Modern planning has a foundational idea: below a certain density, few people walk, bike or take transit. Just about any city-building goal, from supporting main streets to running an effective transit system, requires a certain minimum number of people living within walking distance.
Our research suggests another indicator can do even better job for identifying thresholds after which almost no one walks, bikes or takes transit: road length per resident.
What decides whether someone walks to a corner store isn’t how quickly a crow could get there, but the length of road from one’s front door to the food aisle. Density measures how tightly packed everything is on a 2-dimensional map. Road length per person, in contrast, gets closer to the average distance on the street between each person and anything they might want to get to.
Comparing The Two Indicators
Here’s a table showing what proportion of people got to work everyday by driving for neighbourhoods (census tracts) in nine Canadian cities, Victoria, Calgary, Edmonton, Winnipeg, Oshawa, Windsor, London, Sherbrooke, and Halifax. The outcome is what we’d expect: as density goes up, driving goes down.
The relationship, however, is hardly tight. There are places with very high density where people still rely almost exclusively on driving (top-right), and low-density places where a remarkable number of people find other ways of getting around (bottom-left). The correlation is not strong.
Compare that to road length per resident.
At around 7.5 meters per person, getting to work by any means other than driving collapses completely.
And a similar story can be seen for transit. Most of the high-ridership census tracts are below 7.5 meters per person.
The picture is even clearer for active transportation.
If you want to make sure nobody walks or bikes to work, all you have to do is ensure there is at least 7.5 meters of road between everyone.
So what is this 7.5 meter cliff?
The division can be understood as the difference between medium density and lowest-density suburbs. Here’s an example of a 12+ meter per person suburb in Halifax known as Kingswood.
The indicator could be useful as a tool for helping municipalities prevent the most egregiously inefficient development patterns.
The difference is less clear, however, between more traditional suburbs and downtown, which raises a mystery. Consider three areas of Calgary.
The difference between the downtown (bottom left) and the inner suburbs (top left) is what one would expect. Downtown has a very low road length per person of between 1 and 4.5 meters per person. A bit further north, and the numbers come in at between 4 and 8.
The suburb on the right, however, tells a different story. Marlborough Park has a road efficiency numbers to rival downtown, between 3 and 4 meters per person. Yet, as you can see below, this area doesn’t look like it’d be the most compact of communities. What gives?
Looking at the maps, we realized that there are two ways to achieve a low road length per person. Developers can maintain a tight grid, as in downtown, and put enough people there to support road efficiency. Or, they can abandon the grid, using the fewest possible intersections to minimize the amount of road per household.
As you can see, that can keep maintenance costs down, but it greatly increases the distance residents must walk to get to anything.
We wondered, therefore, if we could “expose the cheaters,” so to speak, by measuring the average number of intersections per kilometre of road for each census tract. If a place has few intersections, pedestrians may have to walk far out of their way to make a short distance, rendering the area unwalkable. Here’s what we found:
For both Calgary and Halifax, 3.8 intersections per kilometre appeared to be another key cutoff. Any fewer intersections than that, and almost no one walked or biked to get to work.
Municipalities should keep this in mind when they design development charges. If they make developers pay for roads but don’t require a minimum number of intersections, they may incentivize neighbourhoods where walking is an unrealistic option.
Not every neighbourhood with less than 7 meters of road per person and more than 3.8 intersections per kilometre had high rates of walking or biking. But almost all the census tracts that supported commuting on foot and bike met one or both of these criteria.
These thresholds suggest useful quantitative minimum standards for municipalities to regulate new development. Reasonable suburban communities of many shapes can be designed with far less than 7 meters of road per person, and that still meet a basic intersection density.
Road length per person, like density, is simple and easy to understand. Measuring every aspect of a neighbourhood will always reveal more, but such broad indices are useful for expressing the basic features of urban form. And the clarity of these thresholds suggests that road length per person is a useful addition to density.