Maps offer neighborhood-level insight into American migration

California's most devastating wildfire - the 2018 Camp Fire, which killed 85 and destroyed nearly 19,000 structures - forced nearly half of all residents living within designated fire perimeters to relocate within a year.

That local exodus is documented by Cornell-led research that mapped annual moves between U.S. neighborhoods from 2010 to 2019 in detail 4,600 times greater than standard public data. Called MIGRATE, the new, publicly available datasetrevealed that most of those displaced remained within the affected county - moves not captured in county-level public migration data aggregated every five years.

The case study illustrates how a finer-grained understanding of migration could help researchers and policymakers investigate a range of issues at the national and local levels, from natural disasters to school funding, affordable housing and economic segregation. About 44 million people move every year in the U.S., but most of those moves are invisible in official data.

"MIGRATE reveals trends that shape daily life that broader data completely miss: rising moves into top-income neighborhoods, racial gaps in upward mobility and local shocks like post-wildfire outmigration," said Maria Fitzpatrick, professor in the Department of Economics and the Cornell Jeb E. Brooks School of Public Policy. "For communities, journalists, researchers and planners, it's a trustworthy, neighborhood-level lens on climate risk, housing pressure and opportunity - available for nonprofit research use."

Fitzpatrick and co-authors detail MIGRATE's development and validation in "Inferring Fine-Grained Migration Patterns Across the United States," published in Nature Communications (fully open access as of early February, after an abstract was posted Dec. 26, 2025). Co-authors are Nikhil Garg, assistant professor of operations research and information engineering at Cornell Tech, the Jacobs Technion-Cornell Institute and Cornell Engineering; Gabriel Agostini, a doctoral student at Cornell Tech; Emma Pierson, assistant professor of computer science at the University of California, Berkeley (formerly Cornell Tech); and Rachel Young, assistant professor at the University of Minnesota.

The researchers sought to addresses limitations in available migration data. Free government sources such as the 10-year census or annual American Community Survey estimates are accurate but produced infrequently or with small samples due to their expense and only reflect moves between counties. Commercial sources mining public and digital data - from voting and homeownership records to cell phone use - to track people's movements may improve granularity but include errors and biases, such as overrepresenting certain regions and incomes.

The MIGRATE team used data and computer science tools like those employed in artificial intelligence to leverage the strengths, and minimize the weaknesses, of publicly available data paired with purchased consumer data. An iterative process of probabilistic modeling produced maps of annual migration between 47 billion pairs of Census Block Groups (CBGs), enabling roughly neighborhood-level views.

Several case studies highlight the methodology's potential to deepen understanding of national and local migration patterns or generate new insights. For example, over the decade studied, MIGRATE showed that nationally, people were more likely to move into high-income neighborhoods, but with large racial disparities. Residents of primarily Asian communities were nearly twice as likely as others to move into top-income neighborhoods, while those from primarily Black communities were twice as likely to move into bottom-income neighborhoods.

"Upward mobility is a core concern for our society and economy. Understanding who can afford to move into higher-income neighborhoods can shape housing policies, investments in education and urban infrastructure, and narratives about residential segregation," Agostini said. "Our dataset enables us to ask questions at a very detailed spatiotemporal level, adding nuances about subpopulations that weren't previously possible."

Zooming in on New York City showed substantial economic variation in mobility patterns that county-level data conceals. People leaving residences in mid- and lower Manhattan, where incomes are higher, moved into city neighborhoods (CBGs) with the highest average incomes, while movers from upper Manhattan moved to lower-income destinations. An analysis of public housing residents found a less mobile population, but evidence of upward mobility among those who left.

In California, residents within the Camp Fire perimeter were at least three times more likely to move than those living near the perimeter or in other parts of Butte County. A similar pattern was seen after the 2017 Tubbs Fire in Napa and Sonoma counties. Publicly available data, however, suggested flat rates of outmigration in the affected counties.

"Access to migration data that can tell you what actually happened at this narrow, fine level could help ensure that resources are targeted most effectively," said Fitzpatrick. "I look forward to seeing how researchers use this data to contribute a wide range of new findings."

The researchers acknowledged support from the National Science Foundation, the Canadian Institute for Advanced Research (CIFAR); NASA; Google; Meta; Amazon; Cornell Tech's Urban Tech Hub; the LinkedIn-Cornell Bowers Strategic Partnership; the Survival and Flourishing Fund; and Open Philanthropy.