw 1960s era faculty members retire just as the children of the baby boom echo reached college age (Bowen and Sosa 1989; National Science Foundation 1989, 1990).

Graduate student enrollments responded fairly strongly and S/E PhD output again increased (Fechter and Gaddy 1998; Shapiro 1999). But, on the demand side, the economic setback of the early 1990s intervened and the much longer-lasting sluggishness and caution in university finances (Zumeta 1998). Again, S/E graduate students completing the PhD faced limited permanent job prospects, the pool of postdoctoral appointees reached unprecedented size (Commission Professionals in Science and Technology 1997; Association of American Universities 1998), and the length of postdoctoral stays grew (Regets 1999).[9] Meanwhile, the number of non-U.S. citizen graduate students and new PhDs in the S/E fields leveled off in the early 1990s after growing for many years (Fechter and Gaddy 1998: 358; Sanderson and Dugoni 1999: 18-19).

3. WHAT IS KNOWN ABOUT THE BEHAVIOR OF “TOP” STUDENTS

There is little to say about how the top tier of young scientists in particular have responded to these ups and downs for the issue has been little studied. Several single-institution studies have suggested that top baccalaureate graduates (by class rank, etc.) from certain elite institutions have been much less likely in recent years than their earlier counterparts to pursue graduate studies in the arts and sciences compared to professional schools (Goheen 1984; Rosovsky 1990). Reviewing a broader range of literature (but little specifically pertaining to the quality of those who pursue science and engineering studies), Bok (1993) draws a similar conclusion.

The most recent thorough empirical study directly pertinent to the issue was published by Hartnett (1985; 1987). He indicated that a motivation for his investigation was that the diminished prospects for academic careers might have hurt the relative attractiveness to top students of PhD studies in the arts and sciences compared to major professional fields (medicine, law, business). Thus, he sought to compare the quality, as best it could be measured, of then-recent cohorts of professional degree recipients in these fields with comparable cohorts of PhD recipients in eight arts and science disciplines, including chemistry, electrical engineering, mathematics and physics among the S/E fields. Because professional degree seekers and PhD seekers usually do not take the same graduate school tests, Hartnett’s procedure was to go back to a comparable national scale that most of them did have in common: the Scholastic Aptitude Test (SAT). After great expense and years of effort, he was able to construct a large sample of degree recipients from several cohorts and identify their SAT scores.[10]

In brief, Hartnett found that, in virtually all the arts and sciences fields (including all the natural sciences and engineering disciplines studied), the PhD recipients held a considerable SAT score advantage over the professional degree recipients and this gap changed little over the degree cohorts he studied: 1966, 1971, 1976, and 1981. The PhD recipients’ advantage was on the order of 25-35 points on each of the SAT scales (quantitative and verbal) for the three most recent cohorts, for which the data were more complete. In order to get test score data for both the PhDs and professional degree recipients, even his most recent cohort had taken the SAT test in the early 1970s so the data, while interesting, are now so old as to be of essentially historical interest. We are left with the question of how S/E has fared in the decision making of more recent cohorts, ideally of entering graduate students rather than completed degree recipients, since PhD completion takes so many years.

The same year Hartnett’s journal article was published, the National Academy of Sciences and its sister organizations convened a steering committee “to examine the issue of whether graduate departments in the sciences, engineering and mathematics are continuing to attract an appropriate share of the brightest students” (National Academy of Sciences, National Academy of Engineering, and Institute of Medicine steering committee report 1988). This group reported that it was quite concerned about the problem but did not have access to enough data or research resources to draw any firm conclusions and simply called for more research “into the nature and causes of the problem” (ibid.).

4. RECENT DEVELOPMENTS: NEW EVIDENCE ABOUT THE INTEREST OF THE “BEST AND BRIGHTEST” IN SCIENCE AND ENGINEERING

Graduate enrollments in virtually all the natural sciences and engineering fields other than biological sciences have been falling since about 1993.[11] Declines between 1993 and 1998 range from nearly six percent in computer science to over twenty-three percent in mathematical sciences (see Figure 1). Significantly, these declines are not confined to U.S. citizens[12] in most disciplines although the declines are generally steeper among citizens. Only in computer science, electrical engineering, and aerospace engineering have there been gains over this period in temporary resident students, and the gain in aerospace engineering did not occur until 1998. Aggregating across all the natural science and engineering fields, the graduate enrollment decline from 1993 to 1998 was 8.6% for U.S. citizens and 3.6% for non-citizens (temporary residents). The negative trend in temporary resident students may signal a new problem – or simply a fact of life – emerging for the U.S. academic science enterprise: we may no longer be able to count on replacing U.S. citizen graduate students with non-citizens.[13]

Figure 1. Changes in the Number of Natural Science & Engineering Graduate Students in Doctorate Granting Institutions, by Broad Field

Within these recent graduate enrollment declines in S/E fields, what can be said about trends in the quality composition of the U.S. citizen group? One piece of indirect evidence is provided by trends in students’ test scores. With the cooperation of the Educational Testing Service, we analyzed trends in the scores of all Graduate Record Examination (GRE) General Test examinees for selected years from 1989 through 1998.[14] Each examinee completes a questionnaire at the time he or she registers for the test that includes a question about intended field of graduate study which is the basis for the trend analyses reported here.[15] Of course, no one can claim with great confidence that high GRE scores are a highly valid predictor of unusual scientific talent or creativity but they are an important indicator used by most graduate department admissions committees and there are few alternatives to choose from with broad national coverage across disciplines and a comparable scale.[16]

Among the U.S. citizen examinees indicating intent to pursue S/E graduate studies, there is a general pattern of modest declines in mean GRE scores between 1989 and 1998, on all three of the General Test scales (see Table 1). On the verbal scale, the decline of 20-27 points in mean scores in the S/E fields over this period is quite comparable to the 26-point decline in mean scores of all examinees (bottom row). On the GRE analytical scale, mean scores of all examinees changed only a little over this period (-3 points). Four of the five S/E field categories experienced decreases greater than this, ranging from a decline of six points among would-be engineers to -9 to -20 points in physical sciences, mathematics, and computer science. However, there was a 7-point gain in the mean analytical score of students indicating intent to pursue graduate studies in the biological sciences. The pattern in the quantitative scale scores is somewhat similar but less steep: among all examinees mean scores fell just one point over the nine year period, while there were declines in three of the five S/E disciplines ranging from –6 points among would-be engineers to –14 points among prospective computer scientists. There was an 11-point decline among prospective physical scientists, virtually no change among those headed for mathematics, and a 10-point gain in mean scores for those seeking to become biological scientists.

Table 1. Mean GRE Scores of U.S. Citizen Examinees, by Intended Field of Graduate Study

Intended Field	Verbal				Quantitative				Analytical
	89	92	95	98	89	92	95	98	89	92	95	98
Biological Sciences	534	526	515	507	587	588	590	597	583	585	598	590
Math Sciences	537	527	516	511	693	694	686	694	637	628	630	620
Physical Sciences	545	534	518	510	643	631	625	632	604	593	606	595
Computer Science	536	528	519	516	650	643	634	636	610	598	602	590
Engineering	523	515	504	499	687	682	676	681	610	600	613	604
Behavioral Sciences	515	508	492	488	518	517	512	516	543	545	550	537
Social Sciences	497	489	475	473	487	485	480	485	514	516	520	511
Art	509	507	501	499	492	494	501	513	532	532	546	545
Other Humanities	567	561	551	547	530	532	531	537	563	567	577	566
Education	463	459	448	447	472	472	474	483	491	496	503	497
Health Science	484	472	458	455	509	503	507	521	527	522	533	529
Applied Biology	486	479	467	462	528	523	523	536	534	531	542	542
Other	494	499	478	477	500	508	510	520	525	535	544	537
Undecided	506	499	485	491	535	532	528	543	548	548	551	548
No Response	500	495	498	479	514	511	529	514	516	519	548	515
ALL FIELDS	507	500	487	481	532	528	525	531	541	541	548	538

Source: Educational Testing Service

These changes are not large (only a small fraction of the GRE standard deviation unit of 100 points). Additionally, mean scores could be misleading and may not be reflective of changes at the top end of the distribution with which we are most concerned. Table 2 shows the proportion of U.S. citizen examinees scoring above 700 on each scale.[17] For all examinees, these proportions of high scorers decreased by 2-3 percentage points over the 9-year period on the verbal and quantitative scales but changed little on the analytical scale. The main differences between the would-be natural scientists and engineers and those interested in other fields shows up in the quantitative scale. Here, decreases in proportions of high-scoring physical scientists and computer scientists drive down the composite figure for S/E by more than four percentage points over the period while there is no change in the proportion of high quantitative scorers headed for non-science fields (“Total: All Other Fields”). Note also that the proportions of high scorers on the quantitative and analytical scales among those headed for biological sciences increased.

Table 2. Proportion of High Scorers (³700) Among U.S. Citizen GRE Examinees, by Intended Field of Graduate Study

Intended Field	Verbal				Quantitative				Analytical
	89	92	95	98	89	92	95	98	89	92	95	98
Biological Sciences	7.6	6.0	4.8	4.1	18.1	17.3	17.9	19.5	17.8	17.0	20.4	20.8
Math Sciences	10.6	9.0	8.6	7.2	57.1	57.4	56.1	59.2	35.5	31.8	33.7	34.3
Physical Sciences	9.7	7.9	6.2	4.3	37.4	32.9	30.4	32.8	25.0	21.2	23.6	23.7
Computer Science	11.6	10.4	8.4	7.1	41.2	39.4	37.5	39.7	27.8	24.5	27.2	26.9
Engineering	5.9	4.8	3.4	3.4	53.5	52.3	50.1	53.6	26.3	23.3	27.0	27.6
TOTAL: NAT SCI & ENGR	8.2	6.7	5.2	4.5	41.6	40.0	35.8	37.2	25.2	22.4	24.8	25.0
Behavioral Sciences	6.4	5.1	3.9	3.6	8.7	7.9	7.3	7.5	11.4	11.1	12.3	12.5
Social Sciences	5.7	4.2	3.4	3.2	5.6	5.0	4.6	5.3	7.3	7.7	8.5	9.2
Art	6.4	5.3	4.5	3.7	5.6	5.6	6.4	7.5	9.3	9.2	12.4	14.6
Other Humanities	14.3	12.0	10.2	9.6	10.0	9.6	9.0	10.2	14.4	14.6	16.5	17.4
Education	2.5	20.	1.6	1.4	4.2	4.3	4.7	4.9	4.7	5.4	6.3	6.7
Health Science	3.0	2.0	1.3	1.1	5.9	4.7	5.0	6.5	7.0	6.3	7.8	9.4
Applied Biology	3.2	2.6	1.7	1.2	7.6	6.0	7.1	8.0	8.2	7.8	10.1	11.2
Other	5.9	5.9	3.7	3.2	7.1	7.5	7.5	8.1	10.0	11.5	12.1	11.9
TOTAL: ALL OTHER FIELDS	5.9	4.8	3.6	3.2	6.8	6.3	6.1	6.8	8.8	9.0	10.1	10.7
Undecided	6.8	5.7	4.3	4.7	14.1	13.0	11.5	14.3	13.8	13.3	13.1	14.8
No Response	7.4	5.6	5.7	4.1	12.1	10.6	12.6	10.3	9.6	9.8	14.0	10.7
ALL U.S. CITIZEN EXAMINEES	6.5	5.2	4.1	3.5	13.7	12.4	11.2	11.9	12.2	11.6	12.7	13.0

Source: Educational Testing Service

The broad patterns over the years from 1989 to 1998 are very similar if we narrow the definition of high scorers to focus on those scoring 750 or above (Table 3).[18] The proportion of very high scorers on the verbal scale is very low and declining similarly among both would-be S/Es and those headed for non-science fields. The proportion of very high scorers on the analytical scale changed little overall but declined slightly among S/Es, and the proportion of very high quantitative scorers is down around four percentage points among prospective S/Es[19] but little changed among those headed for other fields.

Table 3. Proportion of Very High Scorers (³750) Among U.S. Citizen Examinees, by Intended Field of Graduate Study

Intended Field	Verbal				Quantitative				Analytical
	89	92	95	98	89	92	95	98	89	92	95	98
Biological Sciences	2.4	1.7	1.2	1.0	8.2	6.1	6.3	7.6	8.5	8.1	10.4	8.9
Math Sciences	4.3	3.8	3.0	2.2	35.3	35.2	33.6	37.3	21.1	19.3	21.5	18.9
Physical Sciences	3.5	2.7	2.0	1.1	20.5	16.2	15.2	16.3	13.0	11.6	13.2	11.5
Computer Science	4.6	3.7	3.0	1.9	22.6	20.0	20.5	23.4	15.5	14.4	16.6	14.4
Engineering	1.8	1.4	0.9	0.7	30.7	28.2	25.9	29.4	12.9	12.0	14.8	13.8
TOTAL: NAT SCI & ENGR	2.9	2.2	1.6	1.1	23.3	20.8	18.0	19.5	13.0	12.0	13.8	12.1
Behavioral Sciences	2.2	1.5	1.1	1.1	3.6	2.9	2.5	2.8	4.9	5.2	5.9	4.9
Social Sciences	2.0	1.3	0.9	0.9	2.1	1.8	1.4	1.8	2.8	3.3	3.9	3.6
Art	2.1	1.5	1.4	1.4	2.3	2.0	2.0	2.9	3.7	4.3	6.0	6.2
Other Humanities	53.5	4.2	3.1	3.1	4.0	3.5	2.8	3.7	6.2	6.9	8.2	7.5
Education	0.7	0.5	0.4	0.4	1.7	1.6	1.7	1.8	1.6	2.1	2.7	2.4
Health Science	0.8	0.6	0.3	0.3	1.9	1.4	1.3	1.9	2.6	2.4	3.0	3.2
Applied Biology	0.9	0.9	0.4	0.4	2.9	1.8	1.6	2.1	3.0	3.4	4.2	4.3
Other	2.5	1.9	0.8	0.8	2.9	2.8	2.6	3.2	4.4	5.5	6.1	5.0
TOTAL: ALL OTHER FIELDS	2.1	1.5	1.0	1.0	2.6	2.2	1.9	2.4	3.5	4.0	4.6	4.1
Undecided	2.3	1.9	1.3	1.3	6.9	5.7	4.7	6.6	6.5	6.5	6.4	6.4
No Response	2.6	1.8	1.7	1.7	5.4	4.8	5.2	4.6	4.4	4.5	7.0	4.3
ALL U.S. CITIZEN EXAMINEES	2.3	1.7	1.2	1.2	6.7	5.6	4.7	5.2	5.5	5.5	6.2	5.4

Source: Educational Testing Service

Another way to view the data is to ask whether there have been notable changes over time in the distribution among fields of the high (³700) or very high scorers (³750). This analysis asks more directly about the relative attractiveness of S/E studies as compared to other fields to evidently high ability students. Again the findings are very similar whichever cutoff point is used (see Tables 4 and 5). In terms of top scorers on any of the three scales, S/E lost some ground over the last nine years. Comparing the 9-year changes in the shares of high scorers headed for S/E fields to those headed for other designated fields, all the S/E fields except biological sciences have lost ground while non-science fields (see “Total: All Other Fields”) have gained significantly in their attraction (at least to the point of stated intent) of top students.

Table 4. Distribution of High Scoring (³700) U.S. Citizen GRE Examinees, by Intended Field of Graduate Study

	Verbal Verbal				Quantitative Quantitative				Analytical
Intended Field	89	92	95	98	89	92	95	98	89	92	95	98
	%	%	%	%	%	%	%	%	%	%	%	%
Biological Sciences	3.9	33.7	4.6	5.3	4.4	4.4	6.2	7.4	4.9	4.7	6.3	7.3
Math Sciences	2.0	2.0	1.9	1.7	5.0	5.4	4.4	4.0	3.5	5.2	2.3	2.1
Physical Sciences	4.5	4.3	4.2	3.6	8.2	7.5	7.4	7.9	6.1	5.2	5.1	5.2
Computer Science	4.5	3.9	2.9	3.6	7.6	6.1	4.8	5.9	5.8	4.1	3.0	3.7
Engineering	5.7	5.5	3.8	4.1	24.2	25.1	20.2	19.4	13.4	12.0	9.6	9.1
TOTAL: NAT SCI & ENGR	20.5	19.3	17.4	18.3	49.4	48.6	43.0	44.7	33.7	29.2	26.3	27.4
Behavioral Sciences	15.8	16.8	15.8	16.5	10.0	10.9	10.6	10.2	14.8	16.3	15.9	15.5
Social Sciences	8.5	7.9	7.0	7.9	4.0	3.9	3.4	3.8	5.8	6.4	5.6	6.0
Art	1.8	1.8	1.6	1.6	0.8	0.8	0.8	0.9	1.4	1.4	1.4	1.7
Other Humanities	20.9	24.1	22.3	24.1	6.9	8.0	7.0	7.5	11.2	13.2	11.5	11.7
Education	5.8	5.4	5.1	538	4.6	5.0	5.5	5.9	5.9	6.6	6.6	7.4
Health Science	4.8	4.7	5.3	5.9	4.5	4.7	7.1	9.7	6.0	6.7	9.7	12.9
Applied Biology	0.9	0.8	0.7	0.6	1.0	0.8	1.1	1.3	1.2	1.1	1.4	1.7
Other	1.1	1.3	1.2	2.6	0.6	0.7	0.9	2.0	1.0	1.2	1.3	2.7
TOTAL: ALL OTHER FIELDS	59.6	62.8	59.1	65.0	32.3	34.8	36.5	41.3	47.3	53.0	53.3	59.5
Undecided/No Response/Other	19.9	17.9	23.5	16.7	18.3	16.6	20.5	14.0	19.0	17.8	20.4	13.1
TOTAL NUMBER >700 SCORERS	14,161	14,531	11,903	8,639	29,894	34,740	32,962	29,435	26,588	32,381	27,312	32,184

Table 5. Distribution of Very High Scoring (³750) U.S. Citizen GRE Examinees, by Intended Field of Graduate Study

Intended Field	Verbal				Quantitative				Analytical
	89	92	95	98	89	92	95	98	89	92	95	98
	%	%	%	%	%	%	%	%	%	%	%	%
Biological Sciences	3.5	3.2	4.1	5.2	4.1	3.5	5.3	6.5	5.2	4.6	6.6	7.4
Math Sciences	2.3	2.6	2.3	2.1	6.4	7.4	6.4	5.7	4.6	4.1	3.1	2.8
Physical Sciences	4.6	4.5	4.7	3.5	9.1	8.2	8.9	8.9	7.1	6.0	5.8	6.1
Computer Science	5.2	4.3	3.7	4.0	8.5	6.9	6.3	7.9	7.1	5.0	3.8	4.7
Engineering	5.1	5.0	3.5	3.6	28.4	30.2	25.0	24.1	14.5	13.0	10.7	11.0
TOTAL: NAT SCI & ENGR	20.6	19.7	18.3	18.4	56.5	56.2	51.9	53.1	38.4	32.7	30.0	32.0
Behavioral Sciences	15.7	15.7	14.9	16.3	8.5	9.0	8.7	8.6	14.0	16.2	15.5	14.5
Social Sciences	8.6	7.6	6.6	7.1	3.0	3.2	2.6	3.0	4.9	5.8	5.2	5.6
Art	1.7	1.6	1.7	1.6	0.6	0.6	0.6	0.8	1.3	1.4	1.4	1.7
Other Humanities	23.0	26.3	23.7	25.9	5.6	6.5	5.4	6.2	10.6	13.1	11.6	12.0
Education	4.9	4.4	4.7	4.1	3.8	4.0	4.8	5.0	4.4	5.4	5.6	6.5
Health Science	3.7	4.3	4.5	5.5	3.0	3.0	4.4	6.4	5.0	5.3	7.8	10.5
Applied Biology	0.7	0.9	0.6	0.6	0.8	0.5	0.6	0.8	1.0	1.0	1.2	1.5
Other	1.3	1.4	0.9	2.7	0.5	0.6	0.7	1.8	1.0	1.2	1.3	2.7
TOTAL: ALL OTHER FIELDS	59.7	62.1	57.7	63.8	25.8	27.4	27.7	32.5	42.0	49.4	49.6	55.0
Undecided/No Response/Other	19.7	18.2	24.0	17.8	17.7	16.4	20.4	14.4	19.6	17.9	20.4	13.0
TOTAL NUMBER >750 SCORERS	4,919	4,648	3,432	2,146	14,640	15,584	13,726	13,012	12,030	15,443	18,234	13,400

Source (both charts): Educational Testing Service

All these proportions and shares are based on the total number of U.S. citizen GRE examinees in a given year. It is important to note also that the number of GRE test-takers has fallen off considerably in the last few years (Figure 2). The number of U.S. citizen General Test examines was around 219,000 in 1989, climbed steadily to a peak of 301,000 in 1994, then fell in each succeeding year to a level of around 248,000 in 1998. Numbers of examinees by intended graduate study field for the four years for which detailed data were available for this analysis are shown in Table 6. Interestingly, among these four years, the peak year for total examinees was 1995, with about 293,000 examinees (just 2.6% below the 1994 peak). But, among the available years, the peak for examinees reporting intent to study S/E fields was earlier, in 1992, with the 1998 figure 16% lower and slightly below the figure for 1989. Note that biological sciences again run counter to the S/E trend.

Figure 2. Total Number of U.S. Citizens Taking the GRE General Test, by Year

Table 6. Number of U.S. Citizens Taking GRE General Test, by Intended Field of Graduate Study

Intended Field	1989	1992	1995	1998	Change 92 - 98	% Change
Biological Sciences	7,287	8,867	11,516	11,222	2,355	26.6%
Mathematical Sciences	2,641	3,271	2,599	2,007	-1,264	-38.6%
Physical Sciences	6,521	7,928	8,061	7,117	-811	-10.2%
Computer Science	5,501	5,417	4,184	4,393	-1,024	-18.9%
Engineering	13,544	16,687	13,248	10,634	-6,053	-36.3%
TOTAL: NAT SCI & ENGR	35,494	42,170	39,608	35,373	-6,797	-16.1%
Behavioral Sciences	34,686	47,762	48,005	39,802	-7,960	-16.7%
Social Sciences	21,067	27,022	24,693	21,165	-5857	-21.7%
Art	4,094	4,911	4,238	3,692	-1,219	-24.8%
Other Humanities	20,619	29,228	25,930	21,615	-7,613	-26.0%
Education	33,097	40,101	38,633	35,687	-4,414	11.0%
Health Science	22,814	34,342	46,828	44,398	10,056	29.3%
Applied Biology	3,856	4,737	5,028	4,771	34	0.7%
Other	2,608	3,247	3,987	7,157	3,910	120.4%
TOTAL: ALL OTHER FIELDS	142,841	191,350	197,342	178,287	-13,063	-6.8%
Undecided or No Response	75,930	46,177	56,268	34,446	-11,731	-25.4%
TOTAL: U.S. CITIZEN EXAMINEES	218,771	279,697	293,218	248,106	-31,591	-11.3%

Source: Educational Testing Service

Juxtaposing the decline in numbers of test-takers headed for S/E fields with the earlier-discussed modest fall in proportions of would-be S/Es who scored highly produces a picture of apparent decline in interest in S/E among top students that is more complete and more sobering (see Tables 7 and 8). Compared to the 1992 peak, the number of high scorers on the quantitative scale headed for S/E fields in 1998 is down 21-22% (depending upon whether the 700 or 750 score cutoff is used) while the number of such top students headed for other designated fields (“Total: All Other Fields”) has been essentially stable over this period.[20] Moreover, the 20%+ declines in top-scorers headed for S/E fields in aggregate is net of a large gain in such students headed for the biological sciences. All the other S/E field categories show declines in top scorers, ranging from –4% in number of 750+ scorers headed for graduate computer science programs to declines of more than one-third in both 700+ and 750+ scorers headed for mathematics and engineering. Thus, the declines in top GRE scorers headed for graduate studies in the natural sciences and engineering, biological sciences excepted, appear to be consistent (not likely to be a single-year aberration), sizeable, and disproportionate, i.e., larger than the declines in all GRE examinees and in all those who say they are headed for S/E fields.

5. WHERE ARE THE TOP STUDENTS GOING?

If top students appear to be turning away from graduate studies in the sciences and engineering (other than biological sciences), where are they going instead? This is a very important, but complex, question to answer. A clue about a part of the answer may come from examining the trends in the indicated graduate study fields of top GRE scorers, although able students may pursue professions not requiring the GRE, or may opt to enter the workforce.[21] To begin with GRE data, Tables 7 and 8 show that in the “non-science” field set, the only specified field with a notable increase in top scorers is “health science” (ETS terminology).

This field grouping is dominated by applied, master’s-level health professions and includes few basic science specialities or research-oriented doctoral degrees. Thus, we group it with non-science fields, not the S/E fields. The dominant fields, in terms of numbers of graduate students and recent growth in enrollments, making up the health category are speech/language pathology, physical therapy, nursing, veterinary medicine, and public health. The first two professions accounted for 48% of the total gain over 1989 to 1998 in numbers indicating intent to pursue graduate work in the “health sciences” category. Table 7 shows that, among 1998 GRE test-takers scoring above 700 on the quantitative scale, more headed for these applied health professions than for the biological sciences, and the increase in high scorers was about twice as large in absolute terms in the health fields. At the 750+ level (Table 8), the numbers of potential biological scientists and health professionals were similar but the health professions experienced more growth.

Table 7. Number of High Scoring (³700) U.S. Citizens GRE Quantitative Examinees, by Intended Field of Graduate Study

Intended Field	1989	1992	1995	1998	Change 92 - 93	% Change
Biological Sciences	1,315	1,529	2,044	2,178	649	42.2%
Mathematical Sciences	1,495	1,876	1,450	1,177	-699	-37.5%
Physical Sciences	2,451	2,606	2,439	2,325	-281	-10.8%
Computer Science	2,272	2,119	1,582	1,737	-382 -3,010	-18.0% -34.5%
Engineering	7,234	8,720	6,658	5,710
TOTAL: NAT SCI & ENGR	14,768	16,849	14,174	13,127	-3,722	-22.0%
Behavioral Sciences	2,989	3,787	3,494	3,002	-785	-20.7%
Social Sciences	1,196	1,355	1,121	1,119	-236	-17.4%
Art	239	278	264	265	13	4.7%
Other Humanities	2,063	2,779	2,307	2208	-571	-20.5%
Education	1,375	1,737	1,813	1,737	0	0.0%
Health Science	1,345	1,633	2,340	2,855	1,222	74.8%
Applied Biology	299	278	363	383	105	37.8%
Other	179	243	297	589	346	142.4%
TOTAL: ALL OTHER FIELDS	9,686	12,090	11,998	12,157	67	0.6%
Undecided or No Response	5,440	5,801	6,790	4,151	-1,650	-28.4%
TOTAL: U.S. CITIZEN EXAMINEES	29,894	34,740	32,962	29,435	-5,305	-15.3%

Source: Educational Testing Service

Table 8. Number of Very High Scoring (³750) U.S. Citizens GRE Quantitative Examinees, by Intended Field of Graduate Study

Intended Field	1989	1992	1995	1998	Change 92 – 98	% Change
Biological Sciences	586	545	727	846	301	55.2%
Mathematical Sciences	937	1,153	878	742	-411	-35.6%
Physical Sciences	1,332	1,278	1,222	1,158	-120	-9.4%
Computer Science	1,244	1,075	865	1,028	-47	-4.3%
Engineering	4,158	4,706	3,432	3,136	-1,570	-33.4%
TOTAL: NAT SCI & ENGR	8,257	8,757	7,124	6,909	-1,848	-21.1%
Behavioral Sciences	1,244	1,403	1,194	1,119	-284	-20.2%
Social Sciences	439	499	357	390	-109	-21.8%
Art	88	93	82	104	11	11.8%
Other Humanities	820	1,013	741	807	-206	-20.3%
Education	556	623	659	651	28	4.5%
Health Science	439	468	604	833	365	78.0%
Applied Biology	117	78	82	104	26	33.3%
Other	73	94	96	234	140	148.9%
TOTAL: ALL OTHER FIELDS	3,776	4,271	3,815	4,242	-29	-0.7%
Undecided or No Response	2,607	2,556	2,787	1,861	-695	-27.2%
TOTAL: U.S. CITIZEN EXAMINEES	14,640	15,584	13,726	13,012	-2572	-16.5%

Source: Educational Testing Service

Students headed for medicine, law, and business generally take specialized tests designed for the graduate schools of their profession rather than the GRE.[22] It seems plausible that the decline in GRE test taking might be accompanied by a corresponding rise in numbers of examinees taking these other tests. Ideally, one would want to look at such patterns by age or baccalaureate cohort but the available data are limited to recent trends in numbers of professional school applicants, matriculants, test-takers, and mean scores. From these data, law schools seem unlikely to be attracting many students who might otherwise opt for S/E careers. Between 1992 and 1998 total applicants (unduplicated count of individuals applying) to U.S. law schools dropped by 26,000 (27%) and new matriculants decreased slightly. Mean LSAT scores were virtually unchanged over the period while those for matriculants actually fell by two points.[23]

Such indicators as are available regarding interest in medical school and quality of applicants suggest that some more top students may have headed in this direction in recent years. Generally, numbers of top students and total applicants have tended to move in similar directions in graduate professional fields (Adelman 1985; Bok 1993). This pattern may have changed in medicine recently. Medical school applicants (unduplicated number of individuals applying) jumped from around 37,400 in 1992 to 47,800 in 1996, but fell back to 38,500 in 1999.[24] The numbers of new matriculants remained virtually level, at around 16,200 entering medical students annually between 1992 and 1999. This means that, however much interest there is, educational opportunities in medicine are not growing. Both applicant and matriculant mean scores on the MCAT gradually increased throughout the 1990s which does not prove that more top students are taking the test but is suggestive. In short, medicine may have attracted the interest in recent years of more of the top students who could have pursued S/E research careers, but it is not clear that they are all being accommodated in the medical schools since enrollments are flat. Perhaps some of these students are “spilling over” into the non-M.D. health professions mentioned earlier.

There are also mixed signs in the area of graduate business education. The number of U.S. citizen GMAT examinees declined from 154,000 in 1990 to under 129,000 in 1995 before climbing back to over 142,000 in 1997.25 The numbers of S/E majors taking the GMAT followed a similar pattern with over 31,200 examinees in 1990, 23,200 in 1993, and slightly over 27,650 in 1997. Among the S/E majors, the number of computer science majors taking the GMAT decreased by nearly 33 percent, from roughly 4,200 in 1990 to 2,800 in 1997. This decrease was offset by a near 38 percent increase in the number of biological sciences majors, with about 2,950 taking the GMAT in 1990 compared to 4,100 in 1997. This increase in biological sciences majors may reflect the influence of the perceived employment market in businesses related to biotechnology and similar fields.

Significantly, after relative stability between 1990 to 1992, the average overall GMAT scores climbed by over 30 points on a 500-point scale between 1993 and 1999. Likewise, the annual mean score on the quantitative sub-scale increased, with the mean for S/E majors increasing 4.8% while the mean for non-S/E students increased only 1.1%.

Although the GMAT examinees decreased between 1990 and 1997, the numbers of MBAs conferred increased 27 percent.26 The number of degrees conferred grew steadily from roughly 73,000 degrees in 1990 to 97,600 in 1997. Together with the renewed growth in GMAT test-takers with S/E backgrounds, and the increasing scores of this group, the data suggest that graduate business schools are likely providing serious competition to S/E graduate programs for the top students.

6. GENDER AND RACE/ETHNICITY DIFFERENCES

Significantly, the declines in numbers of high GRE scorers headed for S/E graduate studies are concentrated among males and whites. Between 1989 and 1998, numbers of male U.S. citizens indicating intent to pursue graduate studies in S/E and scoring 700 or greater on the GRE quantitative scale have fallen by 4,000 (-19%), as shown in Table 9. The only S/E field showing an increase in interest among male high scorers was biological sciences. Also, the numbers of high scoring males indicating plans for graduate study in other specified fields (i.e., outside natural sciences and engineering) grew by 725, or nearly 13%, as the number of high scorers interested in S/E declined. The number of high scoring women headed for S/E fields has increased, but fairly modestly, by around 300 individuals per year, or less than 10%, for all the S/E fields combined. There were decreases among both women and men in mathematical sciences, physical sciences, computer science, and engineering (though the change was very small for women in engineering). In sum, increases in high-scoring women have not offset the much larger declines in the number of high scoring men headed for S/E. For both sexes, however, biological sciences is an important exception: it continues to attract healthy increases in top scoring students of both sexes.

Table 9. Number of U.S. Citizen Examinees Scoring ³ 700 on GRE Quantitative Test, by Intended Field of Study and Gender*

Intended Field	1989		1998		Change 1989 - 1998		% Change
	Male	Female	Male	Female	Male	Female	Male	Female
Biological Sciences	752	556	1,219	965	467	409	62.1%	73.6%
Math Sciences	1,002	511	757	434	-245	-77	-24.5%	-15.1%
Physical Sciences	1,921	511	1,643	683	-278	172	-14.5%	33.7%
Computer Science	1,859	404	1,514	228	-345	-176	-18.6%	-43.6%
Engineering	6,056	1,166	4,524	1,150	-1,532	-16	-25.3%	-1.4%
TOTAL: NAT SCI & ENGR	11,590	3,148	9,657	3,460	-1,923	312	-16.7%	9.9%
TOTAL: ALL OTHER FIELDS	5,701	3,905	6,426	5,673	725	1,768	12.7%	45.3%
Undecided/No Reponse/Other	5,571	1,809	2,400	1,714	-3,171	-95	-56.9%	-5.3%
TOTAL: NUMBER > 700 SCORERS	22,862	8,862	18,465	10,847	-4,419	1,985	-19.3%	22.4%

*Excludes examinees who did not indicate gender

Source: Educational Testing Service

Turning to the race/ethnicity categories, the numbers of cases are very small so data for all natural science and engineering fields have been combined (Table 10). There have been healthy percentage increases between 1989 and 1988 in the numbers of high scoring students (³700 on the GRE quantitative scale) headed for S/E in each race/ethnicity group: Asian-Americans- +22%; African-Americans- +50%; and Hispanic-Americans- +50% (see lower right panel). But the original bases were quite small, so the additional absolute numbers of high scorers are also small: 283 additional Asian-Americans in 1998 compared to 1989, 203 additional African-Americans and Hispanics combined. This total gain of less than 500 minority high scorers is more than offset by a decline of more than 1,700 white high scorers headed for S/E fields. Also rates of increase were greater in high scorers headed for non-S/E fields (“All Other Fields”) than in those headed for S/E among all the race/ethnicity groups.

Table 10. Number of U.S. Citizens Scoring ³700 on GRE Quantitative Scale, by Intended Broad Area of Graduate Study and Ethnicity,* 1989 and 1998

Broad Field	1989				1998
	White	Asian	Af Am	Hisp	White	Asian	Af Am	Hisp
Nat Sci & Engr	11,985	1,279	151	267	10,250	1,562	226	400
All Other Fields	8,346	433	59	150	9,971	1,100	140	275
Undecided/No Response/Other	3,930	292	37	43	3,076	525	47	89
TOTAL NUMBER ³700 SCORERS	24,261	2,004	247	460	23,297	3,187	413	764
Broad Field	Change between 1989 – 1998				% Change
	White	Asian	Af Am	Hisp	White	Asian	Af Am	Hisp
Nat Sci & Engr	-1,735	283	75	133	-14.5%	22.1%	49.7%	49.8%
All Other Fields	1,625	667	81	125	19.5%	154.0%	137.3%	83.3%
Undecided/No Response/Other	-854	233	10	46	-21.7%	79.8%	27.0%	107.0%
TOTAL NUMBER ³700 SCORERS	-964	1,183	166	304	-4.0%	59.0%	67.2%	66.1%

*Excludes examinees who indicated ethnicity as “Other” or did not answer.

Source: Educational Testing Service.

7. QUALITY TRENDS in NEWLY-ENROLLED GRADUATE STUDENTS

While these trend data are interesting and suggestive, tracking trends in field designations of GRE test-takers is inherently limited in its ability to identify accurately changes in patterns of graduate enrollment by top students. In addition to the limitations on GRE validity as a predictor of scientific potential, we cannot be sure how precisely examinees’ stated intent about graduate field of study matches with who actually enrolls in graduate school. For example, it may be that decreases in quality potential students of the magnitude shown are not large enough to affect enrollments of highly capable students, given selective admissions.

To begin to address this issue, we secured access to a database containing quality indicators27 on successive cohorts of newly-enrolled graduate students in leading departments in ten academic disciplines between 1989 and 1996.28 Designed to study graduate attrition and retention, this data file was developed by the Association of Graduate Schools of the Association of American Universities (AAU), a national organization of leading research universities, and is maintained by the Educational Testing Service. Of the ten academic disciplines, this report will concentrate on the five in natural sciences and engineering: biochemistry, mathematics, physics, chemical engineering and mechanical engineering. Data were not collected on physics and chemical engineering until 1992, so only five year trends are available for these disciplines.

Reporting for individual fields and years is much less than complete so we first eliminated all institutions with substantial reporting gaps that would make meaningful trend analysis not feasible in the fields of interest. The remaining 32 universities were evaluated individually for their GRE score reporting rates in each S/E discipline on both the GRE quantitative and analytical scales.29Since the data file was developed for other purposes, this required substantial effort to isolate the variables relevant for a trend analysis of GRE scores of newly-enrolled graduate students by S/E field. Each institution’s annual reporting rates for the GRE quantitative and analytical scores were evaluated separately for U.S. citizen plus permanent resident first-year graduate students in each S/E discipline.30 This yielded 10 separate sets of institution-by-institution reporting rates, one for each GRE scale for each of the five S/E disciplines.

Our goal was to insure a high level of comparability across years in the institutions being compared. Thus, to be included in the trend analysis for a discipline, an institution had to achieve high rates of individual GRE score reporting over all years of data collection—1989 through 1996 for biochemistry, mathematics, and mechanical engineering, and 1992 through 1996 for chemical engineering and physics. Institutions were retained if they had roughly a 70% or higher individual reporting rate for all years, although in a very few instances, we included an institution with one year below the 70% reporting threshold when there were high reporting rates in all other years. Low reporting rates in one field or on one GRE scale did not exclude an institution from all analyses, instead it was included for the disciplines for which it passed the reporting rate threshold. Between 17 and 24 institutions qualified for analysis in the five S/E disciplines with a total of 10,155 individual records included on the GRE quantitative scale and 10,077 records on the GRE analytical scale. Appendix A lists the institutions included in the trend analyses, by discipline and GRE scale.

7.1 Changes in GRE Quantitative and Analytical Scores of Newly-Enrolled S/E Graduate Students

In general, the mean GRE quantitative and analytical scores for newly enrolled U.S. citizen and permanent resident graduate students in the five S/E disciplines do not reveal consistent upward or downward trends, although there are hints that mean analytical scores may be increasing in both chemical and mechanical engineering. Certainly there is no indication of a downward trend in student mean test scores in any of the disciplines.

The GRE quantitative mean score for graduate students enrolled in biochemistry was 16 points higher in 1996 than in 1989 (see Figure 3). However, the annual mean score changes from 1989 to 1995 were quite small and the 1989 and 1995 mean scores were identical at 680. The 16-point increase actually occurred between 1995 and 1996, and one year clearly does not provide a sufficient basis to draw any conclusion about trend. The mean GRE analytical test scores for biochemistry students mirror the quantitative scores. The mean score was 650 in 1989 and 651 in 1995, but jumped to 679 in 1996. All other yearly score changes were small.

# of individuals (Q)	147	169	165	159	147	189	150	150
# of individuals (A) Scale	143	137	157	159	149	147	189	150

* Universities meeting our GRE Score reporting rate threshold. See Appendix A for institutions included.

Source: Educational Testing Service

Figure 3. Mean GRE Quantitative and Analytical Scores for Newly Enrolled Graduate Students in Biochemistry at 20 Selected AAU Universities,* by Year of Entry

The trend-lines for GRE quantitative and analytical mean scores for mathematics graduate students were also largely flat (see Figure 4). The mean quantitative score was 752 in 1989 and 753 in 1996. Changes during the intervening years were small; and even the 10-point increase between 1991 and 1992 was followed by offsetting decreases in succeeding years. The mean GRE analytical score of mathematics students increased 13 points between 1989 and 1996, starting at 701, peaking at 718 in 1994, and ending at 714 in 1996. Again, there is no clear sign of a trend here.

Only five years of data were available for physics graduate students (1992 – 1996), and again clear trends in the mean GRE scores do not emerge (see figure 5). Although the mean quantitative score started at 753 in 1992 and then dropped each year before leveling at 748 in 1995 and 1996, the decrease of only 5-points is insufficient to conclude there is a meaningful downward trend. The mean analytical score for physics students was 688 in 1992 and 700 in 1996. However, there was little movement after the 15-point gain between 1992 and 1993.

# of individuals (Q)	254	305	279	290	284	267	277	244
# of individuals (A)	249	304	269	289	284	267	276	239

* Universities meeting our GRE Score reporting rate threshold. See Appendix A for institutions included.

Source: Educational Testing Service

Figure 4. Mean GRE Quantitative and Analytical Scores for Newly Enrolled Graduate Students in

Mathematics at 23 Selected AAU Universities,* by Year of Entry

# of individuals (Q)	366	310	255	265	248
# of individuals (A)	366	310	255	265	248

* Universities meeting our GRE Score reporting rate threshold. See Appendix A for institutions included.

Source: Educational Testing Service

Figure 5. Mean GRE Quantitative and Analytical Scores for Newly Enrolled Graduate Students in Physics at 22 Selected AAU Universities,* by Year of Entry

There were also only five years of GRE data (1992 to 1996) available on new graduate students in chemical engineering. The pattern of the mean GRE quantitative test scores is nearly flat, starting at 736 in 1992 and ending at 740 in 1996 (see Figure 6). Over the five years, the mean analytical scores of chemical engineering graduate students increased 37 points, from 679 in 1992 to 716 in 1996. While the mean scores increased each year, 26 points of the increase occurred between 1995 and 1996. The consistent year-to-year gains and the relatively large five-year change imply a meaningful upward trend in this scale at least.

# of individuals (Q)	188	182	174	169	163
# of individuals (A)	188	182	174	171	163

* Universities meeting our GRE Score reporting rate threshold. See Appendix A for institutions included.

Source: Educational Testing Service

Figure 6. Mean GRE Quantitative and Analytical Scores for Newly Enrolled Graduate Students in Chemical Engineering at 19 Selected AAU Universities*, by Year of Entry

# if individuals (Q)	266	236	279	257	341	307	259	238
# of individuals (A)	261	233	263	257	341	305	255	237

* Universities meeting our GRE Score reporting rate threshold. See Appendix A for institutions included.

Source: Educational Testing Service

Figure7. Mean GRE Quantitative and Analytical Scores for Newly Enrolled Graduate Students in Mechanical Engineering at 17 Selected AAU Universities, by Year of Entry

Finally, the patterns of GRE quantitative and analytical test scores of newly enrolled mechanical engineering graduate students resemble those in chemical engineering (see Figure 7). The mean on the GRE quantitative scale was relatively stable, decreasing only 2 points from 737 to 735 between 1989 and 1996. The peak was 744 in 1992 and the lowest mean score was 730 in 1994. The mean analytical scores in this discipline rose from 659 in 1989 to 675 in 1996. The mean score fell below 659 only once (1990), and scores peaked in 1996. Again, the consistent upward trend in this scale is noteworthy.

Overall, mean GRE scores of entering U.S. citizen graduate students in these five S/E disciplines have been fairly stable over the period studied. Certainly, there is no evidence of a significant decline in this measure of student quality.

7.2 Numbers of High and Very High GRE Scorers Enrolled in S/E Disciplines

The AGS/AAU data also permits analysis of trend in numbers of high (³ 700) and very high (³ 750) scorers among newly-enrolled graduate students in the five S/E disciplines. With one clear exception, these departments are evidently not enrolling fewer top students, at least through 1996.

There were modest increases in the numbers of top scoring graduate students enrolling in biochemistry (see Figure 8), although the aggregate numbers are so small it is difficult to draw strong conclusions about trends. Consistent with the undulating pattern of mean GRE quantitative and analytical scores of biochemistry students (see Figure 3), the actual numbers of biochemistry students earning high and very high scores on these scales also rose and fell, but the underlying trends were modestly upward. Also consistent with the mean score pattern, the largest increases in numbers of high-scoring students occurred between 1995 and 1996.

The numbers of graduate students in chemical engineering earning high and very high GRE quantitative scores changed little between 1992 and 1996 while the numbers earning high and very high scores on the analytical scale increased by as much as 75 percent for ³ 700 scorers.

In mathematics, the numbers of new graduate students who had earned high and very high scores on the GRE quantitative and analytical scales were generally stable over the seven years for which data were available. There were 217 high scorers and 170 very high scorers on the quantitative scale in 1989 with 221 high scorers and 169 very high scorers in 1996. On the analytical scale, there were 150 high and 105 very high scorers in 1989, and 163 high and 118 very high scorers in 1996. The annual numbers of top scorers rose or fell slightly each year but no clear increasing or decreasing trends emerged.

In mechanical engineering, there were 208 newly enrolled graduate students earning high scores and 151 earning very high scores on the GRE quantitative scale in 1989. The numbers of top scorers peaked at 269 in 1993 before falling to 192 and 132 respectively in 1996. Although no clear trend emerged, the numbers of these top scorers on the GRE quantitative scale fell a bit each year between 1993 and 1996. Additional years of data might show this pattern to be indicative of a meaningful trend. On the analytical scale, the numbers of high and very high scorers increased minimally between 1989 and 1996.

* High Scorers earned ³ 700 and Very High Scorers Earned ³ 750 on the GRE scales

Source: Educational Testing Service

Figure 8. Number of Newly Enrolled Graduate Students in Biochemistry Who Earned High and

Very High Scores* on the GRE Quantitative and Analytical Scales, by Year of Entry

Of the five S/E fields, only physics showed an appreciable decline in the number of high and very high scorers among the newly enrolled graduate students (see Figure 9). This decline was especially clear for the quantitative scale where there was a 27 percent drop-off in high scorers and a 31 percent decrease in very high scorers over the five years for which data were available. This follows the decreasing trend in total numbers of new physics students enrolling in the AAU departments we examined.

* High Scorers earned ³ 700 and Very High Scorers earned ³ 750 on the GRE scale.

Source: Educational Testing Service

Figure 9. Number of Newly Enrolled Physics Graduate Students Who were High and Very High Scorers* on the GRE Analytical Scale, by Year of Entry

Although the data are limited in coverage and the numbers are small, the picture that emerges is broadly consistent with the earlier analysis of trends in high scorers among all U.S. citizen and permanent resident GRE-takers who plan to enroll in these S/E fields. The total enrollments of graduate students in the five fields in the AAU database have declined, but the numbers of top scorers have remained largely stable, except in physics which shows declining enrollment of top students. Biosciences, such as biochemistry, seem to be attracting an increasing number of top scorers. In general, the concern that fewer top GRE-scorers indicating interest in science would mean fewer top students in entering classes is not borne out in the recent trend data from select AAU departments (physics excepted).

Clearly, it would be desirable to have data on more departments in each field, broader field coverage of the S/E fields, and a continuation of the trend data on newly-enrolled students through more recent years to fully understand the emerging patterns. Since graduate students provide a large part of the academic research workforce in these disciplines, decreased enrollments will eventually lead either to reduced research output or to the need to substitute other “factors of production” for graduate students. Postdoctoral appointees seem the most likely candidates, but other substitutions may also be possible, such as undergraduates, more technicians, perhaps some automation, or possibly different kinds of research problems will be studied. The patterns that emerge will bear close watching.

8. FURTHER ANALYSES

In order to investigate both recent and longer-term trends in the relative attractiveness of graduate studies in S/E compared to professional schools and other careers more fully than has been done in the past, we have secured access to a series of surveys of the senior (undergraduate) classes of colleges and universities who are members of the elite Consortium on Financing of Higher Education (COFHE). (See Appendix B for a listing of the COHFE institutions.) The surveys with adequate and comparable response rates at the institution level for all years cover the senior classes from 1982, 1984, 1989, 1994, and 1998 from six or seven of the COFHE schools (approximately 4,000 students per year).31 The surveys ask students from these elite schools about their immediate and ultimate graduate school plans. COFHE has done some follow-up studies of the 1982 and 1984 cohorts and has found that the initial responses about graduate school plans were reasonably valid for these cohorts at least.32

Our analyses will focus on shifts in the patterns of graduate school plans of S/E majors and high ability students (measured in this data set by self-reported undergraduate grades) compared to all of these elite school graduates. These analyses should at least broaden the picture as to how the graduate school destinations of top students have changed over the years; in particular those who would be prime candidates for graduate S/E studies. To date, the conventional wisdom about the decline in attractiveness of S/E relative to the professions seems to be based largely on single institution studies comparing two cohorts separated somewhat in time (see earlier discussion).

9. POLICY ALTERNATIVES

Clearly, the discussion that follows is preliminary for there is much further research and analysis to be done on this issue. But for purposes of policy discussion here, we will assume that the indications of disproportionate declines in the attractiveness of S/E graduate studies to top students are borne out in further studies. Given that assumption, we will now briefly explore the contours of alternative policy responses.

9.1 Let Market Forces Work

Given the predilections of the U.S. policy system, particularly in regard to labor market policies relating to the highly decentralized and prickly world of higher education, public policymakers might well elect to simply let the market work here. If more of the very best U.S. students elect medicine, business or other professions—or no graduate work at all--so be it, at least until conditions in these alternative fields become so congested that S/E research careers begin to look relatively more attractive. Labor markets are inherently cyclical and many possible policy interventions may do more harm than good (Freeman, 1971).

The problem with leaving the current system alone is that it may be seriously out of balance. The recent report on Trends in the Early Careers of Life Scientists by a committee of the National Research Council’s Commission on Life Sciences (1998) and scholars such as Massy and Goldman (1995) have noted that the U.S. academic research and graduate training enterprise employs graduate students in response more to current research and undergraduate teaching needs than to the ultimate demand for the services of finished PhDs in the labor market. Thus, if demand for PhDs in relatively permanent career positions grows only slowly, as it has in recent years, while the training system grows independently in response to the continued growth in academic R&D funds that appears likely to be available, numbers of high quality applicants to S/E graduate school may continue to fall in response to the long-term prospects for PhDs while new graduates may continue to back up in substantial numbers in the temporary postdoctoral pool.33 These unattractive outcomes would tend to further depress interest in S/E studies among the best students whose career options in business, professions, or information technology, should be the strongest. Such a result, especially if the best U.S. students could not be so readily replaced from abroad as in the past, must eventually affect the quantity and quality of the U.S. research effort, economic competitiveness, and perhaps ultimately the nation’s capacity to understand as a society, economy, and polity the essentials of the scientific age.

In such a scenario, the usual suggestion that prospective graduate students be provided more and better information about their career prospects in various fields, while desirable on principle and now more technically feasible than ever, is unlikely to help attract the best students to science and engineering if the tale told by the market data is unattractive. It seems likely that the best students will also have the best alternative prospects in other fields and, in the past, “hot” fields in terms of general student interest have usually been most attractive to the top cadre of students as well (Adelman, 1985; Bok, 1993).

In response to decreasing interest from top students, universities might well respond to such clear signs of weakness in the graduate student pool by reallocating discretionary resources to enhance graduate assistantships so as to be more attractive to these top students. This is a logical response for them in the current decentralized decision making regime, but to the extent universities are able to do this, does nothing to help on the demand side in terms of PhD career opportunities in S/E. It might simply increase the “logjam” problem at the output end of the graduate education pipeline as universities were tempted to increase total enrollments as they received more good applicants in response to improved student support opportunities. Depending upon the extent of market information provided or available to students, such an approach by itself could even be regarded as misleading to them. More or richer graduate fellowships or assistantships also necessarily mean additional resource costs.

9.2 More Government Support for Graduate Fellowships and Assistantships

This step is of course broadly similar to allowing universities to take the initiative but permits federal agencies more control over how (e.g., more competitive fellowships or higher pay for existing research assistants), in which fields, and to what extent enhancements are made. Thus, the system-wide impact might be greater. To limit public costs and ensure some input from academe, some form of matching incentive grant mechanism might be used. The disadvantages are similar to those that apply to the lassez-faire approach – more or better graduate student support may well add to supply-demand imbalances at the PhD end of the system as more students (not just top students) are attracted – but the costs in this case are borne more directly by the taxpayer.

9.3 Adjust Immigration Policies to Import More Non-Citizen Graduate Students

Federal immigration policies might be employed to encourage more recruiting of competitive S/E graduate students from other countries. As has sometimes been done in the past (although not with much precision), immigration policies might conceivably be more effectively manipulated to respond to shortfalls of highly qualified potential scientists and engineers.34

Such an approach raises the question of whether the U.S. should, as a matter of national policy, raid the young brainpower of the rest of the world, particularly the developing world, primarily to staff its own academic R&D enterprise at low cost. Moreover, signs are emerging that the pool of available non-citizen students may not be as bottomless (or at least as inexpensive) as it once appeared. The direction of scientific capabilities and career prospects in many of the countries that have contributed many S/E graduate students in the past seems to be on the way up, potentially encouraging more students to stay home (Desrusseaux, 1998; Lloyd, 1998). The recent declines in non-citizen graduate students and PhD recipients in most of the S/E disciplines in U.S. universities already noted may be a warning sign of this. In any case, it seems likely to become a problem in the long run.

Moreover, other developed countries are competing more aggressively for S/E graduate students (e.g., Maslen, 1998), which raises the stakes. There are also some potential domestic political problems given perceptions that there are too many foreign graduate students in U.S. universities, especially state universities, being subsidized by public funds (Wilson, 1999). While the subsidy claim is dubious, given what these students contribute to the research enterprise and the fact that they have to compete for admission on the basis of quality, the perceptions are real.

9.4 Demand Side Approaches

If market signals are now being correctly read by top students but there is a market failure (positive spillover) in regard to the long-term social benefit of having at least a stable flow of top U.S.-origin talent into advanced S/E training, then it follows that policy should take stimulative steps on the demand side. The federal government could in principle act to permanently increase demand for PhD scientists and engineers, but this is practically difficult for it requires committing additional resources into the future. Yet, no other actor in the system possesses sufficient leverage to materially affect demand for research scientists and engineers.

One hoary suggestion along this line is that the federal government commit to supporting the research program of virtually every PhD scientist trained in a reputable institution and not otherwise employed in R&D. Moving substantially in this direction would no doubt make career prospects for S/E PhDs much more attractive than they are now and should have a corresponding impact on the supply of U.S. students interested in graduate studies, although not necessarily only on the top talent. While some other countries have arrangements more or less like this via government-sponsored research institutes and the like, they can typically bound their commitments by controlling the supply side much more directly than is the case in the U.S. Such a centralized human resource planning approach seems plainly a poor fit in this country.

A more modest and viable approach that more directly targets the young talent issue would focus on the postdoctoral stage. The federal agencies, mainly NIH and NSF as the largest supporters of postdoctoral fellows and associates, could apply some leverage on universities linked to their` funding to upgrade these positions selectively. With the aid of some matching from institutions now extensively supported by federal funds,35 the idea would be to create a gradually increasing and ultimately significant number of “senior research fellow” positions open on a competitive basis to postdocs who had proven themselves beyond the PhD, over, say, 2-3 years in the standard postdoctoral track. (In the majority of cases, the standard track implies working as a research associate on a faculty member’s grant-supported project. National Research Council 1998).

Candidates for the senior fellow positions would propose a project of their own and need not elect to locate at their current institution. Proposals would be competitive and peer reviewed and positions would be guaranteed for 2-3 years, perhaps renewable once upon submission and review of another proposal. Crucially, the positions would be designed to be competitive with junior faculty positions in terms of salary, benefits and research support. This should be reasonably attractive since a young scientist seeking to build a research record in this type of position would be unencumbered by most of the teaching and other obligations of a faculty member.36 Some individuals might perceive better long-run prospects for a tenure-track or other “career” research position arising from such a post than from beginning a career as an assistant professor (particularly if 5-6 years of support were a realistic possibility). More broadly, the idea would be to increase significantly and predictably the number of attractive relatively junior research positions available so as to expand demand for young scientists enough to affect the decisions of very able students considering S/E careers.

Some of the federal funds for such an initiative could come from reallocation from existing research programs in the same fields – which already support many postdoctoral positions – but substantial new resources would surely be needed to have the desired impact on the market. To try to maximize the effect on decisions of U.S. students, policymakers should consider giving preference in the awards to U.S. citizens although this contradicts pure meritocratic values.

Besides new costs, this proposal has other problems of course. For example: How could laboratory space, now at a premium in many S/E fields, be made available to a significant number of additional, relatively autonomous junior scholars and, at the same time, appropriate collegial arrangements facilitated? Would universities offer overwhelming resistance or could enough incentives and facilitating devices be designed in to make them at least willing to participate? Might not such a program have to be extensively adjusted for the peculiarities of research arrangements in different disciplines? Would this make it too complex to administer? Most importantly, could a program of this kind significantly affect the decisions of the best and brightest young people at a feasible net incremental cost? Although it would expand research positions for young scientists and engineers, such a program could not assure that universities, industry and government created any additional more senior positions. So, ceteris paribus, the career pyramid in S/E research could become even more peaked than it would be without the program, but hopefully with a larger cadre of highly able candidates.

9.5 Addressing the Decline in Interest in S/E by Whites and Males

Turning back to the supply side of the equation for a moment, the apparent fairly sharp decline in attraction of top-scoring whites and males to S/E, while female and minority numbers of top scorers have generally increased, may need to be addressed. This finding suggests that programs designed to interest young people in S/E may need to be broadened in appeal.37 Yet, supply-side approaches alone seem destined to fail unless the poor post-PhD prospects facing even the best young research scientists and engineers are improved.

10. CONCLUSION

The data do not yet permit a firm conclusion that the flow of the very best U.S.-origin students into graduate programs in the natural sciences and engineering is declining, but the evidence is accumulating that this group is turning away from S/E careers, at least in physical sciences and related fields. This poses potential problems for the national interest both in the short term and, at least in physical sciences and related fields, the longer run. It is not clear that the nation can (or should) easily replace these top students with their equivalents from other countries. Yet, if they are not replaced, the academic science enterprise will soon face pressures to reduce graduate admission standards, reduce research output, or substantially restructure itself.

The past pattern of explicit human resource policies for PhD-level scientists, to the extent these have existed at all, has been to work mostly on the supply side of the problem by varying graduate fellowship support and at times encouraging the provision of better market information to prospective students. Indeed, when R & D support or forecasted demand for faculty increases, students do appear to respond readily. But when demand falls off (or does not materialize as predicted), the long production cycle means that many young scientists are “stranded” for extended periods in postdoctoral research/training posts designed to be temporary, with little pay, security or autonomy. This pattern, with the postdoctoral pool building up over a number of years and now extending the total time in PhD plus postdoctoral training for many to 10-12 years, may well be what is discouraging the most able students from S/E studies now.

Following from this analysis, two types of policy suggestions were made. First, efforts to interest more high potential U.S. students in S/E should be broadly focused, not too strictly limited to women and students of color, for the decline in interest by top scoring white males seems to be too steep for a limited focus alone to do enough. Second, the main proposal made here is for the creation of a new matching grant program, strongly encouraged by the federal government’s leverage over major research institutions by virtue of its R & D grants to them, designed to fund the substantial improvement of the status of a significant subset of postdoctoral appointments. The new awards would be competitive, probably renewable once, and targeted at the best of more senior postdoctoral appointees through a rigorous peer-reviewed proposal process. These “senior research fellow” positions would be designed to be competitive to the extent possible with junior faculty positions so as to increase the number of attractive, autonomous early career S/E research positions significantly. While this should add to the aggregate academic research product as a side effect, the main purpose would be to impact the demand side of the market enough to affect the decision making of the most able prospective graduate students of U.S. origin.

To make a lasting difference, these new positions should ideally not be subject to the inevitable up and down cycles of aggregate R & D programs or academic posts. Fiscal partnerships involving universities, industry, and foundations as well as the federal government could help with this but can never make any program invulnerable to budgetary pressures. Nor can a program of this type, even with provision for multiyear support, address the problem of the ultimate narrowing of the career pyramid at the point where the fellowship period ends.38 Still, it should have some value in encouraging those able students with high confidence in their abilities and a serious interest in pursuing S/E research at least through young adulthood who might otherwise never even enter the competition (i.e., enter graduate school), or have their opportunity to contribute to science.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge support for the research underlying this chapter from the Alfred P. Sloan Foundation. The Foundation, however, bears no responsibility for any errors, omissions or assertions made herein. These are the sole responsibility of the authors. The chapter is a preliminary effort in an on-going research program.

References

Adelman, C. (1985). The Standardized Test Scores of College Graduates, 1964-1982. Prepared for the Study Group on Conditions of Excellence in American Higher Education. U.S. Department of Education, National Institute of Education.

Association of American Universities. Committee on Postdoctoral Education (1998, March 31). Report and Recommendations. Washington, DC: Author.

Bok, D. (1993). The Cost of Talent: How Executives and Professionals Are Paid and How It Affects America. New York: Free Press.

Bowen, W. G., and Sosa, J.A. (1989). Prospects for Faculty in the Arts and Sciences: A Study of Factors Affecting Demand and Supply. Princeton, NJ: Princeton University Press.

Breneman, D. W. (1975). Graduate School Adjustments to the ‘New Depression’ in Higher Education. National Board on Graduate Education Technical Report No. 3. Washington, DC: National Academy of Sciences.

Commission on Professionals in Science and Technology. (1997). Postdocs and Career Prospects: A Status Report. Washington, DC: Author.

Consortium on Financing Higher Education. (1988). Five Years Out: Revisiting the COFHE Class of 1982. Cambridge, MA: Author.

Desruisseaux, P. (1998, September 25). “U.S. Colleges Widen Foreign Recruiting to Offset Loss of Students from Asia.” Chronicle of Higher Education, A55-A57.

Fechter, A. E., and Gaddy, C. D. (1998). “Trends in Doctoral Education and Employment.” Higher Education: Handbook of Theory and Research, Vol. XIII. New York: Agathon Press, pp. 353-377.

Fechter, A., and Teitelbaum, M. S. (1997, Spring). “A Fresh Approach to Immigration.” Issues in Science and Technology, 28-32.

Freeman, R. B. (1971). The Market for College-Trained Manpower. Cambridge, MA: Harvard University Press.

Goheen, R. F. (1984). “Cultivating Fresh High-Level Talent in the Humanities.” Paper presented at the annual meeting of the Association of Graduate Schools. Boston, MA.

Grandy, J. (1990). Comparison of Expected with Actual Field of Graduate Study: An Analysis of GRE Survey Data. ETS Research Report 90-17. Princeton, NJ: Educational Testing Service.

Hartnett, R. T. (1987, September/October). “Has There Been a Graduate Student ‘Brain Drain’ in the Arts and Sciences?” Journal of Higher Education 58(5), 562-585.

Hartnett, R. T. (1985). Trends in Student Quality in Doctoral and Professional Education. Talent Trends Project Report, Rutgers University.

Kuhn, T. S. (1962). The Structure of Scientific Revolutions. Chicago: University of Chicago Press.

Lloyd, M. (1998, July 17). “India’s Brightest Technology Graduates Begin to Stanch the Brain Drain.” Chronicle of Higher Education, A57-A58.

Maslen, G. (1998, May 22). “Australia Plans Campaign to Promote Its Colleges to Students from Other Nations.” Chronicle of Higher Education, A52.

Massy, W. F., and Goldman, C. A. (1995). The Production and Utilization of Science and Engineering Doctorates in the United States. Stanford, CA: Stanford Institute for Higher Education Research.

National Academy of Sciences, National Academy of Engineering, and Institute of Medicine. (1988). The Recruitment of the Most Talented Students into Graduate Study in the Sciences, Mathematics, and Engineering: Is There a Problem? Report of a Steering Committee. Washington, DC: Author.

National Research Council. Commission on Life Sciences. Committee on Dimensions, Causes, and Consequences of Recent Trends in the Careers of Life Scientists. (1998). Trends in the Early Careers of Life Scientists. Washington, DC: National Academy Press.

National Research Council. (1969). The Invisible University: Postdoctoral Education in the United States. Washington, DC: National Academy of Sciences.

National Research Council. (1981). Postdoctoral Appointments and Disappointments. Washington, DC: National Academy of Sciences.

National Research Council. (annual publication). Summary Report 19xx: Doctorate Recipients from United States Universities. Washington, DC: National Academy Press.

National Science Foundation. (1989, April 25, working draft). Future Scarcities of Scientists and Engineers: Problems and Solutions. Washington, DC: Author.

National Science Foundation. (1990). The State of Academic Science and Engineering. Arlington, VA: Author.

Regets, M. (1999). “Has the Use of Postdocs Changed?” National Science Foundation, Division of Science Resources Studies Issue Brief. NSF 99-310.

Regets, M. (1998, November 27). “What Follows the Postdoctorate Experience? Employment Patterns of 1993 Postdocs in 1995.” National Science Foundation, Division of Science Resources Studies Issue Brief. NSF 99-307.

Rosovsky, H. (1990). The University: An Owner’s Manual. New York: W. W. Norton.

Sanderson, A. R., and Dugoni, B. (1999). Summary Report 1997: Doctorate Recipients from United States Universities. Chicago: National Opinion Research Center, University of Chicago.

Shapiro, D. (in press). Modeling Supply and Demand for Arts and Sciences Faculty: What Can We Say About the Projections of Bowen and Sosa 10 Years Late? Journal of Higher Education.

Smith, B. L. R. (1990). American Science Policy Since World War II. Washington, DC: The Brookings Institution.

Stricker, L. J., and Rock, D. A. (1993). Examinee Background Characteristics and GRE General Test Performance. GRE Board Report No. 89-07R. Princeton, NJ: Educational Testing Service.

Wilson, R. (1999, May 14). “Ph.D. Programs Face A Paucity of Americans in the Sciences.” Chronicle of Higher Education, A14-AA15.

Zumeta, W. (1985). Extending the Educational Ladder: The Changing Quality and Value of Postdoctoral Study. Lexington, MA: D. C. Heath/Lexington Books.

Zumeta, W. (1998). “State Higher Education Finance and Policy Developments: 1997.” The NEA 1998 Almanac of Higher Education. Washington, DC: National Education Association, pp. 65-92.

Appendix A: AAU/AGS Graduate Institutions Analyzed, by S/E Discipline and

GRE Scale

University	Biochem	Math	Physics	Chem Engr	Mech Engr
California, Berkeley		A, Q	A, Q	A, Q		A, Q
California, Los Angeles			A, Q	A, Q
California, San Diego	A, Q	A, Q	A, Q	A, Q		A, Q
Cal Institute of Tech.	A, Q	A, Q	A, Q	A, Q		A, Q
Columbia	A, Q		A, Q
Cornell	A, Q	A, Q	A, Q	A, Q		A, Q
California, Santa Barbara		A, Q	A, Q	A, Q
Florida	A, Q	A, Q	A, Q	A, Q		A, Q
Illinois	A, Q	A, Q		A, Q		A, Q
Indiana		A, Q	A, Q
Iowa	A, Q	A, Q		A, Q		A, Q
Maryland	A, Q	A, Q	A, Q	A, Q		A, Q
Michigan	A, Q	A, Q	A, Q	A, Q		A, Q
Michigan State	A, Q	A, Q	A, Q
Missouri	A, Q	A, Q	A, Q	A, Q		A, Q
Northwestern		A, Q	A, Q	A, Q		A, Q
Oregon	A, Q	A, Q	A, Q
Pennsylvania			A, Q
Penn. State	A, Q	A, Q	A, Q	A, Q		A, Q
Purdue	A, Q	A, Q	A, Q	A, Q		A, Q
Rutgers	A, Q	A, Q	A, Q	A, Q		A, Q
Southern California	A, Q	A, Q	A, Q
Texas	A, Q	A, Q	A, Q	A, Q		A, Q
Vanderbilt	A, Q	A, Q	A, Q
Washington	A, Q	A, Q	A, Q	A, Q		A, Q
Wisconsin	A, Q	A, Q	A, Q	A, Q

Legend: A – included in analysis of GRE analytical test scores

Q – included in analysis of GRE quantitative test scores

Note: Some institutions do not have doctoral programs in the disciplines studied so this accounts for some of the blanks.

Appendix B: Participant Institutions in the Consortium on Financing Higher Education (COHFE) Senior Surveys

Amherst College

Barnard College

Bryn Mawr

Carleton College

Columbia University

Cornell University

Dartmouth College

Duke University

Georgetown University

Harvard University/Radcliffe College

Johns Hopkins University

Massachusetts Institute of Technology

Mount Holyoke College

Northwestern University

Pomona College

Princeton University

Smith College

Stanford University

Trinity College

University of Chicago

University of Pennsylvania

University of Rochester

Washington University

Wellesley College

Williams College

Yale University

---

1 For purposes of this paper, the natural sciences are defined as the biological sciences, computer sciences, mathematical sciences, and physical sciences. All the engineering disciplines are included within the broad field grouping, engineering. Science and engineering, or S/E, will often be used here to refer to these fields. Sometimes, science or scientists will be used for variety but, unless the context makes clear otherwise, this also refers to the natural sciences and engineering broadly.

2 The goal here is to focus upon PhD-level students for these are the people who will be the major creative engines in the research enterprise. In engineering and to some extent computer science especially however, doctoral students are not easily distinguished from master’s students in spite of the best efforts of the data collection agencies, so precision in analysis is hampered by this limitation.

3 The term “supply/demand imbalances” will be defined with more precision later.

4 We hasten to remind the reader here that we do not claim at this point that the case that graduate programs in S/E are losing ground in attracting top talent is fully proven. At this point, we can simply present some suggestive evidence, describe our research program on the question, and assume the conclusion for purposes of further policy discussion now.

5 See Smith (1990); Fechter and Gaddy (1998). This very brief history necessarily glosses over many fine points and field differences.

6 See National Research Council (1969). It should be noted here that postdoctoral study/research appointments have always been much more common in the biomedical disciplines, chemistry, and physics than in other S/E fields. Thus, the following discussion applies most specifically to those disciplines, although the incidence of these appointments has been growing in the other science and engineering fields too.

7 Both NRC’s (1981) and Zumeta’s (1985) analyses showed that, at least in the early years of their careers, many former postdoctoral appointees had experienced disappointing labor market rewards. Yet, postdoctoral training did show clear signs of adding substantially to appointees’ subsequent research productivity, net of other factors, so it is a plausible investment strategy from a societal point of view. (See Zumeta, 1985: chapter 6.)

8 Measures included stature of graduate institution, registered time-to-the doctorate, receipt of a competitive national fellowship in graduate school, and publication prior to the PhD.

9 Regets (1999) reports that, across all S/E fields, the median annual salary for a postdoctoral appointee in 1995 was just $28,000, half the median salary of recent PhDs in industry and about one-third less than for those in tenure-track academic positions. Thus, these positions are not attractive for a lengthy stay.

10 See Hartnett (1987) for various methodological caveats.

11 This is based on the National Science Foundation’s annual fall Survey of Graduate Students and Postdoctorates. The latest official data are from the fall 1998 survey.

12 Since 1993 this category in the NSF data includes persons with permanent residency visas as well as citizens per se.

13 Nor should the U.S. necessarily do so. See Fechter and Teitelbaum (1997) and the discussion of reliance on international students as a policy option, below.

14 The GRE General Test is the basic set of tests taken by most prospective graduate students in the arts and sciences, consisting of verbal, quantitative, and analytical scales. We are grateful to Robin Durso and Pankaja Narayanan of ETS for their cooperation and assistance with these analyses, and to Jerilee Grandy, consultant to ETS for helpful advice. The years of GRE data available for analysis were 1989, 1992, 1995, and 1998. Each year actually represents a 12-month testing period running from October to October, e.g. the 1989 test year ran from October 1988 through September 1989. For most purposes here, the changes described span either 1988-89 to 1997-98 or 1991-92 to 1997-98 depending upon which span accurately and efficiently captures the changes over time. Tables and graphs covering all of these years are available from the authors.

15 According to ETS, only one thorough follow-up study has been done of a large cohort of GRE examinees to see how many enrolled in graduate school the following year (Grandy, 1990). This study tracked a sample of more than 2,100 individuals who took the GRE General Test in the 1987 testing year through graduate school registration in the fall of 1987. Fifty-six percent of the sample enrolled in the fall after taking the test. (Others no doubt enrolled later.) Of those enrolled in fall 1987, 82% enrolled in the same field they had indicated on the registration questionnaire.

16 These criteria do not hold true of undergraduate grades, for example, which vary widely in meaning across institutions and even over time. Significantly, GRE scores correlate with other plausible indicators of student ability including undergraduate grade point average, college selectivity, and college major, net of examinee background characteristics. See Stricker and Rock (1993).

17 The proportions of all examinees scoring above 700 are shown in the bottom row of the table. This proportion is roughly one-eighth on the analytical and quantitative scales, but less than 5% in recent years on the verbal scale.

18 Note from the bottom row of Table 3 that, by 1998, only about five percent of U.S. citizen examinees scored at this level on the quantitative and analytical scales and just one percent did so on the verbal scale.

19 Note that the generalization about the decline applies overall for the S/E fields, but not to each individual field.

20 The numbers of top scorers (whichever test score cutoff is used) headed for each of the S/E field groupings other than biological sciences were also lower in 1998 than they were nine years earlier (1989).

21 For example, they may choose directly after, or even before, college to take employment in the burgeoning information technology sector.

22 These tests are the Medical College Admission Test (MCAT) for medicine, the Law School Admission Test (LSAT) for law, and the Graduate Management Admission Test (GMAT) for business studies.

23 These data were obtained from the Law School Admission Council.

24 The data reported in this paragraph are from the Association of American Medical Schools.

25 GMAT data were provided by the Educational Testing Service.

26 Graduation data were provided by the American Association of Collegiate Schools of Business.

27 These include individual GRE scores, the potential to add self-reported undergraduate grade point average, and an index of quality of undergraduate institution. At present, we have been able to analyze only the GRE score data.

28 Data collection was extended to 1997 but a much lower rate of institutional responses yielded insufficient data for analysis.

29 We focus on the GRE quantitative and analytical scores because we deem these the most relevant to assessment of S/E talent. It is our understanding that the GRE quantitative and analytical scores are the most utilized by admissions committees in graduate S/E departments.

30 Permanent residents comprise 2% to 6% of the total in each S/E discipline each year. Analyses of yearly mean GRE quantitative and analytical test scores of permanent residents compared to U.S. citizens revealed differences in means of 3 points or less. Hence our analyses are of U.S. citizens and permanent residents combined.

31 By agreement, the individual institutions are not to be identified. Whether six or seven institutions are included depends upon the institution-level response rate criterion applied for each survey year (a 40% or 50% response rate minimum for each survey).

32 For example, 90% of the 1982 respondents who said they planned to pursue an advanced degree immediately after college were so enrolled in February 1983 (COFHE, 1988; 73).

33 Regets (1998; 1999) provides recent evidence that postdoctoral appointment stays are generally getting longer and that many postdocs of recent vintage cite market-related reasons for their appointment.

34 For a thoughtful discussion of the possibilities and pitfalls here, see Fechter and Teitelbaum (1997).

35 In light of the premium on research space and equipment in some disciplines, the institutional match might conceivably be in such in-kind resources. Also, universities would be encouraged to seek private or state funds explicitly for purposes of providing their matching share.

36 Some modest teaching role should be permitted however as this might help the candidate build a

record (thus making the fellow positions more attractive) and would also make the arrangement

more appealing to universities.

37 Needless to say, this does not imply that initiatives that have proven effective in attracting more

women and minority groups to S/E should be discontinued.

38 The narrowing may of course be offset for a given cohort by a sudden jump in cyclical R & D support or industrial or academic demand.

29