National Bureau of Economic Research (NBER)

POLICY AND DATA ISSUES OF THE SCIENTIFIC WORKFORCE

NBER/ SLOAN FOUNDATION WORKSHOP

23 March 2001

Richard B. Freeman (NBER): Organizer

American Association for the Advancement of Science

1200 New York Avenue, NW

Washington, D.C.

On 23 March 2001, the National Bureau of Economic Research (NBER) in conjunction with the Sloan Foundation held a day long conference on the scientific workforce in Washington, D.C. at the American Association for the Advancement of Science. Fifty economists, statisticians, policy analysts, and program directors from the National Science Foundation and other agencies considered the current and future status of the scientific workforce and of national data needed to analyze the situation facing scientists in the job market. The speakers and participants presented a variety of research agendas and perspectives. There was general agreement that this is a crucial period for addressing the promise and shortcomings of the scientific job market in the U.S. and that it will take the best practices in the collection and analysis of data to develop good policy prescriptions.

SESSION I: INTRODUCTION - WHAT ARE THE KEY POLICY AND DATA ISSUES?

Greetings - Prof. Richard B. Freeman, Harvard University/ NBER/ LSE

Key Policy Issues - Dr. Joseph Bordogna, Deputy Director, NSF

Professor Daniel Goroff, Harvard University

After words of welcome from economist Richard Freeman of Harvard University and introductions from each of the participants, Joseph Bordogna of the National Science Foundation (NSF) addressed the long-standing concerns about the future of science and "reliable sources of knowledge" about it. He cited a cover story entitled "Hidden Costs of a Brain Drain" from the Wall Street Journal (1 March 2001) by reporter David Wessel that more foreigners than Americans are studying physics at the graduate level in the U.S. At MIT, nearly forty percent of graduate students come from abroad. More dramatic, Michigan State University has a statistics department with 153 applicants for its graduate program, seven from the United States and 123 from China. "Is it a sign of American success or a symptom of American weakness?," asks Wessel. "Let's not worry," say some; they'll simply stay here, and we'll reap the benefit. However, Bordogna referred to efforts by South Korea and other nations to repatriate all of their PhDs in the U.S., particularly those at middle levels in scientific enterprises. The Koreans were quite successful in getting many to return. He also referred to the efforts of Hong Kong's private sector to set up the Hong Kong University of Science and Technology (begun in the 1980s and opened in 1991) when it was realized that a significant proportion of students were heading to the U.S. for higher education studies.

Among the membership of the IEEE (Institute of Electrical and Electronics Engineers), he reports that of its 360,000 members worldwide, 2/3rds come from the U.S. and 1/3rd from abroad. But among the 60,000 or so student members of the IEEE, the reverse is true (2/3rds come from abroad; 1/3rd are from the U.S.) This might be an indicator of dwindling scientific interest among the young in the U.S. It is of special concern to the NSF because, even though many see this agency mainly as a vehicle to support discovery of new frontiers, its primary mission is to develop the scientific workforce.

Change is happening in many fields at a pace that policymakers could not imagine, leaving programs and research data "in the dust," he declares. Nano-technology could make the information revolution look tame by comparison. Biotechnology is also opening up new horizons. Bordogna suggests that science workers of the future will have to be agile, wholistic, and able to handle complexity. The U.S. needs continuous learning that tightens the relationship of the classroom to the workplace. People will need to know how to cross boundaries. The scientific community will have to reach out better to new pools of talent, particularly "women and minorities." Our central challenge is the workforce. Yet we cannot get where we want to be if we don't know where we are. This conference should supply a context that will help us attain our goals.

Mathematician and former White House staffer Daniel Goroff of Harvard University spoke about the use and misuse of statistics in many public policy initiatives. He provided a humorous case study of Management Focus magazine, with "an actual article" some years ago which claimed that over 90 percent of the 74 CEOs surveyed from Fortune 500 firms grew up with pets: a dog, cat or both. For some, this suggested that pets helped young people develop responsibility and communication skills, and in response some people apparently ran out to purchase pets for their children. (Someone quipped that it was more likely that pets helped CEOs "learn how to tell employees where to sit.") A policymaker might well take this limited knowledge and start the sloganeering or bumper stickers necessary for a national "pet project": "Bow Wow to Know How" or "Fido for High Dough." There are many such misuses of conditional probability in formulating public policy, according to Goroff.

He recalls a series of Washington rules (on the wall of an anonymous colleague) which suggest that statistics may be deployed in deceptive ways:

"The truth is a variable"

"You can't kill a bad idea"

"The facts, although interesting, are irrelevant"

"Chicken Little only has to be right once"

But Goroff cautioned against cynicism. In truth, he is pleased to have met and worked with so many policymakers who are thirsty for reliable numbers and who will try hard to achieve some good with decent data. He discussed how the international TIMSS (Third International Mathematics and Science Study) data on the performance of students showed poor science and math rankings for the United States, especially in the later grades. It made educational performance a more urgent matter. Attempts to shrink class size were launched after an influential experiment and meta-study in Tennessee. There was talk of hiring 100,000 new teachers, a common pattern after a crisis is identified. Goroff noted that not everything done as a result of all the attention was so positive, such as the hiring of unqualified teachers in California and the stuffing of students in trailers.

He discussed the need to police mythologies when confronting data. For instance, in the dog study cited in his opening remarks, researchers would need to go beyond the CEOs who owned dogs. They would have to do many interviews with the vast numbers of people who owned dogs and never ascended to the command posts of power. If the treatment of a problem is to be properly understood, it requires that researchers go beyond the successful cases and that they also seek out those who embark on a different path.

Goroff gave examples of studies that have been valuable in this way. He mentioned the work of Elaine Seymour and Nancy Hewitt, Talking about Leaving: Why Undergraduates Leave the Sciences (Boulder: Westview, 1997), which showed that many of those who leave the sciences are talented people who get out precisely because they are "with it." They are not losers; they are people with more than enough ability to hack it. They switch out for very good reasons.

He also discussed the claims that foreign teaching assistants were driving women out of the science profession allegedly because of cultural differences in the labs that might include retrograde gender attitudes. A recent study indicates that the presence of foreign students is not driving the female exodus from the science labs. Goroff also mentioned Sheila Tobias's work on capable people who were exposed to good science courses, but chose not to go on in that field: They're Not Dumb, They're Different: Stalking the Second Tier (Tucson: Research Corporation, 1990). In Tobias's view, the college science curriculum is designed to find "the top tier" students in science and is content to drive out those who at first glance seem to have a modest aptitude for it, "the second tier." The reigning concept is too often that "scientists are born, not made," she writes. Another study out of Berkeley (Ph.D.'s - Ten Years Later involving 6,000 doctoral recipients from six disciplines, biochemistry, computer science, electrical engineering, English, mathematics, and political science) looking at the 1 July 1982- 30 June1985 degree cohort indicates that many people who left academe still thought their pursuit of a Ph.D. was a good experience.

Neal Lane, the former science advisor to President Clinton, mentioned to Goroff what has been useful to him in meeting politicians on the Hill: data breaking down by each state the federal contribution of resources to science, math, and engineering. Yet the former science advisor wishes he could find studies of how the resources contributed to the human capital in a region; i.e., vivid illustrations and quantification of how significant spending on these fields make it more likely that someone from a specific state or region could secure a good job in, say, high-tech.

Congress through the Government Performance and Results Act now requires all the different agencies to have measurable outcomes for what they are doing. There is greater pressure to measure, and it will be important to find yardsticks that are fair and useful. There are many issues that will require proper analysis: i.e., the unionization of graduate students - are they just students or are they truly employees? A lot is riding on how that particular categorization is determined. The data collection can affect the reality. As Theodore Porter, the author of Trust in Numbers, expresses it: "Public statistics are able to describe social reality partly because they help define it.... The Latin root for validity means power." Goroff spoke of how the French Revolution brought in the mathematicians to impose the metric system, new regimes of measurement, and other standardized administrative tools.

He closed by calling for better coordination so that various research projects can become more than the sum of their parts. He discussed the President's Committee of Advisors on Science and Technology (PCAST) Panel on Educational Technology, which in 1997 addressed the prominence of R&D spending in some fields. However, it is a microscopic percentage in education: less than 1/10th of one percent of the total spending on primary and secondary education, according to the PCAST panel (chaired by David Shaw).

In addition, there is a need for better longitudinal studies. Just as the Framingham heart study has been able to deliver some unexpected findings by a longitudinal approach (i.e., the effectiveness of fish oil in preventing coronary disease), it is possible that a similar approach could improve science and education in the U.S. Many hypotheses can be tested through longitudinal approaches. Again this requires coordination. Just as many studies can be done with nurses because they understand protocols and scientific research, people in science should be particularly good at participating in certain projects.

Goroff closed with some observations of the French novelist Balzac, who has a character express the centrality of data in the modern world: "Numbers are always decisive for societies that are based on personal interest and on money.... Hence, nothing is better for convincing the educated public than a few figures. Our statesmen claim that everything is definitely resolved by figures. Let's go figure."

Thus data should continue to play a critical role in helping to resolve the debates of our time.

SESSION II : MARKET ISSUES

Chair: Richard B. Freeman

Discussants:

Immigration Issues - Professor George Borjas, Harvard Business School

Market for Biologists - Professor Scott Stern, MIT

New Entrants and PIs in Biology - Eric Weinstein, Ph.D., Harvard University

Modeling Market - Dr. Charles Goldman, RAND, California

Professor George Borjas of the Harvard Business School gave an overview of immigrant issues and the U.S. workforce. He explained that from 1996-98 there were approximately 2.4 million legal immigrants to the U.S. Of this total, 1.6 million received the family preference, as these were the immediate relatives of U.S. citizens. Refugees and asylum seekers totalled 295,400, while 285,600 were employment-based immigrants. The diversity visas, which are part of a lottery system, went to 153,100 people out of an applicant pool of 11 million. The chances of winning this lottery are thus far less likely than achieving admission to Harvard. Borjas finds that many winners of the diversity visa are for some reason well-educated and talented. He adds that the skills and wage distribution of immigrants has changed profoundly between 1960 and the cohort of 1996-98. The cohort of 1960 had a wage distribution close to that of the native population. Today the wage distribution is very different, with more of a U-Shape pattern - disproportionately many very low wage skilled immigrants and to a lesser extent disproportionately many high wage/high skill immigrants. The U-Shape pattern was most pronounced in the 1980s and is somewhat more attenuated in the 1990s.

Borjas turned to the controversial H-1B program for specialty occupations. The program is commonly regarded as providing work permits for high-tech workers, but Borjas explained that it is much more wide-ranging. Many fashion models qualify for the H-1B visa. The law requires that a recipient have a college degree, while the length of the H-1B is a maximum of six years. It is often assumed that those with H-1Bs are extraordinarily talented, with many doctorates in the ranks; but the reality is that only 8.1 percent in the October 1999 to February 2000 cohort have a doctorate and the majority only have a bachelor's degree (56.4 percent). The recent limit of the H-1B has been 115,000, though there are many more recipients because of exceptions to the cap for institutions of higher education, related or affiliated nonprofit entities, and non-profit or government research organizations. Slightly over half of the H-1B visas go to those in computer-related occupations. (Borjas also clarifies: "For FY2000, the limit was set at 115,000. The statutes increase the annual limit to 195,000 for 2001, 2002, and 2003. After that date the cap reverts back to 65,000.") India supplies the greatest number of H-1Bs, with 34,381 recipients representing 42.6 percent of the total number of petitions approved (80,786) between October 1999 to February 2000. China was a distant second with 7,987 recipients (9.9 percent of the total). Of the 34,381 Indian recipients, 19,209 came under the cap, while 15,172 Indians had non-cap status. Motorola hired the most H-1Bs, with a total of 618 during this period, followed by Oracle (455), Cisco (398), Mastech (389), Intel (367), and Microsoft (362).

In terms of the training of foreign students, Borjas notes that "in 1994, foreign graduate students earned 12 percent of all master's degrees and 27 percent of all doctor's degrees conferred at U.S. colleges and universities. In 1977, foreign students earned 6 percent of all master's and 11 percent of all doctor's degrees conferred." Among the 224,707 science and engineering faculty in the U.S., twenty percent or 45,009 are foreign-born. India leads with 6,876 or 3.1 percent of all faculty; China provides 4,830 or 2.1 percent; and the United Kingdom, 3,426 (1.5 percent).

Borjas provided a research agenda for the years ahead:

1) What is the impact of the H-1B program on the affected labor market? [Borjas finds virtually no useful studies of these questions. But when 42,563 or 53.5 percent of the approved H-1B petitions between October 1999 and February 2000 are in computer-related occupations, there could well be consequences for the labor market in this field.]

2) How would one measure if there is a "shortage" of high-tech workers?

3) What factors determine the transition to permanent residence of the H-1B's?

4) Do employers "exploit" H-1B workers?

5) Do foreign students crowd out native-born students in S&E programs?

6) What has happened to the post-doctoral labor market in those fields most penetrated by foreign students?

7) Does the structure of the post-doctoral market (e.g., long post-doctoral fellowships) discourage native-born students from entering those fields?

8) Why did this type of post-doctoral market arise?

9) How do foreign students affect the educational process?

10) Do states benefit from subsidizing the education of foreign students in their public universities?

11) Would the United States benefit from expanding high-skill immigration?

Scott Stern of MIT then took up the scientific labor market, with special focus on the discipline of biology. He addressed why science and engineering labor markets are important. In many ways, long-term growth depends on the ability to sustain innovation. In the short term, the supply curve in S&E is historically inelastic. Therefore pure (and non-permanent) R&D subsidies are likely to raise wages with (roughly) fixed effort, rather than increase the level of discovery and innovation. He observes that the desire for cash is not what drives people to enter this field. It is thus necessary to study certain profession-specific institutions. Scientists give up their findings and get the credit for being the first to discover something; then others build upon the discovery. Now it is likely true that some of the appeal of pursuing science for certain students is the circumstance that this field constitutes the cheapest of advanced degrees for an individual. (There happens to be a lot of money available to fund graduate science).

There are many features of science that makes it hard to study for outside researchers. It should be recognized, as Nathan Rosenberg observes, that most disciplines (especially in their modern form) are less than 100 years old. Computer science is especially new. Meanwhile, the distribution of quality in scientific work is variable and skewed. There are multiple market imperfections, for instance between churn and lifetime labor markets. Some firms such as Millennium Pharmaceuticals practice a churn labor market (burning and churning workers in six month cycles), while others like Pfizer, the pharmaceutical giant, seem to believe that when you join the company it is for life. There is also the tenure track versus the post-doc dynamic, dual career ladders versus virtual product development teams.

Stern explores why firms allow life science researchers to participate in academic science (and why this might be associated with high R&D productivity). Many of them are publishing their results, a puzzle because companies usually are seen as guarding information. He provides two explanations for the success of this "science-oriented research approach." He speaks of the Preference effect and the Productivity effect: the former referring to the "taste" for science that engulfs most researchers who can then more easily be recruited by companies and the latter to "R&D productivity gains arising from earlier access to discoveries by external (often university-based) scientists." His abstract adds about postdoctoral biologists: "The results suggest a strong negative relationship between wages and Science. For example, firms which allow their employees to publish extract, on average, over a 20 percent wage discount." Elsewhere in his paper, he adds: "Because letting the scientific community establish its own internal rules for prestige and recognition may reduce the cost of knowledge production, total spillovers from knowledge production into technological innovation may depend on the degree of insulation from commercial incentives." For Stern, this is "a promising area for further study."

Eric Weinstein of Harvard University followed up with a discussion of non-wage signals that pervade science. There is a lot of talk about mutualism, and he referred to a model of bees, pollen, and flowers that supposedly give benefits to everyone in the chain of production. Weinstein elaborated on the periods of growth in science and then the recent retreat into a period afflicted with steady-state performance. To avoid problems with the steady state, he offered an uncoupling proposal that would convert some temporary positions (postdocs) into permanent professional positions. Weinstein has carried out extensive interviews with 18 holders of PI (Principal Investigator) posts based at four different universities.

He then referred to Paul Romer's data on the failure of science programs to provide outcome data on their graduates. Commenting on the findings of his research assistant, Romer wrote in NBER Working Paper No. 7723, "Should the Government Subsidize Supply or Demand in the Market for Scientists and Engineers?" (June 2000):

In response to his 60 initial requests from science and engineering programs, he received not one response giving information about the distribution of salaries for graduates, either in the initial information packet or in response to a follow-up inquiry from him. In contrast, he received salary information for 7 of the 10 business schools in the application packet, and in response to his second request, he was directed to a web page with salary information by one of the three non-respondents from the first round.... Four out of the 10 law schools gave salary information in the application packet and three more of them directed him to this information in response to a second request.

For Weinstein, this is mysterious because one would think that the top school would want to obtain a recruiting advantage by better placement of its students. The key question is whether these are truly students or are they in fact a low-wage workforce. He spoke of functional optimism: professors want to believe that students have a realistic chance of a job so that they don't feel they are exploiting them. The U.S. has created what might be called the Tom Sawyer economy, which refers to the ability of Twain's hero to convince others to paint Aunt Polly's fence for free.

In terms of recruiting the graduate student body, it begins to look more like hiring, rather than admission decisions. They appear to be trying to staff the labs when one listens to their survey responses on admission decisions. Seven out of the 18 surveyed Principal Investigators talk about what social scientists would categorize as "perverse incentives." Holders of PI positions admit that some students are good enough to make a contribution to the lab, but that these young scholars allegedly lack what it takes to ascend to the PI throne themselves. (Their advisors "know" them to be "unfit," and yet it appears that they leave their underlings with false hopes of eventually attaining such success in the future.)

In exploring the economics of opportunity costs, Weinstein closed by showing how the incentive structure favors trainees from developing countries. If U.S. students had a similarly advantageous incentive structure, they might well be pursuing study in science in droves. To get at developments in university labs, Weinstein called on researchers to make use of focus-group methodologies. It will take effective questioning to pierce through some of the illusions and mythologies.

Charles Goldman of RAND presented some of his research with William F. Massy of Stanford, "The production and utilization of science and engineering doctorates in the United States" (1995), work now available in a book from Anker Publishing Company (2001) called The PhD Factory. In an economy with a science-based growth model, there are competing perceptions: the view that unemployed PhDs can be found driving taxis versus the claim that there is a shortage of highly trained scientific talent. Goldman and Massy indicated that doctoral programs responded more to the need for research and teaching assistants than a realistic assessment of the labor market. Their book suggests that universities in the years ahead will pump out "an average 20 percent to 24 percent annual excess" of science and engineering PhDs.

Goldman and Massy have produced data showing that in general "less elite" universities admit more PhD students in order to carry out increases in research demands, while "the more elite departments are admitting more PhD students than their less elite counterparts in response to increases in nonmajor teaching." Thus these scholars have called on elite departments to find "alternative ways of meeting increases in general undergraduate teaching demand," while the less elite departments should consider different approaches to addressing their research demands. Otherwise oversupply will become a permanent feature of PhD output. Their work argues that the IT revolution could create pressures for cut backs in the number of traditional academic jobs, as "universities and colleges are likely to be hiring information and media specialists to leverage the time of professors in producing courseware." Higher education budgets in many states are likely to be squeezed by the expanding expenditures on corrections as well as on health and social welfare. They note that California has constitutional provisions to guard spending on K-12, but higher education could be susceptible to chopping during fiscal crises of the state.

During his presentation, Goldman discussed the oversupply of math PhD graduates in recent times. He spoke of the need for better data collection in several areas, among them: incorporating adjustments due to wages (graduate student stipends, postdoc stipends, faculty salaries, and industry wages). The overall national data base on faculty has to be improved (see this report's conclusion for more details). There should be more work on whether the oversupply of PhDs is displacing workers with less education. He suggested flexible frameworks for policy analysis that allow users to alter the basic model in order to match alternative assumptions. Finally, it will be important to demonstrate which parameters are crucial to behavior and which ones are largely irrelevant. With a better framework in place, researchers can then address specific needs for new data collection.

SESSION III: THE MAJOR DATA SOURCES - What They Have and Plans for Improvement

Chair: Professor Paula Stephan, Georgia State University

Discussants:

SESTAT data/ DRF - Ron Fecso, NSF, Chief Statistician, Science Resource Studies

CPS and BLS Surveys - Anne Polivka, Bureau of Labor Statistics, Research Economist

NCES Surveys - Andrew G. Malizio, National Center for Education Statistics,

Program Director, PLSS

Web-based Surveys - Geoff Davis, 4Charity.com, Senior Software Engineer

GSS and Post-Censal Surveys - Walter Schaffer, NIH, Research Training Officer

Ron Fecso of NSF began with a report on SESTAT (the Scientists and Engineers Statistical Data System), surveys carried out in two-year cycles that are followed up with interviews of certain subsets. SESTAT includes the NSCG (National Survey of College Graduates, first administered in 1993), NSRCG (National Survey of Recent College Graduates, administered biennially since the early 1970s and then parts of it later incorporated into SESTAT), and the SDR (Survey of Doctorate Recipients, sponsored by the NSF and other federal agencies since the 1970s). The integrated SESTAT database represents 12,036,200 individuals (based on the 1995 surveys). "Roughly 70 percent of all S&E-educated individuals in the labor force were employed in non-S&E occupations," according to Nirmala Kannankutty and R. Keith Wilkinson in their NSF report, SESTAT: A Tool for Studying Scientists and Engineers in the United States (April 1999). Fecso explained how the SESTAT data is used for reports and briefs, as well as the ways in which external users can take advantage of both paper and web reports. (See http://sestat.nsf.gov ) There are on-line tabulations and table generators.

He stressed that there are issues of privacy. Essentially, under the terms of the Privacy Act, they have to make it impossible for a researcher to take data and link it to an individual person. He spoke of licensing arrangements with institutions and the possibilities of a researcher visiting directly to look at the data.

Fecso argued that there needs to be better modeling of the data and that some statistical collections from the previous decade's census (1990) are getting stale such as the set used by the Peterson's guide on earnings (which used 1993 data on salaries, age of respondent, and various levels of educational degrees). He indicated that this particular data set, with census information collected every ten years, may not be repeated. He hoped to hear from researchers about what sorts of data collections would be useful to their projects.

Anne Polivka of the BLS discussed three data sets: the Current Population Survey, the Occupational Employment Survey (OES), and the National Compensation Survey. The Current Population Survey is designed to provide information by states; households are in for four months and out for eight and then back in for four. They survey approximately 50,000 households. Fifty percent of the answers come from the respondents of the households, and it is nationally representative data. The CPS data can be merged with other data supplements. There are evolving supplements on computer usage, whether at work or at home. Changes in the short term can be observed. She shows how there have been recent increases in the percentage of women in engineering, while there has been a decline of their participation in computer science. One disadvantage: the sample sizes are relatively sparse in some areas. In certain selected occupations, there are few people represented; i.e., 151 statisticians in the 1999 cumulative totals. It is also hard to determine if the educational attainment of those surveyed matches their occupational field. For instance, among the doctoral degree holders who work in computer science, it is conceivable that there are some English PhDs and others from the humanities.

The OES classifies workers in 22 major occupations and 769 detailed occupations. It too uses a nationally representative sample and can be used for specific regions. However, it has no demographic data, and it excludes the self-employed. The National Compensation Survey contacts 36,000 workplaces. The OMB wanted it in part to analyze government wages in relation to workers in the rest of the economy. Individual wage data and benefits in numerous areas can be observed. Very specific occupations and skill levels can be analyzed. It is not easily available to the general public, but there are facilities for researchers to come in and look at the data.

Andrew Malizio of the National Center for Education Statistics (NCES) reported on the postsecondary studies carried out by his organization. He indicated that the transfer rate of students at college is quite high, as 29 percent move to another educational institution. For the study that ended in 1994, the NCES ran out of money to follow up and study those who were yet to earn their degree. Perhaps some went back to school later in the 1990s and got the degree, but this is speculative. He also explained that data may be skewed by incorrect responses of those surveyed. For instance, it is useful to know how many students take remedial classes at university, but those surveyed may not realize that a course with a feel-good title such as "Leadership 101" is in fact a program to address basic deficits in learning skills.

Percentage of Students reported taking any Remedial Courses (BPS, 1996)

Public 23.3

Less than 2-year 6.8

2-year 26.9

4-year non-doctorate granting 22.9

4-year doctorate granting 14.8

Private, not-for-profit 12.7

Less than 4-year 13.3

4-year non-doctorate granting 15.8

4-year doctorate granting 5.9

Private, for-profit 5.5

Malizio gave background on the Integrated Postsecondary Education Data System (IPEDS), the Postsecondary Longitudinal Studies and Sample Surveys (PLSSS), and the Postsecondary Cooperative System, Analysis, and Dissemination (PCSAD). (See http://nces.ed.gov/surveys ) Of the PLSSS, he provided the research issues pursued by the NPSAS (National Postsecondary Student Aid Study):

How do students and families pay for postsecondary education? How have costs and financial aid packages changed in over the years? Why do some students receive more financial aid than other students from apparently similar backgrounds? How much have students and families borrowed to pay for postsecondary education? How do aided and nonaided students compare on total resources available for education and other expenses? What are the family characteristics of aided and non-aided students? How do costs and financial aid influence students' choices of schools and majors? Why do some students/families not apply for financial aid? How is financial aid related to the academic performance and persistence of first-time entering students?

The last question appears to be explored as well by the BPS (the survey of Beginning Postsecondary Students). The BPS is designed to provide information on "the progress, persistence, and completion of undergraduate programs; undergraduate indebtedness; career entry for vocational programs and degree completers; employment/unemployment patterns; civic participation; and family formation." There is the B&B (Baccalaureate and Beyond) surveys, the "first time longitudinal study of only bachelor's degree graduates" that "... will continue for about 10 years after bachelor's degree completion." He also gave a brief listing of the PEDAR reports on such specific topics as "undergraduates who work while enrolled in postsecondary education; profile of older undergraduates; nontraditional undergraduates; and debt burden and early labor force experiences."

Mathematician Geoff Davis gave a presentation on the possibilities for web-based surveys in order to track the progress of PhDs. He admitted that the self-selected quality of participants could be a pitfall, but he believes that much of the information collected is of value and delivered at a much cheaper cost than other approaches. He finds that the majority of students express satisfaction with their learning experience in graduate school, though a substantial minority have dissenting views in certain areas: training and supervision of teaching assistants is often inadequate; many women and people of color find the atmosphere to be less than supportive; and career guidance and placement is frequently ineffective.

Davis cautioned that web-based surveys are easy to do, but hard to do well. A central problem is that many academic departments fail to track their students. Davis talked about "viral marketing," which refers to the building of chains and word of mouth that expands the network of those surveyed. He gives people and departments benefits from participating, the sharing of alumni mailing lists and suggests that something akin to automatic e-mail forwarding may be a service that will attract people (similar to the alumni associations of some universities). Around 2/3rds learned of the survey from three sources: 30 percent from the department, 20 percent from the institution, and 19 percent from a friend. His project can be accessed at http://survey.nagps.org Davis faced challenges from the audience who worried about privacy issues. What guarantee could Davis give that his data base would not be sold off to, say, marketing firms eager for information on the most highly educated sectors of the U.S. public? He responded that his firm has entered an agreement with a monitoring organization that has agreed to sue Davis and his colleagues should they violate the privacy terms of the survey.

Walter Schaffer, Research Training Officer for the NIH, discussed the GSS and Post-Censal Surveys. He thought that there is too little data on physicians and that some postdocs are not captured in the surveys. PhD production has been relatively constant during the past fifteen years. He expressed skepticism about dire claims that the U.S. is overproducing PhDs by two or three times. Many postdocs, as many as half in some fields, are filled by foreign students. Few science PhDs are out on the streets.

Professor Paula Stephan of Georgia State University argued that more data is needed on the collaborative nature of science. Data that focuses on individuals may miss some crucial developments in science. There has been a substantial growth in co-authorship, especially of papers with four or more authors.

SESSION IV: IMPROVING THE DATA - ANALYSIS - POLICY LINKAGE

Chair: Professor Ronald Oaxaca, University of Arizona

Discussants:

What are the key questions current data can't answer?

Lawrence Burton, NSF, Senior Analyst, SRS,

Human Resources Statistics Program

E-mail responses to the key questions - Richard B. Freeman, Harvard University

Lawrence Burton of the NSF conveyed the challenge and difficulty of obtaining most data. In many ways, he reports, immigration is the last time the government workers can get "a clean shot at the data." Meanwhile, it is way too expensive to expand the scope for some fields. However, many people want more data, and they would shudder at shrinking the studies. (Some economists dissented and thought that the largest surveys could be reduced in size, thus freeing up resources for more specific projects. Yet others at the NSF and elsewhere thought otherwise, explaining [similar to Polivka's presentation above] that some occupations would have too small a sample size to have meaningful data.)

Burton thought one of the areas that may need work concerns attitudinal issues, what motivates certain groups. He spoke of the need for interviews with individuals.

Richard Freeman discussed a survey of economists who study scientific job market issues that addressed three questions: 1) the ideal data for your research; 2) suggestions for the NSF; and 3) key analytic questions you want answered outside of your own research. Among the responses, many wanted better explorations of production of knowledge issues, in particular the boundary between academe and industry. The collaboration issue requires more information: i.e., research families, geographic collaboration, and even the outsourcing of some research abroad. In presenting findings, some desired more links of the data. While many admire SESTAT, they believe it is not easy to use.

Professor Ron Oaxaca thought that some approaches might give better understanding of minorities and women in science, math, and engineering: in particular, longitudinal studies that explored important decision points in the life cycle. It would be valuable to know the characteristics of schools and neighborhoods at points in the life cycle. If all the results appear to be driven by family background, for instance, there might be limits to what public policy can accomplish on certain challenges.

SESSION V: CAREER PATTERNS, PRODUCTIVITY, AND ATTRITION

Chair: Dr. Rolf Lehming

Discussants:

Skill Depreciation - Professor Anne Preston, Haverford College

Career Patterns - Professor Sharon Levin, University of Missouri/St. Louis

What does SESTAT data tell us about career patterns? -

Mark Regets, NSF, Senior Analyst, SRS, Integrated Studies Program

Economist Anne Preston explained the ramifications of skill depreciation on university graduates in the sciences. She measures skill depreciation by looking at the science citation index and calculating what fields have the newest citations (a rough measurement of the obsolescence of previous knowledge). Accelerating rates of skill depreciation seem to push some people out of science. She relied on two data sets: "The first is a set of work histories of 1700 men and women who graduated with degrees in science and engineering from a common public university between 1965 and 1990. The second is a set of citation half lives for journals in some fifty scientific fields annually from 1975 until 1992." She concludes: "Men and women in fields where the rate of acceleration in skill depreciation is one standard deviation above the mean are 31 percent and 46 percent more likely to engage in permanent exit, respectively, than their peers who are in fields with average levels of acceleration in skill depreciation." For the period 1975-1992, "the rate of skill depreciation has been increasing most rapidly in fields such as biology and parasitology and has been decreasing most severely in fields such as physics and astronomy. Furthermore, in 1992 the citing half life of mathematics was over two and a half times the citing half life of biochemistry." The average half-life in math citations was 9.68 years, compared to 3.64 years in biochemistry.

Economist Sharon Levin took up a range of topics that are receiving her research attention. She has been looking at the productivity of scientists as they age and concludes that they do write less in the later years of their careers. Cohort effects and timing of entry into fields can be important: the job market has been good for some cohorts, miserable for others. Women often start out at lower salaries but some benefit from greater rates of leap in wages. Perhaps it is a question of catch-up, but has affirmative action helped some of the closing gap? Levin has done key work on the contributions of foreign-born scientists to the U.S. Many questions are being asked about the foreign-born scientists: Are immigrants a source of strength and contribute disproportionately to U.S. science? Or do they crowd the workplace and discourage native talent (lower wages, increased unemployment)?

Levin finds that foreign-born scientists have made a considerable contribution to U.S. science. As evidence of their disproportionate contribution, she cites her work with Paula Stephan on the membership of the most prestigious scientific academies as of 1980: 19.2 percent of the NAE (National Academy of Engineering) was foreign born, though they were only 13.9 percent of the field. As for the NAS (National Academy of Sciences), 23.8 percent were foreign born, more than their 18.3 percent weight in sciences as a whole. (See Sharon G. Levin and Paula E. Stephan, "Are the Foreign Born a Source of Strength for U.S. Science?," Science, 20 August 1999).

Levin has not done part II of the study: the potential negatives. Do native workers lose? Are many countries suffering from the brain drain? The Economist refers to the talent-sucking vampire, a syndrome that has driven the German Ministry of Education to push for the return of some of the nation's most talented people. Levin observes that U.S. citizens are less likely to hold jobs in the science and engineering fields, but it might be worth exploring if some are today in better careers and more lucrative jobs. Have they been displaced, or is the decision a voluntary one, based on successful search for greener pastures?

Mark Regets of the NSF ran through many data sets. He cautioned that in some cases there is such a focus on the plight of those with doctorates or in doctoral programs that it becomes easy to forget that the PhD cohort is perhaps five percent of the overall scientific labor force. In R&D, they have a special role, but it is still around ten percent.

He spoke of high response rates on the '93, '95, and '97 SESTAT surveys, a success due to relatively high expenditure on carrying them out. He has just begun to take a look at the '99 survey, data that may still be in a raw form.

There are some retrospective questions that help figure out what has happened to a person over time such as why she or he left the previous job. Another fruitful question, according to Regets: Is the person working in a field close to the educational training? There are job satisfaction questions, as well as would you choose the same academic major if you could do it all over again? Job dissatisfaction in many fields seems to peak in one's mid-40s. Are most PhDs in engineering working in their field? Only about half, he reports. PhDs move into management positions at a greater rate than other degree holders. Other data may stand out: the lay-off rate is high for older workers in computer jobs. The growth in postdocs is another prominent development: more pronounced in biological sciences and physics than in psychology. His grand conclusion: SESTAT has a lot of versatility, and nimble researchers should take advantage of this.

SESSION VI: SCIENCE WORKERS IN INDUSTRY - PRODUCTIVITY AND

FIRM ISSUES

Chair: Professor Adam Jaffe

Discussants:

Geographic Distribution Related to University Location - Professor John Bound,

University of Michigan

University-Industry Partnerships - Professor James D. Adams, University of Florida

The Link from University Research to Output - Professor Lee Branstetter,

University of California/Davis

Economist John Bound gave the preliminary results of research that he has carried out with Gabor Kezdi of the University of Michigan and Sarah Turner of the University of Virginia. Entitled "Trade in University Training: Cross-State Variation in the Production and Use of College Educated Labor," their paper tries to elucidate the stock and flow of degree recipients in the United States. While Bound and his colleagues have modestly admitted that their project has had limited success, they have been able to show different rates of dispersion: "MDs and PhDs present the two ends of the spectrum with PhDs 2.5 times more concentrated than MD degree holders." While PhDs show "appreciable clustering," the MD degree holders are spread out because "most physicians are employed in direct patient care rather than in the research sector," thus "their services are not traded across geographic areas and health care services are demanded - at least at a modest level - in all geographic areas." Using the decennial Census surveys and Department of Education data, they have found "a modest association between flow and stock" at the BA level (in contrast to the MD level). They have noticed that certain states such as California and Connecticut are BA importers, while others such as Utah and Vermont regularly export their BA-trained people. The state of Washington shifted to importing BA personnel in the 1980s, while Arizona became an exporter. With further work, they hope to contribute to "the policy discussion about the social return to public investments in higher education." Their tentative conclusion: some benefits can be secured by states with resources directed at the BA level, but individual states are less likely to find payoffs from expanding MD enrollments. Demand for MDs in other states makes it likely that the talent will be "exported."

Professor James D. Adams gave a presentation on "industrial R&D laboratories, with special reference to university partnerships" and the "influence of "industry-university centers." He explained that "per dollar, academic R&D spillovers have a larger effect on laboratory patents than industrial R&D spillovers." But he clarified, the industrial sector is 10-12 times larger in scale than the academic enterprise; so even though the industrial labs have less bang for the buck, they still dominate. He developed understanding of the policy initiatives surrounding "Industry-University Cooperative Research Centers (IUCRCs)," which "have proliferated in recent years." He elaborates, "Their purpose is to encourage universities and especially engineering schools to work more closely with firms." Adams declares that the consulting activities of scientific scholars can be intellectually "enriching," but he also admits that some faculty can be lost to the university while on the job. His findings indicate that "open science" is still practiced at universities. In other words, "firms can approach local universities and still gain access to science and engineering that are close to the cutting edge. This contrasts with the research of other firms, where contractual arrangements that ensure sharing are needed to secure access to findings of business partners, who are often at a considerable distance from the laboratory."

Professor Lee Branstetter discussed his recent work on using patent citations to academic papers as a means of examining the impact of academic science on industrial R&D. In November 2000, Branstetter released an NBER Working Paper on knowledge spillovers and Japanese firms. After many interviews, he concluded that:

Japanese technology "leaks out" through their U.S. subsidiaries. In fact, this "leakage" is sometimes deliberately fostered by the Japanese firms. One Japanese R&D manager based in Silicon Valley described a symposium his company had recently sponsored to publicize some of the firm's more basic R&D. This was done in order to "engage" the local research community, enhance his firm's reputation among local engineers, and assist the firm in forming research partnerships with local academic experts.

This finding may come as a surprise to those committed to a doctrine called "the disincentive effect of spillovers," which argues that private firms cut back R&D when new knowledge flows to rivals.

In his latest work, Branstetter is exploring whether firms benefit from university research: does it enable better inventions and contribute to more productivity? He finds a striking increase in the number of patents in certain fields: drugs, chemicals, and computers. He has created a map of the geographic distribution of patents, and they are clustered heavily in bi-coastal centers, with the exception of some from Texas and the Chicago area. Branstetter is also hoping to provide better understanding of the relationship between increases in research budgets and leaps in scientific output.

RESEARCHERS' DISCUSSION: IDEAS FOR FOLLOW-UP ON THIS WORKSHOP

Richard Freeman convened the conference participants for a final discussion of how to keep the momentum moving forward on this project. He spoke of the possibilities of carrying out some larger scale longitudinal study that had been requested by several speakers. He thought that funding might be sought for graduate students who had interesting projects in this field. Freeman thought that for this project to succeed it must cast a wider net than academic economists by bringing in industry, government, as well as the academic specialists. He thought that there are many interesting questions that could be addressed by getting groups together from disparate backgrounds.

For instance, it might well be valuable to talk to the extraordinary scientific talent at a firm such as Xerox, a company that has come up with so many cutting edge inventions and yet languishes in serious financial crisis today. Recently Business Week (26 February 2001) ran an article by John Carey and Ellen Licking on the biotech company Celera, which helped lay out "the analysis of the entire sequence of the genome - 3 billion bits of DNA." The headline: "Nice Job on the Genome. Now, Let's See Some Profits." The inability of managerial leadership to coordinate scientific accomplishment with commercial success continues to puzzle outside observers, especially for a company such as Xerox that was by many measures the international pacesetter in computer technology. There are many projects of this sort that might give deeper insight into the scientific workforce and modern industry.

Scott Stern, James Adams, and Paula Stephan spoke about creating coherent data bases and improving the data infrastructure. Stephan hoped that inquiry into these topics could become more of a field, which would help prevent things from falling through the cracks, as is sometimes the case today. Elsewhere Charles Goldman and William Massy have pointed out that "basic data on the number of faculty in each discipline at each institution are lacking at the national level. There is no frequent survey of faculty members by field and institution." The National Research Council assembles data once a decade, but it is too infrequent. "Until such a data source is created and maintained," they write in the conclusion of The PhD Factory, "we will not be able to make national estimates of the rates of hiring, promotion, and departure... Major advances in our understanding of faculty careers and labor markets could come with periodic counts of faculty by department." They have also added: "the big need, of course, is for qualitative research into the displacement of lesser degree holders by PhDs when there are positive base employment gaps." By helping to judge whether the PhD adds value to these different kinds of non-academic jobs, the research "if successful.... would inform future quantitative research on the supply and demand for PhDs."

Though there is urgent need to improve the database on academic scientists and PhDs, there was one caveat. In constructing future research projects, scholars should bear in mind, as Mark Regets of the NSF earlier pointed out, that lesser degree holders dominate the scientific workforce and PhDs hold perhaps a five percent share.

Michael Teitelbaum of the Sloan Foundation expressed how many public policy debates are being conducted in the absence of serious research. Too many experts are simply hired guns for interest groups and incapable of giving balanced information. Perhaps even a center could be created that would disseminate effective research on these questions. Earlier George Borjas had pointed out how truly little is known about those who have come in on H-1B visas. A range of inquiries on this subject would have improved the quality of this particular debate.

The participants closed by thanking Jennifer Amadeo-Holl of the NBER (Cambridge, MA) for her considerable contribution to organizing this conference and her assembling a massive briefing book with the key articles and working papers on the scientific workforce. This workshop allowed for a helpful exchange of views between university scholars and key personnel in government agencies. The shape of future projects has yet to be established, but important strides have been taken in uncovering new research agendas for the early twenty-first century.