Special Studies, December 12, 2007
By Helen Fei Liu
Report available to the public as a courtesy of HousingEconomics.com
Vehicle use and emissions have attracted substantial attention in recent years. According to the Energy Information Administration, has the largest impact on global warming of any of the monitored greenhouse gases. About 33 percent of total U.S. greenhouse gas emissions are generated from the transportation sector, and among these, CO2 emissions represents 95 percent of the greenhouse gas emissions from mobile transportation sources.
Concerns about these numbers and their possible implications for climate change issues have prompted states such as California, Massachusetts and Washington require that developers quantify greenhouse gas emissions from vehicle use in large residential projects they are planning. However, these states typically do not provide any guidance on how to perform the calculations, and there is currently no well-established, verifiable method for estimating CO2 emissions or vehicle miles traveled (VMT) for households in a particular development.
This article estimates household gasoline consumption and associated CO2 emissions using the 2001 National Household Travel Survey (NHTS), which is conducted by the Federal Highway Administration (FHWA) within the Department of Transportation. The article uses standard statistical methods to estimate gasoline consumption as a function of the geographic and household characteristics available in the NHTS data. Housing units per block group is used as a proxy for the compactness of a residential subdivision.
The results show that, controlling for the factors available in the NHTS data, gasoline consumption and the associated CO2 emissions decline as the compactness of subdivisions increases. The article also analyzes factors that determine gasoline consumption: VMT, efficiency of vehicles owned and the speed at which they are driven. The results show a clear “congestion” effect: as the compactness of subdivisions increases, vehicles tend to be driven at less efficient speeds. However, this congestion effect is not strong enough to totally offset the reduced VMT, so that the statistical methods employed still estimate that gasoline consumption and the associated CO2 emissions will be lower in more compact development.
The household and housing unit characteristics used in the NAHB model are shown in Table 1. All of them turn out to have a statistically significant impact on gasoline consumption.
In particular, the model finds that gasoline consumption tends to be higher for households that
- are larger
- contain more workers
- have higher incomes
- own their homes
- live in single family homes
- are younger
- are less educated
- have a male householder
- have a white householder
- have a Hispanic householder
Some of these results (such as the finding that larger households use more gasoline) are quite intuitive. Others (such as some of the effects of race and ethnicity on gasoline consumption) are perhaps more surprising. It’s important to remember that the model controls for all of these factors (as well as the factors mentioned in other sections of the article) at the same time. Thus, when the model finds that households with higher incomes tend to consume more gasoline over the course of a year, this is after controlling for the size, incomes, race, etc. of the household—as well as for characteristics of the area in which the home is located, to the extent these are available in the data.
The household and housing unit characteristics controlled for in this article are same ones the U.S. Department of Transportation, the Environmental Protection Agency (EPA), and the Department of Energy control for in their NHTS-based travel forecasting models. Table 1 shows the averages for these characteristics in the data set. The table also shows the estimated gasoline use for a typical subdivision containing 100 households with average household attributes: about 98,319 gallons of gas per year to travel to work or school, shop, pursue leisure activities, etc.
The fact that it’s possible to tease out significant impacts of various household and housing attributes on gasoline consumption doesn’t mean that all the impacts are the same. An additional worker has a stronger impact on gasoline consumption than simply adding an extra person to the household, for example, although both effects are statistically significant. To illustrate this point, Table 2 shows how estimated gasoline consumption for the 100 unit subdivision changes from the 98,319 baseline as the nature of the homes and households in the subdivision changes. For readers interested in more technical details, the appendix table shows the full NAHB statistical model used to estimate gasoline consumption (as well as models using the same variables to estimate VMT, the efficiency of vehicles used, and the efficiency of the average speed driven.)
For the NAHB models used in this article, an attempt was made to use all the geographic information available in the data, which is somewhat limited. For example, the NHTS data don’t identify individual states, but do indicate the four principal Census regions. The gasoline consumption model finds that, all else equal, gasoline consumption tends to be lowest for households living in the Northeast and highest for households living in the South region.
The data set also indicates whether a household is in a Metropolitan Statistical Area (MSA), an urban area, and provides some information about the metro area’s population, although it doesn’t identify individual metropolitan areas specifically. MSAs are defined by the Office of Management and Budget (OMB), and are aggregations of counties based on commuting patterns. In contrast, urban areas (which are defined by the Census Bureau) are basically clusters of relatively small units called block groups that have a population density above a certain threshold. Areas not classified as urban are rural. Areas can be classified as urban or rural whether they are inside or outside of metro areas.
The strongest result the NAHB gasoline consumption model finds among the urban and metro area variables is that households living in urban areas consume less gasoline than households living in rural areas, although for the sake of completeness the model analyzes all urban/metro-size combinations available in the data. For example, the model estimates that a household living in an urban area in an MSA with population size under 1 million consumes about 276 fewer gallons of gas than a household living in a rural area not in an MSA, all else being equal.
The NHTS data also indicate whether a home is located in an MSA with a rail transportation system. There is no information about how close the transportation system comes to a particular home. Nevertheless, the model estimates that a household would on average use about 70 fewer gallons of gas if living in an MSA with rail transportation. While we know that rail commuting is available in a limited number of metro areas and that a minority of households uses rail transportation, the presence of rail transportation reduces gasoline consumption by about 70 gallons per household.
Characteristics on the scale of a metro area, such as its population or whether it has a rail system, are generally beyond the control of an individual developer. However, the data also contain some information about what is in the neighborhood (or “block group”) in which a house is located. Block group is defined by the Census Bureau to capture approximately 500 housing units or about the size of a residential subdivision. We convert housing units per square mile to housing units per acre and call it “compactness of subdivision” to avoid possible confusion over the term block group. The NHTS data don’t show the compactness of the subdivision precisely, but group it into six categories—ranging from under 0.08 units per acre, to over 7.81 units per acre.
Table 3 shows the estimated annual gasoline consumption and CO2 emissions for a subdivision with 100 households, assuming the average household and housing unit characteristics of the entire sample. The variable supplied in the NHTS data and estimated by the NAHB statistical model is gasoline consumption. Gasoline consumption is then converted into CO2 emissions by applying a simple formula obtained from the Energy Information Administration, which is based on the number of carbon atoms in a gallon of gasoline and assumes complete combustion.
Table 3 also shows what happens in each of these areas as compactness of the subdivision changes. A general trend is that the estimated gasoline consumption decreases as the subdivisions become more compact, controlling for the household and geographic factors available in the NHTS data. For example, the estimated gasoline consumption is about 90,700 gallons for a subdivision of 100 households and a density of 1.56 to 4.69 units per acre. As the subdivisions become more compact, the estimated gasoline consumption decreases to less than 80,000 gallons in the case where there are more than 7.81 housing units per acre.
1.56 to 4.69 housing units per acre translates into about 0.21 to 0.64 acres per unit, which is a fairly typical lot size for new construction. About 31 percent of single family detached homes completed in 2006 were built on lots falling into this size range. Nearly 80 percent are on a lot that was 0.64 acre or smaller (Table 4). However, lot sizes will generally be smaller than acres per housing unit measured over a block group or subdivision, because a subdivision will also typically include roads and other public spaces.
Because CO2 emissions are computed as a simple ratio of gasoline consumption, CO2 emissions also decline in Table 3 as the subdivision becomes more compact, controlling for other factors.
It is not possible to control for every conceivable factor with a possible impact on gasoline consumption in a statistical model based on the NHTS data, however. Characteristics such as the presence of concentrated employment centers and their proximity to a particular subdivision, for example, obviously could impact gasoline consumption. But information on employment centers is unavailable in the NHTS data, so a study that accounts for them, or other factors not captured in the NHTS, would first require a new data set to be compiled.
Other Measures of Vehicle Use
This article emphasizes gasoline consumption because that is the NHTS variable most directly related to CO2 emissions. Many other studies, however (even studies based on NHTS) have instead estimated impacts on VMT. Therefore, in order to help shed light on previous studies and show how CO2 emissions are related to other aspects of vehicle use, we have also estimated statistical models that predict VMT, the efficiency of the vehicles owned, and efficiency of the speed with which they’re driven. The complete set of coefficients for the NAHB models that estimate VMT and the efficiency measures are shown in the appendix table, along with the gasoline consumption.
Estimated VMT and efficiency measures are also shown in Table 3. Again these estimates control for a variety of household and housing unit characteristics. The table shows little relationship between efficiency of vehicles owned and subdivision compactness, except that residents in the least dense subdivisions tend to own less efficient vehicles. On the other hand, Table 3 shows a clear relationship between subdivision compactness and the average speeds vehicles are driven. As the subdivision becomes more compact, the estimated results show that vehicles are driven fewer miles, but they tend to be driven at less efficient speeds. However, the congestion effect (less efficient driving speeds) is not strong enough to completely offset the effect of reduced VMT. So on balance, households in more compact development still tend to use less gasoline and thus generate fewer CO2 emissions.
In summary, Table 3 shows how NHTS data can be used to estimate the efficiency of vehicles owned, how far they’re driven, how efficiently they’re driven, the gasoline they consume, and the associated CO2 emissions for a particular subdivision. The table also shows how these variables are related to the compactness of the subdivision (housing units per acre). Keep in mind that compactness of development in this case may be acting in part as a proxy for variables not available in the NHTS data set—such as the presence of bike trails, how well interconnected local roads are, or the characteristics of employment centers near the development. Some of these characteristics are difficult to quantify, and investigating them would require the creation of new (possibly difficult to compile) data sets.
Other greenhouse gases include carbon dioxide (CO2), methane, nitrous oxide, various hydrofluorocarbons, perfluorocarbons, and sulfur hexafluoride. For details, see article “Greenhouse Gas Emissions”.
Source: Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2004, U.S.EPA, 2006.
The NHTS is designed to provide information to assist transportation planners and policy makers who need comprehensive data on travel and transportation patterns in the United States.
The model is called Transportation Analysis and Simulation System (TRANSIMS), which is an integrated system of travel forecasting models designed to give transportation planners information on traffic impacts, congestion, and pollution. The TRANSIMS synthesizes the individual and their activities using census data. These synthetic households are usually generated for a block group, and each individual in the household has an associated set of demographics which are matched closely to the real world. Rather than running the TRANSIMS model, we take a regression approach in this article—using the same set of demographic variables that are in the TRANSIM model.
For more information, see http://tmip.fhwa.dot.gov/transims/.
 The categories seem to break at odd places because in the data set the variable is expressed as housing units per square mile. We convert this to housing units per acre in order to generate a measure that’s easier to visualize, but it does result in categories that break at odd fractions of a housing unit.
The formula is that about 19.36 pounds of CO2 are produced by burning 1 gallon of gasoline (personal communication with the Energy Information Administration).
For households own more than one vehicle, we derive weighted MPG by weighing the MPG by the vehicle miles traveled by each vehicle in a household, and summarize VMT for all vehicles in the household.
The FHWA’s measure of vehicle efficiency adjusts for many factors such as average miles driven per day, seasonal temperature variations, humidity, and road surface conditions. FHWA uses average miles driven per day to categorize most of the driving done as “highway” or “city”. Highway driving is assumed to be characterized by less frequent stops, long trip length and thus greater efficiency; while city driving is assumed to be characterized by more frequent stops and short trip length, and thus lower efficiency.However, the NHTS data do not contain enough information to reproduce the FHWA’s estimates of vehicle efficiency. Instead, we calculate the difference between average trip speed and a theoretically optimal speed. The optimal speed is considered to be about 45 miles per hour for motor vehicles with internal combustion engines (according to the article “Growing Cooler: Evidence on Urban Development and Climate Change” by Reid Ewing, Keith Bartholomew, et al.). Speeds above or below this “sweet spot” should result in lower efficiency and higher gasoline consumption.
Return to Top
Excel Tables (Excel)
See previous Special Studies