What your NSF dollar gets you: understanding of public opinion and technological spinoffs

by Andrew Gelman on May 11, 2012 · 16 comments

in Science

Congressman Flake wants to zero out political science funding from the National Science Foundation. Flake makes two points:

1. “Three out of the four of the grants awarded by the NSF Political Science Program go to the wealthiest universities in the country. Would those who would oppose this amendment have believed that Harvard and Yale would have to close their political science departments if Federal grants are not available for this program? Of course not. These universities and the field of political science will be just fine.”

2. “$600,000 here spent trying to figure out if policymakers actually do what citizens want them to do. I think we can answer that question in about 5 minutes when we vote on this amendment because I can tell you, people out there want us to quit funding projects like this.”

Flake’s first point regards a tradeoff between equity and efficiency. I get NSF grants and I don’t need the money. On the other hand, I think that I produce some useful public goods which would not exist (or would come out years later) if this funding were not coming to me. Harvard, Yale, Columbia, etc. would do just fine if we got no NSF grants, but we wouldn’t be able to do as much: that would be less education, less research, and less service. Is it a good idea to use tax money to help rich institutions do more? That’s not my call.

Flake’s second point is that it’s not useful to try to learn the connections between public opinion and policy. I don’t know why he says that. First off, policymakers are not always aware of public opinion. You might know how many of your constituents are angry enough to send you emails on some issue of the moment, but that doesn’t tell you so much about the attitudes of the silent majority. With the help of NSF funding, we’ve been using multilevel regression and poststratification to estimate public opinion at the state level to learn more about what Americans think. Second, policymakers don’t always do what citizens want them to do! We also have a lot of unelected policymakers in this country. I think it’s valuable to understand the policy areas in which policies are in accord with public opinion, and where there is discord.

Finally, NSF-funded work has spinoffs. The statistical tools that we have and are developing are publicly available and open-source. This means two things. First, anyone can use our methods. All you have to do is download R and you can get going. You don’t have to be at Harvard, Yale, or any university at all to use multilevel regression and poststratification to analyze data from the National Election Study and General Social Survey (these are all NSF-supported projects), and soon you’ll be able to fit all sorts of models in Stan, another federally-funded project. Second, everything we are developing is free, but that does not stop businesses from using our methods and software to innovate. Indeed, after one of my talks, someone came up to me and said that the company he worked for had recently had much success using multilevel regression and poststratification.

NSF funding is for basic research but it’s no surprise that publicly-available research tools developed by our leading university researchers can contribute to the economy.

P.S. Full disclosure: My research has been supported by the National Science Foundation almost continuously for over 30 years. So of course I think it’s a good thing for NSF to be supporting political science research at Harvard, Yale, etc. Some of my reasons for this view are given above.

P.P.S. Some of the commenters below express skepticism. Nobody seems to be claiming that my NSF-funded work isn’t good value for money, but commenters are arguing that I need a better argument in order to convince Congress. That’s fine. The point of this post is not to convince Congress, actually I’d feel a bit uncomfortable doing this since I am personally benefiting from these grants. I wanted to present my perspective, which happens in this case to directly address Rep. Flake’s two points. The rest of you can take it from there. If Congress judges that these projects aren’t worth the money, that’s their call. I’d just like to lay out how our work addresses the concerns laid out above.


Chris May 11, 2012 at 11:00 am

I think you would all be a lot more convincing to Flake if you argued that NSF funding of political science helped us get a better understanding of why excessive government spending occurs, and how NSF funding can therefore help achieve the goal of cutting spending overall.

The “NSF funding is great because it means more money for me” line of argument that’s already popped up several times on this blog couldn’t be more counterproductive.

I imagined that political scientists would be far better skilled at lobbying than this.

RobW May 11, 2012 at 11:55 am


Andrew Gelman May 11, 2012 at 6:49 pm

Chris, Rob:

I’m not trying to lobby here, I’m just trying to tell the truth. If the truth encourages Congress to keep the funding, that’s great. If not, that’s they’re decision. I’m not good at being strategic so I don’t even try.

RobC May 11, 2012 at 11:01 am

I wonder if you could help explain something about NSF political science grants to those of us, including politicians, who are not very familiar with them. Are the grants used entirely to pay actual out-of-pocket expenses of the research (e.g., participation fees for questionnaire respondents and similar study participants, travel expenses, wages and salaries for grad students and non-professional staff, copying and printing costs)? Does any portion go to defray the universities’ general and administrative expenses? Does any portion go to the principal investigator?

The last question is important, I reckon, because faculty members are employed by their universities, and an important part of their professional responsibilities as faculty is to conduct research. That is, of course, the reason why teaching loads alone are not very arduous. (I’m referring here to major universities, not junior colleges.) If faculty members are receiving additional income for, in effect, doing their jobs, that would seem to be worthy of public discussion as to whether it represents an appropriate expenditure of public funds. On the other hand, if the grant money is all being paid to third parties, that’s a useful argument as to why it should continue.

It would also be helpful, I think, to put the $11 million of NSF political science annual grants in some context. What are the sources of political science research funding (university, foundation, NSF, other governmental grants, etc.), and in what amounts? To what extent is the type of research funded by NSF eligible for funding by these other sources? If it is, then the defunding of the NSF political science program would not necessarily mean that the types of studies supported by NSF would not be performed, it would mean they’d compete with other research projects for available funds, and one would assume the losers in that competition would not be the important research projects you and others have discussed but less important projects whose value or methodology is open to greater question.

RobC May 11, 2012 at 5:08 pm

Sorry that no information is response to my questions has been forthcoming, either from Professor Gelman or, for questions I posted similar to those in the third paragraph of my comment, from Professor Sides or Voeten. I’m afraid the logical inference is that the answers wouldn’t advance the case for continued NSF funding. That’s understandable, though one would hope that someplace there would be facts sufficient to permit an informed discussion about public policy.

John Sides May 11, 2012 at 5:49 pm

RobC: My apologies for not answering your questions in a manner timely enough to suit you. Apparently 7 hours is just enough time for you to impute the worst possible motives to us. Without further ado:

Q: Are the grants used entirely to pay actual out-of-pocket expenses of the research (e.g., participation fees for questionnaire respondents and similar study participants, travel expenses, wages and salaries for grad students and non-professional staff, copying and printing costs)? Does any portion go to defray the universities’ general and administrative expenses? Does any portion go to the principal investigator?

A: The majority of most grants pays for research expenses such as data collection, wages for graduate students, and the like. Universities also receive a portion of the grant, known as indirect costs, which constitute part of their general revenue stream. In my experience, the most a principal investigator will personally receive is (pro-rated) salary for a summer or maybe two summers. Faculty salaries are 9-month salaries, which can be supplemented by teaching summer classes, when available. Summer salary from a grant provides similar compensation without the necessity of teaching — thereby, the theory goes, allowing the researcher to focus on the research project that is being funded.

Q: It would also be helpful, I think, to put the $11 million of NSF political science annual grants in some context. What are the sources of political science research funding (university, foundation, NSF, other governmental grants, etc.), and in what amounts? To what extent is the type of research funded by NSF eligible for funding by these other sources?

A: There is no easy way to answer the first question. Political science research is funded by a range of other sources both within government (DoD, NIH) and outside it, such as philanthropic foundations (MacArthur, Carnegie, Russell Sage, Robert Wood Johnson). There is no centralized “count” of how much those sources give to fund political science research.

To answer the second, the NSF provides funding for many projects that would not be eligible for funding from other sources. Projects unrelated to national security or health wouldn’t be funded by DoD or NIH, respectively. Projects unrelated to health policy wouldn’t be funded by RWJ. Other foundations have their own program emphases.

For example, a lot of research in American politics — the subfield I know best — doesn’t easily fit under most other foundation’s mandates. There aren’t very many funding sources for research projects about American political institutions, public opinion, elections, and the like. For this reason, the elimination of NSF funding would put real constraints on political scientists.

RobC May 11, 2012 at 7:58 pm

Thanks for your response. I’m sorry if I was impatient, but since I’d asked both <a href="you and Professor Voeten about the amount of other funding more than a day ago and you’d replied to several other comments but not mine, I feared those questions and my questions to Professor Gelman had all slipped into the same crack.

It’s extraordinary that your professional association, the APSA, doesn’t have a handle on the total funding of political science research; not enough quantitative analysts, I surmise. Very quick research by me about just the sources you mentioned seems to indicate that other sources of funding are an order of magnitude or two greater than the NSF funding of the political science program. That doesn’t mean that the NSF funding isn’t important, but it does suggest that the real concern about defunding isn’t the $11 million but rather the message it sends that political science is not on a par with the other sciences funded by the NSF.

Finally, I hope you’ll forgive my questions about the benefit of NSA grants to principal investigators. Academics, like other government contractors, are subject to scrutiny for what they receive in public funds, and the increased attention to the unequal benefits enjoyed by the rich (defined by the President as those with family incomes over $250,000) and the perceived need to identify potential savings in government expenditures make these issues especially pertinent. As Representative Flake’s assault on wealthy universities makes clear, the President’s populism and the Occupy ethos can have application in all kinds of ways their cheerleaders may not have anticipated.

Andrew Gelman May 11, 2012 at 8:51 pm


I agree this is an important issue; that’s why in the above post I mentioned the tradeoff between equity and efficiency. Congress could, for example, pass a rule restricting NSF funding to investigators with salaries under $100,000. Or they could set a rule of no funding to universities with endowments larger than $1 billion or no funding to universities whose presidents make more than $100,000. Similarly, they could require that government agencies, when buying products such as cars or computers, only buy from companies that pay their CEOs and engineers less than some prespecified amount (perhaps that $100,000 again). I do not know what Rep. Flake’s views are on Occupy Wall Street, but maybe these sorts of rules would make sense from that perspective.

John Sides May 11, 2012 at 10:50 pm

RobC: On brief follow-up: yes, philanthropic foundations have resources. But remember that most foundations are supporting work in a variety of disciplines, not just political science. If I had to guess, the NSF provides political science more than various private foundations combined. I should also note that the NSF supports “infrastructure” — large social science projects like the American National Election Study — to the tune of several million per year, something that private foundations are unlikely to do.

I believe that APSA is trying to do the necessary legwork to count all this up. But it’s harder than you might imagine. It’s likely they will have to do it by contacting individual departments about which scholars have which grants, rather than attempting to do some census of available funding sources. I certainly agree, however, that we need to know.

Newbie May 12, 2012 at 3:08 pm

I think another important point is that NSF funding is technically not as driven by oftentimes partisan political agendas as other types of available funding. Many of the private foundations that fund political science research DO have discrete agendas and want to see confirmation of specific ideas instead of actual exploration of political phenomena. This may be a somewhat paranoid take on our sources of alternative sources of funding, but I don’t think it’s entirely unreasonable. The NSF allow us to conduct objectively important research without the taint of partisanship–well, hopefully…

idiot May 11, 2012 at 11:50 am

“Is it a good idea to use tax money to help rich institutions do more? That’s not my call.”

It is however, the US Congress’ call.

While I like preserving NSF programs, I do express great concern if people always protest cuts to their own programs and try to cast their opposition as “we’re providing a public good”. I think NSF programs must be kept, because they are still very useful…but somewhere down the line, somebody down the line has to lose government funding. It’d probably would be better if the NSF sponsored a program to find cuts within other government programs.

Andrew Gelman May 11, 2012 at 6:54 pm

Indeed, when I wrote “that’s not my call,” I was indeed noting that it is Congress’s call. I’m just providing information here; Congress can decide how to spend our tax money. That’s their job. But I’d prefer they do their job using the best available information. Rep. Flake seemed to be under the impression that it was pointless to fund research on public opinion. I was giving some reasons why studying public opinion is not so easy and is in my opinion a useful way to spend research dollars.

Matt May 12, 2012 at 7:54 am

It’s funny how simple Flake thinks it should be to eliminate $11m in funding for research because “people out there want us to quit funding projects like this,” but he can’t find it in himself to vote to end the Big Oil subsidies, pass the Buffett Rule and a public option, or to do something about global warming, etc.

Joe Jupille May 12, 2012 at 10:11 am

Agreeing with the premise that it might be useful for everyday scientists funded by NSF Political Science to share our stories, here’s mine. From my perspective as the researcher and as an American taxpayer, I believe the process worked very well.

When an Icelandic referendum on “Icesave” debt was announced in January 2010, my colleague David Leblang (University of Virginia) and I saw a scientific opportunity: the chance to understand the political behavior of everyday European citizens around the question of sovereign debt resettlement. Having lived the US crisis ca. 2007-2009, and watching it become global from 2008- present, we knew that understanding what would happen could be very important, not least for the United States.

The referendum would be held sixty days from the announcement, i.e., April 6, 2010, and so the turnaround from insight to implementation was extremely tight. I submitted a proposal to the National Science Foundation’s Political Science Program (SES Directorate), under the RAPID protocol. RAPID recognizes that science sometimes demands quick responses, as, e.g., after natural disasters, or after unique events like a national referendum on debt resettlement. The proposal passed NSF review (in general, perhaps the most rigorous scientific review system in the world) and was funded: “RAPID: A Referendum on Debt: The Political Economy of Icesave,” Award SES 1035102 ($80,032), May 1, 2010-April 30, 2011.

We have found that material self-interest, direct “pocketbook” factors, helped shape peoples’ votes. If this sounds obvious to anyone, it shouldn’t: previous science actually quite strongly suggests that we would not find these effects, which if they operate at all are swamped by more general symbolic attachments and domestic political cues. The American taxpayer should care deeply about understanding the behavior of mass European publics toward sovereign debt, since the politics of ongoing crisis could well transmit strong economic effects across the Atlantic, potentially directly impacting US households. Policymaking, in turn, benefits from these new data and findings addressed directly to an issue of pressing and potentially widespread concern.

John Sides May 12, 2012 at 1:59 pm

Thanks very much for describing your project, Joe.

DrugMonkey May 13, 2012 at 3:44 pm

If it makes you feel any better, every time we get a half decent discussion of some scientific topic related to marijuana, a reliable contingent starts braying about te need to assess social/political policy wrt legalization/enforcement.

Comments on this entry are closed.

Previous post:

Next post: