Next month, the new fiscal year 2024 will be the first year for which most federal agencies and programs will use the 2020 Census’s revised definitions of “urbanized areas” when giving out funding under many grant programs. This means it is a good time to review the changes made in the Census definition of “urban” for the first time in 110 years.

The first federal estimates of urban vs rural population were made following the 1870 Census, when the minimum population threshold for “urban” was set at 8,000 people in an incorporated area. This then dropped to 4,000 people and then popped back to 8,000 people in subsequent Censuses, until a minimum population of 2,500 was used for the 1910 Census and ever since. (The Census stopped using incorporated areas as a requirement for urban areas in 1950).

The new 2020 Census changes the minimum definition of “urban” by getting rid of the flat minimum number and substituting an “either-or.” Instead of just having 2,500 people to qualify for urban status, a place must now have either at least 5,000 people, or a minimum of 2,000 housing units. The Census FAQ document says this has two advantages: “Areas, such as seasonal communities or second-home communities, that are densely developed but have a lower population on Census Day (April 1) can qualify as urban based on the number of housing units,” and “Areas that have higher than average people per housing unit ratios can qualify based on population.”

Since the Census stopped using municipal boundaries in 1950, the question became one of population density – how far apart can 2,500 people (now 5,000 people) live and still be qualified as urban? This question has now been outdated  – instead of population density in a census block, the question is now housing unit density. Per the FAQ, 425 housing units per square mile define the initial urban core, then 200 units per square mile fill in the remainder of the urban area, then 1,275 housing units per square mile ensures each qualifying urban area contains at least one high density nucleus.

Under the new definitions, in the 2020 Census, there were 2,644 urban areas that year across the U.S. and its territories, with populations ranging from 19,426,449 (New York-Jersey City-Newark, NY-NJ) to just 773 persons (Munds Park, AZ). The key to the 14 new urban areas with populations below 2,500 appears to be lots of seasonal housing ­ Munds Park is full of cabins for hunting season, with a total of 2,140 housing units, and the next-least-populous urban area is Fire Island, NY, with a population of 998 but with 3,930 mostly summer homes.

The Census Bureau has prepared two different Excel spreadsheets with which one can download the entire list, along with the 2020 population, number of housing units, land area, water area, and population density of each urban area. It is then a simple matter to add columns for housing density and for average persons per housing unit.

The first spreadsheet shows urban areas across state lines in one consolidated row and can be downloaded here. The second spreadsheet separates the urban areas that cross state lines into separate rows for each state, initially alphabetized by state, and can be downloaded here.

Use in transportation

Mass transit. Generally speaking, federal mass transit programs rely on the Census data but have a 50K population cutoff, per 49 U.S.C. §5302(24): “The term ‘urbanized area’ means an area encompassing a population of not less than 50,000 people that has been defined and designated in the most recent decennial census as an ‘urbanized area’ by the Secretary of Commerce.

Technically, there is now a problem with this. For the 2020 Census, the term “urbanized area” no longer exists – Census has adopted a new term of “urban area” to include what formerly were two different things, urbanized areas and urban clusters. So, strictly speaking, none of the “urbanized areas” that the Federal Transit Administration uses to give out their biggest formula grants exist anymore.

This can probably be lawyered around, but another problem is that the new term is already in use by FTA in 49 U.S.C. §5302(23): “The term ‘urban area’ means an area that includes a municipality or other built-up place that the Secretary, after considering local patterns and trends of urban growth, decides is appropriate for a local public transportation system to serve individuals in the locality.” So, until Congress changes the law, FTA is forced to use specific terms in a very different way than the Census Bureau uses them, which will probably be confusing. This should probably go on the “needed technical corrections” list.

FTA uses Census definitions of urban/urbanized areas much more than do any other USDOT mode. By far the largest annual formula grant program, is the section 5307 “Urbanized area formula grant” program, where funds are apportioned by urbanized area (over 50K in population).

The program has three funding tiers based on population: UZA 1 million and up, UZA 200,000 to 999,999, and UZA 50,000 to 199,999 population, based on the decennial census.

The key is the 200,000 cutoff. For an urban area to add that 200,000th person means that the funding from FTA under the 5307 program is now under local control. UZAs with populations from 50,000 to 199,999 see their formula funding given to the state to use in that area, with the state picking the projects and programs, not the local government.

But there is a downside to getting over the 200K line. Transit providers in urban areas under 200,000 population can use their FTA urban formula grants for operating expenses. Once the 200,000th person moves in and is recorded in a decennial census, it becomes much more difficult for transit providers to use federal funding to pay for operating costs.

Urban areas that were below 200,000 population in the 2010 Census but have moved above the line in the 2020 Census include Clarksville, TN (be sure to take the last train when going there); Santa Barbara, CA; Amarillo, TX; College Station-Bryan, TX; Evansville, IN; Olympia-Lacey, WA; Deltona, FL; Gainesville, FL; Bel Air-Aberdeen, MD; and Fargo, ND. (Huntington, WV shrank in population from 2010 to 2020, but not by enough to drop them below 200,000.)

Highways. The funding formulae don’t take population into account anymore. But the reliance on decennial census population totals, instead of annual Census Bureau estimates, requires analysts to be very wary of relying on data derived from those population totals. Case in point: it took until Highway Statistics 2015 for the Federal Highway Administration to start using the 2010 Census numbers for the urbanized area population in Table HM-72, which gives estimated per capita VMT for urban areas. At some point in the next few years, the Highway Statistics series will switch over to the 2020 Census population totals for urban areas. And this renders the whole per capita part of that series much less useful over time ­ why bother measuring minute year-to-year variations in VMT against a population number that only changes every ten years, and by a much larger margin (usually) than any year or two’s VMT change?

Example: If we are to believe the year-to-year changes in Table HM-72, then the per capita daily VMT in the Las Vegas, NV urban area dropped from 32.5 miles per day in 2014 to 23.2 miles per day in 2015. Which is more likely: that people in Vegas cut their daily driving by one-third overnight, or that this is simply the fact that the denominator in the per capita equation jumped from a population of 1.314 million in the 2014 calculation to 1.886 million in the 2015 calculation as FHWA switched Censuses?

Most competitive highway grant programs use Census totals but have a 200K population cutoff:

• INFRA grants, in 23 U.S.C. §117(i)(3): “the term ‘rural area’ means an area that is outside an urbanized area with a population of over 200,000.”
• RAISE grants, in 49 U.S.C. §6702(1)(7): “the term ‘urbanized area’ means an area with a population of more than 200,000 residents, based on the most recent decennial census.”
• Rural surface transportation grants, in 23 U.S.C. §173(a)(2): “The term ‘rural area’ means an area that is outside an urbanized area with a population of over 200,000.”

Urban areas vs metropolitan statistical areas

Something that even the casual policy wonk needs to understand: the difference between a metropolitan statistical area and an urban area. (Aside from “polis” and “urbs” being the Greek and Latin root words for “city,” respectively.) The two concepts formally diverged in 1950, serve different purposes, and are defined very differently:

At one point, the government tried to use sub-county boundary units to define metropolitan areas, but there was such a variety of good and bad record-keeping within states and federal agencies on such units and their definitions that no standardization was possible, so the metro area boundaries are now county-wide. (Or, in Louisiana, parish-wide. And in Virginia, either independent cities or counties, which also goes for St Louis and Baltimore.) County-equivalent borders are rarely subject to change, and almost all organizations and citizens know where they are.

(But what about New England, you ask, where county governments are either weak or nonexistent (Connecticut), and each state is completely subdivided into towns, which provide all non-judicial government services? Well, the Census also subdivides the six New England states into NECTAs (New England City and Town Areas) based on town boundaries instead of county boundaries. That map is here).

Whereas the boundaries of urban areas are defined by census block, which can be only a few acres in size and hold less than 1,000 people. These boundaries can change drastically from Census to Census depending on where groupings of people have crossed the population or density threshold. But no one is expected to know the number of the census block in which they reside, nor is a federal grant recipient necessarily required to know the specific census blocks in which a federal grant will be spent, which is why OMB sticks to county-equivalent areas for most statistical purposes.

The difference between a metropolitan statistical area and an urbanized area becomes very obvious when you look at Greater Los Angeles. The MSA includes the entirety of two counties: Los Angeles and Orange. But the northern half of Los Angeles County is mostly sparsely inhabited desert (plus two separate urban areas, Palmdale and Santa Clarita), and the eastern half of Orange County transitions to a separate urban area based around Mission Viejo. So the geographic footprint of the Greater L.A. urban area is only one-third the size of the footprint of the Greater L.A. MSA, with the corresponding differences in their population densities.

The difference is easy to see when you look at the nationwide maps, shown below, first of MSA boundaries, and then of urban area boundaries.

(The high-res MSA map can be downloaded here and the high-res UA map can be downloaded here).

Densities

Because of these varying boundaries, it may come as a surprise to some (including this author) that the Los Angeles urban area, famous for its sprawl and auto-dependence, has a 2020 population density of 7,476 people per square mile, highest among the 25 most populous urban areas. The New York City urban area has a population density of only 5,981 people per square mile, 80 percent of L.A.’s.

There are some striking differences between population density and housing density (both in terms of people/units per square mile) amongst the top 25 most populous urban areas in the 2020 Census:

State-level comparisons. Urbanized area analysis has fascinating applications at the state level because there are wide disparities between state urbanization rates (percentage of state population living in an urbanized area) and overall state population density. We are accustomed to looking at small-area, high-density states as being very urban and large-area, low-density states as being very rural. But there are outliers. Two typical small-area, high-density states are Connecticut and Rhode Island, with overall population densities of 744 and 1,040 persons per square mile, respectively. And two of the lowest-density states are Nevada and Utah, with densities of 28 and 40 persons per square mile, respectively.

But these four states have almost identical shares of their population living in urban areas:

At the state level, Census has prepared an Excel spreadsheet showing all 50 state urbanization totals (and DC and PR) for both the 2020 and 2010 Censuses, which can be downloaded here. One inescapable conclusion from that sheet is that the centuries-long trend towards ever-increasing urbanization reversed itself in 2020, with almost all states seeing a slight decline in the share of their population residing in urban areas compared to 2010. Most people assume that this was due to the Census being conducted during COVID, but we will have to wait another decade to be sure. (A county-level spreadsheet is here).

People per housing unit. It is a very simple matter to add a new column to any of these spreadsheets dividing population by housing units in order to get average people per housing unit. Once you sort the list by that variable, two things become clear. The urban areas with the fewest people per housing unit are those like the aforementioned Fire Island, where people own a lot of second homes. (Ocean City, MD, Nags Head, NC, and Big Bear, CA are other such areas.) On the high end, the top reaches of the list are small-population areas with at least one large institution. Here are the ten urban areas in the 2020 Census with the highest persons-per-housing-unit ratios:

Things to remember

Urbanization rates can be very different than population density, so always make sure which you are wanting to talk about.

Urban areas, as defined by the Census Bureau, have borders that change every 10 years depending on where people lived nearer each other in that particular Census, can move as little as a few hundred feet or as much as several miles in any direction every decade, and have fractal-l00king border lines. Metropolitan statistical areas always follow the borders of counties or county-equivalents and do not subdivide counties. Always remember to which you are meaning to refer.