Be Careful What You Wish For: A Cautionary Tale about Budget Doubling
Sometime in the near future, with timing dependent on the economy, the military actions in Iraq and Afghanistan, and other competing demands for government money, Congress will substantially boost R&D spending. It will do so in response to the great challenges facing the United States and the world—global warming, the threat of a global pandemic, rising energy and natural resource prices, and so on—whose solutions depend on increased scientific understanding and technological advance. It will also do so in response to the many reports, especially the National Academy of Sciences’ 2006 Rising Above the Gathering Storm, highlighting the importance of advancements in science and technology to U.S. economic well-being and national security.
Although federal R&D spending relative to gross domestic product has been declining during the past 20 years, between 1998 and 2003 the government increased spending in the biological and life sciences at rates that could presage a future spending boom. The Clinton administration began and the Bush administration completed a doubling of the budget of the National Institutes of Health (NIH).
At first glance, the doubling appeared to be an unalloyed benefit for medical research, but a closer examination reveals that scientists need to be careful what they wish for. The doubling did not appear to produce a dramatic outpouring of high-quality research. It failed to address critical flaws in federal research funding and actually exacerbated some existing problems, especially for younger researchers.
The negative consequences of the rapid run-up in research spending began to be felt immediately after the doubling ended, when the Bush administration and Congress essentially froze the NIH budget, resulting in a sizable drop in real spending. Indeed, one of the key lessons from the doubling experience is that if the aim is to raise aggregate R&D intensity, the United States should increase spending gradually and steadily rather than undertake a one-time surge and subsequent sharp deceleration in spending.
From 1998 to 2003, the NIH budget increased from about $14 billion to $27 billion—twice as rapidly in five years as in the previous decade. Using the Biomedical R&D Price Index, the doubling increased real spending by 66%, or about 12% per year.
Although there are situations in which sharp spending increases are preferable to steady growth, gradual increases generally are more efficient. Gradual buildups produce smaller increases in costs (because it takes time for people and resources to increase to meet the new demand and costs tend to rise nonlinearly in the short run) and avoid large disruptions when the increase decelerates. Although it is difficult to determine whether a more gradual increase in NIH spending would have produced greater scientific output than the five-year doubling, the data are consistent with the notion that the spending surge did less than a gradual buildup of funds might have done. In a November 19, 2007, article in The Scientist, Frederick Sachs noted that the number of biomedical publications from U.S. labs did not accelerate rapidly after 1999, although they did increase steadily (as they had done in the years before 1999). In addition, from 1995 to 2005, the share of U.S. science and engineering articles in the biological and medical sciences did not tilt toward these areas despite their increased share of the nation’s basic research budget, according to 2007 National Science Foundation (NSF) data.
But the big problem with a sharp acceleration of spending occurs when it ends. People and projects get caught in the pipeline. Our analysis here draws on lessons that economists have learned from studying increases in capital spending using the accelerator model of investment in physical capital. In the accelerator model, an increase in demand for output induces firms to seek more capital stock to meet the new demand. This increases investment spending quickly. When firms reach the desired level of capital stock, they reduce investment spending. This process helps explain the volatility of investment that underlies business cycles. The R&D equivalent of demand for output is federal R&D spending, and the equivalent of investment spending is newly hired researchers. We find that the young people who build their skills as graduate students or postdocs during the acceleration phase of spending bear much of the cost of the deceleration.
The deceleration of NIH spending in the aftermath of the budget doubling was particularly brutal. Using the Consumer Price Index deflator, real NIH spending was 6.6% lower in 2007 than in 2004. It is expected to fall 13.4% below the 2004 peak by 2009, according to a 2008 analysis by Howard Garrison and Kimberly McGuire for the Federation of Associated Societies for Experimental Biology. Using the Biomedical R&D Price Index deflator, real spending was down 10.9% through 2007. The drop in the real NIH budget shocked the agency and the bioscience community because it largely undid the increased funding from the doubling.
The deceleration caused a career crisis for the young researchers who obtained their independent research grants during the doubling and for the principal investigators whose probability of continuing a grant or making a successful new application fell. Research labs were pressured to cut staff. NIH, the single largest employer of biomedical researchers in the country, with more than 1,000 principal investigators and 6,000 to 7,000 researchers, cut the number of principal investigators by 9%. The situation was described in the March 7, 2008, Science as “a completely new category of nightmare” by a researcher at the National Institute of Child Health and Development, which was especially hard hit. “The marvellous engine of American biomedical research that was constructed during the last half of the 20th century is being taken apart, piece by piece,” said Robert Weinberg, founder of Whitehead Institute, in the July 2006 Cell.
In economics, the optimal path to a larger stock of capital depends on the adjustment costs. Many models of adjustment use a quadratic cost curve to reflect the fact that when you expand more rapidly, costs rise more than proportionately. If adjustment costs take any convex form, the ideal adjustment path is a gradual movement to the new desired level. Empirical studies estimate that adjustment costs for R&D are substantial as compared to those for other forms of investment. The way research works, with senior scientists running labs in which postdocs and graduate students perform most of the hands-on work, much of the adjustment cost falls on young researchers. An increase in R&D increases the number of graduate students and the number of postdocs hired. During the deceleration phase, a large supply of newly trained researchers compete for jobs when the number of independent research opportunities may be less than when they were attracted to the field. In the United States, indeed, much of the adjustment fell on postdocs, whose numbers increased rapidly during the doubling, with the greatest increase among those born overseas.
Funding and researcher behavior
A second key lesson of the budget doubling is that the way agencies divide budgets between the number and size of research grants will affect researchers’ behavior and thus research output.
Funding agencies and researchers interact in the market for research grants. An agency with a given budget decides how to allocate the budget between the number of grants and the size of grants. Researchers respond to the changed dollar value and number of research awards by applying for grants or engaging in other activities.
During the doubling, NIH increased the average value and number of awards, particularly for new submissions, which include new projects by experienced researchers as well as projects by new investigators. With the success rate of awards stable at roughly 25%—the proportion the agency views as desirable on the basis of the quality of proposals—the number of awards increased proportionate to the number of submissions. From 2003 to 2006, when the budget contracted, NIH maintained the value of awards in real terms and reduced the number of new awards by 20%. But surprisingly, the number of new submissions grew, producing a large drop in the success rate. In 2007, NIH squeezed the budgets of existing projects and raised the number of new awards.
The interesting behavior is the response of researchers to changes by funding agencies in the number of awards granted and thus the chance of winning an award. An increase in the number of awards increases the number of researchers who apply, because the chance of winning increases. As might be expected, researchers responded to the NIH doubling by submitting more proposals. Given higher grant awards and increased numbers of awards (with roughly constant funding rates), the growth of submissions reflects standard supply behavior: positive responses to the incentive of more and more highly valued research awards.
But researchers also increased the number of submissions when NIH research support fell. The number of submissions per new award and per continuation award granted rose from 2003 to 2007 after changing modestly during the doubling period. By 2007, NIH awardees were submitting roughly two proposals to get an award. Fewer investigators gained awards on original proposals, which induced them to amend proposals in response to peer reviews to increase their chances of gaining a research grant. The response of researchers in submitting more than one proposal to funding agencies produced the seemingly odd short-run supply behavior: more proposals with lower expected rewards. Faced with the risk of losing support and closing or contracting their labs, principal investigators made multiple submissions, although over time, normal long-run supply behavior would be expected to lead those who do not gain awards to leave research science and to discourage young people from going on in science. Who wants to spend time writing proposal after proposal with modest probabilities of success? It may also lead to more conservative science, as researchers shy away from the big research questions in favor of manageable topics that fit with prevailing fashions and gain support from study groups.
The message is that to get the most research from their budgets, funding agencies need good knowledge about the likely behavior of researchers to different allocations of funds. NIH, burned by its experience with the doubling and ensuing cutback in funds, hopefully will respond differently to future increases in R&D budgets.
Young researchers take a hit
A third lesson of the doubling experience is that increased R&D spending will not resolve the structural problems of the U.S. scientific endeavor that limit the career prospects of young researchers and arguably discourage riskier transformative projects.
At the heart of the U.S. biomedical science enterprise are the individual (R01) grants that NIH gives to fund individual scientists and their teams of postdoctoral employees and graduate students. The system of funding individual researchers on the basis of unsolicited applications for support comes close enough to an economist’s views of a decentralized market mechanism to suggest that this ought to be an efficient way to conduct research as compared, say, to some central planner mandating research topics. The individual researchers choose the most promising line of research based on local knowledge of their special field. They submit proposals to funding agencies, where expert panels provide independent peer review, ranking proposals in accordance with criteria set out by funding agencies and their perceived quality. Finally, the agency funds as many proposals with high rankings as it can within its budget.
Although there are alternative funding sources in biomedical sciences, NIH is the 800-pound gorilla. For most academic bioscientists, winning an NIH R01 grant is critical to their research careers. It gives young scientists the opportunity to run their own lab rather than work for a senior researcher or abandon research entirely. For scientists with an NIH grant, winning a continuation grant is often an implicit criterion for obtaining tenure at a research university.
It is common to refer to new R01 awardees as young researchers, but this term is a misnomer. Because R01s generally go to scientists who are assistant professors or higher in rank, and because postdoctoral jobs last longer, the average age of a new R01 recipient was 42.9 in 2005, up from 35.2 in 1970 and 37.3 in the mid-1980s. In 1980, 22% of grants went to scientists 35 and younger, but in 2005, only 3% did. In contrast, the proportion of grants going to scientists 45 and older increased from 22% to 77%, and within the 45 and older group, the largest gainers were scientists aged 55 and older.
Most of this change is due to the structure of research and research funding, which gives older investigators substantive advantages in obtaining funding and places younger researchers as postdocs in their labs. Taking account of the distribution of Ph.D. bioscientists by age, the relative odds of a younger scientist gaining an NIH grant as compared to someone 45 and older dropped more than 10-fold. The doubling of research money did not create this problem, which reflects a longer-run trend, but it did not address or solve the problem. The result is considerable malaise among graduate students and postdocs in the life sciences as well as among senior scientists concerned with the health of their field, as has been well documented in a variety of studies. More money is not enough.
Bolstering younger scientists
A final lesson that we derived from the doubling experience is that funding agencies should view research grants as investments in the human capital of the researcher as well as in the production of knowledge, and consequently should support proposals by younger researchers over equivalent proposals by older ones.
There are three reasons for believing that providing greater research support for younger scientists would improve research productivity.
First, scientists may be more creative and productive at younger ages and may be more likely to undertake breakthrough research when they have their own grant support rather than when they work as postdocs in the labs of senior investigators. We use the word “may” here because we have not explored the complicated issue of how productivity changes with age.
Second, supporting scientists earlier in their careers will increase the attractiveness of science and engineering to young people choosing their life’s work. It will do this because the normal discounting of future returns makes money and opportunities received earlier more valuable than money and opportunities received later. If scientists had a better chance to become independent investigators at a younger age, the number of students choosing science would be higher than it is today.
The third reason relates to the likely use of new knowledge uncovered by researchers. A research project produces research findings that are public information. But it also increases the human capital of the researcher, who knows better than anyone else the new outcomes and who probably has better ideas about how to apply them to future research or other activities than other persons. If an older researcher and a younger researcher are equally productive and accrue the same additional knowledge and skills from a research project, the fact that the younger person will have more years to use the new knowledge implies a higher payoff from funding the younger person than from funding the older person. Just as human capital theory says that people should invest in education when they are younger, because they have more years to reap the returns than when they are older, it would be better to award research grants to younger scientists than to otherwise comparable older scientists.
Making adjustments
Future increases in research spending should seek to raise sustainable activity rather than meet some arbitrary target, such as doubling funding, in a short period. There are virtues to a smooth approach to changes in R&D levels, because it takes considerable time to build up human capital, which then has a potentially long period of return. Since budgets are determined annually, the question becomes how Congress can commit to a more stable spending goal or how agencies and universities can offset large changes in funding from budget to budget. One possible way of dealing with this issue is to add extra stabilization overhead funding to R&D grants, with the stipulation that universities or other research institutions place these payments into a fund to provide bridge support for researchers when R&D spending levels off.
To deal with some of the structural problems in R&D funding, our earlier argument that younger investigators have longer careers during which to use newly created knowledge than do equally competent older investigators suggests that future increases should be tilted toward younger scientists. In addition, given multiple applications and the excessive burden on the peer review system, agencies should add program officers and find ways to deal more efficiently with proposals, as indeed both NIH and NSF have begun to do.
In sum, there are lessons from the NIH doubling experience that could make any new boost in research spending more efficacious and could direct funds to ameliorating the structural problem of fostering independence for the young scientists on whom future progress depends.