Progress in understanding the relationship between basic research and economic growth requires high-quality data on patents and patenting. Barriers to acquiring, cleaning, and sharing such data remain a significant hurdle to conducting empirical research on a wide range of topics, including the return on investment to basic science investment, the productivity of scientific teams, regulatory impacts on patenting and innovation, and much more. This grant provides funding to the “Innovation Information Initiative,” or I3, a collaborative project to build a linked series of state-of-the-art, open databases that make high-quality patent data easily available to researchers. Led by Yarn Labs, a not-for-profit spin off of the MIT Media Lab, the project will clean and document existing sources of patent data; create new data products that include a catalog of links between patents and products; disambiguate authors, institutions, funders, and titles; and compile patent citations to the scholarly literature. To facilitate use of these new resources, the team will develop user-friendly interfaces and a series of models, algorithms, and other analysis tools. Outreach plans include organizing an annual research meeting alongside the NBER Summer Institute; an annual meeting to coordinate technical matters; and fellowships for Ph.D. students interested in the rigorous study of patenting.

The Summer Institute run by the National Bureau of Economics Research (NBER) is an annual gathering of empirical economists that takes place over three weeks of July in Cambridge, Massachusetts. It is widely recognized as the most significant meeting of its kind. In 2018, more than 2,200 economists from 427 institutions participated in at least one of over 50 workshops on a wide range of economic topics, from behavioral macroeconomics to innovation and digitization. Indeed, organizers of the Summer Institute work hard to attract the highest quality research and researchers. The selection process is highly competitive. In 2018, for example, over 5,800 submissions were made and fewer than 10% were accepted. The structure and range of activities taking place at the Summer Institute make it a unique venue for cross-fertilization of ideas. Workshops and lunches are deliberately scheduled to overlap in order to increase the likelihood of interaction between researchers interested in related fields. Participant surveys show high satisfaction with the quality and breadth of the presentations, and appreciation for the opportunity to meet other researchers and initiate new collaborations. This grant provides operating support to NBER to continue to host the Institute for three years.

Administrative data can enable a wide variety of new and more accurate statistical work by researchers in academia, government, business, and other organizations. In the United States, no statistical findings are more fundamental for economists and social scientists than those produced by the Census Bureau, including the Decennial estimates of population counts and characteristics. But declining response rates to any kind of survey, not to mention concerns about interference and privacy, have raised serious apprehensions about the accuracy of the upcoming count in 2020. This is an equity issue in addition to a statistical one, since concern is particularly acute about miscounts of historically underrepresented or underprivileged groups.   Funds from this grant support a project by the Massive Data Institute (MDI) at Georgetown to use state-level administrative data to bolster the Decennial Census. MDI will work with various states to identify government sources of administrative data that describe people from across geographic areas and subpopulations; determine how to access the data legally and securely; develop tools to clean and standardize that data; and provide documentation, training, and protocols for use by other states.

An important question in the social sciences is how researchers can responsibly study large datasets containing confidential information about individuals, and how organizations can safely share such data while preserving the privacy of their users. Every query answered inevitably leaks some privacy. The only conceptual framework for specifying, measuring, and controlling such leakage is known as “differential privacy.” Imagine analysts who use a query mechanism to interrogate a dataset held by a trusted curator. An example of a differentially private mechanism would be one that returns an answer after adding a small amount of random noise drawn from a carefully selected distribution. That noise provably limits whether the analyst can even find out whether any given person is in the dataset, let alone anything else about that individual. Sloan has been an early funder of the development and application of this approach. This grant funds efforts by Salil Vadhan of the Harvard Privacy Tools Project to create a library of industrial-strength, open source differential privacy software called OpenDP. Like other open source development communities, OpenDP participants will cooperate to develop trusted, robust, and scalable tools that are easily accessible and adoptable in a wide variety of settings. Vadhan will convene a group of experts and users to guide the project’s overall architecture, features, progress, and sustainability planning.

What constitutes progress in empirical social science research? It is easy to talk about “new knowledge,” but hard to measure it. To evaluate what constitutes new knowledge requires measuring not only what scholars think of an experiment after it is performed, but also what they forecast about the results in advance. This grant supports an initiative by a team led by Stefano DellaVigna at the University of California, Berkeley to build, test, and study an online platform for collecting and cataloguing forecasts and prior beliefs about the results of social science experiments. DellaVigna and his team will write up summaries of intended experiments that forecasters can quickly read and absorb, and will find interesting. Those experts would, in turn, use the developed platform to enter their best estimates about how the experiment will turn out, including their predictions of the effect sizes. These answers can then be compared to the actual results of the experiments. The platform DellaVigna and his collaborators will develop has the potential to allow researchers to capture forecasts, control publication bias, improve the transparency and reproducibility of research designs, and advance the use of Bayesian methods more generally by helping quantify prior probabilities.

Founded with Sloan Foundation support in 2013 and based at MIT, J-PAL North America is a regional hub of the Abdul Latif Jameel Poverty Action Lab, a global network of economic researchers who share a dedication to promoting the use of randomized controlled trials (RCTs) in economics. RCTs are a powerful research methodology in social science, the “gold standard” for drawing robust conclusions about causal relationships. Researchers affiliated with J-PAL North America have completed nearly 100 rigorous studies of a wide range of relevant issues, including the effect of summer jobs on juvenile delinquency, whether you can improve student performance by paying for good grades, what information can reduce consumer reliance on payday loans, and a host of other studies on the topics of voting, crime, discrimination, and health care. This grant provides core operating support to J-PAL North America for the continuation of these and related activities for three years. Funded activities under this grant include a new series of Design Within Reach forums, in which researchers exchange early feedback on the design of planned RCTs; a Short-Term Research Management program that provides hands-on administrative assistance to researchers during the early stages of planning, piloting, and implementing an RCT; an outreach initiative to promote interactions between researchers and policymakers; and an ongoing collaboration with the global J-PAL network to address the challenges and opportunities posed by administrative data.

This grant to the National Bureau of Economic Research (NBER) supports a predoctoral fellowship program that will support Ph.D. students carrying out research in behavioral macroeconomics. Four dissertation fellowships will be awarded to promising graduate students for a period of two years each. Potential dissertation topics include the aggregate implications of limited information, decision-making with flawed models, how expectations are formed, optimal policy design, and the interaction of economic agents with different levels of knowledge or economic sophistication. In addition to the fellowships, grant funds will support an NBER initiative to expand the community of behavioral economists by holding one-week boot camps each year that will train two dozen young scholars in topics in behavioral macroeconomics.

The Abdul Latif Jameel Poverty Action Lab (J-PAL) is a global network of economic researchers who share a dedication to promoting the use of randomized controlled trials (RCTs) in economics as a way of generating real and reliable solutions to important social problems. J-PAL’s 170 affiliated researchers have conducted more than 800 high-quality RCTs since its founding on a wide range of issues including poverty, crime, discrimination, education, and voting behavior. The network has also trained some 1,500 researchers on how to effectively design and implement randomized controlled trials and has launched and promoted an RCT registry now housed at the American Economic Association.   This grant supports a new initiative by J-PAL Global, called Innovations in Data and Experiments (IDEA), which aims to facilitate and improve researcher access to and use of administrative data. J-PAL Global will encourage data holders to work with researchers, provide technical, legal, and practical guidance to researchers who want to work with administrative data, and help avoid duplicate efforts to obtain, clean, and document administrative datasets. Grant funds will support IDEA’s core technical staff, two pilot projects in the United States, and the creation of several handbooks and workshops on using administrative data in economic experiments.