This book was very interesting and worthwhile reading. I was initially worried that a book from 1977 would be hopelessly outdated, particularly given the prevailing trend of road expansion at the period. But this book actually digs quite deeply into the relationships between transportation and land use. This is all the more impressive in a period when data analysis would have been more expensive and difficult than today. They rely largely on data sources from 1970 and earlier, and make an effort to present multivariate analysis in a reasonably intuitive manner.
They start with some very interesting historical analysis, finding that auto-oriented densities tend to be one tenth of historical densities.
The highest urban densities in each period shown fall in the range of 60,000 to 95,000 per square mile of built-up land. This compares to Manhattan's density of 86,000 in 1970 and its peak of about 130,000 around 1910. Historically, people had much less indoor housing space than Manhattanites have today; thus average population densities similar to those of Manhattan could easily exist, even though the density of structures was much lower. One should also recall that in the late nineteenth century the most crowded wards of places such as Prague, Paris, and Manhattan reached the equivalent of 280,000 to 520,000 persons per square mile of built-up land in walk-up tenements. It was this kind of densities that early urban reformers fought against, and that the advent of electric streetcars and rapid transit helped to dissipate.
* Cordon areas of area transportation studies; density within county boundary definition much lower.
Travel demand by urban public transit grew to over 900 miles per capita of the urban population per year by 1922, and thereafter began to decline, dropping to little over 200 miles in the early 1970s. The automobile in the United States began to appear in large numbers (more than a million new cars annually) in 1916, about two decades after it was invented, and before 1922 the amount of urban travel by auto appears already to have surpassed the amount of travel by public transportation. The travel diverted by the automobile from public transportation over the next half century is dwarfed by the amount of new travel which the automobile created. By 1973, urban travel per capita expanded to some 6,500 miles annually, over 10 times what it was in 1900. And total domestic travel per person in the United States exceeded 9,000 miles annually, 20 times what it was in 1900.
Almost 90 percent of this huge amount of travel is by automobile, about 7 percent by air, about 1.7 percent by intercity rail and bus, and about 2 percent by urban public transportation.
Exhibit 0.3 on page 10 is a great graph for understanding how much "10 times more travel" really is. The graph unfortunately exaggerates it by showing "annual passenger-miles" instead of a per capita value; but it's still astonishing how much auto travel dwarfs other modes.
They start out by observing that gross density alone doesn't explain transit use very well. However, the strength of a central business district can be a useful measure. They combine density with the log of office floorspace in the CBD and the presence of rail transit, and find a stronger relationship to transit mode share (p. 26).
They show relationships between trips and density in six U.S. cities in Exhibit 2.4. Interestingly, the total number of linked weekday trips per person drops from about 2.4 at a density of 1 dwelling unit per acre down to 1.3 at a density of 60 dwelling units per acre. I wouldn't be surprised if this is partly caused by their definition of a trip, which excludes walking and cycling. The more interesting part of the graph shows transit trips, and has two distinct sections: from 2 to 7 units per acre, the transit use is low, and then it rises up sharply for higher densities. The authors summarize:
Summarizing, we find that densities in the 2 to 7 dwellings per acre range produced only marginal use of public transportation within major urban areas of the United States in the early 'sixties. We have seen that the average areawide densities of virtually all urbanized areas in the nation fall into this density range. Densities of 7 to 30 dwellings per acre were necessary to sustain significant transit use—in the range of 5 to 40 percent of all trips.
One of their main points here is to show that it makes no sense to think of the average density of an entire urban area, since the densities within that area will vary considerably, and hence the behaviour in different parts of the city will be quite different. They proceed with a discussion of the New York metropolitan area. In Exhibit 2.5, they show a breakdown by mode of trips in New York City, plotted against density. They separate transit trips into two groups: to the CBD, and other transit trips. In the lowest density regions, transit is only used to reach the CBD; as the density rises, local trips are increasingly made by transit, outnumbering trips to the CBD for densities above 7 units/acre.
In Exhibit 2.6, they show a similar modal breakdown, but disaggregate the data further by income group. It's fascinating to see how much less the low income group travels—typically making only 20-60% of the auto trips of the middle income group. Commuter rail is used overwhelmingly by the upper income group, and barely at all by the low income group. Taxis are one of the most popular modes for the upper income groups in the highest density category (200 units/acre). Transit really comes into its own at densities of about 40 per acre, when subway use becomes very popular.
I've reproduced Exhibit 2.6 here. In the exhibit, there are three bars at each density. The left bar is the upper income group (top 25% of households), the middle bar is the middle income group (middle 59% of households) and the right bar is the low income group (bottom 16% of households). I've also plotted the same data for each income group alone.
After this discussion, they move on and make an important distinction about the ways to measure the effects of density:
So far, we have dealt with percentages of people using transit and with transit trips per capita at different densities. However, what matters from the viewpoint of providing transit service is not just trips per person, but the fact that at higher densities there are more persons around to make the trips. [...] The compounding effects of greater density of people and of more trips per person are dramatically evident in the Exhibit. For example, in the New York Region in 1963, trips per square mile by the two rail modes (commuter rail and subway) increase from 12 at a residential density of 0.8 dwellings per acre to an average of 61,452 at a residential density of 200 dwellings per acre. On the log-log graph, the relationship is virtually a straight line.
It can be seen from Exhibit 2.7 that in what one might call the usual range of transit supporting densities—between, say, 7 and 40 dwellings per acre—transit trip ends per square mile range from around 500 to around 10,000, given the service levels of New York and Chicago. Transit trip end density approaches 100,000 per square mile only at residential densities in excess of 200 dwellings per acre in New York City. All these figures represent averages calculated over areas that contain more than ten, and in some cases several hundred, square miles. Trip end densities in particular square miles with heavy concentrations of nonresidential density, notably in Central Business Districts, are of a higher order of magnitude. [...] Trip ends by public transit exceed 100,000 per square mile in the downtowns of Brooklyn, Detroit, Los Angeles and Pittsburgh, and go over 200,000 in the downtowns of Boston, Philadelphia and Chicago. They exceed 600,000 in the central square mile of Midtown Manhattan.
Trip ends by auto are shown below the transit trip ends for comparison; they include the trip destinations of auto drivers, auto passengers, and taxi passengers. It can be seen that the differences in auto travel to these central square miles are much smaller than the differences in transit travel. Generally, auto trip ends do not exceed roughly 160,000 per square mile. Only in Midtown Manhattan does this figure approach 200,000, a result of the saturation of the area with short taxi trips (shown separately), and also a result of heavy street use during other than the usual peak hours.
There appears to be a limit to how many trips by auto a square mile can absorb. This limit is dictated by the character and extent of the pavement available for auto movement and by the extent of the space available for parking. Thus, the peak accumulation of vehicles in a square mile was found not to exceed 23,000 to 27,000 in cities with more than half a million people; Midtown Manhattan's peak accumulation is no different—about 24,000 vehicles per square mile. [...] [A] limitation of auto access into very high density areas does empirically exist, even short of deliberate restraints on environmental grounds. As total destinations per square mile exceed those which can be taken care of by auto, the surplus is handled by public transit. Largely because of this phenomenon, the higher the density of a downtown area, the greater the proportion of trips made by transit. Exhibit 2.8 [reproduced below] demonstrates this in terms of the actual numbers of trips ending in a particular square mile.
I like the above figure. It's intuitively obvious that there's a limit to the number of trips you can pack into an area by automobile, but it's great to see some empirical evidence. In places where regional goals include density (such as Vancouver), it makes it easier to argue against any additional provisions for automobiles.