Fun with Density and Transit Statistics

Conventional wisdom holds that the U.S. is too spread out for workable mass transportation except in a few high-density cities. Urban planning expert Anthony Downs offers this explanation:

But in 2000, at least two thirds of all residents of U.S. urbanized areas lived in settlements with densities of under 4,000 persons per square mile. Those densities are too low for public transit to be effective. Hence their residents are compelled to rely on private vehicles for almost all of their travel, including trips during peak hours.

Traffic: Why It’s Getting Worse, What Government Can Do

Is that accurate? It all depends on statistics and assumptions, both of which are endlessly susceptible to manipulation. For instance, Downs uses “settlements” as the geographic unit of analysis and calculates how many Americans live below a certain density threshold.

That’s one approach, but there are many different approaches. We could use other geographic units and find out the average density in each of them.

Geographic Unit Population Population
% of U.S.
Land Area
(sq mi)
Land Area
% of U.S.
United States 281,421,906 100 3,537,438 100
Metropolitan Areas 225,981,679 80.3 705,790 20
Urbanized Area 192,323,824 68.3 72,022 2

Source: U.S. Census 2000

Looking at the entire United States, what impresses is the vastness and diversity of the land. It’s the America of spacious skies, amber waves of grain, and purple mountain majesties; of trackless tundra, sizzling deserts and rocky badlands.

But look at the urbanized area of the nation and a different picture emerges. Sixty-eight percent of Americans live on just 2 percent of the U.S. land area. Seen in this light, the large majority of the U.S. population is remarkably concentrated on the land. What implication does this have for population density?

Geographic Unit Gross Population Density
(per sq mi)
Gross Housing Density
(per sq mi)
United States 79.6 32.8
Metropolitan Areas 320.2 128.7
Urbanized Area 2,670.4 1,073.1

Source: U.S. Census 2000

Sixty-eight percent of Americans live at an average density of 2,670 persons per square mile. Figuring out what transportation options work at that density is straightforward, but it requires understanding a little bit of jargon.

The census bureau density numbers listed in the tables above are based on gross density. The comment by Anthony Downs is also based on gross density. However, land and transit planners frequently use a different measure called net density.

There’s a significant difference between the two. Gross density is residential density on all land, while net density is residential density on residential land only. That can make a big difference because of all the non-residential land uses that figure into the calculations, like streets, parking lots, commercial and civic buildings, industrial plants, parks, wetlands, etc.

According to the Institute of Transportation Engineers (1989), the average density of U.S. urbanized areas is slightly below the threshold for local bus service.

A minimum level of local bus service (20 daily bus trips in each direction or one bus per hour) is often provided in residential areas averaging 4 to 5 dwelling units per acre. Typically, these residential densities correspond to gross population densities of 3,000 to 4,000 people per square mile.

A Toolbox for Alleviating Traffic Congestion, p. 93

The average density of U.S. urbanized areas is well above the threshold for commuter rail service, but commuter rail also needs a huge job center as a destination.

Commuter rail service, with its high speed, relatively infrequent service (based on a printed schedule rather than regular headways) and greater station spacing is suitable for low density residential areas — 1 to 2 dwelling units per acre. However, the volumes required are only likely in corridors leading to non-residential concentrations of 100 million square feet or more, found only in the nation’s largest cities.

Toolbox, p. 93

Even more suitable for this level of density is paratransit (any type of service that does not use fixed routes). Paratransit includes carpools, vanpools, subscription buses, jitneys, shared-ride taxis and on-demand (route-deviation) services.

For paratransit service modes activity levels and densities lower than the thresholds described above are likely to be more suitable. These modes often depend less on the particular land use pattern found in an area and more on the initiatives of the affected parties. … these modes can be effective, particularly if institutional support is present from large employers with many persons working at one site with identical (and regular) working schedules.

Toolbox, p. 94

Critics of transit are quick to jump on assertions like these, saying that most Americans prefer to drive alone and that transit carries a small percentage of total trips. This brings up the issue of assumptions. The point is not what people are doing today, under current conditions. The point is to explore the capabilities that will be available to Americans under conditions of enduring motor fuel scarcity. Such conditions are likely within the next 10-20 years if not sooner; they are the unavoidable impacts of Peak Oil. Even long-held assumptions can make a sea change when conditions undergo a swift transformation.

High Resolution

Of course, density is not uniform throughout urbanized areas. To see the variations in greater detail, one can use a higher resolution view. To get a higher resolution view, use smaller geographic units of analysis. Many different units are used by commentators and researchers depending on their goals and needs, from the largest (for instance, metro areas) to the smallest (for instance, census blocks).

To see how resolution can affect density, the 1990 statistics for King County, WA make a good case study.

Geographic Unit Number of Geographic Units in County Number of People Living at Density of 6,400 Persons/Square Mile or Higher
King County 1 0
King County Census Tracts 326 40,000
King County Census Blocks 20,992 400,000

Sources: King County, U.S. Census, Hess et al.

The table above shows three ways of looking at the same population in the same land area. At the county level of analysis, the average density is below 6,400 persons per square mile. But when the entire county is viewed with a finer resolution — the county’s 326 census tracts — then it appears that 40,000 people live above 6,400 per square mile density. Viewed with an even finer resolution — the county’s 20,992 census blocks — it appears that 400,000 people live above 6,400 per square mile density. A simple change in resolution alters the density by an order of magnitude.

The density measurement of 6,400 persons per square mile is equal to 10 persons per acre gross. That density can be suitable for an intermediate level of public transit service (1 bus every half hour) under present-day conditions and assumptions. Hess et al. (2001) comment:

Furthermore, these higher density blocks are not only near Seattle, the largest and oldest central city in the county, but are spread throughout the urbanized part of the county. This example demonstrates that both the total number and the spatial distribution of people living at certain densities are highly sensitive to the size and boundaries of the spatial unit of data and analysis.

Measuring Land Use Patterns for Transportation Research


Critics of smart growth and transit love to cite the example of Los Angeles as being the most dense urbanized area in the United States. Imagine auto-oriented L.A. having the highest density — that certainly turns every preconceived notion of density on its head! Or does it?

Actually, it’s just more fun with statistics. Simply by changing the boundaries, several different locations can qualify for “the densest place in America.” For example:

  • The densest CMSA is New York – Northern New Jersey – Long Island at 2,029 persons per square mile. The Los Angeles – Riverside – Orange County CMSA is 482 persons per square mile.
  • The densest PMSA is Jersey City, NJ at 13,043 persons per square mile. The Los Angeles – Long Beach PMSA is 2,344 persons per square mile.
  • The densest county is New York County at 66,940 persons per square mile. Los Angeles County is 2,344 persons per square mile.
  • The densest city is Union City, NJ at 52,978 persons per square mile. The city of Los Angeles is 7,877 persons per square mile.

The study Is Los Angeles more crowded than New York? presents an excellent discussion of and rebuttal to the idea that Los Angeles is the most dense city in the U.S.

Note: CMSA = Consolidated Metropolitan Statistical Area. PMSA = Primary Metropolitan Statistical Area. See glossary below.

Electric Vehicles

Lower density suburbs do have one specific advantage in the race for transportation efficiency. They have adequate space for solar panels and wind turbines that can charge Neighborhood Electric Vehicles (NEVs) and electric motorcycles, electric scooters and electric bicycles.

NEVs are electric cars that are limited to a maximum speed of 25 m.p.h. Being speed-limited allows NEVs to bypass expensive federal regulations, and increases efficiency, range, and safety while also lowering cost. In many states NEVs are legal to drive on streets with a speed limit of 35 m.p.h. or less. Until recently, they looked and functioned like street-ready golf carts, but new models like those from Zenn Cars and Miles Automotive are more like conventional cars with standard weather protection, comfort and ease-of-use features.

NEVs and other low speed vehicles can travel legally on most local streets. To travel on larger suburban arterials, though, they need slow-speed 35 m.p.h. lanes. It’s relatively cheap and quick to designate and convert lanes. One town that is already taking action is Lincoln, CA:

Unlike golf carts, NEVs meet federal and state requirements to drive on residential streets in more than 40 states. Lincoln, which is in danger of violating federal antismog rules, is installing special lanes for them on major roads that have speed limits exceeding 35 mph. The city is also requiring shopping centers to install charging stations for the vehicles.

Thrifty Electric Car Turns On One Town, WSJ, 26 February 2007


Suburbs will be more efficient and resilient if they reconfigure themselves to be walkable, with residences in close proximity to a mix of land uses that provide for everyday needs and activities. But the work of retrofitting will take time and resources, not to mention firm political leadership on the local scale and broad political support. Until then, there are a number of ways that suburban neighborhoods in urbanized areas can meet their basic transportation needs in the near-term, post-Peak Oil era.


Metropolitan Area — Metropolitan areas are made of counties. They are the counties that contain at least one city of 50,000 population (or one urbanized area of 100,000 population) and the adjacent counties that are within commuting distance. Metropolitan areas can include large areas of rural or undeveloped land.

CMSA — Consolidated Metropolitan Statistical Area. An area that qualifies as an metropolitan area and also has a population of one million or more.

PMSA — Primary Metropolitan Statistical Area. A subarea within a CMSA that consists of a large urbanized county or a cluster of counties (or cities and towns in New England) that demonstrate strong internal economic and social links in addition to close ties with the central core of the larger area.

Urbanized Area — Densely settled areas containing at least 50,000 people, and in other places with a population of 2,500 or more. Urbanized areas are made of census blocks that have a minimum population density of 1000 per square mile, plus surrounding census blocks that have a minimum density of 500 per square mile.

Census Tract — Census tracts are small, relatively permanent geographic entities within counties (or the statistical equivalents of counties). Generally, census tracts have between 2,500 and 8,000 residents and boundaries that follow visible features.

Census Block — Census blocks are the smallest geographic areas for which the Census collects decennial census data. Census blocks have a minimum size of 0.7-0.9 acre, and are delineated by physical features such as streets, railroads, water, and legal boundaries.


Fialka, John J., Thrifty Electric Car Turns On One Town. Wall Street Journal, February 26, 2007

Hess, Paul M., Anne V. Moudon and Miles G. Logsdon, “Measuring Land Use Patterns for Transportation Research.” Transportation Research Record 1780, 2001, pp. 17-24

Institute of Transportation Engineers, A Toolbox for Alleviating Traffic Congestion, 1989

King County GIS Center, Census Data Catalog

LincolNEV — Information about the Neighborhood Electric Vehicle plan and program in Lincoln, CA

NEV Portal — Directory of NEV manufacturers

Flaherty, Sandra, Andrea Osgood and Lara Regus, Is Los Angeles more crowded than New York? Using GIS to compare population density in Los Angeles & New York. Prepared for Livable Places, Los Angeles, 2006. See also related materials

Sierra Club — Articles and research about sprawl and transportation

U.S. Census, Census Bureau Map Products

U.S. Census American Fact Finder, Decennial Census Tables

U.S. Census, Geographic Areas Reference Manual and the follow up Census Geographic Changes for Census 2000 + Glossary

U.S. Census 2000, Density Tables (Density Using Land Area For States, Counties, Metropolitan Areas, and Places)

USGS National Atlas, Map of Urbanized Areas

8 responses to “Fun with Density and Transit Statistics

  1. Moshe Braner

    It should also be noted that density needed to support bus service is based on assumptions on how far people will walk to a bus stop. If they’d walk (or bike) twice as far, then the number of people within walking distance of a bus route is doubled. The walking distance will likely increase along with fuel prices. Even more importantly, behind the standard numbers there are assumptions about the percentage of people who will ride the bus if it were available. When fuel prices get high enough, we’ll find that existing transit capacity is overwhelmed. Since getting additional busses will be a bottleneck, and also for the sake of more flexible transit, I expect that the future of public transit is the jitney.

  2. Laurence Aurbach Post author

    I think you’re right about that. Just a few hundred miles from the U.S. are several nations whose rural populations use buses for most of their transportation. They’ll commonly walk a mile or so to catch a bus.

    Even in the U.S. the quarter-mile rule of thumb is too short in some contexts. Recent research by Schlossberg et al. found the mean distance pedestrians will walk to a light rail station is half a mile, especially if the route is convenient, safe and pleasant.

  3. Dan Zack

    Brilliant. Thank you so much for dispelling these myths. The misuse of density statistics has bugged me for quite some time, and I’m glad you shed light on the subject in such an eloquent and thorough way!

  4. karthik

    the density calculations and its variations have left me even more confused. if one were looking for a basis for comparing densities of two urban areas, what would one use to arrive at a realistic estimate?
    realistic meaning — figures which one could use to work out transportation loads, plan neighbourhoods with appropriate residential and other typologies?
    what works?

  5. Laurence Aurbach Post author

    That’s an interesting question. Here’s how transportation planning often works in the U.S.

    Every large urbanized area is required to have a Metropolitan Planning Organization (MPO) that carries out transportation planning, forecasting, and allocation of federal funds. MPOs use geographic units for transportation planning that are called traffic analysis zones (TAZ). TAZs are drawn using streets and roads as boundaries. They are drawn to have roughly equal populations, so in dense areas they will be small (such as a few blocks) and in rural areas they will be large (such as thousands of acres).

    One very useful feature of TAZs is that they include both the residential and work populations. You can see how this leads to some unexpected results in The Density of Traditional Urbanism.

    Roadway planning is relatively uncomplicated. A certain number of residents and jobs are forecast for a TAZ; these generate a certain number of trips; and therefore a certain amount of roadway capacity is required. That’s still the conventional theory, although it has been seriously debated in recent years.

    Transit planning is much less standardized, because transit modes are more diverse and transit ridership is more complex to model. Unlike roadway planning, transit planning has to consider things like the quality of walking routes, feelings of safety, and other hard to quantify issues. A good starting point for learning more might be the report Fixed-Route Transit Ridership Forecasting and Service Planning Methods.

  6. Mark

    Any idea how transit planning is done in Australia – specifically, in the states of Queensland and New South Wales? Especially interested in density estimates for bus service. Thank you!

  7. Drake

    Anthony Downs of the Brookings Institute cites Pushkarev’s and Zupan’s Public Transportation and Land use Policy in his assertions, stating this:

    Past studies, including one published in 1977 by Boris S. Pushkarev and Jeffery M. Zupan, have shown that public transit works best where gross residential densities are above 4,200 persons per square mile; relatively dense housing is clustered close to transit stations or stops; and large numbers of jobs are concentrated in relatively compact business districts.

    I actually have the Public Transportation and Land use Policy right in front of me. And in several sections, Pushkarev and Zupan explicitly stress that gross density is a “very poor estimator of residential density”, and that they don’t use it in their comparisons of cities.

    So that pretty much destroys a huge chunk of Mr. (Dr.?) Down’s argument right there.

Leave a Reply

Your email address will not be published. Required fields are marked *