To examine the association between long-term poverty exposure and census tract-level food insecurity rates across the USA.

Cross-sectional study using nationwide census data. Spatial error models were used to estimate differences in food insecurity rates by long-term poverty exposure, adjusting for rurality–urbanicity.

All census tracts in the USA.

US census tracts with complete data on food insecurity, long-term poverty status and rural–urban classification. Due to historical differences in data availability, sample sizes varied across models.

In the inclusive framework, tracts classified as Persistent Poverty Since 1990 had food insecurity rates 7·2 percentage points higher than other tracts (P < 0·001); tracts classified as Enduring Poverty Since 1980 had rates 7·1 points higher (P < 0·001); and those classified as Enduring Poverty Since 1970 had rates 6·3 points higher (P < 0·001). In the mutually exclusive framework, where each tract was assigned to only one poverty category based on the earliest qualifying decade, food insecurity was 6·3 points higher in tracts with Persistent Poverty only, 4·7 points higher in tracts with Enduring Poverty Since 1980 only, and 6·3 points higher in tracts with Enduring Poverty Since 1970 only (all P < 0·001). All models adjusted for Rural-Urban Commuting Area (RUCA) codes and accounted for residual spatial autocorrelation.

Long-term poverty exposure is strongly associated with elevated food insecurity at the census tract level. These findings underscore the importance of integrating historical measures of disadvantage and the value of tract-level analysis for identifying communities at risk of food insecurity.