Point of View: A fair deal for PhD students and postdocs

In the past three decades (1979–2009), the number of biomedical graduate students in the US doubled, with most of the increase funded by NIH research grants awarded to principal investigators (see Table 1). However, as many as 44% of these students fail to complete their training, and about one in three of those who do obtain PhDs leave research completely (Table 2). This means that only 37%, or slightly more than a third, of the students who start PhDs eventually become researchers, even though the main purpose of a PhD programme is to teach students how to do research.

An efficient system, in my opinion, would produce enough high-quality PhD researchers to fulfil the nation’s research needs, plus a few more. Thus, 10 years after receiving their PhD, about 85% of graduates would directly engage in academic or industrial research, usually after a period as a postdoc; 10% would work in non-research activities related to science; and 5% would opt for careers unrelated to science. Some first-rate PhD programmes come close to the 85% target, but the average PhD programme produces new PhDs with scandalously low efficiency (see Box 1 and Table 3).

Despite this inefficiency, the number of new PhDs still seems to exceed the need for researchers. The reasons, almost certainly, are that universities and principal investigators (PIs) recruit PhD students primarily as cheap labour, ignoring the question of how many PhDs the US needs. Many of my academic colleagues vigorously reject this inference, averring their deep commitment to training and promoting the careers of their PhD students. And despite the evidence, the Workforce working group set up by the NIH waffled on the question of whether the number of PhDs exceeds demand, saying that inadequate tracking prevents the NIH from knowing how many PhD students are supported by research grants and what these students do after they obtain their PhD (NIH, 2012). Denial and pleas of ignorance are delaying tactics, not arguments, but those tactics have stymied attempts to change PhD education.

Universities and PIs recruit PhD students primarily as cheap labour, ignoring the question of how many PhDs the US needs.

The Workforce working group and some academics express concern that one third of biomedical PhDs are employed in non-research positions. However, rather than reduce the number of PhD students, they suggest that PhD training programmes should offer students opportunities to learn the rudiments of other science-related careers (such as biotech, scientific publishing or science policy). I disagree. Why must students be trained for other careers while they struggle to learn the skills that are essential for research? This half-baked notion tries to mask its obvious but unstated goal, which is to justify recruiting more PhD students to work in labs. The same notion, unfortunately, prevents construction of a sustainable biomedical research workforce.

Improving the overall quality of graduate education and selecting better students can remedy the shameful inefficiency of research training, increasing the proportion of PhD students who become scientific investigators. The protracted duration of graduate and postdoctoral training contributes to an equally shameful fact: academic researchers in 2010 received their first NIH grants at an average age of 42, four years older than in 1980 (Workforce report, p29): this means that young scientists are devoting their most creative years to questions posed by older scientists. On average it takes 6.5 years to complete a PhD in the US: however, students can obtain a PhD in just 4.6 years at the Cold Spring Harbor Laboratory (see Box 1). It should be possible for other institutions to match this and, at the same time, accomplish the even more important task of enhancing the quality of PhD training.

The Workforce working group called for modest increases in training grants and for graduate students to be better informed about alternative career options early in their training. Much stronger actions are needed, however, on three fronts.

First, federal training grants should be awarded preferentially to institutions that reform their graduate training programmes and shorten the period required to earn a PhD. Such reforms should make supervision of graduate students a more communal responsibility, less dependent on the judgment of a single faculty member. Emphasizing quality of training and the need to complete it in less than five years, on average, faculty committees should carefully monitor student progress. With few exceptions, NIH funds should not support a PhD student after five years of training.

Second, the NIH should strongly encourage every graduate training programme it funds to institute Master of Science (MS) degrees for all students who satisfactorily complete two years of training, including at least one year of supervised research. Before entering PhD programs, all applicants should know that after the MS degree some students will go on to earn research-based PhDs, while others will choose (or be asked) to pursue a different course. The MS branch-point will allow each student to determine whether research is a desirable and realistic option for them; it will also help faculty to identify those students who are likely to learn enough in the next two years to pursue research careers.

If either the student or the faculty have doubts about the student’s suitability for a career in research, the MS branch-point furnishes a timely escape route. In such cases the university should do its best to help students find further training appropriate for a different career, and some may choose to allow students to switch into selected (non-biomedical) graduate programmes (such as journalism or business administration). Both the PhD programme and the NIH should recognize that re-directing students away from laboratory research can be advantageous for some students, is necessary for effective graduate training of good scientists, and is not a mark of a student’s ineptitude or a university’s callous disregard for students. Compared to a long Darwinian struggle plus a fruitless quest for a good job, taking an MS degree and shifting to another course of study can open avenues to a more satisfying outcome.

Third, the practice of producing PhDs in direct proportion to research grant funding will always be unsustainable, so we must tackle it head-on by completely separating NIH funding for PhD training from the funding for research grants. Here’s how to make a gradual transition: set a starting date, after which each new PhD student funded by a federal training grant will receive that support for a maximum of five years; every PhD student supported by a research grant who graduates or leaves graduate school frees up a ‘slot’, and funds for this slot are transferred from the PI’s research grant into an institutional PhD programme to provide five years of support for a new PhD student. Such a transition could be completed in less than eight years.

A number of problems will need to be overcome if these changes are made. First, it will be necessary to prevent investigators and institutions gaming the system. Second, some schools and departments don’t have existing training grants: as these schools shift funds from research grants to new training programmes, the latter will need to be subject to the same level of review as existing training programmes. Third, it will be necessary to persuade institutions, investigators, the NIH and Congress to transfer a substantial amount of money within the NIH from the budget for research to the budget for training. Although this will not change the overall NIH budget, it will make real costs of PhD training more obvious and reduce the average size of much-loved R01 grants to individual principal investigators. Also, students will have greater autonomy in choosing labs, and investigators less autonomy in hiring workers, so some senior investigators will surely complain (Bourne, 2012; Price, 2013).

Nonetheless, shifting money from research grants to training grants will make the biomedical research enterprise more sustainable in several ways: (i) it will improve the quality of training by making all NIH-funded PhD training subject to rigorous peer review, and it will signal the crucial importance of excellent training to both faculty and students; (ii) it will promote student autonomy and responsibility by freeing students from direct financial support by research supervisors; (iii) it will insulate investigators and institutions from conflicts of interest that tempt them to bend training policy and standards in order to retain cheap workers in their labs: if they do not pay student stipends from their research grants, PIs will not be so motivated to keep students in the lab after they learn how to do research; (iv) it will help to insulate graduate training and new PhDs from future versions of the boom-and-bust cycle of the overall research budget; (v) it will provide better information about the quality and number of PhD students being trained, and allow the NIH to increase or decrease that number in accord with national needs. This last advantage is key. Indeed, a year after the Workforce working group called attention to the number mystery, the NIH is still scrambling to get a better handle on it.

Postdocs bear the heaviest burden of the unsustainable biomedical research enterprise. Over the past three decades, the number of postdocs increased about threefold (see Table 4), but jobs in industry and academic research did not keep pace with this increase, so senior postdocs have collected in an ever-deeper ‘holding tank’. Much of the increase in postdoc numbers was driven by researchers from outside the US. The skills and energy of these non-US researchers have been welcomed across the US, but their presence also helps to keep postdoc salaries at relatively low levels, for both US and non-US researchers. The bottleneck between the holding tank and the small number of permanent research positions also shifted the age profile of NIH-funded investigators: in 1980 18% of NIH-funded investigators were under 36 years old, and only 1% were over 65; by 2009 just 3% were under 36 and 7% were over 65 (NIH, 2012).

The postdoc holding tank parallels a broader problem—the fact that the US produces twice as many STEM (science, technology, engineering, and mathematics) graduates as are needed for STEM-based positions in industry. In other words, the claim that there is a shortage of graduates in these areas in the US is a myth, perpetuated in part by employers who can profit by keeping the salaries of their STEM employees low and by persuading Congress to provide more visas for STEM graduates from other countries (Salzman, 2013).

The related problems of too many postdoctoral researchers and the shifting age profiles of individuals who eventually find permanent positions require decisive action on four fronts.

First, the roles and pay of postdocs need to be changed. Postdocs in institutions that receive research or training grants from the NIH should be called ‘postdoctoral researchers’, not ‘trainees’, and institutions should be obliged to treat them as fully-fledged employees. To signal the demise of the postdoc holding tank, with a few carefully defined exceptions (for instance, career breaks to raise young children), only postdocs who received their PhD (or MD) in the previous five years should be eligible for support on NIH research grants. Staff scientists and faculty researchers would remain eligible for salary support on NIH research grants, but ‘visiting scientists’ and long-term postdocs with other ambiguous job titles would not be eligible. The Workforce working group also suggested increasing pay levels for postdocs supported by NIH research grants, especially in their later years of service. For this excellent recommendation to make a real impact on the size of the holding tank, actual salary increases need to be substantially larger than those the working group proposed.

Second, to plan for its future, the biomedical research enterprise must know how many postdocs it employs and the course of their later careers. (Estimates of the number of postdocs in the US range from 37,000 to 68,000, and the real number may be higher; Workforce report, p 32.) So, the NIH should award grants to help pay administrative costs for monitoring progress and career destinations of postdocs (Rockey, 2012). These grants could also be used to teach skills essential for a career in research, such as scientific writing and communication.

Third, the number of ‘staff scientists’ supported by the NIH should increase. The definition of a staff scientist could be as follows: she/he must have an MS or PhD degree, be able to perform and analyse experimental results with unusual skill in at least one area of special interest to the lab, and be able to teach and help supervise postdocs and PhD students. Universities should create a special staff scientist classification (e.g., salaries higher than postdocs, lower than faculty; benefits like those of other employees; able to apply for grants, but only to support their own salary). Institutions and the NIH should create incentives for bright PhDs to become staff scientists, and for faculty to hire them. Even a modest increase in the number of staff scientists will enhance continuity and the level of research skills in the laboratory workforce. It would also provide academic jobs for young scientists who choose not to compete for research grants, and stabilize research efforts if Congress or NIH decides to decrease the number of PhD students.

Fourth, the US must deal with the growing number of non-US citizens who enter the postdoc population with PhDs earned in the US or elsewhere (Table 4). At present these researchers can be funded by their home country or by an NIH research grant: PhD students from outside the US can also be funded by NIH research grants but not by NIH training grants. Scientists from outside the US bring enormous benefits to the US, but they also swell the postdoc holding tank and depress the market for US citizen scientists because they are often more willing to risk the low pay, long training and fierce competition that deter US citizens from careers in biomedical research (see appendix D of the Workforce report for further details). Moreover, Congress may soon make it easier for non-US-citizen postdocs to obtain visas or citizenship, which will make it even more difficult to achieve a sustainable research enterprise. At the same time, there is evidence that most researchers who enter the US on visas are never sponsored by their employers for citizenship (Salzman, 2013).

The solution, I think, is to use economic incentives to make sure that only the very best non-US citizens are hired to work in research labs, and to increase the likelihood that these researchers will eventually receive citizenship. First, universities should persuade Congress to allow non-US citizen PhD students to be supported by NIH training grants, providing they agree to undertake a subsequent ‘payback’ period of working as a scientist in the US. This would promote more rigorous screening of non-US citizen students entering PhD programmes, and would also enhance the quality of PhD training. For prospective postdocs, it would be useful to require academic institutions (and companies) to pay a modest ‘tax’ (e.g., $7,500) for every non-US postdoc who enters their labs. (Increased postdoc salaries would have a similar effect, but this ‘tax’ would be more effective.) The money raised this way could be used to train PhD students and to keep track of the numbers and career destinations of postdocs.

The changes proposed in this essay will, I believe, improve the quality of PhD training, drain the postdoc holding tank and reduce the age at which researchers get permanent positions and start independent research programmes. Moreover, by breaking the damaging feedback loop that promotes the enlistment of PhD students and postdocs as cheap labour in academic research labs, a clear separation between training programmes and research programmes will help to make the biomedical research enterprise more sustainable.

A clear separation between training programmes and research programmes will help to make the biomedical research enterprise more sustainable.

In practical terms, can these proposals be converted into real changes? The answer depends on whether the key stakeholders in biomedical research and training—investigators, academic institutions, NIH, Congress and so on—learn to cooperate effectively with one another. Each stakeholder group includes vocal sub-groups who either deny existence of any sustainability problem or imagine that the problem will go away as a result of market forces.

Among all the opponents, I worry most about sincere, thoughtful investigators who know from their own experience that mentoring young scientists is a powerful way to meld mature knowledge and youthful creativity into innovation and discovery (see Bourne, 2009, especially chapter 12). Instead of viscerally rejecting this essay’s arguments and proposals, I urge such opponents to consider the following: (i) reducing the numbers of PhD students and postdocs does not mean abolishing them. More likely, the reduction in numbers is likely to be less than 15–20%, so your labs are unlikely to become populated solely by staff scientists and robots. Conversely, however, blocking the reforms proposed above will gravely endanger teaching, mentoring and the sustainability of the entire biomedical research enterprise. Moreover, the changes proposed here are quantitative and reversible: in the event of an economic boom, the NIH budget can grow again and the system can easily produce more PhDs, postdocs and independent scientists. However, if changes are not enacted to make the whole biomedical research enterprise more sustainable, it will struggle to benefit from any growth in budgets.

Finally, let me answer those stakeholders whose oppose change on some or all of the following grounds: the present system has produced what is still the very best national biomedical research effort in the world; the available data are not good enough to support far-reaching change; dramatic actions often produce unwanted consequences. They are saying that we don’t even know for sure that the patient—the biomedical research enterprise—is sick, and in any case treatments may cause harm, so no treatment is justified. Instead, I urge readers to recognize that our patient suffers from a debilitating, potentially fatal disease, with symptoms that already afflict every academic biomedical scientist in the US (Bourne 2013a). We must treat and study the disease simultaneously, beginning now! Critical treatments should be applied in gradual increments, and adjusted in accord with careful monitoring of the patient’s progress. I recommend such gradual approaches to handling PhD training and postdocs (above), and also for addressing the problem of soft-money salaries for faculty researchers (Bourne, 2013b). For decades we missed the diagnosis, which can be denied no longer. Continued dithering will constitute grave malpractice.

The author thanks Alex Gann of the Cold Spring Harbor Laboratory for providing useful information on the PhD training programme at the Watson School of Biological Sciences.

1. Version of Record published: October 1, 2013

© 2013, Bourne

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.