Adriana Kugler, the Georgetown University labor economist and former Federal Reserve Governor, has a simple test for what new economic data the United States should be collecting: ask which questions policymakers, businesses, and households need answered but cannot yet answer in a timely way. In her keynote at the NBER Conference on Innovation in Measuring Employment, she argued that the test points to four gaps in how the labor market is measured, and to a solution that combines public and private data rather than choosing between them.
"As a data nerd and policy wonk, nothing excites me more than talking about data and its implications for policy work," Kugler told the audience.
The timing, she argued, could hardly be more pressing. The 2020s have brought the COVID pandemic, supply-chain disruptions, Russia's invasion of Ukraine, trade wars and tariffs not seen since Smoot-Hawley, oil disruptions not seen since the early 1980s, and technological change driven by AI. "What has been unique in recent years has been having all these supply shocks hit the U.S. economy sometimes simultaneously and at other times sequentially over a very short period of time," she said. Meanwhile, fewer resources are reaching the statistical agencies that produce official data, even as the private sector generates more and better information that could complement it.
Adriana Kugler's lessons from the policy table
Kugler drew on her own time in office. "As a policymaker, first at the World Bank and then at the Federal Reserve, I remember facing key questions we needed to address for which timely data was sometimes not available," she said. When she arrived at the World Bank shortly after Russia's invasion of Ukraine, rising food prices were spreading food insecurity worldwide. She worked with Board colleagues to ask Bank staff for a Global Food and Nutrition Security Dashboard, whose biweekly classification of countries by food insecurity conditions helped the Board prioritize short- and long-term initiatives.
At the Fed, the unemployment rate had begun rising from a 50-year low of 3.4 percent in April 2023; it would peak at 4.5 percent in November 2025. Because recessions are often accompanied by a rapid rise in unemployment, Adriana Kugler asked for a job losses dashboard that paired public data, including weekly unemployment insurance claims and WARN notices, with private sources such as Challenger, Gray and Christmas job cut announcements, natural language processing counts of layoff plans discussed on earnings calls, and online searches about layoffs. She had relied heavily on claims and WARN data as Chief Economist at the Department of Labor during the Great Recession. "My dashboard helped me identify the increased cooling in early 2024 and then the increases in layoffs and unemployment during the Summer of 2024," she said.
Four gaps in the data
Labor market slack. The vacancy-to-unemployment ratio has a break around the 2000s, when job postings moved online, which leaves a fairly short consistent series. Overtime and temporary work help, but Kugler sees earlier signals in internal margins such as the speed of work, absenteeism, and skill mismatches. Public data do not quantify the first two, while HR software platforms capture them through time tracking. "It would be helpful to merge the HR software data with these additional internal margins of slack to identify labor market weakening before it shows at the external margins and it is too late for policy-makers to respond," she said.
Labor force participation. The Current Population Survey identifies discouraged and marginally attached workers and records ill health and family responsibilities as reasons people stay out of the labor force. Less clear during and after the pandemic were the motivations behind excess retirements, adults' withdrawal from work, and weak entry by younger workers. Kugler called for a broader battery of questions on motivations, such as low pay, lack of flexibility, and feeling disrespected at work, to help policymakers design tax incentives and other policies that encourage participation.
AI and workers. Of the four gaps Adriana Kugler identified, the effect of AI on workers has drawn substantial attention from policymakers, yet measurement has focused mostly on adoption and investment. Census surveys of businesses and households have begun asking about AI use, and worker surveys by Gallup and Microsoft ask how people use AI for different tasks and what holds adoption back, including outdated corporate models and managers' limited knowledge of the technology. To judge whether worker anxiety is justified, she said, researchers need to know which tasks will be replaced, which will be enhanced, and which new tasks will be needed. She pointed to efforts that apply natural language processing to job descriptions and real-time openings to map tasks to AI exposure: "This is an example of where AI has been used effectively to analyze private data to answer a timely question which cannot be answered with the public data alone."
Job quality. The Bureau of Labor Statistics has run the Contingent Worker Supplement since 1995, but after 2005 the survey went unconducted for 12 years, and its questions have changed as gig work moved from temp agencies to online platforms. That makes consistent tracking difficult. Kugler highlighted linking administrative tax records to data from platform companies as a way forward. Other aspects of job quality also go unmeasured: the Current Population Survey covers health insurance, retirement benefits, and paid leave, but not vision, dental, disability, or life insurance, and most surveys do not quantify access to career ladders.
Public and private data as complements
Public data are representative, carefully cleaned, seasonally adjusted, and often available over long periods, Kugler noted, but they arrive with lags. Household and payroll survey data come out two to three weeks after collection, with payroll figures revised in later months. The Job Openings and Labor Turnover Survey has an initial lag of about five weeks, and Business Employment Dynamics data on gross job gains and losses arrive about seven months after the quarter ends. Delays of that length, she warned, "could imply that critical information is missing for months and perhaps only available after policy-makers have made decisions that may be incongruous with the actual economic realities."
Private data arrive faster but have their own limits. ADP releases payroll data almost immediately after the reference period, yet its clients employ only about one in five private-sector workers, so ADP reweights its data using the Quarterly Census of Employment and Wages to represent all employers.
The exchange runs both ways. Non-response in public surveys has been trending up, which Kugler attributed partly to the public's reluctance to share information with government agencies and partly to resource constraints at the federal statistical agencies. Outreach has helped at the margin, with in-person visits outperforming repeated reminders. On incentives, she noted a counterintuitive finding: "paying moderate amounts is better than paying higher amounts, which make people suspicious." Response rates nonetheless remain low, and agencies rely on imputation. Kugler argued that payroll data from ADP and from providers serving smaller businesses could complement the official payroll survey, much as private scanner data are being studied alongside official price statistics. One real constraint is that statistical agencies must obtain permission from every company whose data they use.
She also sees room to shrink release lags. Machine learning could check incoming data automatically, speeding releases with no loss of quality, and the birth-death models used to estimate business openings and closings could be estimated more often than twice a year.
For Adriana Kugler, the way forward is to use each source for what it does best. "Private data can offer information not available in public sources," she said, while public data "can allow private sources to use information to make their data representative of the population more broadly."
Adriana Kugler is Professor of Public Policy and Economics at Georgetown University's McCourt School of Public Policy and a Research Associate at the National Bureau of Economic Research. She previously served as a Governor of the Federal Reserve Board, U.S. Executive Director of the World Bank, and Chief Economist at the U.S. Department of Labor.
.png?w=600)
