The census data used more than two dozen categories of commuters, including two- through seven-person carpools, ferry, taxi, and so forth. For this analysis, I collapsed these data into four categories. "Auto" includes people who commute by auto or motorcycle alone or in a carpool. In 1990, only 15 percent of auto commuters carpooled and the average number of people per commuter car was less than 1.1.
"Transit" includes all forms of transit: buses, streetcars, ferries, etc. I included taxis and even included the census category called "other" even though it is not strictly a transit category. "Walk/bike" and "worked at home" are self explanatory.
In the table below, auto, transit, and walk/bike are percentages of all commuters and should add up to 100 percent. Work at home is a percent of all workers, and so is not additive to the others.
The first thing to note is that work-at-home percentages are constant at around 2.5 percent regardless of density. Technology may change the number of people who work at home, but density will not.
The next point is that, at densities below about 6,000 people per square mile, changes in density have an insignificant effect on commuting modes. There may be a slight increase in the number of people using transit, but none in the number walking or bicycling.
Above 6,000 people per square mile the results are different. Doubling density from 6,000 to 12,000 per square mile can result in a nearly 25-percent decrease in auto commuting. Another doubling leads to an even larger decrease in auto commuting. Of course, a doubling of density combined with a 25-percent reduction in per capita driving still leads to a 50 percent increase in traffic (2 times 0.75 equals 1.5).
A few cities may distort the results at the highest densities. New York (whose density is 23,700) dominates the 16,000-24,000 density class, and New Jersey suburbs of New York dominate the two highest classes. Chicago and San Francisco dominate the 12,000-16,000 density class.
These cities all have excellent transit systems, which indicates that density alone is not enough. Eliminate Chicago, San Francisco, and Jersey City from the 12,000-16,000 density class and auto commuting is 75 percent while transit is 17 percent.
But the transit systems alone aren't enough either. Berkeley, Oakland, and other cities on the bay east of San Francisco all have population densities of well above 6,000 people per square mile and most are located near the BART line. Yet less than 20 percent of the commuters from these cities use transit; auto driving is in the mid 70s.
What makes San Francisco, New York, and Chicago different is not just the residential density and comprehensive transit coverage but their very high density employment centers downtown. Cities with high densities and excellent transit but decentralized employment do not do as well.
Table One: Commuting Mode and Time by Density
Density Class Method of Commute Work Commute pop./sq. mi. Auto Transit Walk/Bike Home Minutes 32,000+ 57.9% 28.2% 13.9% 1.5% 26.6 24,000-32,000 30.1% 53.4% 16.5% 2.8% 29.9 16,000-24,000 38.1% 51.1% 10.8% 2.3% 35.2 12,000-16,000 63.0% 29.0% 8.0% 2.1% 29.4 8,000-12,000 73.9% 18.1% 8.0% 2.1% 24.9 6,000-8,000 83.7% 10.9% 5.4% 2.5% 24.1 4,000-6,000 89.2% 6.0% 4.8% 2.5% 22.6 3,000-4,000 90.5% 4.9% 4.7% 2.5% 21.4 2,500-3,000 91.3% 4.4% 4.3% 2.3% 21.4 2,000-2,500 92.5% 3.1% 4.4% 2.3% 20.0 1,500-2,000 93.1% 2.4% 4.5% 2.4% 19.2 1,000-1,500 93.5% 2.2% 4.3% 2.5% 19.5 500-1,000 93.8% 1.8% 4.4% 2.5% 19.9 0-500 93.1% 2.2% 4.7% 3.2% 21.1The last column of the table shows the average amount of time it takes for commuters in each density class to make a trip from home to work or work to home. Higher densities lead to longer commutes, but this is partly because the highest density areas are also the largest in area and so are more likely to have people commute further.
The clear conclusion is this: In order to significantly reduce the amount of driving people do, planners will have to do three things:
These do not sound like terrible biases to me (errors from the first could offset one another, people who walk/cycle only in the summer are still mostly driving). But even if the biases are significant, there is no reason to suspect that they are sensitive to density. If the answers were biased towards cars, they would be equally biased in all densities so my conclusions, that transit usage is not sensitive to density below 6,000 people per square mile, would be unchanged.
Bodzin also points out that the census data deals only with commuting. "Mixed-use zoning can probably contribute to (reducing) other (auto) trips, especially shopping and recreation," he says. But trying to increase walking through high-density mixed-use areas has the same problem as trying to increase transit in most cities: Since walking currently makes up such a small proportion of shopping and recreation trips (almost certainly under 2 percent), doubling or tripling it has little effect on auto driving.
Still, some cities record more walking or bicycle trips than mass transit trips. Such cities can probably effect a greater reduction in auto driving by improving bicycle and pedestrian facilities than by building light rail. Such facilities can be built without making travel more hostile to autos.
If there is a shortage of and demand for mixed-use developments, developers should be allowed to build them. But the Congress on New Urbanism's official policies -- that all future developments should be New Urban and that all existing low-density areas should be "reconfigured" to New Urban standards -- are inappropriately intrusive and are not likely to significantly change travel habits.