The Power of Intersection Density

Intersection density is the number of intersections in an area. It corresponds closely to block size — the greater the intersection density, the smaller the blocks. Small blocks make a neighborhood walkable. This diagram shows three street layouts — extremely walkable, moderately walkable, and unwalkable — with their counts of intersections per square mile:

Intersection density makes surprising news in a study by the formidable academic duo of Reid Ewing and Robert Cervero. They have published Travel and the Built Environment: A Meta-Analysis in the Summer 2010 issue of the Journal of the American Planning Association.

As the title notes, the study is a meta-analysis: a study of 50 other studies about travel and the built environment. The authors look at the results from each of the 50 studies, and then pool all of those results into ten built environment measurements, including intersection density.

Their findings? Of all the built environment measurements, intersection density has the largest effect on walking — more than population density, distance to a store, distance to a transit stop, or jobs within one mile. Intersection density also has large effects on transit use and the amount of driving. The authors comment,

This is surprising, given the emphasis in the qualitative literature on density and diversity, and the relatively limited attention paid to design.

In other words, intersection density is the most important factor for walking and one of the most important factors for increasing transit use and reducing miles driven, but gets relatively little attention in research and in public policy.

The authors report their built environment measures in terms of elasticity. Elasticity can be defined this way: when a built environment measurement changes by a certain percentage, that will cause walking, transit use, or driving measurements to change by a certain percentage. The ratio between the two is the elasticity.

For example, in “Travel and the Built Environment,” the elasticity of intersection density was found to be 0.39 for walking. That means if intersection density is increased 10 percent, walking will increase 3.9 percent. If intersection density is doubled (100 percent increase), walking will increase 39 percent. It was the biggest elasticity found for walking, and also the biggest elasticity found in the entire meta-analysis.

Also surprising was the effect of 4-way intersections on transit use. The number of 4-way intersections in an area can represent the level of street connectivity. But a street layout may provide 4-way intersections along with extremely large, unwalkable blocks. In that case, it may be easier to bike or take a short drive to the transit stop. And in fact, the study did find that 4-way intersections were much more significant for transit use than for walking. The elasticity of 4-way intersections for transit use was 0.29, which was the biggest elasticity for transit use (in a tie with the “distance to the nearest transit stop” measurement).

The study contains many caveats. It is actually a brief primer on the pitfalls and potential disadvantages of meta-analysis. The authors warn that:

  • The sample size (number of studies examined) is small, so the results are only ballpark estimates.
  • Owing to the nature of the studies examined, the level of statistical confidence could not be calculated.
  • Some of the examined studies account for self selection* and some do not; the meta-analysis combines them.
  • Due to all of the above, users should use caution when applying specific elasticities.

* Self selection: Do walkable places cause people to walk, or do people who like to walk chose to live in walkable places? Most nonacademics assume that both are happening, if they bother to think about it at all. But it’s important to academics to pin down these kinds of things.

In addition, intersection density and number of 4-way intersections are not necessarily the most accurate or rigorous measures of walkability and street connectivity. These measurements do not account for intersections that lead to dead ends, bottlenecks in the street layout, or inaccessible gated areas. There are better measurements available, but the intersection density and number of 4-way intersections measurements are relatively easy to compute and can use free databases that have broad geographic coverage. They are favored by academic researchers for those reasons.

Nevertheless, this meta-analysis does a huge service by providing ballpark estimates of the effects of built environment on travel. The study provides a core database, a “seed,” that can be strengthened as more and better research is produced. And it reports surprising, preliminary results about the very large effects of intersection density and connectivity on increased walking, increased transit use, and reduced amount of driving.

26 responses to “The Power of Intersection Density

  1. Daniel Nairn

    I really appreciate this interesting summary of the paper. I hadn’t noticed it yet, but now I’ll certainly want to take a look. I personally like how Virginia DOT has decided to measure connectivity, by a link-to-node ratio. Links are segments of roadway, and nodes are intersections or dead-ends. I would think this would be pretty easy to measure with standard GIS tools. VDOT still had to supplement this measurement tool, because it’s possible to have islands of really great connectivity without external connections to the rest of the system. So any new subdivision must have not only internal connectivity, but a certain number of external connections as well.

    The whole self selection problem may be interesting, but I’m not sure if practitioners need to worry much about it. Whether the desired goals are met through changing behavior or attracting a certain set of people doesn’t much matter as long as the goals are met. I do see that from a larger scale it does matter, if we assume that there is only a certain number of, say, walkers out there. Pulling them to some place is pulling them away from another.

  2. Laurence Aurbach Post author

    Daniel: The Virginia DOT regulations for connectivity are trailblazing, representing a big step forward. Every state DOT should study Virginia’s example and the outcomes of its policy. The link-to-node ratio has also been used in several adopted municipal plans, as described in the APA booklet Planning for Street Connectivity.

    That being said, the link-to-node ratio is not one of my favorites. It has all the disadvantages of intersection density that I mentioned above, plus it has no relationship to block size. So a street layout can have a high link-to-node ratio, but also very large, unwalkable blocks. That’s why the Virginia regs have to include an extra clause that says, “The block layout and other features of the development are designed in such a fashion as to provide reasonably direct pedestrian movement throughout the development and to adjoining property.” But there are no clear, transparent standards given for determining if pedestrian routes are reasonably direct.

    As for the self-selection issue, I agree with you. I wrote about self-selection in Connectivity Part 4 and Connectivity Part 6. Self-selection may or may not play a role in walking behavior; either way, the evidence indicates that policy should support the creation of compact, walkable neighborhoods.

  3. Kaid @ NRDC

    Thanks for writing on this critical analysis. I too am impressed by the findings on intersection density and believe that the attention given the subject in policy dialogue is disproportionately small given its impact on travel and associated emissions.

    That said, it should not be overlooked that the study’s most important (if not its most surprising) finding is that, among the variables, regional/destination accessibility (including proximity to downtown) was by far the most important factor in influencing vehicle miles traveled. As the authors state, “Almost any development in a central location is likely to generate less automobile travel than the best-designed, compact, mixed-use development in a remote location.”

  4. Laurence Aurbach Post author

    Kaid, the study incorporates many data points and covers numerous land use issues. Choosing which finding is most important depends on one’s perspective and priorities.

    If one is focused on VMT, the findings about location will be the most important. However, it is not entirely new information; it confirms the research conclusions that have been rolling in for more than a decade from Cervero and other investigators.

    I have highlighted intersection density because (1) the findings are new, and as the authors note, surprising; and (2) the elasticity of intersection density for walking is large — almost twice as large as the elasticity of destination accessibility for VMT. If one is concerned primarily with walking and transit use, then the findings about intersection density will be most important.

    And if one is a scholarly researcher, the most important aspect of the study is that it establishes a database of studies that can serve as a seed kernel for more extensive and rigorous research.

  5. Dave Murphy

    I’d be interested in seeing if there are any correlations between intersection density and crime, or intersection density and traffic fatalities.

  6. David E

    I wonder whether the intersections need to be with multi-use streets, i.e., ones that allow foot, car, and transit traffic. European cities tend to have a lot of pedestrian-only throughways, such as stairs and blocked-off streets, whereas American cities tend to focus on multi-use or vehicle. If not, building more pedestrian-only streets might be a relatively cheap way to increase intersection density, if that’s a goal.

  7. Laurence Aurbach Post author

    Dave: That’s a softball… For a discussion of connectivity and crime, see Connectivity Part 5. For a discussion of connectivity and traffic crashes, see Connectivity Part 7 and also Two Connectivity Studies for 2008.

    David E.: European cities also have more mixed use, more pedestrian-oriented building frontages, and more travel options. In many European cities, walking, biking and/or transit dominate the transportation system while private autos have a minority share of all trips.

    Installing pedestrian-only streets in American contexts may or may not be a wise option. In the absence of good urban design and convenient transportation options, pedestrian-only streets run the risk of being underused, unprofitable for businesses, or prone to crime. In suburban contexts, pedestrian-only connections perform better than conventional suburbs, but not as well as well-connected urban layouts. The Radburn idea (layouts that are well connectected for pedestrians, poorly connected for vehicles) has been around for 80 years, and after all that time it’s become apparent it has important tradeoffs, both pro and con.

  8. Mike Zimney

    Are there any thresholds for intersection density that would equate to great, good, poor?

    LEED ND requires 140 intersection per square mile.

    Just wondering if there is any consensus for minimum thresholds for intersection density.


  9. Laurence Aurbach Post author

    Mike, the LEED-ND standard of 140 intersections per square mile is an option that projects can use to meet two separate prerequisites. But LEED prerequisites only allow projects to participate; they don’t count toward the project’s rating. The project’s rating is based on the number of credits it earns. So a good place to look for higher performance standards is in the LEED credits.

    The LEED-ND “Street Network” credit gives one point for 300-400 intersections per sq mi, and 2 points for more than 400 intersections per sq mi.

    Another set of thresholds appears in my TND Design Rating Standards:

    Intersections per square mile

    Five stars: More than 330
    Four stars: 290 – 330
    Three stars: 250 – 290
    Two stars: 80 – 250
    One star: Less than 80

    However, it’s important to keep in mind differences in calculation methodology. LEED-ND does not count intersections leading to cul-de-sacs, and does not count alley-alley intersections. LEED-ND does count alley-street intersections, street-rail track intersections, and street-trail intersections (the latter up to 20% of the total).

    My rating system actually subtracts cul-de-sacs from the total. It counts all street intersections and alley intersections. It doesn’t count rail tracks or trails.

    Anyhow, you can see that a place like Venice, Italy is off the charts compared to these thresholds. The TND Design Rating System is calibrated so that 5 stars is equivalent to a well-designed, early twentieth-century, urban neighborhood in the United States. The reasons for that are discussed in the FAQ on page 18.

  10. Mike Zimney

    Thanks Laurence. Do you know if there is a rating scale for percentage of 4-way intersection in terms of great, good, bad?

    I reviewed your TND standards and didn’t see one rating the type of intersection in terms of # per square mile or percentage.


  11. Laurence Aurbach Post author

    Mike, the TND Design Ratings Standards does not differentiate types of intersections, and I am not aware of any rating system that does. One main reason is there’s no evidence that 4-way intersections are universally superior to 3-way intersections. In fact, transporation planning theory and experience says that 3-way intersections have fewer opportunities for crashes (assuming all other factors are equal), and 3-way intersections can be effective traffic calming devices. Many “irregular” or “medieval” street layouts in Europe and elsewhere feature a high proportion of 3-way intersections, while also being extremely walkable and livable.

    And how to account for a layout like Washington DC, where the many diagonal avenues result in 6-way intersections? And traffic roundabouts and traffic circles with 7, 8 or more intersection points? Are these superior to 4-way intersections?

    Maybe, maybe not. It all depends on context and design.

    By the way, Space Sytax measurements like “angular choice” can be fine tools for evaluating the accessibility of irregular street networks. They can account for nuances like bottlenecks in the network, or the angle between streets that run into an intersection.

  12. Pingback: Streetsblog New York City » Walk Score Goes Multimodal With the Addition of Transit Score

  13. Yuri Artibise

    An interesting post on one of the most influential aspects of walkability. One technique to increase intersection density in car centric cities is to take advantage of alleys (especially in older neighborhoods).

  14. MW Brown

    Neat article – I’m looking forward to digging into the links provided. Didn’t Jane Jacobs discuss this issue in “The Death & Life of Great American Cities”? – specifically, one of four criteria for “exuberant diversity” in city streets: “Most blocks must be short; that is, streets and opportunities to turn corners must be frequent.”

    And, “…frequent streets and short blocks are valuable because of the fabric of intricate cross-use that they permit among the users of a city neighborhood.”

    Its good this study is putting some data analysis to Jane’s observations.

    Good stuff.

  15. Pingback: Intersection Density in the GTA « Visualizing Urban Futures

  16. Alex Glueck

    Does anyone know of a website or other kind of reference where you can calculate the intersection density of an area? As this is a pilot credit for LEED v4 it would be a very useful tool

  17. Laurence Aurbach Post author

    Alex, intersection density is not only in a pilot credit, it’s in LEED for Homes v4. And it’s been an integral part of LEED-ND for several years. Here’s the LEED for Homes credit. It gives credit for 90 intersections per square mile within 1/4 mile of the project boundary.

    Manually counting that is usually not too difficult. Also, because the credit has unique rules about what to count, the manual method may be easier than working with a GIS.

    Intersection density appears in several parts of LEED-ND. One thing to be aware of is the counting method may be different in different systems. For example, in LEED for Homes, intersections leading to cul-de-sacs are not counted. In LEED-ND 2009, intersections within pods that are reached through only one entry/exit point are not counted. The latter method often will disqualify more intersections.

    I believe LEED-ND has a Connections Tool that allows projects to count their intersections. I don’t know any details about that or who can use it. ArcGIS plugins are available and some provide free demos, like CommunityViz.

    The Street Smart Walk Score provides a quick and rough intersection density calculation for any address, but the analysis boundary is made by the software.

    You can use the drop-down menu in the H+T Index to display intersection density for any location. The density is displayed in 5 ranges and the map data boundaries are preset.

  18. salma

    I have a question when computing intersection density do you consider all the streetS, i mean those for cars and the ones that separate houses and buildings or only motorized streets. Also how do you count the number of intersection in a roundabout? Do you consider it as one intersection or depending on the number of streets that intersect with the cercle?

  19. Laurence Aurbach Post author

    salma, different rating systems have different methods. For example, LEED for Neighborhood Development includes all streets and trails, motorized and nonmotorized, that are open to the public.

    The TND Design Rating Standards only includes motorized streets and alleys. The thinking behind that is street layouts that are poorly connected for vehicles tend to concentrate traffic on wide, fast arterials that are dangerous and unhealthy for residents and pedestrians. Also, it is too easy to create pedestrian path networks with tiny “blocks” and extraneous intersections, like a college quadrangle criss-crossed by paths.

    Different systems will also treat roundabouts differently. The TND Design Rating Standards uses a one-acre minimum for the area inside the circle. If the area is less than one acre, it is considered one single intersection no matter how many streets converge there. If the area is greater than one acre, the circle can function as a public space or park, so it is considered to be like any other block. In that case every street that intersects the circle is counted as a separate intersection.

  20. Pingback: At the Crossroads: In Order to Create a More Walkable L.A., Start with the Basics. | Streetsblog Los Angeles

  21. salma

    Laurence Aurbach Thank you very much for your answer I appreciate it.
    I do have another question:
    As for pedestrian network, I computed the intersection density:for that I considered the same intersections for motorized network (as sidewalks) and added the intersections for alleys and shortcuts, of course for that case the pedestrians network will seem to have a better intersection density, I read somewhere that there is a correction coefficient for intersections for pedestrians network, do you have any ideas about that?
    Also what is considered a mimimum and a good intersection density per km2?
    Thank you

  22. Laurence Aurbach Post author

    salma, I’m not familiar with a correction coefficient for pedestrian networks. I don’t know what its purpose would be.

    The TND Design Rating Standards do not include pedestrian paths in the intersection density calculation. That’s because a principle of traditional neighborhood design is that all streets are pedestrian friendly and most pedestrian travel will happen on street sidewalks, not on segregated paths or dead ends. Areas that are not vehicular lanes must be larger than 1 acre (0.4 hectare) to count as a block; otherwise elements like traffic islands and roundabouts are considered part of a single intersection.

    The LEED systems do include pedestrian paths when calculating intersection density. That raises the possibility of tiny “blocks” that increase the intersection density in an unjustifiable way. The spaces between paths on a college quadrangle are an example. I once suggested that LEED use a 1-acre (0.4 hectare) minimum when calculating intersection density. I don’t know what if any methods LEED now uses to address that issue.

Leave a Reply

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