Official measures of inflation and consumer spending currently rely on surveys, store visits, and other old-fashioned methods of data collection. Even when there are enough staff and enough responses to generate reasonably representative statistics about various categories of sales, it is still the case that two different systems record prices and quantities in separate and independent ways rather than in simultaneous and more compatible ways. Research led by economist John Haltiwanger at the University of Maryland will demonstrate a more direct approach using the item-level transaction records of private retailers. The project, called Re-Engineering Statistics using Economic Transactions (RESET), leverages point-of-sale data that record prices, quantities, and product descriptions in real time, covering about two-thirds of U.S. retail transactions (over $3?trillion in annual sales).  The team will generate monthly inflation and spending indices on the same schedule and format as official government reports, but with the potential for greater granularity, accuracy, and timeliness.  The project will also provide a blueprint for how federal statistical agencies could adopt more modern methods like this to produce more responsive, cost-effective, and reliable economic indicators.

Consisting of 13 principal agencies as well as dozens of other officials and offices, the federal statistical system is highly decentralized. This institutional structure has widely been seen as convoluted and outdated compared with other countries, resulting in recent bi-partisan calls for reform. Researchers at the Brookings Institution will therefore lead a broad-based effort to develop and analyze options. Since form should follow function, this requires coordination with other researchers—including other Sloan grantees—who are inventing and testing ways to modernize statistical methodologies. Brookings will specifically commission white papers from experts and vet ideas with practitioners, users, and decision-makers from both the private and public sectors. The plan is for the reports and briefing materials generated to help build consensus, comprehension, and commitment about potential new institutional configurations for providing the United States with reliable economic indicators.  

Which empirical findings are truly surprising? Economist Eva Vivalt at the University of Toronto will expand and enhance how the Social Science Prediction Platform crowdsources expert forecasts of research results. The platform allows social scientists to register predictions about the outcomes of planned studies before data are collected. This will highlight which results are truly surprising, help journals recognize the value of null findings, and improve the power calculations used to design empirical experiment. The plan is to enhance the platform’s capabilities and user base, integrating it with research registries and improving its long-term sustainability. By collecting thousands of forecasts across economics, psychology, and other social sciences, the initiative will also generate data for meta-research on expert judgment and forecasting acumen. Expected outcomes include an expanded database of predictions, scholarly publications assessing forecast accuracy, and greater adoption of forecasting as a standard practice to improve research transparency.

A new interdisciplinary research center at MIT, led by economist Sendhil Mullainathan, will focus on using artificial intelligence to augment—rather than mimic or replace—human decision-making. The center will bring together experts in the economic, computational, and behavioral sciences to create systems optimized for the complementary strengths of human and artificial intelligence.  Initial plans call for developing algorithms and tools that work alongside people to address challenges in education, economic mobility, and scientific discovery.  By emphasizing AI as a tool that amplifies human intelligence, the center seeks to shift the research narrative away from automation and toward augmentation.  The goal is not only to produce scholars, experiments, and publications that draw deeply on both computational techniques and behavioral research, but ultimately to design systems that can function as “bicycles for the mind.”

The United States needs reliable, timely, and trustworthy statistics about its economic performance. Beginning with its role in the development of GDP by Simon Kuznets, the National Bureau of Economic Research has been helping answer this need for over a century. Since 2024, the Sloan Foundation has been helping NBER plan launch and incubate a new Economic Measurement Research Institute. The current grant is another step in that process. Former Treasury official Karen Dynan of Harvard will organize a series of research workshops for experts from government, academia, industry, along with other potential funders. These workshops will address stress points in the current production of economic measures, including data gaps caused by declining survey response rates. They will also explore how alternative data sources from the private-sector or from state and local governments could be put to better use. A key example is the Sloan-supported project called RESET (Re-Engineering Statistics using Economic Transactions), which will receive funding for further work on how machine learning can facilitate hedonic price adjustments when calculating inflation and consumption indicators.