Chapter One: Transportation Is Expensive

The Bureau of Labor Statistics, an agency of the U.S. Department of Labor, conducts the Consumer Expenditure Survey (CES) annually. This data is available at the national, regional, and metro area level for 28 selected metro areas. At the national level, the data is further broken down by income quintile, housing tenure, age of reference person, household composition, and other factors. The most recently available data is for 1998 for national-level statistics, and 1997-1998 for regional and metro area statistics. Regional and metro area data are averaged over two years so that the sample sizes are statistically significant.

Our nationwide analysis uses aggregate data for all consumer units. The percentage of expenditures used for transportation is derived by dividing the household expenditures on transportation by total household expenditures. Likewise, expenditures on shelter, food, and other categories was calculated using the same procedure.

As reported in the CES, personal transportation spending includes expenditures on ship- and air-fare. STPP removed these expenditures as they were not reflective of day-to-day transportation. This was done by estimating the percentage of public transportation spent on ship- and air-fare, using regional figures from the CES, and applying those to national figures. At the metro area level, we used the appropriate regional figure from the CES, and applied that percentage to each metro area.

Our breakdown of transportation expenditures was calculated by dividing each transportation expenditure, such as gasoline and motor oil, by total transportation expenditures.

Expenditures by income quintile were calculated using Table 45 of the Consumer Expenditure Survey. For this analysis, we looked at transportation expenditures as a percent of income rather than as a percent of expenditures. This was done to show the different burden transportation expenses place on people of different incomes. However, since most of our analysis sought to determine locational differences, our primary measure, percent of expenditures, helps lessen regional income and income tax disparities that might skew results.

Chapter Two: Where You Live Matters

International Analysis

For our comparison of American cities to cities in other countries, we used Peter Newman and Jeffrey Kenworthy’s analysis of metro areas around the world as published in An International Sourcebook of Automobile Dependence in Cities, 1960-1990 and Sustainability in Cities. Professor Newman and Professor Kenworthy were unable to calculate the percentage of consumer expenditures going to transportation, but they were able to calculate the percentage of Gross Regional Product (GRP) devoted to passenger transportation. This serves as an adequate proxy by which we can compare metro areas and continents to each other.

U.S. Metro Area Analysis

The metro area analysis of the CES data followed the parameters set forth above for the national analysis. The areas covered are U.S. Metropolitan Statistical Areas as defined by the U.S. Census Bureau. The geographical areas covered by MSAs can be found at http://www.census.gov/geo/www/mapGallery/ma_1999.pdf . Rankings are based on the portion of household expenditures that went toward transportation (excluding ship- and air-fare), rather than absolute dollars, in order to avoid income and income tax disparities in different regions. The analysis of income groups was conducted according to the portion of income after taxes that was devoted to transportation.

Intra-Metro Area Analysis

For the analysis of auto cost differences, the Center for Neighborhood Technology relied on automobile cost models developed as part of the Location Efficient Value model. Researchers at the Center for Neighborhood Technology, STPP and the Natural Resources Defense Council developed a formula for assessing the Location Efficiency Value (LEV) of a place: how expensive it is to live in a place based on a set of characteristics. By looking at the demographic characteristics (income and household size), land use (households per residential acre and households per total acre), pedestrian friendliness (existence of a block grid, access to amenities), and transit service (location and frequency) of a place, we can accurately predict how many cars a typical household in a given neighborhood would own, how far that household would drive, and how much that household would spend on transportation¹.

Analysts began with the goal of predicting auto ownership and travel demand for each geographic unit. We performed multiple regression analyses in order to determine the relationship between the different community characteristics listed above, and household automobile ownership and use. The results in Chicago: R2 = 0.963 for Vehicles per household, and R2 = 0.935 for VMT/HH. The R2 figure represents a comparison with available data sources: the 1990 Census auto ownership data and 1995-96 odometer reading data from the Illinois EPA and the California Bureau of Automotive Repair. There were 1 million odometer reading records available in Chicago, 2 million in San Francisco, and 3 million for Los Angeles. For more information about the LEV regression analysis or data sources please contact Peter Haas at the Center For Neighborhood Technology, pmh@cnt.org  or (773) 278-4800.

The LEV model has proven to be a reliable way of predicting automobile ownership and use based on community characteristics, and can be used to predict automobile expenditures at the community level.

The costs of owning and operating a car, which are illustrated by maps in this report, were calculated using the vehicles and VMT per household as predicted by the LEV model. We applied the Federal Highway Administration’s 1991 formula for calculating auto expenses: $2,207 per car per year + 12.7 cents per mile driven, to derive an annual cost per household².

One important note is that all of the Modeled Auto Cost maps published in this report were generated using a standard income and household size. This technique allows us to see more clearly the effect of place, as opposed to income level, on car dependence and cost.

Chapter Three: Sprawl Makes Transportation Expensive

To determine what is influencing transportation expenditures, STPP and affiliated researchers compared the CES metro-level data to a wide variety of transportation, land use, and demographic data available for these metro areas. The most significant correlations are reflected in this report.

Sprawl Factors

The effect of sprawl was calculated using a composite of five land use variables³ available in each metro area represented in the CES (except Anchorage, Alaska). We performed a bivariate correlation of this composite measure and the percent of expenditures spent on transportation. This analysis supports the hypothesis that there is a significant relationship between the land use efficiency index and the percent of expenditures devoted to transportation (R2 = 0.482, significant at the 0.01 level).

The five land-use variables that comprise the composite measure are:

1) Extent of large-lot and scattered suburban development in a metropolitan area: Percentage of the metropolitan area population living in census tracts with densities of less than 750 persons per square mile, or just under one household per half acre. Census tracts with less than 100 persons per square mile were excluded to eliminate rural and largely undeveloped areas.

2) Degree of clustering in a metropolitan area: Standard deviation of population density across census tracts in a metropolitan area. Again, census tracts with less than 100 persons per square mile were excluded to eliminate rural and largely undeveloped areas.

3) The extent of medium-to-high density residential density in a metropolitan area: Percentage of the metropolitan area population living in census tracts with densities greater than 10,000 persons per square mile, or just over six households per acre of total area. In this calculation as well, census tracts with less than 100 persons per square mile were excluded. These are densities which will support transit.

4) Degree of centeredness in a metropolitan area: Population density gradient as a function of distance from the central business district of the metropolitan area’s dominant city, measured by fitting a negative-exponential function to density data for census tracts.

5) Degree of mixing of six sectors—retail, personal services, entertainment, health, education, and other professional service—within subareas of a metropolitan area: An "entropy" formula, often used in travel research, was used to measure the degree of land-use mixing within traffic analysis zones.

The values of the composite sprawl measure range from -3.44 for New York to 1.12 for Tampa. The presence of negative values is a function of the analytical technique employed, factor analysis. The resulting "factor scores" have no intrinsic meaning, but are only meaningful relative to one another.

The weight assigned to each variable was derived via factor analysis with varimax rotation. The first four variables were estimated with data from the 1990 U.S. Census of Population and Housing. The last variable was estimated with data from the 1990 Census Transportation Planning Package.

In addition, to illustrate that the relationship between sprawl and spending on transportation holds true even for the transportation sub-categories, STPP performed an additional analysis. We divided the metro areas surveyed into three groups, according to their sprawl factors. For the highest and lowest groups, we calculated average household spending on car and truck purchases, gasoline, and miscellaneous automobile expenses, and compared these values.

Transportation Choice Factors

We also wanted to look at how public decisions about providing transportation choice would affect personal expenditures. For this analysis, we calculated the transportation choice ratio, which is a measure of the amount of transit service provided, relative to the amount of roadway capacity in a given metro area. Transit service was measured as the hourly revenue service miles of transit per household. In other words, the number of miles traveled by all buses, subway cars, or light-rail cars in an hour, per household. These numbers were available from the Federal Transit Administration’s Transit Database. Roadway capacity was measured by the number of lane-miles of freeways, expressways, principal arterials, and Interstates per household, in a given metro area. Dividing the transit service by the roadway capacity gives the transportation choice ratio for each metro area. Anchorage, Alaska was excluded from this particular analysis because we were unable to find roadway capacity for that metro area.

The Transportation Choice Ratio was found to be highly correlated with the sprawl measure. Performing a bivariate correlation also found a significant relationship between the transportation choice ratio and the percent of expenditures spent on transportation (R2 = 0.336, significant at the 0.01 level).

Using Newman and Kenworthy’s international data, we found that these same patterns held true even across national boundaries. To determine if there was a relationship between transportation balance and household spending on transportation, we examined the ratio between roadway capacity and transit service for three Great Lakes metro areas: Detroit, Chicago, and Toronto. This transportation balance differs slightly from the Transportation Choice Ratio and uses data from Newman and Kenworthy’s database. Comparing this transportation balance to the portion of GDP spent on transportation in these three geographically similar metro areas, demonstrates the relationship between the two variables.

Other Factors

To ensure that the differences among metro areas were not just an artifact of significantly different insurance rates or gasoline prices, STPP acquired average gasoline prices by metro area (from the BLS Consumer Price Index) and average insurance rates by state (from the National Association for Insurance Commissioners). Dividing the metro areas into two groups according to their transportation expenditure ranking, and then comparing average gasoline prices and insurance rates assured us that these factors were of minor significance.

Chapter Four: Expensive Cars and Inconvenient Homes

STPP calculated the changing "equity" of automobile ownership by estimating the approximate depreciation of a new $20,000 car. We used an average depreciation rate, as supplied by FinanCenter. Subtracting the amount owed on the car from the value of the car, adjusted for depreciation gives the approximated equity value of the car. This same methodology was used to calculate home equity, with one major difference. Rather than depreciate, home values increase at a rate of approximately 3.2 percent per year, as estimated by the U.S. Department of Housing and Urban Development.

The location efficiency value (LEV) of a place is based on the modeled auto costs described above. In the LEV maps in this report, each range represents the amount a household saves each month by living in places with varying degrees of location efficiency. The model functions such that there is a typical household income and size, which is described in the legend for each city, and comparisons of cost are performed based on that household’s description.  The representative household is based on the 1997-1998 Consumer Expenditure Survey average. In Chicago, a typical household had an income of $43,000 with 2.6 members. Typical households in Los Angeles had incomes of $50,000 and 2.8 members. For the San Francisco Bay Area, the figures were an income of $56,000 with 2.5 members. As in the auto costs model, the household income and size are variables used to predict vehicle ownership and use for each place.

In the Chicago region, we applied the models at quarter section, a half-mile, by half-mile square. In the San Francisco Bay Area and Los Angeles metro area, we used Traffic Analysis Zones, a unit defined by each region’s Metropolitan Planning Organization. Further information about LEVs can be found at www.locationefficiency.com .

In calculating how lower automobile debt might translate into higher home ownership, we looked at the consequences of lowering automobile credit outstanding by one percent. Here is a walk-through the calculations step-by-step.

The value of an average first home is $113,300, and a first-time homebuyer can expect to make a down payment of about 10 percent, or $11,330. One percent of automobile credit outstanding is $4.657 billion. Therefore, a 1 percent decrease in credit used for cars would be enough money to pay for 411,032 down payments of $11,330. In order to increase the homeownership rate by 1 percent, 1,050,000 households would have to buy a home, and therefore make a down payment. 2.55 percent of total automobile credit is $11.9 billion, which is enough to pay for 1.05 million down payments.