Improving Think Tanks

Today's think tanks have room to improve. Our policy recommendations often restate common generalities with insufficient evidence. They may be politically impractical, but political constraints are rarely addressed. Our public events are usually one-sided discussions. We avoid public debate, and invite the same people to speak over and over again. We tend to work in small groups and rarely collaborate within our own institutions, let alone across them. We build enormous databases for research projects but neglect to publish them as a public resource even after our research is complete. These problems are widely acknowledged among think tankers but we've failed to address them.

We can do better. Here are some ideas:

  1. Prioritize rigorous, relevant policy research. This is, after all, why think tanks exist. We need to develop interesting and important hypotheses, conduct thorough literature reviews, collect large amounts of relevant qualitative and quantitative data, and rigorously test their hypotheses. In a perfect world, we would not align ourselves ideologically or reverse engineer conclusions to our research (this is a common practice today). Our research does not require academic peer-review, but it should still undergo a strong vetting process. We may not be the first to publish on an issue, but we need to get the story right. We should also refuse to work in silos and incorporate research and practices from political science, history, anthropology, data science, and elsewhere.
  2. Offer clear, actionable recommendations. Every publication with policy recommendations should identify the actors (individuals or organizations) that can implement them and what resources (financial, personnel, or otherwise) they may need to do so. If a report is highlighting broader guidelines rather than providing specific policy recommendations, it should say so upfront, and the takeaways should still be clear. But it's usually better to give action items. Leave it to academic political scientists to say interesting things without clear implications.
  3. Collaborate more. We should do so both within our own institutions and across the wider think tank community. Project Atom (CSIS, Stimson, CNAS, NIPP) and Defense Futures Simulator (CSIS, AEI, & War on the Rocks) are great examples of what we can achieve in joint teams. We should of course work effectively in these collaborative projects, meaning we structure contracts to avoid any diffusion of responsibility and ensure adequate funds for all researchers involved. Think tanks should also seek to collaborate more with graphic designers and artists in developing our websites and reports. RAND's Art + Data initiative is a great case study here. Those with limited budgets might work with art students instead of professionals. In general, good think tanks work with everyone, from retired senior policymakers to high school students willing to spend days on data collection. They believe in supporting the next generation of policy analysts—their interns and junior staff with soon be prized researchers at other institutions.
  4. Build more. We should consider ourselves "builders" as well as "analysts." We should develop more open research databases to make research more efficient across the community. It's ridiculous that everyone is fighting for library copies of IISS's latest Military Balance reports. We should instead have a Missile Threat-like website for every important military platform (e.g. tanks, aircraft, and ships). They should exclusively use publicly-available information. Wikipedia is okay for some things, but we can do better. We might turn academic papers into briefings and static graphics into interactive features. We should create tools and resources that empower others to learn and conduct their own policy research.
  5. Create larger, open-source projects for which anyone can contribute. Investigative groups like Bellingcat and Arms Control Wonk are good at this, but few established policy think tanks have followed in their path. We should build request portals for various research projects where anyone can contribute and get credited. In order to keep collaboration possible without degrading quality, research leads should develop robust oversight mechanisms and publications must be thoroughly fact-checked.
  6. Invite more debate. Event moderators should be trained to encourage debate among panelists while keeping conversations cordial, and to clarify precisely the speakers' points of agreement and disagreement. Our publications should steelman opposing arguments, maybe even systematically developing sections in our reports to do so, like those produced in the EA community.
  7. Diversify and amplify our networks. Good think tanks raise the voices of smart academics, activists, and policy wonks who address critical public policy issues. We should invite folks from around the country—especially outside the Beltway—to our events, workshops, and peer review networks. We should actively seek out academics and journalists who use and cite our research and ask them what future products they would be most interested in. In general, we should think more about our function as an intermediary between academia and journalism.
  8. Revamp long and complex grant lifecycles. The grantmaking processes and requirements we have today take days to complete, taking researchers away from the projects that they were hired to work on. We especially need U.S. government grants to streamline processes so we can work more frequently with government agencies. Looking back to bullets #1 and #2, funders should also support collaborative projects between think tanks and database development projects that would fuel interest and wider research on a subject.
  9. Minimize overhead, maximize flexibility. Think tanks must manage their finances wisely. Funding should go primarily to research staff and research enablers like travel, consulting, and publishing. Smaller institutions should avoid shackling themselves with golden handcuffs—aka a fancy DC office. A small office may limit conferencing capabilities, but after Covid-19, public audiences have demonstrated their preference for virtual events. Additionally, there are already enough institutions with hosting capabilities with which smaller teams can join forces.
  10. Be transparent about donors. Think tank research may be sponsored by the U.S. government, foreign governments, philanthropic institutions, and/or defense industry. But every publication should explicitly list its sponsors, and total institutional funding should be available online. Research staff of all levels should be trained on how to pitch projects while maintaining research independence.
  11. Be honest about research novelty. Think tank researchers do not need to only produce novel research. In fact, it's good for our work to build off of academics and journalists. But if we're reproducing older materials, we should say explicitly say so, cite our sources, and explain why our new product is useful.