This post first appeared on the Can We Still Govern? Substack newsletter.
President Trump promised to move federal jobs outside of Washington DC, even though only about 15% of his federal employees are in the broader DC region. He did some of this in his first term, relocating the Bureau of Land Management headquarters to Colorado, and a US Department of Agriculture research office to Kansas City. The result was large scale loss of experienced employees, disruption to government work, and questionable cost savings.
In his second term, Trump is pressing on. The US Forest Service will be relocated to Utah amidst a downsizing of research talent.
The US Department of Agriculture (USDA) recently announced that the Food and Nutrition Service (FNS) would reorganize its 16 nutrition assistance programs into five regional hubs, to be located in Dallas, Denver, Indianapolis, Kansas City, and Raleigh. This represents a relocation of most of its 1,200 employees from its current headquarters in Alexandria, VA and its seven regional offices.
The Trump Administration has framed relocations like this as a way to bring federal services closer to the people they serve. Critics say that it is an effort to induce more resignations, move federal jobs to red states, and closer to conservative stakeholders who can more easily influence federal employees.
In the case of FNS, what is true? Will these new offices be closer to people receiving SNAP benefits? SNAP is the largest of FNS’s 16 nutrition assistance programs and the one most directly tied to low-income household eligibility, making it a reasonable proxy for where FNS services are most needed.
So, let’s take a look at the data.
FNS’s Current and Future Office Locations
The current FNS structure includes seven regional offices plus the headquarters in Arlington, VA, which together are located in counties with approximately 3 million households, about 437,000 of whom receive SNAP benefits—a recipiency rate of 14.2 percent. (Data for Robbinsville, NJ, the seventh current office location, were unavailable, so both figures are slight undercounts. Technical Notes are below that explain the data used here.)1
Under the reorganization, those seven offices consolidate into five new hubs plus headquarters. The cities in that new configuration collectively serve about 1.8 million households, of whom approximately 176,000 receive SNAP—a recipiency rate of 9.7 percent, lower than the current footprint.
Source: American Community Survey; Tom McGovern. Notes: According to the USDA, the Atlanta office will provide retailer and compliance operations.
I also merged these city-level data with 2024 presidential election results, by way of context. Most major cities lean Democratic, so the partisan skew in the table below is not surprising. Worth noting, however, is that all five of the new FNS hub cities are located in states that Donald Trump carried in 2024. Only Denver—and Arlington, VA, if included as a skeleton-crew location—sit in states Kamala Harris won.
What If We Chose Differently?
Obviously, I don’t know how USDA arrived at these five cities. There may be real estate, administrative, or labor market considerations that made them preferable to alternatives.
But the question is worth asking: If the goal is to locate offices closer to SNAP recipients, what would a different set of cities look like?
If we optimized for the sheer number of SNAP-receiving households, we’d end up with the largest cities in the country—New York, Los Angeles, Chicago, Houston, and so on. That’s a somewhat crude measure. More interesting is what happens if we instead choose cities with the highest share of households receiving SNAP. (Alternatively, we could choose more rural areas that have high shares of SNAP recipients, but those areas tend to have smaller populations, so cities remain the more useful comparison).
Source: American Community Survey.
The table below shows the 50 largest US cities ranked by SNAP recipiency rate. The top five—Detroit, Fresno, Philadelphia, Milwaukee, and Cleveland—together serve about 1.6 million households, roughly 430,000 of whom receive SNAP benefits, a recipiency rate of 27.6 percent.
The five new FNS hub cities, highlighted in orange in the table, all rank in the bottom 20 cities by this measure. Even two of the current office cities—Boston and Chicago—reach more SNAP recipients, and at higher rates, than most of the new hubs.
Source: American Community Survey; Tom McGovern.
What the Data Say
The stated rationale for relocating FNS regional offices is to bring services closer to the people they serve. But the data tell a more complicated story. The five new hub cities—Indianapolis, Dallas, Denver, Kansas City, and Raleigh— collectively reach fewer households and a lower share of SNAP recipients than the offices they replace. Meanwhile, cities like Detroit, Fresno, Philadelphia, Milwaukee, and Cleveland—which rank among the highest in SNAP recipiency—are absent from the new map entirely.
None of this is to say the new locations are wrong. Real estate availability, workforce considerations, and operational logistics all matter in decisions like this. And proximity of an office to recipients is only one way to measure how well a federal program serves its constituents.
But if the goal is to be close to the people who rely most on nutrition assistance, the data suggest the new configuration is a step in the wrong direction. The cities with the highest need are largely not the cities getting the offices. The results reinforce suspicions that the relocation is for political reasons, fitting a pattern where Trump has shut off federal resources from blue states. Indeed, 30 of the 50 largest cities shown above are in states Trump carried in 2024, with 8 of the remaining 20 located in California. And already, three-quarters of USDA workers have said they won’t relocate, which is consistent with previous agency moves and will likely have consequences for the quality of processes and services the agency offers.
Technical Notes
SNAP recipiency data come from the Census Bureau’s American Community Survey (ACS), Table B22003, using 2024 one-year estimates, imported into Stata using the getcensus package. The figures reflect the share of households receiving SNAP benefits; SNAP eligibility and benefits are technically determined at the household level, so this is the appropriate unit of analysis. Margins of error are not shown.
Voting data come from Tom McGovern’s Github repo of county-level presidential returns for 2024. Because voting data are reported at the county level rather than the city level, I assign each city to its primary county. This works cleanly for some cities—New York City, for example, maps precisely to five counties (New York, Kings, Queens, Bronx, and Richmond). For other cities, the county is a rougher approximation: King County, WA includes not just Seattle but also Bellevue, Redmond, Renton, and other cities. Where two major cities share a county—Los Angeles and Long Beach both fall in Los Angeles County—I include only the larger city.
I focus on SNAP here. FNS also administers 15 other nutrition assistance programs, including WIC, the National School Lunch Program, and the School Breakfast Program, which are not reflected in this analysis.
The city/county metric is an imperfect measure of geographic closeness. A PUMA-level analysis, or one that calculates the number of SNAP recipients within a some distance or time of each office, would be more precise—and is something I’ve done with Social Security Administration offices in other work. For now, the city-level data are sufficient to get a general sense of the landscape.
Download this Stata file to replicate the results.
