Most of these data are also available for 1993, 1994, and 1995 from the Federal Transit Administration A few of the links on this site do not work for 1996 and 1997 data, so use the previous address for those years.
The analyses discussed in this memo combine 1997 data for all modes and transit agencies in each urban area and compare these with urbanized area data from the 1997 Highway Statistics. Tables HM-71 and HM-72 from this publication contain data about the population, land area, miles of roads, vehicle-miles traveled, and other road information.
The memo also compares transit data with 1990 census data showing the percentage of commuters using autos and transit in each urbanized area as described in Some Numbers on Commuting and Density.
The Census Bureau lookup page can be used to obtain information by urbanized area, but a separate request must be made for each state. If you use this page to look up data for state, county, place, metropolitan statistical area, or urbanized area, ask for P001 (population), P049 (mode of travel to work), and P051 (aggregate commute time).
Note that "metropolitan statistical area," which includes all the land in all counties surrounding each urban area, can be a deceptive land unit when measuring population densities because large portions of many MSAs are rural. The MSA for San Bernardino, for example, includes the entire Mojave Desert extending to the Nevada border. "Urbanized area" works best when dealing with densities.
These data provide comparisons of land areas, densities, road mileages, freeway mileages, freeway lane mileages, daily vehicle-miles traveled, percent of commuters using autos, percent using transit, and various measures of transit service for 235 major urbanized areas. A statistical test known as R-squared can be applied to see if any of these data are correlated. R-squared produces a number between 0 and 1, and a high number suggests that two sets of data are closely related while a low number suggests that they are not.
For example, daily vehicle miles of travel is strongly correlated with the number of freeway lane miles (0.950), road miles (0.946), population (0.933), and total miles of freeway (0.877). It is not closely correlated with the percentage of people who commute by auto (0.184) or density (0.181). In other words, as cities get bigger, no matter how dense they are, total miles driven tend to increase directly proportional to population.
Table One: Correlations Between Daily VMT
% auto commuters 0.184 Freeway miles/capita 0.029 Road miles 0.946 Population 0.933 Freeway miles 0.877 Freeway lane miles 0.950Density has little correlation with any other data. The highest is transit trips per capita (0.222); the R-squareds are less than 0.2 for such things as auto commuters, transit commuters, freeway or road miles, or per capita vehicle-miles traveled.
Table Two: Correlations Between Density
Transit trips/capita 0.222 % auto commuters 0.184 % transit commuters 0.191 Daily VMT 0.181 Pass.mi./capita 0.190 Freeway miles 0.123 Road miles 0.146Urban areas with the densest transit service (revenue miles of service per square mile of land) tend to have the highest number of transit trips per capita (0.738) and transit passenger miles per capita (0.783). There is also a strong correlation between density of service and the percentage of commuters who use transit (0.621). Other measures of transit service, such as revenue miles per capita or revenue hours per capita, result in lower correlations with transit usage.
Table Three: Correlations Between Revenue Miles/Square Mile
Transit trips/capita 0.738 Transit pass.mi/capita 0.783 % transit commuters 0.621 % auto commuters 0.445This suggests that one of the better ways of boosting transit ridership is not to increase population densities but to increase the density of transit service. This could mean increasing the number of transit routes or, more likely, it could mean increasing the frequency of service on existing routes.
Table Four: Correlations Between % Transit commuters
% auto commuters 0.622 Freeway miles 0.556 Road miles 0.534 Freeway lane miles 0.513 DVMT/capita 0.014 Population 0.567 Density 0.191New rail services are often an improvement over the bus services they replace not because they are rails but because they are frequent. Light rail lines usually operate every 5 to 15 minutes, whereas most bus routes operate every 10 to 30 or even 60 minutes.
What role do rails play in attracting transit riders? Twenty-two of the nation's thirty-two most populated metro areas have rail transit services of some kind or another (not counting tourist lines). I estimated rail's proportion of each metro area's transit service by comparing rail vehicle revenue hours with scheduled bus revenue hours (including trolley bus but not demand responsive services). The results ranged from 2 percent for Denver to 72.5 percent for Miami.
In the thirty-two top cities, the correlation between the percent rail service and the percent of commuters using transit was low, 0.387. The correlation between trips per capita and percent rail service was somewhat higher, though still fairly low, at 0.433.
How much of this is due to the few large cities with a long history of rail service: Boston, Chicago, Cleveland, New York, Philadelphia, Pittsburgh, San Francisco, and Washington? (I include Washington because it had a streetcar system immediately preceding its current Metro rail.) For these cities alone, the correlation is 0.439. For cities that have installed rail transit in the past few decades -- Atlanta, Baltimore, Buffalo, Dallas, Denver, Memphis, Portland, Sacramento, San Diego, and St. Louis -- the correlation is 0.437, about the same as the older cities. But add Los Angeles and it falls to 0.393; add Miami and Ft. Lauderdale and it falls to 0.177.
Table Five: Correlations Between Percent Rail and:
% transit commuters 0.387 /1 Transit trips/capita 0.433 /1 Transit trips/capita 0.439 /2 Transit trips/capita 0.437 /3 Transit trips/capita 0.393 /4 Transit trips/capita 0.177 /5 1. 32 largest urbanized areas. 2. Boston, Chicago, Cleveland, New York, Philadelphia, Pittsburgh, San Francisco, and Washington 3. Atlanta, Baltimore, Buffalo, Dallas, Denver, Memphis, Portland, Sacramento, San Diego, and St. Louis 4. Atlanta, Baltimore, Buffalo, Dallas, Denver, Los Angeles, Memphis, Portland, Sacramento, San Diego, and St. Louis 5. Atlanta, Baltimore, Buffalo, Dallas, Denver, Ft. Lauderdale, Los Angeles,Memphis, Miami, Portland, Sacramento, San Diego, and St. LouisMiami and Ft. Lauderdale both have commuter rail services, which are probably not as frequent as a heavy or light rail. Los Angeles, of course, sacrificed bus services to build rails. My tentative conclusion is that it is rail's increased frequencies, not the rails themselves, that boost transit ridership.
The overall conclusion is similar. For those whose goal is to increase transit usage and reduce auto driving, the single most effective policy is to increase the frequency of existing transit services. This will do far more than increasing density and, for a given amount of money, more than building new rail lines.