Perhaps the most important economic statistic calculated by the government is the Consumer Price Index (CPI). The CPI is used as a guide for fiscal and monetary decision-making, as a deflator to compare times series data in constant dollars, and as an inflation-measure to adjust the purchasing power of federal benefits like Social Security. The basic methodology for calculating the CPI, however, has hardly changed since the statistic was first introduced over a century ago. First, using a series of voluntary surveys of businesses and consumers, the Bureau of Labor Statistics attempts to measure average monthly prices in 38 geographical areas for 211 broad consumption categories like men’s shirts. Second, the BLS estimates the proportion that each of these individual categories represents of a typical household’s spending on all 211 categories. The goal is to calculate how much a typical basket of goods and services would cost today, then compare that to the cost of the same basket at the end of some fixed period. Although this may sound straightforward, worrisome and worsening challenges arise from this methodology. First, the time required to collect responses to these surveys and aggregate data means that the CPI is not calculated for months after data is collected. Second, falling response rates to polls and surveys of all kinds means that the CPI calculation is increasingly reliant on an ever-smaller set of data, with consequent worries about accuracy. Third, the metadata collected by surveys about goods and services sold is often insufficient to gauge how much the price increase in, say, a mobile phone, reflects improvements in the phone’s quality or feature-set. Funds from this grant support a project called RESET, Re-Engineering Statistics Using Economic Transactions, led by economist John Haltiwanger (University of Maryland College Park), Matthew Shapiro (University of Michigan) and Ron Jarmin (U.S. Census Bureau). The RESET team has forged data partnerships with major online retailers and market data analysis firms to allow them access to real-time, highly granular information about the prices and features of retail products for sale online. The team proposes to use advanced machine learning techniques to solve the problems with current data collection methods. The RESET partnerships promise to be able to calculate the CPI more quickly, using more comprehensive data, and in a way that promises to give both regulators and researchers more ability to calculate how much of a price variation is due to systemic inflation and how much actually reflects changes in product quality.

Datasets about patents are among the most important tools for studying the economics of science and technology. Patent records do not tell the whole story, of course, but they do provide critical evidence about innovation, regulatory policy, and economic growth. Funds from this grant support the Innovation Information Initiative (“I-Cubed”), a diverse, interdisciplinary community of scholars devoted to enhancing research on the economics of science and productivity by increasing the coverage and usefulness of open patent data. Grant funds will support the I-Cubed networks creation and/or expansion of four innovative resources for researchers: a database of connections between patents and scientific papers; a database of connections between patents and retail products; a portal collecting patent metadata that will allow better disambiguation of inventors and inventions; and a database on patent funding that will distinguish, for example, between patents filed by established firms vs entrepreneurs or start-ups. These resources will allow researchers to ask and answer important questions about the economics of science and innovation with greater confidence. That includes questions about the return on investment to basic scientific research, identifying commonalities among innovators, the relationship between innovation and commercialization, and what sorts of funding and work structures are most supportive of patent production.

Correlation is symmetric. If some variable in a dataset, A, rises and falls in correlation with another variable, B, then B also falls and rises with A. Causal relations, in contrast, have direction. Saying “A causes B” is entirely different from saying “B causes A.” The directionality of causation makes detecting and analyzing causation using only correlations extremely difficult. Statistical methods framed entirely in terms of correlation therefore cannot express, let alone analyze or measure, causal relationships. Traditional data analysis also falls short when based on correlations that are spurious—that is, ones caused by random noise rather than fundamental processes. Big datasets do not make such challenges any easier, either. In fact, the opposite is the case. What’s ultimately needed, if we are to detect, measure, and understand causality, are more robust frameworks that draw on a richer set of concepts specifically designed to detect the inherent directionality of causal inference. Computer scientist Elias Bareinboim of Columbia University is developing just such an analytic framework. Drawing on seminal research conducted with his collaborator, computer scientist Judea Pearl, Bareinboim works with graphical methods designed to represent not only correlations between variables in a dataset, but whether and how other variables in the dataset affect those correlations. This framework allows the analyst to ask rigorous counterfactual queries of a dataset (what would have happened if…) that are essential for understanding causal relations among variables. In addition, Bareinboim’s framework is being adapted for implementation by sophisticated AI and machine learning programs. The goal is to allow them to separate causal relationships in data from non-causal correlations.   Grant funds will support Bareinboim’s research on these and related topics for a period of two years.

The Mathematical Sciences Research Institute (MSRI) is an independent nonprofit research institution and a center for collaborative research. Its events draw thousands of leading scholars from around the world each year, including intensive semesters organized around pairs of specific themes. This grant supports two semester-long programs at MSRI on market and mechanism design and algorithmic fairness. Mechanism design is a field of economics that studies procedures, assignments, and incentives that work other than through markets and prices. The program will advance research on improving the design of real-world transactions such as organ donation and medical student hospital residency matching. Algorithmic fairness, meanwhile, is concerned with understanding and correcting biases in algorithmic decision-making. Recent research has shown that some properties of algorithmic fairness are mathematically incompatible with each other, and the program will investigate what it means for predictive algorithms to be fair in ways that are both well-motivated and computationally feasible. This grant provides support for 3-6 program organizers, 15 senior research professors, 30-35 research members, 8 postdocs and 12 graduate students, for a total of 70 researchers per program, or 140 researchers in total.

Over the past 30 years, behavioral economists have succeeded in cataloging an impressive number of cognitive “biases” that manifest in how individuals make economic decisions.  These describe how real people defy the assumptions made about them in economic models. The big idea is that these biases are uniform enough across decision-makers that they can be incorporated into standard economic models, rendering the models both more accurate and more robustly predictive.  Behavioral macroeconomics is a growing field that seeks to incorporate these insights about human biases into attempts to model whole economies in a more realistic way.    Funds from this grant support a fellowship program run by Yuriy Gorodnichenko at the National Bureau of Economic Research that provides stipends to early career economists interested in conducting research in behavioral macroeconomics.  In addition to supporting the work of two fellows per year, Gorodnichenko runs an intensive every-other-summer “boot camp” to introduce new economics scholars to the concepts, methods, models and findings of behavior macroeconomics.  Topics addressed in the boot camp include the scarcity of attention, decision-making under incomplete information; the formation of expectations; optimal policy design in the presence of informational frictions; and interactions among agents with different levels of knowledge.  

The science of science is a burgeoning new multidisciplinary field that is attempting to rigorously measure and study the factors that drive scientific innovation and productivity.  Drawing from economics, public policy, sociology, history, management science, and information systems theory, researchers working in the science of science explore questions like: Can (just barely) failing to receive a big research grant be better for one’s career than winning one? Are some ways of structuring work in a lab more conducive to higher productivity than others? Can the decay rate of a scientific paper’s citations be predicted well enough to help measure the long-term impact and influence of relatively new work? And do great discoveries arrive randomly during a scientific career, or are there “hot streaks?” This sprawling new research community crosses departmental, methodological, and international boundaries.  Progress will require this community to coalesce around common standards, structures, norms, and infrastructure—particularly regarding data resources.  This grant funds efforts by Dashun Wang, director of Northwestern University’s Center for the Science of Science and Innovation, to help build community within and among science of science researchers.  Grant funds will be used to launch an annual international conference hosted by the National Academy of Sciences in 2022, planning activities for the launch of a new scholarly society dedicated to the Science of Science, and a small program of seed grants and research prizes designed to encourage diversity, data sharing, methodological training, and mentoring.

Seven of the ten most valuable businesses in the world are digital platforms. Their names are familiar to everyone: Apple, Microsoft, Amazon, Alphabet (Google), Alibaba, Facebook, and Tencent. The user base for Facebook alone includes 2.7 billion people, more than the populations of India and China combined. Google processes more than 60% of online searches in the United States, and almost 90% of those in Europe. Such companies not only wield enormous economic power, they have increasing power over our social, political, and personal lives, too.  It is unsurprising then, that lawmakers of all kinds are interested in how to regulate such platforms in a way that would inhibit this power from being excercised contrary to the public good. The economics of these digital platforms, however, is complicated.  First, most of these platforms facilitate two-sided markets, serving two distinct customer bases.  Apple’s app store serves both consumers interested in finding interesting and useful apps, and app developers interested in finding customers to sell their creations to.  In such a situation, what counts as an optimal pricing strategy- and thus what counts as worrisome deviations from it—is complicated.  It may be rational and beneficial, for instance, for Apple to undercharge phone users and make up the loss by overcharging app developers.  Second, digital platforms are often dominated by network effects.  This term refers to those goods or services that become more valuable as more and more people use them.  Vendors want to sell their wares on Amazon because that’s where the customers are, and customers shop on Amazon because so many vendors sell on the site.  Funds from this grant support a project by Fiona Scott Morton at the Tobin Center at Yale to convene a multidisciplinary working group of leading scholars to produce a compelling research agenda that lays out the fundamental theoretical and empirical research needed to advance our understanding of the economics of regulating two-sided platforms.

Funds from this grant provide support for OpenMined, an online community of nearly 12,000 members from academia, industry, and government devoted to advancing privacy-preserving research methods in machine learning and AI development.   The OpenMined community is creating an ecosystem of advanced but accessible cryptographic tools designed to allow machine learning researchers to probe sensitive datasets without the need to copy, move or share any data.  Resources available on the OpenMined website (OpenMined.org) include a beginner’s guide, free classes and tutorials in a dozen languages, blogs and lectures from leading researchers in privacy-preserving research, and open-source coding repositories and projects on such topics as remote execution and federated learning, differential privacy, encrypted computation, and secure natural language processing.  Grant funds provide core operating support for the continued operation and expansion of the OpenMined community for a period of two years.